Steam Turbine Blade Failure Root Cause Analysis Guide: Place Image Here

Steam Turbine Blade Failure Root Cause Analysis Guide
Place Image Here
10305538
10305538
Steam Turbine Blade Failure Root
Cause Analysis Guide
1014137
Final Report, March 2008
EPRI Project Manager

J. Thibault
ELECTRIC POWER RESEARCH INSTITUTE

3420 Hillview Avenue, Palo Alto, California 94304-1338 ▪ PO Box 10412, Palo Alto, California 94303-0813 ▪ USA
800.313.3774 ▪ 650.855.2121 ▪ [email protected] ▪ www.epri.com
10305538
DISCLAIMER OF WARRANTIES AND LIMITATION OF LIABILITIES
THIS DOCUMENT WAS PREPARED BY THE ORGANIZATION(S) NAMED BELOW AS AN
ACCOUNT OF WORK SPONSORED OR COSPONSORED BY THE ELECTRIC POWER RESEARCH
INSTITUTE, INC. (EPRI). NEITHER EPRI, ANY MEMBER OF EPRI, ANY COSPONSOR, THE
ORGANIZATION(S) BELOW, NOR ANY PERSON ACTING ON BEHALF OF ANY OF THEM:
(A) MAKES ANY WARRANTY OR REPRESENTATION WHATSOEVER, EXPRESS OR IMPLIED, (I)

WITH RESPECT TO THE USE OF ANY INFORMATION, APPARATUS, METHOD, PROCESS, OR
SIMILAR ITEM DISCLOSED IN THIS DOCUMENT, INCLUDING MERCHANTABILITY AND FITNESS
FOR A PARTICULAR PURPOSE, OR (II) THAT SUCH USE DOES NOT INFRINGE ON OR
INTERFERE WITH PRIVATELY OWNED RIGHTS, INCLUDING ANY PARTY'S INTELLECTUAL
PROPERTY, OR (III) THAT THIS DOCUMENT IS SUITABLE TO ANY PARTICULAR USER'S
CIRCUMSTANCE; OR
(B) ASSUMES RESPONSIBILITY FOR ANY DAMAGES OR OTHER LIABILITY WHATSOEVER

(INCLUDING ANY CONSEQUENTIAL DAMAGES, EVEN IF EPRI OR ANY EPRI REPRESENTATIVE
HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES) RESULTING FROM YOUR
SELECTION OR USE OF THIS DOCUMENT OR ANY INFORMATION, APPARATUS, METHOD,
PROCESS, OR SIMILAR ITEM DISCLOSED IN THIS DOCUMENT.
ORGANIZATION(S) THAT PREPARED THIS DOCUMENT
Turbine Technology International, Inc.
NOTE
For further information about EPRI, call the EPRI Customer Assistance Center at 800.313.3774 or
e-mail [email protected].
Electric Power Research Institute, EPRI, and TOGETHER…SHAPING THE FUTURE OF ELECTRICITY
are registered service marks of the Electric Power Research Institute, Inc.
Copyright © 2008 Electric Power Research Institute, Inc. All rights reserved.
10305538
CITATIONS
This report was prepared by
Turbine Technology International, Inc.

2024 W. Henrietta Road, Suite 5F
Rochester, NY 14623
Principal Investigators
R. P. Dewey
T. C. Lam
This report describes research sponsored by the Electric Power Research Institute (EPRI).
The report is a corporate document that should be cited in the literature in the following manner:
Steam Turbine Blade Failure Root Cause Analysis Guide. EPRI, Palo Alto, CA: 2008. 1014137.
iii
10305538
10305538
PRODUCT DESCRIPTION
Steam Turbine Blade Failure Root Cause Analysis Guide is a concise reference written for
operators to plan and conduct an investigation into the most probable causes of a steam turbine
blade (bucket) failure. The report provides both an overview and step-by-step approach to
identifying the damage mechanisms most common to turbine blade failures. It proceeds to show
how damage mechanisms are related to the operating history prior to the blade’s failure and how
they are evaluated to establish their role as principal (root) causes versus secondary contributors.
Results and Findings

Over the life of every steam turbine, millions of dollars are spent on proper maintenance,
efficient operation, and routine repairs. Blade failures—from premature cracking to the in-
service catastrophic breakdown of the components—represent the single greatest threat to the
planned maintenance and reliable operation of these large units. Correct and timely diagnosis of
the root cause(s) is fundamental to managing the immediate problem and establishing a long-
term strategy for sister units operating with the same blade.
Challenges and Objectives

The measure of success in any blade failure investigation is the prevention of a repeat incident.
The process of investigation is complicated by the fact that it involves the coordination of
multiple activities and specialized engineering disciplines. This report is meant to provide a
turbine operator with a roadmap of how a typical investigation should proceed, when specialists
should be involved, what they should contribute, and how the evidence is used to establish
corrective action, either with the original equipment supplier or a qualified third party.
Applications, Value, and Use

In a competitive market, operators do not have the luxury of treating blade failures as random
occurrences, with a replacement in kind substituting for a serious attempt to fully understand
what precipitated the breakdown of a specific blade design. Statistics and experience show that if
a turbine blade reliability issue remains uncorrected, it will persist. The associated downtime and
lost power production associated with repeat failures substantially outweigh the cost and time
spent establishing the root cause and a permanent refurbishment or replacement strategy.
EPRI Perspective
The report is meant to remove some of the black box that surrounds the investigation of
unexpected blade problems. It focuses on identifying what, when, and how critical information is
used to minimize speculation and form a technical consensus from which a corrective strategy
with a high probability of success can be devised.
v
10305538
Approach
The methodology and approach described in this report are derived from an experience base of
more than 350 independent failure investigations. Examples are included as an appendix to
demonstrate how certain critical information derived from the method of investigation was
applied to diagnose the most common types of failures.
Keywords
Fatigue
Finite element analysis
Root cause analysis
Steam turbines
vi
10305538
ABSTRACT
This report offers practical assistance in the root cause analysis of steam turbine buckets or
blades. It is written for turbine operators who seek to manage a failure investigation by following
a standard methodology of three phases. Through these three phases, relevant evidence is
gathered and interpreted. The scope of this report includes a practical discussion of immediate
activities that should be taken when a failure occurs. It proceeds through each phase of site
forensics and metallurgical examination, engineering analysis, and the selection and
implementation of permanent corrective action. Presented in the report is the way that evidence
gathered in each phase is interpreted and used to define a successful strategy for replacing or
refurbishing the blades. The report concludes with a series of examples that follow the
methodology in an investigation of common blade failures.
vii
10305538
10305538
ACKNOWLEDGMENTS
Thanks to the following members of the EPRI Steam Turbine Root Cause Analysis Technical
Advisory Group (TAG) for their contributions of time and expertise:
• Ameren: Tom Kordick and Dennis Sullivan
• EPRI: Steven Hesler, Alan Grunsky, and Tom Alley
• First Energy: Phil Schuchter
• Luminant: Steve Brewton
• South Texas Project: Robert Bjune
• Southern Company: Charles Boohaker, Trevor Brooks, and Jim Carlson
• Tennessee Valley Authority: Mark Koss, Scott Turnbow, and Jerry Best
• Xcel Energy: Eric Dupont
EPRI recognizes the contributions of Ronald Munson and Karen Fuentes of Mechanical and
Material Engineering, who were primarily responsible for this report’s section on metallurgical
analysis of blade failures.
Finally, EPRI recognizes the lifetime of work, which is represented throughout the entire report,
of Tony Chungtung Lam. This report represents his final technical effort in a 28-year career that
was devoted to the root cause investigation of steam turbine blade failures.
ix
10305538
10305538
ABBREVIATIONS, ACRONYMS, AND GLOSSARY
OF TERMS
Abbreviations and Acronyms

AISI American Iron and Steel Institute
ASME American Society of Mechanical Engineers
ASTM ASTM International (formerly the American Society for Testing and Materials)
BVM blade vibration measurement
CAD computer-aided design
CAF corrosion-assisted fatigue
CFD computational fluid dynamics
DIN Deutsches Institut für Normung
DOD domestic object damage
EDS energy dispersive X-ray spectroscopy
EPRI Electric Power Research Institute
FATT fracture appearance transition temperature
FEM finite element model
FM fracture mechanics
FOD foreign object damage
HCF high cycle fatigue
HP high-pressure
HTC high-temperature creep
in. inch(es)
IP intermediate-pressure
ksi kips per square inch
LCF low cycle fatigue
xi
10305538
LCM life cycle management
LP low-pressure
mm millimeter(s)
MPa megapascal(s)
MW megawatt(s)
MWe megawatt(s) electric
ND nodal diameter
NDE nondestructive evaluation/nondestructive examination
NPF nozzle passing frequency
OEM original equipment manufacturer
PAA partial arc admission
PDF probability density function
PH precipitation-hardened
QA quality assurance
QC quality control
RCA root cause analysis
rpm revolutions per minute
SCC stress corrosion cracking
SEM scanning electron microscope
SPE solid particle erosion
SS stainless steel
TAG technical advisory group
TMF thermal mechanical fatigue
UT ultrasonic testing
WDE water droplet erosion
WDS wavelength dispersive spectroscopy
xii
10305538
Glossary of Terms
anodic corrosion. Where the environment actually consumes the blade alloy. Metal is dissolved
from crystalline into ionic form, either globally (general) or under deposits (crevice corrosion).
asynchronous vibration. Blade response to dynamic forces at frequencies that are not
harmonics of shaft rotating speed, but involve a more direct interaction of the flow around the
blade airfoil.
blade-bucket. Components and their linkages installed around the circumference of a turbine
drum or disk to convert steam pressure and kinetic energy into a tangential force.
cathodic corrosion. A type of corrosion where metal wastage is not observed, but instead
hydrogen is present that can accelerate the cracking as cathodic stress corrosion cracking or
controlled corrosion fatigue.
corrosion-assisted fatigue (CAF). A damage mechanism referring to the formation of cracks

from corrosives deposited on the material and dynamic stresses from blade vibration,
accumulated during hours of unit operation or during unit cycling.
creep. A damage mechanism involving the formation of voids and cracks from the combined
effect of high temperatures and steady stresses accumulated during hours of unit operation.
damage. The physical consequence of the various mechanisms that are encountered during the
hours and cycles of operation experienced by any turbine, appearing as wear, degradation, or
fatigue.
deformation. A mechanism whereby the load on the blade exceeds the yield strength of the
material, usually as a consequence of a rubbing or object impact event.
design analysis. Analysis wherein computer simulation calculates the operating natural
frequencies and stresses of the blades, then applies these with the appropriate material properties
to estimate life consumption.
design deficiency. Signified by the premature formation of cracks due to excessive steady,
dynamic, or thermal stresses that result in an accelerated rate of fatigue damage to the blades.
detuning. A design practice by which the stiffness and mass of the blades are controlled so that
the natural frequencies maintain a reasonable margin against encountering resonance during
operation.
dynamic stress. Consequence of the blade-disk system response to synchronous or

asynchronous forces present in the steam flow that result in vibration of the modes at their
natural frequencies.
xiii
10305538
erosion. Damage mechanisms involving removal of blade material by the impact of water
droplets entrained in the steam of the low-pressure stages or solid particles entrained in the steam
of high-pressure and intermediate-pressure stages.
failure. When any damage mechanism effectively compromises the structural integrity of the
blades to a point where corrective action is considered necessary to avoid a repeat occurrence.
failure mechanisms. Processes of erosion, corrosion, creep, or fatigue that produce and
accumulate as damage on blades during the operating portion or the start-stop cycle of the unit’s
turbine-generator system.
failure sequence. Four phases used to segregate the development of cracks in steam turbine
blades as (1) incubation, (2) initiation, (3) propagation, and (4) final rupture.
fatigue. Damage mechanisms associated with cycles of start-stop operation (low) and vibration
(high) where loading and unloading result in the accumulation of permanent damage to the
material.
fractography. The study of blade fracture surfaces with the intent to determine the origin of the
crack and evidence of the various damage mechanisms that might have been experienced prior to
failure.
hysteresis map. A plot of the elastic and plastic components derived from a strain life curve.
These components form a repeated loop that can be equated with low cycle fatigue and high
cycle fatigue damage for selected locations within a blade.
immediate activities. Actions taken once the turbine is open for inspection that separate critical
path recovery activities undertaken to restore the unit from the three phases of root cause
investigation.
impact damage. A damage mechanism where objects outside of the turbine (foreign) or self-
generated debris (domestic) strike the blade surface as they pass through the steam path of the
turbine.
incubation. A generic term designating the period of service that precedes the formation of an
original crack on a microscopic (grain boundary) level in a blade.
initiation. Generic term designating the period of service where a crack has first formed in a pit,
erosion notch, void, or other discontinuity found on the blade, or in a localized stress raiser.
in-kind replacement. Installation of an exact duplicate of the blade design that failed as an
interim or permanent solution for damaged blades or failed rows.
interim solution. Corrective action taken to expedite the unit’s return to service while a
permanent solution is identified through the process of investigation and/or while it is secured
and scheduled for installation.
xiv
10305538
metallurgical analysis. Physical examination of intact damaged or failed blades to establish
their chemistry, mechanical properties, and any evidence related to the initiation and propagation
of cracks.
methodology. Investigative approach of up to three possible phases that guide the assembly of
evidence, engineering analysis, and identification of corrective action when a turbine blade
failure occurs.
modification. Adjustments made to correct a design deficiency in the original blade design or to
mitigate a debilitating condition within the stage that is encountered during operation.
off-design. Non-normal baseload operating conditions within or which exceed the allowable
limits or criteria that are specified for the turbine-generator by the original equipment
manufacturer.
overload. A damage mechanism where the ultimate tensile limits of the blade material are
exceeded by a severe operational transient event, such as overspeed operation or an impact event.
partial steam admission. A mode of operation and root cause of blade failure involving high
cycle fatigue caused by the transient dynamic stresses that are produced when steam flow is fed
to the control stage by separate valves.
permanent solution. A corrective action in the form of repair, replacement, or a refurbishment

designed to prevent a repeat occurrence of the same failure over the active service life of the
blades.
propagation. A generic term designating the period of service in which an initial crack advances
through the material at a deliberate, predictable rate after exceeding the material’s stress intensity
threshold.
random failure. An event not attributable to the selected design material or design stresses, or
which occurred as a consequence of improper installation, controls, or operating practices.
recovery scenarios. Parallel but separate logistical activities associated with replacement, repair,
or removal of the damaged blades to restore the units to service in the most expeditious manner
possible.
resonance. A root cause of blade failure involving high cycle fatigue, promoted by synchronous
vibration of modes with limited detuning, which results in a sympathetic response and maximum
dynamic stress.
root cause. A source related to the operation of the turbine that promotes one or more damage
mechanisms acting on critical locations in the rotating blades that are primarily responsible for
their failure.
xv
10305538
root cause investigation. The application of a generic methodology designed to consolidate the
results from inspection, analysis, and operating history into a reasonable account of how and why
a blade failure occurred.
rupture. A generic term designating the period of service in which a crack reaches the fracture
toughness of the material and a catastrophic separation of the blade is predicted.
secondary contributor. A source related to the operation of the turbine that promotes associated
mechanisms acting on the blades but that is not considered responsible for its failure.
speed cycling. A root cause of blade failure involving low cycle fatigue, promoted by localized
steady stresses in critical locations that result each time the unit experiences a start-stop cycle.
stalled flutter. A root cause of blade failure involving high cycle fatigue, promoted by
asynchronous vibration and dynamic stress produced at low load and high backpressure flow
conditions of the turbine-generator.
steady stress. A consequence of the blade-disk system response to the centrifugal forces applied
to the blade and steam-bending forces applied to the airfoil at a constant speed of rotation.
strain life curve. Test data from material specimens fitted to establish a relationship between
true plastic strain amplitude and the number of reversals before crack initiation is predicted.
stress corrosion. A damage mechanism involving the formation of cracks from the combined
effect of corrosive products deposited on the material and steady stresses produced during full-
speed operation, or from residual stress effects.
synchronous vibration. A blade response to dynamic forces that are present in the steam flow
that occur at frequencies that are harmonics of shaft rotating speed, that is, 1x, 2x, 3x, and so on.
thermal mechanical fatigue. A damage mechanism involving the formation of cracks from the
combined effect of thermal transient stresses produced during the startup or shutdown of a high-
temperature stage.
torsional system interaction. A root cause of a blade failure the results from a sympathetic
reaction to torsional oscillations of the rotor-train system originating as electric unbalances at the
generator.
unstalled flutter. A root cause of a blade failure involving high cycle fatigue, promoted by
asynchronous vibration and dynamic stress produced at high load and high backpressure flow
conditions of the turbine-generator.
xvi
10305538
CONTENTS
1 INTRODUCTION TO ROOT CAUSE ANALYSIS ..................................................................1-1

Organization of This Report ..................................................................................................1-1
First Response: Recovery Versus Investigation....................................................................1-5
Immediate Recovery Scenarios ............................................................................................1-7
Moving from an Interim to a Permanent Solution ................................................................1-10
Cost and Risk of Repeat Failures........................................................................................1-11
Summarizing the Objective for an RCA...............................................................................1-14
2 DAMAGE MECHANISMS AND THEIR ROOT CAUSES.......................................................2-1

Cause Elimination Process ...................................................................................................2-2
Failure Sequence ..................................................................................................................2-5
LCF Caused by Start-Stop Cycling .....................................................................................2-11
EPRI References on LCF ...............................................................................................2-11
Definition.........................................................................................................................2-11
Damage Estimation ........................................................................................................2-13
Design Factors and Issues .............................................................................................2-13
HCF Caused by Blade Vibration .........................................................................................2-13
EPRI References on HCF ..............................................................................................2-13
Definition.........................................................................................................................2-13
Different Sources............................................................................................................2-14
HCF Caused by Synchronous Vibration.........................................................................2-14
Definition....................................................................................................................2-15
Design Factors...........................................................................................................2-15
HCF Caused by Asynchronous Vibration .......................................................................2-17
Definition....................................................................................................................2-17
Design Factors and Issues ........................................................................................2-17
xvii
10305538
HCF from Torsional System Interaction .........................................................................2-19
Definition....................................................................................................................2-19
Definition....................................................................................................................2-21
Stress Corrosion from Chemical Attack and Steady Stress ................................................2-23
EPRI References on SCC ..............................................................................................2-23
Definition.........................................................................................................................2-23
Corrosion Fatigue from Chemical Attack and Dynamic Stress ...........................................2-25
EPRI References on CAF...............................................................................................2-25
Definition.........................................................................................................................2-25
Erosion from Water Droplets or Solid Particles ...................................................................2-27
EPRI References on WDE and SPE ..............................................................................2-27
Definition.........................................................................................................................2-27
Material Creep from High Temperature and Steady Stress ................................................2-29
EPRI References on HTC ..............................................................................................2-29
Definition.........................................................................................................................2-29
TMF from Transient Thermal Gradients ..............................................................................2-31
EPRI References on TMF ..............................................................................................2-31
Definition.........................................................................................................................2-31
Other Mechanisms: Overload, Deformation, Rubbing, and Impact.....................................2-33
EPRI References............................................................................................................2-33
Definition.........................................................................................................................2-33
Closing Remarks .................................................................................................................2-34
xviii
10305538
3 METHODOLOGY AND OBJECTIVES ...................................................................................3-1
Flowchart of Phases and Tasks in the Methodology.............................................................3-1
Overview of Three Phases of Investigation...........................................................................3-3
Phase 1: Assemble Failure Evidence: Tasks 1–5 ............................................................3-3
Phase 2: Perform Engineering Analysis: Tasks 1–9 ........................................................3-4
Phase 3: Develop a Corrective Strategy: Tasks 1–5........................................................3-5
Objectives of Tasks in Each Phase.......................................................................................3-5
Compile Inspection, Metallurgical, and Unit Operating Data ............................................3-5
Compile Unit Operating Data............................................................................................3-5
Compile Metallurgical Data...............................................................................................3-5
Evaluate Industrywide Data (Historical Problems) ...........................................................3-6
Make a Preliminary Identification of Failure Mechanisms ................................................3-6
Define Additional Material Properties ...............................................................................3-6
Define Stage Geometry ....................................................................................................3-7
Generate FEM ..................................................................................................................3-7
Calculate Steady Stresses ...............................................................................................3-7
Calculate Frequencies and Mode Shapes........................................................................3-8
Calculate Dynamic Stresses.............................................................................................3-8
Validate Model with Modal or Spin Pit Data .....................................................................3-9
Estimate Time and Probability to Fatigue and Fracture ...................................................3-9
Iterate Evaluation of Proposed Solutions .......................................................................3-10
Timing Issues ......................................................................................................................3-10
Comments on the Technologies and Their Potential Limitations ........................................3-11
Metallurgy .......................................................................................................................3-11
Finite Element Analysis ..................................................................................................3-12
FM ..................................................................................................................................3-12
Probabilistic Analysis......................................................................................................3-13
4 IMMEDIATE RESPONSE AND ON-SITE FORENSICS.........................................................4-1

Pre-Inspection Action Items ..................................................................................................4-2
Immediate Action Items When Turbine Is Open....................................................................4-3
Visible Inspection of the Parts ...............................................................................................4-4
Evidence Related to Installation ............................................................................................4-4
Basic Design Information Required.......................................................................................4-6
Items Specific to the Turbine(s).............................................................................................4-6
xix
10305538
Items Specific to the Blade and Row Involved ......................................................................4-8
Relevant Operating Data.......................................................................................................4-8
Final Comments ..................................................................................................................4-10
5 OFF-SITE FORENSICS: METALLURGICAL EXAMINATION ..............................................5-1

Basic Goal of the Laboratory Analysis Process ....................................................................5-2
Visual Examinations, Dimensional Examinations, and NDE .................................................5-5
Visual Examination ...........................................................................................................5-5
Collection and Preservation of Deposits/Residues...........................................................5-5
NDE ..................................................................................................................................5-6
Dimensional Examination .................................................................................................5-6
Determination of the Chemical Composition of the Component ...........................................5-7
Metallographic Examination ..................................................................................................5-8
Microstructure.................................................................................................................5-10
Hardness ........................................................................................................................5-11
Mechanical Properties Testing ............................................................................................5-12
Tensile Strength, Yield Strength, Elongation, and Reduction in Area ............................5-12
Impact Strength ..............................................................................................................5-12
FATT...............................................................................................................................5-13
Fatigue Properties ..........................................................................................................5-13
Elevated Temperature Properties...................................................................................5-13
Basic Features of Fractures ................................................................................................5-14
Documentation of Fracture Surfaces..............................................................................5-16
Low Magnification Examination .................................................................................5-16
High-Resolution Fractographic Examination..............................................................5-17
General Characteristics ..................................................................................................5-17
Incubation Through Rupture................................................................................................5-18
Features Associated with Incubation..............................................................................5-18
Features Associated with Initiation .................................................................................5-19
Features Associated with Crack Propagation.................................................................5-24
Features Associated with Final Rupture.........................................................................5-26
Basic Features of Non-Fractures ........................................................................................5-27
Creep..............................................................................................................................5-27
Corrosion ........................................................................................................................5-27
Erosion ...........................................................................................................................5-28
xx
10305538
Wear or Rubbing ............................................................................................................5-28
Impact Damage ..............................................................................................................5-28
Interpretation of Results: Relation to RCA ..........................................................................5-29
Examples from Blade Failures ............................................................................................5-32
6 GENERIC DATA.....................................................................................................................6-1
Information Sought ................................................................................................................6-1
Interpreting Statistical Data ...................................................................................................6-2
Similar Versus Same Blade Designs.....................................................................................6-3
Potential Sources ..................................................................................................................6-4
Summary Comments.............................................................................................................6-4
7 WEIGHING PHASE 1 EVIDENCE..........................................................................................7-1

Reasons to Execute a Replacement Strategy Without Further Analysis ..............................7-1
Random Failures ..............................................................................................................7-2
Replacement with a Demonstrated Redesign ..................................................................7-2
When to Proceed with an Engineering Analysis (Phase 2) ...................................................7-5
Permanent Strategies Sought from Phase 2 Investigation....................................................7-6
Final Comments ....................................................................................................................7-7
8 PHASE 2: ENGINEERING ANALYSIS ..................................................................................8-1

General Overview..................................................................................................................8-2
Plant Support: Geometry, Operation, and Installation Records ............................................8-2
Material Properties Needed..............................................................................................8-3
Operating Records and Reports Needed .........................................................................8-4
Disk/Rotor and Blade Geometry.......................................................................................8-6
Finite Element Modeling........................................................................................................8-9
Calculating Steady-State Stresses......................................................................................8-12
Natural Frequencies and Mode Shapes..............................................................................8-16
Dynamic Model...............................................................................................................8-18
Presentation of Frequency Results ................................................................................8-18
Mode Shapes .................................................................................................................8-20
Special Circumstances and Considerations for Marginally Detuned Blades ..................8-21
Summary Comments......................................................................................................8-25
Calculating Dynamic Stress ................................................................................................8-25
Further Comments on Stimulus and Damping ...............................................................8-28
xxi
10305538
Model Verification by Frequency and Vibration Testing ......................................................8-30
How Different Test Results Are Applied .........................................................................8-33
Life Prediction and Probability of Failure.............................................................................8-38
Crack Initiation: LCF .......................................................................................................8-38
Crack Initiation: HCF ......................................................................................................8-41
Crack Initiation: HTC ......................................................................................................8-43
Crack Propagation: SCC, CAF, Erosion, and Fretting....................................................8-45
Probability of Failure.......................................................................................................8-48
Final Comments on Phase 2 Activities................................................................................8-50
9 PHASE 3: IDENTIFYING A REPLACEMENT STRATEGY ...................................................9-1

Available Longer Term Strategies .........................................................................................9-1
Dealing with Design-Related Root Causes ...........................................................................9-2
Dealing with Operating-Related Root Causes.......................................................................9-4
Dealing with Environment-Related Root Causes ..................................................................9-5
Dealing with Material-Related Root Causes..........................................................................9-6
Salvaging Damaged or Worn Blades ....................................................................................9-7
Supporting Inspection Criteria .............................................................................................9-10
Application of Coatings........................................................................................................9-10
10 POST-FAILURE GUIDANCE .............................................................................................10-1

Upgrading Successfully.......................................................................................................10-1
Reviewing Bid Packages.....................................................................................................10-2
Validation of Replacements.................................................................................................10-4
Installation and Inspection...................................................................................................10-5
Monitoring to Detect an Impending Failure .........................................................................10-5
11 CLOSING REMARKS ........................................................................................................11-1

Additional Resources to Assist in Using Guidelines............................................................11-1
Litigation..............................................................................................................................11-2
12 REFERENCES ...................................................................................................................12-1
A CHECKLISTS ....................................................................................................................... A-1
xxii
10305538
B CASE STUDIES.................................................................................................................... B-1
B.1 L-1 Blade Tiewire/Airfoil Failure .................................................................................. B-2
B.2 L-2 Shroud/Shank Failure ........................................................................................... B-3
B.3 L-0 Blade Root Attachment Cracking.......................................................................... B-4
B.4 L-0 Blade Root Attachment Cracking from Asynchronous Vibration .......................... B-5
B.5 L-0 Blade Integral Cover Failure ................................................................................. B-6
B.6 L-0 Blade Airfoil Notching ........................................................................................... B-7
B.7 IP Airfoil Failure .......................................................................................................... B-8
B.8 HP 1 Root Cracking .................................................................................................... B-9
xxiii
10305538
10305538
LIST OF FIGURES
Figure 1-1 The Three Fundamental Steps in a Turbine Blade Failure Investigation..................1-3
Figure 1-2 Flowchart to Guide Root Cause Investigation of Turbine Blade Failures .................1-4
Figure 1-3 Critical Information to Be Immediately Secured When Failure Occurs.....................1-7
Figure 1-4 Three Basic Recovery Scenarios Involving Blades ..................................................1-8
Figure 1-5 Results from 1981 EPRI Blade Failure Survey.......................................................1-12
Figure 2-1 Timing and Failure Sequence Associated with Blade Damage Mechanisms...........2-5
Figure 2-2 The Local Strain Approach to Predicting Crack Initiation .........................................2-8
Figure 2-3 The Fracture Mechanics Approach to Predicting Crack Propagation.....................2-10
Figure 2-4 Partial Admission Loading of Control Stage Blades ...............................................2-21
Figure 3-1 Methodology to Guide a Steam Turbine Blade Failure Investigation........................3-2
Figure 4-1 Information Sought from Initial Visual Inspection of Turbine Components ...............4-4
Figure 4-2 Inspection of Items Related to Installation................................................................4-5
Figure 4-3 Basic Design Information Sought from an Open Turbine .........................................4-6
Figure 4-4 Examples of Damage Resulting from High-Energy Impact Mechanisms .................4-7
Figure 5-1 Typical Laboratory Sample.......................................................................................5-3
Figure 5-2 Post-Failure Damage on Failure Sample .................................................................5-3
Figure 5-3 Post-Failure Damage from Inadequately Peened Tenons .......................................5-4
Figure 5-4 Three Steps of Metallurgical Examination ................................................................5-9
Figure 5-5 Photomicrograph of AISI 403 Showing Grain Boundary Carbide Precipitates .......5-10
Figure 5-6 SCC Cracking in an L-0 Blade Adjacent to a Brazed Shroud.................................5-11
Figure 5-7 Typical Features Found on Fatigue-Fracture Surface ............................................5-14
Figure 5-8 CAF Fracture ..........................................................................................................5-15
Figure 5-9 Fracture Surface of a Ti 6-4 Blade..........................................................................5-16
Figure 5-10 Failure from HCF Initiating at Corrosion Pit on Leading Edge..............................5-17
Figure 5-11 Features of Crack Origins: Embrittlement, Inclusions, and Defects .....................5-20
Figure 5-12 Crack Origins: HCF and Fretting ..........................................................................5-21
Figure 5-13 Features of Crack Origins: CAF ...........................................................................5-22
Figure 5-14 Features of Crack Origins: SCC and Erosion .......................................................5-23
Figure 5-15 Features of LCF and HCF Crack Growth .............................................................5-25
Figure 5-16 Examples of Root Cracking Axial Direction ..........................................................5-33
Figure 5-17 Examples of Root Cracking Tangential Direction .................................................5-34
Figure 5-18 Examples of Airfoil Failure: L-1 Blades.................................................................5-35
xxv
10305538
Figure 5-19 Examples of Airfoil Cracking: LP L-0 Blades ........................................................5-36
Figure 5-20 Examples of Airfoil Cracking: IP Blades ...............................................................5-37
Figure 5-21 Examples of Welded Lashing Cracking and Failure: LP L-0 and L-1 Blades .......5-38
Figure 5-22 Examples of Tenon/Cover Failures: L-1 and L-4 Blades ......................................5-39
Figure 5-23 Examples of Root Integral Cover Failure: HP 1 and L-0 Blades ..........................5-40
Figure 8-1 Basic Measurements Taken from the Rotor .............................................................8-6
Figure 8-2 Basic Measurements Taken from Blade and Stage Information ..............................8-8
Figure 8-3 Typical Features Represented in the Blade Disk FEM ...........................................8-10
Figure 8-4 Typical Locations of High Steady Stress ................................................................8-13
Figure 8-5 Typical Stress-Strain Hysteresis Map Used to Define LCF/HCF Parameters ........8-15
Figure 8-6 Mode Shapes of Bladed Disks ...............................................................................8-17
Figure 8-7 Different Frequency Displays .................................................................................8-19
Figure 8-8 Potential Consequence of Excessive Grid Fluctuations .........................................8-22
Figure 8-9 Axial Entry Root Contact Effect on Frequencies ....................................................8-24
Figure 8-10 The Effect of Detuning on Dynamic Stress...........................................................8-27
Figure 8-11 Damping Ratios Versus Blade Response ............................................................8-29
Figure 8-12 Example of Frequency Results from Impact Tests of Single Blades ....................8-31
Figure 8-13 Examples of Impact and Shaker Test Frequency Data ........................................8-34
Figure 8-14 Examples from a BVM Test..................................................................................8-36
Figure 8-15 LCF Initiation Estimation.......................................................................................8-40
Figure 8-16 HCF Initiation Estimation ......................................................................................8-42
Figure 8-17 HTC Damage Process..........................................................................................8-44
Figure 8-18 Fracture Mechanics Models .................................................................................8-46
Figure 8-19 Fracture Mechanics Models .................................................................................8-47
Figure 8-20 Probabilistic Analysis of Turbine Blades...............................................................8-49
Figure 9-1 Evaluating Material Removal Limits .........................................................................9-8
xxvi
10305538
LIST OF TABLES
Table 1-1 Basic Organization of This Report .............................................................................1-2

Table 1-2 Overall Statistics Associated with Turbine Blade Failures, 1970–1981 ...................1-11
Table 2-1 Summary of Section 2 Contents ................................................................................2-1
Table 2-2 Mechanisms and Root Causes of Steam Turbine Blade Failures .............................2-2
Table 2-3 LCF Characteristics in RCA .....................................................................................2-12
Table 2-4 HCF Characteristics from Synchronous Vibration Used in RCA..............................2-16
Table 2-5 HCF Characteristics of Asynchronous Vibration in RCA .........................................2-18
Table 2-6 HCF Characteristics of Torsional Interaction in RCA...............................................2-20
Table 2-7 HCF Characteristics of PAA in RCA ........................................................................2-22
Table 2-8 SCC Characteristics in RCA ....................................................................................2-24
Table 2-9 CAF Characteristics in RCA ....................................................................................2-26
Table 2-10 Erosion (SPE or WDE) Characteristics in RCA .....................................................2-28
Table 2-11 HTC Characteristics in RCA ..................................................................................2-30
Table 2-12 TMF Characteristics in RCA ..................................................................................2-32
Table 3-2 Objectives for Each Phase of Investigation and the General Timing Involved ..........3-3
Table 5-2 Typical Compositions of Turbine Blade Alloys...........................................................5-7
Table 7-2 Alternatives for Planning and Management of Blade Failures ...................................7-4
Table 8-2 Material Properties Used in Steam Turbine RCA ......................................................8-3
Table 8-3 Additional Information Required to Support RCA ......................................................8-4
Table 10-1 Summary of Section 10 Contents ..........................................................................10-1
Table A-1 Checklists Provided in Appendix A........................................................................... A-1
Table A-2 Checklist of Pre-Inspection Action Items .................................................................. A-2
Table A-3 Checklist of Immediate Action Items When Turbine Is First Opened ....................... A-3
Table A-4 Checklist of Visible Inspection of Damaged Row and Evidence Related to
Installation ......................................................................................................................... A-4
xxvii
10305538
Table A-5 Checklist to Guide Identification of Basic Design Information .................................. A-5
Table A-6 Checklist to Guide Identification of Items Specific to the Blade and Row ................ A-6
Table A-7 Checklist of Relevant Operating Data to Assemble for Review ............................... A-7
Table A-8 Checklist to Guide Blade Material Analysis .............................................................. A-8
Table A-9 Checklist for Fractographic Examination of Cracks.................................................. A-9
Table A-10 Checklist for Evaluating Preliminary Evidence ..................................................... A-11
Table A-11 Checklist of Information Provided by the Operator to Support RCA..................... A-12
Table A-12 Checklist of Information Related to Steady Stresses ........................................... A-13
Table A-13 Checklist of Information for Frequencies and Dynamic Stresses ......................... A-14
Table A-14 Checklist of Information for Life Prediction and Probability of Failure .................. A-16
Table A-15 Checklist for Identifying and Selecting a Corrective Strategy............................... A-17
Table B-1 Case Studies of Blade Failures ................................................................................ B-1
xxviii
10305538
1
INTRODUCTION TO ROOT CAUSE ANALYSIS
This report on root cause analysis (RCA) of steam turbine blade failures was written for plant
managers and the operators of large steam turbine-generators to help them effectively resolve
problems caused when a rotating blade failure intrudes in the day-to-day operations of their
units. As applied in this report, the root cause investigation involves a methodology of up to
three phases of related activities that seek to identify the principal driving source that prompts
the damage mechanisms of fatigue, corrosion, erosion, or wear to result in a failure. In the
context of this report, failure is defined as the discovery of unexpected damage or wear that
sufficiently compromises the operator’s confidence that the unit can be returned to service
simply by reinstalling the original design.
A successful RCA is measured as the prevention of a repeat failure. Corrective action is defined
either as a permanent solution, which almost always involves replacing the original row of
blades, or an interim solution, such as removing the row, repairing damage to the blades, or
installing the original design. Interim solutions are used to gain time if design changes to the
blade are required. Modifications are adjustments made based on results of the RCA either to
correct a design deficiency in the original blade design or to mitigate a debilitating condition in
the stage that is related to how the turbine is operated. A design deficiency is signified by the
premature formation of cracks and is the result of excessive steady, dynamic, or thermal stresses
that result in accelerated low cycle fatigue (LCF) or high cycle fatigue (HCF) damage to the
blades. Finally, the phases of RCA identified within this report are intended for pursuit by a
separate, plant-assigned team of specialists who are decoupled from the normal critical path
recovery activities associated with returning the unit back to service.
Organization of This Report
A summary of the section topics is presented in Table 1-1. It starts with a list of immediate
activities that should be undertaken whenever a failure occurs. Basic information is itemized that
should be obtained once the turbine is open for inspection. This separates the subsequent efforts
pursuant to an RCA from the logistical (critical path) recovery activities for restoring the unit to
active service. The investigative activities are then organized into a methodology of three basic
phases: (1) assembling evidence, (2) conducting an engineering analysis, and (3) developing a
corrective solution (see Figure 1-1). These phases are further divided into a flowchart of tasks,
both technical and non-technical (see Figure 1-2). Within each of these tasks, evidence is
collected to make a reasonable diagnosis of the root cause.
1-1
10305538
Introduction to Root Cause Analysis
Table 1-1
Basic Organization of This Report
Introduction. Presented are the objectives of a root cause investigation; the role of the operator;
the timing, logistics, and intermediate strategies; what is reasonable to expect from the original
1
equipment manufacturer (OEM); and the independent technical expertise that might be needed to
assist the plant.
Damage Mechanisms and Their Root Causes. The majority of turbine blades fail as a result of
2 a limited number of causes. Section 2 is a reference to supplement the interpretation of
information obtained in the different steps of the investigation.
Methodology and Objectives. A standardized methodology of three phases provides a roadmap

3 for managing a failure investigation from start to finish. At each phase, a decision to stop and
execute a strategy or proceed with the investigation is possible.
Immediate Response and On-Site Forensics. Describes what information should be secured
4 prior to and after the unit is opened for inspection. Itemized lists are presented relating to visible
condition of the blades, details of the design, and relevant operating data to be assembled.
Off-Site Forensics: Metallurgical Examination. Describes how samples of the blade material
5 and fractured or damaged parts are analyzed. Itemized lists relate to the chemistry, properties,
and evidence on the fracture surface sought in the examination.
Generic Data. Discusses how to decide if industrywide or published experience on a blade

6 replacement can be applied directly to the failure at hand or whether the root cause investigation
should proceed to its logical conclusion.
Weighing Phase 1 Evidence. Reviews how the assembled evidence is used to determine
7 whether to replace in-kind, to accept an available field repair or alternative replacement design, or
to proceed with the remaining steps of the investigation.
Phase 2: Engineering Analysis. Discusses how to use an independent analysis when the root
8 cause or proposed strategy from the initial level of investigation remains inconclusive. Four basic
activities of modeling, frequency and stress analysis, life prediction, and validation are reviewed.
Phase 3: Identifying a Replacement Strategy. Offers guidance on whether the RCA results
mean that a new design is required, whether a repair can be used to salvage superficially
9
damaged blades, or if alternative inspection criteria might be applied to manage the remaining life
of the blades.
Post-Failure Guidance. Offers additional guidance on using the knowledge and experience
10 gained from a blade failure to improve the chances of procuring and installing reliable blades from
OEMs or third-party suppliers.
1-2
10305538
Figure 1-1
The Three Fundamental Steps in a Turbine Blade Failure Investigation
1-3
10305538
Figure 1-2
Flowchart to Guide Root Cause Investigation of Turbine Blade Failures
1-4
10305538
General guidance is presented on how to decide whether to execute a permanent or interim

strategy and/or whether to proceed to the next phase of investigation. The section concludes with
a discussion on historical data that reflect the risks associated with proceeding to operate with an
unverified solution.
The remaining sections (Sections 2–11) discuss each phase and task of the methodology
presented in the flowchart in Figure 1-2 in specific detail. A section devoted to each one of these
tasks features an itemized list of specific questions and background discussions to further explain
why the answer (and information on which it is based) is relevant toward making a decision to
conclude the investigation or proceed through the remaining phases. The questions identified in
these sections are summarized as tables in Appendix A to provide a simplified blueprint and
checklist for each step that corresponds to the detailed discussions, tables, and figures offered
within the broader report.
First Response: Recovery Versus Investigation
Experience has shown that the logistics of getting a unit back on-line will invariably conflict
with the time required to complete a thorough root cause investigation. When an RCA is being
considered, the plant manager should be advised from the start that the investigative process is
one that will involve weeks, not days, to reconcile. It should therefore be part of the recovery
plan to have the investigative work performed in parallel with but separate from the logistical
activities required to order and replace damaged parts.
If a failure occurs, the following three activities should be immediately undertaken once the unit
is down and opened:
1. Make an extensive photographic record of the turbine when it was first opened.
While the blades are still in the rotor, rub and witness marks should be recorded both
photographically and through sketching. Each photograph should have reference marks so
that the location being photographed can be ascertained later. Video is an excellent tool for
recording overall rotor condition, but it should not totally replace high-quality digital images.
Note any patterns of cracking or failure around the circumference of the disk. Make a count
of the number of blades in the row and their grouping arrangement. If the row involves
blades that have interlocking or interfacing covers, document the state or condition of any
interfacing surfaces between portions of the blade that connect them with adjacent blades. If
gaps are meant to exist at zero revolutions per minute (rpm), record the number present. If
covers are meant to be pre-loaded, note any gaps that are apparent or blades that appear to be
loose. If the blades do not contain visible markings that can be used to determine the relative
location of the blades after removal, each blade should be marked with a paint pen or other
permanent marking method so that the relative position of the blades can be determined after
removal.
2. Secure and protect samples of failed blades or remaining parts removed from the turbine.
These will be needed so that they can be sent in an uncontaminated condition for detailed
metallurgical investigation and fractography. Do not let these be handled carelessly, and do
not allow them to disappear from the site without authorization. If the blades are to be
1-5
10305538
shipped to a laboratory, they should be individually wrapped for shipment to prevent cross-
contamination and damage to the blade parts. If there are fracture surfaces, mating surfaces
should not be mated for shipping. Mating fractures should never be placed back in contact
with one another because this will result in damage to the delicate fracture features. The
fractures should be covered with bubble wrap, rags, or a soft packing material to prevent
damage to the surface. Steam turbine blade materials will not rust; thus, the fracture surfaces
should not be covered with grease or protective coating. However, desiccant should be added
to the shipping container to prevent moisture from reacting with in situ blade deposits and
continuing any acidic corrosion reactions.
3. Obtain basic dimensions of the rotor disk profile and the blade row’s outer tip diameter.
In addition to a sample of the failed blade that can be reverse engineered, these are critical
details on the geometry and dimensions that will facilitate the accuracy of any finite element
model (FEM) that might be prepared to perform an engineering analysis of the blade’s
stresses and operating natural frequencies. If sample pieces of covers and tiewires are not
recovered, also take dimensions from blades in the turbine.
As illustrated in Figure 1-3, each activity is meant to capture a certain piece of critical evidence
that might no longer be available after repairs are instituted on the damaged row and/or turbine.
1-6
10305538
Figure 1-3
Critical Information to Be Immediately Secured When Failure Occurs
Immediate Recovery Scenarios
As summarized in Figure 1-4, one of three recovery scenarios is likely to be executed to quickly
deal with the row that failed: replace the row, repair it, or remove it.
1-7
10305538
Figure 1-4
Three Basic Recovery Scenarios Involving Blades
1-8
10305538
The relationship between these three immediate action scenarios to the objectives of the larger
RCA is as follows:
• Scenario 1: repair row. Field repairs are sometimes possible if the damage is confined to a
portion of the blade—generally, the tip. Cutting off damaged covers or eroded portions of
blades and corresponding sections 180º apart has been successfully performed in situ to
address situations where a maintenance outage is scheduled in the immediate future.
Examples are presented in Appendix B. However, when such interim strategies are
employed, this should be recognized by the plant as a decision not to pursue all three phases
of the RCA in the traditional sequence. Rather, it jumps directly to the second phase and can
sometimes move completion of an FEM into the critical path associated with recovery action
items, particularly if low-pressure (LP) blades are involved. The model is used to predict
when changes to the original mass and stiffness of the bladed-disk assembly might
dangerously increase local stresses or affect the original tuning of the blades used to prevent
a condition of resonant vibration.
• Scenario 2: remove row. Removal of a damaged row and installation of a pressure plate
have been used in the past to buy time while a permanent fix is designed. It is not usually
necessary, except in cases where there is a concern that the disk attachments might have
become so weakened that it is questionable if they can retain the blades. In this third
scenario, the replacement strategy might involve changes that extend beyond the blades to
the rotor/disk. This action has severe penalties associated with it. First, time is required for
plates to be designed and manufactured. If used to supplement a last stage row of 30-inch
(30-in.) (762-millimeter [762-mm]) blades, the plates are not likely to be cheap. The other
more obvious penalty is the cost of operating at a reduced output. In this scenario, there is an
implied commitment by the plant to either refurbish the stage or possibly replace the entire
rotor. Given the capital investment of either, an RCA should still be seriously considered so
that the results can be used as a basis to select and/or qualify the replacement.
• Scenario 3: replace row. If the failure were genuinely a random event, a replacement in
kind would be the correct and permanent solution. If the failure has historical precedence, a
modified design might also be available and represent a permanent solution. However, until
the RCA is concluded, any replacement should be considered an immediate response and
interim strategy. In the chance that an upgraded design happens to be available for
installation by the OEM, this should still be technically treated as an interim solution
depending on the final results of the RCA. It is the responsibility of the investigative team to
ensure that the replacement is still verified after the unit is restored to service to ensure that
the original problem has been corrected. If the outcome of the investigation is to conclude
that changes to the original design have not extended the service life significantly, the results
can then advise the operator how much time the replacement has provided, and this time
frame is then used to engineer a permanent solution.
It is strongly recommended that the team concentrating on turbine recovery be separate from a
second team that does the RCA. Otherwise, the RCA team members can be put under enormous
pressure to cut corners and not follow the process that has been identified to solve the problem.
1-9
10305538
Moving from an Interim to a Permanent Solution
Referring back to Figure 1-2, the RCA can end within the first phase of investigation if it can be
determined that the failure was not related to a weakness of the blade design. Such a
determination is normally made for one of the four following obvious reasons:
• Review of the operating history revealed an extreme off-design condition of prolonged
operation, that is, an obvious anomaly (see Section 4).
• Inspection revealed that failure was initiated by a high-energy impact of some sort, that is,
foreign object damage (FOD), water hammer, and so on (see Section 4).
• Evidence shows that the blades were damaged as a consequence of failure in some other
system of the main turbine (see Section 4).
• Metallurgical analysis found that the material was deficient in composition or in mechanical
properties (see Section 5).
When such a determination is possible, a replacement in kind is normally available to expedite
the restoration of the unit unless the blade is a particularly old design. If the failure has some pre-
history (see Section 6), the OEM might also have a modified design readily available that can be
used. If the engineering that the OEM provides is sufficiently detailed and persuasive to accept
on face value, the replacement with an upgraded design might also be considered as permanent
and the RCA terminated in a single phase. Weighing the merits of evidence gathered in Phase 1
is addressed at length in Section 7.
However, if the results from the first phase of investigation are inconclusive, inconsistent, or not
definitive enough to explain why the blades failed, it is recommended to proceed through the
second phase of engineering analysis (see Section 8). The essence of this phase is identification
of localized stresses—steady, dynamic, and/or thermal—in critical locations, which ultimately
determine how much service the blade design can be expected to provide. Maximum stresses
govern when and where cracks will first form and how quickly they will propagate. They are
generally concentrated in the fillets, notches, holes, or grooves found in any given blade design.
Knowing the operating frequencies of the bladed disk is of importance only in that they exert a
profound impact on the magnitude of dynamic stresses resulting in different locations on the
blades. If the OEM is unwilling to demonstrate the effectiveness of modifications made in an
upgraded design, engineering analysis is recommended to verify its potential reliability (see
Section 9). In either scenario, the replacement (in-kind or upgraded) would remain an interim
until authenticated by the results of a Phase 2 type of engineering analysis.
Because the results of an RCA normally follow the critical path of recovery efforts, there is
always the tendency to assume that an in-kind or upgrade replacement is permanent, therefore
making any further investigation a waste of time and money. If the failure is genuinely the
random consequence of external forces, and not a problem of the design itself, this gamble with
the unit might be successful. However, the cost to correctly diagnose the root cause of a turbine
blade failure should be weighed against the risk of a repeat failure. A measure of these odds
associated with steam turbine blade failures was taken by the Electric Power Research Institute
(EPRI) two decades ago, but the painful lessons experienced by the industry at that time remain
as valid today.
1-10
10305538
Cost and Risk of Repeat Failures
The risks of not following through with a conclusive identification of the original root cause are
tangible, particularly where a replacement in kind is used as an expedient. In 1981, EPRI
sponsored an industrywide survey of steam turbine blade failures as a response to the general
perception that with the dramatic increase in a new generation of steam turbine-generators during
the 1960s and 1970s, there had been a commensurate increase in reported failures. The survey
focused on a population of 495 nuclear and fossil units of 300 megawatts electric (MWe) and
larger operated within the United States [1]. Published in 1985, the results documented the
perceived increase in failures and the cost to the industry in both replacement parts and lost
power production (see Table 1-2).
Table 1-2
Overall Statistics Associated with Turbine Blade Failures, 1970–1981
Item Forced Scheduled Total

Number of outages 156 267 423
Downtime in hours 89,205 127,175 216,830
Blade replacement costs $45,657,000 $41,799,000 $87,456,000
Estimated lost power cost $562,271,000 $602,461,000 $1,164,732,000
Average cost per outage $3,896,974 $2,412,959 $2,960,255
More revealing was the common responses to many of the failures, which were to presume that
they were a random event and to replace the blades using the same design. This was also the
most expeditious action that could be undertaken because replacements could be made
immediately available if the plant were willing to pay an expediting charge. As shown in Figure
1-5, when this strategy was adopted, a repeat failure occurred 45% of the reported time. In some
cases, failures persisted through a series of as many as nine outages.
1-11
10305538
Figure 1-5
Results from 1981 EPRI Blade Failure Survey
1-12
10305538
Several lessons from the survey still hold true today, including the following:
• Many failures were originally unrecognized because they were classified in narrowest terms
as not resulting in a catastrophic separation leading to a forced outage. Instead, evidence of
premature, repeated cracking was treated as maintenance rather than a design problem, with
regular inspection intervals being reduced to two years in some reported incidents.
• The replace-in-kind and multiple failure scenarios reflected a trial by error approach at a time
when analytical methods for examining the critical features of new designs were becoming
available to the designers but beyond the means of most operators. In certain cases, once it
became clear that the problem had reached serious proportions, such advanced techniques
were applied and the problems were resolved.
• Problems associated with specific blade designs were often not unique to any one operator
but frequently appeared within a number of fleets, regardless of the unit’s configuration or
power output. In other words, the problems were often stage-specific, signifying that features
of the design played a predominant role.
• A vast majority of the problems were associated with the last two rows of the LP turbine. For
L-0 rows, this susceptibility was in part attributed to the scaling up from a maximum 26-in.-
long (660-mm-long) blade extending up to 52 in. (1321 mm). For the L-1 rows (and the third
and fourth rows in nuclear units), the wet steam environment compounded the problem
caused by scaling.
• In high-pressure (HP) and intermediate-pressure (IP) turbines, the control stage and first
reheat rows were also predominantly affected. This is more understandable because both
stages are exposed to the high temperatures and pressures of the steam path. Reheat stages
can rely on longer blades. A number of original problems were also due to erosion from solid
particles entrained within the steam.
In the 25 years since the survey was completed, the majority of these large units continue to
serve the power generation industry. Blade failures continue to occur, predominantly affecting
the same rows. The reasons that failures persist are actually not complicated or unexpected; they
include the following:
• Many of the initial problems involved marginal detuning of critical blade modes, which were
eventually corrected. Today, the role of HCF has re-emerged in regards to the damage
tolerance a design can accommodate. Blades that have served reliably for 15–20 years now
unexpectedly fail because pitting, corrosion, FOD, and so on initiate a crack in a location
where even a nominal vibratory stress can take and propagate it to failure.
• Because many of the problems occurred in stress concentrations formed by tenon covers and
tiewire/lashing lugs, upgrades were designed that eliminated these weak links. This involved
integral interlocking shrouds, snubbers, and so on to form the row into a continuously linked
system, or reliance on completely freestanding blades. Either represents a radical departure
from the older linking strategies. While the length of the blade may be correctly represented
as essentially unchanged, the dynamic properties will be substantially changed and should
therefore undergo the types of qualification tests their predecessors underwent.
1-13
10305538
• With deregulation and the advancement of computational fluid dynamics (CFD), a premium
has been placed on replacements that are more efficient than their predecessors. Efforts to
streamline the flow throughout the steam path have been particularly focused on the last few
stages of the LP turbine. These stages are where alterations to the airfoil can often produce
the biggest return in terms of power output, but they are also the rows most susceptible to
fatigue problems and erosion, corrosion, or wear. In particular, a favored method of upgrade
has been to increase the annulus of the last stage by extending the airfoil by one or more
inches. Again, although this additional length may sound small, it can profoundly alter the
well-known dynamic properties of the preceding design and increase the stresses in the root
attachments because of the additional mass.
The principal lesson for the twenty-first-century operator is that one can reasonably expect that
blade failures will persist, either as the steam environment degrades traditional designs or as new
designs are introduced to produce more power. The objective of the root cause investigation is to
preempt this viscous cycle. As illustrated by the statistics, compromising on the scope of
investigation for the sake of expediting the logistics associated with returning the unit to service
is not correcting the problem but simply resetting the clock. In an era of deregulation, however,
operators do not have the same luxury of treating any incident as a random event, nor are they
advised to do so. In a competitive market, no plant can afford to cope with repeat failures
extending through multiple outages, as was done in the 1980s.
Summarizing the Objective for an RCA
The principal reason for undertaking a root cause investigation is to prevent a repeat failure. A
secondary goal is to assist in defining a practical strategy for restoring the unit to reliable service
as soon as feasibly possible. It is important that these priorities not be reversed. Confidence in a
repair or replacement strategy is achieved by understanding (as well as possible) how aspects
contained within the three basic parameters of design, manufacturing, and operation might have
interacted to result in the necessity to prematurely repair or replace the original rows of blades.
From an operator’s perspective, a failure can be considered premature based on a perceived need
to replace blades in fewer than 100,000 hours of continuous service or 5000 starts. This need is
normally signified by the suspicion or appearance of cracks that were caught by inspection or
resulted in a catastrophic failure. Historically, many successfully designed rows have continued
to operate reliably for up to 30 years.
Because all blades vibrate while in service, the formation of cracks is generally considered to
signify the beginning of the end, or the point where it can be reasonably assumed that the
inherent reliability expected by the original blade’s designer could be seriously compromised.
Mechanically, the constant presence of a dynamic stress means that once a crack has formed,
there is a significant increase in the chance for HCF to take control and propagate it to a critical
size where rupture becomes inevitable. Estimations of the crack initiation life from both LCF and
HCF are therefore fundamental measures in deciding whether a design modification is called for.
Running to final rupture while the row is in service truly represents a catastrophic situation,
particularly if a significant portion of a large turbine blade is released at full speed. The resulting
unbalance can literally wreck the unit before it is brought to a halt.
1-14
10305538
This does not mean that the risks of operating damaged blades cannot be managed for a limited
period, that is, to the next scheduled outage. In some cases, blades are replaced in kind with
available spares with the understanding that the intent is to reset the clock and gain time until a
permanent solution can be made available. In situations dealing with erosion, corrosion, or FOD,
the risks can be calculated or defined in probabilistic terms to provide some sense of how much
longer damaged blades might remain in service. Neither scenario, however, constitutes a final
solution, but represents an interim strategy. The primary objective for following these guidelines
remains to correct the original problem, not cope with it.
In summary, the intention of this report is not to teach blade design, metallurgy, manufacturing,
or analysis. Instead, it shows how a root cause investigation proceeds through a standardized
methodology whereby initial speculation is systematically replaced with hard evidence. From
this evidence, reasonable conclusions can be drawn. In the sections dedicated to each phase,
emphasis is placed on what facts should be assembled and how they can be used to identify the
mechanisms and root causes from the list of those that are most often responsible for blade
failures in steam turbines.
The guidance offered in this report is a consolidation of statistical information assembled from a
review of nearly 500 incidents of steam turbine blade failures, complemented by experience
drawn from more than 400 separate independent technical investigations performed over the last
quarter-century. Like the turbine itself, first introduced by Parsons at the turn of the nineteenth
century, the basic methodology and the root causes it is designed to identify have actually
changed little in the last 25 years.
The most dramatic change has not been in the technology or features of modern blade designs,
but in the affordability of sophisticated investigative tools, which are now available to the public
at large. Affordability has resulted in the following developments:
• Calculations of bladed disk structures, which originally required supercomputers using in-
house finite element programs supplemented by expensive field tests, are no longer the
exclusive prerogative of the turbine manufacturers.
• Metallurgical analysis with scanning electron microscopes (SEMs) can provide critical
details left on the fracture surfaces of failed parts.
• Assembly of relevant operating data has been facilitated with the digital recording systems
now installed on most large steam turbines.
As will be shown, these tools provide enough critical details on which a reasonably accurate root
cause diagnosis can be based. Once the root cause is correctly identified and understood, the
original analysis further provides a baseline of critical engineering information to establish the
effectiveness of changes in design, materials, or operating patterns. If pursued in this way, the
operator and plant can prevent the plant from becoming a repeat statistic.
1-15
10305538
10305538
2
DAMAGE MECHANISMS AND THEIR ROOT CAUSES
Table 2-1
Summary of Section 2 Contents
Subject:
Most blades fail from one of a limited number of causes. This section features a limited review of the
most common mechanisms that an operator is likely to encounter in pursuing a blade failure
investigation. It is meant as a reference section, where the damage mechanism and its sources are
defined in preparation for subsequent sections, which deal with the metallurgical and engineering
analysis of blade failure.
Topics Covered:
2.1 – Cause Elimination Process
2.2 – Failure Sequence
2.3 – LCF Caused by Start-Stop Cycling
2.4 – HCF Caused by Blade Vibration
2.5 – Stress Corrosion from Chemical Attack and Steady Stress
2.6 – Corrosion Fatigue from Chemical Attack and Dynamic Stress
2.7 – Erosion from Water Droplets or Solid Particles
2.8 – Material Creep from High Temperature and Steady Stress
2.9 – TMF from Transient Thermal Gradients
2.10 – Other Mechanisms: Overload, Deformation, Rubbing, and Impact
Tables/Figures:
Table 2-2 provides a summary of the principal mechanisms and associated root causes that are
most commonly observed on steam turbine failures.
Tables 2-3 through 2-12 provide an illustrated summary of the evidence and characteristics
associated with each of the specific damage mechanisms listed in Table 2-1.
Figure 2-1 illustrates how the timing and failure sequence associated with different blade damage
mechanisms is typically reflected on the fracture surface.
Figures 2-2 and 2-3 illustrate the local strain and fracture mechanics (FM) concepts as they apply to
the estimation of crack initiation and crack propagation of steam turbine blades.
2-1
10305538
Damage Mechanisms and Their Root Causes
Cause Elimination Process
Throughout the investigative process, evidence is gathered that is ultimately used to confirm or
eliminate one or more of the most prevalent root causes of turbine blade failures. As Table 2-2
indicates, the list is relatively manageable, with many of the primary mechanisms inherently
specific to the stage (environment) of the turbine in which they operate. Other mechanisms are
those where, for the most part, diagnosis of the root cause and actions to take are straightforward
and normally do not require further analysis to identify a corrective action. Two prominent
published resources referenced in Table 2-2 provide additional details on the mechanisms and
their consequences beyond the overview provided in this section.
Table 2-2
Mechanisms and Root Causes of Steam Turbine Blade Failures
Turbine Stage Principal Mechanism Root Cause Ref 1 Ref 2

V1,
All All LCF Start-stop cycling of unit
Section 1
V1,
All All HCF Synchronous vibration V2, 20-1
Section 5
V2, 21-1
Asynchronous vibration
LP L-0,-1 HCF
(flutter)
LP L-0 HCF System torsional resonance
Stress corrosion cracking
LP Wet Chemistry and steady stress V2, 26-1
(SCC) V1,
Corrosion-assisted Section 6
LP Wet Chemistry and dynamic stress V2, 24-1
fatigue (CAF)
Water droplet erosion V1,
LP Wet Excessive moisture V2, 27-1
(WDE) Section 3
High-temperature creep Temperature and steady V1,
HP, IP All(a)
(HTC) stress Section 4
V2, 16-1
(a) Thermal mechanical V1,
HP, IP All Thermal excursions
fatigue (TMF) Section 5
Solid particle erosion V1,
HP, IP Initial Steam quality and treatment V2, 17-1
(SPE) Section 4
V1,
HP Row 1 HCF Partial arc steam admission V2, 21:2
Section 5
2-2
10305538
Table 2-2 (continued)

Mechanisms and Root Causes of Steam Turbine Blade Failures
Turbine Stage Other Mechanisms Root Cause Ref 2 Ref 3

Overload Overspeed, transient events V2, 32-1 V1,
All All
Section 7
Deformation Axial/radial rubs V2, 32-1 V1,
All All
Section 7
High energy impact Foreign objects, water V2, 28-1 V2,
All All
Section 9
1
Turbine Steam Path Damage: Theory and Practice, Volume 2 [2].
2
W. Sanders, Turbine Steam Path, Volume 1: Turbine Steam Path Maintenance and Repair [3].
(a)
Rows that operate at temperatures >900ºF (482°C).
The process of elimination should start immediately. Substantial progress should be possible
once facts obtained from the operating records, inspection records, and metallurgical analysis
become available. After this initial paring down is complete, stresses obtained from a structural
analysis of the design are needed to further differentiate the remaining root causes from
secondary contributors. From this differentiation, an appropriate strategy is determined.
Often, operators are advised that their failures are a product of LCF, HCF, SCC, and so on. They
are then left to relate these damage mechanisms to the operational counterpart or source on their
turbine. This section is meant to bridge this gap. It begins with the most common mechanisms
and ends with those that are infrequent but known to have caused turbine blade failures. Review
of the mechanisms is confined to the context of resolving steam turbine failures, that is, how they
are related to the source or root cause to explain how they effectively compromise the structural
integrity of any particular blade design.
In the next section, the following items are noted:

• Elimination. Every mechanism is initially considered but not automatically evaluated in all
turbine blade failure investigations. The intention is to pare down the list to minimize
unnecessary time and effort spent pursing root causes that can be reasonably eliminated.
• Combination. Blade failures often are not the result of any single damage mechanism, but
occur from a sequence where the overall process of failure is dominated by different
mechanisms at different times.
• Process. Pursuant to this concept of sequence, the total life of a blade known to have failed is
divided for the purpose of consistency and discussion within this report into several phases—
incubation, initiation, propagation, and rupture.
• Material degradation. Nearly all failures end (or would end if not caught) in rupture as a
consequence of HCF. In the phase designated as propagation, it is defined as when an
existing crack extends to a depth where even a nominal (detuned) dynamic stress can take
control of the process. It is signified by when the threshold stress intensity—not the rupture
strength—of the material is exceeded.
2-3
10305538
The exception to the stress intensity threshold limit is when a failure occurs due to overload, that
is, an initial overspeed test. When an overspeed check is performed, the temporary additional
force overloads the remaining uncracked portion of the blade and ruptures the material. It is
therefore recommended that overspeed testing be performed during unit shutdown instead of
startup whenever possible, at higher temperatures, and with increased fracture toughness range of
the material.
2-4
10305538
Failure Sequence
Figure 2-1 shows four practical phases of crack development where different mechanism(s) tend
to dominate.
Figure 2-1
Timing and Failure Sequence Associated with Blade Damage Mechanisms
2-5
10305538
Incubation is used as a generic term to designate the period of service preceding the actual
formation of an initial crack. For SCC or CAF, this phase of pre-crack formation accounts for
chemical attack of the grains/grain boundaries as a result of the presence of corrosive products
that are deposited at some time in the service life of the blades. For LCF or HCF, this is the
period when the plastic strain or permanent deformation left from repeated cycles eventually
forms an initial crack at the surface of the material. For either form of erosion (WDE or SPE),
this phase would be the period when water droplets or particles removed material, forming
trenches, pits, or other physical distress to the material surface. In HTC or TMF, this would be
the period preceding the dislocation of grain boundaries in a manner where initial voids were
eventually formed. Under normal circumstances, any of these processes should take years.
Premature cracks from LCF, HCF, TMF, or HTC suggest that the material was substandard or
the design stresses were unexpectedly high. Premature corrosion cracking, erosion, or pitting
suggests a material-related problem or an operating-related issue that has unnaturally accelerated
these physical or chemical processes.
In the initiation phase, once a pit, notch, void, or other surface discontinuity has formed, it acts
like a micro-stress riser. If the surrounding stress field is high enough, the discontinuity will form
a true crack and start the second phase of initiation. Corrosion-assisted cracks tend to be
intergranular in nature, where fatigue cracks reflect a transgranular pattern as they extend
through the material. For turbine blades, crack incubation and initiation represent the safe portion
of total operating life. Well-designed blades will remain within these two phases for their entire
service life because of the safety factors and design margins that are present to accommodate the
damage caused during operation. A failed blade has moved into one or both of the remaining
phases.
Propagation is a reflection of the damage tolerance inherent within the design. Crack propagation
from SCC, LCF, TMF, or HTC usually involves a slow, deliberate, and relatively constant
extension of the crack. Further erosion or corrosion (pitting) does not normally play a significant
role in the propagation of cracks unless either condition is unusually severe—for example, the
rate at which the material is being physically removed exceeds the formation of surface cracks. If
initiation and propagation are both a result of HCF, the remaining life is rapidly consumed and
normally catastrophic as a result of the hundreds of cycles at which the modes naturally vibrate.
Rupture is the signature of a truly catastrophic failure when the remaining uncracked portion of
the blade was overwhelmed by the centrifugal stresses produced at rotating speed. Usually, it is
evident on the fracture surface as a clear demarcation from the smooth transgranular appearance
left by fatigue. In root failures of large steam blades, the relative proportion of the fracture
surface associated with rupture is often much larger than for crack propagation. This is a visual
indication of the tremendous forces that these blades must withstand.
In terms of quantifying the proportion of life consumed within each of the four phases, it is the
first phase (incubation) that is usually the most difficult to estimate. Most life prediction
techniques pick up the story at the second or third phase, that is, estimating how many hours or
cycle cracks are needed to initiate and/or propagate a crack based on the local stresses,
temperatures, and selected material properties. To evaluate LCF and HCF, crack initiation life is
estimated using the local strain approach and a strain-versus-life curve (see Figure 2-2)
2-6
10305538
developed from test specimens of the appropriate material [4, 5]. This is an important approach
to know because it can identify whether crack initiation is a direct result of excessive stresses
being produced as a function of the blade design itself. When applied with the results from the
structural analysis, a well-designed blade will produce estimated numbers of start-stop cycles
and vibratory hours of cumulative operation (representing both LCF and HCF initiation life),
which is more than sufficient to accommodate the actual cycles and hours a blade might
experience in 20–30 years of service. If the outcome is otherwise, it signifies a need for the
original design to be changed; that is, a replacement in kind will be equally vulnerable to the
original root cause of the problem.
2-7
10305538
Figure 2-2
The Local Strain Approach to Predicting Crack Initiation
2-8
10305538
Conversely, if the outcome indicates that in its undamaged (new) condition, the design is more
than capable of accommodating the LCF and HCF damage that it would have experienced based
on the hours and starts identified prior to failure. This would indicate that the root cause is a
damage tolerance issue and not design-related (presuming the material is not deficient). Because
crack extension is usually more deliberate, an FM approach is more appropriately applied to
evaluate the potential life that may remain due to CAF, erosion damage, or SCC in the third
phase associated with propagation. In an investigation, this is done by assuming that a crack has
been formed and assigning it an initial size or depth. Unless inspection records can show
otherwise, it is reasonable to assume an initial size on the order of 20 mils (0.5 mm) or less, as
this is below what current nondestructive evaluation (NDE) technology is able to identify with
any certainty.
In such scenarios, the threshold or end of the third phase is used as a conservative limit when
dynamic stress and HCF assume control of the propagation. Referring to the crack growth rate
(da/dN) curve of any blade alloy (see Figure 2-3), even at the most miniscule growth rates
associated with the threshold stress intensity of the respective material, the number of cycles that
blades vibrate (hundreds of times per second) will cause the crack to reach the fracture
toughness of the material and the final phase of rupture usually well before the next scheduled
outage can intercede. This explains why the majority of any crack surface involved in a blade
failure is normally evidence of HCF. Evidence of other participating damage mechanisms is
generally confined to a relatively small region around the origin, often less than 50–100 mils
(1.3–2.5 mm), depending on the stress within the immediate region.
2-9
10305538
Figure 2-3
The FM Approach to Predicting Crack Propagation
2-10
10305538
It should be mentioned that fatigue is the leading cause of steam turbine blade failures.
Determining the causes of the cyclic stresses that lead to fatigue is a complicated technical
challenge that is discussed fully later in the report (see Section 8). From a metallurgical
standpoint, a thorough study of the surfaces resulting from the fatigue damage can divulge much
about the cause. The study of these fracture surfaces is called fractography. An important issue
in fractography is determining the origin of the fatigue crack. Generally, the fatigue crack will
grow in concentric rings radiating from the crack origin. The location of the origin is usually
obvious during visual examination. The origin can then be located and studied in detail.
Corrosion can accelerate the fracture of turbine blades by two basic processes—anodic corrosion
(general wastage is present) and cathodically controlled cracking. In the first process (wastage),
the environment actually consumes the blade alloy. This is characterized as an anodic corrosion.
Anodic corrosion simply means that a metal is being dissolved from crystalline into ionic form.
This can occur globally, referred to as general corrosion, or locally as either under-deposit
corrosion or crevice corrosion. Visible metal loss is observed. This corrosion can lead to loss of
geometry (usually a drop in thermodynamic efficiency) or, more commonly, as discrete pits or
grooves that can act as stress raisers to initiate fatigue cracking. Once a crack is started, the
causative agent for the corrosion can accelerate the crack propagation rate. This mechanism is
typically called corrosion fatigue if the stresses are cyclic, or anodic SCC if the stresses are
steady-state.
In the second process (corrosion-assisted cracking), corrosion is occurring, but metal wastage is
not observed. The byproduct of this corrosion—hydrogen—is present and can accelerate the
cracking. This mechanism is cathodic SCC or, in some cases, cathodically controlled corrosion
fatigue. The exact mechanism is not always discernible, thus both processes are included in the
term corrosion-assisted cracking.
LCF Caused by Start-Stop Cycling
EPRI References on LCF

Additional discussion of LCF in steam turbine blades can be found in Sections 20, 21, and 22 of
Definition
LCF is one of the two fundamental mechanisms (along with HCF) that should always be
suspected until proven otherwise (see Table 2-3). At full speed, 50-hertz or 60-hertz blades are
subjected to substantial and significant centrifugal forces. Thus, each time a unit ramps up and
shuts down, the centrifugal forces load and unload the blades. This results in a magnitude and
distribution of stress that remains fairly constant or steady during full speed. Maximum steady-
state stresses naturally occur in radii or sharp transitions of the blade geometry, that is, in the
roots, transition of the platform to the airfoil, in snubbers or tiewires, and in shrouds. Because
they are primarily a consequence of centrifugal force, the highest stresses are regularly found in
the root attachments (fir tree fillets or finger pinholes). Typically, LCF occurs within 103 cycles
at stresses approaching the yield strength of the material.
2-11
10305538
Table 2-3
LCF Characteristics in RCA
Relationship of the Evidence

Basis Graphic
to the Mechanism
Evidence of regular beach marks

on a microscopic or macroscopic
Evidence unique to LCF
scale, confined around the crack
origin.
Similar cracks on multiple blades in

Evidence usually pointing to locations where maximum steady
LCF stresses occur, such as holes and
fillets.
Cracks formed in regions of lower

Evidence that LCF is not the steady stress than found in
cause locations such as root attachments
or tenon covers.
Stress corrosion and creep cracking

are also promoted by steady stress
Mechanisms confused with LCF
and tend to affect multiple blades in
the row.
Blades are not operated in a wet or

When to suspect that LCF is the
high-temperature environment.
cause
Crack growth is transgranular.
2-12
10305538
Damage Estimation
To estimate the damage caused with each start-stop cycle, analysis calculates the steady stress as
input to an LCF model to and from a hysteresis loop of strain versus cycles for a completed
start-stop cycle. To properly assess the maximum strain range produced, the cycle should define
the amount of plastic yielding (if any), the residual stress produced as a consequence, and the
mean stress of the cycle (see Equation 8-1). The cycle should normally include the permanent
consequences of an initial overspeed that most blades experience. With the appropriate material
fatigue properties derived from specimens cycled under controlled laboratory conditions, this
start-stop cycle can be conveniently equated to the number of turbine start-stop cycles (see
Equations 8-4 and 8-5) in terms of predicting when cracks might initiate in localized regions of
stress concentration. The number of start-stop and overspeed cycles is the operating parameter
that can be related to the mechanism of LCF as a root cause.
Design Factors and Issues
For most steam turbine blades, the root attachments of the largest, longest LP blades should
allow tens of thousands of predicted cycles, or well below the several hundred cycles a large unit
will normally experience in a lifetime of baseload operation. However, on older designs, the
damage produced from high steady stresses would not be apparent if the start-stop cycles were
limited to 10–15 per year. LCF can be made a problem if the unit is over-cycled. Some large
steam turbines meant to operate with blades designed for baseload service subsequently used for
peaking service might unexpectedly develop cracks.
HCF Caused by Blade Vibration
EPRI References on HCF
Additional discussion of HCF in steam turbine blades can be found in Sections 20, 21, and 22 of
Definition
HCF is always a natural suspect when a blade fails. The challenge is to determine if it is a root
cause or secondary contributor. It can be the exclusive mechanism responsible for all four
phases of crack development; in other situations, HCF will play a role in the sequence that
ultimately leads to a failure. Vibration is the source or root cause of HCF in turbine blades.
Obtained from a design analysis, the amplitude of dynamic stress is used in conjunction with the
aforementioned steady stress at the location of cracking to define the magnitude of the strain that
occurs during continuous periods of operation. The appropriate material fatigue properties are
derived from cycled specimens under controlled conditions. Initiation life from HCF can be
equated to the hours of operation preceding a failure (see Equation 8-6). HCF failures typically
occur within 108 cycles or fewer at low stress values, below the yield strength of the material.
2-13
10305538
Damage Estimation
To estimate HCF initiation life requires knowing the frequency of vibration for the mode
involved. In turn, this produces a reasonable estimate of the magnitude of dynamic stress (see
Equations 8-2 and 8-3). As a rule, vibratory stresses in a well-designed blade are generally on
the order of a few kips per square inch (ksi) or less. This results in an estimated HCF life of
more than enough cycles to accommodate the vibratory damage over 20–30 years of continuous
operation. It is when stresses approach 8–9 ksi (55–62 megapascals [MPa]), particularly in
regions where they coincide with a high steady stress (which acts as a mean), that they can pose
a serious threat to the integrity of the design.
The purpose of this report is not the complex subject of blade dynamics, but instead its
consequences. In an RCA of turbine blades, the two principal root causes responsible for most
HCF failures are referred to as synchronous and asynchronous, each of which can be directly
related to a different operational source. In a pure HCF failure, synchronous vibration is
generally referred to as resonance and asynchronous vibration is often called flutter. In either
scenario, the source is derived from the fluid forces imparted by the steam to the rotating airfoils
and manifested as a dynamic stress, which promotes and controls the amount of damage caused
by HCF. Referred to as stimulus or excitation, the sources of each are distinct, and therefore can
be used to differentiate the root cause at the completion of an investigation despite the fact that
all are a consequence of the same mechanism.
Different Sources
Distinct from other blade damage mechanisms, HCF can be related to several different sources
or root causes depending on the type of vibration involved. Related to steam turbine blade
failures, there are generally four sources of vibration that are of concern at the start of an
investigation: (1) synchronous vibration, (2) asynchronous vibration, (3) torsional system
vibration, and (4) periodic vibration from sequential steam loading. Of the four types, the first is
common to all blades. The second and third predominantly afflict L-0 blades. The fourth is
exclusive to control stage blades that rely on partial arc admission (PAA).
HCF Caused by Synchronous Vibration
Because of unavoidable physical non-uniformities of the turbine, such as casing splits, steam
extractions, and so on, dynamic forces are produced within the flowing steam at frequencies
synchronous to the rotor speed. For steam turbines, these per-revolution (per-rev) forces occur at
frequencies that are multiples of 50 hertz and 60 hertz, sometimes referred to as engine orders.
During operation, the blade modes respond to these synchronous forces at their natural
frequencies of vibration. The response of a blade (and dynamic stress that results) is
predominantly affected by the proximity of the mode’s frequency to the nearest per-rev forcing
and mode shape.
2-14
10305538
For LP turbine blades, synchronous vibration is usually a concern for any blade modes
(displacement that the structure assumes) whose frequencies fall within the first 10–12 engine
orders or harmonics of rotating speed, that is, 1x, 2x, 3x, and so on.
For HP blades, the principal modes of concern are those near the nozzle passing frequency
(NPF).
Definition
Resonance describes a condition of maximum response and maximum dynamic stress. It is

therefore a less technical term that is often used to describe the root cause under the more
generic umbrella of synchronous vibration. In design analysis, its counterpart is the non-
resonant or detuned response of the blade. If calculated and plotted, the detuned stress conforms
to the same pattern or distribution of stress for the particular mode of vibration that is involved,
but at a reduced magnitude relative to the resonant condition.
Design Factors
To add further perspective of the importance placed on whether a mode operates in a resonant
condition or not, in a detuned condition, dynamic stresses are generally calculated to be on the
order of 1 ksi (7 MPa) or less. In a resonant condition, these stresses can be increased by as
much as 10 to 20 times. For typical blade alloys, dynamic stress approaching 8–10 ksi (55–
60 MPa) in locations of high steady stress can initiate an HCF crack in a relatively short period
of operation. Hence, if a sufficient margin of detuning is not maintained, the risk of an HCF
failure is high. Features of HCF caused by synchronous vibration are highlighted in Table 2-4.
2-15
10305538
Table 2-4
HCF Characteristics from Synchronous Vibration Used in RCA
Relationship of the
Basis Graphic
Evidence to the Mechanism
Rapid crack growth leading to failure in

days, weeks, or months. Damage appears
Evidence unique to HCF
as increasingly distinctive bands (beach
caused by resonance
marks) emanating from the crack origin up
to the point of rupture.
Pure HCF failures that occur in the airfoil or

failures involving a select number of blades.
Evidence usually pointing to
Propagation is transgranular in appearance,
HCF caused by resonance
usually with no evidence of marks left by
oxidation of the exposed crack surface.
Cracks grow unusually long without

fracturing, indicating that dynamic stresses
Evidence that HCF from
are relatively low. Metallurgical examination
resonance is not the cause
indicates that cracks originated in pits,
fretting marks, or other surface defects.
Cracks due to HCF from asynchronous

Mechanisms confused with sources (flutter), torsional system
HCF from resonance interaction, or partial arc operation
(exclusive to HP control stage blades).
Blades fail in a very short period of operation

after they are installed. Failures are
When to suspect that HCF
selective, favoring one or two blades closest
from resonance is the root
to resonance. Cracks initiate in locations
cause
other than locations of maximum steady
stress.
2-16
10305538
HCF Caused by Asynchronous Vibration
Definition
HCF damage whose root cause is asynchronous vibration involves a more direct interaction of
the blade airfoil with the flow around it. It is a transient condition of vibration that occurs in
steam turbines during limited periods of operation that might be termed off-design—for
example, at the extremes of loads and exhaust pressures from a turbine’s designed state line.
From the perspective of investigating turbine blade failures, asynchronous vibration is a generic
description of the response that differentiates it from its counterpart related to the per-rev forces.
Flutter is used colloquially to describe the root cause for asynchronous vibration of blades,
although research into literature on the phenomenon will quickly produce myriad technical
distinctions, including buffeting, unstalled flutter, stalled flutter, and rotating stall. Stalled
flutter is associated with low-load/high-backpressure combinations, and unstalled flutter is
associated with high-load/high-backpressure conditions. In the era of deregulation, stalled flutter
is becoming more common. At partial load operation during non-peak periods of power
consumption, the off-design operating conditions can result in the incidence angle of the steam
flow directed to the individual profiles that form the stacking line of the blade airfoil becoming
negative. Because LP blades are tapered and twisted, this condition tends to start at the tip and
extend down the airfoil as the conditions worsen. As this deviation in the incidence angle
increases, the flow on the pressure side of the airfoil tends to separate, stall, and recirculate. At
some point, the vortices approximate to the airfoil will produce sufficient excitation to cause it
to react or respond with a significant and sudden increase of vibration, akin to the sudden
flapping a stop sign exhibits in swirling winds.
Although transient in nature, the dynamic stresses produced can become very high, depending
on the amount of interaction that occurs between the airfoil and the flow. As a source or cause of
blade failures, the danger posed by asynchronous vibration results because a significant amount
of HCF damage can be produced in a relatively short time. Because the damage mechanism is
HCF, the features left on the fracture surface can appear the same as if they were caused by
resonant (synchronous) vibration. This possible confusion would primarily occur in failures of
an L-0 row, where operating records found events that approached or exceeded the allowable
envelope, and where the damage was from a relatively small number of prolonged events. If the
historical record is not definitive enough to rule out the occurrence of such events and the
fractographic features are similar to a resonant fatigue failure, the results of a frequency analysis
remain the means to determine which type of vibration is most likely responsible. In other
words, if the modes are well detuned, asynchronous vibration is the next likely source of HCF to
naturally consider.
Features of HCF caused by asynchronous vibration involved with a steam turbine blade failure
are highlighted in Table 2-5.
2-17
10305538
Table 2-5
HCF Characteristics of Asynchronous Vibration in RCA

Basis Graphic
to the Mechanism
If caused by one or more prolonged

events, evidence of regular beach marks
on a macroscopic scale and fatigue
Evidence unique to HCF from
striations on a microscopic scale,
asynchronous vibration
confined around the crack origin. If due to
intermittent events, marks might be
regularly spaced.
Similar cracks affecting multiple blades

forming in locations corresponding to
HCF from asynchronous
regions of maximum steady stresses
vibration
occur, such as holes and fillets.
The unit has no record of operating in a

Evidence that HCF from
condition where the load and
asynchronous vibration is not
backpressure limits imposed by the OEM
the cause
are approached or were exceeded.
Cracks due to HCF from synchronous

Mechanisms confused with
sources (resonance), torsional system
HCF from asynchronous
interaction, or partial arc operation
vibration
L-0 blades are involved. Cracks do not

originate from pits. Crack growth is
When to suspect that HCF from
transgranular. The unit is periodically
asynchronous vibration is the
exposed to operating under conditions of
root cause
low load/low backpressure or high
load/high backpressure.
2-18
10305538
HCF from Torsional System Interaction
Definition
A third source of HCF associated with dynamic stress is produced by the cause referred to as
system torsional interaction or system resonance. This is a rare cause of blade failures, but it has
been known to happen. Although the mechanism and damage left on the fracture surface are
from HCF, the source of vibration is electromechanical and not associated with the steam flow.
Vibration is caused by a sympathetic reaction of the blade row to torsional oscillations of the
entire rotor drive train system. These originate from electrical unbalances or transients imposed
on the generator.
Sometimes referred to as torsional resonance, this should not be confused when talking about
resonant vibration of the blade modes. This root cause of HCF damage involves the response of
the entire system. Like critical speeds, the entire system is designed so that the torsional natural
frequencies are detuned from the 50-hertz or 60-hertz multiples of the generator phase to
minimize the reaction or response to vibratory oscillations produced whenever such electrical
transients are inadvertently experienced. If a torsional mode of the system is not detuned, it can
amplify the oscillation. If the peak of the system mode shape coincides with one of the rotating
stages of the LP turbines, the shaft oscillation can result in excessive dynamic stresses within the
blades over the duration of the transient.
Unlike a synchronous vibration problem, failure from torsional system resonance is precluded
by properly detuning the entire rotor system, not by making changes to the blade rows. If the
blade modes are well detuned and there is no evidence of off-design periods of operation,
torsional vibration can be suspected. To establish this directly rather than circumstantially
requires a rotordynamic as well as a blade model, based on sufficient physical information so
that all of the mass, stiffness, and bearing coefficients forming the system are accurately
defined.
Such information has normally been available only to the turbine OEM. Alternative rotor
suppliers will generally overcome this problem by reverse engineering the necessary
information from the original system, then running a system model with their retrofit rotors
included. Evidence of a potential concern raised from such an analysis is apparent if a large
mass is strategically added as part of the rotor replacement. Features associated with the
investigation of torsional system interaction that are suspected to be involved with blade failures
are shown in Table 2-6.
2-19
10305538
Table 2-6
HCF Characteristics of Torsional Interaction in RCA

Basis Graphic
to the Mechanism
Problem is confined to the last or

second-to-last LP blade. Multiple blades
are cracked, typically in the root
torsional system interaction
attachments. Unit does not have many
start-stop cycles.
A new rotor has been installed on an

original turbine-generator system that
has changed the torsional natural
HCF due to torsional system
frequencies of the system. Proper
interaction
adjustments to the system were not
made to detune the system mode.
HCF cracks formed in a system that has

Evidence that HCF is due to
been in service for a significant period
torsional system interaction is
with no reported problems involving the
not the cause
torsional frequencies.
Cracks from HCF from synchronous

sources (resonance), asynchronous
HCF due to torsional system
vibration (flutter), or partial arc operation
interaction
Failure occurs shortly after a new rotor

design is installed in an original turbine-
generator system, causing the system
from torsional system
frequencies to need recalculation and
interaction is the root cause
detuning. Evidence of system detuning is
from large masses added to the shaft.
2-20
10305538
HCF Caused by Partial Steam Admission
Definition
A fourth root cause of blade failure involving the mechanism of HCF is exclusive to the initial
stage of high-pressure turbines in designs where the steam flow is fed to the stage by separate
control valves. PAA is used as a means to minimize throttling losses and to improve heat rate at
low-load operation.
Although the forcing produced by different control valve settings is synchronous with the rotor
speed, it is not applied to the blades in the form of engine orders derived from casing joints,
extractions, and so on. Instead, it is a periodic loading and unloading of the steady steam-bending
forces that occur as the blades pass through opened and closed arcs of the steam admission
nozzle block (see Figure 2-4). The blades experience the loading and unloading sequence as a
series of stepwise transient forces. Because of the frequency with which this occurs (number of
arcs times 50 hertz or 60 hertz), the consequence is a dynamic stress related to the rotating speed
of the rotor that might promote damage to the blades in the form of HCF.
Figure 2-4
Partial Admission Loading of Control Stage Blades
Because of the high pressures and temperatures these blades must typically accommodate, they
are generally substantial enough to withstand a reasonable amount of dynamic stress before a
problem might develop. In circumstances where SPE is attacking the structure and thereby
reducing its overall strength or forming notches that can form cracks, PAA might further
accelerate this failure process. Features associated with the investigation of PAA operation that
are suspected to be involved with blade failures are shown in Table 2-7.
2-21
10305538
Table 2-7
HCF Characteristics of PAA in RCA

Basis Graphic
to the Mechanism
Failure occurs in the control stage of the HP

turbine where partial admission loading is
partial arc operation
applied.
Evidence usually pointing to Similar cracks on multiple blades in

HCF due to partial arc locations where maximum steady stresses
operation occur, principally in the root attachments.
Evidence that HCF is due to Cracks formed in regions where significant

partial arc operation is not the erosion from solid particles might have
cause reduced the strength of the original design.

Cracks due to HCF from synchronous
HCF due to partial arc
sources (resonance).
operation
Failures occur in control stages where

partial admission loading is routinely
from partial arc operation is the
practiced. Crack growth is transgranular and
root cause
unassisted by damage from SPE.
2-22
10305538
Stress Corrosion from Chemical Attack and Steady Stress
EPRI References on SCC
Additional discussion of SCC in steam turbine blades can be found in Section 26 of Turbine
Steam Path Damage: Theory and Practice, Volume 2 [2].
Definition
The term corrosion implies that the initiation phase has been assisted by environmental factors
that attack the inherent resistance of the material. Typically, neither should be suspected unless
(a) the blades operate below the Wilson line, (b) the blades have been in service for an extended
period, (c) there is reason to believe that the stage operating environment was allowed to
become particularly aggressive and/or, (d) there is detectable evidence at the crack origin
derived from visual inspection or metallurgical analysis of the blades. Like LCF, the formation
and growth of stress corrosion cracks are related to the steady-state stresses produced during
operation at full speed. However, the growth rate of SCC is not a function of the number of
start-stop cycles that the blades experience or differences in the magnitude of stress beyond a
certain threshold.
Damage Estimation
The growth rate of SCC is deliberate and predictable based on specimen tests under controlled
conditions. It is the incubation phase—a function of yield strength, temperature, and the relevant
properties of the material—that is normally not distinguished from crack imitation when an
investigation involves SCC. During the incubation phase, a disruption of the steam chemistry
results in deposits left on exposed surfaces of the blade or disk attachments that in the presence
of moisture, attack the grain boundaries of the exposed material surface. Centrifugal forces
exerted during steady-state operation assist to pull them apart, further facilitating the penetration
of contaminants and extending them as branching cracks. For stress corrosion cracks to form,
the stress must exceed only the material threshold. Once formed, tests indicate that the
magnitude of steady stress around them does not play a major role in the rate at which stress
corrosion cracks propagate.
For steam turbine blades operating in the wet LP rows, the stresses produced from centrifugal
force are generally more than enough in the attachments to facilitate the initiation of cracks if
the other criteria are met. For this reason, SCC often starts in locations of lower stress, but
where contaminants can be deposited and concentrated. Although blade roots often experience
higher stresses than their attachments, the problem is more associated with the rotor disks and
older style keyways because of the higher chrome content of the blade alloys and/or the fact that
the blades are replaceable. However, if left in service long enough, blades can develop SCC.
Features of SCC associated with steam turbine blade failures are highlighted in Table 2-8.
2-23
10305538
Table 2-8
SCC Characteristics in RCA

Basis Graphic
to the Mechanism
Evidence of multiple branching cracks

formed on a microscopic scale on multiple
Evidence unique to SCC
blades or attachments operated in regions
of wet steam.

locations where corrosive products can
become deposited or trapped, such as
SCC
holes and fillets or beneath shrouds, along
tenons.
Cracks are formed distinctly in regions of

Evidence that SCC is not the maximum steady stress on blades not
cause exposed or operated in a wet steam
environment.
LCF and creep cracking are also

promoted by steady stress and tend to
Mechanisms confused with affect multiple blades within the same row.
SCC Creep is associated with high-temperature
rows. LCF forms in stages where no
moisture is present.
Branching cracks are found on multiple LP

blades operated in the same row that are
exposed to a wet steam environment.
Cracks take years to develop and are
When to suspect that SCC is usually found by inspection before fracture
the cause occurs.
2-24
10305538
Corrosion Fatigue from Chemical Attack and Dynamic Stress
EPRI References on CAF
Additional discussion of CAF in steam turbine blades can be found in Section 24 of Turbine
Definition
CAF is generally related to dynamic stress, HCF, and, again, the number of hours a blade has
experienced. Microscopic or macroscopic pits, inclusions, crevices, erosion notches, and so on
create highly localized stress raisers or singularity at their bottom. By doing so, they circumvent
the time otherwise required to initiate a crack exclusively from HCF. Once a crack is started, the
remaining life is then dictated by its further propagation to a critical size.
Damage Estimation
With CAF, further propagation is dictated by the stress intensity factors along the front of the
crack. In turbine blades, this is affected by the magnitude of dynamic stress present in the region.
CAF is distinctive in that it makes regions vulnerable to HCF that are otherwise immune to the
nominal dynamic stresses routinely experienced by a well-designed, undamaged blade. In this
context, a well-designed blade is one where either the potentially damaging modes are well
detuned to minimize the dynamic stress from normal vibration, or the design itself is rigid
enough to resist responding to the forces present in the steam flow.
To establish the root cause of a CAF failure, a frequency analysis plays a critical role in
identifying the source of the vibratory stress in the region where cracks formed. Inspection and
metallurgical reports are used to characterize the average pit sizes present within the region.
Coupled with FM, they can be used to determine at what size HCF might take control of pits in
that region.
The probability for CAF to develop is related to the hours the respective material is exposed to
the dynamic stresses after a chemical attack has occurred and/or been allowed to persist
uncorrected. Over time, CAF can become a problem if higher modes of vibration (beyond the
initial three) are not detuned.
Features associated with CAF related to the investigation of steam turbine blade failures are
highlighted in Table 2-9.
2-25
10305538
Table 2-9
CAF Characteristics in RCA

Basis Graphic
to the Mechanism
Evidence of a pit or crevice found at the

Evidence unique to CAF
origin of the crack.
Similar cracks forming on blades that have

otherwise served reliably for many years.
Evidence usually pointing to Cracks form in locations where dynamic
CAF stresses from non-resonant vibration of
detuned modes would otherwise be
tolerated.
Cracks formed in blades where there is no

evidence of corrosion, pitting, and so on,
Evidence that CAF is not the
particularly at the crack origin. Analysis
cause
indicates high potential for synchronous or
asynchronous vibration to occur.
Cracks initiated by fretting and/or HCF can

form on blades where nominal deposits are
detected on the exposed surfaces
approximate to the origin. Cracks would
initiate without assistance of pits.
CAF
Blades have operated reliably in a wet steam

environment for many years before failure
occurs, often on the airfoil in a region of
otherwise low (detuned) dynamic stress.
When to suspect that CAF is
Visible pitting is evident and/or found in the
the cause
metallurgical examination.
2-26
10305538
Erosion from Water Droplets or Solid Particles
EPRI References on WDE and SPE
Additional discussion of WDE and SPE in steam turbine blades can be found in Sections 2 and
17 of Turbine Steam Path Damage: Theory and Practice, Volume 2 [2].
Definition
Both WDE and SPE are distinct from other blade failure mechanisms in that damage is the result
of the physical removal of material during unit operation. WDE normally affects last stage
blades. SPE is a phenomenon that predominantly attacks first HP and first IP (reheat) stages, but
it can migrate to other stages downstream in particularly severe cases. Physically, WDE is
caused by the impact of particles (oxide scale) or large droplets against the blade material. For
SPE, the oxide scale is exfoliated from the high-temperature surfaces within the boiler system.
Damage Estimation
Relative to erosion, the period assigned as incubation in the overall development and growth of a
crack is the period during which material is being removed. Eventually, a notch is formed that
might develop or be considered an initial crack, that is, the next phase referred to as crack
initiation. The time to complete this initial phase is difficult to predict for either type because the
erosion process involves a number of variables. The magnitude of stress at impact is a function
of the particle or droplet density, diameter, impact angle, and velocity. The damage rate itself is
also a changing process because as the surface becomes irregular and fissures or notches
develop, the effect of the impact is reduced and the rate is thereby decreased. The investigation
of erosion as the root cause of a failure normally starts at the second phase, or the assumption
that the notch is sufficiently large enough to form an initial crack. The remaining life is then
determined using FM.
In field repair schemes, analysis provides guidance on how much damaged material might be
ground away or removed, thereby eliminating any initial cracks and returning the process to the
incubation phase. In such strategies, the critical or controlling parameter is that the fundamental
natural frequencies are not significantly affected, creating an HCF fatigue failure problem from
resonant vibration. In the longer run, an erosion problem is addressed as a sequence. The first
corrective action is to take actions that will better remove or minimize the number of particles or
water droplets entrained in the steam. The second is to provide adequate protection. Design
changes meant to increase LP blade life sometimes involve improving the original detuning of
higher order modes, which tend to produce their maximum dynamic stress in the upper portions
of the airfoil that are most exposed to WDE. Features associated with SPE or WDE related to the
investigation of steam turbine blade failures are highlighted in Table 2-10.
2-27
10305538
Table 2-10
Erosion (SPE or WDE) Characteristics in RCA

Basis Graphic
to the Mechanism
Evidence of material loss normally

Evidence unique to SPE or typically found on the leading edge of the
WDE blade airfoil around stellite if used as a
protection.
SPE is typically found on the initial HP

Evidence usually pointing to and IP reheat stages. WDE normally
SPE or WDE affects the last LP row. Material is visibly
missing from the airfoil.
Cracks formed in blades not exposed to

Evidence that erosion is not the particles entrained in the steam or which
cause form in undamaged or intact regions of
the blades.
Stress corrosion and creep cracking are

also promoted by steady stress and tend
SPE or WDE
to affect multiple blades in the row.
Blades are operated in a wet or high-

temperature environment. Visible loss of
When to suspect that SPE or
material is apparent in the forms of
WDE is the root cause
grooves, notches, or trenches from which
cracks can form.
2-28
10305538
Material Creep from High Temperature and Steady Stress
EPRI References on HTC
Additional discussion of HTC in steam turbine blades can be found in Section 16 of Turbine
Definition
The root causes of HTC leading to rupture are stress and temperature accumulated during
continuous periods (hours) of turbine operation. The principal or driving stress is steady state due
to centrifugal force at full speed. The range of temperature for creep to be of any serious threat to
the turbine blades is normally greater than 900°F (482°C). In steam turbines, it is therefore
normally confined to the initial stages of the HP turbine unless a problem with the turbine flow
path design causes steam to stagnate or recirculate in a manner whereby the temperature is
allowed to build up in a localized region.
Damage Estimation
In HTC, the process leading to rupture start with voids forming as the material relaxes. This can be
considered the incubation stage of crack development. If the conditions of stress and temperature
are maintained, these voids grow and eventually link together, forming initial cracks. Crack growth
during this phase is normally deliberate, occurring at a constant rate. If the blades are not removed,
at some point the vibratory stresses will assume control and propagate the cracks to critical size,
resulting in rupture. Traditionally, designers have relied on the Larson-Miller parameter, which is
used in conjunction with accelerated tests of material specimens to estimate how long it will be
before creep rupture occurs. Instead of directly calculating the creep rupture life, a design criterion
is selected, with the design criterion meant to approximate when the tertiary phase (final rupture)
will be approached as a cumulative percentage of creep strain. To directly investigate the process
of creep, a transient thermal analysis is required. The analysis calculates the rate of creep strain as
a function of time. In turn, this can be plotted against the design criterion to determine the
proportion of creep damage that occurred for the selected location.
Creep damage manifests itself as the elongation or distortion of the material. Depending upon
features of the blade design, material relaxation might be confined to regions of peak stress and
strain, eventually forming cracks, and it might be experienced over large regions of the airfoil.
Generally, the creep will lead to deformation that allows the rotating blade to contact the axial or
radial sealing surfaces, and the resultant rubbing will lead to turbine shutdown through vibration.
In most cases, the rotor attachment points (rotor steeples) will fail by creep before the blades.
This is because the blade alloys are more highly alloyed than the rotor steel and have superior
creep resistance as compared to the rotor material. Features associated with HTC related to the
investigation of steam turbine blade failures are highlighted in Table 2-11.
2-29
10305538
Table 2-11
HTC Characteristics in RCA
Relationship of the
Basis Graphic
Formation of microvoids between grain

Evidence unique to HTC
boundaries that eventually link and form
rupture
cracks in advanced state.
Similar cracks found on multiple blades in

same general locations where maximum
steady stresses and temperatures occur,
such as holes and fillets.
HTC rupture
Cracks formed in regions of lower steady

Evidence that HTC is not the stress and/or in regions on blades where
cause operating temperatures are below 750ºF
(399ºC).
TMF is also a high-temperature mechanism

that can form cracks in regions of high stress
HTC
and temperature.
Blades operate in temperatures greater than

750ºF (399ºC). Blades have not seen many
When to suspect that HTC is start-stop cycles, or thermal transient
the root cause stresses are not sufficient to do much. TMF
damage and cracking occur at locations of
high, steady stress.
2-30
10305538
TMF from Transient Thermal Gradients
EPRI References on TMF
Additional discussion of TMF in steam turbine blades can be found in Section 16 of Turbine
Definition
The mechanism of TMF is caused by the differential between the thermal strains produced as
regions of the blade warm or cool at different rates. TMF represents the cumulative portion of
strain that is retained by the material at a given location with the completion of each start-to-stop
cycle of operation. If the thermal transient is accelerated too quickly, there is a potential for the
resulting damage to form cracks in localized regions of the design. As a damage mechanism,
TMF behaves like LCF. It is normally more of a problem for first row turbine blades, such as
those operated in gas turbines, where the stage operating temperatures are >1500–2000°F
(816–1093°C).
Damage Estimation
TMF is promoted by transient thermal stresses, which are reflected as local strains within the
blade as it goes through during a complete start-stop cycle. Similar to LCF, the estimation of
TMF life generally does not separate out the incubation phase, but uses the total strain range for
a completed thermal cycle to equate the number of start-stop cycles before cracks are predicted
to initiate. The strain-range cycles are determined by empirical tests of material specimens over a
range of operating temperatures.
To account for this type of damage in a failure investigation, an analysis is relied on to produce a
strain temperature map for each of the thermal/load conditions that the blade would encounter.
The rate of TMF is strongly influenced by the total strain range and peak temperature
experienced during an operating cycle. If the time to accommodate temperature swings is
shortened, the strain range is often increased, producing more severe TMF damage within
locations where peak strains occur.
As a rule, TMF is not a serious concern for most steam turbine blades, but predominantly a
mechanism that causes cracks to form in gas turbine blades, particularly those operated in the
initial rows where cooling passages create highly concentrated stress/strain gradients. It is the
parameter that controls the dwell time that is required for HP rotors to allow the thermal
differential stress to reach an equilibrium, thereby minimizing the strain gradient they produce.
Features associated with TMF related to the investigation of steam turbine blade failures are
highlighted in Table 2-12.
2-31
10305538
Table 2-12
TMF Characteristics in RCA
Relationship of the
Basis Graphic
Formation of cracks in high-

temperature blades that have usually
Evidence unique to TMF
been subjected to a large number of
start-stop cycles.

Evidence usually pointing to locations where maximum transient
TMF thermal stresses—such as holes and
fillets—occur.
Cracks formed in blades that operate

at relatively low temperatures, are not
Evidence that TMF is not the
cycled frequently, and do not
cause
experience a large thermal strain
during run-up or run-down of the unit.
Creep cracking is also a high-

Mechanisms confused with temperature cracking mechanism that
TMF tends to affect multiple blades at the
same locations within the row.
Blades are operated in a high-

temperature environment. Units are
When to suspect that TMF is cycled frequently. Cracks form in
the cause regions of maximum thermal strain,
such as the platform or the leading
edge.
2-32
10305538
Other Mechanisms: Overload, Deformation, Rubbing, and Impact
EPRI References
Additional discussion of other damage mechanisms of steam turbine blades can be found in
Sections 28 and 33 of Turbine Steam Path Damage: Theory and Practice, Volume 2 [2].
Definition
Overload describes a situation where the operating loads exceed the ultimate tensile strength of
the material. Deformation occurs when the load exceeds the yield strength of the material. Wear
or rubbing is abrasive metal loss or frictional heating that results in metallurgical damage.
Impact occurs when foreign objects (outside the turbine) or domestic objects (self-generated
debris) strike the blade and damage the surface.
Damage Estimation
Unlike the principal mechanisms, their identification and actions to take are generally
straightforward. As such, they do not normally require an intensive investigation or corrective
changes to the blade design.
Overload failure of steam turbine blading as the cause for an event is rare. Blades are typically
designed with significant safety margins, and tensile overload is not a common occurrence.
Overload damage can occur as a result of severe operational transients, such as axial or radial
rubbing when the airfoils strike the casing or turbine overspeed. Deformation of blading occurs,
but it is also seldom a cause for failure. As with overload, deformation is usually the result of a
rubbing or object impact event. Blading alloys have good ductility at operating temperatures and
will deform considerably before failure. However, because of the close clearances between
blades and stationary parts, minimal deformation can be tolerated before a catastrophic overload
occurs.
Another common occurrence is airfoil fracture from the impact of debris, either foreign or
domestic in nature. The fracture surface produced can have arrest marks, especially if the
impacting event is cyclic in nature—multiple impacts from random events. This fracture
appearance can look like fatigue, but the fracture surface is generally rougher and more textured.
Fractographic analysis by a trained investigator can easily differentiate between fatigue and
impact damage. A version of impact damage is also due to water induction. It generally results in
a geometric distortion of the affected blades. Three potential mechanisms are normally involved:
(1) thermal shock or high density of the water (cool vapor) relative to the steam, (2) overload
from the impact of particles or slugs of water, or (3) rubbing wear caused by thermal distortion.
Root causes are the result of inadequate or improper feedwater control, drainage, warmup
procedures, malfunctioning valves, or leaking tubes.
2-33
10305538
Closing Remarks
In a failure investigation, the damage from each of the aforementioned mechanisms can be
approximated with information obtained from three separate sources: (1) the pre-failure
operating records; (2) operating stresses—steady, dynamic, and/or thermal if the high-
temperature stages are involved; and (3) material tests performed under conditions that are
specific to each mechanism. Because of their nature, each of these separate damage mechanisms
can be conveniently related to an identifiable source associated with the turbine.
In general, when the evidence indicates that the problem is initiated by LCF or HCF, this points
to a problem with the blade design. If the material is up to its required standards, and/or the
installation is performed correctly, it is reasonable to suspect that operating stresses are
exceeding what the original designer assumed or calculated. In such scenarios, corrective action
is likely going to require replacement of the row with a modification designed to reduce or
minimize steady, dynamic, and/or thermal stresses by a significant amount.
The role of the original blade design and the operating stresses it produces within susceptible
locations in the assembled components can play a secondary role for failures involving erosion
or corrosion. If the assembled row results in marginal operating stresses, corrective strategy
might not involve modifications to the blade, but instead address weaknesses within the larger
turbine systems that control chemistry or moisture removal. In such circumstances, limited
grinding and blending might allow damaged blades to remain in service, and/or protective
coatings might be considered as a refurbishment strategy. If stresses are not necessarily large
enough to initiate a crack but are not marginal, changes to the original design might be necessary
to improve the limited damage tolerance that these stresses produce.
As a rule, any estimates of life consumption are not to be viewed as absolute numbers, but
weighed in terms of orders of magnitude. In other words, a well-designed blade put through the
technical evaluation described in this report should produce results that suggest it can
accommodate tens or hundreds of thousands of start-stop cycles and millions of hours of
vibratory stress before crack initiation is predicted.
2-34
10305538
3
METHODOLOGY AND OBJECTIVES
Table 3-1
Subject:
Presented in this section is a standard methodology consisting of three phases that are
recommended in this report as a basic roadmap for managing an investigation from start to finish.
Each phase of this methodology has been further broken down into a sequence of mutually related
tasks. These tasks direct the assembly of basic evidence and how it is used to either decide to stop
and execute an available strategy or proceed to the next step. The section concludes with a
discussion of the general timing involved in completing each phase and some comments on the
various technologies that are normally involved. Explicit details on the activities performed within
each of the three phases of the RCA methodology are presented in subsequent sections of the
report.
Topics Covered:
3.1 – Flowchart of Phases and Tasks in the Methodology
3.2 – Overview of Three Phases of Investigation
3.3 – Objectives of Tasks in Each Phase
3.4 – Timing Issues
3.5 – Comments on the Technologies and Their Potential Limitations
Tables/Figures:
Table 3-2 summarizes the objectives of each phase of the turbine blade RCA and the estimated
timing involved in completing each. It represents the touchstone or framework around which the
entire report is structured.
Figure 3-1 illustrates the three phases of a standard turbine blade RCA and the separate tasks
involved within each phase. Decision points regarding the decision to execute a strategy based on
the available information or to proceed with the next phase are identified.
Flowchart of Phases and Tasks in the Methodology
Many flowcharts used to successfully investigate a turbine blade failure are available in
published cases. Whether arrayed in a fishbone format or as descending flowcharts, they all
include, basically, the general tasks that revolve around site forensics, metallurgical analysis, and
blade design analysis. The methodology recommended in this report and shown in Figure 3-1
(repeated from Figure 1-2) is a derivative of a 14-step procedure from Turbine Steam Path
Damage: Theory and Practice, Volume 1 [2].
3-1
10305538
Methodology and Objectives
Figure 3-1
Methodology to Guide a Steam Turbine Blade Failure Investigation
3-2
10305538
The objectives for each phase are summarized in Table 3-2. The three phases are further divided
into a series of mutually related tasks that are each intended to produce evidence (data), which
are used to make a determination as to what should be done regarding the original blade that
failed. At the end of each phase is a logical break point where the operator can review the status
and decide if sufficient information is available to confidently execute a repair or replacement
strategy, presuming one is available.
Table 3-2
Objectives for Each Phase of Investigation and the General Timing Involved
Phase Objectives Timing Issues

Estimated duration: 4 to 12 weeks
Prevent a repeat occurrence of the Activities within the RCA take time as data are
Overall assembled and parts are transported for
failure on the original or sister units
RCA measurements, testing, or analysis. If activities in
that operate the blade that failed.
tasks 1 and 2 are initiated in parallel, the timing can
be reduced by two to four weeks.
Prior to the failure and when the Estimated duration: 3 to 4 weeks
turbine is inspected, assemble
The metallurgical analysis is normally the critical path
Phase operating data on the quality of the
item in this step of the RCA. Retrieving, transporting,
1 blade material and evidence of
and testing samples can take two to four weeks.
mechanisms left on any recovered
samples.
Identify whether the root cause of the The structural analysis (finite element analysis) is
blade failure is a design or damage normally the critical path item. If drawings are not
Phase tolerance issue. Achieved by available, a sample blade must be reverse
2 calculating the operating stresses engineered. Modeling takes one to two weeks.
and life consumption they produce in Calculations take two to four weeks, depending on
locations where cracks have formed. the complexity of design and whether thermal
stresses are involved.
Use a computer model to represent
and evaluate the potential benefits of Most modifications are not major redesigns unless a
Phase completely new replacement (upgrade) is available.
recommended changes to the original
3 Adjustments to an original model normally take about
design proposed by the OEM and/or
based on results of the RCA. one week. Calculations are rerun, and results are
tabulated for comparison to original design results.
Overview of Three Phases of Investigation
Phase 1: Assemble Failure Evidence: Tasks 1–5

Phase 1 is the starting point of every failure investigation. The work is principally to establish
basic facts surrounding the failure event. The outcome is a decision as to whether sufficient
evidence exists to adopt a readily available repair or replacement strategy. The site forensics
(documenting the failures and visible condition of the blades as well as assembling the pre-
failure service history) is accomplished by relying on qualified personnel from the plant. Trained
specialists are needed to perform the metallurgical examination and any nondestructive testing.
3-3
10305538
Phase 1 stops short of pursuing a detailed examination of the blade design in terms of stresses,
frequencies, and so on. At its conclusion, a replacement in kind would be the most likely
corrective action if it were decided not to proceed with the investigation. Such a determination
would normally be based on one of the following four obvious reasons:
• Review of the operating history reveals an extreme off-design condition of prolonged
operation, that is, an obvious anomaly.
• Inspection reveals that failure was initiated by a high-energy impact of some sort (that is,
FOD, water hammer, and so on).
• Evidence shows that the blades were damaged as a consequence of failure in some other
system of the main turbine.
• Metallurgical analysis found that the material was deficient in composition or in mechanical
properties.
In essence, each of these reasons would support a conclusion that the failure was a genuinely
random event. As such, the root cause was not directly related to the design of the blades, nor
would it be corrected by making modifications or adjustments to the design. Instead, some
external, unexpected force introduced during service exceeded the strength and safety factors of
the original blade design, or the material itself proved to be substandard. It is emphasized that the
evidence supporting this conclusion should be unimpeachable. A random event would
correspond to circumstances whereby the event occurred relatively soon after the blades had
been thoroughly inspected and/or of a design that had truly operated problem-free for a
significant period of service (several years) before the event occurred, but not long enough for
significant aging or wear to have played a role.
Because preparing the FEM in Phase 2 is a significant undertaking in terms of spent time, it
might be prudent to initiate this step in parallel with those identified in Phase 1 rather than
adhere strictly to the sequence in which they are presented. This is recommended if the
immediate evidence indicates the possibility that aspects of the blade design are involved.
Phase 2: Perform Engineering Analysis: Tasks 1–9
Presuming that no obvious reason for failure is evident, the investigation proceeds to the next
phase of nine tasks. A major activity within this second phase involves the construction and use
of an FEM of the bladed-disk-rotor segment to simulate the response of the design to the
mechanical and thermal forces it might have experienced during operation. Typically, such a
model is used to define three key items for the assembled row: stresses, natural frequencies, and
mode shapes. Once available, stresses (steady, dynamic, and thermal) are fed into the appropriate
life assessment models along with the respective material fatigue properties to estimate the
number of cycles or hours associated with crack initiation, crack propagation, and probability of
failure.
3-4
10305538
Phase 3: Develop a Corrective Strategy: Tasks 1–5
Phase 3 repeats portions of the engineering analysis by introducing any proposed modification to
the design or materials to the FEM and recalculating both stress and life. The need for this phase
presumes that the root cause is not predominantly controlled by the way the blades were operated
or that control systems for the turbine are not modified to eliminate the source of excessive
damage or unusually high stresses. Modifications can range from limited repairs of damaged
regions to a complete replacement with a blade where features have been physically changed. In
this final phase, the investment in the FEM produced from the root cause investigation is
leveraged by not only demonstrating how the original design was compromised, but by
quantitatively gauging how much the design would benefit in terms of predicted life from
proposed changes in material, geometry, or avoidance of certain operating scenarios.
Objectives of Tasks in Each Phase
The following presents a broad overview of each task performed within the three phases that
might ultimately make up a complete RCA. Aspects and features associated with each of these
tasks are discussed at length in subsequent sections of this report.
Compile Inspection, Metallurgical, and Unit Operating Data
Phase 1 is preceded and includes those tasks identified as immediate action items, discussed in
Section 1, “Immediate Recovery Scenarios.” In addition to securing parts, taking photographs, and
taking measurements of the blade outer tip diameter and rotor/disk profile, any available inspection
records are sought to give some sense of how quickly a problem might have developed. This is
particularly true if the root cause is related to the operating environment versus the operating
stresses (for example, erosion- or corrosion-related mechanisms versus HCF or LCF). If available,
they should provide some basis to characterize the perceived condition of the blades before the
failure occurred. For example, was there any visible evidence of pitting, fretting, rubbing, and so
on, or did the blades appear to be in reasonably good condition?
Compile Unit Operating Data
Operating data relevant to a failure investigation always start with a reasonable estimate of the
total start-stop cycles and the service hours the blades might have seen. Anomalies involve
atypical events, such as when the unit approached or exceeded the allowable envelope of load
and backpressure, experienced excessive grid disturbances, experienced a steam chemistry
upset, or was exposed to an unusual transient event (thermal, electrical, or mechanical).
Compile Metallurgical Data
The purpose of the metallurgical examination is to establish that the material was not
substandard and to seek evidence from the fracture surface that can be compared and correlated
to confirm the initial conclusions drawn from operating history and inspection records. That is,
3-5
10305538
what was the condition of the material surface on a microscopic level from the point of initiation
to final rupture? As noted, there should be some correlation between time-dependent evidence,
such as beach marks, pitting, or corrosion, and the record of operating anomalies and results of
inspection records from the prior period of operating history.
Evaluate Industrywide Data (Historical Problems)
If a previous failure investigation is similar enough to the case at hand, an opinion can be
formed with sufficient confidence to accept the same remedy without proceeding beyond this
task. Alternatively, if an operator finds a significant record of pre-history events associated with
the blade, it is indicative of a situation where emphasis was placed on a trial-and-error approach
instead of a concerted effort to investigate and resolve the problem. As evident in the
aforementioned industry survey, some of the problems associated with specific blade designs
clearly had a generic impact among the fleets in which they had been installed. Given enough
time, most of these problems were eventually resolved. However, multiple design iterations
were often involved in these efforts. This can make it difficult for the operator to draw a direct
relationship between this broad experience and the specific design that failed.
Make a Preliminary Identification of Failure Mechanisms
The objective of this task is to immediately begin paring down the list of potential causes using
all available evidence gathered in Phase 1. Referring to the previous section of this report (see
Table 2-1), the mechanisms and root causes of blade failures can be consolidated into a fairly
manageable list. In turn, this list can be pared down even further depending on where and in
what turbine the blade failure occurred. It is emphasized that any conclusions drawn at this step
are preliminary and should remain subject to change as more information becomes available. In
other words, this is the point in an investigation where speculation is still appropriate, and those
forming the investigative team should not be intimidated by observations that might prove to be
opinion rather than fact.
Define Additional Material Properties
The material properties referenced are those used to estimate life consumption for different
mechanisms by the initiation and propagation models, not the basic properties of elasticity and
density applied in the FEM. Fatigue properties for the most conventional blade alloys are
generally available in published sources. For LP blades, they generally fall within three basic
classifications: 12% chrome (403 stainless steel [SS]), 17-4 precipitation-hardened (PH), and
titanium. For HP blades, 410SS and 422SS are most prevalent. Although specific properties for
an OEM’s particular alloy are always sought and used if available, the distinctions between
coefficients and properties are generally not critical for the purpose of making reasonable
estimates of life consumption.
3-6
10305538
Define Stage Geometry
The objective of this task is to provide sufficient details on the geometry and dimensions of the
blade design and its linkage to other blades in the row (tiewires, shrouds, and so on) in order to
construct a finite element of the fully assembled row. The geometry of the blade is usually
obtained through the cooperation of the OEM and/or by means of direct measurements or counts
obtained while the turbine is open and by reverse engineering available sample parts. In reverse
engineering, a sample is usually sent out to obtain critical details of the root geometry that might
not otherwise be accessible when the blades are mounted in the turbine. Although the disk
profile can reasonably be estimated and scaled from a cross-section diagram of the rotor and the
exact height of the blade is known, the amount of overlap between the blade root and disk root
attachment is not so easily extracted from the cross-section diagram. For long blades, where
0.25–0.5 in. (6.4–12.7 mm) in length could have a significant effect on the calculated natural
frequencies, the best means to establish the correct length of the assembled stage is to register
the outer tip diameter. For last stage blades, this dimension can sometimes be found on lifting
diagrams. If the rotor is available—as it usually is when a failure occurs—a circumferential tape
measure will serve this purpose. If all else fails, the OEM can usually be persuaded to release
this single bit of information.
Generate FEM
FEM is the principal tool or workhorse in the second phase. Its primary purpose is to calculate
the magnitude of operating stresses at rotating speed, which are necessary as input to the life
consumption models. Therefore, the model does not provide a definitive answer as to whether a
blade design will fail, but supplies critical information that is normally available only to the
OEM. Cyclic symmetry to form a 360º blade-disk model is necessary in order to properly
compute the nodal diameter modes and the natural frequencies of the assembled structure.
Interfacing surfaces, such as those at the root attachments, between interlocking covers, and so
on, are also properly treated by using gap elements, particularly in locations approximate to
where cracks have initiated. If not included, the stresses produced at these interfacing surfaces
will not be correctly calculated. Finally, it should be kept in mind that the results of the FEM
analysis are not definitive in themselves, but are used as input to the appropriate life assessment
models. Given the inherent variability and uncertainty that the fatigue properties exhibit even
when tested under controlled laboratory conditions, nominal differences between predicted
stresses are not going to significantly affect the outcome or the interpretation of it.
Calculate Steady Stresses
Steady-state stresses from centrifugal and steam-bending loads are calculated for two basic
purposes. The first purpose is to identify locations susceptible to the damage mechanisms that
they drive, for example, LCF, SCC, TMF, and so on. The values are ultimately used as input to
the respective damage models. The second purpose is to establish the magnitude of mean stress
at locations subject to dynamic stress and HCF. Once the FEM has been completed, a limited
period will be required to put it through two initial checks by the analyst. The first confirms that
the model will run successfully, that is, a numerical solution is accomplished. The second check
3-7
10305538
is to see if the elements making up the model are responding in a reasonable manner, that is, the
results make sense. Problems of this type, which are subtle enough to allow a solution, are
normally caught when the distribution of steady stresses throughout the blade or group can be
examined. An anomaly in the distribution of steady stress or an usually high stress in a region
where there is no local geometric feature to amplify it often indicates that some corrective action
is required to refine the FEM before proceeding with further analyses.
Calculate Frequencies and Mode Shapes
In assessing the root cause of HCF in longer LP blades, it is of paramount importance to know if
the operating frequencies of the assembled row are well detuned, marginally detuned, or
potentially resonant. For shorter blades like those found on HP and IP turbines, knowing the
natural frequencies is still important because they are usually designed to avoid coinciding with
the NPF produced by the upstream stationary airfoils and to withstand the dynamic stresses that
are produced even in an assumed resonant condition. Results of the frequency analysis are
summarized in two useful graphic displays: the Campbell diagram and the interference diagram.
The Campbell diagram shows the families of modes from zero rpm to full-speed operation. The
interference diagram shows the individual nodal diameters at full-speed operation. Both
represent the same data from different viewpoints. A second item produced in Step 10 is the
mode shapes of the blade disks. These are the relative displacements that occur at each selected
frequency of vibration. Ultimately, they are used to calculate dynamic stresses and distinguish
between modes that could promote HCF damage.
Calculate Dynamic Stresses
The objective of this task is to extend the insight produced from the frequency analysis by
estimating the magnitude of stress that might be experienced by the blade at locations where
cracks initiated and to identify the modes most likely to be responsible. If the frequency analysis
indicates a reasonable possibility for a mode to operate in a resonant condition, this task
determines whether the stresses are high enough to represent a threat. If the problem is related to
damage tolerance, the results from the detuned stress analysis provide a basis to estimate the
critical crack size of a pit, erosion notch, FOD, and so on that might be tolerated in a region
where such damage preceded the formation of cracks. In either scenario, the stress results
provide input to the life models and help to identify the mode most likely responsible for
initiating or driving the crack. Like the steady-state results, the dynamic stresses should correlate
with the metallurgical evidence in terms of maximum stresses generally corresponding to the
locations where cracks formed.
The calculation of dynamic stress is typically the most controversial aspect of the entire analysis
because it is the one calculation that must proceed on several assumed inputs, as follows:
• The operating frequency of the mode of interest
• An estimate of the amount of damping present within the design
• The magnitude of stimulus present in the nearest per-rev force contained within the flowing
steam (unless the flow is modeled directly with CFD)
3-8
10305538
As subjective as each of these critical parameters might be, it is absolutely essential to make a
reasonable assessment of the dynamic stresses that a blade design might experience for any root
cause investigation to be valid. Fortunately, each of these critical parameters falls within
reasonable ranges, with the at-speed natural frequency being dominant in terms of the
magnitude of dynamic stress that is produced—hence, the importance placed on the availability
of any frequency or test data that might be used to authenticate the FEM calculations.
Validate Model with Modal or Spin Pit Data
Although predicted magnitudes of steady-state stress can be reasonably checked with a simple
load-over-area formula, how well the FEM represents the dynamic aspects of the structure is
best evaluated by comparison and correlation with frequency test results. Results from two types
of tests are normally sought. Modal testing normally refers to a test performed on the blades at
zero rpm at room temperature. The tests can be on individual blades (suspended or in a test
fixture) or on blades installed in the disk/rotor. Spin pit data look to capture the rotational effects
on the natural frequencies in a controlled, non-steam environment, whereby frequencies are
monitored by strain gauges throughout the range of zero rpm to full speed. The most elaborate
and expensive version of this test is in situ, wherein selected blades on an actual turbine are
instrumented to record the frequency responses, but in the actual steam environment. Most
recently, non-contacting systems for blade vibration measurement (BVM) are being applied to
minimize the expense of an in situ telemetry arrangement and to include the effect of the
flowing steam.
Estimate Time and Probability to Fatigue and Fracture
This represents the final task of the engineering analysis. The assembled evidence is
consolidated into a comprehensive story in which all of the various aspects produce a consistent
explanation for the failure. Two models are involved: a life estimation model and a probabilistic
model. Life estimation models are essentially any deterministic formulation whereby key items
(such as stress and appropriate material properties obtained from tests under controlled
operating conditions) are applied to evaluate the rate of life consumption. The resulting
projections are then compared to the number of hours and starts to see if the mechanism could
account for the formation of cracks within the known period of service. This initial estimate is
generally good enough to confirm the root cause if one is dealing with pure fatigue problems. If
the problem involves the assistance of erosion or corrosion, the estimate of crack propagation
life or damage tolerance might also be initially estimated with a straightforward deterministic
formula that relies on edge correction factors, stresses obtained from the analysis, and the
appropriate material properties to identify at what threshold size the respective mode of
vibration could take control of the propagation process, either in a detuned or resonant
condition. If the issue of damage tolerance is particularly critical, a more precise assessment is
completed by inserting crack tip elements at the location of interest in the FEM and
systematically computing the stress intensity factor along the crack front (rather than
approximating it with an edge correction factor). Again, the critical crack size should correspond
to measured defects, pits, and so on identified in the metallurgical analysis near or at the crack
origin.
3-9
10305538
Iterate Evaluation of Proposed Solutions
Once the root cause investigation is completed, the operator has at his or her disposal the tools
necessary to rerun the engineering model as a means to calibrate the effectiveness of any
proposed change to the design, operation, or materials presently involved in the failure. This is
typically broken out as a third phase of an investigation, because it is usually guided by
recommendations from the RCA and involves discussions of the results with the technical
support personnel, operating personnel, and, at some point, engineers from the OEM. Reflecting
changes on the FEM, recalculating the impact on stresses and frequencies, and feeding this
information back into the life and probabilistic models allows an improvement to be
quantitatively assessed and evaluated relative to the original design evaluation results. This uses
a recognized strength of finite element modeling—namely, it is very reliable at reflecting the
commensurate effect of relative differences, where absolute results might not be possible due to
a large number of unknown variables or factors that cannot be avoided.
Timing Issues
As mentioned, it is not necessarily recommended that the two subsequent phases—engineering

analysis and corrective action—be deferred until the first decision point is reached. However,
many of the activities associated with the engineering analysis require some serious preparations
and are dependent on the completion of preceding tasks. For example, consider the following:
• No finite element modeling or analysis can be performed until the geometry of the part is
physically defined. Reverse engineering often means that parts must either be shipped to a
shop or arrangements made to bring specialists and their equipment on site. This should be
accomplished within the first week of any investigation.
• Depending on the complexity of the design and the region of interest (that is, the area where
cracks formed), an experienced analyst should be able to have a reliable model working in
one to two weeks. However, modeling cannot start until basic information on the blade, disk,
and rotor geometry has been secured. The tasks of analysis cannot be completed until the
stage operating conditions have been provided and the history of prior service has been
assembled. Again, both of these should be accomplished in parallel with reverse engineering
so that all are completed within the first week.
• The individual calculations also involve more than just number crunching. An experienced
analyst will routinely plan time to revisit the model and fine-tune the original grid or certain
critical assumptions that go into the calculations of stresses and frequencies. Running the
stress and frequencies calculations and reviewing them for authenticity can normally be
accomplished in one to two weeks, depending how much adjustment and refinement are
necessary to produce reasonable results.
3-10
10305538
• To obtain test data for comparison with predicted frequencies inherently requires time.
Logistics normally intrude when making arrangements to have specialists on site to perform
modal frequency tests. Delays are often involved when data are supplied with the cooperation
of the OEM. Again, if frequency data are sought to verify the FEM, efforts to secure these
data should be initiated immediately.
• If a database of material properties of commonly used alloys is not readily available, or if the
alloy is relatively new, time will be required to assemble and process the data necessary to
convert the stresses derived from the analysis into fatigue, fracture, and probability estimates.
If deficient material fatigue properties are suspected, additional weeks will be required to
prepare specimens from sample blades and have them independently tested.
General experience has shown that a normal root cause investigation involving all three phases
and their respective tasks is unlikely to be completed in less than four weeks, even if the first two
phases are initiated in parallel. Six to eight weeks is usually more realistic when reviews and
discussions of the results between the separate parties are involved. Invariably, questions are
raised and logistical issues become involved that are guided in part by the findings of the
investigation. Unless an FEM is already available from a prior study, efforts to further expedite
the analysis to a schedule of less than four weeks invites trouble. For each task of calculations
itemized within Phase 2 of the methodology, a reasonable amount of time is necessary to check
the results produced with the model and to see whether they make sense. Decisions involving
mesh density, constraints, and so on remain subjective and at the discretion of the analyst.
Incorrectly applied constraints and/or boundary conditions can cause the frequencies to be off or
create artificial stress concentrations. In turn, these errors will be carried through the life
consumption estimates. For the operator and/or manager of an investigation, it is important to
recognize that the FEM is ultimately only a tool, and it does not make decisions on the results it
produces. Time for review is essential to ensure that the structural model is correct.
Comments on the Technologies and Their Potential Limitations

To understand why it is important to recognize that the method of investigation is an integrated
program that inherently requires a number of qualified specialists, some of the limitations
associated with each are mentioned.
Metallurgy
Depending on the condition of the fracture surface itself, it is not always possible to conclusively
identify the presence of all mechanisms that might have been involved, particularly if the event
was catastrophic. As stressed throughout this report, protection of the fracture surface is critical.
The absence of evidence such as fatigue does not necessarily preclude the possibility that it
played a role in the formation or propagation of a crack. It is also not always possible to
definitively identify the origin of a crack, particularly if the exposed surface of the region is
being regularly removed or degraded in some manner, that is, the origin has been obliterated
before failure signaled the presence of a crack. Another limitation of any metallurgical
examination is that it cannot definitively identify the specific source of HCF when evidence left
on the fracture surface is found. Given the high frequencies at which blades vibrate, it is not
reasonably possible to establish without the assistance of analysis what individual mode of
vibration promoted the initiation and/or the propagation of a crack.
3-11
10305538
Finite Element Analysis
With the advent of more powerful computers, model size (number of elements and mesh density)
is not as much of a problem as it was in the 1980s. However, it is critical that sufficient detail is
included at locations where cracks have formed to ensure that the predicted stresses are
reasonably correct. A major limitation of any analysis is that it is ultimately dependent on the
input, assumptions, and boundary conditions that are assigned by the individual to represent the
rotating row under simulated operating conditions. In particular, the dynamic model used to
calculate natural frequencies and define the magnitude of dynamic stresses that control the rate
of HCF requires a significant amount of practice and experience. The challenge of modeling a
blade-disk system is further compounded when interlocking or interfacing features, such as
snubbers, covers, and tiewires, are involved. If these are incorrectly represented, they will
introduce artificial stiffnesses or fixed points of contact that result in errors in stresses and/or
predicted frequencies.
The subjectivity of any model to decisions by the analyst can also be abused to produce results
that defy the fact that a failure occurred, that is, stresses are low and frequencies are well
detuned, so the problem must lie elsewhere. This can be accomplished by assuming an
unreasonably low stimulus and/or excessive damping to minimize the magnitude of predicted
dynamic stress in the region of interest. Subtle changes in boundary conditions or reasonable
adjustments of Young’s modulus can shift critical modes by a nominal amount to reduce the
dynamic stress that might otherwise be produced in a resonant condition. As emphasized in the
procedure, comparing and verifying the predicted frequencies are often the best ways to check
the validity of a model. There is one word of caution in this regard. Tests to measure blade
modes of vibration are controlled by which sensors are used and how they are oriented or
applied. In other words, arrangements to conclusively identify the frequency of the first or
second mode of vibration might not be optimized to detect higher modes. Generally, if the first
two or three modes correspond well, evidence of higher modes can usually be found.
FM
As with finite element analysis, FM models are ultimately subject to the same assumptions and
judgments made by the analyst who prepares the model. Again, the capability to introduce crack
tip elements into the FEM is a major improvement over the use of edge correction factors in simple
formulas because it can accommodate the actual stress field produced at the location of interest,
which in turbine blades is often highly three-dimensional in nature. The major uncertainty
associated with any FM study remains the actual stress intensity threshold assigned to represent the
fracture properties of a given blade alloy. Because of the high frequencies of vibration involved,
even at the smallest measured growth rate found on the da/dN curve for a given material, there are
more than a sufficient number of cycles within a year to grow an initial crack to its critical size. As
such, the determining factor relevant to blade failure investigation is exactly where the stress
intensity starts. As evident on most da/dN curves, the difficulty and cost of testing within this
region normally limit the number of data points by which the curve can be fitted or extended to a
3-12
10305538
logical start point. Another feature evident in any da/dN plot is the inherent variability of the data
points even under the controlled conditions of a laboratory environment. The potential uncertainty
of all of these factors should be recognized whenever damage tolerance limits are investigated. In
other words, like finite element stresses or frequencies, there are no single absolute values, but
numbers that should fall within a reasonable range.
Probabilistic Analysis
Probabilistic analysis in the root cause investigation of blade failures is an attempt to eliminate
some of the subjectivity in the selection of critical parameters, such as stimulus factor, damping,
and material properties, that are used in the otherwise discrete life models. If properly applied,
the results of the probabilistic analysis should reinforce the results from the discrete life
predictions and provide additional insight as to how quickly the odds of a problem occurring
worsen over a measure of unit operation, that is, hours or starts. The probabilistic treatment more
correctly recognizes that despite the best efforts of the designer and the operator, random chance
can always intercede and cause a failure. With this said, a probabilistic model is like any other
simulation in that it can be abused if certain assumptions are applied. Although the approach
deals with probability density functions (PDFs) rather than discrete values, the odds for a failure
to occur can be reduced or favored based on the mean value and coefficient of variation assumed
to define the PDF. This is particularly true if a marginal detuning condition is involved. Also,
like FM, the input associated with the relevant material properties that govern the life model of
interest (LCF, HCF, SCC, and so on) is dependent on the amount and consistency of statistical
data that have been assembled from available sources. As with any statistically based parameter,
more data are always welcome, and too often, sufficient data are not available. This is less of a
problem for most conventional steam turbine blade alloys because much research was performed
and published in the 1980s and 1990s. It is a major impediment to the probabilistic analysis of
gas turbine blades, which rely on directionally solidified and single crystal materials.
3-13
10305538
10305538
4
IMMEDIATE RESPONSE AND ON-SITE FORENSICS
Table 4-1
Subject:
Once a blade failure is suspected or confirmed, Phase 1 of the investigation begins with a series of
tasks performed primarily by the operator. These include the three immediate actions previously
discussed in Section 1, “First Response: Recovery Versus Investigation.” In addition, certain
background data highlighted in this section of the report should be assembled from records
maintained at the plant and on the turbine deck once the unit is open for inspection. Attention should
be given to coordinate the collection of information that documents the features of the damage with
the logistical efforts of getting the turbine restored to an operating condition so that a permanent
record of the as-found condition of the turbine is made before any circumstantial information is
removed or destroyed.
Topics Covered:
4.1 – Pre-Inspection Action Items
4.2 – Immediate Action Items When Turbine Is Open
4.3 – Visible Inspection of the Parts
4.4 – Evidence Related to Installation
4.5 – Basic Design Information Required
4.6 – Items Specific to the Turbine(s)
4.7 – Items Specific to the Blade and Row Involved
4.8 – Relevant Operating Data
Tables/Figures:
Figures 4-1 through 4-3 illustrate the items sought from a visual inspection of the turbine, items
related to the installation of the blades, and basic design information when the turbine is open.
Figure 4-4 illustrates examples of blade damage resulting from impact rather than fatigue, corrosion,
or erosion.
Figure 4-5 illustrates the potential importance and consequences that excessive grid fluctuations
can have as a source of HCF.
4-1
10305538
Immediate Response and On-Site Forensics
Pre-Inspection Action Items

Pre-inspection activities consist primarily of reasonable efforts to assemble and maintain
documentation that is complete as possible of (a) evidence prior to when the failure was
identified and (b) features that are apparent when the row is inspected. The following pre-
inspection questions are designed to maximize use of the time before visible evidence can be
physically obtained from the turbine:
• Was there any evidence or early warning of the problem?
• Is there any reason to suspect that blade failure was a consequence of another problem?
• Would the operating conditions at the time be considered typical or unusual?
• Is a current heat balance of the unit available?
• What spares are available for any of the rows?
Once a failure is suspected, attempts should be made to see whether there was a possibility to
have foreseen the problem as it developed, either based on the readings produced by the standard
instrumentation found on most turbine-generators or by a qualitative sense of the unit’s response
to the immediate demands of the grid. Any reason to suspect that the damage associated with the
turbine blades is a consequence of some other system should also be initially considered until
evidence to the contrary (or lack thereof) is produced and confirmed. A check should be initiated
to determine whether any spare blades are readily available.
To minimize the time spent transitioning from the initial phase of investigation to a more
detailed analysis, a file should be started that contains the basic information that any third-party
investigator might eventually require. Such a file should start with a current or original heat
balance and cross-sectional diagrams of the turbines. The heat balance diagram is used in the
structural analysis to identify the relevant parameters at the inlet and outlet to the turbine where
failure is suspected as a basis for characterizing the steam flow and temperatures for the specific
stage(s) of interest. The cross-section can be used to quickly identify relevant features of the
turbine, such as extractions and the number of blades per disk, and to provide a reasonable
definition of the rotor profile and disk section corresponding to the row in question.
The file should also include the results of any prior inspections, whether they were visual or by
NDE. If these reports reveal no prior indicators of a problem, this actually might be as relevant as
evidence of indications, pitting, erosion, or unusual wear patterns. If the RCA is pursued, at some
point the investigators will need to define a point from the date the rows were first installed.
As both an investigative and logistics action item, immediate efforts should be undertaken to
identify any spare blades that are available as possible replacements, presuming the damage to
the turbine is not severe or catastrophic. For an investigation, the reason to locate spare blades is
to preserve the contingency for pursuing an RCA by having the means to reverse engineer basic
features of the blade geometry. When there is a high probability that a blade failure has occurred
and a particular row is suspected, reverse engineering should be undertaken right away if spare
blades are available. This can save several days of critical path time waiting to extract parts from
the damaged rotor and/or shipping them for measurement after it is decided to pursue the second
phase of analysis.
4-2
10305538
If spare parts are unavailable, a short list of plants operating similar units should be prepared and
inquiries made to see if there is an alternative to either the OEM or third-party shops for
supplying replacements. Options to proceed independently of the OEM timeline will be severely
compromised if no blade samples are secured before or after the inspection.
If the failure was sufficiently catastrophic to wreck a good portion of the machine, external
forensics will need to include sufficient documentation in the form of photographs and records that
can assist in the determination of whether the wreck was the consequence of a blade failure or that
the damage to the turbines was consequential. Generally, there are few types of failures that can
literally tear apart a unit to the degree that couplings and bearings are torn from their mountings
and shafts are found bent. Usually, an unbalance of such an order is caused by the disk frag-
menting or a substantial portion of a large blade being suddenly released. However, in the wake of
such a catastrophic event, the scrutiny will be intense and necessarily involve several parties. The
best means for the plant to protect its interests without obstructing the process of restoring the unit
to its original condition is to take a liberal number of photographs. These will prevent time being
wasted on speculation of what landed where or what was damaged and what was not.
Immediate Action Items When Turbine Is Open
As noted in Section 1, three immediate action items should be undertaken once the turbine is
open to start the assembly of evidence, as follows:
1. Multiple photographs should be taken before anything is moved or removed. This includes
both wide-angle and close-up photos of any cracks or fractures as well as photos that show
the general condition of the blades. If notable erosion or corrosion is evident, these details
should be captured in a way that their actual severity can be reasonably estimated. Photos are
important because they can reveal clues to an analyst or metallurgist that reinforce their
findings. They can also limit speculation on how pitted, corroded, or eroded the blades
appeared during subsequent discussions over the role that steam chemistry or quality might
have played.
2. Any samples of failed or cracked blades that are found should be protected from
contamination before they are examined or removed from the rotor. Particular attention
should be taken to cover and protect any exposed portion of the fracture surfaces. Oils and
salts can compromise the chemical analysis performed by the metallurgist. Often, fracture
samples are limited, so particular care should be taken in their removal and transport. Marks
left on the pieces resulting from their forced extrication can leave confusing evidence. A
dropped sample can obliterate critical information on the surface. Such accidents inherently
favor landing on the region where the origin is located.
3. No blade samples or parts should leave without the plant’s approval. This cannot be stressed
firmly enough. Control of the site and anything that leaves it should remain the standing
policy from the start. If there is only one sample of a failed blade, it is recommended that the
operator keep it and either have an independent metallurgist perform the analysis or make
arrangements with the OEM to witness their test and to have the specimen returned after the
examination is complete. If multiple examples of cracked blades are found, these should be
documented and divided between the OEM and the plant so that they preserve the
contingency for having an independent investigation performed.
4-3
10305538
Visible Inspection of the Parts
Once the casing is removed and the turbine is available for direct inspection, a careful record of the
condition of the row or row(s) should be made, again by documenting them with plenty of
photographs. Questions to be answered by this documentation are highlighted in Figure 4-1. A
general overview or impression of the as-found condition for the rows is sought that might
reinforce the evidence found on a detailed examination of the fracture surface. To focus all
subsequent actions, an initial determination must be made as to which row most likely failed first.
If the consequential damage makes this difficult, it is reasonable to proceed on the assumption that
the first to fail was the row farthest upstream. Unless definitive evidence can be produced from
prior inspection reports, it must be initially assumed that the time frame bounding the failure
process extends to when the parts were originally installed. A more definitive role that erosion,
corrosion, or other defects might have played will be sought from the metallurgical analysis.
Figure 4-1
Information Sought from Initial Visual Inspection of Turbine Components
In the visible inspection, a qualitative opinion is sought as to the relative appearance of the
blades. Notching or wire cutting can create localized cracks, whereas uniform material loss
might be more tolerable. If visible distress is apparent, photos should be taken that typify the
general state of the blades as they were first found.
Evidence Related to Installation
Specific questions to be resolved from this documentation are highlighted in Figure 4-2. They
involve certain operations performed when installing blades that are not freestanding, which can
affect the magnitude of stresses at these critical interfaces. Inconsistency in the quality of welds
or brazes can affect the scatter in blade frequencies. For designs that depend on forming a
continuous ring at full speed, gaps that do not close or remain open can result in frequency/HCF
issues. For grouped blades, cracks forming in tenons and shrouds might be influenced by the
quality of the peening that was performed. General documentation is sought that can be used to
4-4
10305538
assess the apparent quality and control performed during installation of the rows, with particular
attention to critical locations, for example, welds, brazes, and interfacing surfaces. In particular,
visible evidence is sought that might distinguish the row where failure occurred from other rows
where it did not.
Figure 4-2
Inspection of Items Related to Installation
Generally, welds or brazes that are performed around the disk to form blades into groups should
be uniform and controlled. Excessive or non-uniform amounts of material can affect the
frequencies of the blade or be an indicator of the possibility for localized work hardening of the
material in the vicinity of where the welding or brazing was performed.
Management of gaps and pre-loads is critical to any design where it is intended that the row form
and maintain an interlocking system at full speed. Out-of-tolerance fits can result in an unequal
sharing of loads among multiple pairs of hooks or fingers in the root attachments. Initial gaps
that are too wide might fail to close as the tips of the blades untwist due to centrifugal forces.
Gaps that close but achieve only a marginal force because the initial size was too large can result
in damage from friction (overheating) or fretting on the contact surfaces. A check of the gap
sizes, their apparent uniformity around the circumference of the wheel, and the condition of the
contact surfaces can prove to be extremely relevant in terms of how to model the structure and
the stresses that might be produced as a consequence. For shorter blades that likewise rely on
maintaining a continuous link during operation but that do not have the length to untwist like
longer last stage blades, a reverse condition is established: the blades are installed so that when
assembled, a pre-load is created between specified contact surfaces in the cover. If the pre-load is
insufficient, the centrifugal forces will relieve or open the gap. For such rows, their initial state
should be checked for evidence of their condition after installation.
Problems originating in peened covers can involve a failure where the peening was not
performed correctly. This can sometimes be the consequence of features of the original designs
not functioning properly—for example, shims or side grips, which are meant to keep the blades
from moving inward as the peening is performed. If the blades can move, the resulting line of
4-5
10305538
contact between the tenon and cover might be marginal but not apparent to the installer. In
instances where cracks are found in the cover or tenon, an assessment of the quality and
uniformity of the peening should be attempted to assist in supporting or ruling out this
possibility. Depending on how severely damaged the row is, intact samples of the peened cover
might be removed for further examination, that is, sectioning to confirm whether a proper line of
contact with the tenon was achieved.
Basic Design Information Required
Questions to be answered by this portion of the documentation are highlighted in Figure 4-3. The
length of last stage LP blades can classify most large steam turbines.
Figure 4-3
Basic Design Information Sought from an Open Turbine
Some basic information on the type of material, number of blades in the row, and features of the
linking arrangements after the rows are assembled will be needed to do the following:
• Determine the relevancy of any prior events to the specific design that failed
• Identify whether spares of the same type can be procured
• If an analyst is approached, allow them to assess whether any prior investigative work might
be used to expedite or assist in modeling the row
Items Specific to the Turbine(s)
Gathering of visible evidence naturally proceeds from an overall perspective of the turbine
condition to focus on more specific details of the failure. Before proceeding to the specific row,
general evidence should be sought that indicates whether a feature of the unit configuration or
flow arrangement played any role, or whether the problem is more generic to the blade design
4-6
10305538
itself. Problems related to steady-state stress tend to be more pervasive and widespread and favor
the leading or trailing blades within a group. Dynamic (frequency-related) problems are more
selective and tend to be more random, reflecting the nominal differences in frequencies. The
number of cracks found is often evidence of how long the problem has been developing.
Questions to be answered by this initial documentation should include the following:

• Does the evidence indicate damage caused by a non-fatigue-related mechanism?
• Was fatigue (cracking) of several blades found, or was fracture confined to one blade?
• Was the failure confined to one row and/or one turbine?
• Around the disk, were cracked or fractured blades found?
Non-fatigue-related high-energy mechanisms, such as FOD (see Figure 4-4) and water hammer,
might be considered random events, not requiring a detailed analysis.
Figure 4-4
Examples of Damage Resulting from High-Energy Impact Mechanisms
Depending on the location and extent of FOD, this might still promote the consideration of a
detailed analysis, but it would change the objective from establishing the root cause to an
assessment of the remaining life if the blade is not immediately removed, or to establish repair
(blending/grinding) limits. Blades damaged by water hammer (slugs of water that get into the
flow) tend to be physically or plastically deformed rather than fractured. Any apparent pattern
where only one row of several found within the same unit was damaged might suggest features
of the flow arrangement; water extraction or the condensing system might play a role. The
apparent fact that fatigue cracks were not caught before fracture occurred will further indicate the
overall urgency (time available) to find a corrective solution for similar rows.
4-7
10305538
Items Specific to the Blade and Row Involved
The location around the row and on the blade provides an additional clue as to what mechanisms
might play a role. The number of blades found damaged is an indicator of the duration or
severity of the mechanism(s) at work. Patterns, particularly in grouped blades, can assist in
revealing the mode responsible for HCF damage. These can indicate a natural progression of the
original problem or a second problem resulting from the first. Some of the questions to be
answered by this documentation include the following:
• Where did the fracture or fatigue generally occur on the blade(s)?
• How many blades were found cracked among each row?
• How many blades were found cracked in the same location?
• Were any blades found cracked in different locations?
LCF problems will naturally conform to locations of maximum steady stress, and they are
particularly fond of the fillets within root attachments, welded lashing lugs, brazed tiewires, and
peened covers. Failures or cracking found above the platform often points to a higher mode of
vibration, particularly if there is widespread presence of pits or erosion damage within these
regions. Sometimes, secondary problems appear when the original grouping is compromised. If a
blade is left freestanding as a consequence of tiewire or shroud failures, the change in stiffness
can result in a new resonant condition that was a direct consequence of the initial failure.
Evidence of such a breakdown in the form of cracked tiewires, blades detached from their
covers, or excessive gaps should be pursued in any analysis by comparing the frequencies of the
blades if left in a freestanding state. This can assist in establishing the proper sequence of events
that led to a final fracture.
Relevant Operating Data
To complete the assembly of relevant documentation available at the plant, some basic details of
the duty experienced by the turbine blades prior to their failure are assembled. Questions to be
answered by this documentation include the following:
• How many start-stop cycles did the turbine row experience?
• How many hours did the turbine row experience before failure was identified?
• Was there any significant event that might have disrupted the steam chemistry prior to
failure?
• When was the row last inspected, and was there any evidence of erosion or corrosion?
• How often did the unit approach or exceed the OEM limits of load versus backpressure?
• How often did the unit experience any out of specification (out-of-spec) rotation due to
severe grid disturbances?
• Prior to failure, did the unit experience any particularly severe electric transients?
4-8
10305538
Information on the number of start-stop cycles is a fundamental measure to determine whether

LCF could be responsible for the initiation of cracks. The total number of hours that the blades
operated is similarly important in determining whether HCF, SCC, or CAF could be responsible.
If SCC or CAF is involved, the time when contamination occurred is used to assist in estimating
the rate of subsequent fatigue. Chemistry unbalances or contaminating events, such as condenser
tube leaks or feedwater problems, can often be traced to a specific breakdown in other parts of
the system. Mechanisms such as SCC grow at a deliberate rate. Crack lengths determined in the
subsequent metallurgical examination can sometimes be correlated to such events.
Erosion/corrosion problems are driven by the particulates entrained in the steam. Because the
rate is not readily predictable, the last time the blades were inspected can be used to provide a
gross estimate of how quickly the problem developed. Estimating the rate of a sudden or
premature severe erosion or corrosion problem might become critical if the root cause
investigation changes to one of determining how much remaining time the damaged parts might
remain in service and/or whether their reliable service might be restored to some degree with
some remedial action. In such instances, the rate of damage can only be estimated based on the
interval between inspections where evidence of the problem would have been noticeable.
The number of times backpressure limits were approached or exceeded is a crucial fact in
determining if the failure was potentially caused by asynchronous vibration (flutter). The
possibility for any such events to have occurred needs to be identified in terms of the number of
events and the estimated duration for each. A heat balance is needed to further interpret this
information. Evidence indicating this possibility will eventually save time if analysis indicates
that the blade modes are well detuned and there is no evidence that erosion or corrosion
contributed to the formation of cracks. If there is no possibility that the operating backpressure
and load conditions were approached or exceeded, this is an indication that the calculation of
resonant frequencies is incorrect, or that other sources must be responsible for the HCF.
Grid-related problems are a concern for units whose rotating speeds fluctuate by more than ±1.0–
1.5 hertz. In regional arrangements where large unbalances are unavoidable, the trip limits might
need to be extended to keep the unit continuously on-line. When such speed changes occur, the
per-rev multiples of running speed will change proportionally, but the natural frequencies of the
blades will be affected as a squared value of the shaft speed—that is, they will be slightly
increased or lowered as a reflection of the change in stress stiffening, but not proportionally with
the nearest per-revs. If fundamental modes are marginally detuned, this additional margin can
cause a per-rev to shift enough to create a temporary resonant or near-resonant condition, thereby
increasing the dynamic stresses and HCF damage.
Torsional system interaction as a root cause of a blade is a problem more likely to occur on units
where original rotors are mixed with new designs, thereby changing the original torsional natural
frequencies of the turbine-generator system. Evidence of this possibility is sought in the form of
corrective masses added to the rotor drive train as part of a relatively new rotor upgrade where a
substantial amount of successful operating time has not yet verified the modified system
torsional frequency calculations. It is important to further recognize that any such calculations be
of the modified system in use, and not just generically related.
4-9
10305538
Final Comments
As highlighted in this section, the operator is generally best suited to undertake these first tasks
because much of the information is contained or controlled within the confines of the plant.
Ultimately, the objective for this documentation is to establish a reasonable record of what
occurred, what was found, and the specific blade design that is involved. The absence of
evidence, such as no indications of exceeding backpressure limits or serious upsets to the quality
of the steam chemistry, can be as important as evidence to the contrary. To metallurgists and
analysts who might become involved later in the investigation, the facts assembled by the plant
should correspond to the results of their subsequent studies. For example, if SCC is clearly
identified in a metallurgical examination, it is counterproductive to have evidence of a tube leak
emerge after the fact.
It should be emphasized that the initial documentation assembled in accordance with the
questions raised throughout this section is not expected to be absolute or definitive, except in the
cases where the failure is clearly a consequence of a random cause or event. For example, the
damage left by FOD is normally indisputable. If a crack or fracture is initiated from such damage
and no evidence is found on other undamaged blades or rows, it is reasonable to conclude that
this was an anomalous event that does not need to be investigated further.
More likely, the documentation assembled from the site is going to be somewhat inconclusive,
often incomplete, and sometimes contradictory at first glance. The incompleteness of records,
particularly historical records on prior failures, is often a frustrating impediment, but it is not
inconsistent with the fact that once a failure is resolved, everyone wants to forget it and move
forward. The recollections of senior technicians or engineers and their personal journals are often
more informative than official records. Inspection reports are also sometimes vague or
incomplete because at the time the inspection was performed, the potential problem was not
evident. Depending on the design, the location where cracks eventually formed might not be
readily accessible. Removing blades to thoroughly inspect them might be cost-prohibitive and
violates the basic maintenance rule of “don’t fix what ain’t broke.” This rule is sometimes
grounded in truth, particularly for axial entry type blades that are installed through a notch. The
frequencies of these rows can be significantly affected by the tightness of the final fit achieved
around the circumference. A row of tightly fitted blades that has operated for years without a
problem can suddenly fail if it was removed and then reinstalled incorrectly. Phased array
systems are being developed to overcome this limitation.
Contradictory evidence is often a consequence of the sequence from incubation to rupture that
was previously discussed, where damage is produced by different mechanisms at different stages
of crack development. Additional information obtained from the metallurgical and structural
analysis should be able to help sort out such apparent conflicts in subsequent steps of the
investigation.
4-10
10305538
Finally, it is important that any suspected non-normal events be identified so that they are not
discovered later, forcing the investigation to retrace its steps and revisit some of the original
assumptions on which it has been proceeding. As noted, if an anomaly is overlooked, it is likely
that eventually, the anomaly will be found when the metallurgical analysis is completed, or
worse, when the design analysis is completed.
A classic example of this involves the waste of weeks of engineering time spent checking the
authenticity of the FEM used in an investigation that proceeded on the basis that no damage was
apparent around the pinholes of a blade root attachment where cracks had formed and a
catastrophic failure resulted. Despite attempts to improve the meshing, adjust the boundary
conditions, and recalculate the natural frequencies, the stress analysis correctly indicated that
neither LCF nor HCF could explain the failure given the known timeframe of service.
Confronted with this finding, a photograph was eventually produced that showed significant
dents left around the edges of the pinholes as a consequence of hammering in the pins. Given
their locations and sizes, the resulting dents were more than sufficient to act as localized stress
concentrations and to accelerate the fatigue failure timeline.
4-11
10305538
10305538
5
OFF-SITE FORENSICS: METALLURGICAL
EXAMINATION
Table 5-1
Subject:
This section features the basic metallurgical examinations that should be performed on samples
taken from the row. Using examples of fatigued or fractured specimens, it offers a description of how
different failure mechanisms typically appear on both a microscopic and macroscopic scale. It
concludes with a discussion of how this information is used to separate possible root causes from
secondary or passive contributors.
Topics Covered:
5.1 – Basic Goal of the Laboratory Analysis Process
5.2 – Visual Examinations, Dimensional Examinations, and NDE
5.3 – Determination of the Chemical Composition of the Component
5.4 – Metallographic Examination
5.5 – Mechanical Properties Testing
5.6 – Basic Features of Fractures
5.7 – Incubation Through Rupture
5.8 – Basic Features of Non-Fractures
5.9 – Interpretation of Results: Relation to RCA
5.10 – Examples from Blade Failures
Tables/Figures:
Table 5-2 lists the typical chemical composition of turbine blade alloys.
Figures 5-1 through 5-3 illustrate the conditions and features of different types of failures.
Figure 5-4 illustrates the three steps of a metallurgical examination.
Figures 5-5 through 5-10 illustrate features left on the fracture surface of different materials caused
by different blade mechanisms.
Figures 5-11 through 5-23 illustrate examples of features left on fracture surfaces of blades that
failed in different locations of the blade, for example, at the root, airfoil, lashing lug, tenon, and
covers.
5-1
10305538
Off-Site Forensics: Metallurgical Examination
Basic Goal of the Laboratory Analysis Process
Metallurgists are fond of saying “the metal doesn’t lie.” This is reasonably true as long as the
interpreted metallurgical results are not underreported, overstated, or extrapolated into the arena
where analysis is better equipped to handle the interpretation of certain data. The metal does not
lie, but it does not tell the whole story. Because blades are a distinctive type of component,
subject to a mixture of steady and dynamic stresses and operated in a potentially aggressive
environment, it is strongly recommended that those who are tasked to perform the metallurgical
examination have some prior experience and a basic understanding of how different blade
designs can react to their particular stage environment.
The first objective in an analysis of the blade materials is to eliminate substandard material as a
source or root cause of the blade failures. Substandard material can be the result of improper
manufacturing or operational exposure. Specific answers sought from the material analysis
include the following:
• Did the chemistry conform to OEM and/or general industry standards?
• Did the material’s tensile strength conform to OEM and/or general industry standards?
• Did the material’s yield strength conform to OEM and/or general industry standards?
• Did the material’s ductility conform to OEM and/or general industry standards?
• Did the material’s hardness conform to OEM and/or general industry standards?
• Did the impact strength conform to OEM and/or general industry standards?
• Was the microstructure observed consistent with general industry standards for the material?
• Were the dimensions consistent with the drawings or exemplar blade?
If the answer to all of these questions is yes, the blade manufacturing process likely did not
contribute to the failure. If the answer to any of the questions is no, it is possible that
manufacturing issues contributed to the failure, and future analysis is necessary to determine if the
observed material anomalies are due to manufacturing or operational factors. One difficulty that is
generally encountered in answering these questions is that the OEM is not usually willing to share
its specification requirements. There are industry guidelines available through ASTM International
(ASTM) and others that have been published by industry organizations such as EPRI.
It should not be immediately assumed that deficiencies found in the chemical makeup and/or
with some of the basic material properties automatically explain why a failure occurred. Out-of-
spec properties might be only a further contributor to a stress-related problem by weakening the
resistance to damage or fatigue. If the corresponding design stresses are too low to form a crack,
the out-of-spec property or properties might not prove to be the root cause. Familiarity with the
basic mechanisms that affect blades and their potential sources is most likely to result in the
operator receiving true insight as to what might have transpired, rather than a report of
photographs that are documentation but not interpretation. As with the on-site information that is
assembled after a failure, the evidence left on the blade material might not be absolute or
decisive. The condition of the specimen might have been severely compromised, particularly if
the failure was catastrophic.
5-2
10305538
Figure 5-1 shows a typical laboratory specimen. The specimen might have been contaminated as
it was passed through other parts of the turbine or when it was retrieved. Figure 5-2 shows an
impact mark witnessed on a fracture surface caused by post-failure handling.
Figure 5-1
Typical Laboratory Sample
Figure 5-2
Post-Failure Damage on Failure Sample
A metallurgist familiar with examining failed blades will be aware of these factors, and he or she
should therefore be able to recognize them and know how to properly cope with them in terms of
the ambiguity they might cause when preparing final conclusions on the findings of the
examination.
5-3
10305538
An essential part of the failure analysis process is the metallurgical laboratory investigation. This
investigation consists of a number of discrete tests or steps. The accuracy and thoroughness of
this testing will have dramatic impact on the overall process, not least of which is the
interpretation of the data collected. There is no set protocol for this testing. Just as with any well-
stocked toolbox, there are many tools that can be used by the engineer; any effort requires certain
tools. Some tools are not necessary for a particular job, but other tools can literally be worn out
on one project. The skill of the person wielding the tool is also critical. A poorly performed test
is more damaging than no test at all. This section will discuss those tools and how they are used.
The first issue in laboratory investigation is what hardware is available to analyze. In an ideal
world, the entire turbine would be available for examination in the as-failed condition. In most
cases, however, the investigation is limited to a failed component selected by one person, and
that component might not have the essential information needed to solve the puzzle. Often, the
real key to the investigation resides not in the failed component but in exemplar components that
have seen similar service. For example, a fractured turbine blade will likely be damaged by post-
failure events, whereas a sister blade that has not separated but has a crack caused by similar in-
service conditions will have a trove of useful information.
Another example is in blades with attached shroud bands. The condition of the shroud band (is it
tight and/or crack-free?) can shed considerable light on the likely pre-failure condition of the
failed blade. Figure 5-3 shows a shroud band that was improperly peened. The loose shroud led
to a fatigue failure.
Figure 5-3
Post-Failure Damage from Inadequately Peened Tenons
5-4
10305538
When a rotating component fails, it does considerable damage to the stage downstream. For this
reason, detailed analysis of downstream hardware must be carefully considered versus the
expected information gained. It is also not unusual for the blade to damage the stationary row
immediately upstream. Again, the value in analyzing upstream components must be carefully
weighed.
Visual Examinations, Dimensional Examinations, and NDE
Visual Examination
Keeping in mind that subsequent analysis and testing of the blade after the visual examination
will be destructive in nature, the initial visual examination and documentation will be the only
time in which the failure analysis engineer will be able to observe the blade in its as-failed
condition. Sufficient documentation must be conducted during the visual examination so that the
blade can be reconstructed later. Once the blade(s) reach the laboratory, all parts available for
examination should be carefully examined and photographically documented in their as-received
condition. Any identifying stampings or markings should be recorded and compared to those
collected in the field, if available. The type, severity, and location of damage to the blade should
be noted and recorded through photographs, sketches, and measurements and cataloged. The
condition and general features of the fracture should be documented as part of the initial visual
examination. If the blade contains deposits, care should be taken during the visual examination to
minimize disturbing deposits and the parts should be protected against cross-contamination.
Wear or contact marks should also be noted. There is often fretting or abrasion at the fixation
point that will indicate any blade root vibratory displacement during operation.
Collection and Preservation of Deposits/Residues
Deposits on blade and fracture surfaces can be very useful in characterizing the environmental
conditions present during operation. However, careful collection procedures are required to
prevent contamination that can lead to erroneous results. Deposits can be collected by scraping
the surface or by applying acetate replica tape to the surface and pulling the deposits from the
surface. Scraping is typically used to collect deposits from blade surfaces normally exposed
during routine operation. Adhesive tape can be used if acetate replica tape is not available. This
tape is commonly used to collect deposits from fracture surfaces.
Collecting deposits by scraping is best accomplished by placing the sample on a clean piece of
paper and gently scraping the surface with a wooden Popsicle stick or glass rod. The collected
deposits can then be transferred from the clean paper to a vial with a lid for storage. Fresh paper
and Popsicle® sticks should be used for each deposit collected to prevent cross-contamination of
the deposit samples. If a glass rod is used to collect samples, it should be thoroughly cleaned and
rinsed with alcohol between deposit collections. If the deposits are particularly tenacious, a metal
rod or carbide tool of known composition can be used to collect the deposits. Again, the rod
should be thoroughly cleaned between collections.
5-5
10305538
If the deposits are collected by applying tape to the surface, care must be taken in handling the
tape so as not to contaminate the tape. Gloves are recommended when handling the collection
tape. The tape should be firmly applied to the deposit-bearing surface and then gently removed.
It is recommended that once it is removed, the tape should be applied to a clean glass slide and
the slide placed in a vial with a lid for storage.
Often in major turbine failure, miscellaneous debris will be collected from the condenser or
crossover piping. This debris usually contains blade fragments that might not be useful in the
analysis. Often, especially in L-0 or L-1 failure, the initial liberated airfoil will survive in pristine
condition, whereas the remaining root is subjected to impact from domestic objects. Given that,
the miscellaneous debris should be evaluated. Visual and magnetic separation is useful for
locating and recovering blade components.
NDE
NDE methods are typically employed in the field to identify the existence—and, in some cases,
the dimensions—of flaws or cracks. During failure analysis, these techniques are particularly
helpful in identifying the location, size, and orientation of the crack or flaw. With this
information, a sectioning plan can be developed to accomplish the removal or opening of the
crack or flaw in such a way as to minimize damage to the crack or flaw surface to be analyzed
once it is opened.
There are many types of NDE techniques—ultrasonic testing (UT), liquid and dye penetrant, wet
and dry magnetic particle, and so on. Most if not all of these methods require exposing the part to
be inspected to contaminants. UT requires the application of couplant, penetrant testing requires
the use of dyes and developers, and magnetic particle testing requires the application of powders
and fluids. If these methods are to be used to inspect the blades, any deposit samples should be
collected prior to the nondestructive testing to avoid contamination by NDE test methods.
Dimensional Examination
General part dimensions should be collected on the failed blade or blade pieces and an exemplar
blade, if available. If reverse engineering is required, the pieces can be measured with any of the
modern coordinate measuring devices and these data archived. This will also permit FEMs to be
constructed. If available, the blade drawing should be reviewed. Comparisons of like dimensions
will aid in determining whether the failed blade was different from unfailed blades or design
dimensions.
One possible consideration at this point in the investigation is whether exemplar blade samples
should be modal-tested. This ring test can be used to truth-check the FEA model, but it must be
done with great skill, insight, and knowledge of the limitation of the testing (see Section 8,
“Model Verification by Frequency and Vibration Testing”). In any case, the exemplar blade
might not have the same modal signature as the failed blade. If the exemplar is cracked, the
modal testing is not practical.
5-6
10305538
Determination of the Chemical Composition of the Component
Chemical testing of the exemplar blade will provide baseline compositional data. The chemical
composition along with the mechanical properties will assist in defining the blade material
specification if it is unknown. For all of the diversity in turbine blading design, there is very little
variation in turbine blade alloys. Generally, there are three classes of blading alloys, as follows:
• Martensitic stainless steels
• Precipitation-hardened: martensitic stainless steels
• Titanium
All of these classes are quite similar in behavior and damage mechanism susceptibility. Even the
harmonic response characteristics are similar, that is, the ratio of the dynamic modulus to the
density is nearly the same. Table 5-2 summarizes the basic composition of the martensitic
stainless steel blade alloys.
Table 5-2
Typical Compositions of Turbine Blade Alloys
Blade Alloys (Nominal Composition or Range)

X20Cr13
Element AISI 403 (DIN Jethete M
AISI 410 AISI 422 17-4 Ph H46 Ti 6-4
(Note 1) 1.4021 152
[Note 2])
Carbon 0.15 0.15 0.22 0.16-0.25 0.07 Max 0.08-0.15 0.18 0.02-0.04
Manganese 1.00 1.00 0.69 1.5 Max 1.00 0.50-0.90 0.65
Silicon 0.50 1.00 0.50 1.0 Max 1.00 0.35 0.40
Nickel 0.60 NS 0.76 3.0-5.0 2.0-3.0 0.45
Chromium 11.5-13 11.5-13 11-12.5 12.0-14.0 15.0-17.5 11.0-12.5 10.0-14.0
Molybdenum 0.93 0.50 1.5-2.0 0.65
Iron Balance Balance Balance Balance Balance Balance Balance 0.10-0.8
Tungsten 0.97
Copper 3.0-5.0 0.50 Max
Vanadium 0.21 0.25-0.40 0.30 4.1-4.5
Columbium + 0.15-0.45 0.40
tantalum
Aluminum 6.2-6.6
Oxygen 0.12-0.20
0.002-
Hydrogen
0.006
Nitrogen 0.01-0.05 0.01-0.019
Titanium Balance
Notes:
1
AISI is the American Iron and Steel Institute.
2
DIN is the Deutsches Institut für Normung (the German Institute for Standardization).
5-7
10305538
Metallographic Examination
The metallographic examination of the blade hardware should occur in three steps, as shown in
Figure 5-4. First, the exemplar should be sectioned to characterize its condition, hardness, and
composition. A section should be taken near the same position of the suspected origin of the
failed blade. This examination will determine if defects, in-service damage, or metallurgical
anomalies are present. Step 1 will educate the investigator on what might be found on the failed
blade.
5-8
10305538
Figure 5-4
Three Steps of Metallurgical Examination
5-9
10305538
Step 2 will be the metallographic examination of the failed blade, remote from the fracture. This
section should be removed about 1 in. (25 mm) from the fracture, after which the examiner
verifies that there is no difference from the exemplar blade. The final step is metallographic
sectioning through the origin. Step 3 will be one of the last metallurgical tests completed.
Because this test is destructive to the fracture surface, it should occur only after the fractography
has been completed and technically reviewed and all parties concur on the interpretation. This is
often the major error made by the inexperienced investigator.
Metals used in turbine alloys respond quite predictably to damage. The fractures produced are
permanent records of the damage and loads encountered. The trained investigator can interpret
the features and establish with high reliability the loading that led to the failures. The
macroscopic and microscopic observation of the fracture surfaces produced in the failed
hardware can be compared to exemplar fracture surfaces documented in the literature and
produced by exemplar testing. Correct interpretation of these data can identify the mechanism of
damage.
Microstructure
The general microstructure should be examined and photographically documented, typically at

100X and 1000X. The microstructures at the core of the blade cross-section and at the blade
surface should be examined to establish the baseline microstructure of a good blade. The overall
composition and apparent consistency of the material grain boundaries should also be examined.
The precipitation of certain elements or the presence of notable inclusions from the original
fabrication versus the environment can indicate that the material was exposed to an unanticipated
operating environment, which weakened it. Figure 5-5 shows an unusual grain boundary
precipitate in type 403SS turbine blade alloy. These typically appear as carbides.
Figure 5-5
Photomicrograph of AISI 403 Showing Grain Boundary Carbide Precipitates
5-10
10305538
For high-temperature materials, the formation of voids that might or might not be linked is a sign
of creep. The magnitude of temperature that a material was exposed to can often be reasonably
estimated from tests. Such tests can be used to calibrate thermal models and creep models
pursued in high-temperature failures, or it might indicate that an unexpectedly high-temperature
condition was encountered and experienced by the blades.
Regarding origin-specific microstructure, once the origin is located on the failed blade, the
corresponding area on the exemplar blade should be examined to determine if manufacturing and
forming processes produce microstructural anomalies at these locations that could assist in crack
initiation.
Hardness
Globally, general hardness should be measured and recorded for comparison to hardness
measurements collected from similar areas on the failed blade. Again, the exemplar blade
hardness will provide a baseline for comparison to the failed blade. The global hardness should
include a survey of the blade looking at hardness in the airfoil and in the root.
Just as the microstructure of the exemplar blade will be examined in the area corresponding to the
failure initiation area in the failed blade, hardness profiles or traverses should be conducted in the
area corresponding to the failure initiation area. Detailed hardness readings corresponding to the
failure initiation area will reveal if manufacturing and forming processes produce hard or soft areas
that might be more prone to cracking. This is especially useful on LP turbine blades that have
leading-edge heat treatment for erosion resistance or at the location of brazed tiewires or brazed
stellite shields. If not properly done, these welds and brazes can alter the microstructure and lower
the fatigue resistance. Figure 5-6 shows a blade that cracked adjacent to a brazed tip cap.
Figure 5-6
SCC Cracking in an L-0 Blade Adjacent to a Brazed Shroud
5-11
10305538
Mechanical Properties Testing
Mechanical properties testing will define the material strength for various loading conditions.
Testing of the exemplar blade will provide baseline strength data and assist in defining the likely
blade material specification if it is unknown. Standard tests can include tensile strength, yield
strength, elongation, reduction in area, impact strength, fracture appearance transition
temperature (FATT), fatigue properties, and elevated temperature properties.
Tensile Strength, Yield Strength, Elongation, and Reduction in Area
Tensile testing is a common test and reveals much about the strength of the blade alloy. Tensile
strength reflects the material's ability to carry load without rupturing in uniaxial slow straining.
Tensile strength is temperature-dependent, with the tensile strength decreasing with increasing
temperature. Normally represented as a 0.2% offset, the yield strength is the point at which stress
versus strain is no longer proportional. Corresponding to the stress analysis, this is the limit applied
to elastic stress values that signifies that plasticity in the material has likely resulted. The
permanent deformation (plastic strain) is equated to the amount of fatigue damage caused each
cycle (see Section 8, “Life Prediction and Probability of Failure”). For HCF, the loading can be
less than yield strength, so no plasticity is necessarily involved in the initiation phase of crack
development. Yield strength is also temperature-dependent, making the stage in which failure
occurred a significant piece of information that should correspond to the findings of the
metallurgical examination. The tensile test also reveals important information about the material’s
ductility (its ability to absorb energy without fracture). A tensile test generates a percentage
elongation and a percentage reduction in area. The percent elongation is the amount that a
specimen stretches before failure occurs. The percentage reduction in area is the amount of
necking or thinning that occurs before fracture.
All of these values can be determined by a tensile test, commonly performed to ASTM E8
specifications. One to three test specimens are machined from the available blade material and
tested at room temperature to determine the mechanical properties of the blade material. All of
these mechanical properties, in addition to the chemical composition, are used to define the
material specification. That is to say, if the blade material specification is unknown, it can
usually be determined by comparing the results of the compositional analysis and the mechanical
properties tests to standard blade material specifications.
Impact Strength
Impact strength measures the strength of the material under impact, or rapid load application,
conditions. For material specification purposes, these tests are commonly conducted at room
temperature. Most blade material specifications include room temperature impact strength.
Impact testing of blades is valuable in the later stages of an LP turbine, where temperatures can
be quite low. Impact strength is temperature-dependent, with lower strengths experienced as the
temperature decreases. A common impact test is the Charpy test performed to ASTM E23. This
5-12
10305538
testing can be accomplished at ambient temperatures or at lower temperatures if the blade is

operated at lower temperatures. Impact strength can be important on blades that are weld-
repaired or have hardened leading edges.
FATT
FATT is the fracture appearance transition temperature that defines the temperature at which the
material transitions from ductile failure to brittle failure. This is determined by conducting
impact tests at varying temperatures. The impact test specimen fracture surfaces are examined,
and the percentage of ductile versus brittle fracture is measured. Typically, FATT50-50 is reported.
This is the point at which 50% of the fracture is ductile and 50% of the fracture is brittle. This
number is compared to the operating temperature range of the blade to determine if the blade
operates at a temperature at which the material will behave in a brittle or ductile manner. It
should be noted that this is usually not a problem with blade materials and more commonly
affects older rotor materials.
Fatigue Properties
Fatigue is a common failure mode for blades, so fatigue data for most blade materials are
available in various handbooks but not typically as part of the material specifications. Fatigue
properties are typically defined in two ways—fatigue endurance limit and fracture toughness.
The fatigue endurance limit is a cyclic stress level below which the material will last indefinitely.
From a practical standpoint, it is a stress level below which standardized fatigue specimens have
not failed after a specified number of cycles. Rotating beam tests, where an hourglass-shaped
specimen is rotated while a load is applied to one end of the sample, are typically used to
determine the endurance limit. The number of cycles to failure is recorded. Specimens are tested
at various loads to produce a fatigue curve. Iron-based materials will produce a curve that
flattens typically between 106 and 108 cycles. This defines the endurance limit.
Fracture toughness measures the material’s resistance to cracking under fatigue loading. Fracture
toughness test specimens are relatively large, requiring a cube of material that is typically not
available within a turbine blade. Fracture toughness, along with corresponding crack growth
data, is used to estimate the time (number of cycles) between crack initiation (from an assumed
flaw size) to failure under cyclic loading.
Elevated Temperature Properties
Tensile testing at elevated temperatures is useful in evaluating HP and IP blading materials.

Generally, there is no specification for the tensile properties at elevated temperatures in the blade
materials specifications, but the information can help establish the root cause for the ruptures.
Tensile strength defines the strength of the material at failure. Yield strength defines the point
at which the material begins to plastically deform. Elongation and reduction in area define
the distortion of the material prior to failure. The test specimens needed to determine
5-13
10305538
high-temperature properties are the same as those for room-temperature testing; the tests are just
conducted in a furnace at elevated temperatures instead of room temperature. Typically, the
tensile strength and yield strength decrease as the temperature increases, whereas the elongation
and reduction in area typically increase with temperature.
Creep refers to failure as a result of applied stress at an elevated temperature for an extended
period. Testing determines a material’s time to failure at a given stress and temperature
combination. A test specimen similar to the specimen used for tensile tests is loaded with a
known dead weight and enclosed in a furnace at a known temperature. The specimen is left with
the dead weight at temperature, and the time to failure is recorded. Typically, multiple load and
temperature combinations are tested to develop a creep curve. If any two parameters of the three
parameters are known, the third can be estimated using the creep curve. For example, if the
operating temperature and time at temperature to failure are known, the applied stress can be
estimated by referring to the creep curve. Typically, creep curves will not be developed during a
failure investigation—instead, published data will be used. Additionally, creep testing of failed
HP and IP blading is also of limited value because these values cannot be related to new
properties. The uncertainties in applying accelerated tests to judge the remaining life of suspect
blades is of marginal value and can lead to reliance on erroneous data.
Basic Features of Fractures
The most important step in the examination of fatigue fracture surfaces is the visual and
binocular examination of the fracture surfaces. Many features of these surfaces can reveal
extremely useful information about the mechanism of fracture. For example, high-cycle fatigue
(those cracks dominated by low mean stresses and frequent but low amplitude cyclic stress
[vibratory]) tend to have a single planar crack front and single initiation point. These fractures
have a very fine texture and are smooth. On the other hand, LCF cracks (those cracks dominated
by large magnitude forces acting periodically) tend to have multiple origins and multi-plane
fracture paths. The fracture surfaces are rough and have a distinctive texture. Figure 5-7
illustrates this well.
Figure 5-7
Typical Features Found on Fatigue-Fracture Surface
5-14
10305538
Corrosion fatigue cracking, especially in a strongly corrosive environment, tends to be

intergranular and have very indistinct arrest marks. Figure 5-8 shows a corrosion fatigue crack in
an LP turbine blade that propagated completely in an intergranular model. Unfortunately, in the
real world, actual fractures are mixtures of all of these features. It is not unusual for a crack to
begin by LCF and progress by HCF. Investigators have to understand what they are observing
and relate it accurately to the stress/dynamic analyst. Note that corrosion effects could have an
intergranular appearance, such as from deposits, even though the crack mode is transgranular.
Unfortunately, because of their microstructure, the martensitic steels used in turbine blades (and
the alpha beta titanium alloy Ti 6-4) do not leave distinct striations that would indicate individual
cycle crack progression. Figure 5-9 shows a titanium blade failure by HCF, but no fatigue arrest
marks are visually evident. The martensitic microstructure cleaves and tears as the crack
progresses, and the resultant tears do not leave the classic striation marks that are evident when
examining austenitic alloys or aluminum.
Figure 5-8
CAF Fracture
5-15
10305538
Figure 5-9
Fracture Surface of a Ti 6-4 Blade
Another issue with steam turbine blades is that the resultant fractures are immediately placed into
a steam environment. The steam oxidizes the fracture surface, and the resultant oxides mask or
can obliterate the fracture surfaces. Another challenge encountered with fractographic
interpretation of the surfaces is that they can rub together during the cyclic loading experienced
by the blade. This mechanical damage can make the surface difficult to interpret.
Documentation of Fracture Surfaces
When tests are conducted on material removed from the exemplar blade, the fractures should be
examined and documented as well as the in-service fracture(s). The laboratory fractures were
produced under known conditions and can be compared to the in-service fracture(s). Although
many fracture features of various materials under various failure modes have been documented
and cataloged, fractures generated from exemplar material are preferable.
Low Magnification Examination
Fracture features should be documented at low magnification because some features are visible
at lower magnification. Fatigue beach marks (clam shell markings) are not visible at higher
magnifications. The general fracture condition (smooth and silky, shiny, faceted, and so on)
should be noted and documented. Figure 5-10 shows a typical macro-photo of an LP blade that
failed. Evident is a large corrosion pit at the leading edge. The macro-photograph yields a
tremendous amount of information about the fracture history.
5-16
10305538
Figure 5-10
Failure from HCF Initiating at Corrosion Pit on Leading Edge
High-Resolution Fractographic Examination
Finer fractographic features, such as fatigue striations and intergranular facets, require higher
magnification examination. Typically, these examinations are conducted using an SEM. Most
SEMs are equipped with chemical analytical equipment (energy dispersive X-ray spectroscopy
[EDS]) or wavelength dispersive spectroscopy (WDS) that can determine not only the elements
within the base metal, but also in any deposits on the fracture surface. Fracture surfaces should
be examined and analyzed before cleaning to determine the general condition and features of the
fracture surface. It will also allow for analysis of the composition of deposits on the fracture
surface. After cleaning, the surface should be reexamined and documented. Cleaning conducted
to remove deposits or oxides should be progressive in nature, beginning with mild detergents and
progressing to inhibited acids if needed. The fracture features should be examined and
documented after each cleaning step.
General Characteristics
Metals used in turbine alloys respond quite predictably to damage. The fractures produced are
permanent records of the damage and loads encountered. The trained investigator can interpret
the features and establish with high reliability the loading that led to the failures. The
macroscopic and microscopic observation of the fracture surfaces produced in the failed
hardware can be compared to exemplar fracture surfaces documented in the literature and
produced by exemplar testing. Correct interpretation of these data can ultimately identify the
mechanism(s) that were responsible for the observed damage.
5-17
10305538
The following are general characteristics of fractures in turbine blades:

• Overload. An overload fracture of a turbine blade has very distinctive features. The fracture
exhibits plasticity and usually a shear lip that is visible by visual examination. Examination
of the fracture in the SEM will generally reveal a cup-and-cone type fracture morphology,
which is typical of tensile overload. There is usually directionality to the fracture texture,
which will indicate the direction of the force causing the fracture. Fatigue fractures will have
overload regions. The size of these regions is proportional to the load applied at the time of
failure. This information can be quite useful.
• Blade deformation. Blade deformation has many causes, but the physical appearance of the
deformation is easily interpreted. It is very helpful to know the rotor rotation direction in
order to evaluate the cause for the deformation. Water induction or cool vapor induction can
cause non-directional distortion in that the blades will deform by thermal influences, as
opposed to mechanical influences. The blade airfoils will have plastic deformation and metal
distortion that are visible through optical examination.
• Fouling. Fouling of a turbine blade generally will not lead to blade fracture, but it can lead to
abnormal outlet pressures in the turbine, disruption of blade aerodynamics, and performance
losses. Fouling is most often caused by significant water chemistry upsets, especially at
seawater entrances into condensers. Fouling can lead to blade corrosion. The deposits can
trap corrosive species near the blade surfaces. These species can cause pitting, which can
lead to SCC or corrosion fatigue. These deposits can be removed from the airfoils by wet
cleaning or blasting, but removal from the blade roots is much more difficult. The blade
fixation design is a critical factor in the ability to flush the deposits from these crevices. In
most cases, blade removal is the most conservative approach. A less desirable approach is
rotor immersion in a liquid bath that is replenished based upon conductivity.
• Fatigue. Corresponding to the previous discussion of damage mechanisms that affect turbine
blades, the critical features sought from a detailed analysis of a blade fracture can generally
be related directly to the four stages of crack development: incubation, initiation,
propagation, and final rupture. It should be noted from the start that the relative zones on a
surviving fracture surface associated with each of these stages are not proportional, that is,
the portion of a crack reflecting each stage of development is normally very disproportional.
For blades, the most important evidence associated with incubation and initiation is often
confined to a very small area around the point of origin. However, the majority of fracture
will be due to propagation and final rupture. A detailed study of each area using various
analytical tools is an essential part of the investigation (see Section 8).
Incubation Through Rupture
Features Associated with Incubation
Roughly 50–70% of fatigue life is expended in crack initiation. During this time, the material is
being locally damaged, but the damage is on the sub-microscopic level. This damage is not
detectable by conventional nondestructive inspection or metallography. During this time, there is
local disruption of the blade’s crystal structure. Although it is possible to detect this damage
5-18
10305538
using analytical procedures, it is impractical. The lesson to be learned is that if a single blade in a
stage has failed by fatigue, adjacent blades are likely to have fatigue damage, too, even if NDE
does not identify cracked blades. In anodic CAF, the incubation time must include the time to
nucleate and grow a corrosion pit. This time is extremely variable. It depends upon the blade
alloy and the environment, the nature, and density of the overlying deposits. Generally, the
higher the strength of the blade alloy, the shorter the time to reach a critical pit size.
Features Associated with Initiation
Identification of the potential damage mechanisms starts with the initiation area commonly
referred to as the crack origin. Figures 5-11 through 5-14 show a series of examples selected
from different blade failures. As indicated by these examples, many are not exclusively due to
fatigue, but involve rows that have operated successfully for several years or longer. In rows that
have operated for some time, there is always a strong possibility of finding some evidence of
material degradation, such as corrosion or erosion, leading to deterioration of the fatigue
resistance over time. Steam turbine blading is subjected to cyclic stressing by design; as long as
the fatigue resistance of the blade is sufficient to resist these stresses, no failures will initiate.
However, environmental degradation can degrade the components by reducing the load-carrying
area (erosion or corrosion), raising the effective stresses, or creating stress concentration notches
(impact marks, erosion, or pitting corrosion). The presence of these is notable, but not
necessarily definitive. If a particular blade mode is vibrating in a manner such that the dynamic
stresses are moderate (not sufficient to initiate a crack, but large enough to propagate an existing
crack), they will naturally find any discontinuity within the region of maximum stress and favor
the largest.
5-19
10305538
Figure 5-11
Features of Crack Origins: Embrittlement, Inclusions, and Defects
5-20
10305538
Figure 5-12
Crack Origins: HCF and Fretting
5-21
10305538
Figure 5-13
Features of Crack Origins: CAF
5-22
10305538
Figure 5-14
Features of Crack Origins: SCC and Erosion
5-23
10305538
It is not always necessary to have a distinctive feature or defect at the origin. As shown in several
examples, the large centrifugal forces involved (due to the size or mass of a blade and speed of
rotation) often reduce the overall allowable crack size to a very small portion of the fracture
surface. The corresponding steady-state stress is therefore identified by the location of the origin
identified in the metallurgical examination. Because all blades vibrate, evidence of HCF is to be
expected, with the exception of high-energy impact types of failures. However, the growth rate
might be so small (on the borderline of the material’s threshold stress intensity) that in the region
associated with initiation, distinctive striations left by vibration might not be readily apparent,
certainly not to the degree found on the rest of the surface. This should be recognized so as not to
exclude the role that HCF might have played in the formation of the original crack, particularly if
no evidence of corrosion or erosion is found.
Features Associated with Crack Propagation
Moving beyond the origin, LCF and HCF fatigue normally exhibit a distinctive appearance, with
some features that are specific to the different alloys most commonly used in blades. Generic
features of these two mechanisms are shown in Figure 5-15. The phase of propagation often
dominates the visible portion of the crack surface. However, for turbine blades, the time
associated with this part of the crack is generally a small portion of the overall remaining life. As
a crack extends into and through the blade, the beach marks often become more pronounced,
even visible. This increase in visibility is offset somewhat by the microstructural influence. The
martensitic needles prevalent in the microstructure tend to delaminate under the advancing crack,
and the delamination spacing can be easily mistaken for striations. This is usually a reflection of
an increase in dynamic stress being affected by the corresponding change in blade stiffness. As
the vibratory stress increases, so does the rate of crack growth with each cycle. For turbine
blades, when HCF from a synchronous source of vibration assumes control of crack growth, time
is generally short and fracture is inevitable, unless luck happens to schedule an inspection that
coincides with this transition.
5-24
10305538
Figure 5-15
Features of LCF and HCF Crack Growth
5-25
10305538
Sometimes ridges are apparent over the zone of propagation. This might be caused by a
condition of resonance, which is temporarily created and then detuned as the crack advances. It
is a consequence of the blade’s natural frequencies migrating through the per-rev harmonics as
the mass and stiffness of the failing blade are changed. Beach marks are also often visible to the
naked eye. These discolorations, bands, or marks are left when a notable load change occurs.
Sometimes their number can be related to a distinctive type of operating event or condition. This
exercise must consider that only about 50–70% of a part’s fatigue life is represented by the
propagation portion of the fatigue crack. Thirty percent to 50% of the life is used in incubation
and initiation. Thus, during the first third to half of the unit’s operating history, events might not
be visible on the fracture surface. These arrest marks can be traced backward from final fracture
to estimate life events, but care must be exercised. For this reason, if start-stop cycling and LCF
are significant contributors to the formation of a crack, many of the individual marks confined
around the origin might be obscured or not fully registered on the micro level. In other words,
the visible evidence of distinct beach marks does not necessarily indicate exactly how many such
load changes or start-stop cycles were experienced from initiation to rupture.
Finally, although the trajectory of a crack is often a good indicator of the potential mode that
might be responsible for driving it, it is ultimately the role of analysis to identify the frequency of
the mode that is responsible. Estimating a natural frequency of vibration based on the HCF
striations left on the fracture surface is an exercise in futility because of the high frequencies at
which blades normally vibrate. If a catastrophic failure has occurred, the crack will end in
rupture, indicating the point when the remaining intact portion of the blade could no longer
sustain the stresses imposed by centrifugal force. It is therefore not unusual to find the topology
of a crack to include a disproportionate amount of the surface attributable to pure rupture as the
origin moves down toward the root of the blade. A large overload area is particularly apparent
for failures of large last stage and L-1 blades that have large centrifugal loads.
Depending on the source of fatigue, cracks in blade roots with multiple fingers or pairs of hooks
are often apparent at several different locations. Because of their redundancy, as one or more of
these features starts to fail, the proportion of load they carried is naturally shifted or redistributed
to the remaining intact attachments. This can involve a sequence of failures within the individual
root, evident as separations mixed with larger and smaller cracks. Nominal differences in blade
tolerances will also cause some slight differences in the ultimate disposition of the centrifugal
loads during operation and in blade natural frequencies. Therefore, in a blade root failure, evidence
of cracks formed within similar features should be sought. If multiple cracks are found, further
analysis might be required to consider the relationship between these redundant attachment
features.
Features Associated with Final Rupture
Fatigue failures or any type of progressive cracking will be terminated by a tensile overload
separation of the part. This overload occurs when the progressing crack consumes enough of the
part cross-section that the mean load exceeds the load-carrying capacity of the remaining cross-
section. The size of the final overload area is proportional to the load at fracture. In LCF-
dominated failures, the area tends to be larger than in HCF-dominated failures. Another
significant feature is the angle of the final shear relative to the part axis: this can indicate the
direction of the load at fracture.
5-26
10305538
Basic Features of Non-Fractures
Creep
Fracture of steam turbine blades by creep is extremely rare. The creep process in the common
blade alloys involves a considerable amount of plastic deformation as precursor to the creep.
This deformation would cause the blade airfoil or root to elongate, leading to tip radial rubbing
or blade lean (axial rubbing), which usually results in shutdown of the turbine before rupture can
occur. It is also in that the alloy of construction of rotors is usually low-alloy steel that has less
creep resistance than the blade alloy. All factors being equal, the rotor will have damage before
the blade.
Creep in turbine blades progresses from deformation (elongation) to microstructural void

formation to cracking. Metallographic examination can detect the void (cavity) formation and the
cracking.
Corrosion
Corrosion damage in steam turbine blades will generally occur as pitting. Most often, chlorides
will cause the pitting. General corrosion of steam turbine blading is rare. The blades are stainless
steels containing more than 11.5% chromium and, as such, are fairly resistant to general
corrosion. Pitting can and will develop when deposits accumulate on the blades, and the deposits
concentrate corrodent sufficiently to disrupt the passive oxide layer on the blades and lead to the
formation of an occluded corrosion cell.
The mechanism for corrosion in the blades is generally well understood. Both metallography and
SEM/EDS analysis can usually identify the source of the corrodent. Blade pitting damage is very
harmful. The pits can and will act as initiation sites for fatigue cracks. The corrodent in the pits
can accelerate the fatigue cracking. Both mechanisms are classified as corrosion fatigue.
Traces of deposits or oxide films or the presence of elements not indigenous to the blade
materials are evidence that corresponds to or corroborates possible contamination of the steam
chemistry at some time during the service history of the row. The ambiguity associated with
timing is intentional. Corrosive products naturally tend to seek out and concentrate in the
crevices of covers, tiewire or pinned root holes, joints like snubbers, or contacting surfaces such
as those found in integral covers and fir tree or dovetail roots. They remain even after the source
of the contamination is discovered and corrected. However, in themselves, these events and the
presence of films or deposits do not establish contamination as a root cause. If stress in the
region is low, widespread pitting or surface damage can be sustained without forming cracks.
More definitive information about the role contamination played is normally sought from an
examination of the fracture itself, with particular attention to the origin.
5-27
10305538
Erosion
Erosion is mechanical wastage of the blade airfoil by either solid particles entrained in the steam
or water droplets condensing from the steam. Erosion alone can make blades unserviceable
because reduction in airfoil cross-sections can cause efficiency losses and, eventually, blade
overload. Usually, erosion damage can and will act as a strong stress concentrator for any
stresses on the blades, either centrifugal or harmonic in nature.
SEM examination can differentiate WDE from particulate erosion. There are morphological
differences in their appearance, and in particulate erosion, there will be particulate embedded in
the blade surface. These embedded particles can be identified by EDS analysis. Metallographic
examination can confirm WDE. Cross-sections across the damaged areas will show tiny areas of
plastic deformation caused by implosion of the water droplets on impact.
Wear or Rubbing
Mechanical damage to blades caused by wear and rubbing is often optically visible, but it can be
confirmed by metallographic examination. Frictional heat from the rubbing will cause a
significant alteration to the microstructure at the rubbing location. This alteration can be
confirmed by metallography as transformed microstructure or a localized marked change in
metal hardness. Depending upon the severity of the rubbing, the metal can be tempered by the
heat that is generated. The hardness (or local recrystallization) forms a hardened layer (un-
tempered martensite) lying over a softened (tempered) layer.
Impact Damage
Impact damage to the blade airfoils is readily apparent by visual examination. The important task
is to identify if the objects causing the damage are domestic (that is, domestic object damage
[DOD]) or foreign (that is, FOD). Objects from outside the turbine are called foreign. These can
include any tools or debris left in the turbine during maintenance or fabrication. Domestic
objects are pieces of the turbine that have been liberated for whatever reason and flow
downstream, striking blades and nozzles in their path. As a general rule, the first row with impact
damage is the proximate entry point for the object. One exception is that the damage is usually
visible on the trailing edges of either the nozzle or rotating element immediately upstream.
Forensic investigation can usually identify the object causing the damage with some degree of
reliability. Two features make this identification possible. As the object strikes the blade, it
leaves an impact mark that has geometric features replicating the object. For example, a bolt will
leave thread imprints (one can even determine the thread pitch), and a hex bolt head will leave
120º angular indents. Another factor in determining the identity of the offending object is that
when the object strikes the blade, some of it will transfer to the blade airfoil. This transferred
metal can be analyzed using an SEM/EDS system to identify the composition of the object. This
can be compared to possible suspected parts and conclusions formed.
5-28
10305538
Interpretation of Results: Relation to RCA
At the closure of the laboratory analysis, the RCA team must interpret the meaning of the
findings. Hopefully, the laboratory investigation supports the parallel engineering analysis.
Features obtained by scanning the origin with an electron microscope are sought to make a
fundamental determination as to whether the phase of initiation was primarily the result of pure
fatigue (that is, excessive steady or dynamic stresses) or a process that involved corrosion,
erosion, or embrittlement around the origin. Fatigue points the finger to aspects of the blade
design. Corrosion or erosion raise questions regarding the operator’s control over chemistry and
moisture removal. Embrittlement or superficial damage left in stress-sensitive locations are
issues associated with the quality assurance (QA)/QC involving manufacture and installation of
the blades. Fretting is also a reflection of a looseness between mating surfaces that should by
design have maintained a firm contact during operation. Eventually, the identified origin should
be compared against the locations of peak dynamic stresses produced by the different
fundamental modes to see which corresponds. Specific answers sought from examination of the
fracture surface include the following:
• Where did the crack originate relative to the orientation of the blade airfoil?
• Was the origin of the crack apparent, and if so, was it of measurable size?
• Was the immediate zone of cracking around the origin transgranular or intergranular?
• Were oxides or corrosive products detected on the crack origin or surface?
• Was there any evidence that the material might have been over-tempered or overheated?
• Is there any evidence of fretting approximate to the origin?
• If pitting was evident around the origin, what was the average diameter/depth?
• Is there a transition point evident at some distance from the origin?
• Is the appearance of the crack surface typical of how the alloy looks when exposed to
fatigue?
• Was there any evidence that the initial crack might have been present for a significant
period?
• Is there evidence of beach marks or striations along the crack trajectory?
• Is there any point along the crack trajectory that indicated a change in applied loads?
The origin and crack trajectory identify where to look for the presence of a high steady stress
and/or the mode of vibration that might have promoted fatigue damage. If the initiating
mechanism is LCF or HCF from synchronous vibration, maximum stresses predicted from a
structural analysis should correspond to these sites.
Metallurgically, LCF can appear as beach marks on the fracture surface, signifying the
significant change in load experienced by the blade, normally when the unit goes from zero to
full speed and then back to zero. Growth due to LCF is fairly deliberate—normally, less than a
few mils with each cycle. Therefore, evidence of it might be confined to the area immediately
5-29
10305538
within the origin. This is particularly true for large steam turbine blades, where the steady
stresses produced during operation in these regions are often high. It is often LCF that will
transition the stress singularity created at the base of a pit, flaw, or surface indentation into a true
crack, with all of this occurring on a microscopic level. Occasionally, LCF will dominate the
initial crack growth. This is often corroborated by an analysis that indicates that the blades are
well detuned, resulting in relatively low dynamic stresses. Cracks formed and driven by LCF
particularly favor root attachments for which features of the geometry—such as pins, hooks, and
fingers—carry a majority of the overall blade from the platform up. Beach marks or staining can
highlight such demarcation points. Interim damage between such marks is typically not LCF,
but HCF extending the crack between start-stops.
The portion of a crack due to HCF will have a transgranular appearance, with marks emanating
from the crack origin and in the direction of the crack trajectory, reflecting the displacement
assumed for the mode of vibration that is predominantly responsible for the growth over that
portion. These marks tend to expand in width, and the surface between them becomes grainier as
the crack extends away from the origin. In a pure fatigue HCF failure, there is no substantial
evidence of corrosion or erosion causing a discontinuity at the origin. Pure HCF failures can
originate at machining marks, nicks, or other physical discontinuities. These features do not
necessarily play a determinant role in the formation of cracks, but simply reflect the most
favorable points within the zone of dynamic stress for initiation to occur. Supported by analysis,
the root cause is usually the result of synchronous vibration of a mode.
As a transitory condition, HCF due to asynchronous vibration such as flutter can result in a beach
mark type of pattern across the crack path. In conjunction with the results of the frequency and
stress analysis, this pattern is often very useful in further distinguishing the root cause where
evidence of HCF found on the fracture surface is the commonly shared mechanism. Rather than
a continuous march toward destruction, the pattern suggests a condition of high dynamic stress
that is only intermittently encountered over the course of extended periods of operation. In other
words, the crack is initiated and then driven for the duration of each event. It arrests when the
period of asynchronous vibration ends, then starts up again the next time the combination of load
and backpressures that causes the blade to vibrate is encountered.
As for any guidance offered in a blade failure investigation, there is always an exception. For
example, if a single asynchronous vibration event were particularly severe and of prolonged
duration, the abnormally high dynamic stress could both initiate and propagate the crack,
resulting in a final fracture, presuming the period of off-design operation was long enough. If the
same single event terminated but left a crack large enough so that the reduced dynamic stress
under normal operating conditions exceeded the threshold stress intensity of the material, the
result on the fracture surface would appear as a continuation of HCF. In such circumstances, it is
necessary to know if there is any evidence of off-design events and to know whether the
fundamental modes are well detuned in order to discern which of the two sources of HCF is the
root cause.
5-30
10305538
Examination of the origin should reveal if a distinct microscopic stress raiser facilitated the
initial formation of a crack. In such instances, an estimate of the pit size (width and depth)
should be made from the metallurgical examination to correlate against subsequent estimates of
threshold crack size made by means of FM models. Evidence is also sought of whether cracking
might have been promoted by a chemical process that attacked the cohesion of the grain
boundaries, versus the transgranular appearance usually left by fatigue. This can establish
whether corrosion or erosion can be ruled out or included as contributors or root causes.
Evidence in the condition of the grain boundaries is also sought; if found, it suggests that a
localized heat-affected zone might have made the material at the crack origin more vulnerable to
fatigue.
Fretting is often used as a convenient catchall to point to a source of damage that is operating-
related. Fretting is a mechanism that tends to produce pits, which act as stress concentrations.
The visual appearance of fretting damage is therefore very similar to corrosive pitting, and the
distinction between them is often difficult to identify. Fretting damage is characterized by
smooth, polished surfaces with intermingled pits. Fretting will produce a reddish oxide residue
that looks like jeweler’s rouge. This residue is completely iron oxide, that is, hematite. Pits from
corrosion are quite different in appearance, and the corrosive residues may be different in
composition. A feature of fretting is that it tends to form adjacent to the contact surfaces, rather
than in locations of maximum stress. When compared to the stress analysis results, the origin
identified in the metallurgical examination can help make this distinction. If evidence of fretting
is found between interlocking shrouds, special analytical studies might be considered to identify
the borderline when it is predicted that initial gaps will not close. The results would be expected
to correspond to evidence (gap sizes) assembled from the unit during initial installation or after
the failure.
Depending on the size of the blade and location of the crack, its length can be a relatively small
portion of the overall area defined by the blade geometry. It is not unusual to find the topology of
a crack that starts in fir tree roots of large, last stage blades to include a disproportionate amount
of the surface attributable to pure rupture. This is a reflection of the tremendous loads and forces
that these blades must withstand at full-speed operation. It further reflects the importance for the
pairs of hooks or fingers to be manufactured within precise tolerances in order to be certain that
the nominal gaps will naturally close and thereby correctly distribute the loads and forces
throughout the attachment. For this reason, large cracks (>0.5 in. [13 mm]) are not ordinarily
found when synchronous vibration plays a role in the failure. Large cracks are generally an
indication of a location on the blade where the steady loads and forces are marginal and/or an
indication that the dynamic stress in the region is also low. A large crack will often have
evidence of deposits or oxides on its surface that further testify to a low-stress field. When marks
are periodically found in such long cracks, this can be a further indicator of an asynchronous
vibration problem in which the damaging stress is abnormally increased, it moves the crack, then
growth arrests as the stresses return to normal levels. This is distinctive of the onrushing pattern
of striations leading up to rupture that HCF due to synchronous vibration typically exhibits.
5-31
10305538
Examples from Blade Failures
A series of examples are presented that show evidence of the issues previously described. The
failures are organized to start with examples that show features of those that occur in the blade
root, then move up to the airfoil, through the lashing lugs or tiewires, and finally to the covers.
As indicated by the survey statistics, failures of LP blades predominate, particularly those of the
L-0 and L-1 stages. Of note in these examples are the following points:
• The origin of the failures is not always consistent, even though the designs appear similar.
Figure 5-16 shows two root failures on fir tree designs that occurred on opposite sides, one
caused by SCC, the other by fretting and HCF. Even when failures occur at the same location
on the root, the region of unstable crack growth can be very different. This is shown in
Figure 5-17, where both failures were attributed to HCF from asynchronous vibration. In
these cases, the evidence left on the fracture surface was used to correlate with the two
different off-design operating histories, that is, one involving a series of intermittent
excursions, and one a single prolonged period of operation.
5-32
10305538
Figure 5-16
Examples of Root Cracking Axial Direction
5-33
10305538
Figure 5-17
Examples of Root Cracking Tangential Direction
• Similar to root failures, airfoil failures similar to those shown in Figure 5-18 through Figure
5-20 can start on the leading edge, trailing edge, or in the middle of the airfoil. In contrast to
the root failures, because less mass is supported by the airfoil above the fracture, the fracture
surface is often proportionally larger. However, the evidence left on this surface
predominantly reflects evidence of unstable, rapid growth due to HCF. As with root failures,
the origin remains key in distinguishing whether the crack formed from fatigue or was
assisted by erosion or corrosion.
5-34
10305538
Figure 5-18
Examples of Airfoil Failure: L-1 Blades
5-35
10305538
Figure 5-19
Examples of Airfoil Cracking: LP L-0 Blades
5-36
10305538
Figure 5-20
Examples of Airfoil Cracking: IP Blades
5-37
10305538
• Tiewires, lashing lugs, covers, and even integral shrouds are natural locations where stress
concentrations can form and cause failures, as shown in Figure 5-21 through Figure 5-23. As
noted in the tiewire-lug examples, cracks formed in these regions are sometimes treated as
maintenance problems rather than symptoms of a poorly detuned vibratory mode that is
overstressing the connection. Repairs such as welding or brazing can worsen the problem by
reducing the original strength of the material.
Figure 5-21
Examples of Welded Lashing Cracking and Failure: LP L-0 and L-1 Blades
5-38
10305538
Figure 5-22
Examples of Tenon/Cover Failures: L-1 and L-4 Blades
5-39
10305538
Figure 5-23
Examples of Root Integral Cover Failure: HP 1 and L-0 Blades
5-40
10305538
6
GENERIC DATA
Table 6-1
Subject:
As part of Phase 1, this section provides guidance associated with the use and application of
published information on related failures, which might preclude the need for proceeding with a
detailed engineering analysis. Discussed in this section is what type of information should be
sought, what sources are available, how to determine if the information is directly relevant to the
case at hand, and how it can be used in the final Phase 1 decision process described in the next
section of this report.
Topics Covered:
6.1 – Information Sought
6.2 – Interpreting Statistical Data
6.3 – Similar Versus Same Blade Designs
6.4 – Potential Sources
Tables/Figures:
None
Information Sought
Pursuant to an individual case of blade failure, general information is sought that indicates the
following:
• Whether there is any precedence for the problem
• Whether the present blade design is a consequence or replacement of earlier failures
• Whether a root cause of the failure has been completed
The principal reason for including this task in the initial phase of investigation is to determine if
a root cause investigation of the failure has actually been performed. If it has, is there sufficient
information available from the investigation to adopt the recommended solution without
pursuing the second phase of failure analysis? The possibility of immediately installing new,
reliable blades could represent an enormous savings in downtime and/or prevent the need to
reschedule an outage to perform this replacement while an interim solution is applied.
6-1
10305538
Generic Data
The two greatest obstacles to applying historical data are the lack of published information on
turbine blade failures and determination of whether any such data are truly design-specific.
Previously in this report, an industrywide survey conducted by EPRI was referenced to identify
basic trends that are expected to continue because long steam turbines are the basic means for
producing power. Conducted in the early 1980s, the information gathered on specific blade
designs (OEM, stage, and length) is somewhat outdated. Many of the designs that represent
scaled-up versions of the turbines of the 1940s and 1950s have been replaced. Many of the
problems reported to the survey involved those commonly referred to as infant mortality in
Weibull’s analysis of failures. Today, populations of these original blades still in service are
more likely to experience long-term wear and degradation problems. They are more often subject
to damage tolerance studies than root cause investigations because when it is determined that
they are at risk, they are replaced with new designs.
Interpreting Statistical Data
At some point in any investigation, an operator will be offered some statistical background
related to the problem. Other than the “never seen before” comment that must invariably be
stated, the second obstacle that such information presents is determining whether the data are
truly relevant or not. This relevancy extends to information that the OEM might produce that
reflects QA/QC processes and results that the design was reportedly subjected to before being
placed in service, along with enormous numbers of accumulated hours in service, with only a
few failures in all of this history.
Generally, these overall statistics are not meant to be misleading, but often honestly reflect the
fact that the vast majority of blades that are manufactured and installed in turbines do not fail.
This is a reflection of several facts involving blade design, as follows:
• Most OEM QA criteria and rules are adequate in that they catch the majority of problems that
might be caused by anomalies from tolerances, machining, or materials.
• The majority of designs with borderline issues that slip through the QA rules and criteria still
survive because the three basic controlling parameters of material strength, stress, and
environment do not come together (much like the locus formed in the SCC circles).
• Few designs reflect radical departures from the general experience amassed on steam
turbines over the last century. In cases where a design is altered in a dramatic manner (such
as replacing a tenon cover arrangement with an integral shroud), its serviceability is normally
determined by careful retesting.
• The materials presently used in steam turbine blades have been around for a while. Their
properties have been thoroughly documented, and their weaknesses are generally well
understood. Blades made from radically different materials, such as single crystal, or
directionally solidified are primarily found in land-based gas turbines.
• Because they represent such a small portion of a total population, truly generic failures often
take time before the numbers of reported events distinguishes itself in the statistics as more
than a random event.
6-2
10305538
Generic Data
To give some perspective of when a design problem might be considered statistically significant,
if the randomly combined factors of stress and material properties produce a chance of failure of
1:10,000 or lower, military standards recommended to govern system design safety programs
would classify such odds as remote or acceptable [7]. Operators of nuclear units will recognize
this relative chance as the same, which governs the acceptable limits or probability for a missile
failure to occur from a disk burst.
Similar Versus Same Blade Designs
The key to using any statistical or historical background is whether the information on hand is
genuinely representative of the design that failed. Although blade length is often a good indicator
that two designs have a high chance of being identical, this should never be automatically
assumed. Blades that are exposed to the most aggressive operating environments and therefore
suffer the greatest number of problems are also the ones that OEMs will adjust, modify, and
improve. Unannounced changes in certain tolerances of design features that make up an
assembled row of blades can often be discovered when samples are sent to independent third-
party suppliers or analysts for comparison to the dimensions and geometries that they have on
file. Sometimes, differences are discovered that were made to expeditiously correct non-
conformance problems during the manufacture or assembly of the rotor.
In some cases, the term same design refers strictly to the geometry of the individual blades, but
not how they are ultimately grouped when assembled around the turbine disk, or by what
devices. Same design family is a term that can mean that the generic features of the individual
blade are to scale, but not exactly to the same specific dimensions. Blades identical in geometry,
material, and length can behave differently when in service if they are regrouped using a cover of
shorter or longer length.
For the plant, it is therefore often necessary to establish that the blades at issue are identical. This
requirement has two potential consequences when used to make a decision whether to trust a
solution based on indirect but apparently related evidence. First, it will isolate the design and
service history that are not directly relevant to the design in hand, which failed. Second, it
requires any replacement strategies involving third-party suppliers to be based on a confirmation
of blades in kind based on reverse engineered samples.
Although the second situation would seem to involve common sense, it should be noted that in
any failure scenario, critical path time associated with turbine recovery is always lingering in the
background. A premium is given to any options that might save significant time in returning the
original turbine or unit to service. Many third-party suppliers have the electronic files on hand
for the most commonly replaced blades. However, the assumption that the design coming out of
the unit is identical to one previously manufactured has turned an advantage in reducing delivery
time into a new set of problems when it is discovered that certain details are visibly different.
Presuming that the obligations and liabilities in the original purchase order for producing in-kind
duplicates can be waived, the operator is left to determine whether the consequences of the
changes are as nominal as they appear.
6-3
10305538
Generic Data
Potential Sources
Generally, the two sources most likely to have detailed information on specific types of steam
turbine blade failures are the OEMs and insurers. If convinced that they have fixed a problem
and can offer a superior replacement, the OEMs often will supply a great deal of useful
information on the obsolete product. In particular, the operator is looking for details that
establish that the original design is the same and specific details from a root cause investigation
in terms of stresses, frequencies, and/or protective measures that are reflected in aspects of the
new design.
Insurers also have the opportunity to assemble a great deal of information that is specific to the
problems of steam turbine blades. Sorting this information by blade length and row can further
refine the statistics to those that might correspond to the present failure. In addition, if allowed
access to these statistics, the operator would want to know the date of the failure and some idea
of the duration that the rows had been in service. A dateline of events indicates whether the
number of reported incidents is increasing, stabilizing, or decreasing relative to the time of the
present investigation. In other words, is there any evidence that the problem has been fixed?
Timing (hours/starts) between installation and failure is often a more useful perspective in that it
can be directly correlated or compared to life consumption models.
Beyond that, the lack of an explicit definition that differentiates iterations of the same blade
design might not be available to the insurers as a means to narrow their statistical data to those
cases that are relevant. Another problem evident in the statistical survey is that if only
catastrophic occurrences are reported, the true extent of a problem is not going to be immediately
apparent in the resulting data. As mentioned earlier in the report, repeated cracking of welded or
floating tiewires was often considered and handled as a maintenance issue, whereas in fact it was
symptomatic of a design-related problem with the operating frequencies. Similarly, cracks from
LCF or SCC are normally caught before they are allowed to fracture. In the context of this
report, these are also considered a failure because they might prompt a root cause investigation.
Although EPRI did a number of fundamental surveys of blade failures, bearing failures, disk
SCC, and so on, most of these were performed prior to deregulation, when exchange of
information was promoted or at least allowed. The possibility of updating these databases is
severely limited in the age of competition. However, databases are available that identify units
that share common turbine designs. These can be useful in identifying operators who have a
common interest in relating to any failure experiences. In some circumstances, when a number of
operators found themselves confronted by the same problem, working groups formed to cost-
share the root cause investigation and cross-reference their field experience.
Summary Comments
A source of potentially useful or relevant information can sometimes be found in publications

where individual root cause investigations are discussed. Sources include the proceedings of
steam turbine reliability and power generation conferences held by EPRI, the American Society
of Mechanical Engineers (ASME), the American Power Conference, and the annual PowerGen
Conference, to name a few of the most prominent that feature power generation equipment.
6-4
10305538
Generic Data
Papers submitted by the prominent OEMs often focus on the development of their most recent
designs. Within these papers, issues (or problems) are sometimes discussed that might be of
potential relevance to diagnosing a failure at hand. Discussions of newly developed rotor and
blade products often help define the features of the technology and generic steps that the OEM
has undertaken to qualify the new components. If nothing else, these establish some timeline and
background on how long certain features of the blade designs have actually been in active
service.
Case histories describing the results of independent blade failures are often found scattered
within the proceedings of the aforementioned conferences. The chance for one of these to be
directly related to a case at hand is not great. However, they do offer insight as to how different
problems manifest themselves, and they further illuminate some of the investigative techniques
discussed in this report.
6-5
10305538
10305538
7
WEIGHING PHASE 1 EVIDENCE
Table 7-1
Subject:
This section represents the end point of Phase 1, providing additional guidance on how to weigh the
evidence at hand in order to decide if a detailed engineering analysis (Phase 2) is still required. If
the decision is no, it means that modification to the original blade design is considered unnecessary
or that a modified replacement has been satisfactorily demonstrated. It further presumes that a
proposed corrective strategy is available to be executed as soon as possible, without further
investigation. If a decision is made to proceed with a detailed engineering analysis, additional
guidance is provided in the next section of this report.
Topics Covered:
7.1 – Reasons to Execute a Replacement Strategy Without Further Analysis
7.2 – When to Proceed with an Engineering Analysis (Phase 2)
7.3 – Permanent Strategies Sought from Phase 2 Investigation
Tables/Figures:
Table 7-2 provides a summary of alternative strategies used in the planning and management of
blade failures.
Reasons to Execute a Replacement Strategy Without Further Analysis
A decision to execute a proposed replacement strategy (without further analysis) should be made
for one of two reasons only. The first possible reason is that the evidence at hand genuinely
indicates that the failure was the random product of external events that by chance compromised
the inherent integrity of the original design. The second possible reason is that the evidence at
hand clearly demonstrates that the root cause of the failure has been thoroughly investigated and
that the proposed replacement has established a successful record of service.
7-1
10305538
Weighing Phase 1 Evidence
Random Failures
In the first possibility where a detailed analysis is not required, supporting evidence gathered
from the operating history, site forensics, and/or metallurgical analysis of samples from the row
that failed clearly demonstrate what caused the blade failure and how. Under such circumstances,
the root cause will be attributable to one of the following:
• An external operating event that would have caused the blade operating stresses to exceed
their design limits
• A failure in QA/QC that allowed substandard materials (chemistry or properties) to be
improperly used
• Obvious defects or damage from foreign objects that got into the steam path
• A failure in QA/QC associated with an aspect of the row’s installation that was critical to the
reliable operation of the design
As stated in the previous sections associated with site forensics and metallurgical analysis, the
evidence at hand should be unambiguous and logical. Unambiguous means that a reasonably
acceptable cause for the distress found on the blades can be corrected without changing anything
in the original design. Evidence of some off-design operation or material that might have been
work-hardened can be relevant, but it does not meet this level of certainty.
Logical means that there is no contradiction between elements of what was seen when the
turbine was opened, what was learned about the operating history prior to the failure, and what
the metallurgical analysis revealed about the chemistry and strength of the material. For
example, if asynchronous vibration or torsional system interaction is blamed, the operating
records should back this up, and the fracture surface should reveal a pattern of HCF that
corresponds in a reasonable manner with the number or duration when such events might have
been experienced. If the crack originated in pits or multiple cracks were found in blades spread
around the disk, it would suggest the possibility of CAF or LCF.
Replacement with a Demonstrated Redesign
In the second possibility, where a detailed analysis is not required, two conditions are met. First,
it is presumed that the evaluation of prior industry experience was conclusive enough to
determine that the investigative efforts are directly transferable to the problem at hand. Second,
the proposed replacement has sufficient operating experience for the plant to anticipate that it
will perform with equivalent reliability. In this second outcome, some modification to the design
is expected. The change might be as simple as regrouping the number of blades around the row
or as radical as a complete change of the design from the root attachment up. Where any change
is reflected in the proposed replacement, the supporting documentation should clearly
demonstrate with technical detail how the change was selected and why it is expected to work.
The reason for this is simple. Whenever changes are introduced, there is always a chance that by
doing so, an unanticipated problem might be created. Often, correction of a problem of repeated
cracking at highly stressed features is attempted by simply eliminating the feature. Shroud
7-2
10305538
cracks, tiewire hole cracks, and tenon cracks can be indisputably corrected by changing to an
integral shroud or a freestanding blade design. However, even an innocuously basic change, such
as regrouping a blade from five to six blades per group, can significantly adjust the original
natural frequencies to the degree that unless due diligence has been performed to qualify the
effect of these changes, a new condition of resonant vibration might be encountered.
Documentation that demonstrates that the OEM has addressed this concern is sought in the form
of details of comparative stresses and frequencies. As a consequence of the modification, it
should be clearly shown in non-normalized plots how localized stresses of the replacement
remain low enough to mitigate the premature initiation of fatigue cracks throughout the modified
design. Particular attention is placed on the remaining locations of stress concentration.
Frequencies (again, non-normalized) should demonstrate that the principal modes of vibration
remain detuned. In this regard, generic evidence involving similar designs or design families,
normalized stresses, or Campbell diagrams with the frequency scale hidden should not be
considered definitive. Test data that are from a similar design but not an exact duplicate of the
proposed blade replacement should also not be considered to be definitive.
In addition to technical documentation, the decision to execute the replacement strategy without
further examination or investigation should be supplemented with unimpeachable evidence that
the new design has operated successfully without symptoms of the original failure or evidence of
new problems. Given all of the advantages that can be gained from a detailed analysis, the most
conclusive evidence of a truly corrected problem is still a period of successful field service. This
is because no analysis can anticipate or represent all of the variables that might affect the
reliability of blades in a fleet of turbines where the operating conditions and service duty are
unique. Often, establishing the service history of a new design again depends on the evidence (or
lack thereof) obtained in Phase 1. When reviewing rotor replacement bid packages, the list of
successful applications of similar technology consisting of several hundred rows and rotors is
often reduced to a handful when simple distinctions are made between 50-hertz and 60-hertz
applications, last stage blade lengths of the LP turbines, and the actual installation date (not the
purchase date) when the rotors started seeing service.
As a rule, synchronous vibration problems (failures caused by inadequate detuning of critical

modes) will often make themselves known in six months or less. Particularly severe LCF
problems (indicating an excessive steady stress load on the attachments) have been known to
appear as early as several hundred to a few thousand start-stops. Presuming a normal operating
environment and moderate levels of stress, time-dependent creep and SCC problems are
normally expected to take 100,000–150,000 hours to manifest. Erosion-assisted fatigue and CAF
problems are less predictable because they depend on the severity of the agents present within
the steam. No timelines can be assigned to problems from asynchronous vibration or torsional
system interaction because they are more dependent on the duration and severity of each event as
well as the number that might have occurred.
A summary of approaches to life cycle management (LCM) of LP turbine blade failures is

presented in Table 7-2.
7-3
10305538
Table 7-2
Alternatives for Planning and Management of Blade Failures [6]
Alternatives, Strategies, Contingencies TYPE FEATURES ECONOMICS Action is recommended:
If a failure has occurred in hours, days, or weeks, a design weakness is likely to
MANAGEMEMENT ISSUE: LP turbine require a modification or correction. Otherwise, damage tolerance limits based
blade failures on detailed analysis need to be available when indications or cracks become
Improve Historical Reliability

large enough to detect by conventional NDE methods, particularly if damage is
Estimated Direct Costs (in

CONTINGENCIES
ALTERNATIVES
More Power Generation

CAUSE: HCF: Inadequate design. such that a limited refurbishment might be possible to salvage multiple rows of
STRATEGIES
Reduce Historical Risk

LCF,SCC: stress, cycling, chemistry. otherwise structurally sound blades.
Potential Net Power

Improve Availability
Downtime (Hours)
TIMING: HCF: >2000 hours. LCF, SCC:
130,000–175,000 hours.
Preventive
Corrective
dollars)
FACTORS: Resonance, hours, cycles, PRINCIPAL BENEFITS POTENTIAL RISKS
FOD, off-design operation.
0 Continue NDE: Use finite element Y N N N N 24–36 25K - Contingencies A1–A3 all depend New designs with no service history are
analysis results to guide inspection. on having a detailed stress analysis more likely to experience premature
CONTINUE TO RETIREMENT
of the parts. failures.
1 Premature Failure: Replace with Y Y N Y N 1008– 300K– - Involves minor modification to Success depends on establishing
LCM PLAN A
corrected design based on RCA . 1344 750K correct a design deficiency. May be correct root cause. Replacement in kind
able to salvage rows. is not advised.
2 Pitting, FOD: Perform limited grinding in Y 24–72 50K - Avoids major outage extension and Depends on knowing tolerance limits,
regions of maximum stress. defers purchase of one or more amount of conservatism in design, and
blade rows. extent of damage.
3 LCF, SCC: Remove blades. Install Y 168– -0.1 Avoids major outage extension and Requires major work to rotor.
pressure plates (assuming plate is 336 250K– to defers purchase of one or more Purchasing of pressure plates.
available). 500K -0.5 blade rows. Significant loss of power.
1 Replace rows with original design. Y N N N N 672– 300K– - Reset original service life to Not considered a corrective option
REPLACE/UPGRADE
840 750K historical rates and duration. Same where a premature design failure is
level of maintenance. involved.
LCM PLAN B
2 Replace rows with new structural design Y Y N Y N 672– 300K– - Reduce regular maintenance costs Unverified changes to the original design
features. 840 750K (repairs). Aerodynamic design may cause a new or potentially more
unchanged. damaging problem.
3 Replace rows with an upgraded design. Y Y N Y Y 672– 400K– 0.1 Aerodynamic improvements are May require major work to unit, possibly
840 1000K to coupled with improved materials, to modify end of LP section to
0.5 stress, and service life. accommodate design.
References: Potential risks associated with premature failure of new designs and replacement in kind are based on experiences published in the EPRI report Survey of Steam Turbine Blade Failures (CS-399). Weld
repairs are discussed in the EPPRI report State-of-the-Art Weld Repair Technology for Rotating Components, Volume 2: Repair of Steam Turbine Blading (TR-107021-V2). Procurement and inspection criteria for
blades are found in the EPRI report Guidelines for Reducing the Time and Cost of Turbine-Generator Maintenance Overhauls and Inspections, Volume 4: Blade-Bucket Procurement and Refurbishment Procedures;
Volume 6: HP/IP Blade Design Audit and Inspection Criteria;, Volume 7: LP Blade and Disk Design Audit and Inspection Criteria (TR-114128).
Technical considerations/comments: After three to six months of operations, the major concern for blades/buckets is related to issues of age and degradation, for example, SCC, LCF, and erosion. To run, repair, or
refurbish these parts requires analysis of damage tolerance limits coupled with results from regular inspections. It is recommended that FEMs be prepared in advance of outages so that the probability and risks can
be evaluated as soon as inspection data are available.
7-4
10305538
When to Proceed with an Engineering Analysis (Phase 2)
It should normally be decided that a detailed evaluation of the blade design is warranted if one
or more of the following conditions are met:
• A search of the operating records reveals no event that would explain conclusively why the
blade failed.
• The metallurgical analysis indicates that the material was not substandard and reports
evidence of HCF on the fracture surface.
• Attempts to find similar experiences where the problem was clearly corrected have been
unsuccessful or found mixed results.
• The principal reason for replacing the blades with the same design is logistical, a reaction to
the urgency of restoring the unit to service.
• The metallurgical analysis identifies evidence of HCF, but it cannot further identify the
specific mode of vibration that is responsible.
Often, circumstantial evidence gathered in Phase 1 that is not of sufficient detail to inform a
conclusive judgment as to the root cause can still assist in framing the specific issues to be
addressed with Phase 2. Examples of this are as follows:
• A number of circumstantial clues can point to LCF as a potential mechanism. Because LCF
is promoted by steady stress, often a large number of blades are found cracked, rather than a
select few. This is because the steady stress is generally on the same order of magnitude for
all of the blades assembled within a row. The location of cracking is also revealing. Root
attachments, lashing lugs, and cover-tenons are all more susceptible to LCF than the airfoil
proper. Evidence of multiple blades cracking at these locations would also point to an LCF
problem. Because growth due to LCF is relatively slow and deliberate, problems are often
found by inspection, before catastrophic failure occurs.
• Similar to LCF, SCC in an advanced state tends to appear on a large number of blades rather
than a select few because it is promoted by steady stress. Circumstantial evidence that might
be used to differentiate SCC from LCF is that it is related to hours, not cycles, and thus tends
to take a significant amount of time to develop, particularly in a unit that does not cycle
frequently. SCC does not necessarily form in the regions of maximum stress, but it is often
found on the lands or grooves of the disk attachment corresponding to the blade root
attachments. These are the crevices that allow precipitates to concentrate. The disk material
is less resistant because of its lower chrome content, and there even might be some galvanic
action involved between the two dissimilar materials. SCC can also occur in other crevice
locations, such as tiewire holes and tenons beneath the shrouds. SCC requires the presence of
moisture (hence, it tends to form in the rows operated approximate to the Wilson line),
whereas LCF does not.
• Blade failures caused exclusively by resonance (synchronous vibration) tend to be more
selective because the magnitude of dynamic stress is very dependent on the proximity of a
mode to the nearest forcing frequency. Cracks forming in the blade above the platform are
often symptomatic of a higher mode of vibration being involved, although they can also
7-5
10305538
produce cracks in the root attachments. Because no two blades are exact duplicates due to
slight differences from materials, manufacturing, and assembly tolerances, this makes one or
more blades act as lead indicators of a widespread frequency problem. Because a blade
failure from resonant vibration and HCF invariably results in fracture and release of some
portion of the blade, the failure is usually discovered by either a step increase in vibration or
when the unit is automatically tripped off-line. In other words, the cracks form and propagate
so quickly that evidence of the fatigue damage to other blades might not be immediately
apparent, particularly if the location affected is highly stressed, reducing the critical crack
length to the microscopic scale.
• HCF fatigue from resonance is also more commonly anticipated for longer LP blades, where
resonant stresses can become extremely high. In shorter blades like those found in HP and IP
turbines, resonant stresses are less of a concern because the steam pressures involved demand
shorter, less tapered designs. This produces higher natural frequencies that fall within a
regime where the nearest per-rev forcings (with the exception of NPF) are traditionally
reduced. The lower stimulus is matched with a more rugged structure, thereby making it
more capable of accommodating any dynamic stresses that are produced. This is a general
perception held within the industry, but it is not necessarily true. Cases of high-pressure
blades failing within months of installation have occurred.
• Resonance is a non-linear condition and tends to be an all-or-nothing source of HCF.
Typically, it will make itself known in a relatively short period if a fundamental mode of
vibration is being excited. However, it can sometimes affect a row or design that has
otherwise served reliably for many years if a critical mode is only marginally detuned by a
few hertz so as not to be in a resonant condition when first installed. Because frequencies are
controlled by the stiffness and mass of the blades, older designs that rely on connecting
devices can sometimes be affected as deposits build up and stiffen the interfaces between
these features. If any modes are marginally detuned, this might cause the frequency to
migrate and result in a resonant fatigue failure.
• Asynchronous vibration (flutter) tends to be a problem exclusive to large last stage turbine
blade designs. Technically described as a condition of instability or negative damping, it is a
potential suspect when failure involves blades that are freestanding or that do not have covers
at the tips to resist the vibratory motion and/or provide frictional damping. Because the first
mode is involved, cracks often form in the fillets of roots that have fir tree type designs and
axial entry blades that have pinned root designs. If the damage was accumulated from a
larger number of limited events—which is more often the case—failures attributable to
flutter often share characteristics of those involving LCF. They tend to involve a number of
blades around the circumference rather than a discrete few.
Permanent Strategies Sought from Phase 2 Investigation
To correct an LCF problem, a design change is necessary to reduce or relieve the steady stresses
within these locations. An unstated concern over LCF is often reflected in replacements that have
enlarged fillets or radii, reduced the weight of the blade, repositioned the airfoil on the platform,
reduced the thickness of covers, or eliminated sharp transitions in the blade geometry.
7-6
10305538
In any circumstance where resonant vibration is suspected as the root cause, the critical modes
responsible for initiating cracks must be adjusted so that their frequencies no longer coincide
with a multiple of running speed. There are multiple approaches to achieve this, all of which
essentially require an adjustment to the mass and/or stiffness of the design.
Critical to any detuning strategy is recognition that the blade disk is a dynamic system. Changes
to affect one mode can change others. Also, any changes made must adequately account for the
influence of stress-stiffening and spin-softening that causes the frequencies while the blades are
at rest from being different from what they are at full speed. Sometimes, changes characterized
as upgrades to optimize blade efficiency must also be reviewed. When changes in the profiles
forming the stacking line within the airfoil are made, they can affect the mass and stiffness of the
original blade.
In a similar vein, blades that are supplied by third parties who rely exclusively on reverse
engineering samples will define manufacturing tolerances from these samples, without any
knowledge of their operating natural frequencies. Inadvertently, the change in tolerances can end
up causing a mode to become resonant. If the blade-to-blade tolerances are held to more
consistent standards, as is possible with today’s technology, the number of blades affected can be
significantly increased.
Asynchronous vibration problems are not as straightforward in terms of designing a corrective

strategy. More often, asynchronous problems are corrected by identifying zones of
corresponding load and backpressures that should be avoided. In principal, corrective measures
associated with flutter involve the introduction of more damping with tiewires, covers, and so on
or by mixed tuning of the blades, that is, where the blades are purposely installed to have
monomial differences in their frequencies.
Because of their higher chrome content, SCC is traditionally considered a disk, not a blade,
problem. However, if blades are left in service long enough (25–30 years), they have been
known to develop SCC. An entire industry has formed to correct SCC in rotors and disk
attachments. In some cases, an enlargement of fillets and radii in root attachments is apparent as
part of a replacement design. However, because the stresses in these attachments inherently
exceed the KISCC, strategies concerned with addressing SCC are generally indicated by the
addition of coatings or use of more resistant materials to inhibit or extend the incubation phase.
Issues associated with erosion and corrosion are often corrected by enhancing the water
extraction or filtering systems, controlling the steam quality, or adopting protective measures for
the regions on the blades that have proven susceptible to such damage.
Final Comments
For large steam turbine-generators, the determination of whether to proceed with a detailed
analysis or defer this until the inspection, operating history, and metallurgical analysis evidence
becomes available is often a luxury that time does not permit. As a rule, truly random failures are
fairly evident and clearly supported by some critical facts that are indisputable. With some luck,
the operator might also find that they are not the first to have encountered the problem, and the
7-7
10305538
work to develop a reliable replacement is done. However, in cases where there is any doubt
about the diagnosis or effectiveness of the fix, it is recommended that the operator provide for
the remaining steps of a typical blade failure investigation and initiate these as soon as possible,
given the time that is generally required to do the basic work involved. Even if the OEM is
undertaking an investigation, it is in the plant’s best interest to pursue these steps independently
so that it retains control over the situation, particularly the time involved to identify an interim or
permanent solution.
If there is any vagueness in the explanation given in terms of the root cause, or evidence from the
three initial sources is contradictory rather than complementary, it is the role of analysis to
resolve these conflicts. If the proposed replacement has only a limited history of service, it is
possible to predict only what can be expected from the design modifications by an extension of
the analysis with life models.
Finally, if it is determined that the problem is related to premature erosion or corrosion, the
initial focus of the next phase of analysis should be to assess damage tolerance limits versus root
cause. In such a scenario, the basic steps outlined in the next section will remain the same; only
the immediate objective will change from identifying a replacement strategy to identifying an
interim solution until such a strategy becomes available.
7-8
10305538
8
PHASE 2: ENGINEERING ANALYSIS
Table 8-1
Subject:
The multiple steps in Phase 2 are meant to identify operating stresses (steady, dynamic, or thermal)
and to estimate life consumption as crack initiation or propagation. These results can then be
correlated against the operating history, inspection records, and metallurgical findings assembled in
the previous phase to identify and/or confirm the root cause(s) of the blade failure. Also discussed is
the support required from the plant, details of the calculations that should be performed, and tests
that can be used to verify the analytical model and the results they produce.
Topics Covered:
8.1 – General Overview
8.2 – Plant Support: Geometry, Operation, and Installation Records
8.3 – Finite Element Modeling
8.4 – Calculating Steady-State Stresses
8.5 – Natural Frequencies and Mode Shapes
8.6 – Calculating Dynamic Stress
8.7 – Model Verification by Frequency and Vibration Testing
8.8 – Life Prediction and Probability of Failure
Tables/Figures:
Table 8-2 provides a summary of material properties that are typically required as input to the finite
element or life prediction models corresponding to the different mechanisms shown in Table 3-2.
Table 8-3 itemizes the information that the plant should assemble and provide in support of an
independent engineering analysis of a turbine bladed disk.
Figures 8-1 and 8-2 illustrate the basic dimensions that should be obtained when the row is exposed
for inspection, that is, a cross-section of the rotor/disk profile and blade outer tip diameter, and what
additional information is needed to describe the assembled stage.
Figures 8-3 through 8-4 show examples of critical regions in the FEM where maximum stresses
often occur.
Figures 8-5 through 8-8 illustrate the dynamic behavior of bladed disks, input used to represent
damping, and the corresponding plots used to identify the potential for resonant vibration to occur.
Figures 8-9 through 8-11 show examples of various frequency and vibration tests that are often
applied to verify the predicted natural frequencies of bladed disk assemblies.
Figures 8-12 through 8-20 illustrate the process of estimating crack initiation, HTC, and crack
propagation in different regions of the blade design with examples of applying the life models.
8-1
10305538
Phase 2: Engineering Analysis
General Overview
Nine tasks within the original methodology are identified as Phase 2 activities. Pursuit of this
phase has been determined as necessary because the original failure cannot be explained as
simply a random event, nor has sufficient information been produced by the OEM or a search of
published literature that would sufficiently demonstrate a viable replacement strategy. At issue in
this phase is whether features of the blade design itself might have played a principal or
secondary role in the failure that occurred. Analysis is used to provide critical information on the
stresses that occur throughout the bladed disk under simulated operating conditions.
Stresses are the key data and focus of work performed in this phase. They are initially calculated
for baseload operating conditions, with particular attention to the location(s) where distress was
identified from the site inspection and metallurgical inspection. Depending on what these results
show, additional calculations might be considered to quantify the effect of any off-design
operation identified from review of the pre-failure service history or of certain critical tolerances
where gaps were meant to close and/or contacts achieved.
The multiple steps of Phase 2 follow the following sequence of three interrelated activities:
1. Assemble relevant input to develop the FEM. The objectives of these tasks are focused on
setting up a structural analysis, defining the appropriate materials involved, and tailoring the
calculations in accordance with the circumstances surrounding the failure.
2. Calculate natural frequencies and operating stresses. The objectives of these tasks are to
simulate and quantify the response of the blade design to steady, dynamic, and thermal loads
(if necessary). The frequency analysis is used as a means to verify the model and calibrate
the potential magnitude of the vibratory stresses.
3. Predict life consumption. The objective of this task is to take the stresses at critical locations
and estimate the damage they would cause as a means to separate the principal mechanisms
and root causes from secondary contributors.
Plant Support: Geometry, Operation, and Installation Records
The technical examination always starts by establishing features of the design and a basic
understanding of the circumstances surrounding the failure. Much of this data should already be
available, unless it has been decided to initiate phases 1 and 2 simultaneously for the sake of
expediency. The original data assembled in Phase 1 should be forwarded to assist in determining
how long the specific design has been in service and/or how many changes have been made to it.
Technical specialists will further extend or supplement this information by relying on in-house
databases of materials and design, as required to build the FEM and run the battery of
calculations. To support a detailed analysis, basic questions that an operator can expect include
the following:
• What is the blade material?
• What are the blade length, model number, and/or vintage (age) of the design?
• What are the basic dimensions of the blade, including tiewires, shrouds, snubbers, and so on?
8-2
10305538
• What is the blade count, grouping, outer tip diameter, and general profile of the disk/rotor?
• For blades with integral shrouds, what is the specified contact at zero rpm?
Material Properties Needed
Table 8-2 presents a standard list of material properties that are usually required as input to the
FEM or as input to the respective life model to characterize the different mechanisms originally
shown on Table 2-2. Typical data for two common turbine alloys are also shown.
Table 8-2
Material Properties Used in Steam Turbine RCA
Parameter Symbol Unit 403 12% Cr 17-4 PH Use
Young’s modulus E ksi 29,000 29,000 FEM
Poisson’s ratio ν 0.3 0.3 FEM
Density ρ lb/in3 0.283 0.283 FEM
Tensile strength σult ksi 135.4 150.0 FEM
Yield strength σy ksi 110 130 FEM
Hardening coefficient K ksi 156.1 240.1 LCF model
Hardening exponent n 0.12294 0.093 LCF model
Fatigue ductility coefficient εf in/in 0.24655 0.1680 LCF model
Fatigue ductility exponent c -0.5456 -0.72624 LCF model
Fatigue strength coefficient σf ksi 131.2 175.4 HCF model
Fatigue strength exponent b -0.0641 -0.704 HCF model
Fracture toughness ΔKIC ksi-in1/2 200 245 FM model
Threshold stress intensity ΔKth ksi-in1/2 3.0 7.76 FM model
Stress corrosion resistance KISCC ksi-in1/2 35 30 SCC model
To estimate LCF and HCF initiation life, the basic properties and coefficients are sought to
define a strain-life curve of the material (see Figure 2-2). These are used in conjunction with the
local strain approach to using stresses obtained from the FEM at selected locations of interest
and estimating the number of cycles or hours before initiation is predicted. Locations of interest
are regions where cracks formed, identified from the visual and metallurgical inspections
performed in the previous phase.
To characterize the remaining life associated with propagation, the threshold stress intensity is
typically the most important parameter (see Figure 2-3). For special types of propagation
problems (such as SCC), the stress intensity threshold is identified as KISCC to account for the
effect of the environment. High-temperature damage mechanisms, such as material creep and
8-3
10305538
TMF, are evaluated with different life models that use thermal stress obtained from the analysis,
with parameters defined from tests performed at various temperatures. Notably, these tests are
costly and much more difficult to perform than standard LCF and HCF tests. SCC and creep
tests are often accelerated by exposing the specimens to concentrated solutions or high stresses
at elevated temperatures because the normal process at the lower end of the curve would take
years to complete.
Mechanisms where the initial damage is dominated by a product of the environment, such as
corrosion fatigue, WDE, and SPE, are generally not characterized as material-related properties,
but more often treated as formulations that attempt to model the process of material degradation
or loss. Relative to the investigation of a blade failure, these are pursued as crack propagation
issues, using either the present or a projected condition of the damaged airfoil as a basis to
determine when HCF will take control and fail the blade, thereby relying on the material
properties that are defined and available. If the rate of erosion is the driving force, it is usually
best estimated from prior inspection records.
Operating Records and Reports Needed
Table 8-3 lists the principal items that are sought during the course of any normal blade failure
investigation.
Table 8-3
Additional Information Required to Support RCA
Item Use
1 Heat balance A heat balance diagram of the turbine is typically found in its thermal kit and is
diagram to define used to establish the operating conditions for the specific stage of interest.
stage loads and
temperatures
2 Turbine cross- A cross-section of the turbine is normally found in the operator’s manual and
section that normally provides sufficient details on the disk/rotor geometry to complement
shows turbine the information obtained by reverse engineering a sample blade.
configuration and
rotor/disk profile
3 Blade count to A basic count of the number of blades in the row is needed to assemble the
complete the final FEM.
finite element
blade/disk model
4 History of If rows from spare rotors were swapped out or new rows were replaced, these
starts/hours to details should be identified; the total number of unit hours can be misleading. A
establish limits comparative record of starts and hours for rows in any sister unit’s hours and
for life starts should also be assembled to assist in determining how threatened other
assessment rows might be and/or if there is any trend apparent between the hours/starts
models and discovery of the blade problem.
8-4
10305538
Table 8-3 (continued)

Additional Information Required to Support RCA
Item Use
5 Record of off- Off-design conditions of interest are generally related to events of low load and
design events to high backpressure or high load and high backpressure that approached or
consider or exceeded the envelope prescribed for the unit. For HP blades operating in
eliminate as a control stages, a breakdown of the hours spent at different arc admissions
possible cause should be assembled. For units where significant grid instability is possible, an
effort should be made to record how many times and for how long the unit
might have operated in a frequency range beyond 1.5 hertz of rotating speed.
Other events that might be relevant would include failure of the condensing
system (resulting in contamination of the steam), breakdown of the drainage
system causing excessive water to be present in the steam, and other
anomalies that might affect the turbine blades.
6 Metallurgical Reports are sought on the immediate failure and on any preceding failures that
reports to confirm were originally dismissed as random events. Crack or pit depth measurements
or eliminate establish initial limits for any FM models, and they assess the probability for a
potential failure as the crack grows in size.
mechanisms
7 NDE test results It should be noted that if the last NDE prior to the failure reported no detectable
as input to indications, this information is just as useful and revealing when these
damage calculations are performed. If the problem involves premature erosion,
tolerance/risk knowing the period between the last inspections and/or if erosion was noted
models provides a basis to estimate the rate at which it is progressing.
8 Campbell This should be requested from the OEM as a means to compare against the
diagram (from predicted frequencies from the independent analysis.
OEM) to verify
frequency
calculations
9 Failure reports It is more efficient to make the investigator fully aware of work that preceded
(OEM) to their current efforts so that any such explanations can be included as part of
consider previous the final report. There is sometimes a tendency to withhold results from
facts and findings previous studies in order to test the consultant. If a competent investigator is
involved, his or her work will not be biased by any previous efforts made to
explain a failure. When discrepancies occur between parallel efforts, they
ultimately require time to explore and resolve. Almost always, significant
differences between results can be explained and reconciled.
10 Material test Referred to are material tests performed on samples taken from the original
results to input to row of blades that failed. Normally, the analysis relies on data assembled from
life assessment published sources, which is usually satisfactory to proceed with an analysis.
models However, if the metallurgical analysis indicates that the fatigue properties
might have been significantly affected, a report on such a test would provide
three important pieces of information. First, it would advise the analyst of the
possibility that the failure might be related to the material. Second, it should
provide the correct factors and parameters used in the initiation and
propagation models. Third, it should allow the data points to be compared to
the general properties obtained from the literature.
8-5
10305538
Disk/Rotor and Blade Geometry

If possible, the basic shape of the disk/rotor profile should be measured from an available rotor
in accordance with Figure 8-1. If an accessible rotor is not available, this portion can be modeled
by scaling from a reasonably accurate cross-section drawing. In this approach, the disk
attachments are defined as a reflection of the corresponding blade root geometry with applied
tolerances. The outer tip diameter of the assembled row should be measured off the exposed
rotor as a check of the overall length of the row from centerline to tip when the blade is
assembled in the disk attachment. For last stage blades, this dimension can sometimes be found
on lifting diagrams. It can be estimated from a cross-section diagram if there is no other
alternative, but this is not preferable because nominal differences in the estimated diameter of an
assembled row can have an important influence on the calculated natural frequencies.
Figure 8-1
Basic Measurements Taken from the Rotor
8-6
10305538
The most critical item that the plant must be prepared to secure if they intend to involve a third
party is specific details that describe the geometry of the blade itself. The quickest and most
efficient way for this to occur is to have appropriate electronic files transferred directly to the
analyst. Legitimate third parties are normally willing to work under non-disclosure arrangements
that will protect these data as long as the operator is not excluded from any direct contact with
the consulting party to discuss results obtained through the use of these data. This should be
sufficient protection for the OEM, particularly if the third parties do not make or sell turbines or
blades.
Considering that most third-party consultants who specialize in RCA do not design or
manufacture turbines, an OEM’s resistance to release drawings for the purpose of developing an
FEM might be viewed as an attempt to forestall or delay any outside attempts to investigate their
failure. Issues over the lack of control for proprietary information that might be released are
easily handled through confidentiality arrangements. Under such arrangements, the operator or
plant does not even have to be given direct access to the information, nor do any drawings or
dimensions need to be published in order to present and plot results from the finite element
analysis. Because the principal dimensions of any blade can be easily reverse engineered, the
only information contained on an OEM’s computer-aided design (CAD) files that might be
considered commercially sensitive would be tolerances identified at critical interfaces. This
information can easily be stripped out of any files that are sent to support the development of the
FEM.
If such cooperation is not forthcoming, reverse engineering a sample blade is normally sufficient
to allow an independent FEM of the blade to be constructed (see Figure 8-2). Such samples do
not have to be undamaged as long as their original geometry is not too severely compromised;
this way, when the analyst builds the model, the corrections caused by FOD, erosion, or other
superficial damage can be determined from the overall geometry. In some instances, parts not
removed from the rotor can be reverse engineered. However, the extent and accuracy of such
data can be compromised by limited accessibility issues that occur as a consequence of the
location of the row or the type of root attachment involved.
8-7
10305538
Figure 8-2
Basic Measurements Taken from Blade and Stage Information
8-8
10305538
For example, the root attachments of axial entry blades that are assembled through a notch will
not be available for measurement. The basic dimensions of tangential entry blades can be
estimated if both ends are exposed, but these, too, will not be as exact as desired. Limitations
such as these can sometimes be overlooked if the objective is to examine the relative effect of
damage that is confined to the exposed portions of the blade. However, to legitimately examine
all of the potential root causes of a failure, the model must define the blade, disk, and all
connecting devices or linkages as distinctly separate components in order to correctly simulate
their response when they are represented as an assembled row. For shrouded designs,
information or samples are needed that define the width, length, and thickness of the shroud as
well as shape and location of the tenon holes. Where tiewires are involved, the operator should
identify the diameter, lengths, and whether the wire is split.
Given the remarkable ease and sophistication of some laser-based systems that are now
available on the market, the potential investment that a plant might make is trivial in order to
generate and store CAD files of the most critical steam turbine blades at convenient outages.
Confidentiality and non-disclosure arrangements should also allow any relevant information—
metallurgical or analytical—assembled in Phase 1 to be transferred to the third-party analyst.
Although operators sometimes view this as giving away the answer or hold back the results as a
means to judge the expertise of their consulting party, this is inherently the wrong way to start an
independent investigation and is usually counterproductive. Although the series of calculations
described in the steps of this procedure appear routine, much time is spent in planning the model.
As emphasized when selecting a third-party metallurgist, the analyst should be able to
demonstrate his or her qualifications by presenting a resume of past investigations. Because the
investigation of blade failures in not a commonly practiced consulting service, such a resume
should not be difficult to produce. Emphasis should also be placed not on experience in finite
element modeling, but on the capability to diagnosis and correct blade failures.
Finite Element Modeling
The role of the FEM is essentially to provide a detailed profile of the magnitude and distribution
of the stresses throughout a bladed disk under selectively imposed simulated operating
conditions. To this end, a number of critical parameters should be evident within the model in
order for the operator to be able to assume that it is sufficiently accurate to serve this purpose.
The parameters include the following:
• Details of the entire assembly sufficient to produce a model that represents a completely
assembled rotor/disk blade row
• Sufficient detail (element mesh density) in critical regions of stress concentration, with
particular attention to locations where cracks have been identified (see Figure 8-3)
• Gap elements to represent any contacting surfaces between the root and disk attachment,
floating tiewires, and the airfoil or the interlocking surfaces of adjacent integral shrouds
8-9
10305538
Figure 8-3
Typical Features Represented in the Blade Disk FEM
8-10
10305538
As discussed in Campbell’s groundbreaking treatise on the vibration of turbine blades [8], the
action of an assembled row consists of two vibratory motions: (1) the displacements or shapes
assumed by the disk or rotor and (2) the motions of the blades or groups superimposed on these
disk modes. Therefore, to correctly simulate and calculate the all-important natural frequencies
of the system or row, the model must be able to represent both. Frequencies are all-important
because the calculation of dynamic stress is highly dependent on the amount of detuning that is
available to prevent a condition of resonant vibration.
The challenge involved in modeling steam turbine blades is not the physical shape of the blade or
the number of elements used, but the proper selection of boundary conditions and the correct
treatment of any interfacing surfaces. Most steam turbine stages can be explicitly represented as
a segment of the bladed disk, with the segment large enough to represent the largest repeatable
arrangement of blades on the rim. If the blades are freestanding, the segment can be reduced to
one blade. If the blades are organized into distinct groups, the segment should be expanded to
accommodate the largest number. The segment is then represented in the computer as a complete
360º structure by means of cyclic symmetry, which is defined by boundary conditions applied at
the two faces of the disk profile. The segment might need to be further expanded if a floating
tiewire arrangement across several groups is failing. From the operator’s perspective, the most
critical feature in judging a model is that a fully assembled row is represented in the frequency
analysis. Otherwise, the natural frequencies and mode shapes might not be correctly calculated.
This is particularly true where the blades are installed on relatively flexible disks.
Mesh density is a favorite point of argument when analysts approach blades from the different
perspectives of validating a design versus investigating a failure. If competent analysts are
involved, this concern is generally a canard. Analysts who have routinely modeled blades and
used them successfully to investigate blade failures have a general awareness of where the point
of convergence lies around certain features, such as root attachments, tie wires, shrouds, and so
on. Over the course of many years, they should have the further advantage and insight gained
from having their models and techniques regularly checked with the comparison of frequency
data.
As computer processing becomes more powerful, the need to keep the mesh or grid of elements
within a manageable limit has changed dramatically since the 1980s, when element selection and
placement in the model were required skills, not just luxuries. Today, automeshing routines
allow areas and volumes to be defined and filled by the finite element program to general criteria
prescribed by the analyst. Both techniques used to build the model are valid, and neither has a
decided advantage over the other in terms of producing more accurate results as long as the
density is sufficient in the region of critical interest. At some point, the increase in mesh density
reaches convergence—that is, the original prediction of stress is only marginally changed. The
uncontrolled increase in element density within regions of no interest (such as the cover/tenons
where root cracking has occurred) will only add unnecessary computational time required to
complete a single calculation. In a failure investigation, the model is a tool. If it becomes too
unwieldy or difficult to modify when parametric studies are involved to investigate the
sensitivity of critical tolerances, additional, unnecessary definition will become an impediment.
8-11
10305538
A final word on finite element modeling as it relates to failure investigation and analysis: many
designers rely on an analysis to provide a basic snapshot or definitive set of stresses and
frequencies for use with design criteria. In such applications, there is no premium placed on
keeping the model within a size so that it can complete a solution within a run of several hours.
In fact, as the model is expanded to include more stages, one large calculation can suffice in the
place of one model. In support of a root cause investigation, the purpose of the FEM is
completely the opposite: its only purpose is to be a tool that can be efficiently and effectively run
many times within a reasonable period. This feature is essential given the number of different
aspects that often must be checked. If a model is so unwieldy that it takes days to complete one
analysis, it becomes an obstacle rather than an asset in facilitating the identification of the root
cause. As previously mentioned, there is more than enough processing power available in most
personal computers to allow adequate definition of local features found in steam turbine blades
to provide efficiency without sacrificing accuracy.
Calculating Steady-State Stresses
Steady stresses represent the response of the blade-disk system to the centrifugal forces
(computed based on the mass of the blade and speed of rotation) and, to a lesser extent, the
steam-bending forces applied to the airfoil. They drive the damage mechanisms of LCF and
SCC, affect the blade operating frequencies, and provide a mean stress in the estimate of HCF.
Steady stresses should ultimately be adjusted to account for the relaxation that occurs from any
localized yielding. Thermal stresses can generally be ignored in stages that operate at less than
750–800°F (400–425°C). At full load, steady-state stresses will drive the formation of creep
cracks. Specific answers sought from the steady-state analysis include the following:
• What are the magnitude and distribution of steady stresses throughout the blade?
• How high are the stresses at the failure origin?
• Where do the mechanical stresses exceed the yield strength of the material?
• Are the thermal stresses high enough to cause creep damage?
In the FEM, steam-bending forces are approximate as surface loads, using information derived
from the heat balance diagram and measured shape of the airfoil. Centrifugal load represents the
majority of forces that are experienced by a set of blades. They are established as a function of
the blade mass, geometry, density, and rotating speed of the turbine. In high-pressure rows,
thermal stresses are calculated as a function of the temperatures when they reach equilibrium (a
steady-state condition) at baseload operation.
Typically, the highest steady stresses are expected to occur in the root attachments because these
carry the bulk of the blade and normally feature geometric concentrations (see Figure 8-4, Part
A). Other locations of concentration, such as covers, tenons, tiewires, and so on, should exhibit
higher-than-average steady-state stresses similar to those predicted throughout the airfoil (see
Figure 8-4, parts B and C).
8-12
10305538
Figure 8-4
Typical Locations of High Steady Stress
8-13
10305538
As shown, maximum stress values at the locations most susceptible to fatigue damage from start-
stop cycling are highly localized. This explains why the damage models can ultimately be
focused on one or more selected locations (normally, where cracks have formed) rather than
having to consider the entire blade. To obtain an initial sense of the potential sensitivity of such
locations, the initial steady-stress analysis will often first presume that no localized yielding of
the material is allowed. Instead, the stresses are calculated as if the blade were infinitely elastic.
This allows the locations where the elastic stresses exceed the yield strength of the alloy to be
quickly identified as those most susceptible to LCF. If these coincide with the location or
locations where cracks initiated, this indicates that LCF might play a significant role in their
formation. The amount of life consumed each time the blade is loaded and unloaded is
determined later in the analytical process.
Predicted values of stress that do exceed the yield strength must eventually be adjusted to reflect
the reduction that occurs from plastic relaxation of the material. The result is sometimes referred
to as the true stress in order to distinguish it from the elastic values. The consequence of
localized yielding can be handled in two ways. In structures where the relaxation is not
concentrated or localized, an elasto-plastic analysis is performed wherein a series of iterations
produces the true stress. Because steady stress in turbine blades tends to be highly localized,
there is an alternative whereby values of true stress and strain (σ−ε) derived from the elastic
values (σe) produced by the original finite element analysis are solved using the following
equations:
Eq. 8-1a
Eq. 8-1b
Equation 8-1a is known as Neuber’s hyperbola, and the constant term (σe) on the left is the
elastic stress calculated from the finite element analysis. Equation 8-1b is the stress-strain curve
for the blade material. The two terms on the right of the second equation represent the proportion
of elastic strain and plastic strain, respectively. As shown in Figure 8-5, the equations are
programmed to prepare a hysteresis map that graphically represents the loading and unloading
that occurs at the location originally represented by a specific calculated elastic stress.
8-14
10305538
Figure 8-5
Typical Stress-Strain Hysteresis Map Used to Define LCF/HCF Parameters
The example in Figure 8-5 starts with an elastic stress of 205 ksi (1413 MPa) and includes the
consequence of an initial 20% overspeed test, which is typically performed by most manufacturers.
Plastic deformation during this overspeed results in an initial true stress of 100 ksi (690 MPa).
After the turbine stops, a residual stress of 66.9 ksi (461 MPa) is retained at the location. When the
unit returns to full speed, it produces a maximum stress of 81.3 ksi (560 MPa).
8-15
10305538
The stress range and mean stress are both used as input by which LCF and HCF initiation life are
estimated. The equivalent (true) stress previously corresponding to the elastic stresses should
then be summarized for one or more locations throughout the blade, with particular attention to
regions of stress concentration. The stress range and mean stress are generally defined for any
regions of significant localized yielding indicated by the finite element analysis.
Natural Frequencies and Mode Shapes
Because HCF is expected to play some role in any blade failure, it is absolutely necessary to
attempt a reasonable approximation of the magnitude of dynamic stresses that occur under
normal or off-design circumstances. Answers to the following specific questions are sought from
the frequency analysis:
• What are the frequencies of the fundamental mode families at zero rpm?
• How do stress stiffening and spin softening affect these frequencies at full speed?
• What mode shapes does the blade-disk system produce at full speed?
• What modes, if any, operate in a potentially resonant condition?
The magnitude of dynamic stress in steam turbine blades is primarily a consequence of three
parameters: (1) frequency ratio, (2) damping, and (3) forcing (or stimulus). This task in the
engineering analysis focuses on the most influential of these three parameters, which is the
frequency ratio—the ratio produced by the at-speed natural frequencies of the blade disk divided
by the frequency nearest the engine order (per-rev harmonic of forcing). The natural frequencies
are further sorted and classified by plotting the shapes or deflections that they assume when they
vibrate as nodal diameter modes. Nodal diameter refers to the pattern of displacement of the
rotor/disk, and mode refers to the pattern of displacement that the blades assume around the
circumference of the disk, as shown in Figure 8-6.
8-16
10305538
Figure 8-6
Mode Shapes of Bladed Disks
Because the role of operating natural frequencies is so critical to the success of any blade RCA,
further discussion is given regarding critical features required in the dynamic model, common
formats used to present these results, and special considerations associated with certain blade
designs that can make the frequency calculation a challenge.
8-17
10305538
Dynamic Model
To calculate the natural frequencies, a dynamic model of the 360º rotating blade row is typically
extrapolated from the segment model previously used to obtain steady stresses. Reducing the
segment to master degrees of freedom and then forming these into a dynamic model of the entire
row does this. The choice of these master degrees of freedom is where experience is crucial in
terms of producing accurate results. As mentioned, dynamically, the blade-disk system consists
of two types of vibratory motions in which the displacement of the blades is superimposed on
the flexure of the disk. Modes involved with continuous structures like a disk are referenced by
the corresponding number of diameters that can be drawn through the axis that correspond to the
number of flexures that occur around the disk rim. Blades or blade groups, which act like
cantilever beams fixed to the rim of the disk, assume vibratory motions in which direction of
displacement tends to predominate. Often, the first four fundamental blade modes are referenced
by the predominant direction of their motion relative to the rotor/disk as the tangential, axial,
torsional, and second bending modes, as shown in Figure 8-6.
The dynamics of a blade-disk structure inherently produce many vibratory modes. First, for any
given blade-disk arrangement, the maximum number of nodal diameter modes that the disk will
produce is N/2, where N represents the number of groups or freestanding blades in the assembled
row. For example, if 196 blades in a row are organized into 28 groups (28 x 7), the maximum
rotor/disk nodal diameter associated with this arrangement would be 28/2, or 14. Each of the
resulting 14 nodal diameters would have a corresponding tangential, axial, torsional, and second
bending blade mode occurring at different frequencies.
In other words, a frequency table of the 28 group arrangement would have four fundamental
mode families of 14 nodal diameter modes each, or a total of 56 modes to consider in deciding
whether the potential for resonance might exist at full-speed operation. For drum type or
monoblock rotors where the flexibility of a disk is negligible, it is still recommended that the
individual families of blade modes be tabulated against the respective number of nodal diameters
to maintain this conventional nomenclature to identify modes of specific interest.
Presentation of Frequency Results
Although the complexity of the dynamics for a blade-disk system sounds intimidating when
described, in fact, the calculated results can be simplified into a readily understandable set of
corresponding plots that quickly organize the data and highlight the modes to be of further
concern when calculating dynamic stress.
The conventional Campbell diagram shows the families of modes as a sequence of lines that
extend from zero rpm to full speed (see Figure 8-7). They reflect the rise in frequencies with
speed as a consequence of the aforementioned stress stiffening and spin softening. The engine
orders or per-rev forces are plotted as a series of lines emanating from the origin. Normally, a
line parallel to the Y-axis indicates the baseload speed of the turbine at 15, 30, 50, or 60 hertz.
The points at which the modes cross this line indicate their respective position from the nearest
per-rev. Their proximity is a reflection of the potential for a resonant condition to occur.
8-18
10305538
Figure 8-7
Different Frequency Displays
8-19
10305538
A corresponding plot of the same information viewed from a different perspective is the
interference diagram (see Figure 8-7). The essential difference between it and the Campbell
diagram is that it provides a broader view of the modes by focusing exclusively on their
frequencies at full speed, rather than over the range of zero to full speed. On the diagram, the X-
axis is the sequence of nodal diameters, whereas the Y-axis remains as hertz. The individual
families are plotted against each corresponding nodal diameter number. In other words, all 56
frequencies of the aforementioned 28-group arrangement are shown. Forcing is reduced to a
single line (because the speed is now constant instead of a range). The single impulse line is
formed by the corresponding frequencies at 1/rev, 2/rev, 3/rev, and so on. For a 60-hertz turbine,
the impulse line would intersect the 1ND axis at 60 hertz, the 2ND axis at 120 hertz, and so
forth. In this plot, a condition of potential resonance is indicated only where the impulse line
consisting of per-rev forces coincided with any nodal diameter mode within any of the families.
Resonance is signified wherever the lines representing the natural frequencies of each blade
mode family coincide with a per-rev harmonic on the Campbell diagram, or the impulse line on
the interference diagram. The number of hertz anticipated to exist at full speed indicates
detuning. As a rule, a 10-hertz margin is generally considered to signify a well-detuned mode of
vibration, where the statistical chance of resonance is likely to be remote. For long blades where
the first four mode families are generally within a range of 500 hertz, this translates into a
frequency ratio of ±3–5%. As shown in the two plots, despite the large number of vibratory
modes, only a few are of any potential relevance or importance in the failure investigation, that
is, normally only one mode of each family. Where a condition of resonance is signified, this does
not automatically mean that the dynamic stresses will be high enough to initiate a crack. This
determination is made in a subsequent step in which the stresses are calculated for modes of
interest identified in the Campbell and interference diagrams.
Mode Shapes
A second series of plots that should be produced as output from the frequency analysis are the
shapes or displacements that are produced when the particular mode is excited. Each mode—
tangential, axial, torsional, and second bending—will assume a unique pattern. The location
where each mode is likely to produce a maximum dynamic stress on the blade is a further clue as
to which mode was responsible for the formation or propagation of cracks in specific locations of
the design. Where two modes might produce a maximum stress at the same critical location, the
role of the two would be differentiated based on their predicted margin of detuning. Generally,
this approach is how a case where synchronous vibration as the root cause of HCF initiation or
propagation starts to form.
Conversely, if the results of the frequency analysis indicate that the fundamental modes of
vibration all appear to be well detuned, this would cause the investigator to question whether
vibration was a secondary contributor, depending on LCF, erosion, corrosion, or impact damage
to precipitate the chain of events leading to failure. If there was no evidence of any of these in
the metallurgical analysis or the LCF life analysis performed in a later task, this would next point
to evidence of HCF as coming from an asynchronous problem such as flutter, to be supported by
further evidence that such an off-design operating event might actually have occurred.
8-20
10305538
Special Circumstances and Considerations for Marginally Detuned Blades
In situations where stability of the grid might be a problem, an additional frequency calculation
might be needed to check the potential consequence of a maximum drift above or below the
rotating speed (see Figure 8-8). In such a scenario, the nominal drift in rotating speed will have
less effect on the original natural frequencies. However, the per-rev harmonics will remain at
multiples of the current rpm. This can create a situation where they migrate enough to coincide
with a marginally detuned mode and create an intermittent situation of resonant vibration.
8-21
10305538
Figure 8-8
Potential Consequence of Excessive Grid Fluctuations
8-22
10305538
When calculating the frequencies of axial entry blades, the amount of contact assumed between
the adjacent surfaces of the platform and root can become important (see Figure 8-9). If the
surfaces are slightly tapered to account for the angle or orientation created when they are
arranged in a circle and they are tightly packed when the last blades are installed at the notch, the
additional stiffness and rigidity can significantly change the predicted modes of vibration when
compared to the results of a model where a loose fit is assumed, or a line fit is assumed (to
represent a non-tapered set of blades). If the crevices formed by a dovetail installation are
exposed to the steam flow, the buildup of deposits can increase the stiffness of a row that might
not originally have been tightly packed, causing the frequencies to migrate over time. This
migration might be enough to shift a marginally detuned mode into resonance. When the analyst
investigates these types of blade designs, the potential consequence of these effects might require
him or her to simulate the extremes of both states to assess if the range could make a well-
detuned mode resonant, thus pointing to installation as playing a role in the root cause of a
fatigue failure.
8-23
10305538
Figure 8-9
Axial Entry Root Contact Effect on Frequencies
8-24
10305538
Similarly, for blades that rely on untwist or a pre-load to establish and maintain the operating
row as a continuously linked ring, the capability to simulate the results of an extreme opposite
situation might be required to see what might happen to the margin of detuning for any modes if
the gaps failed to close or the pre-load was relaxed and a single blade or row was allowed to
become freestanding. In cases where the integrity of the continuously formed link might have
been compromised, it is usually assumed that the result is so unpredictable between all blades
being linked and all blades being in a freestanding state that the risk of a resonant condition is
unacceptably high. Such a finding, however, would focus future efforts on strengthening the
original linkage rather than detuning a particular mode of vibration.
Summary Comments
Accurately calculating the natural frequencies of a rotating blade row is normally the most
challenging task in any RCA, and if not accomplished successfully, it can essentially
compromise the entire investigation. The challenge is in part due to the complex behavior of
bladed disk structures, which produce families of modes for any given row of blades. In addition,
these frequencies are subjected to two simultaneous effects known as spin softening and stress
stiffening. The effect is non-linear and can therefore only be predicted either by analysis or by
direct measurement to ascertain what the frequencies will be at full-speed operation. Knowing
these frequencies is critical to any root cause investigation. This is particularly true when dealing
with longer turbine blades, where the margin of detuning can profoundly influence the
magnitude of dynamic stresses and HCF produced as a consequence.
Calculating Dynamic Stress
Dynamic stresses are calculated as a basis for estimating the amount of HCF damage that might
be caused in locations where cracks formed. Dynamic stress is approached and viewed from
three perspectives, represented by the following questions that this task should address:
• How high can the stress become if it is assumed that a mode is resonant?
• Where does the maximum stress occur for each mode relative to the crack origin?
• How much would the stress be reduced from detuning the mode?
8-25
10305538
Generally, a profile of vibratory stresses is obtained for each of the first four or five families of
blade modes. The modes of particular interest are those identified in the frequency analysis as
being potentially resonant and/or having a maximum stress coinciding with the location where
cracks formed. Dynamic stresses are first calculated with the FEM for an assumed condition of
resonance. A straightforward formula is then used to reflect the benefits of damping and
detuning to lower the stress. As shown in Equation 8-2 and Equation 8-3, detuned dynamic
stress (σd) is obtained by applying an amplification factor (λ) for a calculated resonant stress
(σR) as
Eq. 8-2
Eq. 8-3
The amplification factor is based on a frequency ratio of (η), under damping (ξ) assumed in the
resonant stress analysis. (As noted earlier in the report, damping [ξ] for these LP turbine blades
is generally assigned at 2–5%.) The frequency ratio (η) is the relative proximity of the
calculated natural frequency to the nearest forcing frequency.
In the example shown in Figure 8-10, the resonant and detuned dynamic stresses for two
different modes of vibration are calculated to show how this information can further assist in
identifying which of the two is more likely to be responsible for observed HCF damage. In an
assumed resonant condition, the frequency ratio (η) equals one, whereas the detuned frequency
ratio is based on the calculated at-speed frequency of the mode. Damping (ξ) is assigned at 0.3
critical.
8-26
10305538
Figure 8-10
The Effect of Detuning on Dynamic Stress
In the original finite element calculation of resonant dynamic stress, the magnitude of stimulus
applied to represent the potential force of the nearest per-rev is generally a ratio of the steam-
bending force. Based on published tests, NPF is normally assumed to have a stimulus ratio of
10%. The first 10 per-rev harmonics can reasonably be assumed at a stimulus ratio of 2–5%. It
should be noted that the magnitude of stress is directly proportional to the assumed stimulus
ratio, but that the overall distribution of dynamic stress produced by the response of any given
mode does not change, regardless of whether 2%, 3%, 4%, and so on is applied.
8-27
10305538
After exercising these formulas, it soon becomes clear that of the three aforementioned
parameters, the margin represented by the frequency ratio most profoundly affects the potential
magnitude of dynamic stress that the location of interest might experience. In an assumed
condition of resonance, calculated dynamic stresses on the order of 10 ksi (69 MPa) and higher
can be produced. With even a nominal amount of detuning, dynamic stresses are reduced to 1 ksi
(7 MPa) or less. This is why so much emphasis is placed on the frequency analysis and its
authentication with test data, preferably provided by the OEM.
Further Comments on Stimulus and Damping
Because HCF plays such a critical role in any steam turbine blade root cause investigation,
technical arguments often fixate on its calculation. Unlike the frequency calculation, which can
be verified with test data, the magnitude of dynamic stress is subject to assigned values of
stimulus and damping. As such, there is always some room to dispute the choices made. It is
therefore worth knowing what these values are meant to represent in the calculation and what
are considered reasonable limits (maximum or minimum), based on tests performed to research
these parameters. As will be further shown, the subjectivity of these values is sometimes
negated by the predominant role that frequency ratio has on the calculation of dynamic stress.
In any turbine stage, the non-uniform flow field around the circumference of the stage is
normally the predominant source of dynamic excitation that promotes blade vibration during
operation of the turbine. A lifting force on the airfoil tends to excite the flap modes, whereas the
drag force excites the edge modes. Because of the geometric variation inherent to any specific
turbine stage, the absolute magnitude of dynamic forcing that results from the physical features
and conditions for a specific row of blades can be exactly determined only by in situ testing. In
lieu of such rare and costly measurements, the magnitude of a given per-rev force is typically
approximated as a stimulus ratio, or as the ratio of the dynamic forcing amplitude over the
steady bending force. Dynamic excitation at NPF is estimated to be about 10% of the calculated
steady forcing. The stimulus at lower per-rev frequencies (below NPF) is generally considered
as lower because these excitations are the result of pressure non-uniformities around the
circumference, both upstream and downstream of the rotating blades. Therefore, as input to the
resonant response calculation, stimulus is represented as 1–5% of the steam-bending force,
assigned to each per-rev harmonic up to the tenth engine order. For higher order per-revs, a 1%
stimulus is generally assumed.
With regard to damping, the primary source for blades—aerodynamic damping—is a strong
function of the inter-blade phase angle, inlet Mach number, and the reduced frequency. A
damping ratio of percentage over the critical damping is applied to calculate the dynamic
response under normal operating conditions. The relative effect of different damping ratios on
the amplification of dynamic stress at resonance is illustrated in Figure 8-11. The aerodynamic
damping applied to calculate dynamic stress for LP turbine blades is typically in a range of 0.2–
0.5%. It is worth noting that in a suspected case of resonant vibration, the value selected within
this range to represent this input parameter becomes less relevant. As shown in Figure 8-11,
damping measurements made at corresponding levels of dynamic stress illustrate how damping
increases commensurate to the magnitude of dynamic stress. Although additional damping can
help to mitigate the resulting dynamic stresses, if the response of the mode is sympathetic
8-28
10305538
enough and the corresponding stress is high, it can essentially overwhelm the available damping
and result in a fatigue failure. Aerodynamic damping can become negative under a stalled
condition of flow. If negative aerodynamic damping overwhelms the positive mechanical
damping, the design becomes unstable. This is in reference to conditions of asynchronous
vibration that are not normally addressed as part of the standard investigation associated with
characterizing dynamic stresses.
Figure 8-11
Damping Ratios Versus Blade Response
In plots of dynamic stress, the general distribution produced for a blade’s mode, like those
shown in Figure 8-10, will not change as a product of the stimulus. The effect of stimulus ratio
is linear. Thus, if a dynamic stress of 1 ksi (7 MPa) is calculated using a conservative 1%
stimulus ratio, the stress would be increased to 5 ksi (34 MPa) using the most liberal ratio of
5%. For longer, LP blades, it is not unusual to calculate the resonant stresses of higher modes to
be relatively low, just as it is also not unusual for the first tangential mode to produce dynamic
stresses on the order of 20 ksi (138 MPa) or higher in an assumed condition of resonance. Both
results indicate why designers assume that they must detune the first fundamental mode, and
sometimes choose to ignore higher order modes (beyond the third mode) or those whose
frequencies reside above the 10x forcing. This can also explain why blades fail after they are
damaged by erosion, corrosion, or other forms of wear. The nominal stresses produced by
resonance of these higher modes are insufficient to initiate cracks, but they are high enough to
propagate them when they form as a consequence of such damage.
8-29
10305538
For short turbine blades, the resonant stresses are generally used to represent the dynamic
condition they might reasonably encounter. Because of their size, they vibrate at much higher
frequencies than their longer later-stage LP counterparts (at thousands versus hundreds of hertz).
As such, it is impossible to design them so that their frequencies can be expected to avoid the
nearest per-rev forces. Instead, they are designed to avoid NPF and accommodate the worst-case
scenario of dynamic stress that might be encountered. As noted, the assumed condition of
resonance is somewhat offset by the lower stimulus ratio that is assumed for these high per-revs.
Model Verification by Frequency and Vibration Testing
Given the aforementioned importance and role of the frequency analysis, one of the most
accepted ways to confirm the authenticity of a blade-disk FEM is to use frequency test data as a
basis of validation and refinements if the comparison of test frequencies and predicted
frequencies is significantly different. Frequency testing seeks the answers to the following
questions:
• What is the range of scatter between individual blades before they are assembled as a row?
• What are the frequencies of the nodal diameter modes at zero rpm?
• How does stress stiffening and spin softening affect these frequencies at full speed?
In the world of mathematical certainty, the frequencies produced with the FEM are distinct for
each mode of vibration. This is because the model can represent each blade as being exactly
identical in both geometry and material composition and at all points of contact representative of
the final condition after assembly. In the real world, a small variation inherently affects each of
these three parameters, causing the frequencies when tested to fall within a range about a mean
value (see Figure 8-12). The general uncertainty that can therefore be associated with any
calculated natural frequency is typically estimated at 2–3%. In other words, a correlation to
within this degree of accuracy would be considered evidence of a reasonably well-modeled
bladed disk.
8-30
10305538
Figure 8-12
Example of Frequency Results from Impact Tests of Single Blades
8-31
10305538
Frequency testing is often considered to be a definitive check because of the predominant role
that the frequency ratio has in terms of affecting the synchronous dynamic stresses that occur
under normal operating conditions. Whether modes are marginally detuned or not can become
the final point of contention between the OEM and independent failure analysts. Frequency data
can resolve such arguments.
Frequency data come in several forms. Their ultimate value as a means to verify the FEM is
consistent with the level of difficulty and expense that is taken to bring the data as close to the
operating state of the blade row as possible. Frequency tests and the data they produce can be
organized into the following five levels of increasing difficulty, accuracy, cost, and expense:
• Impact testing of individual blades. Sometimes referred to as rap testing, this is often
performed as a QA/QC operation whereby the blades are mounted in a fixture or suspended
to represent a free-free state. An accelerometer is mounted on the blade and then struck
repeatedly with a specialized hammer that has an accelerometer mounted on it. The resulting
signal is processed to show the frequencies of vibration for several modes. These are
tabulated to form a distribution about a mean value for each mode. An allowable range is set
to identify and catch blades whose dynamic response indicates that it might not behave as
predicted. A version of this is to test the frequencies of blades installed on the disk, either as
groups or individual structures.
• Shaker testing of assembled rows. Representing the next level of sophistication and
required expertise, sweep testing involves the use of a portable shaker attached to the
disk/rotor section of the row of interest, with an accelerometer mounted around the
circumference and dual channel analyzer. The shaker produces a controlled frequency of
excitation that is swept through a frequency range where the fundamental nodal diameter
modes are predicted to reside. If the shaker can produce sufficient excitation, the analyzer
will be able to process the data as the individual nodal diameter modes within each family
respond. In turn, these are directly correlated to the interference diagram of zero rpm
frequencies.
• Spin pit testing of rotating blades. The OEM usually performs these tests where a row of
blades is instrumented with numbers of strain gauges that are positioned and oriented to
capture the response of selected fundamental modes, that is, within the first two or three
families. An artificial source of excitation is provided by means of a magnet or air jet in
proximity to the moving row. Signals are brought through a slip ring assembly. The
centrifugal forces produced in a spin pit test cause the interlocking features of the system to
react as they would in the actual steam turbine. Thus, snubbers, floating tiewires, and so on
assume their operating position. Strain gauges allow a direct measurement of frequencies to
be made. Corrected for temperatures, the results can be compared directly with the at-speed
predicted frequencies of the FEM.
• BVM of rotating rows. BVM systems are a compromise between the shaker tests and spin
pit/strain gauge tests performed by OEMs. They include rotational effects but simplify the
initial test setup and eliminate the significant challenges of mounting gauges on blades
securely enough to withstand the centrifugal forces at full rotating speed. This is
accomplished by relying on a series of non-contacting sensors arrayed around the
circumference of the stage. The sensors are used to detect differences in the time of arrival
8-32
10305538
for each blade and process these as tip displacements (peak-to-peak). Processing algorithms
are then applied to identify the modes of vibration associated with these displacements.
Although the sensors allow every blade to be monitored, the measurement technique is
indirect.
• In situ testing of rotating blades. In situ testing is the Cadillac of turbine vibration
measurements. Similar to spin pit testing, a selected number of blades in the row are
instrumented with strain gauges to allow direct measurements to be taken under the actual
stage conditions produced at operation of turbine. Signals are extracted from the stage by
means of telemetry for external recording and processing. During startup, excitation is
artificially introduced by means of simple steam jet to provide a waterfall plot commensurate
with the Campbell diagram. The results can be directly compared with the at-speed
frequencies of the FEM.
How Different Test Results Are Applied
Impact data are useful and easy to obtain, but it must be recognized that as described, they do not
represent the row in a fully assembled condition, or with any of the at-speed operating conditions
of the stage—such as temperature, stiffening, and so on—imposed on the blades. To correlate
impact test data with the FEM, a separate model of a single blade must be extracted from the
blade-disk model used in the investigation to recalculate the frequencies either in a free-free state
of suspension or with the simulated mass of the test fixture added. It should be further noted that
if a fixture is used for blades that have relatively small root attachments, the hydraulic system
normally applied to secure each part in the fixture should be adjusted accordingly so as not to
overload the small root attachments and inadvertently damage them. The amount of dispersion
apparent between the individual blades can also be used to verify this assumption in a
probabilistic model. If blades are tested in this manner when installed in a row, information to
distinguish the individual nodal diameter frequencies confined within a band will not be
available unless a shaker-sweep test is included. In any test of an assembled row at zero rpm, the
blades must be firmly fixed in their root attachments by artificial means (normally, with the
careful application of Locktite®). Otherwise, the non-linearity from an unspecified amount of
looseness will produce data that are useless for correlation with predicted results.
If performed correctly, shaker testing is much more informative than impact tests of individual
blades because it is intended to also excite and identify the frequencies of the disk corresponding
to the frequencies of each blade mode (see Figure 8-13). This can be more directly correlated to
the zero rpm results. Adjustments or refinements are then made to the FEM, and the at-speed
frequencies are recalculated with the expectation that the relative changes will be reflected in the
full-speed positions of the critical frequencies. Although shaker tests are generally affordable,
they do require undisturbed access to the rotor, preferably out of the bearings on jack stands, and
are more likely to work on rows with flexible disks, not drum type or monoblock rotors, which
are too massive for the shaker to excite sufficiently and bring out the disk modes.
8-33
10305538
Figure 8-13
Examples of Impact and Shaker Test Frequency Data
8-34
10305538
BVM systems have been under development since the 1980s to provide a more cost-effective
alternative for monitoring rows of blades under field operating conditions. The principal
advantage of a BVM is that it limits the time spent and enormous challenges of securing strain
gauges to a limited number of blades and finding a means to bring the lead wires from the
rotating stage to a transmission point. A BVM can also sequentially examine each of the
individual blades in a row if they are freestanding. The principal measurement obtained from a
BVM is the amount of blade tip displacement, based on processing the time of arrival data
obtained as each blade passes in sequence an array of non-contacting proximity sensors (see
Figure 8-14). Generally, a minimum of three sensors is required to transform time-of-arrival
measurements into meaningful values of tip displacement, although more sensors might be
required if a particular mode is the object of the test. The principal limitation of the BVM in
regard to frequency testing and model validation is that it inherently relies on data processing
algorithms to identify modes associated with tip displacement patterns. If the vibration is not
particularly severe, the modes contributing to the overall magnitude of registered displacement
can not be identified with confidence.
8-35
10305538
Figure 8-14
Examples from a BVM Test
8-36
10305538
Because the time-of-arrival system is tied to the rotating speed of the turbine, there is also a
limitation imposed that makes it least sensitive as the frequencies come closer to coinciding
exactly with the per-rev harmonics, that is, a condition of resonance. As such, BVM is more
adapted to detecting asynchronous conditions, such as stalled or unstalled flutter, that are
suspected under certain conditions of off-design operation (see Figure 8-14). In tandem with the
finite element analysis, the tip displacements obtained with the BVM can be applied to develop
transform functions to calibrate the dynamic stresses at locations throughout the structure. In
turn, these stresses can be input to the HCF life model to estimate the relative amount of damage
caused by the event.
The two principal advantages of spin pit testing are that it includes the effect of centrifugal forces
in the same manner as expected in the operating turbine and that it provides the capability to
measure the frequencies of vibration directly, through strain gauges mounted on the blades. In a
failure investigation, disputes over the authenticity or validity of FEMs can be resolved by
requiring the OEM to produce these results in the form of a Campbell diagram, which can then
be compared with the Campbell diagram from the frequency analysis derived from Task 10. The
OEM should have these results at their disposal to allow such a comparison.
We mention two final words of caution when this is attempted. First, blade designers generally
focus on detuning frequencies that fall within the first 10–12 engine orders, with particular
attention to positioning (detuning) the first mode family. The orientation and position of gauges
can affect the magnitude of response that is detected for different modes. In some investigations
where higher order modes are suspected to be the root cause of HCF damage, their response
might not be registered in the test data. These modes are sometimes overlooked because of their
higher frequencies, but they have been known to cause fatigue problems. The fact that they are
not apparent in the OEM’s frequency data does not mean that they do not exist or are not
responsive. The relative correlation with lower, more prominent modes observed in the test
would be expected to carry through in verifying the calculation of higher mode frequencies.
The second word of caution involves establishing the specific design that was tested. Spin pit or
wheel box tests are expensive enterprises. As such, it should be made clear if the results are from
the specific, exact design that was modeled, or from a generic design used to qualify a family of
scaled blades.
Strain gauge telemetry tests are probably the most definitive that can be performed to clearly
establish the dynamic behavior of the blades under actual operating conditions. Unfortunately,
such tests are extremely expensive and usually must be completed within a limited window of
time, dictated by the life span of the transmitter batteries and survival of the gauges within the
highly aggressive environment. Setting up such tests is something of a lost art because they
require the gauges and leads to be secured and shielded, then run to a point where transmitters
can be located and an antenna installed for data retrieval and signal processing. Several tests
performed by EPRI in the 1980s and 1990s were able to overcome these obstacles, and the
results were used to verify frequency predictions made by the BLADE-ST software to assist its
development [9, 10].
8-37
10305538
Life Prediction and Probability of Failure
This is where all of the relevant facts associated with operating history, material properties, and
design stresses are brought together and used as a basis to separate root causes from secondary
contributors. As indicated by the sequence of the remaining sections of Section 8, the path
within this final step should always start with an assessment of crack initiation, first by LCF and
then by HCF. These results are fundamental to deciding whether features of the blade design
itself are the likely root cause or whether they are secondary contributors. Specific questions that
this task should address include the following:
• How many start-stop cycles before a crack might initiate because of LCF?
• How many hours of operation before a crack might initiate because of HCF?
• In a wet steam environment, how many hours before a crack might form because of SCC?
• What size of a defect can the design potentially tolerate at the crack origin?
• What are the chances for a failure to occur as starts or hours accumulate?
• What are the chances for an initial crack to propagate from HCF as it increases in size?
With the exception of an HTC analysis, two approaches are typically used to completing this:
(1) deterministic life prediction models that rely on discrete values selected by the investigator
and (2) a probabilistic treatment whereby the same input is objectively selected from a statistical
database and randomly applied to estimate the odds of a failure. The remaining sections of
Section 8 discuss each of theses techniques in its normal sequence. Even in situations where the
prior evidence indicates the strong possibility that erosion, corrosion, or impact damage played a
critical role, the assessment of LCF and HCF should always be performed. The results justify
the elimination of over-cycling or excessive blade vibration as principal root causes, reinforcing
the decision to transition from initiation to damage tolerance.
Crack Initiation: LCF
To estimate LCF crack initiation life, the amount of plastic strain (εp) must first be estimated
based on Manson’s life prediction model, based on the stress range (Δσ/2) and plastic strain that
occurs each time the material is cycled. The stress range (mean + residual) is shown on the
hysteresis map (see Figure 8-5) previously developed for locations of interest from the steady
stress analysis. Using Manson’s model, plastic strain (εp) is calculated as
Eq. 8-4
8-38
10305538
where n is the strain hardening exponent and K is the hardening coefficient, both fatigue
properties obtained from testing specimens of the material. With the plastic strain in hand, the
LCF life (Nf) for any location is calculated by rearranging the terms of the following equation:
Eq. 8-5
Equation 8-5 is referred to as an LCF model, where fatigue properties εf, σf, b, and c are obtained
from tests performed on specimens of the material. As illustrated by the example shown in
Figure 8-15, crack initiation due to LCF occurs when the damage (1/Nf) reaches 100%. For most
steam turbine blades, stresses produced in the root attachments of the largest LP blades should
still allow for tens of thousands estimated cycles or well above the several hundred cycles a large
unit will normally experience in a lifetime of baseload operation. LCF can be made a problem if
the unit is over cycled. Some large steam turbines meant to operate with blades designed for
baseload service that is subsequently used for peaking service may unexpectedly develop cracks.
8-39
10305538
Figure 8-15
LCF Initiation Estimation
8-40
10305538
Crack Initiation: HCF
To estimate HCF damage as a source of crack initiation, fatigue properties σf and b from the
strain life curve are used. Based on HCF life, Nf (the number of cycles) can be estimated from
the following equation:
Eq. 8-6
The equation is referred to as an HCF model, where the mean stress (σmean) is the same as that
applied in the previous calculation of plastic strain (see Equation 8-4) and the dynamic stress
(σd) is either the resonant or detuned stress, whichever is considered appropriate based on the
proximity of the natural frequency to the nearest per-rev. The model is applied to any locations
where the maximum dynamic stress produced by one of the first four fundamental modes
coincides with where cracks formed. As illustrated by the example shown in Figure 8-16, crack
initiation occurs when the damage (1/Nf) reaches 100%.
8-41
10305538
Figure 8-16
HCF Initiation Estimation
8-42
10305538
Both resonant and detuned dynamic stresses obtained from the previous step can be substituted
to show the consequence of detuning and resonance in terms of predicted fatigue life. Generally,
detuned stresses produce estimates on the order of 1010 cycles or greater. This can be compared
to the pre-failure operating time by multiplying the respective frequency of vibration x 60
seconds x 60 minutes x 24 hours x 365 days x 20 years to get a sense of whether the predicted
HCF life was even approached. In general, stresses on the order of 1 ksi (7 MPa) or less
(detuned) will produce predictions that are essentially interpreted as infinite (in terms of high
cycle initiation life, not propagation life). Stresses on the order of 3–5 ksi (21–34 MPa) are also
usually not sufficient to produce enough damage to initiate a crack. Stresses exceeding 7–8 ksi
(48–55 MPa) or higher (which typically occur only in a resonant or near-resonant condition) can
significantly reduce the HCF life.
Given the uncertainties involved with dynamic stress, results from the HCF are generally judged
by orders of magnitude, rather than as absolute cycles or hours of continuous operation. If an
infinite number of cycles is not estimated, this is usually interpreted as signifying a potential
HCF problem, pointing to the respective mode of vibration and limited detuning as the root cause
or source of the failure.
Crack Initiation: HTC
High-temperature mechanisms, such as creep, can typically be discounted for stages that operate
at 900ºF (482ºC) or less. The root cause of creep is temperature and steady stress over time.
Creep damage in HP or reheat steam turbine blades typically manifests itself as the elongation or
distortion of the material. To describe the behavior of material at elevated temperatures, the
process of creep is generally separated into three categories, as shown in Figure 8-17. In Figure
8-17, it should be noted that the times associated with both the primary and tertiary stages are
expanded for the sake of clarity.
8-43
10305538
Figure 8-17
HTC Damage Process
8-44
10305538
As a rule, the tertiary phase (where plastic instability of the material like necking occurs) is a
small proportion of the total time to rupture. Operating in this region for any extended period is
generally not considered feasible. Instead of directly calculating the creep rupture life shown in
Figure 8-17, a design criterion is often selected that is meant to approximate when the tertiary
phase occurs. This is then compared to the creep strain over time predicted with an FEM. The
strain-to-failure approach relies on a creep material curve generated from uniaxial creep material
specimen tests conducted under controlled conditions. An iterative process using a finite element
thermal model obtains creep strain throughout the blade, especially in regions of high stress. It is
assumed that a significant risk of failure is posed when the amount of creep is likely to produce
crack initiation. Creep is therefore measured in terms of a cumulative percentage of creep strain.
Crack Propagation: SCC, CAF, Erosion, and Fretting
In the evaluation of SCC, erosion, or CAF, the metallurgical evidence establishes the presence of
these initial damage mechanisms. This finding of why the blade failed is supported by the results
of applications for the two previous life models, which indicates that it was not the result of LCF,
HCF, or creep. At some point, regardless of what initiates a crack, should it continue to enlarge,
the vibratory stress will eventually exceed the stress intensity threshold of the material along
portions of the crack front. Once HCF takes full control, the process of further propagation to
final rupture will accelerate (because of the high frequencies at which blades vibrate). FM is
therefore used to complete the story by explaining how the blade failed and when this was likely
to have occurred relative to the damage sustained by the design at the location of interest.
Crack propagation from an existing flaw under an alternating load (in this case, blade vibration
from a selected mode) is governed by the stress intensity factor (ΔK). ΔK is a function of the
crack geometry and the applied stress. For a plate with a semi-elliptical surface crack (a) under a
uniform stress (Δσ), ΔK can be estimated with a simple model and edge correction factor as
Eq. 8-7
When the stress intensity along the crack front exceeds the material threshold value of the
material, HCF will take control of propagation. By rearranging the previous equation, the
threshold crack size can be estimated as
Eq. 8-8
8-45
10305538
When the results of the equation are estimated and plotted for increments of dynamic stress, the
graph (see Figure 8-18) illustrates that when the dynamic stress exceeds 8–10 ksi (55–69 MPa)
(a resonant condition), this magnitude is sufficient to propagate even a small indication at the site
of interest, that is, from indications on the order of less than 10 mils (0.3 mm). In a near-resonant
condition (3-5 ksi [21–34 MPa]), indications of 0.02 in. (0.5 mm) or greater could form but still
be detected before HCF took control. If the dynamic stress is less than 1 ksi (7 MPa) (a well-
detuned mode), the structure is not expected to fail as a consequence of HCF, even in the
presence of relatively large cracks (those larger than 0.25 in. [6.4 mm]).
Figure 8-18
FM Models
In this initial example, a simple correction factor of 1.12 is assumed. The correction factor and
equation should be adjusted to represent cracks formed at different locations within a blade, as
shown in Figure 8-19.
8-46
10305538
Figure 8-19
FM Models
8-47
10305538
In cases where it is decided that the dynamic stress is likely to have played a secondary role in
the promotion of cracks, a more precise calculation of the critical crack size is warranted.
Recognizing that in blades, the local geometry and the stress field are more complicated than
assumed in the previous FM formulations, crack tip stress intensity, K, is calculated specifically
along the crack front by modifying the original FEM. The stress singularity is enforced along the
crack front by placing linear stress wedge elements in the FEM, with quarter points orientated
toward the crack tip. Figure 8-19 shows this technique.
Probability of Failure
To further assess the potential for a problem to develop on the specific blade design, a
probabilistic analysis is performed (see Figure 8-20). This is an extension of the snapshot
provided by the application of the aforementioned LCF, HCF, and FM models. Instead of
discrete values of stress or material fatigue properties being subjectively assigned in the
respective equations, they are instead represented as statistical distributions. A Monte Carlo
simulation is set up to randomly combine these parameters thousands of times. The results are
tracked over cumulative numbers of starts or hours to reflect the percentage of instances where
high stresses and/or limitations of the material combined, signifying that a failure would occur.
The odds or chances are also weighed for successively larger initial crack sizes to correspond
with the deterministic FM model results.
8-48
10305538
Figure 8-20
Probabilistic Analysis of Turbine Blades
8-49
10305538
In past investigations where components have been subjected to this procedure, a well-designed
blade will generally result in a predicted chance of failure on the order of 1x10-4 (1 in 10,000) or
lower over a reasonable measure of service (cumulative hours or starts). Military Standard 882C,
which establishes system safety program requirements for marine and aircraft turbines, would
classify these odds as remote, but with the unlikely possibility that failure would occur as a
random event within the lifetime of an item. This level of risk is consistent with U.S. Department
of Energy requirements for nuclear turbine rotors, in which the allowable chances for generation
of a missile from a failure are also 1x10-5–1x10-4.
Characterization of critical material properties should be based on the amount of scatter

registered in laboratory results of material tests. For HCF, the PDF of the fatigue strength
coefficient (σf) is used to describe the randomness of the material’s HCF fatigue strength. For
LCF, the (εf ) is similarly used.
Because dynamic stress and HCF remain the governing criteria in initiation and propagation,
critical parameters of stimulus, damping, and frequency ratio are represented as statistical
distributions, guided by the results of FEM analysis. Once descriptions of the random forcing,
damping, and frequency have been established, a PDF of dynamic stress for the selected mode of
interest can then be constructed and provided as a source of random input to the probabilistic
model. Using this procedure, a possible distribution of dynamic stress is produced by a mean
stress of a known standard deviation.
As with the results produced from the life models, the odds produced from the probabilistic
applications are viewed in relative, not absolute, terms. Each steam turbine is ultimately a unique
problem because of the way it is controlled and operated. In some instances, additional factors
can be included to assess particular features of the problem, such as erosion notch sizes in
control stage airfoils or yield strength and temperature in SCC problems. Even in such instances,
the role of the probabilistic model is to reinforce the original conclusions drawn from the LCF or
HCF life assessments. In assessing problems initiated by erosion or corrosion, its additional
value is that it can advise the operator of the relative chances of a fatigue failure as either
condition worsens.
Final Comments on Phase 2 Activities
As this section reveals, although specialized skills and experience are required to analyze blades,
interpretation of the results produced from these studies is fairly straightforward. Given the cost
and time involved, many blade designs are not rigorously analyzed. When such studies are
performed, the objective is not to diagnose a failure, but to check that the results fall within
established criteria. In the wake of a field failure, it is not unreasonable to question whether these
criteria might have missed something that they were not developed to register.
Finally, if qualified specialists are employed, the basic results they independently produce will
generally agree. Finite element analysis is a sufficiently mature tool. As discussed, critical
assumptions, particularly those that go into the dynamic analysis of the blade-disk structure, have
reasonable bounds that limit the discretion of the analyst to make a design survive or fail.
Probabilistic techniques can further remove the potential subjectivity that an analyst might
8-50
10305538
impose on the calculations and results. When viewed against the service history and
metallurgical analysis, the facts will typically converge toward an identification of potential
causes and/or sources that have not already been eliminated and end speculation surrounding a
key item, that is, the magnitude of stresses that ultimately transform operating conditions into
forms of damage common to steam turbine blades.
Presuming that experienced specialists have been used, the availability of stress and frequency
results provides a natural means to focus remaining discussions between the operator and the
OEM in terms of forming a final consensus on what to do based on agreement as to what caused
the failure. Disagreements in results obtained from a structural analysis generally revolve around
either the validity of the operating frequencies or the magnitude of the dynamic stresses. Because
material properties are generally available in the published literature and the service history prior
to failure is readily definable, stresses are the third-easiest item to dispute as input to various life
prediction models. It is therefore important that trained specialists from within an OEM’s
organization or from independent sources perform these calculations. Generally, these efforts are
performed in parallel, although the OEM should be invited to assist if the operator contracts an
independent organization.
8-51
10305538
10305538
9
PHASE 3: IDENTIFYING A REPLACEMENT STRATEGY
Table 9-1
Subject:
This section discusses a final phase of a typical blade failure investigation. In Phase 3, it is decided
what type of permanent strategy is appropriate based on the diagnosis of root cause. It addresses
when a new design is considered necessary as opposed to the alternatives, which include replacing
the blades in kind or repairing superficially damaged blades.
Topics Covered:
9.1 – Available Longer Term Strategies
9.2 – Dealing with Design-Related Root Causes
9.3 – Dealing with Operating-Related Root Causes
9.4 – Dealing with Environment-Related Root Causes
9.5 – Dealing with Material-Related Root Causes
9.6 – Salvaging Damaged or Worn Blades
9.7 – Supporting Inspection Criteria
9.8 – Application of Coatings
Tables/Figures:
Figure 9-1 illustrates common examples of LP leading edge erosion and how they can be modeled
to establish erosion removal limits.
Available Longer Term Strategies

If the tasks in Phase 2 have been successfully completed, the operator should have sufficient
confidence in the assignment of a root cause and a clear understanding of how the original failure
occurred. The objective of Phase 3 is to further leverage this insight and understanding to
execute a strategy that has a high probability of not resulting in a repeat occurrence of the same
problem. In cases where adjustments to the blade design are considered to be part of the solution,
a secondary objective is to check that any such changes are not going to reinforce a weakness of
the original design, thereby creating a new set of problems. Toward this goal, the efforts
involved with Phase 3 are meant to answer one or all of the following questions:
• Is a new design necessary?
• Could changes in unit control systems or operating limits better correct the root cause?
• Can a repair salvage superficially damaged blades?
• What additional inspection criteria might be used to manage the remaining life?
9-1
10305538
Phase 3: Identifying a Replacement Strategy
If it is determined that the original problem is a direct consequence of excessive stresses (steady,
dynamic, or thermal), some alteration of the design is necessary. A replacement in kind will only
have the same susceptibility and chance of developing a similar problem. This is particularly true
for synchronous vibration problems where inadequate detuning margins will continue to produce
fatigue cracks unless changes to the original mass or stiffness of the assembled row are made.
If the role of HCF or LCF is one of participation and not initiation, solutions other than
modifying the original blade design might be more productive. For example, solutions involving
premature erosion or corrosion might involve changes to the systems that control or contribute
particles, droplets, or contaminants in the steam flow.
When talking about replacements, the term new can be misleading in that it naturally implies the
replacement will represent the most recent blade technology the OEM can offer. However, in
selecting a repair or replacement strategy, a new design is really represented by any alteration to
the original blade where significant modifications are made to the root attachment, airfoil, or
grouping arrangement, or the blade material itself. Particularly for longer blades, any
adjustments to the mass or stiffness of the original blade-disk/rotor system will have an effect on
the natural frequencies.
Modifications proposed to the blade root attachments are less common because they require the
original disk to be removed or require sufficient material within original serrations to
accommodate removal and contouring. Generally, changes involving the root attachment region
of a turbine row are meant to remove damaged disk material or reposition the damaged locations
so that they are away from the high stresses formed in notches or fillets. Even a drop-notch
replacement applied to remove SCC from the original disk rim attachments normally retains the
same blade root geometry, although the airfoil must be extended to compensate for the reduction
in disk diameter. In an HP blade row, an original fir tree root was replaced with a pinned finger
design, again to allow removal of rim material damaged over time by creep. The creep damage to
the rim caused the original tolerances prescribed for the fir tree blades to result in gaps that were
larger than prescribed for the original design.
Most proposed design changes are subtle, retaining the original root attachment and confining
adjustments to the platform and above. As such, they particularly lend themselves to re-using the
original FEM and eliminating much of the initial setup work associated with the steps in the
Phase 2 analysis. Depending on the extensiveness of the changes, this can result in an economy
of savings of 50–75% in the time and cost to evaluate proposed modifications relative to the
initial investigation. Included in this potential reduction of time and cost is the fact that Phase 3
efforts primarily evaluate suitability, not investigate root cause. In design-related problems, once
the critical parameters are identified, gauging the potential of a proposed modification is usually
quite straightforward.
Dealing with Design-Related Root Causes
Problems that are truly design-related will typically require the original rows of blades to be
replaced, even if the corrective measures involve regrouping or detuning strategies. Failures
prompted by excessive steady, dynamic, or thermal stress are the consequence of accumulated
9-2
10305538
fatigue damage. Because the all-important phase of initiation is often confined to regions below
the detectable threshold, microscopic cracks might have already formed that will not be caught
by inspection before the blades fail. In original blades that have not yet failed, this fatigue
damage is present, but it cannot be quantified certainly enough to know whether they might
continue to serve a normal lifetime now that the original problem has been found and corrected.
The changes offered in replacement designs are either corrective or preventive. Corrections are
usually intended to improve the original detuning of specific modes of vibration without
upsetting the other modes, or to relieve the magnitude of steady stresses in critical joints or
fillets. Preventive measures essentially eliminate the problem area completely.
Not only are freestanding blades or blades with integral covers easier to manufacture, but they
completely eliminate the possibilities for cracking in peened and shrouded designs, and they
reduce the stresses caused by shrouds extended over several blades. Elimination of tiewires is
also often promoted as having the additional advantage of reducing losses; that is, any
obstructions in the steam flow passage improve the efficiency of a design.
However, there is a trade-off in these scenarios. Freestanding blades are more susceptible to
asynchronous vibration. The tolerance fits of blades that rely on interlocking integral covers must
be carefully designed. In particular, the initial gaps have to be controlled in a manner that allows
the blades to be installed so that the contacts or gaps are reasonably uniform around the
circumference; this way, the contact loads and forces at full speed are correct. Each of these
poses significant design challenges as the length of the blade is increased. Any mass added to the
tip will increase the loads on the attachments. The tips of many of the longest last stage LP
blades can be severely twisted to accommodate the steam flow in this outermost region. Any
integral cover strategy still has to be large enough to bridge the tip-to-tip spacing without
overloading the airfoil.
If a generic problem becomes apparent enough to force the OEM to go back to the drawing
board, improvements in performance too often become the featured aspect of the new design,
emphasizing attractive paybacks as part of a replacement strategy. For the operator, it is critical
to remember that any significant change to the mass or stiffness of an original blade design,
particularly LP turbine blades, will affect the operating frequencies and corresponding magnitude
of dynamic stresses. It is to the operator’s advantage to quantify the relative impact of changes to
the airfoil length, stacking line of profiles, and tip in terms of its structural behavior and
increased efficiency.
In this regard, the key indicators highlighted in the previous phase of the root cause investigation
are similar to those used to evaluate any design change ( that is, frequencies, stresses, and
initiation life). The indicators are as follows:
• Frequencies control detuning and, therefore, dynamic stress and the potential for HCF to
become a problem.
• Frequencies plotted as Campbell and interference diagrams allow the analytical predictions
by the OEM or a third party to be authenticated.
9-3
10305538
• Steady stresses control the rate at which LCF is accumulated, and it affects the susceptibility
to damage from HCF, SCC, CAF, or erosion.
• Fatigue damage is disproportional to stress. Nominal changes in stress, therefore, might not
result in a significant improvement in service life or damage tolerance.
Dealing with Operating-Related Root Causes
When the root cause is attributed to an operating-related issue, these problems sometimes cannot
be corrected with adjustments to the original blade design. More often, these problems are
addressed by preventing the unit from encountering the conditions that resulted in failure of a
row. Among the operating issues that might be linked to blade failure are asynchronous
vibration, torsional system interaction, localized grid swings or disturbances, or PAA, as
discussed in more detail in the following:
• Problems caused by asynchronous vibration are normally addressed by readjusting zones of
avoidance (combinations of loads and backpressures) to minimize the potential for flutter.
The propensity for flutter or stability of a turbine blade design is reflected in the damping
available to it and general features that inhibit airfoil flexibility. Long, thin, freestanding
blades are most susceptible because they have less available damping and/or lack
intermediate structures on the airfoil or tip to resist the vibratory motion caused by adverse
flow conditions. Correcting these problems by design represents both a significant
investment and technical challenge, because the analysis of flutter is presently a cutting-edge
calculation, meaning that a program of field testing is necessarily going to be involved to
qualify a flutter-resistant replacement.
• Problems caused by torsional system interaction require an improvement in the detuning of
the entire turbine-generator system because it is the resonant response of the system modes
that is the root cause of the dynamic stress in the row of blades that failed, not the individual
modes of the particular row. Corrective measures normally involve adding substantial masses
to the system. The original blades might need to be replaced (to account for fatigue damage
experienced prior to correction), but adjustments to their design will not correct a fatigue
problem caused by a resonant vibration problem of the drive train.
• Problems caused by localized grid swings or disturbances are normally prevented by
maintaining trip settings on the unit so that the maximum drift in shaft RPM is confined
within a range of ±0.5 hertz. In situations where this is not possible, problems appear when a
fundamental mode of vibration is marginally detuned.
• In this scenario, a detuning strategy can be reasonably applied without requiring a major
redesign of the original blade. Problems caused by PAA are normally controlled by limiting
the period at which partial arcs are maintained, or by revising the steam admission system to
the control stage so that the transient effects of passing from closed to open arcs is
minimized.
9-4
10305538
It should be mentioned that the preventive measures associated with imposing more limits for
off-design load and backpressure combinations do not necessarily represent a correct solution.
As a rule, the last stage turbine blades should be able to run within fairly reasonable
backpressure ranges, both low and high, without encountering an asynchronous problem.
Unfortunately, in the era of deregulation, units that were designed and operated for base load are
now frequently run at part-load when demand is low. This has resulted in a notable increase in
the frequency of asynchronous problems compared to those documented in the 1980 survey. It is
also an indication that correctly defining these extremes or limits is not nearly as straightforward
as detuning a synchronous vibration problem.
Despite these challenges, at some point, it needs to be recognized that if the limits become so
restrictive that the unit cannot operate economically, a redesign is necessary no matter what the
cost and challenge it might represent to the OEM.
Dealing with Environment-Related Root Causes

Problems related to the stage environment are generally a consequence of contamination,
excessive moisture, or the failure to prevent particles from becoming entrained in the steam flow.
It is logical to first address these at their source by protecting steam quality and ensuring that
moisture/particle removal systems are not under-designed and are functioning properly.
Normally, the first measure taken directly on the blades affected by these problems is to enhance
their resistance to the formation of pits, trenches, or cracks. The second measure is to minimize
stresses in locations most susceptible to these damage mechanisms. In these cases, however,
stresses are rarely the predominant mechanism that controls the time spent in the initiation of
cracks. Therefore, a strategy recommended to reduce stresses in locations often most susceptible
to these problems is not going to eliminate the root cause, but only mitigate the consequences by
improving the tolerance of the region to any future damage.
Of the three dominant environmental-related problems, SCC is the least affected by changes to
the design stresses. The threshold stress intensity for SCC to form is well below the magnitude of
stress produced in most blade/disk rim attachments at full speed. Residual stresses might also be
of sufficient magnitude to cause SCC. Once SCC starts, tests have also shown that the growth
rate is fairly deliberate, affected more by the temperature and yield strength of the material than
the operating stresses around the cracks. Corrective measures for SCC are therefore focused on
eliminating the source of contamination, upgrading the material, using protective coatings.
Blades are less susceptible to SCC because of their higher chrome content, but they can exhibit
SCC if the conditions of moisture and temperature are satisfied or exceeded due to problems of
the steam flow.
In this regard, the benefits of shot peening remain a source of continued debate. The advantage
of peening is that it produces a layer on the surface material that is in compression. As such, it
provides additional resistance to the detrimental effects of the environment. The principal
limitation of shot peening is the limited depth to which this compressed layer can be formed, and
that beneath it the residual tensile stresses will be increased. In wet steam stages, erosion or
pitting can compromise this layer, which might be only 10 mils (0.3 mm) or less thick. If the
immediate layer is penetrated by a pit or trench, the tensile stresses will favor rather than retard
the formation and further propagation of a crack.
9-5
10305538
Prevention of failures from CAF is sometimes possible if the source of the dynamic stress
promoting the formation of cracks from pits scattered over a portion of the blade is caused by a
mode that is inadequately detuned. In vintage designs that suddenly start to have fatigue failures
after years of reliable service, this can indicate the involvement of higher modes—such as the
fourth and fifth modes—that were not originally detuned by design. As indicated from the
discussion of blade dynamics, usually, a well-detuned mode will produce dynamic stresses on
the order of 1 ksi (7 MPa) or less. Higher order modes or fundamental modes that are on the edge
of the preferred detuning limits can produce stresses on the order of two or three times higher.
Although not high enough to pose an initiation problem, they are significantly larger than if they
were detuned. Adjustments to tweak the airfoil can address these nominal detuning issues and be
of benefit as a second line of defense against CAF failures.
Normal erosion is generally prevented by strategies to harden the leading edge so that it can
handle the impact of the water droplets as they strike the surface. Different strategies are
preferred among the OEMs. Their relative reliability is generally less a function of design
(stresses) and more the quality of the workmanship.
Dealing with Material-Related Root Causes
On occasions, the solution to a problem involves changing the original material, with the most
notable being the substitution of titanium for conventional blade alloys, particularly blades made
from 17-4PH. Used in the longer LP turbine rows, titanium is viable because it is strong and its
reduced density can significantly lower stresses on the root attachments. Root steady stresses are
generally the governing design parameter that determines the allowable length for last stage
blades. An understated convenience of titanium is that its lower density combined with its lighter
weight tends to preserve the original frequencies and tuning margins established with the heavier
and denser steels, making it easier to employ as a substitute when compared to materials that
would require significant redesign to detune critical modes.
Unfortunately for titanium, its original introduction in the 1980s to correct root-cracking
problems in LP turbine rows was compromised by problems involved with the limited detuning
of critical modes. The physical restrictions imposed by the original design carried this problem
over when titanium was substituted. Combined with the additional cost, the potential of titanium
languished for some time. This has changed as the drive to squeeze more power from turbine-
generators has naturally focused on the last stage of the LP turbines. The drive to extend their
length and physical limits on how massive a blade root can become has required the use of a
lighter material. The cost of titanium has also come down.
In summary, in any blade-cracking problem where high steady stresses play a significant role,
titanium can be considered a viable strategy. Again, this can be quantified to see how much the
substitution of titanium will reduce steady stress in a region like the root attachments, reflected
as the commensurate improvement in predicted LCF life. To improve damage tolerance against
erosion and corrosion, the resistive properties of titanium are also subject to debate, although it is
conventional wisdom that titanium is superior to the conventional stainless steels. A more
tangible advantage of titanium can be derived from the analysis in terms of the reduction in mean
9-6
10305538
stress that it produces in regions that have proven susceptible to pitting, erosion, or corrosion
when compared to those produced by the original steel. As shown in the HCF life model, a
reduction in mean stress will increase the tolerance to any dynamic stresses relating to the
initiation or propagation of cracks.
Another corrective strategy associated with steam turbine blades is the substitution of Jethete
material for 403 stainless materials. The principal reason for this substitution was to eliminate
the manufacturing challenges and failure problems posed by stellite shielding strategies applied
to protect the leading edges of last stage blades against WDE (another example of a design
prevention strategy whereby the troublesome device is eliminated). Jethete has the stated
advantage of allowing the leading edge to be hardened by a carefully controlled process of
heating. As with similar strategies, its success as a corrective solution is dependent on strict
control of the flame hardening process and consideration of how it might be affected after the
zone of hardening is compromised over time. There have been reported instances of situations
that suggest that the crevices formed from erosion more easily transition from the phase of
incubation to crack initiation. In the presence of nominal dynamic stresses produced from
resonance of higher order modes, they would have less tolerance than the original 403 stainless
base metals around the stellite nose. In other words, a post-failure evaluation of new blades made
from Jethete might focus on how well the dynamic stresses in these regions are minimized,
although the root cause for the formation of initial cracks was attributed to erosion.
Salvaging Damaged or Worn Blades

With the exception of failures that are design-related (where HCF or LCF initiation is by
excessive stresses), situations where the root cause is initial damage by erosion or corrosion have
the potential to be refurbished and retained for further service. The success of any such
refurbishment strategy is dependent on achieving two basic criteria: (1) the damaged region must
be sufficiently blended and dressed to remove any residual traces of cracks that might have
formed, and (2) the amount of material cannot substantially increase the steady, dynamic, or
thermal stresses in a manner that would significantly accelerate initiation by fatigue or creep.
Of the two criteria, the first is often the more difficult to establish compliance with. The
consequence of removing material confined to portions of the blade airfoil can be readily
examined using the original FEM (see Figure 9-1). This is often done either by choosing a worst-
case scenario made obvious by inspection of the row, or by parametrically approaching the
possible amount of material that might be removed in terms of succeeding depths of milling or
grinding. In either scenario, the stress results are compared to the original analysis to see if the
relative differences appear significant. In addition to the stresses, the natural frequencies are
examined to be certain that the possibility of creating a resonant vibration condition is
minimized.
9-7
10305538
Figure 9-1
Evaluating Material Removal Limits
9-8
10305538
When a minimal amount of surface material is removed, the results of the relative comparison
are viewed as not eliminating the problem, but instead resetting the clock back to when the
blades were undamaged. It is presumed that the mechanism that originally caused erosion,
pitting, or SCC remains unaffected. As such, the effectiveness of the refurbishment can be
expected to last on the same order as before the problem manifested itself, presuming the stresses
are relatively of the same magnitude. This assessment further presumes that any accumulation of
HCF or LCF was nominal.
In situations where correction of the damage would result in a pronounced (visible) change to the
original geometry, more attention should be given to the local stresses created by the
discontinuity and the consequence for the frequencies of vibration. Such scenarios often involve
advanced conditions of erosion that form gradual notches in the unprotected metal on the leading
edge or extreme cases where a significant notch is formed on the airfoils or a massive amount of
material is missing. These are normally problems of the first HP turbine rows and last LP turbine
rows, although an artificially created steam jet can notch any row.
As a general rule, the lower the damage is on the airfoil, the more likely it is to produce a
significant increase in the steady stresses. If the end of the notch after blending is too deep, a
LCF problem might be created whereby subsequent start-stop cycles initiate a crack and thereby
increase the risk of an HCF failure. If the damage is mostly confined to the tip, the temporary
effect of removing the tip or a portion of it might allow the turbine to run until the next
scheduled outage. Although such cuts are inherently not going to create a stress concentration
and steady stress problem, be sure to always have any material removal be guided by the results
of a frequency analysis. On LP blades, deep notches are inviting locations where such cuts
should be made; however, they might also be the worst height to take away mass. A frequency
analysis should always be performed to advise where to make the cut so that the modes of
vibration remain detuned.
Returning to the first criterion, it is paramount to the success of any refurbishment that the depth
of material removal and subsequent blending effectively eliminate any residual traces of cracks
that formed but not yet reached a length that was detectable or led to a failure of the blade. As
stated, a realistic objective of any refurbishment plan is to return the process or sequence of
damage to the incubation phase. Any crack tip left in the repaired zone will compromise this
fundamental premise governing the success of the refurbishment. The process of blending or
dressing can sometimes hide such indications rather than remove them. In considering whether a
refurbishment strategy is feasible, the extent or severity of apparent damage on the rows
involved should be appraised. Evidence of particularly deep cracks or pits widespread over a
majority of the blades increases the possibility that either the amount of material to be removed
is likely to prove excessive or the risk of leaving a portion of a residual crack is high.
Presuming that the damage is nominal, the objective in the final analysis is to use the FEM to
establish how deep a repair might go (grinding) to ensure that any superficial damage is removed
without compromising the basic integrity of the structures.
9-9
10305538
Supporting Inspection Criteria
Although NDE examination can detect indications down to a range of 20–30 mils (0.5–0.8 mm),
these would still permit cracks of sufficient size to compromise the operating life of the original
design. As a minimum, it is recommended that in addition to dye penetrant tests, the critical
regions established as susceptible to cracks should be dressed off; that is, about 30–50 mils
(0.8–1.25 mm) of the surface in the general area should be removed if possible.
The results from the examination can also be used in making a determination of what action
should be taken for sister units operating similar rows of the same turbine blade design. If the
root cause concludes that the original failure was design-related, the number of starts or hours of
continuous operation produced from the original analysis provides a perspective of how long
similar blades might be expected to operate before succumbing to the same problem. The results
from the probabilistic analysis should further assist by indicating how the odds of a failure
change as operating history increases.
The results from a basic stress and frequency analysis also show where inspection efforts should
be focused for a particular design, that is, the locations that are most sensitive to any damage
because they coincide with regions of maximum stress. In particular, for blades that are known to
have marginal tolerance to any damage, the results might be used to guide their regular
inspection and/or locate areas where cleaning and polishing are adopted as routine maintenance.
In scenarios involving refurbishment or repair strategies, the application of finite element

analysis has generally been discussed. In some instances, sensitivity studies used to guide the
removal of damaged material are extended to determine at what point the integrity of the blade
design is likely to be compromised (that is, at what point the erosion advances to when the
resulting loss of mass will shift a detuned frequency enough that it has a high probability of
becoming resonant). A similar examination of gradual notching due to SPE can be performed to
determine the point where the increasingly unsupported mass of the cover is likely to increase
airfoil stresses to the point where there is a high risk of initiation and rupture.
Application of Coatings
Coatings often represent a low-risk strategy because they essentially conform to the original
features of the blade design in a manner that their effect on the mass or stiffness is negligible. In
other words, they cannot hurt and might help. Coatings are more likely to be effective in
preventing failures where a root cause is corrosion-related, such as SCC or CAF. In these two
problems, coatings are applied to address the sequence of crack formation at its initial stage of
incubation, that is, as a first line of defense. As a preventive measure against erosion- or
corrosion-related problems, coatings are effective only as long as they remain intact on the
surface. In a particularly severe erosion scenario, the additional operating life they provide will
probably be negligible. If the root cause of the failure is design-related, coating the surface (like
shot peening) will have little consequence in slowing or preventing the formation of cracks
caused by excessive stresses. In addition, the impact on future NDE should be considered prior
to implementing a coating strategy.
9-10
10305538
10
POST-FAILURE GUIDANCE
Table 10-1
Subject:
This section offers some additional guidance on how the knowledge gained from a root cause
investigation of a steam turbine blade failure can be transferred to improve the chances of procuring
and installing reliable blades from OEMs or third-party suppliers.
Topics Covered:
10.1 – Upgrading Successfully
10.2 – Reviewing Bid Packages
10.3 – Validation of Replacements
10.4 – Installation and Inspection
10.5 – Monitoring to Detect an Impending Failure
Upgrading Successfully
Often, given the considerable capital expense associated with purchasing multiple LP or HP
rotors, it is problems associated with the critical blade rows (first HP, first IP, L-0, L-1) that
initiate their serious consideration. Disenchantment caused by generic design problems
associated with an original turbine prompt an operator to consider alternative suppliers that offer
a steam path of completely different technology. Increasingly frequent problems associated with
normal age and wear indicate a need to install new rows to replace those whose time is up. The
irony in either scenario is that although these reliability issues are often the underlying motive
for seeking new rotors with upgraded turbine blade technology, the final selection often becomes
focused on performance improvements featured in the upgrade package.
Performance improvements are a tangible benefit that should not be discounted in the selection
of new blades. However, the operator should not lose sight of the fact that a catastrophic failure
will quickly drain the benefits of additional power output. This said, there is no reason to suspect
that the rotor replacement technology packages currently offered by major OEMs are not going
to perform reliably. The only caveat is that many times, the specific rotors and blades that are
being considered have only a limited operating history because they are new technology,
10-1
10305538
Post-Failure Guidance
involving larger blades or steam paths with additional rows to maximize performance
improvements. Referring back to the lessons drawn from the experience of the preceding
generation of turbines, stresses do not scale linearly, and sometimes, original design criteria must
be revised to catch problems that are subtle rather than self-evident.
Toward this objective, the experience of surviving a blade failure can be used to help in the
selection and qualification of a replacement package consisting of new blade designs. As part of
a replacement strategy, the prior experience list should always be verified to determine how
many of these cited examples involve exactly the same rotor and blades.
Reviewing Bid Packages
In the competitive market of rotor replacements, many operators often attempt to make a
technical assessment of the competing products at the initial selection phase. Although
procurement specifications can assist in identifying what technical information is sought and
what criteria should be applied to QA/QC parts, the level of detail necessary to make a Phase 2
assessment is never available in the form previously identified in the root cause investigation:
specific blade dimensions that could be used to model the designs, detailed stresses that are not
normalized, or frequency diagrams that clearly identify the margins of detuning. In reviewing
alternatives from a technical standpoint, it is a more realistic goal to assemble enough
information about the rotor/blade technology to determine answers to the following questions:
• Are the appropriate materials selected for the different operating stages?
• How many rows are involved in the new turbine?
• Which rows are detuned and which rows are meant to be able to operate in a resonant
condition?
• What strategies are used to control the at-speed frequencies and stresses on the blades?
• How are the blades attached to the rotor in each stage?
• What methods are used to protect rows that operate after the Wilson line of the turbine?
• What is the height of the last stage of the LP turbine?
It is normal for a bid package to list the essential chemical composition and mechanical
properties expected within the various components that form the rotor. These are generally
helpful in identifying the basic ASTM class and criteria that the blades would fall within.
Keeping in mind that a goal of most replacements is to avoid a problem associated with one or
more rows of the original rotor, indications are sought of relevant differences in the new
technology. For example, the number of rows is an indirect indicator of (a) whether the designs
share the same basic heritage as the original blades, (b) whether the number generally
corresponds to the amount of projected improvement in additional power output, and (c) whether
the maintenance requirements are likely to increase with more rows composed of more blades.
10-2
10305538
Because most rotor replacement strategies are cost-justified based on potential increases in
output relative to the original rotor, changes to the original steam path have to be made to
increase efficiency. The magnitude of changes should correspond to the potential gain in power
output. Many efficiency improvements come from improved sealing systems and leakage
control. Tweaking the blade airfoils of the initial rows can contribute to an overall increase in
power output. In most upgrade scenarios, it is the last stage that is the primary row that is
modified. Being the last stage, and sometimes involving as many as six rows of the same design
within the unit, increases in last stage blade length of 1–2 in. (25–51 mm) can provide a sizable
portion of an overall performance improvement guarantee. In other scenarios, often where a
competing product is offered for replacement, the new OEM will prefer to completely redesign
the flow path anyway. As such, they might increase the total number of stages to better manage
the flow. As indicated, more blades should not be perceived as a disadvantage, but they should
be considered in terms of what they might add in maintenance.
In large steam turbines, the last three rows of the LP turbine will rely on blades long enough that
their modes of vibration will fall within the first 10–12 harmonics of running speed and,
therefore, must be detuned to prevent resonance. Statistically, detuned blades are the more likely
to encounter problems, so it is useful to identify these rows. In concert with this is how the OEM
prefers to manage or control these frequencies. As reviewed in the previous sections of this
report, no particular design or detuning strategy is overtly superior. What is sought from this
information is whether the OEM favors freestanding designs, integrally shrouded designs, tip
linkage devices, or conventional peened/tenon covers. Rows that require tiewires or snubbers
should also be identified.
Certain root attachment designs are also favored by different OEMs. As a general observation,
finger type pinned root attachments offer the redundancy of multiple fingers to protect against their
catastrophic failure before cracks in the attachments are discovered. Although fir tree attachments
also have multiple pairs of hooks, these must be sufficiently sized and the tolerances carefully
maintained to ensure that the loads are evenly distributed. The size and width of tangential entry
blade roots are more limited by the available space on the disk rim. The advantage with tangential
entry blades is that they are normally easier to remove and inspect, particularly in the last stages
where deposits tend to form. The stresses produced from larger finger root blades will also yield or
distort the pins, adding to the difficulty and cost in removing them for inspection.
Methods used to protect the last stage from erosion have also been discussed. Generally, they
involve either flame hardening, brazing, on stellite shields or welding a formed length of stellite
to form the leading edge. Again, the objective is to identify the method preferred rather than to
distinguish a superior method. The effectiveness of any of these three strategies is predominantly
controlled by the manufacturing process and not the design.
Finally, the length of the last stage turbine blade is sought as the simplest means of identifying
how much of a long list of prior operating experience is directly relevant to the turbine and
blades being offered. Such information can be found in EPRI publications [11]. This is
particularly useful in establishing when designs involving new technology were actually
introduced. Without the advantage of a structural analysis to validate the claims of reliability for
proposed replacements, actual duration of service is an extremely useful indicator of infant
mortality problems that attack the blades.
10-3
10305538
Validation of Replacements
If possible, a plant should include cooperation of the OEM as part of the final letter of intent to
purchase, whereby they agree to provide sufficient dimensional information as required to
perform an engineering analysis of selected blade rows. This arrangement does not have to
include the distribution of dimensional information to the plant, but only to the third party tasked
to make the FEM and perform the relevant stress, frequency, and life calculations.
In such circumstances, the intent of the examination is not the same as that of the Phase 2 effort.
It is more closely aligned with the goals of Phase 3, that is, to verify that stresses are well
managed. Well-managed implies that the maximum steady stresses in the root are well
distributed among the pairs of attachment hooks or number of pinned fingers. The loads imposed
on the root attachment by the airfoil are generally centered with the platform, rather than
favoring one end over another. As noted several times, dynamic stresses are considered well
managed if the fundamental modes are detuned. In both scenarios, the calculated stresses applied
to the respective HCF and LCF models should produce results that indicate more than enough
cycles before there is any chance for cracks to form.
For high-temperature blades, a creep calculation might also be undertaken, again to identify
whether the formation of cracks or lifting problems sometimes experienced by older designs
have been mitigated.
Given the costs associated with any analysis, rows and problems exhibited over the historical
record of the original rotor(s) operation should be given first priority. If an original rotor is being
replaced because it was routinely experiencing cracks in covers or tiewires, developed SCC over
time, or suffered premature erosion damage to the blades, these problems should single out the
rows that should be examined in detail.
Rows that are detuned and/or operate in a wet steam environment should be made a second
priority. As a rule, the blades on the back end and front end of the steam path are the ones most
likely to be subjected to the highest stresses and the extremes imposed on the operating
environment. Therefore, for LP rotor replacements, the last stage is often the most likely
candidate to seek a detailed evaluation, followed by the L-1 row, depending on its length. On HP
rotors, the first stage is the one to select.
It is emphasized to both management and the OEM that any technical qualification is not
expected to identify a design problem. It is presumed that the OEM’s blade technology packages
have been put through a rigorous process of design analysis and testing. The purpose of this
examination is one of due diligence on behalf of the plant.
10-4
10305538
Installation and Inspection
The comments and observations regarding the role that certain installation requirements and
allowable criteria can play in compromising the reliability of a blade design remain just as valid
for new replacements, particularly as continuous linking strategies involving integral shrouds
have become popular with all OEMs. As such, the following guidance is offered when obtaining
and installing replacements:
• In the original QC process, blade material should be checked in terms of chemistry and
mechanical properties. Frequency tests should be used to indicate dissimilarities that are not
immediately apparent by looking at the individual parts, and to reject parts that display a
response that deviates significantly and cannot be corrected.
• Tolerances in root attachments are critical. If the gaps between pairs of load bearing hooks
are not uniformly maintained to within certain minimal criteria (usually within ±0.5 mils [0.1
mm]), the centrifugal forces at full speed will not be sufficient to close them. The result will
be to have the loads redistributed among the remaining pairs where contact is achieved.
Stresses in the fillets can be increased substantially as a consequence, potentially reducing
the LCF life in the worst possible location of the blade.
• Tolerances between interlocking and interfacing covers are equally important. For LP blades,
the initial gap is sized to accommodate the predicted magnitude of untwist that centrifugal
force is expected to produce on the airfoil. If the gap is undersized, it could result in
excessive stress in the shroud. If the gap is oversized it could result in marginal contact and
fretting or no contact, causing all of the design frequencies to be invalidated. Not only should
the installed row be inspected against allowable initial gap criteria, but the uniformity of the
gaps should also be checked.
• Conversely, many shorter blades that do not have much untwist of the airfoil are installed in
a manner that a pre-load is created throughout the row of sufficient magnitude to maintain a
firm contact at full speed. Again the purpose is to increase the fundamental frequencies
minimize the capability to vibrate that might otherwise be produced if the blades were left
freestanding.
Monitoring to Detect an Impending Failure
Despite progress in the technology of non-contacting BVM systems, there has yet to be
demonstrated a reasonably inexpensive, straightforward technology that can warn of an
impending blade failure. Although vibration is routinely measured through the bearings of the
turbine, the individual frequencies of different turbine rows are simply not detectable with
sufficient sensitivity to detect impending separation of a blade given the transmission path that is
involved. Even if the cost of conventional strain gauge telemetry systems could be overlooked,
the limited operating period caused by the harsh operating environment and requirement to
instrument every blade would make this test method impractical.
10-5
10305538
Non-contacting BVM systems have overcome this problem, but they sacrifice the precision and
accuracy of direct analog measurements for longevity and simplicity of installation. Because of
the nature of the time-of-arrival technique on which these non-contact systems depend, their
ability to detect resonant vibration is limited, and they are better suited to catch an intermittent
condition of asynchronous vibration.
The primary limitation on detecting an imminent catastrophic problem for blades on the stages
most likely to fail is the relatively high stresses involved. As an assessment of damage tolerance
for any large steam blades will reveal, given the magnitude of steady and dynamic forces
involved, a survivable crack, particularly in the airfoil, is relatively small. Large cracks
(approaching 100–200 mils [2.5–5.1 mm]) exceed the threshold where HCF will take control and
fail the blade. For the sake of argument, assume that such a crack had formed in a tangential
direction on the lower portion of the airfoil of a freestanding blade. The impact of such a small
crack would be unlikely to affect the original stiffness long enough to cause the frequency to
shift in such a manner that it would stand out from the range of frequencies produced by a row of
blades.
Using BVM technology to detect an initial synchronous problem might be possible if the issues
associated with aliasing of the signals can be overcome. (Synchronous vibration generates no
time-of-arrival variability over multiple revolutions at a constant RPM because the blade arrives
early or late by the identical amount on every revolution. In signal processing, this would be
described as aliasing to dc. To identify blade response at synchronous frequencies, special data
processing algorithms and an optimum number of sensors must be used.)
Using BVM monitoring to detect and map an asynchronous vibration problem is more feasible
for unshrouded blades. However, catching cracked blades before HCF takes control is highly
unlikely, presuming that every critical stage could be routinely instrumented in a cost-effective
manner.
10-6
10305538
11
CLOSING REMARKS
Additional Resources to Assist in Using Guidelines
Although managing or preventing a steam turbine blade failure might sound like a daunting task
to operators who have little or no experience in the design of these critical components, this
report is intended to show that the opposite is true for the following reasons:
• When blades fail in steam turbines, the main possible reasons can be represented by a concise
list that can quickly be reduced with a limited amount of forensic information.
• The techniques for evaluating the role that features of the blade design play in the failure are
relatively straightforward.
• Stress (or strain) as a function of load, frequency, or temperature is the critical parameter
sought in a technical examination to establish whether the design plays a primary or
secondary role when used in conjunction with simple fatigue life models.
• If the problem is design-related, actions taken will require modifications. When examined,
these would be expected to significantly reduce the critical stress by an amount that would
produce a magnitude change in predicted fatigue life.
• If the design plays a secondary role in circumstances where intermittent operating stresses
are responsible for the initiation of cracks, the off-design operating condition must be sought
out and corrected.
• If corrosion or erosion is involved, operating stresses should be checked and/or minimized to
provide maximum damage tolerance.
• If the operating stresses prove to be marginal or reasonably low, a protective coating must be
applied and/or the source of contamination, moisture, or particles must be addressed.
Included as appendices are lists that summarize the basis questions raised at each step of the
methodology and a series of cases that illustrate their application to a variety of the failures
mentioned throughout this report. The lists are meant to act as a convenient blueprint that walks
through the individual steps without the discussions or illustrations found within the associated
sections of the report. The cases are meant to show how quickly the investigation process tends
to become focused when the subject is a specific type of failure.
11-1
10305538
Closing Remarks
Litigation
Just as recovery time is always lingering in the background of a root cause investigation, so is the
potential for litigation. Normally, with a successful resolution of the immediate problem,
arrangements are made that are enough to satisfy the operator and OEM. As indicated in the
aforementioned list, circumstances where litigation is likely to hold the OEM responsible are
those that are (a) directly related to the blade design, (b) related to the design or under-design of
a controlling system that compromised the normal steam environment, or (c) in the form of
operating criteria that allowed the blades to encounter forces that they were not designed to
handle.
In such circumstances, the set of guidelines that is applied is different from that contained in this
report. The techniques and methodology presented herein are meant to make a reasonable
assessment of the most likely source or root cause for a failure in a deliberate but expeditious
manner. The goal is to restore the unit to service as quickly as possible and to prevent repeat
occurrence of the original failure.
In litigation, the emphasis shifts to establishing with as much certainty as possible the veracity of
the diagnosis and predominant responsibility of the identified source or root cause. Because the
stresses and life consumption models are used to establish responsibility, the veracity of the
diagnosis is normally challenged by questioning the assumptions made at each step of a Phase 2
type of examination. This starts with questioning the validity of the material properties that were
taken from published sources rather than from samples of the actual blade material. Next, the
critical role of frequencies or temperatures is questioned. Frequencies are tricky to calculate
because of stress-stiffening and spin-softening effects as well as the way that boundary
conditions are assigned to represent interfaces within the assembly. Several hertz can mean the
difference between a well-detuned mode and a legitimate HCF problem. Temperatures rely on
extrapolations of inter-stage conditions defined from known parameters at the inlet and exit of
the respective turbine steam path. Dynamic stresses are ultimately based on presumed conditions
of stimulus and damping. As noted, only an in situ test of the machine at issue can definitively
establish the absolute response of the rows under the approximate operating conditions.
These issues are not raised to dissuade operators from pursuing litigation, particularly in
scenarios where repeated failures persist. They are meant to highlight the differences in both the
approach and level of effort that are involved. Unlike a root cause investigation, the amount
invested in analysis is often compounded by a factor of two to four in order to demonstrate that
parameters such as material properties, mechanical properties, fatigue properties, and tolerance
deviations do not seriously influence or change the evidence on which the final root cause is
assigned. In fact, the potential success for winning litigation is often directly commensurate to
the degree of sophistication apparent in an operator’s technical case. Unlike a root cause
investigation, the process is usually a lengthy one. In addition to building an initial case, because
pre-trial deposition raises new issues or areas of speculation, the analysis has to be revisited in
order to prepare a viable response.
11-2
10305538
12
REFERENCES
1. Survey of Steam Turbine Blade Failures. EPRI, Palo Alto, CA: 1985. CS-3891.
2. Turbine Steam Path Damage: Theory and Practice. EPRI, Palo Alto, CA: 1999. TR-108943-
VI and TR-108943-V2.
3. W. P. Sanders, Turbine Steam Path – Volumes 1 and 2: Maintenance & Repair; Volumes IIIa
and IIIb: Mechanical Design and Manufacture. PennWell, Tulsa, OK: 2004.
4. S. S. Manson and M. H. Hirschberg, “Fatigue Behavior in Strain Cycling in the Low and
Intermediate Threshold Range.” Proceedings of the 10th Sagamore Army Research
Conference, Syracuse University Press, Syracuse, NY (1964).
5. D. F. Socie and J. Morrow, “Review of Contemporary Approaches to Fatigue Damage
Analysis.” Risk and Failure Analysis for Improved Performance and Reliability. Plenum
Publishing Corporation, 1980.
6. Life Cycle Management Planning Sourcebooks, Volume 8: Main Turbine. EPRI, Palo Alto,
CA: 2004. 1009071.
7. “Military Standard System Safety Program Requirements.” United States Department of
Defense, Washington, D.C.: 1993. MIL-STD-882C.
8. W. Campbell, “The Protection of Steam Turbine Disk Wheels From Axial Vibration.”
American Society of Mechanical Engineers, New York, NY: 1924.
9. W. A. Burton and R. J. Ortolano, “Field Testing of Long Arc and Continuous Shrouding of
L-1 Turbine Blade at SCE.” Paper presented at the EPRI symposium Steam Turbine Blade
Reliability Seminar and Workshop, Los Angeles, CA (1986). CS-5085.
10. S. H. Hesler and J. P. Marshall, “Diagnosis and Correction of Recurring Failures in L-2 LP
Turbine Stage,” Proceedings of Steam and Combustion Turbine Blading Conference and
Workshop. Orlando, FL (1992). TR-102061.
11. Guidelines for Reducing the Time and Cost of Turbine-Generator Maintenance Overhauls
and Inspections, Volume 4: Turbine-Generator Component Procurement Specifications.
EPRI, Palo Alto, CA: 2007. 1014729.
12-1
10305538
10305538
A
CHECKLISTS
Presented in this appendix is a series of checklists that consolidate the questions raised in the
various sections associated with each phase and step of the recommended methodology. The
objective and purpose associated with each are also summarized. The intention is to provide a
concise summary of the guidance offered within each step to act as a blueprint or simple
reference during the course of an investigation.
Table A-1
Checklists Provided in Appendix A
Table Number Subject

Table A-2 Pre-Inspection Action Items
Table A-3 Immediate Action Items When Turbine Is First Opened

Table A-4 Visible Inspection of Damaged Row and Evidence Related to Installation
Table A-5 Basic Design Information
Table A-6 Items Specific to the Blade and Row
Table A-7 Relevant Operating Data to Assemble for Review
Table A-8 Blade Material Analysis
Table A-9 Fractographic Examination of Cracks
Table A-10 Evaluating Preliminary Evidence
Table A-11 Information Provided by the Operator to Support RCA
Table A-12 Information Related to Steady Stresses
Table A-13 Information Related to Frequencies and Dynamic Stresses
Table A-14 Information for Life Prediction and Probability of Failure
Table A-15 Identifying and Selecting a Corrective Strategy
A-1
10305538
Checklists
Table A-2
Checklist of Pre-Inspection Action Items
Phase 1: Pre-Inspection Action Items Refer to Section 4, “Pre-Inspection Action Items”
Specific Issues Objectives Comments
Was there any evidence or

early warning that might be A qualitative opinion is sought that might indicate when
1 If the failure causes a wreck of the unit, photographs
related to problems on the the problem started to develop.
blades? should be taken that document the location and
condition before any parts are removed. A standing
Is there any reason to rule should be enforced that any failed blade parts are
suspect that the failure was Did any conditions or circumstances associated with to be protected from contamination and not removed
2
a consequence of another the unit precede the failure? from the machine or the site without expressed
problem? permission of the manager. Initial efforts should be
undertaken to determine what spare blades are
Would the operating
available, particularly for the last two or three rows of
conditions at the time failure An opinion is sought as to whether any external
3 the LP turbine or the first rows of the HP/IP turbine.
occurred be considered conditions might have played a role in the failure.
typical or unusual?
Is the most current or The heat balance diagram will provide basic Any analysis will request that this information be
4 representative heat balance parameters used to establish the flows and provided. If there have been upgrades or the
diagram available? temperatures for the stage in question. conditions have changed, this should be established.
Are sample blades for any If damage is confined, available replacements might be Options to limit the outage or proceed independently
5 rows presently available for able to buy some time and/or should be reverse of the OEM’s timeline are severely limited if no
reverse engineering? engineered immediately. samples are secured before or after the inspection.
A-2
10305538
Checklists
Table A-3
Checklist of Immediate Action Items When Turbine Is First Opened
Refer to Section 4, “Immediate Action Items When

Phase 1: Immediate Action Once Turbine Is Opened
Turbine is Open”
Was there any
evidence or early
A qualitative opinion is sought that might indicate when the
1 warning that might be If the failure causes a wreck of the unit, photographs
problem started to develop.
related to problems on should be taken that document the location and
the blades? condition before any parts are removed. A standing
Is there any reason to rule should be enforced that any failed blade parts are
suspect that the failure Did any conditions or circumstances associated with the unit to be protected from contamination and not removed
2
was a consequence of precede the failure? from the machine or the site without expressed
another problem? permission of the manager. Initial efforts should be
Would the operating undertaken to determine what spare blades are
conditions at the time available, particularly for the last two or three rows of
An opinion is sought as to whether any external conditions the LP turbine or the first rows of the HP/IP turbine.
3 failure occurred be
might have played a role in the failure.
considered typical or
unusual?
Is the most current or
The heat balance diagram will provide basic parameters used Any analysis will request that this information be
representative heat
4 to establish the flows and temperatures for the stage in provided. If there have been upgrades or the
balance diagram
question. conditions have changed, this should be established.
available?
Are sample blades for

If damage is confined, available replacements might be able Options to limit the outage or proceed independently
any rows presently
5 to buy some time and/or should be reverse engineered of the OEM’s timeline are severely limited if no
available for reverse
immediately. samples are secured before or after the inspection.
engineering?
A-3
10305538
Checklists
Table A-4
Checklist of Visible Inspection of Damaged Row and Evidence Related to Installation
Refer to Section 4, “Visible Inspection of the Parts”

Phase 1: Inspection and Evidence Related to Installation
and “Evidence Related to Installation”
A determination is needed as to which row the

From which row did the initial
1 cracking was confined to or contained a fractured
failure appear to originate?
blade. An initial determination must be made as to which row
most likely failed first. If the consequential damage
When was the last time that Evidence is sought that otherwise might indicate a makes this difficult, proceed on the assumption that
2 the cracked region was shorter timeframe for the failure process than when the first to fail was the row furthest upstream. Unless
inspected? they were installed. definitive evidence can be produced from prior
inspection reports, it must be initially assumed that the
What was the general timeframe bounding the failure process extends to
A qualitative opinion is sought as to the relative
3 condition of the turbine blade when the parts were originally installed. A more
appearance of the blades.
materials? definitive role that erosion, corrosion, or other defects
Was any erosion concentrated Notching or wire cutting can create localized cracks, may have played will be obtained from the
4 or uniformly spread, and whereas uniform material loss might be more metallurgical analysis. Sought here is a general
where? tolerable. overview or impression of the as-found condition of the
rows that might reinforce the evidence found on a
Is there any visible evidence detailed examination of the fracture surface.
If visible distress is apparent, photos should be taken
5 of pitting, erosion, or corrosion
that typify the general state of the blades.
on any rows?
How uniform did regions

Inconsistency in the quality of welds or brazes can
Problems originating in peened covers might involve a
6 involving welds and brazing
affect the scatter in blade frequencies. failure for the peening to have been performed
appear?
correctly. In such instances, an assessment of the
How uniform were gaps on quality and uniformity should be attempted to assist in
For designs that depend on forming a continuous ring supporting or ruling out this factor. Management of
integral covers? Were there
7 at speed, gaps that do not close or remain open can gaps is critical to any design that intends to form or
gaps apparent in pre-loaded
cause frequency issues. maintain the row in an interlocking system at full
covers?
speed. Initial gaps too large will fail to close as the
Was their uniformity apparent Cracks forming in tenons and shrouds can be blades twist. If insufficient pre-loads are applied, the
8 in the quality of peening or influenced by the quality of the peening that was untwist will open the gap.
shaping any connections? performed.
A-4
10305538
Checklists
Table A-5
Checklist to Guide Identification of Basic Design Information
Refer to Section 4, “Basic Design Information

Phase 1: Basic Design Information
Required”
What is the length of the blade This information is generally used to identify
Most large steam turbines can be identified by
1 in the row where the corresponding units and blade designs that have a
the length of their last stage blade.
fatigue/fracture occurred? similar heritage.
What material are the blades This information is used to establish the basic material
2 Some basic information on the type of material,
manufactured from? and mechanical properties associated with the design.
number of blades in the row, and features of the
How many blades are installed Blade counts will be necessary as part of input supplied linking arrangements after the rows are
3 assembled will assist in determining how
on the row that failed? to support any third-party analysis.
relevant any prior reported events are to the
How the blades are grouped or linked is fundamental to specific design that failed, whether spares of the
How are the blades organized same type can be procured.
4 understanding how frequencies are meant to be
(linked) around the disk?
controlled.
A-5
10305538
Checklists
Table A-6
Checklist to Guide Identification of Items Specific to the Blade and Row
Refer to Sections 4, “Items Specific to the

Phase 1: Items Specific to Turbine, Blade, and Row Turbine(s)” and “Items Specific to the Blade
and Row Involved”
Does any visible evidence Problems associated with FOD, water hammer, and so FOD can change the objective of further analysis
1 suggest a non-fatigue-related on might be considered random events that do not from an investigation of root cause to an
mechanism? require analysis. assessment of remaining life if the blade is not
immediately removed, or to establishment of
Was fatigue or a fracture of one Evidence is sought indicating whether the mechanism is repair (blending/grinding) limits. General
2 evidence is sought that indicates whether a
or more blades clearly evident? purely fatigue-driven or could involve other issues.
feature of the unit configuration or flow
arrangement might have played a role and
Was the failure confined to one Apparent patterns suggest how flow arrangement, water
3 whether the problem is more generic to the
flow or one turbine? extraction, or the condenser system might play a role.
blade design itself. Problems related to steady-
state stress tend to be more pervasive and
If cracks are not caught before fracture occurred, it will widespread and favor the leading or trailing
How many, if any, blades
4 dictate the overall urgency (time available) to find a blades within a group. Dynamic (frequency-
actually failed catastrophically?
corrective solution. related) problems are more selective and tend to
be more random, reflecting the nominal
Where, specifically, on each Overall patterns provide clues as to whether the differences in frequencies. The number of cracks
5 disk/wheel were cracked or problem is related to steady-state or dynamic found is often evidence of how long the problem
failed blades found? characteristics of the row. has been developing.
Where did the fatigue or fracture The location on the blade provides an additional clue as LCF problems will naturally conform to locations
6 of maximum steady stress, and they are
generally occur on the blade? to what mechanisms might play a role.
particularly fond of the root attachments.
How many blades were found to The number of blades found damaged is an indicator of Failures or cracking above the platform point to
7 modes of vibration, particularly if there is
be cracked among each row? the duration or severity of the mechanism(s) at work.
widespread presence of pits or erosion damage
How many blades were found to within these regions. Sometimes, secondary
Patterns, particularly in grouped blades, can assist in problems appear when the original grouping is
8 be cracked in the same
revealing the mode responsible for HCF damage. compromised. The change in natural
location?
frequencies, often if a blade is left freestanding
as a consequence of tiewire or shroud failures,
Were any blades found cracked These can indicate a natural progression of the original results in a new resonant condition that was a
9
in different locations? problem or a second problem resulting from the first. consequence of the initial failure.
A-6
10305538
Checklists
Table A-7
Checklist of Relevant Operating Data to Assemble for Review
Phase 1: Relevant Operating Data Refer to Section 4, “Relevant Operating Data”

How many start-stop cycles did This information is a fundamental input in

1 the turbine row experience prior to determining whether LCF could be Sometimes, care must be taken when a fleet of similar
the identification of failure? responsible for the initiation of cracks. turbine rotors is involved and/or spare rotors are routinely
installed to facilitate outages. Estimates of duty cycling
This information is fundamental in (starts and hours) should be approximate and do not have
How many hours did the turbine
determining whether HCF, SCC, or CAF to be absolute. They will be correlated against life
2 row experience prior to the
could be responsible for the initiation of predictions to see whether they correspond.
identification of failure?
cracks.
Was there any significant event If SCC or CAF is involved, knowing the timing Consistency is sought between the metallurgical findings
3 that might have disrupted the of any contamination assists in estimating the and the operating history. SCC also grows at a deliberate
steam chemistry prior to failure? rate of subsequent fatigue. rate. Crack lengths can correspond to such events.
When was the row last inspected, Erosion/corrosion problems are driven by the If relevant to assess replacement timing, the rate of an
4 and was prior evidence of erosion particulates entrained in the steam. The rate erosion or corrosion problem can be estimated based on
or corrosion noted? is not readily predictable. the interval between inspections.
Any such events need to be identified in terms of their

How often did the unit approach This is a crucial factor in determining if the
number and estimated duration for each. A heat balance
5 or exceed the OEM limits of load failure was potentially caused by
for the unit will be needed to further interpret this
versus backpressure? asynchronous vibration (flutter).
information.
How often did the unit experience When modes are marginally detuned, if the These types of problems are general for units that can
6 any out-of-spec rotation from shaft speed is excessively changed, it can have their rotating speed fluctuate by more than ±1.0–
severe grid disturbances? shift a per-rev, causing resonance. 1.5 hertz.
Prior to failure, did the unit A failure involving torsional system These types of problems are more likely to occur on units
7 experience any particularly severe interaction would seek to find corroborating where original rotors are mixed with new designs, thereby
electric transients? evidence of any such events. changing the torsional natural frequencies.
A-7
10305538
Checklists
Table A-8
Checklist to Guide Blade Material Analysis
Refer to Section, Section 5, "Determination of

Phase 1: Chemistry and Material Analysis the Chemical Composition of the Component"
and "Metallographic Examination"
Did the blade material chemistry conform Often, materials are provided by several third-
Eliminate substandard material as a source or
1 to OEM and/or general industry party suppliers, within which the immediate
root cause of the blade failures.
standards? QA/QC responsibility resides.
Is there e Evidence is sought that would correspond to For any turbine that has been in service for a
2 while, some evidence of deposits is inevitably
vidence of inclusions or byproducts in the either a source of steam contamination or
material grain microstructure? improper processing of material. going to be present. It should also not be
assumed that deficiencies found in the chemical
Provide evidence that corresponds to or makeup and/or some of the basic material
Were any traces and/or deposits noted
3 corroborates possible contamination of steam properties automatically explain why a failure
on the surfaces of the blade? occurred.
chemistry.
Did tests of the material tensile strength Check the material's mechanical strength and Out-of-spec properties can contribute by
4 conform to OEM and/or general industry ability to carry load without rupturing. Over weakening the resistance to damage or fatigue
standards? temperature range. but still not necessarily prove to be the root
cause if the design stresses are borderline. If
Did tests of the material yield strength Normally represented as a 0.2% offset, check portions of the blades are exposed to localized
5 conform to OEM and/or general industry where stress versus strain is no longer heating caused by welding, brazing, or flame
standards? proportional. Over temperature range. hardening, evidence of the quality of the
workmanship and the control of the heat-affected
Did tests of the material’s ductility Check the material's ability to deform under zone is sought.
6 conform to OEM and/or general industry load without fracturing, normally done at room
standards? temperature.
Did tests of the material’s hardness
Check the material's ability to resist wear and
7 conform to OEM and/or general industry
indentation, normally at room temperature.
standards?
Did tests of the impact strength conform

Check the material's ability to withstand impact
8 to OEM and/or general industry
loads, normally at room temperature.
standards?
A-8
10305538
Checklists
Table A-9
Checklist for Fractographic Examination of Cracks
Refer to Section 5, Section 5, "Mechanical

Phase 1: Examination of Fracture Surfaces Properties Testing"
Where did the crack originate The origin and crack trajectory are
Eventually, the identified origin will be compared against the
1 relative to the orientation of the indications of the mode of vibration that
location of peak modal stresses to see which correspond.
blade airfoil? might have promoted fatigue damage.
Was the origin of the crack Did a distinctly large microscopic stress
2 apparent, and if so, was it of raiser (that is, a pit, notch, or inclusion
measurable size? machining mark) facilitate the initial crack?
Features obtained by scanning the origin with an electron
Was the immediate zone of microscope are sought to make a fundamental determination
Evidence is sought of a chemical process of whether it appears that the phase of initiation was primarily
cracking around the origin
3 that attacked the cohesion of the grain due to pure fatigue—that is, excessive steady or dynamic
transgranular or intergranular in
boundaries versus fatigue. stresses—or if there is any evidence that the process involved
nature?
corrosion, erosion, or embrittlement around the origin. Fatigue
Were oxides or corrosive Evidence is sought that will rule out or points the finger at aspects of the design. Corrosion or erosion
4 products detected on the crack include corrosion or erosion as possible raises questions regarding the operator's control over
origin or surface? causes or contributors. chemistry and moisture removal. Embrittlement or superficial
damage left in stress-sensitive locations are issues associated
Was there any evidence that Evidence is sought that suggests a with the QA/QC that was exercised during manufacture and
5 the material might have been localized heat-affected zone might have installation of the blades. Fretting is also a reflection of a
over-tempered or overheated? made the origin more vulnerable to fatigue. looseness between mating surfaces that should by design
have maintained a firm contact during operation.
Fretting suggests a degree of looseness
Is there any evidence of fretting
6 between contacting surfaces, possibly from
approximate to the origin?
inadequate pre-loads or interlocking.
If pitting was around the origin, If pitting initiated the crack, an estimate of The formation of cracks within a relatively broad region of the
7 what was the average the general size relative to the origin is airfoil is sometimes dictated by the presence of unusually
diameter/depth? sought. large pits randomly scattered in the stress field.
A-9
10305538
Checklists
Table A-9 (continued)

Checklist for Fractographic Examination of Cracks
Refer to Section 5Section 5, "Mechanical

Phase 1: Examination of Fracture Surfaces Properties Testing"
Is there a transition point The point where the initial crack reached
8 evident at some distance from critical size is a reflection of the dynamic
the origin? stress that was present.
Large steam turbine blades tend to have relatively small
critical crack sizes. This is often apparent as a transition in
Does the overall crack surface
Fatigue usually exhibits a distinctive visible fatigue beach marks. If high dynamic stresses are
9 appear typical of how the alloy
appearance on different blade alloys. involved, such as those produced by resonance, the portion
looks when exposed to fatigue?
associated with crack initiation might not be discernable
If the surface around the origin appears because it is confined to within the initial 10 mils (0.3 mm) of
Was there any evidence that
tarnished or discolored, this might indicate the origin. If the problem is associated with damage tolerance
10 the initial crack was present for or detuned stresses, the immediate 50–100 mils (1.3–2.5 mm)
an initial period of low stress and slow
a significant period? from the origin will often reflect the initially slow process of
growth or corrosion-related effects.
crack extension due to low dynamic stresses. If the crack is
Is there evidence of beach These features can indicate a step change unusually deep before rupture occurs, with regular beach
11 marks or striations along the in load. Staining suggests that the crack
marks apparent, this could suggest a transitory situation of
crack trajectory? was open long enough for oxides to form.
high dynamic stresses that are periodically driving the crack.
As cracks propagate, they change the Evidence is sought to correlate against the predicted
Did any point along the crack stiffness of the blade. Sometimes, ridges operating frequencies.
12 trajectory indicate a change in temporarily appear as resonance or crack
applied loads? direction changes related to changes in
loading or blade vibration mode.
A-10
10305538
Checklists
Table A-10
Checklist for Evaluating Preliminary Evidence
Phase 1: Weighing Preliminary Evidence Refer to Section 7
Failure may be considered premature

Why did this failure based on need to replace blades in less Scenario 1: Replacement in-kind is viable if forensic evidence
1 clearly demonstrates (1) some operating event that caused the
happen now? than 100,000 hours of continuous service
or 5,000 starts. operating stresses to exceed design limits, (2) a failure in QA/QC
that allowed substandard materials to be used, (3) obvious defects
Based on forensic evidence, decide failure or damage from foreign objects that got into the steam path or (4) a
What caused it to was either genuinely a random event, or failure in QA/QC associated with an aspect of the row’s installation.
2
happen? begin paring down the list of potential
causes.
Scenario 2: Replace with new design if investigation of similar
Decide if replacement in-kind of damaged failure is conclusive enough to accept an OEM’s proposed solution.
What can be done
3 blades or new design offered by OEM is a
immediately?
viable response or likely to fail again.
Scenario 3: Evidence is ambiguous with some contradictions.
What needs to be done to Determine if root cause of the failure has Operator advised to move forward and invest in the subsequent
4 prevent it from happening been investigated, identified and/or steps of detailed analysis.
again? proposed strategy has worked successfully.
A-11
10305538
Checklists
Table A-11
Checklist of Information Provided by the Operator to Support RCA
Refer to Sections 8, "Plant Support" and

PHASE 2: Geometry, Operation, and Installation
"Finite Element Modeling"
Basic Issues Objectives Comments
To help the analyst identify specific

What are the relevant blade material The properties for most blade alloys are generally
1 properties required as input to the finite
mechanical and fatigue properties? available in published sources.
element and life prediction models.
Help analyst determine if he or she has OEMs tend to reference designs as broader
What is the blade length, model number, worked on the specific design, whether it generic families. Any field experience relevant in a
2
and/or vintage (age) of the design? has been in service a while, and/or if any failure investigation is that associated with a
prior changes are known. specific (identical) design.
Sufficient information or a sample part for Generally, a limited amount of information is

What are the basic dimensions of the
reverse engineering is required to allow an required to supplement the dimensions or estimate
3 blade, including tiewires, shrouds,
independent FEM of the blade to be from a sample blade. If pieces from the failure are
snubbers, and so on?
constructed. available, these can be used. For shrouds,
information is needed on width, length, and
What is the blade count, grouping, outer To calculate blade-disk natural frequencies, thickness. For tiewires, the analyst needs the
4 tip diameter, and general profile of the a segment of the row from shaft centerline to diameter, lengths, and presence of splits. A
disk/rotor? blade tip must be constructed. disk/rotor profile can be represented from a cross-
sectional drawing. Disk attachments are a
Long blades might have an initial gap that reflection of blade root with applied tolerances.
For blades with integral shrouds, what is
5 falls within an allowable range. Short blades The outer tip diameter should be measured off
the specified contact at zero rpm? rotor.
often apply a pre-load to maintain contact.
A-12
10305538
Checklists
Table A-12
Checklist of Information Related to Steady Stresses
Refer to Section 8, "Calculating Steady-

Phase 2: Steady Stresses
State Stresses"
Preliminary check of FEM. The highest Steady stresses include both centrifugal and
What is the magnitude and distribution of
1 stress is expected to occur in the root steam-bending forces. Steady stress drives the
steady stresses in the blade?
attachments. mechanisms of LCF and SCC, affects the
blade operating frequencies, and provides a
See where maximum values occur and how mean stress in the estimate of HCF. Maximum
What is the magnitude of steady stress at the
2 localized they are in locations of design stress values should be summarized for
crack origin?
concentration. different regions of design. In regions where
localized yielding is predicted and at the origin
Identify how much yielding there is at the of any cracks, steady stresses should
Where do the mechanical stresses exceed eventually be adjusted to account for the
3 initiation site. Provide input to calculate
the yield strength of the material? relaxation that occurs. This will be used to
LCF, HCF, and/or SCC initiation life.
assess LCF initiation life. Thermal stresses
can generally be ignored in stages that operate
Make a preliminary determination as to at > 800ºF (427ºC). At full load, in combination
Are the thermal stresses high enough to have whether the source of cracking is possibly with the steady stress, they will drive the
4
caused significant creep damage? the result of a combination of high stress formation of voids that eventually link over time
and temperature. to form creep cracks.
A-13
10305538
Checklists
Table A-13
Checklist of Information for Frequencies and Dynamic Stresses
Refer to Sections 8, "Natural Frequencies and Mode

Phase 2: Frequencies and Dynamic Stress Shapes," "Calculating Dynamic Stress," and "Model
Verification by Frequency and Vibration Testing"
Provide a table of frequencies without the To calculate natural frequencies, a dynamic model of the
What are the blade natural 360º rotating blade row is extrapolated from the segment
1 influence of stress stiffening for possible
frequencies at zero rpm? model previously used to obtain steady stresses. The
comparison to test results.
blade-disk reflects two vibratory motions: the
Prepare Campbell and interference diagrams displacement of the blades is superimposed on the
What are the blade natural flexure of the disk nodal diameters. First, four
2 to show margin of detuning and potential
frequencies at full speed? fundamental blade mode families are referenced by the
resonance.
predominant direction of their motion relative to the
rotor/disk nodal diameter as the tangential, axial,
What mode shapes does the Identify the deflections associated with each torsional, and second bending modes. Calculated
3 blade-disk system produce at full family of modes and where they are likely to frequencies are graphically shown as Campbell and
speed? produce highest stress. interference diagrams.
Identify the frequency ratio (per-rev/natural Modes with margin less than 10 hertz from the nearest
What modes operate in a
4 frequency) to identify margin of detuning per-rev forcing are considered potentially resonant and
potentially resonant condition?
and/or potential sources of HCF. prime candidates for further examination.
If tested, how and what is the Provide a limited means to verify authenticity Frequency testing is considered a definitive check
5 range of scatter found between of a finite element single blade model, either because of the predominant role that the frequency ratio
individual blades? suspended or mounted in fixture. has in terms of affecting the synchronous dynamic
stresses that occur under normal operating conditions.
Provide a means to verify authenticity of the Whether modes are marginally detuned or not can
If tested at zero rpm assembled
finite element blade-disk model in an become a point of contention between the OEM and
6 on rotor, what are the frequencies
assembled state without influence of independent failure analysts. Frequency data are a
of the nodal diameter modes?
centrifugal or bending forces. means by which such arguments can be resolved.
Uncertainty associated with any calculated natural
If tested, how do stress stiffening Provide a means to verify authenticity of the frequency is typically estimated at 2–3%. In other words,
and spin softening affect finite element blade-disk model in an a correlation to within this degree of accuracy would be
7
frequencies at full-speed assembled state at speed in wheel box or in considered evidence of a well-modeled bladed disk.
operation? operating unit.
A-14
10305538
Checklists
Table A-13 (continued)

Checklist of Information for Frequencies and Dynamic Stresses
Refer to Sections 8, "Natural Frequencies and Mode

Phase 2: Frequencies and Dynamic Stress Shapes," "Calculating Dynamic Stress," and "Model
Verification by Frequency and Vibration Testing"
Identify the distribution or pattern of dynamic

How high could the dynamic Dynamic stresses are calculated with finite element
stress produced for the different mode
8 stress become if a condition of analysis for each mode, based on applied stimulus and
families and maximum that might potentially
resonance is presumed? damping, first in an assumed resonant condition and
occur.
then in a detuned condition. Damping is normally
assigned within a range of 0.2–0.5 critical, based on
Where does the maximum stress Identify the maximum potential dynamic features of the design. NPF is normally assumed to have
9 for each mode occur relative to stress produced by each mode corresponding a stimulus ratio of 10%. The first 10 per-rev harmonics
the crack origin? to the failure origin. can reasonably be assumed at a stimulus ratio of 2–5%.
To complement the resonant stress, detuned dynamic
stress is estimated by means of a formula that includes
How much would the dynamic Evaluate the impact of the predicted
the mean stress at the location of interest, the resonant
10 stress be reduced by detuning the frequency ratio in terms of reducing dynamic
dynamic stress, and an amplification factor.
mode of vibration? stress, with attention to location of cracking.
A-15
10305538
Checklists
Table A-14
Checklist of Information for Life Prediction and Probability of Failure
Refer to Section 8, "Life Prediction and

Phase 2: Life Prediction and Probability of Failure
Probability of Failure"
At the origin, how many start-stop Evaluate the potential role that LCF and start- If the predicted LCF is less than 10,000 starts, the
1 cycles will there be before a crack stop cycles might have played to initiate cracks steady stresses might be too high and need to be
might initiate due to LCF? at the origin. reduced by local changes to the design. If the
HCF life is on the order of 10E10 cycles or fewer,
At the origin, how many hours of Evaluate the potential role that HCF and the mode driving it might be the root cause and
2 operation will there be before a crack vibration of a specific mode might have played to the detuning needs to be improved. Serviced
might initiate due to HCF? initiate cracks at the origin. blades should be replaced with modified design.
If operated in a wet steam environment, Evaluate the potential for SCC to initiate cracks If metallurgical analysis establishes the presence
3 how many hours will there be before a at the origin as a consequence of the steady of these initial damage mechanisms, this is
crack might form due to SCC? stresses and operating hours. supported by the finding that it was not due to
LCF, HCF, or creep. Regardless of what initiates
a crack, should it continue to enlarge, vibratory
Presuming that a crack was formed by erosion, stress will eventually exceed the stress intensity
What size of defect can the design threshold along portions of the crack front. Once
4 corrosion, or fatigue, evaluate when vibratory
potentially tolerate at the crack origin? HCF takes full control, the process of further
stress would take control.
propagation to final rupture will accelerate (due to
the high frequencies at which blades vibrate). FM
Assess how the odds of a failure change over
What are the chances for a failure to is therefore used to complete the story by
5 accumulated hours or starts, if parameters are
occur as starts or hours accumulate? explaining how the blade failed and when this was
defined as ranges and randomly combined.
likely to have occurred relative to the damage
sustained by the design at the location of interest.
What are the chances for a crack to A Monte Carlo simulation is performed to
Assess how the odds of a failure change as an
6 propagate from HCF as it increases in reinforce the original findings and give an idea of
initial crack increases in size at the origin.
size? how likelihood increases with time.
A-16
10305538
Checklists
Table A-15
Checklist for Identifying and Selecting a Corrective Strategy
Phase 3: Identify and Select Corrective Strategy Refer to Section 9
Basic Issue Objectives Comments
Do the present rows of blades Execute a replacement strategy when modifications Any cracking initiated or controlled by excessive
1 need to be replaced, and replaced to the present design are considered necessary to stress requires changes in the design (geometry,
with a new design? prevent a repeat occurrence. arrangement, or materials). If the root cause is
related to unit operation or controls, changes
should be considered to minimize the
consequences. Effectiveness of changes is
Could changes in control systems
Execute a corrective strategy that might or might not based on comparison of stresses, frequencies,
2 or operating limits correct the
benefit from changes to the original blade design. and life for the modified versus original design.
identified root cause?
Can a repair salvage blades that Include limited field repairs as part of a strategy to
3 appear to be only superficially restore original blades to serviceable condition and Refurbishment would be localized
damaged? use until replaced with new. blending/grinding, with the possibility of keeping
the sets in service, supplemented by routine
inspections. Inspection criteria would be on
parametric studies to determine at what point
What additional inspection criteria Monitor erosion damage, relying on limits established damage might make the design susceptible to
4 might be used to manage the from analysis to signify when blades are at high risk HCF failure.
remaining life of serviced blades? of failure.
A-17
10305538
10305538
B
CASE STUDIES
A series of case studies is presented to show how the methodology was applied to a variety of
common blade failures. The cases are presented in the order shown in Table B-1.
Table B-1
Case Studies of Blade Failures
No. Mechanisms Subject of Investigation
1 HCF/resonance L-1 blade tiewire/airfoil failure

2 HCF/resonance L-2 shroud/shank failure
3 SCC/LCF L-0 blade root attachment cracking

4 HCF/flutter L-0 blade root failure
5 HCF/fretting L-0 blade integral cover tip failure
6 HCF/erosion L-0 blade airfoil notching

7 HCF/resonance IP 1 airfoil failure
8 HCF/fretting HP 1 root cracking
The cases were selected to highlight a variety of blades and operating scenarios, as well as a
variety of the mechanisms discussed within the report. Consistent with the practices outlined
within the report, each case proceeded through the entire three-phase methodology, with site
forensics, metallurgical examination, and design analysis systematically performed. Only the
most relevant features in each phase are presented in order to focus on those aspects that were
ultimately used to identify the different root causes and/or contributors. Some of the cases
involved catastrophic failures, others, the discovery of cracks during routine inspections. The
cases were selected to further illustrate scenarios where design modifications were and were not
required, as well as instances where interim strategies of repair or refurbishment were feasible
based on the identification of the root causes.
B-1
10305538
Case Studies
B.1 L-1 Blade Tiewire/Airfoil Failure
Phase 1: After less than six months of operation, a

catastrophic failure occurred in a 400-megawatt
(MW), 50-hertz steam turbine, resulting in a forced
outage. Inspection of the unit revealed a fractured
blade in the L-1 row, with several additional blades
also cracked. All cracks initiated at a three o’clock
location on the edge of the tiewire hole. The blades
were ~24 in. (610 mm) long and assembled in groups
of five by means of a riveted cover and floating
tiewire. The blade material was a 12% chrome
stainless steel. The fracture principally showed
evidence of HCF. Metallurgical analysis indicated
that it had a normal martensite structure. Examination
of the fracture found that cracks initiated at the corner
of the tiewire hole on the convex side of the blade
profile and propagated toward the concave side of the
profile. The observed beach marks and striations were
attributed to HCF, consistent with the short period of
operation. The fracture mode was exclusively
transgranular in appearance. No signs of fretting,
corrosion, or pits were found at the crack origin. No
oxidation deposits were found on the fracture surface.
Phase 2: Analysis of the design calculated that the

second bending mode resided only a few hertz from
the eighth per-rev harmonic of forcing. This result
indicated a high probability for operating in
resonance. In an assumed resonant condition, a
significant dynamic stress was produced in the
tiewire hole, which corresponded to the initiation
site found on the cracked and failed blades. The root
cause of the failure was ultimately attributed to
resonance of this mode, with the recommendation
that a detuning strategy be applied.
Phase 3: In accordance with the recommendation, a

replacement blade with an integral cover and mid-
span snubber that eliminated both the tiewire and
riveted cover was considered. Analysis of the
replacement indicated that the detuning problems
associated with the original design should be
corrected using the design with the more advanced
types of linkages at the mid-span and covers.
B-2
10305538
Case Studies
B.2 L-2 Shroud/Shank Failure
Phase 1: An L-2 row experienced a sequence of

failures spread over a period of 20 years on a 150-
MW turbine. The bucket was made from a 422
stainless steel with a tempered martensite
microstructure. The buckets were organized into
groups of four and five. Fracture of the bucket
shank was thought to be preceded by failure of the
tenon-shroud. Metallurgical examination of the
most recent failure of the blade shank/platform
revealed that cracks had initiated on the leading
edge side of the shank, starting within the cross-key
slot just below the deformed region created by the
cross-key staking. The fracture surface was typical
of HCF promoted by a low stress. The blue and
gray staining was the result of oxides formed on the
surface. The beach marks indicate a change in
operation of the machine and the extent of crack
growth at various points in time. Based on a review
of the unit operating data, it was concluded that they
reflected load variations in the months preceding
the catastrophic failure.
Phase 2: The source of crack initiation was verified

by analysis to be the result of HCF from resonant
vibration of the fundamental tangential mode. The
vibratory stress produced from this mode was also
verified as sufficient to promote HCF initiation at
the keyhole in the shank. Based on the calculated
dynamic stresses, an HCF life assessment indicated
that they were of sufficient magnitude to have
caused the trailing blade of the group to initiate a
crack and fail from HCF, even without a failure of
the cover occurring first.
Phase 3: Given the root design involved and the

challenges this imposed in terms of controlling the
natural frequencies and amount of detuning
available to the critical first mode, a long arc shroud
was recommended. Using this approach, the
original four and five blade groups were extended to
form groups of seven in order to detune the critical
tangential mode. The regrouping strategy was
guided by the results of the frequency analysis to
select a final configuration.
B-3
10305538
Case Studies
B.3 L-0 Blade Root Attachment Cracking
Phase 1: An eddy current test of an 850-MW unit

revealed that 67 out of 480 blades had developed cracks
on the pressure side in the top and middle hooks of
several L-0 rows consisting of blades ~33.5 in. (851
mm) long. Indications were found in the notches of all
three attachments, favoring the uppermost notch and the
trailing blade of the five blade groups. A review of the
operating records revealed that eight years prior to the
discovery of indications, the unit had changed from
baseload operation to an average of 30–50% of its rated
capacity. Metallurgical examination revealed that
multiple parallel cracks had developed along the bottom
of the root attachment notch, forming in the base of
corrosive pits, which were filled primarily with iron
oxides. Transgranular cleavage with some faint marks
indicated that the cracks started from the surface, rather
than propagating from the fillet corner of the root where
the maximum steady stress occurs. Initiation was
attributed to stress corrosion further assisted by LCF.
Phase 2: As part of the RCA, it was determined that

under periods of low load operation, the corresponding
change in inlet temperature could cause the Wilson line
(wet zone) to be shifted toward the L-0 stage, exposing
the inlet ends to an alternating wet-dry environment.
The corresponding distribution of steady stress among
the individual blade attachments formed into a group
corresponded well with the distribution of indications.
This was consistent with the metallurgical findings that
crack initiation was caused by stress corrosion assisted
by LCF and further propagated in a stable manner,
primarily by LCF.
Phase 3: Based on the root cause diagnosis, no design

modification was recommended. Instead, FM was used
with the stress analysis to estimate a maximum
allowable crack depth for comparison against the NDE
data. Based on the comparison, it was predicted that the
blades could remain in service for six more months
(until the next scheduled outage) if start-stop cycling
was minimized.
B-4
10305538
Case Studies
B.4 L-0 Blade Root Attachment Cracking from Asynchronous Vibration
Phase 1: After running uneventfully for 200,000

running hours and 600–700 starts, a catastrophic
fracture occurred on an L-0 blade operated in a 400-
MW unit. Upon inspection, cracks were found on
five other blades in addition to the failure. All of the
cracks initiated on the pressure side of the root just
below the platform corner. Of note from the
operating records was a recent three-month period
of low-load/high-backpressure operation. The
metallurgical examination revealed that the cracks
propagated toward the suction side, but no obvious
indications of HCF were evident in the first phase
of crack initiation and growth. Instead, a transition
phase of stable growth was attributed to SCC and
LCF that occurred over a period of many years until
the crack reached about 0.08 in. (2 mm). At this
point, the evidence indicated that HCF took control
of crack propagation.
Phase 2: Analysis indicated that the frequencies of

the blade were well detuned, ruling out the
possibility of resonance as the root cause. This was
consistent with the long period of operation
registered for the blades preceding the failure. The
magnitude of steady-state operating stresses in this
location concurred with the role LCF could have
played in initially growing the crack. FM then
determined that for the 0.08-in. (2-mm) transition
crack size evident in the origin point, a dynamic
stress of 4–5 ksi (28–34 MPa) was required for
HCF to take control. Because all of the modes were
well detuned, any dynamic stresses produced from
synchronous vibration were well below this
threshold. Using data supplied from the operating
records on the three-month period of low-
load/high-backpressure operation, further analysis
determined that the source of this dynamic stress
was most likely a condition of stalled flutter.
Phase 3: Based on the root cause diagnosis, no

modification to the blade was recommended. The
damaged blades were replaced in kind, and an
independent operating criteria chart was prepared to
minimize the chances of encountering the same type
of harmful event.
B-5
10305538
Case Studies
B.5 L-0 Blade Integral Cover Failure
Phase 1: After less than a year of service, a failure

occurred in a last stage blade in a 500-MW steam
turbine that is part of a combined cycle
arrangement. Inspection revealed that one cover had
failed catastrophically, triggering the outage.
Several cracked covers were randomly scattered
around the row. Evidence of overheating and wear
was noted on several of the contact surfaces. A
significant difference between initial gaps from
cover to cover was documented throughout the row.
A metallurgical examination reported that the
material was within its allowable specifications of
chemistry and heat treatment. The discoloration was
attributed to heavy frictional rubbing that led to
overheating and smearing of the hard-faced surface.
The intergranular fracture surface and fatigue marks
were indicative of HCF.
Phase 2: Analysis identified the fourth mode of

vibration to be potentially resonant. However, the
stress from this mode was insufficient in itself to
initiate a crack. Further studies indicated that if the
initial gap between adjacent blades at zero rpm was
too large, the resulting contact forces at operating
speed could become substantially reduced. At a
predicted threshold, this could allow the contact
surfaces to slip or vibrate because of the forces
present from the fourth mode. The resulting
slippage would result in overheating and fretting
damage, consistent with what was observed in the
metallurgical examination. FM further confirmed
that the same dynamic stress would make this
region susceptible to any such nominal damage.
Phase 3: Based on the results of the RCA,

specifications to establish a minimum allowable
tolerance limit to govern the inspection of gaps in
replacement and existing blades in sister units.
Further specifications were provided to improve the
detuning of the critical fourth mode and thereby
increase the tolerance to any further damage at this
critical feature of the design.
B-6
10305538
Case Studies
B.6 L-0 Blade Airfoil Notching
Phase 1: Within a four-year period of service,

concentrated WDE had formed a prominent notch
within each of the last stage blades operated on a
145-MW steam turbine. As a result of the erosion,
three tips had already separated, causing significant
damage to the condenser. It was determined that a
cost-effective field repair would be to operate the
rows for two more years if the remaining tips could
be removed. The primary concern was awareness
that such a retrofit would significantly change the
original blade-disk frequencies and result in more
fatigue failures.
Phase 2: First, analysis confirmed that there was a

high risk of failure for the blades in their present
condition. Next, it was examined whether a cut
could be made at a point above the notch so as not
to put a fundamental group mode into resonance.
The assessment was performed by establishing the
structural properties of the original, undamaged
blade design; comparing these to those predicted for
the damaged blades; and then considering three
retrofit strategies, for example, removing the tip at
three locations (blade heights). The purpose of this
approach was to understand the limitations of the
original blade design, suggest a cause of the eroded
blade tip failures, and identify an optimum repair.
Because the study represented critical path
downtime for the plant, the failure analysis was not
undertaken to the extent of a traditional root cause
investigation, but limited to a reasonable attempt to
ensure that the most obvious cause of erosion
notching was not masking some other failure
mechanism.
Phase 3: The parametric studies found that removal

of the tips exactly at the notch was not advisable
because this would create a resonant condition as a
result of the loss of mass. Instead, tip removal
slightly below the notch was ultimately
recommended as an expedient until the blades could
be replaced.
B-7
10305538
Case Studies
B.7 IP Airfoil Failure
Phase 1: Shown is an example of an IP blade failure

that occurred on a 650-MW steam turbine in a row
that had operated reliably for nearly 20 years.
Inspection of the row showed that the airfoil had
separated just above the platform. The fracture
principally showed evidence of HCF. An SEM
revealed that the crack started to propagate from the
suction side of the failed blade. No single origin
point was evident; instead, it was concluded that
normal deterioration of the surface over years of
constant operation resulted in the formation of
micro-crevices (25–35 mils [0.6–0.9 mm]), some of
which formed in regions of maximum dynamic
stress.
Phase 2: As expected, the steady stresses in the

location of fracture were insufficient to represent a
threat of initiation due to LCF. The frequency and
dynamic analysis identified two potentially resonant
modes. That could produce a dynamic stress on the
order of 4–5 ksi (28–34 MPa). FM confirmed that
nominal surface damage in the presence of a
resonant dynamic stress of 4–5 ksi (28–34 MPa).
Microstructural deterioration of the blade material
could cause this region to become susceptible to the
nominal dynamic stresses. Analysis indicated that
the damage tolerance limit in the region of
maximum stress was only about ~35 mils (0.9 mm)
deep. It was concluded that both initiation and
propagation were the result of a combination of the
environment and stress produced by resonance of
the fifth and thirteenth nodal diameter modes.
Phase 3: Based on the results of the RCA, a

refurbishment strategy was determined whereby
limited removal of mass from the cover could
improve the detuning of the critical modes. In
conjunction with this modification to the covers,
superficial damage accumulated over time in the
critical region of maximum dynamic stress was
ground and polished away.
B-8
10305538
Case Studies
B.8 HP 1 Root Cracking
Phase 1: As part of a scheduled refurbishment plan,

significant cracks were discovered in the root
attachment of the Row 5 blades operated in the
high-pressure turbine of a 600-MW unit. The cracks
occurred along the edge of the hook-bearing area.
They propagate along the fillet radii and later
change direction and proceed into the shank.
Inspection revealed that many rows in both the HP
and the IP rotors were found to have become loose,
that is, no pre-twist was evident. A metallurgical
examination of the fractures identified fretting
grooves approximate to the blade/disk contact
surfaces as the initiating mechanism. The fretting
was attributed to differential thermal expansion
between the rotor and the blade roots as a
consequence of insufficient fit. Evidence on the
crack surface suggested propagation was the result
of HCF.
Phase 2: The investigation focused on establishing

the root cause as a basis for guiding the rotor
refurbishment. Analysis showed that where the
crack initiated (approximate to the fillet of the T-
hook design) was not at the location of maximum
steady stress, indicating that LCF was not the
source of initiation. When installed, a pre-twist was
used to achieve a continuous interlock between
adjacent blades. Further analysis revealed that if this
required pre-twist was not achieved, certain
frequencies would be put in a resonant condition of
vibration. The increase in dynamic stress was
sufficient to propagate cracks from the origins
created by the fretting. Crack Origin
Crack Tip
Phase 3: A series of studies were subsequently
performed to produce replacement blades that
would remain in a detuned condition, even if the
pre-load condition were to be relieved over time.
Limited detuning was guided by the frequency
results, and applied using a skim cut to reduce the
thickness of the cover by a nominal amount.
B-9
10305538
10305538
10305538
Export Control Restrictions The Electric Power Research Institute (EPRI), with major
Access to and use of EPRI Intellectual Property is granted with the locations in Palo Alto, California; Charlotte, North Carolina; and
specific understanding and requirement that responsibility for ensur- Knoxville, Tennessee, was established in 1973 as an independent,
ing full compliance with all applicable U.S. and foreign export laws nonprofit center for public interest energy and environmental
and regulations is being undertaken by you and your company. This research. EPRI brings together members, participants, the Institute’s
includes an obligation to ensure that any individual receiving access scientists and engineers, and other leading experts to work
hereunder who is not a U.S. citizen or permanent U.S. resident is collaboratively on solutions to the challenges of electric power. These
permitted access under applicable U.S. and foreign export laws and solutions span nearly every area of electricity generation, delivery,
regulations. In the event you are uncertain whether you or your com- and use, including health, safety, and environment. EPRI’s members
pany may lawfully obtain access to this EPRI Intellectual Property, you represent over 90% of the electricity generated in the United States.
acknowledge that it is your obligation to consult with your company’s International participation represents nearly 15% of EPRI’s total
legal counsel to determine whether this access is lawful. Although research, development, and demonstration program.
EPRI may make available on a case-by-case basis an informal as-
Together...Shaping the Future of Electricity
sessment of the applicable U.S. export classification for specific EPRI
Intellectual Property, you and your company acknowledge that this
assessment is solely for informational purposes and not for reliance
purposes. You and your company acknowledge that it is still the ob-
ligation of you and your company to make your own assessment
of the applicable U.S. export classification and ensure compliance
accordingly. You and your company understand and acknowledge
your obligations to make a prompt report to EPRI and the appropriate
authorities regarding any access to or use of EPRI Intellectual Prop-
erty hereunder that may be in violation of applicable U.S. or foreign
export laws or regulations.
Program:
Steam Turbines, Generators, and Balance-of-Plant
© 2008 Electric Power Research Institute (EPRI), Inc. All rights reserved. Electric Power
Research Institute, EPRI, and TOGETHER...SHAPING THE FUTURE OF ELECTRICITY are
registered service marks of the Electric Power Research Institute, Inc.
Printed on recycled paper in the United States of America
1014137
Electric Power Research Institute

3420 Hillview Avenue, Palo Alto, California 94304-1338 • PO Box 10412, Palo Alto, California 94303-0813 USA
800.313.3774 • 650.855.2121 • [email protected] • www.epri.com
10305538

Steam Turbine Blade Failure Root Cause Analysis Guide: Place Image Here

Uploaded by

Copyright:

Available Formats

Steam Turbine Blade Failure Root Cause Analysis Guide: Place Image Here

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Steam Turbine Blade Failure Root Cause Analysis Guide: Place Image Here

Uploaded by

Copyright:

Available Formats

Steam Turbine Blade Failure Root Cause Analysis Guide

Place Image Here

Final Report, March 2008

EPRI Project Manager

ELECTRIC POWER RESEARCH INSTITUTE

(A) MAKES ANY WARRANTY OR REPRESENTATION WHATSOEVER, EXPRESS OR IMPLIED, (I)

(B) ASSUMES RESPONSIBILITY FOR ANY DAMAGES OR OTHER LIABILITY WHATSOEVER

ORGANIZATION(S) THAT PREPARED THIS DOCUMENT

Turbine Technology International, Inc.

This report was prepared by

Turbine Technology International, Inc.

Results and Findings

Challenges and Objectives

Applications, Value, and Use

Abbreviations and Acronyms

corrosion-assisted fatigue (CAF). A damage mechanism referring to the formation of cracks

dynamic stress. Consequence of the blade-disk system response to synchronous or

permanent solution. A corrective action in the form of repair, replacement, or a refurbishment

1 INTRODUCTION TO ROOT CAUSE ANALYSIS ..................................................................1-1

2 DAMAGE MECHANISMS AND THEIR ROOT CAUSES.......................................................2-1

4 IMMEDIATE RESPONSE AND ON-SITE FORENSICS.........................................................4-1

5 OFF-SITE FORENSICS: METALLURGICAL EXAMINATION ..............................................5-1

7 WEIGHING PHASE 1 EVIDENCE..........................................................................................7-1

8 PHASE 2: ENGINEERING ANALYSIS ..................................................................................8-1

9 PHASE 3: IDENTIFYING A REPLACEMENT STRATEGY ...................................................9-1

10 POST-FAILURE GUIDANCE .............................................................................................10-1

11 CLOSING REMARKS ........................................................................................................11-1

A CHECKLISTS ....................................................................................................................... A-1

Table 1-1 Basic Organization of This Report .............................................................................1-2

Organization of This Report

Methodology and Objectives. A standardized methodology of three phases provides a roadmap

Generic Data. Discusses how to decide if industrywide or published experience on a blade

General guidance is presented on how to decide whether to execute a permanent or interim

First Response: Recovery Versus Investigation

Immediate Recovery Scenarios

Moving from an Interim to a Permanent Solution

Cost and Risk of Repeat Failures

Item Forced Scheduled Total

Summarizing the Objective for an RCA

Cause Elimination Process

Turbine Stage Principal Mechanism Root Cause Ref 1 Ref 2

Table 2-2 (continued)

Turbine Stage Other Mechanisms Root Cause Ref 2 Ref 3

In the next section, the following items are noted:

LCF Caused by Start-Stop Cycling

EPRI References on LCF

Relationship of the Evidence

Evidence of regular beach marks

Similar cracks on multiple blades in

Cracks formed in regions of lower

Stress corrosion and creep cracking

Blades are not operated in a wet or

Design Factors and Issues

HCF Caused by Blade Vibration

EPRI References on HCF

Design Factors and Issues

HCF Caused by Synchronous Vibration

Resonance describes a condition of maximum response and maximum dynamic stress. It is

Rapid crack growth leading to failure in