Life Management and Maintenance For Power Plants Vol. 2010

Keywords: thermal power plant, nuclear plant,
VTT SYMPOSIUM 265 boiler, turbine, condition, life, fuel, maintenance,

inspection, monitoring, risk, reliability, material,
damage, corrosion, fatigue, creep, degradation
Baltica VIII
Life Management and Maintenance
for Power Plants
Vol. 2
Helsinki–Stockholm–Helsinki
18–20 May, 2010
Edited by Pertti Auerkari & Juha Veivo
Organised by
VTT
ISBN 978-951-38-7593-2 (soft back ed.)
ISSN 0357–9387 (soft back ed.)
ISBN 978-951-38-7594-7 (URL: http://www.vtt.fi/publications/index.jsp)
ISSN 1455–0873 (URL: http://www.vtt.fi/publications/index.jsp)
Copyright © VTT 2010
JULKAISIJA – UTGIVARE – PUBLISHER
VTT, Vuorimiehentie 5, PL 1000, 02044 VTT

puh. vaihde 020 722 111, faksi 020 722 4374
VTT, Bergsmansvägen 5, PB 1000, 02044 VTT

tel. växel 020 722 111, fax 020 722 4374
VTT Technical Research Centre of Finland

Vuorimiehentie 5, P.O. Box 1000, FI-02044 VTT, Finland
phone internat. +358 20 722 111, fax + 358 20 722 4374
Edita Prima Oy, Helsinki 2010

Preface
The Baltica Conferences have since 1988 witnessed a continuous change in the
environment for the technology and business of power plants. Apart from the
shift towards large actors and more unified practices, also grids and businesses
are increasingly connected by the developments in the political landscape,
technology, and the ability of people to contact each other anywhere. Meanwhile,
some aspects may simply come around: the first Baltica Conference took place
two years after Chernobyl, and new nuclear seems to be rising again. One driver
of change is global warming, easy or not it is to remember after a hard winter in
Europe and a decade of steady world mean temperature.
450 0,6
IPCC 2007 Now
400
World mean
0,4
350
CO2 (ppm)
0,2
ΔT (°C)
300
250 0,0
200
-0,2
150
-400000 -300000 -200000 -100000 0
-0,4
Years ago
1880 1920 1960 2000
To start an ice age – something not everybody want – requires in air less than
300 ppm CO2. Hydro, wind, other renewables and nuclear can push, but the
world is not easily or quickly weaned from coal, oil and gas, or taught to apply
carbon capture. The current level is 390 ppm CO2 and rising, and the
countermeasures provide a background for BALTICA VIII.
The editors wish to thank all authors, referees, organisers and the Board of the
Conference for their invaluable help in preparing for the event and the
proceedings. Financial and other contributions by the supporting and sponsoring
organisations are also gratefully acknowledged.
Espoo, 18 May 2010
Pertti Auerkari & Juha Veivo
Editors
3
Contents
Preface 3
Session 3. Combustion plant – Biomass & waste boilers
Benchmarking for plant maintenance 6
Kari Komonen, Toni Ahonen and Susanna Kunttu
Managing corrosion in biomass boilers: benefits and limitations of coatings 22
Satu Tuurna, Tommi Varis, Kimmo Ruusuvuori, Stefan Holmström,
Jorma Salonen, Pertti Auerkari, Tuomo Kinnunen, Patrik Yrjas, Risto Finne,
Matti Nupponen, Ulla McNiven, Hannu Ahonen and Ari Kapulainen
Selecting the right material for recovery boiler superheaters 37
Esa K. Vakkilainen and Pekka Pohjanne
Oxyfuel combustion: Oxidation performance of steels in simulated oxyfuel
conditions 52
Satu Tuurna, Pekka Pohjanne, Sanni Yli-Olli, Tuomo Kinnunen
and Petra Jauhiainen
Session 5. Combustion plant – Fossil & co-fired boilers
Safety issues arising from the corrosion-fatigue of waterwall tubes 63
John M. Brear, Paul Jarvis and Stephen Scully
Life assessment for vintage boilers 78
Pertti Auerkari, Jorma Salonen, Stefan Holmström, Leila Laaksonen, Ulla
McNiven, Olli Lehtinen, Sari Mäkinen, Ville Väänänen and Reino Nikkarila
Session 7. Combustion plant – High temperature components
Italian standardization in the field of defect evaluation, fracture mechanics and
fitness for service of pressure equipment 90
Corrado Delle Site, Carlo Fossati, Giuseppe Giannelli, Claudio Sampietri
and Andrea Tonti
Defect integrity in components subjected to cyclic loading at high temperatures 104
Stuart Holdsworth
Life assessment and extension for gas turbine blades 117
Pertti Auerkari, Jorma Salonen, Juhani Rantala, Sari Mäkinen,
Ilkka Karvonen, Heikki Tanttari, Pekka Kangas and Jukka Taattola
Fire incidents in underground coal storage 128
Juha Sipilä and Pertti Auerkari
4
Development and Implementation of i-LAS (indentation based turbine
remaining life assessment system) 143
Woosung Choi, Geewook Song and Jaeraeyang Koo
Creep damage and long term life of steam line components: case X20 154
Stefan Holmström, Pertti Auerkari, Jorma Salonen, Anssi Laukkanen,
Juhani Rantala, Reino Nikkarila and Jarmo Junttanen
Remaining lifetime assessment of power plant steam boilers 167
Vaclav Liska and Vaclav Mentl
Creep strain, damage and life prediction for welded 0.5CMV steel 176
Juhani Rantala, Pertti Auerkari, Stefan Holmström, Jorma Salonen and
Anssi Laukkanen
Different types of cracking of P91 steel weld joints after long-term creep tests 193
Dagmar Jandová, Josef Kasl and Eva Chvostová
Session 10. Cross-cutting issues – New methods & tools
New mechanical testing equipment for testing in real and simulated service
environments (HIPS) 209
Pekka Moilanen and Stefan Holmström
Characterization of true stress-strain behavior using optical monitoring 221
Tom Andersson, Lauri Elers, Anssi Laukkanen and Pekka Nevasma
Non-contact measurement of oxide thickness with a laser ultrasonic technique 233
Jonne Haapalainen, Joona Eskelinen, Tarja Jäppinen, Stefan Sandlin and
Edward Hæggström
On-line flaw growth monitoring in high temperature plant 239
Ian Atkinson, Stephen Kelly, Philip Lay and Colin Brett
Experience on the periodic wet bar diagnosis of water-cooled generator
stator windings 253
Hee-Soo Kim and Yong-Chae Bae
A breakthrough in ultrasonic detection and sizing of partially closed cracks? 270
Stefan Sandlin
Development of nondestructive techniques for fuel cladding evaluation
in hot cell 294
Tarja Jäppinen and Stefan Sandlin
5
Benchmarking for plant maintenance
Kari Komonen, Toni Ahonen and Susanna Kunttu

Espoo, Finland
Abstract
The product of the project, e-Famemain, is a new kind of tool for benchmarking,
which is based on many years’ research efforts within Finnish industry. It helps
to evaluate plants’ performance in operations and maintenance by making
industrial plants comparable with the aid of statistical methods. The system is
updated continually and automatically. It carries out automatically multivariate
statistical analysis when data is entered into system, and many other statistical
operations. Many studies within Finnish industry during the last ten years have
revealed clear causalities between various performance indicators. In addition,
these causalities should be taken into account when utilising benchmarking or
forecasting indicator values e.g. for new investments. The benchmarking system
consists of five sections: data input section, positioning section, locating differences
section, best practices and planning section and finally statistical tables.
1. Introduction
The significance of the availability of the production equipment, from the
competitive advantage point of view, has increased considerably during the last
years. The stronger emphasis on the safety and environmental matters has also
increased the importance of the maintenance function. In many industrial
branches, the significance of the maintenance function has grown and it has
become a factor, which makes the difference. The availability of production
equipment is very often less than 90% and OEE (overall equipment efficiency)
less than 80%.
6
According to the management of many industrial enterprises ‘benchmarking’

is a powerful tool for the development of the plant operations. However, among
many problems concerning benchmarking a challenge is to make companies or
plants comparable. This challenge is sometimes used as an excuse to explain
away the worse ranking in comparison. In this paper, we describe the system of
benchmarking, which aims to make various production units comparable with
the aid of statistical methods based on many years’ theoretical and empirical
research. The positioning of the production systems is one of instruments
introduced in this paper, and the other tools of the system lean on the results of
the positioning tool.
The success of the maintenance function can evaluated (simplifying) with the
aid three factors: the performance of production function, cost-effectiveness of
the maintenance function and the quality of maintenance processes. The product
of the maintenance activities can be measured e.g. with the aid of overall
equipment effectiveness (OEE), availability, lost production due to maintenance
or reliability of the machinery. Typical measures for the cost-effectiveness of the
maintenance function are ‘maintenance costs in relation to plant replacement
value’, ‘maintenance costs in relation to production volume’ or ‘maintenance
costs in relation to production capacity’. The quality of the maintenance
processes can be evaluated with such the indicators as planning rate, customer
satisfaction, employee satisfaction, accident rate or hazard rate etc.
The objectives of this research and development project have required a lot of
pre-work:
• Standardisation of concepts and key figures needed in benchmarking
• Data collection
• Research work in the area of causalities between various key
performance indicators
• Methods to identify the best plants
• Research to specify relationships between the best practices and performance
indicators
• Methods to transform the research results into the properties of this
benchmarking system
• Design of system architecture and programming of the web-tool
(Komonen 2006).
7
2. Web-tool for benchmarking

The benchmarking system consists of four sections or tools (Komonen 2005).
1. ‘Data input section’ is an electronic form to insert the data of the plant in
the database. An incorporated test tool will give the user a notice in the case
of the most probable mistakes.
2. In the ‘positioning section’ there is an option to evaluate the success of the
plant’s maintenance function branch-wise (by industrial sector) with the aid
of several indicators such as OEE, availability of machinery, production
losses due to maintenance, maintenance costs as a percentage of estimated
plant replacement value and many other.
3. In the ‘locating differences section’ it is possible to compare the plant’s
mode of operations with the successful plants, with the less successful
plants and with the average plant in the industrial sector in question.
Comparison can be made with the aid of several economic, technical and
organisational indicators.
4. In the ‘best practices and planning section’ the user of the benchmarking
system may find more hints of, which kind of modes of operations lead to
successful results. The structure of the tool is illustrated in Figure 1.
5. Statistical tables is a section where floating averages for all key
performance indicators are presented.
The system is automatic and runs all the automatic calculations getting
commands from the user-code and password entered in the system. All the
needed calculations are carried out when changes occur or the user commands to
execute any task. Simulations are also possible options, for example, for
planning purposes. The determination of the variables, which should be taken
into account when the system executes automatic multivariate analyses for the
positioning, is an expert’s job. The same applies to the determination of the
impact (direction and strength) of the modes of operations on the selected
objectives. The users of the system are able to see their own data only and
statistical key figures concerning the industrial branch in question.
8
Testing
Electronic form module Database
Continual statistical Determination Statistical

Positioning estimation of positioning of exogenous analysis,
equations branch-wise variables Expert work
Automatic grouping
into succussful
and less succussful
cases
Comparison with the mean
values of all the plants
Locating Continual
differences Comparison with the mean automatic
values of the best plants calculation
Comparison with the mean
values of the worst plants
Pareto grouping of Statistical

Best practices normalised differences and analysis,
and planning direction and strengh Expert work
of impact
Figure 1. Schematic description of benchmarking system.
3. Features of benchmarking system
3.1 Data input
‘Data input section’ serves the user an electronic form to insert data concerning
the plant’s maintenance function. An incorporated test tool will give the user a
notice in the case of the most probable mistakes. Data entered by the user is
precondition for the broader use of this benchmarking system. A part of the data
input sheet is illustrated in Figure 2.
All the entered data is confidential and no-one but the organisation which has
entered the data in question can see it. The users of the system are able to
compare branch-wise their own performance with the total figures and with the
figures concerning their own country.
The successful use of this benchmarking system demands a high hit-rate of the
most essential data. When this data has not been entered into the system, the
program informs of missing data. The system informs the users also in the case
of incorrect data input.
9
For each indicator, there exists also the info-box, which gives additional
information concerning the indicator in question. The users can see additional
information, when pointing the word “info” with the cursor arrow. On the left of
the input boxes, the plants can see the values, they have earlier entered into the
database. It is possible to interrupt data input when needed and continue it later.
Figure 2. The illustration of the data input page.
3.2 Positioning
With the aid of the positioning results the sample of plants in the benchmarking
data-base can be split down into two groups: successful plants and unsuccessful
plants. These two groups can be used when trying to identify causes for
successful or less successful performance in the case of the individual plant.
In the ‘positioning section’ the users are able to evaluate the success of their
maintenance function branch-wise (by industrial sector) with the aid of several
indicators such as availability of machinery, production losses due to maintenance,
maintenance costs as a % of estimated plant replacement value and many other.
10
The above mentioned indicators are affected by several external influencing

factors which are beyond the scope of the plant management or maintenance
managers to determine. But still they have a considerable impact on the performance
indicators. Examples of exogenous influencing factors are process severity, plant
size, utilisation rate, age of the plant, technology. These factors are taken into
account when evaluating one’s success with the aid of multivariate statistical
models. The conclusion page of the positioning is illustrated in Figure 2.
Figure 3. The conclusion page of the positioning.
As stated earlier in the positioning section, various plants are made comparable
with the aid of exogenous factors. These factors should be taken into account
when evaluating one’s success. The benchmarking tool will take care of this
analysis automatically for the user. The developed software calls the plants’ data
from the database, takes into account influencing factors and calculates for the
users an expected, predicted value for each performance indicator the user has
chosen. The users can also make simulations, how things would change if they
were able to change factors which are or are not under their own control. This
benchmarking software also groups all plants into successful and less successful
subgroups (Komonen 2006).
11
The conclusion page tells the users their position, their success, measured with
the aid of the performance indicator they have chosen. Their position has been
shown graphically and numerically. In the graphic presentation the result of the
positioning is marked with the aid of a green plot. In the horizontal co-ordinate
predicted values are shown. These values take into account the impact of
external and internal influencing factors (exogenous factors). In the vertical co-
ordinate, the actual values of the plant are shown. The 95% confidence limits are
plotted with red lines. The confidence limits give the plant in question
information whether expressed deviation is significant or not. The position of the
green plot marking plant’s performance can be changed with the aid of the
simulating tool introduced in the previous page.
3.3 Performance indicators
The simplified system of the key performance indicators can be classified in a

hierarchical manner. The following structure is the modified version of those
ones presented earlier by Komonen (e.g. 2002a).
(1) business objectives
(2) first level objectives of production
(3) second level objectives of production
(4) objectives for maintenance
(5) exogenous factors for production and maintenance (external factors)
(6) intermediate internal objectives for maintenance (follow-up variables)
(7) action variables of maintenance function (means for improvement)
(8) internal descriptive (explanatory) variables.
Some of the variables demand more explanation. The maintenance objectives
are corresponding metrics for the performance of the maintenance function: e.g.
availability of machinery due to maintenance, sum of unavailability costs and
maintenance costs (“dependability costs”), maintenance costs as a % of estimated
plant replacement value (PRV)). Exogenous independent factors are indicators,
which help management to evaluate the state of technology or business environment
and at the same time are beyond the scope of production or maintenance
managers to determine, but which have a considerable impact on the other
objectives. Examples of these kinds of factors are the utilisation rate of the
production machinery, production volume, the amount of the production equipment,
the integration level of the production process etc. Intermediate internal
12
objectives such as MTTF or MTTR are performance indicators, which should be

measured and followed up, since they may give more information about the
development needs. These variables are objectives, because production or
maintenance managers cannot influence them directly. The action variables of
the maintenance function are tools of production or maintenance managers,
with the aid of which objectives are achieved. Preventive and improvement
maintenance, outsourcing, operator maintenance etc. are good examples of these
tools (e.g. Komonen 2002, 2002b).
Exogenous independent variables have the major role in the benchmarking
system, because they offer an option to make various plants comparable, are
basis for the positioning of plants and give a tool to divide the plants in the
sample into the groups of successful and less successful production units.
Therefore studies concerning causalities between the objective performance
indicators and exogenous factors are valuable. The aim is to isolate two sources of
variation in the objective indicators: exogenous factors and modes of operations.
As mentioned above different kinds of production units or plants can be made
comparable with the aid of exogenous factors. Thus, for example, objectives
such as ‘overall equipment effectiveness’ (OEE) or availability are dependent on
the external factors, which cannot be influenced in the shorter time span.
In the case of maintenance costs in relation to equipment replacement value
exogenous factors may be the following:
• Integration level of production system (which has, in practice, often
parallel impacts to maintenance costs as replacement value) (-)
• scale (replacement value of production equipment ) (-)
• production volume / production equipment (replacement value) (+)
• shift work rate, operating rate (+)
• Industry –dummy variables (industry specific factors) (-, +).
In the case of maintenance costs in relation to capacity of the production line in
question, the effect of exogenous factors may be the following:
• replacement value (-)
• capacity (-)
• capacity / replacement value (-)
• integration level (-)
• production volume (+).
13
Also other key performance indicators seem to be a function of exogenous

factors. For example, ‘breakdown hours in relation to plant size’ is a function of
the plant size. There is a straight scale effect present. This phenomenon is
depicted in Figure 4. The causality between plant size and a number of
breakdown hours is nonlinear and therefore we have used log transformation.
We could present more empirical evidence for this kind of causalities, but in this
paper we have to limit evidence in two examples.
log (breakdown hours / plant replacment value)
Log (Plant replacement value)
Figure 4. Example of causalities between key performance indicators: ‘Breakdown hours

in relation to plant size’ is a function of the plant size.
In the case of the food industry as an empirical evidence, multivariate regression

analysis was carried out in order to find out the impact of exogenous factors. It
was interesting to notice that the variables having the most significant impact on
the relative cost level were plant size and degree of integration (Table 1). The
determining power of model was moderate (0,542), which is not bad taking into
account the heterogeneity of the branch (Komonen 2006b).
14
Table 1. The variables having an impact on relative maintenance costs in the food industry.
Processing of food and liquid (N = 34)

Dependent variable: Production equipment, Maintenance costs / Replacement value
Adjusted R² = ,542 F(4,29) = 10,773 p < ,00002
t(29) p-level
Intercept 3,865 0,0006
Replacement value -4,386 0,00014
Degree of integration -2,275 0,030
Plant turnover / plant replacement value 1,859 0,073
Operating rate 1,331 0,193
The causal relationship between the relative maintenance costs and estimated
variables is demonstrated in the scatter diagram (Figure 5), where predicted and
observed values equal on the diagonal. Cases above the diagonal represent worse
performance than expected (Komonen 2006).
Processing of food and liquids

Dependent variable:
Production equipment: Maintenance cost / Replacement value
.
.
.
Observed Values
Predicted Values
. 95% confidence
Figure 5. The predicting power of independent variables on relative maintenance costs in

the processing of food and liquid.
15
3.4 Locating differences
Applying positioning to all the cases in the sample it is possible to compare

industry-wise the plant’s mode of operations with the average plant or with the
successful plants or with the less successful plants. Comparison can be made
with the aid of several economic, technical and organisational indicators.
Examples of the modes of operations are the amount of operator maintenance,
preventive maintenance, contracting and improvements or the use of criticality
analyses. Following the techniques introduced above it is possible to find some
hints of which kind of modes of operations lead to successful results. Since the
proper way to improvements depends on the plant’s present positioning results
and present mode of operations. The conclusion page of “locating differences
section” is illustrated in Figure 6.
The objective of this analysis is to find probable causes for deviations from
the expected (predicted) values in the positioning section. The benchmarking
tool will take care of this analysis automatically for the user. The software calls
all the needed data from data base and carries out positioning for each case and
in this way groups all the cases into successful and less successful subgroups.
Finally, the tool calculates means and deviations for the groups of all cases,
successful plants and less successful plants and for each performance indicator
the user has chosen. Plants’ own actual values for the selected indicator is marked
with a green plot. The user can also make simulations, how things would change
if he were able to change factors which are or are not under his own control. As
soon as the differences in modes of operations have been identified it is possible
to investigate which modes of operations are effective and which are not.
16
Figure 6. The conclusion page of locating differences.
3.5 Best practices and planning
In ‘best practices and planning section’ the user can find some hints of, which
kind of modes of operations lead to successful results. Since the proper way to
improvement depends on the user’s present results and present mode of operations
this section includes also management tool to help in the planning process.
The user is able to compare the plant’s mode of operations with the successful
plants. The comparison is the summary of the 10 largest deviations in the form
of Pareto-diagram and the comparison can be carried out with the aid of several
economic, technical and organisational indicators. The objective of this analysis
is to find the course of actions which should be carried out in order to improve the
performance of objective variables (e.g. OEE, availability or cost effectiveness).
Deviations can be positive or negative values.
Since the statistical performance of the successful companies is based on the
results of positioning, those results are presented here once again in order to
make interpretation easier. Thus, the simulation tool is also available in this
section. The differences between the mean values of the most successful plants
17
and the users’ own actual indicator values have been normalised. The graphical
conclusions have been presented in Figure 7. The effect of the indicators in
question on the dependent variables (e.g. OEE, availability or cost effectiveness)
has been indicated in the same table (direction and strength). The causalities
between independent and dependent variables have been studied with the aid of
the sample entered into data base. Clear causalities could not always be
identified (Komonen 2006).
Figure 7. The first conclusion page of best practices and planning section.
The strength of impact has been presented with the aid of five levels: no impact,
small influence, medium influence, significant influence and very significant
influence. The direction of the effect denotes what happens to an objective value
(e.g. availability) if the numerical value of an independent indicator (indicator
which depicts the mode of operations) increases.
In the following page there is the summary of the largest deviations and their
influence on the performance indicators presented in 2 × 2 table (Figure 8). The
greatest attention should be focused on the factors, which indicates the largest
deviations and strongest impact. Further, less attention should be focused on the
18
factors having smaller deviations and smaller impact. If e.g. availability and
maintenance costs are studied simultaneously, actions needed may be conflicting. In
the case focus should be directed to those activities which have positive impact
on both the factors (Komonen 2006).
Figure 8. The second conclusion page of best practices and planning section.
4. Summary
Several studies during the last 17 years in Finnish industry have revealed clear
causalities between certain performance objectives and exogenous variables (e.g.
Komonen 1998). These causalities are valid in industrial maintenance, power
plants as well as, in many cases, in logistics and operations. In addition, these
causalities should be taken into account when utilising benchmarking or
forecasting. The introduced web-based benchmarking tool is based on the above
mentioned research activities concerning causalities between objective variables
and exogenous influencing factors and on the research concerning causalities
between the successful plants (determined with the aid of the positioning) and
19
modes of operations determined by action variables such as preventive maintenance

or training of personnel. The results of the first mentioned causalities (objectives
and exogenous factors) have been clear and the samples have been sufficiently large.
The needed actions to improve the benchmarking tool are identified. However,
the most probable obstacle to improve the characteristics of the system is the
lack data. The total number of the plants is surely large enough, but there is a
lack of cases for the other than the most common indicators in a specific sector
of the industry. In order to draw correct conclusions when using those indicators,
more research concerning causal relationships between objectives and
exogenous factors (and also action variables) should be carried out.
Regardless of the above mentioned problems some clear results have been
achieved. In several industrial branches, the means which improve the product of
maintenance activities such as OEE or availability increase the relative costs of
the maintenance function and vice versa. Sometimes the results are opposite to
expectations. Therefore, the optimisation of the improvement efforts is crucial.
The quality of data and sufficient data are also crucial issues when improving
benchmarking tools. In empirical research there is always trade off between the
detailed data and the number of cases in sample. The ability of the companies to
deliver data for a large number of sophisticated indicators is limited.
The introduced benchmarking system is quantitative in nature. There is surely
a need to add qualitative features to the system. However, as soon as the scope
of the tool is broadened to the qualitative field there is a need to change research
strategies, and that means challenges in the architecture of the benchmarking
tool and also problems with plants’ willingness to deliver data.
References
Komonen, K. 1998. The structure and effectiveness of industrial maintenance. Espoo,
Helsinki University of Technology, Department of Industrial Economics, Ph.D.
Dissertation. (In Finnish).
Komonen, K. 2002. A cost model of industrial maintenance for profitability analysis and
benchmarking. International Journal of Production Economics, Vol. 79, pp. 15–31.
Komonen, K. 2002b. Views on performance indicators and benchmarking in industrial

maintenance. Maintenance 5, pp. 52–56.
20
Komonen, K. 2005. E-FAMEMAIN Benchmarking tool for industrial maintenance – a web

based planning tool. Helsinki, Finnish Maintenance Society, p. 22 p. (In Finnish).
Komonen, K. 2006. E-FAMEMAIN: A new kind of web-based benchmarking tool.

Conference Proceedings. Euromaintenance 2004, Basel.
Komonen K. 2006b. Availability guarantees and physical assets management: Empirical

evidence of the impact of some underlying factors. Conference Proceedings.
RAMS 2006. Newport Beach, CA, USA.
21
Managing corrosion in biomass boilers: benefits and limitations of coatings
Managing corrosion in biomass boilers:

benefits and limitations of coatings
Satu Tuurna1, Tommi Varis1, Kimmo Ruusuvuori1, Stefan Holmström1,

Jorma Salonen1, Pertti Auerkari1, Tuomo Kinnunen1, Patrik Yrjas2,
Risto Finne3, Matti Nupponen4, Ulla McNiven5,
Hannu Ahonen5 and Ari Kapulainen6
1
VTT Technical Research Centre of Finland, Espoo, Finland
2
Åbo Akademi, Turku, Finland
3
Kuopion Konepaja, Kuopio, Finland
4
Turku Energia, Turku, Finland
5
Fortum, Espoo, Finland
6
Ekokem, Riihimäki, Finland
Abstract
Oxidation and corrosion of materials in highly demanding applications is critical

for longterm performance of materials used in power generation. Energy
production in biomass fired boilers is increasing rapidly due to the advantages of
CO2 neutrality and renewability. Fluidised bed combustors and grate fired boilers
have proved the most reliable technologies for burning biomass since its fuel
characteristics are variable. The fly ash of biomass is relatively corrosive due to
its composition containing high concentrations of chemically active compounds
of alkali, sulphur and chlorine, and it also may contain erosive components
especially in the case of fluidised bed boilers. Due to the potential for severe
chlorine induced corrosion, plant operating temperatures and efficiency must be
limited. To overcome material wastage more alloyed materials are needed. The
high cost and sometimes hard to manufacture bulk materials could be replaced
with coatings, which can be applied on the more economic low alloy steel
having the proper mechanical properties. The paper discusses the benefits and
limitations of thermal sprayed coatings in biomass boilers. The applied examples
22
have included iron and nickel based HVOF and arc sprayed coatings subjected to
verification field testing in boiler testing under in aggressive biofuel conditions.
The coatings have shown good corrosion resistance in long-term field tests.
1. Introduction
Thermal sprayed coatings have been used several years successfully as
protective coatings against erosion in coal-fired power boilers. As the variety of
dirty fuels used (waste, biomass) in power generation increases and the steam
temperatures are increasing with higher efficiency demands, the need for better
materials and protection become more important. The rate of corrosion in
biomass combustion is much higher than in conventional coal fired power plants
due to the heterogeneous fuel quality. Biomass deposits are usually rich of
potassium, silicon and calcium, and chlorine and sulphur content can be also
high. Chlorine can accelerate corrosion resulting in oxidation, metal wastage,
internal attack, void formation and loose scales. HCl alone in the flue gas can
not cause severe attack, but it may provide scale failure and enhance the
corrosion rate. Severe corrosion may occur when chlorine is present in the
deposits which are in contact with the metal scale. Alkali salt mixtures have low
melting points and tend to melt on tube surfaces that increases the ash deposition
on the tubes. The presence of alkali chlorides can cause enhanced corrosion
already below the melting point of the deposit. The corrosion may be further
accelerated by SO2, which can cause sulphation of alkali chlorides liberating
HCl or Cl2 gas near the metal surface. Fouling and slagging seem to be worsened
by the presence of chlorine which increases the mobility of inorganic
compounds. Many times the major mechanism of material wastage in biomass
boilers has been detected as erosion stimulated corrosion and corrosion
accelerated erosion, especially in the case of fluidised bed boilers [1–6].
Operating temperatures and efficiency are limited to maximise the lifetime of
boiler components due to the potential for severe chlorine induced corrosion. In
order to minimise material related risks, normally used low alloy steels have
been replaced with high alloy steels in more aggressive conditions. These alloys
normally contain high amounts of chromium and nickel for enhancing high
temperature corrosion resistance. However, many times a combination of good
mechanical properties and superior corrosion resistance is mutually exclusive.
The high cost and sometimes hard to manufacture bulk materials could be
23
replaced with coatings, which can be applied on the more economic low alloy
steel having proper mechanical properties.
In demanding corrosion environments it is not sufficient to just specify alloy
and coating method. To obtain reliable coatings the whole coating production
needs to be considered including alloy selection, spray parameters, surface
preparation and base metal properties. The high velocity oxy fuel (HVOF)
process, arc spraying, and spray with fuse are the most promising techniques to
optimise both coating properties and economical aspects [7]. HVOF process is
used in many industries due to flexibility, cost effectiveness and the superior
quality of coatings [8]. The HVOF coatings are considered as relatively
homogeneous and dense, but always contain some oxides and porosity (< 1%) at
splat boundaries. These inhomogeneities affect the corrosion protection ability
of the coatings. Aggressive elements from the surrounding environment can
proceed through a coating causing the corrosion of coated component at least in
some extent. The critical issue is to achieve high integrity coatings without
significant particle oxidation during spraying. A cheaper option that is also more
easily adaptable to on-site coating is arc spraying, although it will result in a
higher porosity level (< 5%) [9, 10]. Thermal sprayed fused coatings are dense
and very well adhered to substrate, but the applicability of these coatings in
boiler plants is challenging and the production is mostly carried out in
workshops. However, some companies are specialised on on-site manufacturing.
Nickel-chromium based alloys have been widely used as protective coating
systems in environments with high temperatures [11]. To improve the erosion
resistance, the standard practice is to add hard particles such as carbides (often
chromium carbides) or borides to a metal matrix (e.g. nickel-chromium or iron-
chromium) coating. For best erosion resistance the carbide content can exceed
50–70%, however high carbide content can reduce ductility and resistance to
cyclic service [12, 13].
The corrosion behaviour of different HVOF and arc sprayed coatings were
tested in long-term field tests. The results of these tests are represented. The
overall benefits and limitations of coatings in biomass boilers are discussed.
2. In-service boiler testing

Coatings were applied in three boiler applications. Boiler I was a fluidized bed
boiler fired with solid biomass, mainly with peat mixed with wood-based fuels.
The location of interest was selected within the economiser of the boiler, at a
24
position where ash blockages and tube failures have been recorded. The base
material of the ∅48.5 × 5.05 mm (measured dimensions) test tube was 16Mo3.
The HVOF thermal spray process using a DJ hybrid with hydrogen and Amperit
584.1 powder was applied to produce an about 200 μm thick 75% Cr3C2–25%
NiCr coating on the tube, Figure 1. The selected type of coating is mainly
intended for applications against erosion, but with the high chromium and nickel
contents (Ni-20%Cr) of the matrix also shows considerable resistance to corrosion.
A maximum service temperature of 870°C is indicated by the powder manufacturer
for Cr3C2-NiCr coating. A short section at each end of the tubes was left
uncoated to facilitate appropriate tube attachment by welding. The coated length
of the tube was about 1 m. The coated tube was installed in the economiser,
Figure 2. The targeted time of exposure was set at about one calendar year.
Figure 1. An as-coated tube attached to a rotating fixture.
After about one year of exposure, the coated test tube was removed for post-
exposure analysis. For post-exposure analysis, the coated sample tube was
visually inspected and then cut to rings for more detailed cross-sectional
investigation. The cross-section ring samples were prepared for metallographic
inspection by grinding, polishing and etching, and for deposit analysis without
aqueous media. The inspections were made by light optical (LOM) and scanning
electron microscopy (SEM), and the external deposit was also analysed by using
25
energy dispersive X-ray spectroscopy (EDS). Similar procedures were also used
for post-exposure analysis for the other two cases.
Figure 2. A coated tube installed into an economiser in boiler 2.
Boiler II was also a fluidized bed boiler with solid biomass, mainly wood-based
fuels mixed with small amounts of peat. Test tubes were located at two
positions, at cold and hot economiser, where tube failures have been recorded.
The base materials of the test tubes were P235 in the cold end and P265 in the
hot end of the economiser. The measured dimensions of the test tubes and
applied coatings are presented in Table 1. Both the HVOF process using a DJ
Hybrid and arc spray using a Smart Arc were applied to produce about 300 μm
and 400 µm thick coatings, respectively, on the tubes. As in previous case, the
targeted time of exposure was set to about one calendar year. After about one
year of exposure, the coated test tubes were removed for post-exposure analysis.
Boiler III was a waste-fired grate boiler. The location of interest was selected
within the superheater area of the boiler. The base material of the superheater
tubes was 16Mo3 (nominal measures ∅38 × 5 mm). Both the HVOF and spray
and fused coatings were about 300 μm thick, Table 1. Also in this case the
coated length of the tubes was about 1 m and a short section at each end of the
tubes was left uncoated for welding. As in previous cases, the coated test tubes
were removed for post-exposure analysis after one year of exposure.
26
Table 1. Summary of produced coatings for different boilers. AM = amorphous.
Boiler Powder Tube

Coating type Composition Spray device
appl. manufact. material
Cr3C2- Sulzer, HVOF 16Mo3
Boiler I Cr3C2-25(Ni20Cr)
(Ni20Cr) WOKA 7302 (DJ Hybrid) ∅48.5×5.05
Praxair, Ni-46Cr-1.1Fe-0.5Si- HVOF P235
NiCr 50-50 Boiler II
1260F-10 0.1CMnO (DJ Hybrid) ∅38.3×4.2
Nanosteel, Fe-20Cr-10W-5Mo- HVOF P235
AM Fe-Cr Boiler II
SHS7170 5B-5Mn-2C-2Si (DJ Hybrid) ∅38.6×4.3
Praxair, Ni-46Cr-1.1Fe-0.5Si- HVOF P265
NiCr 50-50 Boiler II
1260F-10 0.1CMnO (DJ Hybrid) ∅38.7×5.1
Nanosteel, Fe-26Cr-5W-5B-4Mo- Arc spray P265
AM Fe-Cr Boiler II
SHS7170 4Mn-3Si-2C (Smart Arc) ∅39.5×5.5
Fe-27.2Cr-10.7Ni- HVOF 16Mo3
Fe-27Cr-2C Boiler III Durum
3.9Mo-1.4Si-2C (DJ Hybrid) ∅38×5
Nanosteel, Fe-25Cr-15W-12Nb- HVOF 16Mo3
AM Fe-Cr Boiler III
SHS9172 6Mo-5B-4C-3Mn-2Si (DJ Hybrid) ∅38×5
Ni-16Cr-4Si-4B-
Sulzer Metco, Flame spray 16Mo3
Ni-16Cr Boiler III 4Fe-2.4Cu-2.4Mo-
Metco 19E + fused ∅38×5
2.4W-0.5C
Ni-23Cr-5Fe-10Mo- HVOF 16Mo3
IN625 Boiler III Plasmalloy
4Nb-TiAlCMnSi (DJ Hybrid) ∅38×5
Ni-24Cr-16.5Mo- HVOF 16Mo3
Alloy 59 Boiler III
1.5Fe-0.5Al (DJ Hybrid) ∅38×5
3. Results of field testing

All the coated sample tubes were covered after a year exposure by a light brown
deposit. In the first boiler application, no clear signs of significant wall thinning,
defects of the coating or other damage were found on the external surfaces. The
general appearance of a tube cross-section including the deposit and coating is
shown in Figure 3. The coating seems to have suffered little if any wear,
corrosion or other damage during service. Regarding the potential impact of
corrosion, fireside element mapping of the coating showed only slight possible
penetration of the deposit elements into the outer coating layers up to depth of
27
about 10 μm. The results of EDS analysis taken from the deposit of the sample
tube are shown in Figure 4. The results showed high contents of compounds like
Ca, Al and Si, which are typical for deposit in peat and wood firing. The
observed K content was higher than Na content, which is also typically for
wood. Concentration of chlorine (more than 1%) is probably well compensated
by the much higher S content.
Figure 3. Cross-section of the Cr3C2-(Ni20Cr) coating (thickness 150–170 μm) below the
fireside deposit.
The visual inspection of hot economiser tubes of the second boiler application
revealed undamaged coating surfaces whereas corrosion pitting could be seen on
un-coated tube surfaces. Coatings resisted conditions quite well in the conditions
where uncoated tubes lost even 2 mm of their thickness. Deepest pits were found
on the windward side at the positions of ten and two o´clock. Some differences
between the coatings were observed in cross-sectional study, Figure 5. HVOF
sprayed nickel-chromium coatings seemed undamaged after a one year service.
Some cracking and spallation of amorphous iron-chromium coatings could be
observed, especially in the case of arc sprayed coating. However, all coatings
were still attached to the tube surfaces.
28
Figure 4. EDS spectrum of the deposit formed on the test tube in Boiler I. The peaks
show presence and relative amount of the indicated elements.
The visual inspection of the external surfaces of the sample tubes in the third
boiler applications did not reveal any signs of coating damages. The cross-
sectional studies showed that all the coatings were principally in a good
condition, Figure 6. At the inspection points, no peeling of coatings, no
corrosion of the coating structure or tube below the coating was found.
However, in some points a thin continuous oxide layer was formed at the
coating-tube interface. This phenomenon was usually connected to the vertical
cracking of the coating (Figure 6b) but not in all cases. Although cracks were
found, no corrosion products were observed inside the cracks. Some minor
cracks near the coating surface were also observed in the spray and fused
coating. In all cases uncoated tube material was corroded and corrosion
penetrated near the coating edge to around 100 μm, Figure 6c. The typical thinning
depth of tubes was around 100 μm.
29
a) b)
c)
Figure 5. Tube cross-sections after one year service from Boiler II. a) Uncoated tube,
b) HVOF sprayed nickel-chromium coating and c) arc sprayed iron-chromium coating.
30
a) b)
c)
Figure 6. Cross-sections of the coated tubes from Boiler III. a) HVOF sprayed Ni-24Cr-16Mo,
b) cracked HVOF sprayed Fe-25Cr and c) corroded tube surface near the coating edge.
4. Discussion
The service exposures were conducted in selected biomass and waste fired
boilers where deposit formation or tube failures were observed in the past. In the
case of boiler I, the applied coating performed well but the uncoated section was
also not experiencing much damage or wall thinning. It was concluded that in
this boiler the fuel and the process were sufficiently well controlled even without
additional protection from coatings. In Boiler II, no appreciable corrosion or
other damage was observed in the HVOF coated tubes under conditions where
31
uncoated tubes suffered severe material loss up to millimetres within one year of
service. While fully protective for the test period, some cracking and spallation
was observed in an amorphous iron-chromium coating, especially when
produced with arc spraying. Similar cracks were also found on the amorphous
coating from Boiler III. The importance of matching the CTE’s in the coating
and tube material is critical to avoid the tensile stress development in the
coatings during the boiler heat up. Tensile stresses may easily break the coating
having relatively brittle nature. The brittle nature of the thermal sprayed
amorphous arc sprayed material is seen in Figure 7, where horizontal and
through lamella cracks are visible even in the assprayed coating. Higher brittleness
compared to HVOF coating (Figure 7a) may arise from the microstructure where
arc sprayed coating typically has limited lamella cohesion due to the cracks
between lamellas as shown in Figure 7. The brittle nature of the coating makes it
also susceptible to failures in any kind of impacts during preparation. In particular
Ni-50Cr HVOF coatings have performed well, without any detected symptoms
of damage.
a) b)
Figure 7. Typical microstructural features of a) HVOF sprayed and b) arc sprayed coatings.
The coatings in Boiler III performed well and no corrosion was observed in the
coated areas. Some coatings, particularly amorphous coating and spray and
fused coating, had some cracks, but no corrosion products were observed inside
the coatings. Fusion treatment might form some stresses in the coating structure,
which facilitate crack formation. The spray and fused coating composed of
different phases and no clear lamellar structure which is typical for sprayed
coatings was present. The cracks are suggested to facilitate the formation of
32
oxide layers at the coating-tube interface. The oxide layers were also observed
below the sound coatings. This may be a result from the cracks situating nearby,
but not in the exact inspection point, or from the coating structure and
composition. Some structures may allow easier oxygen penetration although the
microscopical observation does not reveal any significant differences.
The good results suggest that the benefit of coatings could be extended to
boilers and locations with even more severe service conditions. Also, the observed
good performance of the tested coatings may provide opportunities for cheaper
coatings, and for this purpose further investigation with field testing is underway.
5. Benefits and limitations of coatings

The advantages of thermal spraying are that it is relatively inexpensive, part size
is not a concern and deposition rate is high. Relatively thin coating layers may
show an economical benefit provided that they retain sufficient protective
qualities. Unlike new bulk metals, coatings can be used as non-standard materials
with highly alloyed tailored compositions that have not been established in the
current material standards for pressure equipment. The experience on the
practical value of fireside thermal spray coatings has been rather mixed, and the
resulting operator attitudes sometimes reserved. However, improvements have
been reported in the materials, processes and quality aspects of such coatings,
e.g. [14–16]. And also the study related to this paper has shown that coatings can
withstand conditions where bare tube materials suffer severe material loss. For
successful service on the fireside, the coating should be thick and resistant
enough not to wear off before a specified minimum period of boiler service. The
coating should remain protective without peeling or exfoliation in the boiler
environment up to a specified time, and should allow for fabrication at
competitive cost so that the intended function (e.g. heat transfer) of the coated
component remains at the specified level for the specified time.
The adhesion of thermal spray coating is mainly mechanical, thus coating
removal and recoating are possible instead of replacement of whole component. The
surface heat-up is quite limited (max 200°C) and the potential for the distortion and
change of mechanical properties of tubes during coating process is very small.
The coating production is very sensitive for manufacturing parameters, and
incorrect parameters produce porous structures with weak adhesion. Even
though the coating material performance itself may be advantageous the chlorine
penetration to the base material-coating interface or thermo-mechanical compatibility
33
can cause unexpected damages. Based on the experiments it became evident that
the mechanical compatibility is one of the key issues in the material selection,
thus most of the tested materials can withstand the corrosion conditions but
suffered the cracking. Different kinds of surface modification techniques or post-
spray treatments, e.g. laser melting or sealants, can be used to improve the
coating quality and properties, such as density, hardness and toughness of
coatings [17–19]. These, however, increase production costs in some degree.
Coating selection should be based on life-cycle costs and the material selection
is case specific. The benefit from the investment to the coated solution should be
significantly larger than the cost of repeated replacement. Except perhaps in case
of simple wall panels, the component should be probably designed as a coated
system, to obtain maximum benefit from the attainable coating quality. In this
way suitable coatings could facilitate the use of more aggressive fuels and firing
conditions than what are the practical limits otherwise.
6. Conclusions
On-site experiments in three Finnish boiler plants for up to one year each at a
time have demonstrated a range of coating materials that promise to improve the
expected life of boiler tubes under corrosive/erosive service. Requirements of
further development are seen in up-scaling the coatings from the test scale of
about one square meter to that of thousands of square meters or more in plant,
essentially without loss in quality.
The results can be applied for current and future boilers to improve the
durability and life expectancy of the fireside heat transfer surfaces. For optimal
benefit in many cases, such solutions are likely to require some design
modification of the components. However, when the combustion conditions are
severe and result in rapid loss of fireside material, the economical benefits from
the coatings can be substantial by more than doubling the expected life.
Acknowledgements
The valuable technical and financial support of Fortum (Espoo, Finland), Turku
Energia (Turku, Finland) and Ekokem (Riihimäki, Finland) are gratefully
acknowledged. The financial support by Tekes (Finnish Funding Agency for
Technology and Innovation) and VTT is also acknowledged.
34
References
1. Pettersson, R., Flyg, J. & Viklund, P. High temperature corrosion under simulated
biomass deposit conditions. Corrosion Engineering Science and Technology,
Vol. 43 (2008), pp. 123–128.
2. van Lith, S. C., Frandsen, F. J., Montgomery, M., Vilhelmsen, T. & Jensen, S. A.
Lab-scale investigation of deposit-induced corrosion of superheater materials
under simulated biomass-firing conditions. Part 1: Exposure at 560°C. Energy &
Fuels, Vol. 23 (2009), pp. 3457–3468.
3. Nielsen, H. P., Frandsen, F. J., Dam-Johansen, K. & Baxter, L. L. The implications of

chlorine-associated corrosion on the operation of biomass-fired boilers. Progress
in Energy and Combustion Science, Vol. 26 (2000), pp. 283–298.
4. Ma, W. & Rotter, S. Overview on the chlorine origin of MSW and Cl-originated
corrosion during MSW & RDF combustion process. Second International
Environment and Public Health Track Shanghai, China, 2008. IEEEXplore.
5. Montgomery, M., Vilhelmsen, T. & Jensen, S. A. Potential high temperature corrosion

problems due to co-firing of biomass and fossil fuels. Materials and Corrosion,
Vol. 59 (2008), pp. 783–793.
6. Wang, B.-Q. Erosion-corrosion of coatings by biomass-fired boiler fly ash. Wear, Vol.
188 (1995), pp. 40–48.
7. Heath, G. R., Heimgartner, P., Irons, G., Miller, R. & Gustafsson, S. An assessment of
thermal spray coating technologies for high temperature corrosion protection.
Materials Science Forum, Vols. 251–254 (1997), pp. 809–816.
8. Stokes, J. & Looney, L. Residual stress in HVOF thermally sprayed thick deposits.
Surface and Coating Technology, Vols. 177–178 (2004), pp. 18–23.
9. Oksa, M., Kärki, J., Varis, T. & Turunen, E. Corrosion resistance of HVOF coatings in a biofuel
boiler plant. EUROCORR ´05, Lissabon, Portugal, September 4–8, 2005, 421.
10. Uusitalo, M., Vuoristo, P. & Mäntylä, T. High temperature corrosion of coatings and
boiler steels below chlorine-containing salt deposits. Corrosion Science, Vol. 46
(2004), pp. 1311–1331.
35
11. Higuera Hidalgo, V., Belzunce Varela, J., Carriles Menéndez, A. & Poveda Martinez,
S. High temperature erosion wear of flame and plasma-sprayed nickel-
chromium coatings under simulated coal-fired boiler atmospheres. Wear, Vol.
247 (2001), pp. 214–222.
12. Totemaier, T., Wright, R. & Swank, W. Microstructure and stresses in HVOF sprayed iron
alumide coatings. Journal of Thermal Spray Technology, Vol. 11 (2002), pp. 400–408.
13. Wang, B. Q. & Seitz, M. Comparison in erosion behavior of iron-base coatings sprayed
by three different arc-spray processes. Wear, Vol. 11 (2001), pp. 755–761.
14. Singh, H., Puri, D. & Prakash, S. Some studies on hot corrosion performance of
plasma sprayed coatings on a Fe-based superalloy. Surface and Coatings
Technology, Vol. 192 (2005), pp. 27–38.
15. Sidhu, T. S., Prakash, S. & Agrawal, R. D. Hot corrosion studies of HVOF NiCrBSi
and Stellite-6 coatings on a Ni-based superalloy in an actual industrial
environment of a coal fired boiler. Surface and Coating Technologies, Vol. 201
(2006), pp. 1602–1628.
16. Sundararajan, T., Kuroda, S. & Abe, F. Steam oxidation resistance of two-layered Ni–
Cr and Al APS coating for USC boiler applications. Corrosion Science, Vol. 47
(2005), pp. 1129–1147.
17. Oksa, M., Turunen, E. & Varis, T. Sealing of thermally sprayed coatings. Surface
Engineering, Vol. 20 (2004), pp. 251–254.
18. Tuominen, J., Vuoristo, P., Mäntylä, T., Kylmälahti, M., Vihinen, J. & Andersson, P. H.
Improving corrosion properties of high-velocity oxy-fuel sprayed Inconel 625 by
using a high-power continuous wave neodymium-doped yttrium aluminum
garnet laser. Journal of Thermal Spray Technology, Vol. 9 (2000), pp. 513–519.
19. Sidhu, B. S., Puri, D. & Prakash, S. Mechanical and metallurgical properties of
plasma sprayed and laser remelted Ni–20Cr and Stellite-6 coatings. Journal of
Materials Processing Technology, Vol. 159 (2005), pp. 347–355.
36
Selecting the right material for recovery boiler superheaters
Selecting the right material for recovery

boiler superheaters
Esa K. Vakkilainen and Pekka Pohjanne*

Lappeenranta University of Technology
Lappeenranta, Finland
*VTT Technical Research Centre of Finland
Espoo, Finland
Abstract
The aim of this paper is to examine the selection of superheater materials for
recovery boilers. Fireside wastage in recovery boilers is mainly caused by
formation of corrosive deposits, typically with aggressive molten alkali phases.
Fireside corrosion can also be caused by gaseous components and is then usually
associated with a reducing atmosphere. This active oxidation can affect
superheater tubes even at temperatures lower than the first melting temperature.
There is a wide array of available superheater materials to choose from and a
few newer materials to be considered. The operating conditions and deposit
properties must be looked at when material selection is done. The material
choice has a significant effect to the cost of the recovery boiler, as the selection
of the final superheater material can change the price by a factor of 10.
1. Introduction
The recovery boiler is fired with evaporated black liquor that is the spent
cooking liquor from the kraft process. The current average black liquor dry
solids content in Finland is around 80% DS. The black liquor is a good fuel
since it contains more than half of the energy content of the wood fed into the
digester. Recovery boiler carries out several functions, it burns the organic
material that is contained in the black liquor to generate high pressure steam and
37
nowadays also increasing amounts of electricity; it regenerates the used chemicals

in black liquor, and reduces some waste streams in an environmentally friendly
way [1]. The trend to increase electricity generation is forcing to increase steam
values. If the steam pressure and temperature can be increased to 104 bar and
520°C from typical current values of 80 bar and 480°C the electricity generation
capacity of the recovery boiler plant increases about 7% [2]. A typical new
recovery boiler is shown in Figure 1.
A key issue for the recovery boiler is the high availability need. And when
recovery boilers are designed one of the most difficult questions that arise is;
what kind of materials should one use for different parts of the recovery boiler.
In future this is expected to become even more complicated since the demand to
increase steam value increase the corrosion risk especially in the superheaters.
When selecting superheater materials one needs to consider several factors [4, 5].
Set design conditions (pressure, temperature, corrosion, erosion) determine the
boundary conditions. Tube material mechanical properties (strength) determine
the tube thickness. Manufacturing properties (weldability, need for heat
treatment) determine the required time to manufacture a piece of heat transfer
surface. Cost of material and relevant manufacturing cost give the total cost for
heat transfer surface. Often also material availability (time the material can be
delivered for manufacturing) plays a significant role. One should not forget that
variation of process values during operation (e.g. temperature spikes and liquor
chemistry) can greatly affect corrosion rate and strength of some materials but
not others.
38
Figure 1. Typical modern recovery boiler. The superheater tubes are colored red. [3].
2. Recovery boiler processes

Recovery boiler processes efficiently inorganic and organic chemicals in the
black liquor. Efficient inorganic chemicals processing can be seen as high
reduction rate. The furnace also disposes of all organics in black liquor. This
means stable and complete combustion. Reduction (removal of oxygen) and
combustion (reaction with oxygen) are opposite reactions and it is difficult to
achieve both at same unit operation, furnace.
Black liquors contain inorganic chemicals. During combustion and reduction
reactions significant amounts of alkali compounds vaporize. Small char
fragments and liquor particles may entrain to flue gas flow. These phenomena
cause deposits on heat transfer surfaces that increase heat transfer resistance and
39
thus reduce boiler performance. Even with improved air systems and advances
in recovery boiler design, fouling remains one of the big operating problems.
Corrosion is the main problem that limits the ability of kraft recovery boiler to
produce electricity [4]. In e.g. coal fired boilers much higher superheater
temperatures are typically used. In comparison to coal fired boilers kraft
recovery boilers have higher rates of alkali metals, chloride in gaseous form and
often highly reducing conditions caused by carryover particles On the other hand
levels of some high temperature corrosion causing elements like antimony,
vanadium and zinc are typically low.
Fireside corrosion in recovery boilers is caused by formation of deposits,
which have corrosive properties. Recovery boiler ash in contact with the tubes is
normally solid. Main compounds affecting recovery boiler corrosion are alkali,
chloride and sulfur (sulfidity). Sodium and potassium may enhance formation of
ash deposits with low melting points 500–600°C. Contact of tube surface with
molten alkali salts causes rapid corrosion. Active oxidation can affect superheater
tubes even at temperatures lower than the first melting temperature. Therefore
operating the recovery boiler without gaseous chlorine at superheaters is
essential. As sulfidity increases the polysulfide content of deposits on the heat
transfer surfaces can increase which increases corrosion [6].
3. Superheater failures
Recovery boiler superheaters suffer from corrosion, design issues and operation
issues. Operation problems that cause superheater failure can be [7]
- Condensate blocked tubes during start up
- Water from hydrotest not being evacuated during start up
- Water carryover from the drum
- Desuperheater spray water quality control issues
- Sootblower action
- High temperature cycling caused by poor superheating control.
Table 1 shows reported BLRBAC data on significant superheater faults in American
Recovery boilers. As can be seen there are several reasons for faults and this
places significant difficulties on selection of proper material. The reason for
large number of fatigue faults is the constant swinging caused by the recovery
boiler Sootblowing equipment operation. The second biggest failure type is then
hot corrosion.
40
Table 1. Typical recovery boiler superheater faults [8].
Fault type % fault % boiler

Fatigue 32.4 32.1
Hot corrosion 28.0 34.9
Tie in breakage 20.5 31.1
No fault given 12.5 23.6
Welding 9.2 15.1
Overheating 7.5 13.2
Roof sealing 7.1 12.3
Leak above the roof 4.6 7.5
Moisture 2.9 5.7
4. Superheater corrosion
The superheater corrosion behavior depends on three factors [9]: 1) The tube
surface temperature, 2) the deposit chemisty and 3) the local atmosphere at the
vicinity of tubes.
The main steam temperature is the main parameter that affects the choice of
superheater materials, this because the corrosion rate increases as the superheater
steam temperature increases, Figure 2. It should be noted that the superheater
corrosion rates are not uniform, but depend strongly on the local conditions.
However, the overall increasing effect of temperature to superheater corrosion is
evident. Steam side heat transfer coefficients in typical recovery boiler
superheaters are low. Superheater surface temperature can be tens of degrees
higher than the bulk steam temperature, Figure 3. A rough estimate is that in
areas were radiation is present the metal temperatures are about 50°C higher
than the steam temperature [10]. This is taken into account when positioning
superheaters. The corrosion risk of hottest superheaters can be reduced by
placing behind the nose or protecting it by screen tubes.
The superheater panels can also be designed so that the steam temperature in
the outer tubes is lower than in the middle to balance the high heat flux and
higher corrosivity. This is achieved by using shorter routes for outside bends,
which results to lower pressure loss, higher mass flow and lower end
temperature, Figure 4. This procedure is in more detail described in [11].
41
180
160
Corrosion rate [mills/y]
140
120
100
80
60
40
20
0
460 480 500 520 540 560
Temperature [ºC]
Figure 2. Effect of temperature on superheater corrosion rates [10].
200
180
In radiation
160 Behind nose
140
Tsurface - Tsteam, oC
120
100
80
60
40
20
0
0 5 10 15 20 25 30 35
Steam velocity, m/s
Figure 3. Effect of steam velocity on superheater surface temperature [1].
42
Flue gas
Secondary Tertiary
‘Cooler edge tubeb'
Figure 4. Cooler edge design to mitigate corrosion.
Melting behavior of deposits in kraft recovery boiler heat transfer surfaces

determines many of the deposit properties. The first melting temperature T0
(FMT) is the temperature at which the melt appears in a recovery boiler deposit,
Figure 5. The temperature where deposit melting appears marks also the region
where fouling and corrosion problems occur. Understanding and predicting first
melting temperature is particularly important in terms of superheater corrosion
[12]. Sodium and especially potassium lower the first melting point. Also
carbonate and sulfide may lower the FMT, but their effect is smaller than that of
the potassium [9]. Hot corrosion failures occur typically at the lower bends of
the hottest superheaters.
Recently it was noted that fast corrosion can occur even below the first
melting temperature, Figure 8. Reason to this is active oxidation. Chloride in
fuel forms NaCl, Cl2 and HCl in gaseous from. In active oxidation these gases
react with iron to FeCl2 [13]. This when in contact with oxygen, reacts back to
Fe2O3 releasing gaseous chlorine. Reformed iron oxide layer is porous and does
not offer protection from corrosion. Released chlorine can react with fresh iron
oxide and the cycle repeats. This process needs continuous gaseous chlorine
43
outside the metal surface to proceed. Other necessities for the process are high
temperature and reducing conditions [10]. In recovery boilers process upset that
produce high SO2 levels may ignite oxidation [10]. Active oxidation has been
studied a lot in connection with waste burning and recovery boilers. It can be
identified from sharp chloride layer between deposit and the uncorroded tube [14].
Figure 5. First melting temperature (T0) of recovery boiler superheater deposit.
Figure 6. Corrosion rate as function of the first melting temperature (T0) of recovery boiler
superheater deposit [15].
44
5. Selection of superheater materials

Typical primary superheater materials, when they are protected from direct
furnace radiation are carbon steel (SA/A-210 Gr A-1, S235JRG2) which can be
used up to 350–400°C. Secondary and tertiary superheater materials contain
often 1 to 3% Cr to improve corrosion resistance. Examples from these are
13CrMo44 and T22/10CrMo910) which are used up to 400–480°C. Corrosion
resistance is increased with chromium alloying and martensitic (X10CrMoVNb9-1,
X20CrMoV11-1) and austenitic (X1NiCrMoCu31-27-4, NiCr22Mo9Nb) steels
are used in hottest and most corrosive superheaters up to 500–520°C.
The material choice has a significant effect to the recovery boiler cost as the
selection of the final superheater material can change the price in ratio of 1 to
10, Table 3. Usually this is solved by using the most corrosion resistant and
expensive materials only in locations where needed, e.g austenitic stainless
steels in the lower bends or at the outer tubes in the panels.
45
Table 2. Materials used for recovery boiler superheaters.
Cr Ni Mo Si Cu Al Mn C Fe Other
15Mo3 0.3 0.3 0.3 0.26 0.3 0.65 0.16 97.6
13CrMo44 1.0 0.3 0.3 0.26 0.3 0.55 0.14 97.1
10CrMo910 2.3 1.00 0.50 96.1
X10CrMoVNb91 9.0 0.3 1.00 0.40 0.50 88.7 V, Nb
HCM 12 12.0 1.00 86.9 V, Nb
X20CrMoV121 12.0 0.5 1.00 0.50 0.50 0.23 85.2 V
AISI 347 17.5 10.3 1.30 0.70 1.60 0.08 68.4 Nb, Ta
AISI 321 18.5 10.2 0.75 2.00 0.06 68.5
AISI 309 23.0 14.5 0.70 1.50 0.10 60.2 P
HR2M 22.2 14.4 1.49 0.55 3.14 0.03 58.1 N
YUS170 24.4 13.2 1.50 0.78 0.58 0.02 59.5 N, P
AISI 310 25.0 20.5 0.50 0.08 53.8
HR3C 25.0 29.5 0.42 1.28 1.10 0.06 42.6 N
AC66 27.0 32.0 0.25 0.02 0.8 0.06 39.2 Ce, Nb
Sanicro 28 27.0 31.0 3.50 1.00 0.05 0.01 37.3
Sanicro 38 21.0 38.0 2.50 0.30 1.70 0.80 35.6 Ti
Alloy 825 22.0 39.0 3.50 0.36 0.50 34.5 Ti
HR11N 28.5 41.2 1.06 0.12 0.50 0.01 28.6 N
Super 625 21.0 52.7 8.97 0.21 0.28 0.20 0.01 15.8 Nb
Sanicro 65 21.0 61.0 8.40 0.35 0.38 8.8
Alloy 625 21.0 58.0 9.0 0.50 0.4 0.50 0.10 5.0 Nb,Ti
Alloy 690 29.0 61.5 0.1 0.03 9
46
Table 3. Recovery boiler superheater costs calculated for SHIII for boiler in Figure 1.
Material Tube Manufacturing Total

€/kg k€ €/kg k€ M€
SA/A-210 0.9 300 5 1 500 1.8
16Mo3 1 300 5 1 500 1.8
13CrMo44 1.2 400 5 1 500 1.9
10CrMo910 1.5 500 5 1 500 2
T24 1.8 600 5 1 500 2.1
X10CrMoVNb91 (T91) 4.2 1 400 8 2 500 3.9
AISI 347 6 1 900 9 3 000 4.9
AISI 310 9 2 900 11 3 500 6.4
HR3C 9 2 900 11 3 500 6.4
AC66 20 6 500 14 4 500 11
Sanicro 28 24 7 800 15 4 800 12.6
Sanicro 38 24 12 900 15 4 800 12.6
Sanicro 63 40 17 5 400 18.3
6. Future recovery boilers

Typical recovery boilers fire black liquor at a dry solid of 80% and operate at a
steam pressure of 80 bar and a temperature of 480°C to reduce superheater
corrosion and ensure boiler availability. In the future the situation is expected to
change. Reason to this is the aim and need to increase CO2 free, green
electricity. This requires that the steam values are increased maybe even to a
level of 160 bar and 550°C, which increase material temperatures to a level of
580–600°C. At present a number of modern boilers exist in Sweden, Finland and
Japan that operate at more than 100 bar and 500°C. No current recovery boiler
operates at higher pressure than 120 bar and higher temperature than 520°C.
Special attention is paid to the superheaters, whose corrosion resistance is the
major concern. The traditional heat resistant CrMo-steels and austenitic stainless
are not anymore applicable in these temperatures, because of their corrosion
resistance is insufficient. Tests have shown that the corrosion rates of the
austenitic stainless steel AISI 321, 347, 316 and even 310 increase to so high
47
level at temperature clearly beyond 500°C that their use should be avoided. New
materials solutions to this are looked from the highly alloyed austenitic iron and
nickel based alloys, which can be used either as mono or composite tube.
Composite tubes are considered attractive, because they are not susceptible to
steam/water side stress corrosion cracking, which is one of the Achilles heels of
austenitic stainless steels.
Stainless steels:
AISI 347 has been widely used in current superheaters typically as short pieces
in most corrosive places. Parts manufactured from AISI 347 have no record of
corrosion failure in Finland. In high temperature boilers, where improved
intergranular corrosion resistance is needed alloys like 309 and 310, with higher
amounts of chromium should be used. For this Sumitomo has developed new
grades HR2M and HR3C [16, 17]. Nippon Steel answer to this is 25Cr14Ni
(YUS 170) modified (low C and Si with small amounts of Mo and Ni) to resist
intergranular attack (IGA) [18, 19]. Welding Service Inc (WSI) uses for
recovery boiler superheaters Unifuse 310 weldoverlay coating. The experiences
have been mainly positive. In example cases Unifuse 310 is reported to last more
than two years without visible damage. In the same conditions the old steel
(T11) corroded 3.9 mm/year [20].
High Cr – high Ni – Fe base alloys:
Sanicro 28 is currently used in recovery boiler superheaters as compound tube. It

has a good corrosion resistance because of high Cr- (27%) and Ni- (31%)
content, other alloying elements being 3.5% Mo and 1.% Cu. AC66 (1.4877) is
like Sanicro 28, but it doesn’t contain Cu and it is Nb-stabilised because of the
higher carbon content. This material has not been widely used in recovery
boilers, but in comparable corrosion conditions at biomass boilers it has been
proven as one of the best materials.
High Cr – high Mo – Ni base alloys:
Materials belonging to this category are e.g. Sanicro 63, Sanicro 67, Super 625.
These high molybdenum grades have excellent service records from waste
incineration. When selection these materials special attention should be paid on
manufacturing as well as on long-term stability and ageing at high temperatures
48
[21]. WSI has also developed and used Unifuse 52 type welded coating (28% Cr,
14% Fe with minor amounts of Al and Ti) for use in coal fired boilers [20].
Coating can be welded to carbon or austenitic steel tube.
7. Conclusions
Recovery boiler operation in a pulp mill is a complex process with several
interrelated parameters and almost infinite variables. Therefore the selection of
materials for long time trouble free service is not an easy task.
There is a wide array of available superheater materials to choose from. There
are even a few newer materials that should be considered. The operation conditions
and deposit properties must be looked at when material selection is done. The
material choice has a significant effect to the recovery boiler cost as the selection
of the final superheater material can change the price in ratio of 1 to 10.
Understanding and predicting first melting temperature is particularly
important in terms of superheater corrosion. Contact of tube surface with molten
alkali salts causes fast corrosion even with the best of materials. Chloride
corrosion can affect superheater tubes even at temperatures lower than the first
melting temperature. Therefore operating the recovery boiler without gaseous
chlorine at superheaters is essential.
Acknowledgements
Parts of this paper were from research funded by the Tekes – the Finnish
Funding Agency for Technology and Innovation which is gratefully
acknowledged.
References
1. Vakkilainen, E. K. Kraft recovery boilers – Principles and practice. Suomen Sooda-
kattilayhdistys ry. Helsinki, Finland: Valopaino Oy, 2005. 246 p. ISBN 952-91-
8603-7.
2. Vakkilainen, E. K. Future of recovery boiler technology. 40th Anniversary International

Recovery Boiler Conference. Finnish Recovery Boiler Committee, Haikko
Manor, Porvoo, May 12–14, 2004. Pp. 51–58.
49
3. Tikka, M. UPM Kymin uusi talteenottolinja. (UPM Kymi new recovery line). Sooda-
kattilapäivä, 29 October 2008, Finnish Recovery Boiler Committee. (In Finnish).
4. Bruno, F. Superheater corrosion in kraft recovery boilers (Korrosion hos överhettare i

sodapannor. En översikt och diskussion). Stockholm, Sweden: Stiftelsen för
Värmeteknisk Forskning, 1997. Technical Report SVF/602. 134 p.
5. Bruno, F. On the influence of chlorides and sulphurous compounds on the corrosion of

superheater tubes in boilers with special consideration on Kraft recovery boilers.
Värmeforsk Service AB, 1999. Technical Report VF-664. 36 p.
6. Yli-Olli, S., Pohjanne, P., Heikinheimo, L., Kinnunen, T. & Pankakoski, P. H. Sulphidation
of advanced kraft recovery boiler materials. International Chemical Recovery
Conference – Efficiency and Energy Management. Quebec City, Canada, May 29
– June 1, 2007. Pulp and Paper Technical Association of Canada. Pp. 273–278.
7. McMillan, J. Superheater problems, their causes & solutions. 2004 International

Chemical Recovery Conference. Charleston, South Carolina, USA, June 6–10
2004. Pp. 953–964.
8. Grace, T. M. & Clement, J. L. Investigation of superheater design and performance,

Part 2. 2009 TAPPI Engineering, Pulping & Environmental Conference. 13 p.
9. Salmenoja, K. Superheater corrosion in modern recovery boilers. 40th Anniversary

International Recovery Boiler Conference. Finnish Recovery Boiler Committee,
Haikko Manor, Porvoo, May 12–14, 2004. Pp. 103–113.
10. LaFond, J. F., Verloop, A. & Walsh, A. R. Engineering analysis of recovery boiler
superheater corrosion. Tappi Journal, Vol. 75 (1992) 6, pp. 101–106.
11. Tran, H., Gonsko, M. & Mao, X. Effect of composition on the first melting temperature of
fireside deposits in recovery boilers. Tappi Journal, Vol. 82 (1999) 9, pp. 93–100.
12. Tran, H. Recovery boiler corrosion. In: Adams, T. N. (Ed.). Kraft recovery boilers.
Atlanta, GA, USA: Tappi Press, 1997. Pp. 283–324.
13. Grabke, H. J., Reese, E. & Spigel, M. The effects of chlorides, hydrogen chloride, and
sulfur dioxide in the oxidation of steels below deposits. Corrosion Science, Vol. 37
(1995) 7, pp. 1023–1043.
50
14. Nishio, T., Matsumoto, H., Shinohara, M. & Arakawa, Y. Influence of ash composition
on corrosion rate of superheater tubes. Proceedings of 2001 Tappi Engineering,
Finishing & Converting Conference. Atlanta, GA: Tappi Press. 6 p.
15. Salmenoja, K. & Tuiremo, J. Achievements in the control of superheater corrosion in

black liquor recovery boilers. Proceedings of 2001 Tappi Engineering, Finishing
& Converting Conference. Atlanta, GA: Tappi Press. 9 p.
16. Sumitomo Metal Ind. Ltd. Comparison of new materials for boiler tube and pipe.
Sumitomo, 904 F-No. 2878 Rev. 4, April 1998. 3 p.
17. Otsuka, N. Soodakattilapäivät 10.10.2002, Marina Congress Center, Helsinki, Finland.
18. Kaneko, S., Takatuka, H., Arakawa, Y. & Tanaka R. Mitsubishi-CBC high efficiency
chemical recovery boiler power generation system. 31st Pulp and Paper Annual
Meeting, ABTCP, Sao Paolo, Brazil, October 19–23, 1998. Pp. 693–704.
19. Matsumoto, H., Notomi, A., Nishio, T., Arakawa, Y. & Takatsuka H. Advanced
technology for corrosion resistant materials for recovery boiler. 1998 International
Chemical Recovery Conference. Pp. 51–60.
20. Lai, G. & Blogg, N. Unifuse cladding for surface protection against corrosion and
erosion/corrosion in power boilers and waste heat recovery systems. 2003
Tappi Fall Meeting. 19 p.
21. Kawahare, Y. & Kaihara, Y. Recent trends in corrosion resistant tube materials and
improvements of corrosion environments in WTE plants. Corrosion 2001. NACE,
Paper No. 01173.
51
Oxyfuel combustion: Oxidation performance of steels in simulated oxyfuel conditions
Oxyfuel combustion: Oxidation

performance of steels in simulated
oxyfuel conditions
Satu Tuurna, Pekka Pohjanne, Sanni Yli-Olli, Tuomo Kinnunen

and Petra Jauhiainen
Espoo, Finland
Abstract
Carbon capture and storage (CCS) is shown to be one of the key technologies for
CO2 emissions reduction in numerous energy scenario analyses. In addition to
moving towards emission free energy production, improved energy efficiency is
essential in order to reach emission reduction targets. Oxyfuel combustion can
be expected to differ from combustion in air by e.g. modified distribution of
fireside temperatures, much reduced NOx but increased levels of fireside CO2,
SO2 and water levels due to extensive flue gas recirculation. Increased flue gas
recirculation may increase the concentration of a number of contaminants in the
deposited ash and promote fouling and corrosion. In the paper, the oxidation
performance of superheater steels has been determined at 580 and 650ºC under
simulated oxyfuel firing conditions. The results are compared to corresponding
results from simulated air-fired coal combustion conditions.
1. Introduction
The demand for reduced emissions requires the development of efficient
combustion technologies suited for carbon capture and storage (CCS) systems.
Oxyfuel combustion is seen as one of the major options for CO2 capture for
future clean coal technologies. It can be in principle adapted to both new and
52
retrofit units with similar fuel flexibility as in conventional coal-fired plants

[1, 2 ].
Compared to conventional air-fired combustion, the oxyfuel process will use
oxygen or oxygen-enriched air to reduce the nitrogen in the flue gas and to
increase the CO2 content for easier capture. Much increased flue gas
recirculation is used to avoid overheating. Even if the thermal conditions are
similar in the oxyfuel and air fired systems, recirculation and reduced flue gas
volume will increase the concentration of not only CO2 but also other flue gas
constituents and impurities such as CO, SOx, Cl, H2O, Hg, As, Se etc. in the
oxyfuel process. The resulting concentrations naturally depend on the fuel and
process details, but imply a potential challenge to the high temperature corrosion
resistance of boiler surfaces. Volatile elements that often serve as bonding
agents for ash deposits will increase the risk of fouling and corrosion. The higher
O2 concentration may also contribute to higher SO2 to SO3 conversion with
significant effect on boiler corrosion. As the currently available information on
the corrosion rates under oxyfuel conditions is very limited [3–5], corrosion
testing of existing boiler materials under high CO2 combustion environments
with the relevant oxygen, water vapour and impurity concentrations is very
much needed. This should be supported and complemented by field-testing
preferably in a well characterised coal fired boiler before and after retrofitting to
oxyfuel combustion, to investigate the potential risk of increased corrosion rates.
2. Materials and methods

The oxidation performance of alloys T22 (10CrMo9-10), X20 (X20CrMoV11-1),
347HFG (TP347HFG), and HR3C was investigated under simulated combustion
conditions. The chemical compositions of materials are given in Table 1. Rectangle
shape specimens with dimensions of 12 × 18 × 3 mm3 were machined from thick
walled tubes. The samples were bright polished with SiC paper and ultrasonically
degreased in ethanol.
The multi-sample exposure tests were carried out under isothermal conditions
at temperatures of 580 and 650°C in gases simulating oxy-fired and air-fired
coal combustion. Steel T22 was tested only at 580°C, because 650°C is so much
over its design values. High chromium steel HR3C was tested only at 650°C.
The exposure tests were carried out in horizontal tube furnaces with premixed
gases. The gas flow used was 10 l/h. The gases were passed through a humidifier
unit at an appropriate temperature to obtain desired moisture content in the gas
53
mixture. Table 2 summarises the test conditions. The specimens were weighed to
obtain mass changes. The thickness and microstructure of the oxide layers were
evaluated by optical and scanning electron microscopy (SEM) combined with
energy dispersive spectroscopy (EDS).
Table 1. Chemical composition (in wt.%) of the test materials.
Alloy Fe Cr Ni Mo Other
T22 bal. 2.16 0.06 0.92 0.48 Mn, 0.26 Si
X20 bal. 11.2 0.48 0.86 0.2 C, 0.49 Mn, 0.3 V
347HFG bal. 18.3 11.7 0.23 0.07 C, 1.64 Mn, 0.92 Nb, 0.33 Cu
HR3C bal. 25.5 19.4 0.11 1.1 Mn, 0.39 Si, 0.45 Nb
Table 2. Test conditions for laboratory tests.
Simulated Vol. % Temp. Duration

system
N2 O2 CO2 H2O Ar °C h
Air-fired 73.7 1.8 16.0 8.5 - 580, 650 168, 500, 1 000
Oxyfuel - 3.6 60.0 30.0 6.4 580, 650 168, 500, 1 000
3. Results
SEM/EDS analyses of cross-sections revealed that the oxide layer formed at
580°C on the T22 steels had a triple layer structure, Figure 1. The outermost
layer composition corresponded to hematite (Fe2O3). A layer beneath that
consisted of magnetite (Fe3O4). The inner scale growing inwards from the
original sample surface consisted of magnetite type oxide, partly an iron-
chromium spinel, probably mixture of Fe3O4 – (FeCr)3O4. In addition, there was
an internal oxidation zone at metal surface. After 1 000 h in the air combustion
environment the thickness of hematite layer increased at the expense of
magnetite layer compared to the oxide layers formed in oxyfuel combustion
environment.
54
a) b)
Figure 1. Optical micrographs of T22 cross-sections after exposure at 580°C in a) air

combustion and b) oxyfuel combustion environment for 1 000 h.
From micrographs of X20 (Figure 2) it can be seen that also in this case oxide
scales formed had a triple layer structure containing hematite, magnetite and
mixed iron-chromium oxide with small amounts of alloying elements like
molybdenum. An internal oxidation zone could be found at metal surface, but in
this case it was discontinuous. The outer hematite layer was much thicker on
samples exposured in air-fired coal combustion conditions than in oxyfuel
conditions.
Oxide layers formed on steel X20 had a two layer structure at 650°C. No
spallation was observed. The outward growing scale consisted of hematite with
only local traces of magnetite. The inward growing layer consisted of the mixed
oxide of iron and chromium with minor amounts of alloying elements from the
alloy. Figure 4 represents the EDS analyses of formed oxide scales. Some
carbon enrichment was found near the interface between inner and outer oxide
layer, and also inside the inner layer after 1 000 h exposure in oxyfuel
combustion environment (Figure 4b). The inward growing layers contained quite
much porosity, Figure 3. There were no significant structural differences
between oxides formed in air and oxyfuel combustion environments at 650°C.
No internal oxidation zone was found beneath the oxide scale.
55
a) b)
Figure 2. Optical micrographs of X20 cross-sections after 1 000 h at 580°C in a) air

combustion and b) oxyfuel combustion environment.
a) b)
Figure 3. Optical micrographs of X20 cross-sections after 1 000 h at 650°C in a) air

With short exposure times only very thin or no oxide scale was formed on alloy
347HFG. After 1 000 h exposure at 580°C in air combustion environment the
oxide layer was still relatively thin with a two-layer structure. The layer
thickness increased when exposure temperature was increased up to 650°C. The
oxides formed in oxyfuel conditions had a nodular structure; even at 650°C the
oxide scale was uneven around the sample although a continuous scale was
found some parts of the sample, Figure 5. EDS analyses showed that the outer
layer was iron oxide whereas the inner layer contained mainly mixed oxide of
iron, chromium and nickel.
56
100
Cr K
80 Fe K
CK
OK
Relatively frequency
60
40
20
0
0 10 20 30 40 50 60 70
-20
Distance from oxide surface [μm]
a)
100
interface between outer
and inner oxide scale
80
60
CK
40 OK
Cr K
Fe K
20
0
0 10 20 30 40 50 60 70
-20
b)
Figure 4. EDS line scan analyses representing main components of oxide scales formed on
X20 after 1 000 h at 650°C in a) air combustion and b) oxyfuel combustion environment.
57
a) b)
Figure 5. Optical micrographs of 347HFG cross-sections after 1 000 h at 650°C in a) air

Very thin oxides were formed on the surfaces of HR3C steel at 650°C. After
1 000 h exposure some nodule formation, Figure 6, was observed on the HR3C
surfaces in both test conditions. Figure 7 shows the cross-sections of exposed
HR3C samples. The oxide was composed of two layers, the outer layer
contained iron oxide with small amounts of chromium and the inner layer
contained mixed oxide of chromium, nickel and iron.
Figure 6. SEM image of HR3C steel surface after 1 000 h exposure at 650°C.
58
a) b)
Figure 7. Optical micrographs of HR3C cross-sections after exposure at 650°C in a) air

combustion and b) oxyfuel combustion environment for 1 000 h.
90
OK
80
Cr K
70 Fe K
Ni K
60
CK
50
40
v
30
20
10
-10 0 5 10 15 20 25 30
Figure 8. EDS analysis of oxides formed on HR3C after 1 000 h at 650°C in oxyfuel
conditions.
Figure 9 summarises the weight gain data for each alloy after 1 000 h at 580 and
650°C in the both test conditions. The oxidation rate of steels T22 and X20 was
substantially similar at 580°C. Both steels were heavily oxidised. Steels T22 and
X20 were marginally less oxidised in simulated oxyfuel condition than in air
condition. At 650°C the oxidation rate (weight gain) of steel X20 was
59
substantially similar than at 580°C, although it appears to be marginally better in

air than in oxyfuel conditions. Steel 347HFG has approximately one order of
magnitude less weight gain at 580°C, and it seems to be somewhat better in
simulated oxyfuel conditions. At 650°C in air 347HFG is clearly more oxidised.
However, in oxyfuel conditions the difference between 580°C and 650°C in
oxidation is less. Alloy HR3C has clearly less weight gain than other test
materials. The results show one order of magnitude less weight gain under
oxyfuel condition in comparison with testing under air.
10,00
AIR
Weight gain [mg/cm ]
2
OXY
1,00
0,10
0,01
T22 X20 347HFG X20 347HFG HR3C
580°C 650°C
Figure 9. Weight gain of different samples exposed 1 000 h at 580 and 650°C in air-fired
and oxyfuel conditions.
4. Discussion and conclusions

Metallographical analysis of low alloyed samples showed that the exposure
conditions have effect on the oxide scale formation and structure. In reference
air-fired coal combustion conditions at 580°C the outermost hematite layer
thickness increased and magnetite layer decreased as a function of exposure time
as expected, Figure 2. The increase of hematite layer thickness in oxyfuel
conditions was much slower. At 650°C the difference between test conditions
was smaller. Oxidation rate at higher temperatures is higher and reactions occur
much faster.
60
The oxidation rate of steels T22 and X20 was substantially similar at 580°C.
Both steels were heavily oxidised. The oxidation rate of both steels X20 and
347HFG was somewhat lower in the simulated oxyfuel conditions than in the
simulated air-fired coal combustion conditions at 580°C. At 650°C the oxidation
rate of steel X20 was substantially similar than at 580°C, although it appears to
be marginally better in air than in oxyfuel conditions. The significant difference
of oxide structures of X20 formed at different exposure temperatures was the
porous internal layer formed at 650°C (Figure 3). Porosity may weaken the
protective behaviour of scale and accelerate breakaway oxidation. Oxide
spallation of can also affect on the observed oxide thickness and mass change
results. However, in this study clear spallation was not observed. These issues
bring some uncertainty to the results, for a mass change and oxide thickness.
The oxidation of 347HFG steel was less than low alloyed steels. 347HFG
seems to be somewhat better in simulated oxyfuel conditions. The oxidation
started with nodule formation, which began from the grain boundaries. With
longer exposure times and higher temperature the nodules grew together forming
a continuous layer. This is typical behaviour for alloyed materials. The smaller
grain size guarantees the easier formation of a protective oxide because of the
slow bulk diffusion of chromium. The oxidation of HR3C samples in both
atmospheres was very slow, especially under oxyfuel conditions, only some
nodules were found on the exposed surfaces.
In this study the test conditions were purely oxidising. The situation is likely to
change when some impurities, like SO2 and HCl, are added to exposure gases. The
metallurgical characterisation of samples revealed that some carbon enrichment
inside the oxide layer had occurred after 1 000 h exposure in oxyfuel conditions.
Acknowledgements
Financial support from VTT Technical research Centre of Finland and Tekes is
gratefully acknowledged. The authors thank the skillful assistance of Arto
Kukkonen, Taru Lehtikuusi, Marketta Mattila, Kirsti Nöjd and Päivi Varis.
References
1. Kung, S. C, Tanzosh, J. M. & McDonald, D. K. Fireside corrosion study using B&W
clean environment development facility for oxy-coal combustion systems. Proc.
61
of the 5th International Conference on Advances in Materials Technology for

Fossil Power Plants, 2008. Pp. 982–992.
2. Jordal, K., Anheden, M., Yan, J. & Strömberg, L. Oxyfuel combustion for coal-fired
power generation with CO2 capture – opportunities and challenges. Greenhouse
Gas Control Technologies, 2005. Vol. 7, pp. 201–209.
3. Covino, B. B., Matthes, S. A. & Bullard, S. J. Effect of oxyfuel combustion on superheater

corrosion. Corrosion 2008 Conference & Expo, March 16–19, New Orleans, LA,
2008. 9 p.
4. Hünert, D., Schulz, W. & Kranzmann, A. Corrosion of steels in H2O-CO2 atmospheres

at temperatures between 500°C and 700°C. ICPWS XV, Berlin, 2008.
5. Pirón Abellán, J., Olszewski, T., Penkalla, H. J., Meier, G. H., Singheiser L. &
Quadakkers, W. J. Scale formation mechanisms of martensitic steels in high
CO2/H2O-containing gases simulating oxyfuel environments. Materials at High
Temperatures, 2009. Vol. 26, pp. 63–72.
62
Safety issues arising from the corrosion-fatigue of waterwall tubes
Safety issues arising from the corrosion-

fatigue of waterwall tubes
John M. Brear*, Paul Jarvis*and Stephen Scully+

*Stress Engineering Services (Europe) Limited, Chichester, UK
+
Electricity Supply Board, Dublin, Ireland
Abstract
An incidence of waterwall tube failures, one leading to a significant steam

release external to the boiler, has highlighted the need for rigorous risk
assessment of such events. Initial review of the utility’s experience revealed one
of their boiler designs as having had a greater incidence of corrosion-fatigue
issues in waterwall tubing than the others. These units were treated as a priority.
To address failure likelihood, fracture mechanics calculations were performed.
These studies defined the necessary inspection coverage and sensitivity required
to underwrite safe operation at various potential loads.
Personnel safety was considered the most important consequence of failure.
Accordingly, potential steam releases were modelled to define safe exclusion
zones within the boiler house. Standard calculations were found to be non-
conservative; more exact formulae were needed to give realistic results.
Using the results of these studies, the utility was able to draw up a realistic
inspection plan. Safe operating pressures and appropriate exclusion zones were
defined for each boiler, and for a range of operational scenarios.
These tactics have allowed the utility to inspect the boilers in turn and to
repair all significant defects in the waterwall tubes, whilst maintaining a good
overall power output
In parallel, a root-cause investigation was performed to identify the factors
contributing to the failures. Where possible, causative influences were reduced
or mitigated so as to reduce the likelihood of failure whilst allowing increased
flexibility of boiler operation.
63
1. Introduction
Historically, waterwall tube life has been predicted on the basis of a measured
corrosion rate and the flow strength of the tube material. Provided this was
referred to the area of maximum corrosion – often the ebullition zone –
satisfactory results were obtained [eg 1, 2]. However, moves to increased cyclic
operation have led to an increasing importance of less predictable processes,
such as corrosion-fatigue.
Two sister once-through-type boilers, of 250MWe capacity, showed a higher
incidence of corrosion-fatigue failures than did other units in the utility.
Generally these manifested as leaks, but in 2005 a burst occurred, causing a
significant steam release outside the boiler (Figure 1). The operating pressure
was immediately reduced from 170 to 120 bar on the sister unit. A risk assessment
was put in place, including investigation of failure likelihood, of failure
consequence, of the variation of these with operational conditions, and of the root
cause. Later, this exercise was extended to other boiler designs within the utility.
Figure 1. The 2005 tube-burst, showing window blow-out failure.
This paper highlights the salient features of this investigation, centring on the
most critical units but including results from work on other boilers. It draws on
work performed within the utility and collaboratively between the authors’ two
organisations.
64
2. Failure likelihood
2.1 Failure analysis
The failure was a window blow-out. Examination of the fracture surfaces

indicated that there had been a pre-existing crack-like defect along the lower
fracture surface, growing from the bore of the tube. The fracture was deepest
where it was coincident with an old attachment weld, being effectively through-
wall at this point. A section (Figure 2) confirmed that the defect was a corrosion-
fatigue crack. The two circumferential cracks that allowed the tube material to
tear away were relatively clean with some plastic deformation evident. Several
sections through the top portion of the fracture found no evidence of a pre-
existing defect, but rather signs of deformation consistent with the material
being torn away due to overload.
Corrosion fatigue crack
Final fracture
Membrane wall weld
Figure 2. Section through the lower portion of the rupture.
2.2 Fracture mechanics
Stress analysis:
The tubes experience: the pressure stress, modified by the constraint of the
membrane wall; local thermal stresses associated with heat transfer through the
tube wall; long-range thermal stresses associated with the temperature distribution
in the panel; and system loads associated with supports and attachments. Only
the first is readily quantifiable.
65
The constraint effect of the membrane wall was investigated, but no suitable
hand calculations were found to enable it to be quantified. As the analysis
progressed, it was noted that the observed crack growth could be explained
solely on the basis of the hoop stress, assuming lower-bound materials
properties. The consequences after final fracture were influenced only by boiler
pressure tube bore dimensions. It was not considered necessary, therefore, to
perform a more elaborate stress analysis at this stage.
It should, however, be noted that the actual distribution of cracks and their
growth and fracture behaviour will be affected by the other sources of loading
identified above. This should be borne in mind when modifications are made to
components of this type and age.
Materials properties:
The tube material is 15Mo3 (Material Number 1.5415) and materials properties
have been taken from the most relevant standard contemporary with the
construction of the boiler: DIN 17 177, May 1979 [3]. Physical properties were
obtained from the corresponding Mannesmann data-sheet 405R [4].
Conservative fracture toughness values, appropriate to corrosive environments,
have been taken from API 579 [5] and UK nuclear practice.
The assessment has been performed using lower-bound properties. Whilst it
might be argued that actual properties are likely to be better, it is also recognised
that there are additional sources of loading that have not been explicitly
considered. The subsequent sections demonstrate that the failure can be
explained under a combination of the pressure stress and minimum materials
properties, thus other loads and material property variations can be taken to be in
balance and do not need further consideration at present.
Crack behaviour:
Any longitudinal defect initiating at the bore of a tube will extend through-
thickness and axially. A point will be reached where the remaining ligament
ahead of the crack fails instantaneously and the crack penetrates the wall.
Depending on the crack dimensions and the loading, subsequent growth in the
longitudinal direction may be stable, resulting in a leak, or it might be unstable,
leading to an instantaneous burst. The following sections address these various
aspects of crack behaviour, in accord with recognised practice [5, 6, 7].
66
Through-wall crack growth:
The failure analysis confirms a corrosion fatigue mechanism. However, it

appears that the process is corrosion dominated as the cracks are irregular and
show alternating periods of crack growth and corrosion. At present, there are no
data available to allow a predictive crack-growth analysis; however some
estimates may be made from historical experience.
Through-wall crack stability:
A conventional fracture-mechanics analysis was performed to determine critical

crack sizes for through-wall fracture. The load cases considered were based on
the operating and test pressures of interest. A range of initial crack lengths from
4 to 240 mm was covered and a semi-elliptical crack geometry assumed. The
surface behaviour is similar at both crack sizes, but instability at the deepest
point becomes more brittle with increasing crack length, due to the decreasing
constraint.
Instability of a 120 mm crack at the service temperature and an operating
pressure of 170 bar is predicted at a depth of 2.86 mm, and the subsequent fast
fracture through-wall is predicted to be by a predominantly ductile plastic
collapse. These predictions are fully consistent with the observed failure.
Leak-before-break
A semi-elliptical crack of large aspect ratio will, once it penetrates the wall,
assume an approximately rectangular geometry. Its stability with respect to
longitudinal growth may now be assessed in exactly the same way as the
through-wall behaviour was analysed.
At the service temperature and the operating and test pressures of interest, the
lower constraint on growth in this scenario leads to a more mixed failure mode
than was predicted in the through-wall direction. For the hydrostatic test
pressures the increased flow stress and considerably reduced fracture toughness
at ambient temperature lead to a predominantly brittle failure mode and a
reduced critical crack length. Figure 3 defines the calculated leak-before-break
boundary in terms of critical crack length as a function of service pressure.
67
120
100
Critical cracl length, 2c, mm
80
60
40
LBB boundary at service
conditions
20
LBB boundary at hydrotest
conditions
0
0 20 40 60 80 100 120 140 160 180 200 220
Pressure, bar
Figure 3. Leak-before-break boundary.
At the normal operating temperature and pressure, a crack of 50 mm or longer

will, once it penetrates the wall, lead to a burst. This is consistent with the recent
failure, where a 120 mm long defect led to a burst. The observed final failure
mode, where a panel was blown out of the wall, is not uncommon in conditions
of dynamic fracture and constrained displacements. The tubes are constrained by
the membrane wall, and this will tend to inhibit the more usual ‘fish-mouth’
opening associated with ductile bursts. The adjacent 45° bends and the welded
burner support stay will also have some influence. Under constrained dynamic
fracture conditions, the 45° shear lip causes the fracture line to curve and
become circumferential. High bending moments are then generated which cause
the far edge of the panel to fail by ductile overload, as seen here.
These predictions are also consistent with the earlier failure history where all
reported leaks appear to be associated with short crack lengths. It should be
noted, however, that the majority of the earlier leaks appear to be located at
welds, where microstructural effects may limit the crack size.
At service temperature and a reduced operating pressure of 120 bar, the
critical crack length is predicted to be a little below 70 mm. To achieve a critical
length of 100 mm would require the pressure to be reduced to below 80 bar.
68
Combined results:
Putting these parts of the analysis together allows the total stability behaviour of
the cracks to be described, see Figure 4. The broadly horizontal lines are failure
loci derived from the through-wall crack stability analysis; each represents a
different operating or test pressure. A crack of dimensions below the appropriate
failure locus is predicted to be stable: i.e. static or growing by a progressive
mechanism such as corrosion fatigue. A crack of dimensions above the failure
locus is predicted to be unstable and will penetrate the tube wall instantaneously.
The leak-before-break boundary is derived from Figure 3. Any crack whose
growth behaviour causes it to cross the appropriate failure locus at a point to the
left of this boundary will generate a leak; any crack whose growth behaviour
crosses the failure locus to the right of this boundary will generate a burst.
4.0
70 bar
80
3.5
Crack depth, a, mm
100
120
170
3.0 188
200
Failure 2005
2.5 LBB boundary -

service conditions
2.0
0 50 100 150 200 250
Crack length, 2c, mm
Figure 4. Total crack stability predictions at service conditions
The significance of safety valve and hydrostatic tests
The sister boiler was subject to a safety valve test, at the service temperature, in
which the operating pressure was raised to 188 bar. No tube failures occurred. It
is therefore reasonable to assume that no defect existed in the tubes, at that date,
of dimensions above the 188 bar failure locus shown in Figure 4. In particular,
considering the leak-before-break limit of 45 mm length at 188 bar:
69
• Any defect present in the tubes of depth greater than 2.75 mm must have
had a length less than 45 mm. At any operating pressure below 188 bar such
a defect is predicted to generate a leak.
• Any defect of length greater than 45 mm present in the tubes must have had
a depth less than 2.75 mm. At the current reduced operating pressure of
120 bar, such a defect needs to grow to a depth of 3.20 mm before it reaches
a burst condition (and would also need to be of length greater than 68 mm).
• It is prudent to assume that the waterwall tubes of this sister boiler contain
defects of depth close to 2.75 mm and length in excess of 68 mm. Such
defects are predicted to generate bursts when they penetrate the wall.
Regrettably there is no information available at present to predict the
growth rate a priori.
Both boilers underwent hydrostatic tests at ambient temperature in the two years
prior to the burst. The sister unit was tested in 2003 at 70 bar and 200 bar; the
boiler in which the burst occurred in 2004 at 200 bar. In each case some leaks
occurred during testing and some shortly after, generally in the range 1 600 –
3 700 h.
The calculated failure locus for hydrostatic testing at 200 bar is broadly
comparable with that for normal service for cracks up to 100 mm length, but
falls below it at greater crack lengths. Thus, despite the differences in failure
mode predicted in the crack stability analysis, this test should provide a realistic
screen for defects that would be immediately critical at the service condition.
However, it will only become conservative and provide a margin for crack
growth for defects above 100 mm in length.
Consideration of the burst event in the light of the previous hydrostatic test
history does allow some estimate of crack growth behaviour in the potential
burst regime. The failure locus for hydrostatic tests at 200 bar implies that any
crack of 120 mm length must have been less than 2.80 mm depth at the time of
testing, in comparison to a critical depth of 2.86 mm at the normal service
condition. The crack that led to the failure must therefore have grown from no
more than 2.80 mm to the predicted critical value of 2.86 mm in the 10 454
calendar hours between the hydrostatic test and the failure event. This gives a
lower (non-conservative) bound to the crack growth rate.
A generally accepted conservative Paris law for fatigue crack growth in a
corrosive environment gives the crack extension in metres per cycle as:
70
da/dN = 1.45x10-12 . ΔK3.5 (1)

where ΔK is the cyclic stress intensity range. Assuming a 24 hour cycle from
zero to the normal pressure stress, this law predicts a period of 9,576 hours for a
120 mm long defect to grow from 2.80 to 2.86 mm. This is within 10% of the
observed time, indicating that this scenario is at least plausible, though not
excluding the large number of other possible combinations of cycle length and
stress range that would give similarly close results.
It is noted that, at the time of the tube burst, the sister boiler had served 15,120
hours from its comparable hydrostatic test without incident at the normal service
condition. This would indicate that at the time of its hydrostatic test the most
significant defect present in the sister boiler was smaller than the one which led
to the tube burst.
Given that the above fatigue crack growth law is demonstrably compatible
with the failure history, it may be applied to the sister unit. The previous
calculations indicate that the reference defect condition is defined by the safety-
valve test in 2005, at 188 bar and the service temperature, rather than by the
hydrostatic test in 2003 at 200 bar and ambient temperature. Thus the crack
growth regime to be considered is from 2.74 mm in 2003, at the service
temperature and selected operating pressure. Predicted crack growth behaviour
for this sister boiler is shown in Figure 5, which demonstrates the increase in
crack growth time achieved by reduction of the operating pressure.
It must be stressed that these estimates of crack growth behaviour are
illustrative, rather than indicative, let alone definitive. Many assumptions have
been made from a relatively small amount of hard data. In particular, the
irregular crack growth rates evidenced by the metallography have not been
addressed. Nonetheless, reasonable assumptions on loading and materials
properties have produced predictions of defect stability and development that are
compatible with observation and experience.
71
4.0
170bar 120bar 100bar 80bar

3.5
Crack depth, a, mm
3.0
2.5
2.0
0 200000 400000 600000 800000 1000000 1200000
Time since last pressure test, h
Figure 5. Predicted crack growth behaviour for the sister unit.
3. Failure consequences
Whilst there is a significant impact on availability due the frequency of
waterwall failures, undoubtedly the danger to human life is the dominant
consequence. It was therefore decided that a rigorous analysis of the potential
steam release scenarios should be undertaken, in order to define safe exclusion
zones around these and other boilers where tube failures had occurred.
Type of release:
As the failure situation addressed is that of a waterwall, it is necessary to

consider possible water, steam and two-phase releases. Whilst the potential
ballistic trajectory of a discharged water slug is considerable, it can readily be
shown that instantaneous vaporisation would occur at the rupture point. Similar
considerations apply to a two-phase release. It is therefore only necessary to
examine a pure steam release in detail, as water and mixed discharges would
immediately converge to that situation.
72
Characteristics of a steam release:
At any realistic boiler operating pressures it can be shown that a steam release
would be choked, i.e. sonic in nature (that is, the local velocity of escaping
steam has reached the speed of sound at the local conditions). Calculations of
sonic release rates were initially performed in a manner consistent with the
methods given in recognised hazard assessment procedures [8, 9]. However the
results were considered unrealistic in the light of known experience. It was noted
that the formulae adopted by these procedures assume ideal gas behaviour; this
is not appropriate for steam at these conditions. In particular such formulae
implicitly underestimate the density of the steam, and thus the mass discharge
rate. Accordingly, more accurate models have been used here, following standard
texts on fluid dynamics [10, 11]. These take the density of the steam from
standard steam tables, rather than calculate it on the basis of molecular weight,
pressure and temperature. In all cases, a full rupture of the tube has been assumed.
Mass and temperature profile of the release:
The release characteristics – mass release rate, velocity – were calculated for a
range of operating pressures. As the models suggest, the mass release is
approximately proportional to the operating pressure. At operating pressures
above 60 bar, the mass released in three minutes exceeds the 10 000 lbs criterion
of API 580 [8] and the releases are therefore classed as ‘instantaneous’. Assessment
is thus based on the immediate effects of the release, rather than on subsequent
dispersion of the released material. It should be noted, however, that the total
fluid inventory in these drum-type boilers is large and therefore a considerable
amount of material will be released after this three minute period.
The corresponding heat released from the steam to the air and the quantity of
air required to cool the resultant mix to various temperatures were also
calculated. Standard relationships have been used to determine the distances
over which this occurs [9, 12, 13]. At a pressure of 120 bar, the exclusion zone
size is calculated to be 12.5 metres. It is noted that this distance is insensitive to
boiler pressure. This is a consequence of the significant contribution of the latent
heat of vaporisation to the total heat of the released steam – particularly at lower
saturation temperatures. Thus, in this range of operating conditions, the air/steam
ratio required to achieve any given temperature below 100°C is only mildly
affected by the release conditions.
73
In simple terms, therefore, the mass released (and therefore the heat released)
is more-or-less proportional to the operating pressure, but the distance over
which the temperature falls to a tolerable level is broadly constant.
100
80
Temperature, °C
60
40
20
0
0 10 20 30 40
Distance, m
Figure 6. Decrease in temperature of the release with distance from the rupture
Figure 6 gives the variation in temperature of the steam/air jet as a function of

distance from the release point, for an operating pressure of 120 bar. It is seen
that below 10 metres the temperature rises rapidly and exceeds 60°C at 9 metres.
Significant injury, even fatality, might be expected on such exposure. Brief
exposure to a temperature of 50°C may be tolerable (12.5 metres) but for anyone
involved in work, and therefore possibly restricted from rapid escape, a limit of
40°C (19 metres) is considered more appropriate. This last figure corresponds to
the distance between the unit’s boilers.
Exclusion zones should be planned recognizing that situations might arise
where personnel are required briefly to move towards the rupture point in
effecting their escape.
4. Root-cause analysis
The root-cause analysis comprised a detailed review of operating, inspection and
maintenance history in conjunction with a detailed failure analysis. This was
followed by a programme of controlled operation with strain gauges, additional
74
thermocouples in place (Figure 7) along with potentiometers to monitor boiler

movements. This enabled various operational modes to be studied and control
procedures implemented that would allow the desired operational flexibility
whilst minimising cyclic stresses within the waterwall.
Figure 7. Typical strain gauge installation.
The root-cause analysis identified a number of contributing factors. It was

demonstrated that the problem is largely a legacy of a period of poor water
chemistry and boiler operational difficulties during a period of high cycling and
load following in the 1980s. This was rectified in the early 1990’s but pitting and
cracking had initiated in the tubes. Whilst the poor water chemistry contributed
to the environmental influence on the corrosion-fatigue problem, the stress
cycles were influenced by a number of elements. These were primarily
generated during the periods when the feedwater flow was disrupted and there
was a mismatch between the thermal input and flow. Flow disruptions can be
caused by a number of issues, e.g. HP bypass openings, turbine trips and
economiser steaming. Once the flow is disrupted, or reduced below what is
needed for thermodynamic stability, the metal temperature rapidly rises. Thermal
differentials give rise to high stresses in each affected tube and neighbouring
75
tubes. A further stress cycle is imposed when fluid flow is re-established and the
tube cools. In conjunction with the pressure hoop stress these are the prime
stress generators for the axial cracking as observed.
This resulted in marked reduction in the severity and frequency of these flow
disruptions.
In the immediate period after the burst event, operation was kept as benign as
possible. However, over time the data generated by the additional instrumentation
identified the stress cycles, their causes were determined and efforts made to
reduce or eliminate them. By carrying out controlled test programmes, it was
possible to introduce increasingly flexible patterns without significantly
increasing the number or severity of flow disruptions and resultant stress cycles.
5. Closure
The work summarised in this paper has led to a much improved understanding of
the corrosion-fatigue issue in these boilers, and thus to more confident and
robust management plans.
The failure likelihood assessment established clear links between critical crack
size, including leak-before-break behaviour, and operating conditions. Use of
past inspection and test history enabled approximate crack growth behaviour to
be predicted, and thus defined the timescales for inspection and the safe operating
pressures to be maintained until such screening inspections could be performed.
The failure consequence assessment gave a rigorous basis for establishing
exclusion zones around the boiler for normal operation, and pressure limits for
times when it was necessary for personnel to be in closer proximity – eg when
working on neighbouring units. It also demonstrated the non-conservatism of
some standard release-rate formulae when applied to steam. Better formulae
have been implemented and are recommended for steam-generating plant.
The root-cause assessment has identified both historical and current causative
factors and allowed implementation of an investigation programme to address
these in detail. Specific stress-raising events have been identified and control
procedures have been put in place.
Acknowledgements
The authors are grateful to many colleagues within ESB for their cooperation in
the investigations, and to them, the utility’s insurers and the UK Health and
76
Safety Executive for useful discussions. This paper is published with the
permission of ESB Management and the Directors of SES Europe.
References
1. CEGB, The Control of Fireside Corrosion – various editions.
2. American Petroleum Institute, Calculation of Tube Thicknesses for Fired Heaters. API
Recommended Practice RP530 – various editions.
3. DIN Standard 17 177, May 1979. 15Mo3 Steel, Material Number 1.5415.
4. Mannesmann Steelworks Datasheet 405R, 1975. 15Mo3 Steel, Material Number

1.5415.
5. American Petroleum Institute, ‘Fitness for service’ API Recommended Practice 579,
First Edition, 2000.
6. British Standard BS7910, Guide on Methods for Assessing the Acceptability of Flaws in
Structures.
7. CEA Report DMT 96.096 A16: Guide for Defect Assessment and Leak-Before-Break
Analysis.
8. American Petroleum Institute, Base Resource Document on Risk-Based-Inspection,

Supplement to API Recommended Practice RP580/581.
9. Technica Ltd, Techniques for Assessing Industrial Hazards, World Bank Technical
Paper WTP 55.
10. A.C. Walshaw and D.A. Jobson, Mechanics of Fluids, Third edition 1980.
11. G.F.C Rogers and Y.R. Mayhew, Engineering Thermodynamics Work and Heat
Transfer, second edition 1967.
12. Albertson, Dai, Jensen, Rouse, Trans AMCE.
13. Hill, J Fluid Mech.
77
Life assessment for vintage boilers
Pertti Auerkari, Jorma Salonen, Stefan Holmström

VTT, Espoo, Finland
Leila Laaksonen*, Ulla McNiven, Olli Lehtinen*
Fortum, Naantali*/Jyväskylä, Finland
Sari Mäkinen, Ville Väänänen
Helsingin Energia, Helsinki, Finland
Reino Nikkarila
Inspecta, Espoo, Finland
Abstract
The high temperature sections of power and combined heat and power (CHP)
boilers are designed for some de facto expected but finite life. Several significant
damage mechanisms must be taken into account, particularly in superheaters and
reheaters that are designed for creep but also suffer from thermal degradation,
external erosion and corrosion, and internal steam oxidation that will gradually
increase the metal temperature. As rising temperature tends to accelerate all
damage mechanisms, major effects can be expected from the internal oxide
growth. The feedback loop is taken into account in procedures to predict
superheater life from known tube dimensions, time in operation, and other initial
data. The oxide effect can be mitigated by internal cleaning, but in-service
spallation of a relatively thick internal oxide can also become significant in old
plants. Spallation can create problems but extends the tube life by keeping the
wall cooler than with an adherent oxide. Examples are shown for boilers with
more than 150 000 h of service.
78
1. Introduction
Superheaters and reheaters experience the highest material temperatures of all
boiler internals that operate as heat exchangers. The mechanical loading by
internal pressure at high temperatures requires design for finite creep life, and the
fireside and steam side environments also promote wall thinning through corrosion,
oxidation and erosion. As the real service history will not be well known at the
time of design, the differences in the design assumptions, manufacturing/assembly
and later operation may offer potential for life extension, or for some surprises,
when compared to the nominal “design life”. Typically, condition monitoring and
maintenance actions are applied to support the life management so that the
expected life (that itself may vary in time) will be achieved.
Two features can be assumed to dominate in the evolution of high temperature
damage in superheaters and reheaters. First, increasing temperature will enhance
the rates of practically all damage mechanisms that may be involved (creep,
fatigue, erosion, external and internal oxidation/corrosion, material degradation).
Secondly, the growing internal oxide will progressively increase the material
temperature and thereby accelerate most damage mechanisms, shortening tube
life. The design practices are usually based on creep life, with allowances for
wall thinning due to the other mechanisms. As wall thinning can be significant
in the hot end sections, the initial wall thickness is high and therefore the initial
stress (and creep rate) is low, typically well below 30–50 MPa in older plants.
When this stress with wall thinning arrives to a range of significant creep rate,
the expected total damage rate also depends on the pressure level. In super-
heaters loaded by the pressure of live steam, relatively small additional wall
thinning with the accompanied rise in material temperature can drive the creep
rate relatively quickly towards the final failure. Together with operational
fluctuations or events of overheating, this is thought to be one of the main
reasons why creep failures in superheaters so rarely show creep cavitation
damage of the type that is common in welded steam lines. Reheater tubes show
cavitation damage more frequently, because the lower pressure level and initially
thinner wall can result in longer times spent in the required stress range.
Naturally, this can be also prevented by short term events like overheating.
In addition, the patterns of expected damage will also depend on material.
This is partly due to the likely ranges of application and operating values, but
also due to the composition, i.e. intrinsic features of the material type. Examples
of such characteristic features are listed in Table 1 for a few classical super-
79
heater/reheater steels. The rate of internal oxidation in steam is much reduced

with increasing chromium content of the steel, i.e. in the order of the steels
shown in Table 1.
Table 1. Features of internal damage in some superheater/reheater steels.
Material type Features of internal damage EN designation

Mo-steel Degradation, graphitization, (cavitation) 16Mo3
1Cr-0.5Mo Degradation, (cavitation) 13CrMo4-5
2.25Cr-1Mo Degradation 10CrMo9-10
11%Cr-steel Degradation (but relatively resistant) X20CrMoV11-1
1)
Austenitic (316H) σ-phase formation, SCC X6CrNiMo17-13-2
1) SCC = stress corrosion cracking
Consequently, any useful conclusions that may be drawn from the observed state
of the material are also material specific. One of the major quantities sought in
condition monitoring in this sense is the temperature history at the location of
interest, since this will largely determine the rates of damage accumulation. With
a rate model bounded by the initial, current and limit states of all influential
damage mechanisms, tube life can be then predicted and modified also to future
alternatives of the expected service history.
Apart from the model itself, this requires that:
- initial and current tube dimensions are reasonably well known/measured
at worst locations
- initial and current operating pressures and temperatures are taken at the same
locations, e.g. from operational records, and/or inferred from tube samples
- the rates of influential damage mechanisms are combined in the model.
As the true service conditions are not really known at the time of design, the
implementation of design, fabrication and assembly is likely to include many
conservative aspects that could on average translate into life extension.
However, this is far from certain and only in-service measurements and
inspections can confirm the actual state of the tubes. The long service life of
typical plants also requires upgrading well before the end of life, although this
may apply more on e.g. control systems than on the principal mechanical
80
components. Nevertheless, also superheater and reheater tubes are typical

components that could be subject to maintenance operations such as internal
cleaning (to remove scales and lower tube material temperatures) and local
renewal. It is sometimes challenging to keep track on the changes so that tube
sampling for condition and life assessment remains representative of the worst
areas of the boiler.
Although creep design with relatively new steels has often used optimistic
standard data [1, 2], more important sources of uncertainty are in the input data
of life assessment. Examples are shown below on the impact of errors,
uncertainties and material/boiler specific features in the assessed condition and
life of superheaters/reheaters.
Also other comparable effects may appear when using guidelines based on
shorter term data than the actual time in service. Here particular cases of low-
alloy steels are taken as examples, with long term internal oxidation, micro-
structural instabilities and operational data not necessarily agreeing for the
assessment of the temperature history. Various sources of bias can be expected
and have been observed for different approaches of assessment, with typical
consequences in the predicted (creep) life. By properly accounting for the
expected bias, it is believed that reasonable and safe but not overly conservative
estimates of the residual life can be achieved. This also applies to cases where it
is otherwise difficult to rely on one single source to indicate the actual
temperatures or damage levels. Examples are given for cases where the time in
service is approaching or exceeding 200 000 h.
2. Life of superheaters and reheaters

The hottest sections of superheaters often have metal temperatures exceeding
that of the steam by some 15–50°C. For a fixed steam temperature such a wide
range of metal temperatures can cover a very large range in the expected tube
life, as 10°C increase may approximately halve the tube life. It helps a little that
the operating stress may be dominated by the fairly well controlled internal
pressure, as superheaters and reheaters are made of relatively flexible small size
tubing with low system constraints at least outside attachments. However, the
tubes are subjected to the fireside environment, resulting in through-wall
temperature gradients and non-constant material temperatures in addition to wall
thinning by external oxidation, hot corrosion and wear (erosion) by the flue gas.
Finally, the internal oxide layer is gradually growing and by its insulating effect
81
resulting in progressively increasing material temperature of the tube, if

approximately constant heat flow is extracted from the superheater/reheater
section. To counter the effect of the internal oxide, this layer may be removed by
chemical cleaning, as has been done in the first example described below.
A reheater tube leak (∅ 70 × 4 mm, Figure 1a) was observed in a coal fired
boiler after 185 000 h of service. Cross-section of the tube at this location
showed through-wall creep cracking with abundant cavitation damage (Figure
1b). The differences in the microstructures of the tube (Figure 2) can explain
how the measured steam temperature after reheater was only about 465°C, while
the material temperature on the hotter side had apparently exceeded 500°C. For
the material of the tube (16Mo3), this was sufficient to result in creep failure
after fairly extensive time in service.
a) b)
Figure 1. Cross-sections of the failed reheater tube; a) main failure surface; and
b) initiating internally oxidised creep cracks on tube surface (scale bars 100 µm).
a) b)
Figure 2. Microstructure of the tube material outside the immediate failure area:
a) fireside, showing graphitization; b) opposite side (scale bars 50 µm).
82
Further investigation of unfailed tube samples suggested that the remaining

tubes were apparently subjected to less severe service conditions than the failed
tube. Life prediction for the tubes was based on a simplified effective
temperature history, essentially reducing the three past oxide cleaning episodes
to a single effective cycle and then to a natural oxide growth prediction from the
time of assessment onwards (Figure 3). An in-house software tool was used to
combine the effects of stress, temperature and creep strength evolution to obtain
a predicted life. In this case the tube temperatures were estimated from internal
oxide thickness, which however fails to indicate the temperature history before
the most recent tube cleaning (at 138 000 h of service). For the more distant past,
the micro-structural evolution was used to assess the tube temperatures. An in-
house study including 16Mo3 suggests that a Larson-Miller type of expression
can describe the effects of time and temperature to a common master curve with
constant features in the microstructures [2].
Temperature (°C)
Tox
Ts
t1 t2 t3 ta tr
Temperature (°C)
Teff
Tox
Ts
t0 t3 ta tr
Time (h)
Figure 3. Reduction of service history to simplified cycles; Teff = effective tube

temperature in past service, Tox = tube temperature from internal oxide; Ts = steam
temperature; t1 to t3 = time to oxide removal, ta = time of assessment, tr = predicted end of
life.
83
However, the microstructure (Figure 2) of the tube clearly indicated that

graphitization of the material has taken place on the hotter side of the tube,
resulting in a structure nearly completely lacking the eutectoid and grain
boundary carbides. Due to somewhat faster transformation under creep loading
and because such carbides are still present in the reference micrographs taken
from shorter term laboratory annealing, direct comparison of the microstructural
appearance (or hardness) can overpredict the effective service temperature. In
this case the tube temperature must have been between the recent (oxide indicated)
and higher, microstructure indicated level. The actual past effective temperature
before the last tube cleaning was estimated to have been about 520 ± 5°C.
The predicted minimum additional service life, taking also into account the
other potential damage processes in addition to creep, was still several years
(about 35 000 h). The predicted life should be inherently somewhat conservative
due to e.g. assumptions made on the distribution of pressure in time. The re-
heater section was included in the inspection program for damage monitoring.
For this purpose, a combination of replica inspections, diametral and wall
thickness measurements have been applied at selected locations. The section was
re-inspected after 15 000 h of additional service, with no indications to
contradict the previous assessment of predicted life.
Another example was taken from the reheater of another coal-fired plant after
about 180 000 h of service but no internal oxide cleaning. In this case the
material is 10CrMo9-10, for which the microstructural state should work
relatively well to describe the effective service temperatures (Figure 5). The
oxide has grown to a considerable thickness of about 0.4 mm (Figure 6), but
here it is the oxide based temperature assessment that is unreliable due to partial
oxide spalling. Therefore, microstructure (and hardness) was used to establish
the effective past service temperature at about 575–600°C. This is not far from
the design temperature of 579°C. This can be compared with the measured mean
steam temperature of 530 ± 12°C from a one-year sample of in-service
measurements. Using the same tools and principles as above, the minimum
residual life of comparable tubes was predicted to be 50 000 h. The reheater
section was included in a follow-up inspection program to monitor its
performance. After more than 20 000 h of additional service, no incident has
contradicted the predicted life.
84
Figure 4. Reheater tube sample (surface oxide removed), with indications of hot corrosion
and thermal fatigue.
Figure 5. Microstructure of a reheater tube, scale bar 10 µm.
85
Figure 6. Sectioned internal oxide; scale bar 100 µm.
3. Discussion and conclusions

While creep strength may provide the long-term integrity for superheater and
reheater tubes, life of these tubes can be radically shortened by parallel damage
mechanisms acting simultaneously and in combination with creep. Even
mechanisms like thermal degradation that are included in the standard creep
strength values are not explicitly accounted for by design, as internal oxidation
will progressively increase the tube metal temperatures. Increasing metal
temperature will accelerate all damage mechanisms such as creep, fatigue, and
fireside and internal corrosion/oxidation/erosion of the tubes.
However, the effects of temperature can also be used to support life
assessment by assessing the effective tube metal temperatures from the observed
state of an extracted tube sample. This approach is fairly well established [3], but
not without pitfalls as demonstrated above. Indications of the thermal history as
the observed state of e.g. tube microstructure, hardness, oxide thickness or direct
thermo-couple measurements can each suffer from various sources of error or
86
uncertainty. Some of the error sources are listed in Table 2. The above example
of steel 16Mo3 graphitization and resulting loss of carbides in the microstructure
is rather extreme, but it is common to see that microstructural features are
different in long term service and in samples heat treated to a presumed similar
state according to a time-temperature scaling parameter. This is partly because
different features in the microstructure may show differences in their
temperature dependence, and partly because by increasing the temperature to
compensate for reduced time of exposure one may arrive at a different region of
stability or different phase volume fractions. The other example on internal
oxide spallation shows a common phenomenon that is generally indicated when
the oxide is unrealistically thin, jagged or with a thinner outer than inner layer [4].
Table 2. Error sources in estimating the tube metal temperatures of superheater/reheater.
Method Sources of error/uncertainty Notes
Microstructure Limited coverage of t-T range E.g. graphitization of 16Mo3
Hardness Not established for all materials Widely used for 10CrMo9-10
Internal oxide Oxide spalling, oxide variation Indicated by layer imbalance
Thermocouples Drift in service, installation limits Often K-type (chromel-alumel)
Diffusion couples Non-standard installation e.g. PETIT couples
Such complexity is not uncommon in the measured data. Contradictory

indications may exist of the actual service conditions, but it is often possible at
least to rank them in order of credibility. Naturally, a single source of
information will remove the contradictions but is still not ideal. Much attention
should be paid to evaluating the temperature history, which will have a major
influence in the resulting predicted life.
The examples from reheaters of coal fired plants also demonstrated the
existing potential for material-specific features that can introduce bias in life
prediction. Nevertheless, so far there are no indications to suggest that the
predicted service life would be overly short or long in the example cases.
Moreover, oxide spallation and possible worst case censoring in sampling are
likely to extend life.
In the efficient supercritical boilers of future plant, superheaters will operate at
considerably higher pressures (up to 300–400 bar) and steam temperatures (600–
87
750°C). The differences in operational values will be reflected in materials and

relative roles of damage mechanisms, at least in some detail (Table 3).
Table 3. Expected differences in the damage mechanisms of end superheaters in vintage

boilers (ferritic steels) and future boilers (austenitic to Ni-based materials, supercritical).
Mechanism Vintage plant Future plant Notes
Creep Steam ≤ 550–580°C ∼600–720°C Design base
Fatigue Usually not critical Cycling limits Headers more limiting
Erosion (fireside) High flow High flow Not temp. limiting
Corrosion (fireside) Limits by fuel 1) Coal/gas only? Often temp. limiting
Oxidation (internal) Max ∼ 0.5–1 mm Max << 1 mm 2) Increases tube temp.
Material degradation Rarely limiting Rarely limiting? Reduced strength 3)
Combined total Mostly established Limited experience -
1) also through fouling

2) frequent spallation of external oxide layers
3) largely included in standard creep strength values except for overheating
While the operating temperature and pressure levels will continue to increase in
the top-of-the-line coal-fired boilers, this will not happen to similar extent in
future waste or biomass boilers that need to avoid excessive fireside fouling and
corrosion. For the extreme cases of corrosive municipal or industrial waste
incineration, the temperatures may remain so low (≤ 400°C) that creep is of
negligible importance. In spite of different materials and operating conditions
that require adaptation in the tools of life assessment, and persistent conflicting
damage information that is unlikely to disappear, the basic principles of
superheater life assessment are likely to remain similar in the future. With
gradually improving monitoring techniques and other new technology, the
available background information can be expected to support the life
assessments even better than today, independently of the boiler type.
Old plants will continue to operate under conventional service conditions
widely in the world, while new comparable and modified plants will also mature
to become new vintage plants. Consequently, the needs for long term life
assessment will not fade quickly even for the conventional range of materials,
88
fuels, and boiler designs. The critical issues will also remain, such as the quality
of data on long term material properties, accurately determined initial and current
states of the components of interest, understanding of the dominant damage
mechanisms, and their impact under the conditions of past and future service.
References
1. Holdsworth, S. R. & Merckling, G. ECCC developments in the assessment of creep-
rupture data. Proc. 6th Int. Charles Parsons Conf. on Engineering Issues in
Turbine Machinery, Power Plant & Renewables. Trinity College, Dublin, 16–18
Sept, 2003.
2. Auerkari, P., Salonen, J., Rantala, J., Holmström, S., McNiven, U., Lehtinen, O.,
Pihkakoski, M. & Nikkarila, R. Experiences on lifing of boiler plants. EPRI
International Conference on Advances in Condition and Remaining Life
Assessment for Fossil Power Plant. Louisville, USA, 16–18 October, 2006.
3. Salonen, J. & Auerkari, P. Microstructural degradation of boiler tube steels under long
term exposure to high temperature. Espoo: VTT, 1996. VTT Publications 280.
20 p. + app.
4. Knödler, R. & Straub, S. Growth of oxide scales during steam oxidation at 650°C. VGB
Powertech, 2008. Vol. 88, No. 10, pp. 66–70.
89
Italian standardization in the field of defect evaluation, fracture mechanics and fitness for
service of pressure equipment
Italian standardization in the field of defect

evaluation, fracture mechanics and fitness
for service of pressure equipment
Corrado Delle Site, Carlo Fossati*, Giuseppe Giannelli**,

Claudio Sampietri* and Andrea Tonti
ISPESL, Rome, Italy
*CSA/Exova, Milan, Italy
**ISPEL, Milan, Italy
Abstract
Typically the design standards and codes (DIN, ASME, API, AFNOR ...)
provide rules for the design, construction, inspection, testing and verification of
new pressure vessels, pipes, tanks, components of machinery, other components.
These codes often do not consider the fact that the component is subject to an
inevitable degradation during service and that during inspections following
manufacturing pressure test and putting into service problems can be found.
Indeed defects in structural components of plant and machinery can arise both
during construction but also during their lifetime (e.g. cracks, corrosion damage,
exposure to fire, etc.). The Fitness for Service (FFS) Assessment procedures,
developed as codes, standards, referenced guidelines periodically reviewed by
committees of recognized experts, were born just to cover this aspect and are set
to become increasingly important for analysis of component integrity. At the
same time, the existence of clear procedures becomes essential because of two
reasons: on one hand the growth of the competition between different plant
operators and, on the other hand, the competition between the different providers
of services (including FFS analysis, checks on plants, materials characterization,
etc.).
90
Italian standardization in the field of defect evaluation,Life
fracture mechanics
assessment and fitness
for vintage for
boilers
Currently, the Italian working group “FFS” is conducting a comparative study

of the major European and international standards and procedures in light of the
industrial situation and national laws, in order to draft an appropriate guideline
for the application of the “FFS” methodology in Italy. The principles of this
guideline will be presented and discussed.
1. Introduction
Engineering structures may contain imperfections (pores, flaws, defects, local
thin area or cracks) during the fabrication stage or during the service life [1]. The
degradation of materials in components of plants necessitate procedures and data
for assessing their service performance for safe operation and avoiding costly
outrages. If a flaw is found in a structural component during in-service inspection,
an assessment is needed in accordance with a respective code whether the flaw
should be removed (repaired) or component replaced or flaw is being left in the
component to continue to use the component. In late years several procedures
have been proposed to address this issue, avoiding any safety condition decrease.
Therefore, engineering assessment procedures containing analytical expressions
need to be further developed and unified – particularly in Europe – to assess
(primarily to provide conservative estimation of the critical condition) the structural
significance of the flaws or damages.
Damage of pressurized components may have several origins. Damage growth
can be driven by different mechanisms. Several factors can affect the behaviour
of the damaged component: application area, type of the component, loading
conditions and others. Main failure mechanisms are fracture, fatigue, creep and
corrosion. Also interactions among them need to be considered [2]. Industrial
sectors like nuclear power, petrochemical, chemical, etc. have already
established FFS procedures for their applications.
FFS procedures can be used to determine the size limits of defects due to
above mentioned damage mechanisms. This issue can result in cost savings in
operation. These procedures can lead to an assessment of component containing
defect to ensure its structural integrity during its foreseen life and its re-
inspection interval.
The outcome of the assessment is a decision to operate with or without
monitoring, to repair or to replace the component under investigation. Even
components manufactured by good manufacturing practices and procedures may
develop cracks or may contain various kinds of defects; FFS assessment can
91
Life assessment
service forequipment
of pressure vintage boilers
provide structural integrity information, a re-schedule of re-inspection intervals

as well as actions for life extension.
Each module of FFS analysis include a stage about defect detection and thus non-
destructive examination. Therefore it is important to establish a good maintenance
program and periodical non destructive examination to find out defects as soon as
possible. Non destructive examination is essential for fitness for service in the case
size and shape of the defect is required as an input for the subsequent analysis. The
level of inspection to comply with the whole procedure must be chosen considering
that FFS procedures are applied first to reduce the costs, avoiding unnecessary
repairs. So, in order to apply correctly a FFS procedure it is necessary to determine
the flaw length, flaw height, flaw position, flaw orientation, whether the flaw cross
section is planar or it is not planar. The defects needs to be identified as “surface
breaking defects” and “embedded defects”. In order to find surface breaking defects
the following methods can be used: visual testing, penetrant testing, magnetic
particle testing, eddy current testing, potential drop, radiographic testing, ultrasonic
testing. For embedded defects the following methods can be applied: radiographic
testing, ultrasonic testing, eddy current testing, D.C. potential drop testing.
Ultrasonic testing is useful to provide the height measurement. Eddy current and
potential drop can provide the cross-sectional area of the flaw.
2. Methods
Defect stability, the growth mechanism, the evolution of the flaw are essential
factors for the FFS procedure to apply. Activated flaws need to be carefully
evaluated during service, because they can seriously affect the safety of plant
working conditions. FFS procedures are established to provide better design
rules, good practices for fabrication and advanced tools for prevention of in-
service failures related to creep, fatigue, fracture and corrosion. These
procedures are continuously updated, modelling procedures are also improved
with new material properties and technology upgrading.
One of the latest procedure established in Europe is FITNET MK8 [2]. This
procedure is based, according to above mentioned principles, to well known
standards and good practices, like:
• SINTAP ‘Structural Integrity Assessment Procedures for European
Industry’, Brite-Euram Project No. BE95-1426.
92
Life assessment for vintage
service of pressure boilers
equipment
• BS 7910 ‘Guide on Methods for Assessing the Acceptability of Flaws in

Metallic Structures’, 2005.
• R6 ‘Assessment of the Integrity of Structures Containing Defects’, British
Energy Generation Report R/H/R6, Revision 4, 2001.
• API 579 ‘Recommended Practice for Fitness-For-Service’, Edition March
2000, American Petroleum Institute.
In case of using material properties in the assessment, a reference to a material
standard is required. Testing of the material is not covered by the FITNET
procedure.
Simplified models are needed to provide a standardized and reliable procedure
to establish whether a defect is stable or not. This issue leads to a schematic
approach and the use of flowcharts as much as it is possible. This can result in an
improved reliability of the method.
National and international standards are widely applied. The rules and
standards mentioned before are continuously tested, by application in the
pressure components.
2.1 Most widely applied standards
The Italian working group about fitness for service is mainly working on three
main international standard: the American API 579, the British BS7910 and the
European FITNET procedure MK8. The American API 579 was originally
developed for petrochemical applications. Now it is applied also in other sectors,
like power generation, with suitable modifications. The BS7910 was developed
after the publication of British Energy R5 and R6, related to flaw assessment for
high temperature conditions (R5) and flaw assessment in general conditions
(without addressing of temperature issues). The FITNET procedure was
developed by experts from 16 European countries, plus the contribution of
experts from Japan, from USA and from Korea. It is conceived for the
assessment of structures with welds, containing various kind of flaws.
2.1.1 API 579
In API 579, life extension is regulated by residual lifetime and re-inspection

interval, with a suitable safety factor. NDEs are relevant for flaw monitoring,
indications for actions, health and safety for personnel [3]. The main damage
93
Life assessment
mechanism considered are: brittle fracture, corrosion, pitting, hydrogen

blistering, welding misalignment, cracking, creep, firing, undercut and scaling. For
each of the damage mechanism described above, there is a specific assessment
procedure.
2.1.1.1 Brittle fracture
API 579 provides procedures for evaluating the resistance brittle fracture of
existing pressure equipment (vessels, piping and storage tanks) in carbon steel
and low alloy steels. It also provides the testing procedure to determine the
normal operation of starting and stopping.
2.1.1.2 Corrosion
API 579 provides procedures for the evaluation of general corrosion. The
thickness value used for the assessment may be to read in a "point" or in a
detailed profile. It also provides a methodology to be used in case the reduction
in thickness is treated as localized corrosion. This part also provides techniques
for the evaluation of isolated or grouped defects made by thickness decreasing
located in pressurized components. For evaluation the details of the thickness
profiles are required. Procedures can also be used to assess individual craters of
corrosion and “blisters”.
2.1.1.3 Pitting
The standard provides procedures for the evaluation of "pitting corrosion"

extended and/or localized. Procedures can also be used to evaluate a network
aimed to “blister”.
2.1.1.4 Hydrogen blistering
The standard provides procedures for the evaluation of isolated or grouped

“blisters” and damage from hydrogen (HIC/SOHIC) including those located in
the welding and structural discontinuities such as transitions in the shell,
stiffening rings and flanges.
94
equipment
2.1.1.5 Welding misalignment
Procedures are provided for the evaluation of efforts arising from geometrical
discontinuities in the shell, including the misalignment of welding and/or
distortion of a coating.
2.1.1.6 Cracking
The Standard provides procedures for the assessment of cracks. Solutions for
stress intensity factors and reference stress (load limit) are respectively in Annex
C and Annex D. Methods to evaluate the residual stress as required by the
assessment process are described in the Annex E. The characteristics of the
material required in the evaluation are provided in Annex F. Recommendations
for the assessment of the development of cracks, including attention to
environmental conditions, are included in the procedure.
2.1.1.7 Creep
The standard also include procedures for evaluating the remaining life of a
component operating under creep conditions. The characteristics of materials
required for the assessment are shown in the Annex F. The methods of analysis
for the assessment of crack development are included, considering the attention
to environmental conditions.
2.1.1.8 Firing
This procedure provides the evaluation of equipment which have been damaged
by fire. For the evaluation provides a methodology based on experimental values
of heat exposure during the fire. For the damage assessment of components the
procedures described in other parts of this norm are used.
2.1.1.9 Undercut and bulge
The procedure provides techniques for the assessment of bulges, undercuts and
combination thereof.
95
Life assessment
2.1.1.10 Scaling
It provides procedures for the evaluation of stripping including those located in the
welds and structural discontinuity in the shell, such as stiffening rings and flanges.
2.1.2 FITNET MK8
FITNET was a thematic networking project designed to develop a European

Fitness-for-Service Procedure (FFS) for assessing the structural integrity of
metallic welded or non-welded structures transmitting loads. It deals with
defects known or postulated to be present in a structure together with the
possible growth of such defects by a range of mechanisms (fracture, fatigue,
creep and corrosion) and the assessment techniques required to evaluate failure
risk. It is intended to provide a scheme into which results from European,
national programmes and other contributions are harnessed to the aim of a
fitness-for-service procedure based on European technology [1].
The FITNET Procedure covers four major failure modes for advanced design
and safety in all industrial sectors. It also includes the recent advances in
international, national standards as well as industrial codes such as API 579, BS
7910, British Energy R5 and R6 [1].
2.1.2.1 Fracture module
The FITNET Fracture Module is based on fracture mechanics principles and the
purpose of the analysis is to determine the significance, in terms of fracture and
plastic collapse, of flaws postulated or present in structures and components. The
procedure is based on the principle that failure is deemed to occur when the
applied driving force acting to extend a crack exceeds the material’s ability to
resist the extension of that crack. This material ‘property’ is called the material’s
fracture toughness or fracture resistance.
The principles of the FITNET fracture analysis method are:
• A hierarchical structure based on the quality of available data inputs
• Decreasing conservatism with increasing data quality
• Detailed guidance on determination of characteristic input values such as
fracture toughness
96
• Analysis in terms of a Failure Assessment Diagram (FAD) or Crack

Driving Force (CDF)
• Method of incorporating the effect of weld strength mismatch
• Guidance on the treatment of constraint in fracture assessment
• Compendia of stress intensity factors, limit load solutions and weld
residual stress profiles.
2.1.2.2 Fatigue Module
The FITNET fatigue module provides a series of assessment routes, for

evaluating the fatigue damage due to cyclic or fluctuating loads.
Two basic scenarios are foreseen:
a) There is no pre-existing flaw, and the goal of the analysis is to determine
the accumulation of fatigue damage at a critical location. In this case the
basic approach is to determine the fluctuating stress range at the location
in question and to relate this to appropriate fatigue life curves.
b) A real or postulated flaw is present, and the goal of the analysis is to
determine the fatigue life.
2.1.2.3 Creep Module
The FITNET Creep Module specifies methods for assessing defects in structures
operating at high temperatures and under creep-fatigue loading conditions. The
information required to perform an assessment are:
• the operating conditions
• the nature of the defects
• materials data
• structural calculations to correlate materials data with the behaviour of
complex structures.
The procedure includes the structural calculations required to assess whether a
given defect will grow to an unacceptable size in a given service life under a
given loading history. A step-by-step procedure is provided to perform these
assessments and methods for following each step are specified.
97
Life assessment
There are 13 steps in the creep module of the FITNET Procedure, as follows:
• STEP 1 Establish Cause of Cracking and Characterise Initial Defect
• STEP 2. Define Service Conditions
• STEP 3. Collect Materials Data
• STEP 4. Perform Basic Stress Analysis
• STEP 5. Check Stability under Time-Independent Loads
• STEP 6. Check Significance of Creep and Fatigue
• STEP 7. Calculate Rupture Life based on the Initial Defect Size
• STEP 8. Calculate Initiation Time
• STEP 9. Calculate Crack Size after Growth
• STEP 10. Re-Calculate Rupture Life after Crack Growth
• STEP 11. Check Stability under Time-Independent Loads after Crack Growth
• STEP 12. Assess Significance of Results
• STEP 13. Report Results.
2.1.2.4 Corrosion Damage Module
The corrosion damage module of the FITNET procedure provides guidelines on

the appropriate steps to take when an environmental assisted, either by stress
corrosion or corrosion fatigue, crack as well as when local thin area has been
detected in service and an assessment of the implications for structural integrity
has to be done. This module deals primarily with the Fitness-for-Service
assessments of damage types due to:
A) Environment assisted cracking (EAC):
a) Stress corrosion cracking
b) Corrosion fatigue
B) Local Thinned Area (LTA).
2.1.3 BS 7910
The fracture mechanics based fitness-for-purpose (FFP) approach enables the

significance of flaws to be assessed in terms of structural integrity. The concept
has been widely used by the British standards PD6493. The document has
recently been revised and is now published as BS 7910 “Guide on methods for
assessing the acceptability of flaws in metallic structures” [4].
98
equipment
BS 7910 includes the information required for assessment in terms of defect

characteristics and dimensions, stresses and material properties. It gives also the
procedures for assessment of fracture, fatigue, flaws under creep conditions and
other modes of failure. The annexes contain normative procedures for dealing
with certain situations: combined direct and shear stresses, determination of
fracture toughness from variable materials data and informative data like
residual stress distributions for welded joints, weld strength mismatch, proof
testing and warm pre-stressing, etc.
BS 7910 gives procedures for assessing fatigue crack growth based on quality
factors and crack growth calculation. A single procedure is given for assessing
flaws at high temperature and corrosion. Further assessment is foreseen if initial
results are not favourable. There are three levels for the assessment of fracture
based on failure assessment diagram concept:
• Level 1 is a screening procedure and the most conservative.
• Level 2 is material specific and estimates the interaction between fracture
and plasticity.
• Level 3 involves a direct calculation of plasticity effects.
A significant computation of stresses and fracture parameters is often necessary.
BS 7910 is intended to apply to equipment manufactured by different design
codes and materials, while API 579 is based around ASME design and materials.
Specific stress and materials data is required even for level 1 fracture
assessment. As a result, use of BS 7910 generally requires personnel experienced
in FFS assessment with access to appropriate data and/or testing facilities. The
code outlines methods for assessing the acceptability of flaws in all types of
structures and components. The types of flaws which can be assessed are planar
flaws, non-planar flaws and shape imperfections.
There are seven modes of failure:
1. Failure by fracture and plastic collapse
2. Damage by fatigue
3. Damage by creep and creep fatigue
4. Damage by leakage of containment vessels
5. Damage by erosion/corrosion
6. Damage by environment assisted cracking
7. Failure by instability.
99
3. Results, main differences among main

standards
The API are typical Anglo-Saxon standards which leave very little uncertainty to
use them and do not necessarily require that the user is an expert in the various
damage mechanisms. The BS even with very similar approach assumes that the
user is a very experienced technician. BS 7910, like FITNET MK8, is intended
to apply to equipment manufactured to different design codes and materials,
unlike API 579 which is based around ASME design and materials.
While BS7910 and API 579 are national standards, FITNET Thematic
Network (TN) has been prepared together by 62 experts from 16 European
countries and 41 organisations. Additionally, experts from Japan, USA, Korea
and Norway have provided in-kind contributions to the development of FITNET
FFS Procedure. National programmes and in-kind contributions are harnessed to
the common goal of a fitness-for-service procedure based on European advanced
technology. Establishing an Italian national guideline, to be applied in Italian
plants, all these differences need to be considered.
4. Discussion, Italian guidelines about Fitness

For Service
It is established that the Italian FFS evaluation of a pressure equipment will be
conducted on the basis of the results of the following actions to perform:
• preliminary analysis that takes account of the project, history of exercise
equipment, data of previous inspections, accident and/or repairs
• characterization of the defect
• identification of mechanisms of damage
• calculation.
4.1 Preliminary analysis
First there is preliminary assessments based on the following items which shall
be available to the user and/or manufacturer.
100
equipment
4.1.1 Original data project
The original data needed for project evaluation FFS in general depend on the
type of defect and/or damage mechanism identified in any case at least the
following information must be available:
• project drawings and/or spreadsheet detailing the size (including
tolerances and thicknesses of corrosion), materials used, the efficiency of
welding and standard project,
• in the case of piping, isometric drawings,
• design loads under various conditions (including pressure, temperature,
thermal loads under transient conditions, occasional loads such as wind,
snow, earthquake, etc.),
• process fluids.
4.1.2 Service data
Service data are necessary for the assessment as they are linked to the damage
mechanism and/or defect type; at least the following data need to be collected:
• Service history, pressure and temperature
• Working hours, number of start ups and shut downs, kind of process
history, environmental conditions (last two issues are not always required)
• Suitable documentation about each relevant variation of service conditions
during the life of the equipment
• Suitable documentation about each relevant variation of service conditions
already scheduled for future life of the equipment.
4.1.3 Maintenance data
Maintenance data necessary for FFS assessment are generally related to the
defect kind and/or damage mechanism; at least the following data are considered
essential:
• Outcomes from manufacturing inspections
• Outcomes from previous inspection during service
101
• Outcomes from previous FFS assessment and/or life extension assessment

• Summary of corrective actions performed on the equipment (reparations,
modifications and replacements).
4.1.4 Defect characterization by NDEs
Main parameters of each defect found in the pressure equipment need to be

assessed. Size, orientation, position and possible causes need to be establish
before performing the damage mechanism characterization. All NDE to apply
are evaluated before and during the inspection to apply the most suitable
procedures for the case under investigation.
4.1.5 Damage mechanisms
The draft of this part is still under development, but in principle it is foreseen to
individuate the cause of each damage mechanism. The guidelines for the
preparation of this part are according to FITNET MK8, so there are no large
constraints for the expert.
4.1.6 Calculations
On the basis of faults, the found mechanism of damage and the component under
investigation, the method/rule/procedure more suitable for the FFS analysis
among those provided for in paragraph 2 must be identified. If more damage
mechanisms can coexist simultaneously, the FFS analysis must be performed for
all mechanisms. The method/rule/procedure chosen must clearly show the limits
of applicability and limitation. Where there is more than one level of study
analysis, it should normally be performed starting from the lowest (most
conservative approach) and it could go to the next level where the analysis is not
satisfactory. The required input data necessary for assessing FFS in general
depend on the method/rule/procedure and level of study chosen. The
method/rule/procedure chosen must clearly show the list of all data input
required, regarding the characteristics of the component (dimensions, materials,
etc.), the design conditions, operation and proof that the component is subject
(pressure, temperature, nature of fluids, etc.), running time, characteristics of
materials with particular reference to those related to faults and the mechanism
of damage found. Particularly, as regards the characteristics of the materials, it is
102
equipment
appropriate that the method/ rule/procedure adopted will show these characteristics
or indicate explicitly how to obtain them.
5. Conclusions
The Italian working group “FFS” is conducting a comparative study of the major
European and international standards and procedures in light of the industrial
situation and national laws, in order to draft an appropriate guideline for the
application of the “FFS” methodology in Italy. The principles of this guideline
have been presented and discussed. The Italian procedure needs to integrate the
European procedure FITNET MK8, with API 579, as the stake holders in Italy
are from petrochemical application as well as from power plants. For
petrochemical application they are more familiar with API and thus the
transition needs to be smooth, without big jumps to the new procedure. This will
lead to a better accepted procedure. The principle has been established to follow
a single standard, once the working group has fixed the residual life, it is
necessary also to fix a suitable safety factor to determine the appropriate
inspection interval.
References
1. FITNET Final Technical Report, Version 27th Nov. 06.
2. FITNET, Fitness for Service (FFS) Procedure. Revision MK8, Vol. 1, January 2008.
3. API 579. Recommended Practice for Fitness-For-Service. Edition March 2000,

American Petroleum Institute.
4. Rastogi, R. An Introduction to BS 7910: The UK guide on methods for assessing the

acceptability of flaws in metallic structures. Indian Nuclear Society Lecture on
Welding, NDE and Integrity Assessment, September 18–22, 2006.
103
Defect integrity in components subjected to cyclic loading at high temperatures
Defect integrity in components subjected to

cyclic loading at high temperatures
Stuart Holdsworth
EMPA: Swiss Federal Laboratories for Material Testing & Research
Dübendorf, Switzerland
Abstract
Cracks established during component manufacture or service subsequently

propagate during high temperature duty due to the combined influences of
thermo-mechanical transients and creep loading. As a consequence of the
diverse loading conditions which can be responsible for high temperature crack
propagation in service, the assessment of defect integrity in such circumstances
is covered by more than one approach, each relying on a variety of material
property input data. These approaches and the respective properties required for
their implementation are reviewed. In addition, their application to the
assessment of high temperature defect integrity is examined.
Nomenclature
a, Δa Crack length, crack extension
A(T,υ,th) Material constant in mid-ΔK fatigue crack growth rate laws; dependent on
metallurgical condition, temperature, environment, frequency and hold time
A20 Material constant for 20°C in mid-ΔK fatigue crack growth rate power law
b Crack size exponent in short-crack creep-fatigue crack growth rate equations
B’ Constant in short-crack creep-fatigue crack growth rate equations (6,7)
C* Parameter characterising stress and strain rate fields at tip of crack in
material deforming due to creep
da/dN Crack growth per cycle
(da/dN)C Creep crack growth per cycle
(da/dN)F Fatigue crack growth per cycle
(da/dt)C Creep crack growth per unit time
Creep damage accumulated at surface in single cycle
104
DC, Total creep damage, total creep damage at surface

D(εu) Constant in creep crack growth equation (5); dependent on material and
temperature through dependence on creep rupture ductility
E, ET, E20 Elastic modulus, elastic modulus at temperature, T, elastic modulus at 20°C
HSFCG High strain fatigue crack growth
J, ΔJ Parameter characterising stress and strain fields at tip of crack in material
deforming plastically, cyclic J
K, ΔK Stress intensity factor, stress intensity factor range
Kmin, Kmax Minimum, maximum stress intensity factor
Kc Critical stress intensity factor responsible for unstable fracture
ΔKeff Part of ΔK responsible for crack opening in transients involving a
compressive component of loading [3]
ΔKeq Equivalent cyclic stress intensity factor [3]
ΔKth Fatigue crack growth threshold
LEFM Linear elastic fracture mechanics
LSFCG Low strain fatigue crack growth
m Exponent in mid-ΔK fatigue crack growth rate power law equations
N Number of cycles
q Cyclic strain exponent in short-crack creep-fatigue crack growth rate
equations (6,7)
rp Size of cyclic plastic zone
R Load ratio (Kmin/Kmax)
Rm Tensile strength
t, th Time, hold time
tm Maximum allowable time at temperature
T, Tref Temperature, reference temperature
α1, α2 Empirical constants in equation (8)
ε, Δ ε Strain, strain range
Instantaneous equivalent creep strain rate during the hold time
, Creep rupture ductility, uniaxial and multiaxial
Exponent in creep crack growth equation (5)
σ, Δσ Stress, stress range
σm Mean stress
σref Reference stress
υ Frequency (dN/dt in Figure 4)
1. Introduction
Cracks established during component manufacture or service may subsequently
propagate during high temperature duty as a consequence of the combined
effects of cyclic and creep loading. The operating conditions responsible for
crack growth at high temperatures are diverse, ranging from predominantly
cyclic to mainly steady loading with infrequent off-load transients. The fatigue
component of crack extension may be a consequence of high frequency (time
105
independent) or low frequency (time dependent) transients, applied in either load

or displacement control. Moreover, cyclic loading can be linear-elastic, elasto-
plastic or highly plastic. The creep component of cracking may be the result of
primary (directly applied) and/or secondary (self-equilibrating) loading. As a
consequence of the very diverse conditions which can be responsible for crack
propagation at high temperatures, the assessment of defect integrity is currently
covered by more than one approach, each with its own material property input
data requirements (e.g. Table 1). These are reviewed in the following paper.
An important consideration in the characterisation of creep-fatigue crack
growth rates and the assessment of high temperature components is the size of
the crack relative to the size of the cyclic plastic zone at the surface. In the
following review, long cracks are those whose size exceeds rp, and short cracks
are those whose size is less than that of rp.
Table 1. Data requirements for the assessment of crack growth at high temperatures.
Higher Frequencies, Lower Frequencies,

Load Control Displacement Control
Early (Short) Crack
da/dN(ΔJ,R,T,υ,th) da/dN(Δε,T,υ,th)
Growth (a < rp)
da/dN(ΔKeff, R,T,υ,th) +
Long Crack Growth da/dN(ΔKeq, T,υ,th) +
(1/υ).da/dt(C*,T) ;
(a > rp) (1/υ).da/dt(C*,T)
ΔKth(R,T,υ,th)
2. Long crack growth
2.1 General
Cyclic crack growth is conventionally considered in terms of three regimes

(Figure 1). These are: i) a low-ΔK regime close to the fatigue crack growth
threshold, ii) a mid-ΔK regime in which propagation rates are modelled by a
power law (equation 1), and iii) a high-ΔK regime in which Kmax approaches Kc
(or σref approaches Rm), i.e. the condition(s) responsible for unstable fracture.
(1)
In equation (1), A(T,υ) and m are material constants dependent on temperature,
environment and frequency (below a limiting value).
106
1E-1
LOW- Δ K REGIME MID- Δ K REGIME
1E-2 Sensitive to : Sensitive to :
- material condition - temperature
HSFCG REGIME
- load ratio - environment
1E-3
- temperature - frequency / hold time
- environment Insensitive to :
da/dN, m/cycle .
1E-4
- frequency / - material condition
hold time - load ratio
1E-5 m
1E-6
LSFCG REGIME
HIGH- Δ K REGIME
1E-7 Sensitive to :
- material condition
1E-8 m - load ratio
da/dN = A(T, υ ) .Δ K
- temperature
1E-9 - frequency / hold time
- size/geometry
1E-10
ΔKth ΔK, ΔKeff
Figure 1. Cyclic crack growth rate regimes [1].
At low-ΔK levels close to ΔKth, the magnitude of da/dN is very sensitive to small
increases in ΔK and dependent on the same factors which influence ΔKth, these
being: material, microstructure and yield strength, temperature, environment and
load ratio (R). Propagation rates in the mid-ΔK regime are less sensitive to
microstructure and mean stress (R) effects.
In the high-ΔK regime, da/dN becomes increasingly sensitive to the level of
ΔK and, in particular Kmax (and/or σref,max) as Kc (and/or plastic collapse) is
approached. Depending on the deformation and fracture characteristics of the
material, crack growth rates can be strongly influenced by size and geometry. In
these circumstances, a simple LEFM defined ΔK is not the most effective
correlating parameter and alternative energy based cyclic loading parameters such
as ΔJ or ΔKeq are employed in equation (1) [2, 3]. In addition to the factors already
mentioned, da/dN in the high-ΔK regime is strongly dependent on microstructure,
mean stress, temperature, environment and frequency (strain rate).
At elevated temperatures, the da/dN(ΔK) diagram may be alternatively split
into two crack growth regimes (Figure 1) [1]. In the low strain fatigue crack
growth (LSFCG) regime, load/displacement transients result in linear elastic
loading cycles and low to mid-ΔK crack growth rates for which ΔK (or ΔKeff)
still provides the most appropriate correlating parameter. Load/displacement
107
transients responsible for cyclic loading involving a degree of general yield (in
particular in tension) are referred to as high strain fatigue crack growth (HSFCG)
cycles. HSFCG rates are due to apparent ΔKs and are influenced by whether
deformation is load or displacement controlled, in particular in the material’s
creep regime. The overlap shown between the LSFCG and HSFCG regimes in
Figure 1 is due to the fact that higher ΔKs can be generated under linear elastic
conditions with displacement controlled loading because of shakedown into
compression.
Fatigue crack growth rates are increasingly sensitive to frequency with
increasing temperature.
1E-6
da/dN F(15MPa√m)T, m/cycle
1E-7
1CrMoV
9CrMoVNb
IN738LC
1E-8
3
(E20/ET) .(da/dNF)20
1E-9
0 200 400 600 800 1000
o
TEMPERATURE, C
Figure 2. Effect of temperature on high frequency cyclic crack growth rate [4].
2.2 High frequency
At loading frequencies above those for which time dependent effects are
influential (i.e. at elevated temperatures where oxidation is not responsible for
significant crack tip oxide-wedging and/or creep damage enhancement), the
factors affecting da/dN(ΔK) behaviour are summarised in the previous section.
For such conditions, the effect of temperature may be quantified in the mid-ΔK
regime by rewriting equation (1) to give:
(2)
108
where A20 is equal to A(T,υ) at ambient temperature in Equation 1. The

effectiveness of this approximation is shown in Figure 2 for a number of high
temperature materials [4]. It is clear that even at relatively high frequencies,
there is a temperature (dependent on material) above which da/dN becomes
increasingly influenced by time dependent thermally activated processes.
Figure 3. Effect of frequency on high temperature LSFCG rates in 1CrMoV steel [5].
2.3 Low frequency
At lower frequencies, oxidation and creep interaction effects become in-

creasingly more influential at high temperatures (e.g. Figure 3). In load control,
below a limiting frequency, crack growth rates may be regarded as being
dominated by time dependent crack growth mechanisms (e.g. creep crack
growth in Figure 4). The high temperature behaviour shown in Figure 4 is
typical for many engineering alloys subject to load controlled cycling in the mid-
ΔK regime.
In contrast, crack growth rates in the low-ΔK regime can reduce and ΔKth
values increase with decreasing frequency due to oxide wedging and premature
crack closure (Figure 3).
At high temperatures, the high growth rates associated with the high-ΔK or
HSFCG regimes (Figure 1) can be generated either as a consequence of
109
relatively high magnitude cyclic loading applied remotely to a long crack (e.g.
[3, 6]), or (more usually at an initial stage of thermal fatigue crack development)
as high strain transients applied locally to a small crack contained in the cyclic
plastic strain field (e.g. [7], see section 3). In the former case, cyclic crack
growth rate behaviour is modelled using a modified form of equation (1), e.g.
da/dN = A(T,υ).(ΔK)m (3)
where ΔKeq is ΔKeff for purely elastic loading. Acknowledgement of the
dependence of A on hold time reflects the potential dependence of this parameter
on associated oxidation as well as creep damage [3].
1E-4
(da/ dt )C. dt/ dN

da /dN(30MPa√m ) , m/cycle
1E-5 CREEP
CRACK
GROWTH
FATIGUE CRACK
GROWTH REGIME
1E-6 CREEP-
FATIGUE 9CrMoVNb, E911,
CRACK GROWTH 600oC
IN738LC, 850oC
1E-7
1E-5 1E-4 1E-3 1E-2 1E-1 1E+0 1E+1 1E+2
FREQUENCY, dN /dt , Hz
Figure 4. Effect of frequency on high temperature LSFCG rates in various alloys [4].
The high temperature crack growth properties required for the defect assessment
of components subject to fatigue cycles involving hold (steady operating)
periods may be derived from pure fatigue and pure creep crack growth rate data
in a construction of the form given in Figure 4 when the loading is directly
applied (i.e. load controlled). In such circumstances, the effective frequency may
be simply determined from a knowledge of the total cycle time (i.e. transient +
hold time). Alternatively, creep-fatigue crack growth behaviour is analytically
modelled on the basis of fatigue and creep crack growth rate characteristics for
the material, i.e.
(da/dN)total = (da/dN)F + (da/dN)C (4)
110
In equation (4), (da/dN)F is given by equation (3), where A(T,υ,th) may be

influenced by creep and oxidation damage through its dependence on frequency
and prior hold time [3] and (da/dN)C is given by equation (5).
(5)
Any enhancement of the total growth rate per cycle due to creep-fatigue-
oxidation interaction is covered by A(T,υ,th) which is determined experimentally
(e.g. Figure 5).
Figure 5. Long crack cyclic/hold creep-fatigue crack growth test data for 2¼CrMo cast
turbine steel at 538/565°C [3].
3. Short crack growth

The HSFCG rates associated with small cracks contained in local cyclic plastic
strain fields (typically a ≤ 5 mm) are most effectively modelled as a function of
Δε [7–11] (e.g. Figure 6), i.e.
(6)
111
For relatively short cracks contained in high cyclic plastic strain fields,
(da/dN)total is effectively modelled using a refinement of equation (6) [7, 10, 11],
(7)
1E-1
ε
th
th=16h
t
da/dN, mm/cycle
1E-2 th=0.5h
th=0
1E-3
o
Cast 1¼CrMoV, 550 C
Δa = 0.5mm
1E-4
0.1 1 10
Δε, %
Figure 6. Comparison of crack growth rates after 0.5 mm crack extension from notch root
in large SENB feature specimen creep. fatigue tests on cast 1¼CrMoV steel at 550°C [7].
The effect of hold time on HSFCG rate for a cast 1¼CrMoV steel at 550°C is
shown in Figure 6. In this example, DC was modelled as a function of hold time,
i.e. equation (8), but in a formal assessment would be determined in terms of
time fraction or ductility exhaustion (e.g. equation (11), section 4.3).
(8)
where α1 and α2 are empirical constants.

It is evident that the material property data required for the defect assessment
of high temperature components subject to creep-fatigue loading can be strongly
dependent on the specific operating conditions relating to the practical
application under consideration. This means that the rigorous creep-fatigue
defect assessment of a component can require a significant investment in the
determination of appropriate material property data. It is therefore important to
demonstrate that both creep and fatigue loading are significant at the critical
feature to be assessed.
112
4. Defect assessment
4.1 Significant creep and fatigue
A creep-fatigue crack growth assessment is only necessary when both creep and
fatigue are shown to be significant [12].
Creep is significant if the sum of the ratios of hold time to maximum
allowable time at the reference temperature, for the total number of cycles, is
greater than or equal to unity, i.e.
(9)
The values of tm depend on material, crack size and temperature [13]. For
materials with εu ≥10%, tm is taken to be the time required to achieve an
accumulated creep strain of 0.2% at a stress level equal to the reference stress.
Alternatively for εu > 10%, tm is determined on the basis of a creep strain of 0.2.εu.
Fatigue is significant if cyclic loading influences the development of creep
damage. This is likely if the elastic range exceeds the sum of the steady state
creep stress and the stress to cause yield at the other extreme of the cycle [12].
Fatigue is also considered significant if the estimated crack growth due to cyclic
loading exceeds 10% of the calculated creep crack growth.
4.2 Creep-fatigue interaction
When both creep and cyclic loading are shown to be significant, the extent of
creep-fatigue interaction should be determined. As a generality, the effect of
creep damage on fatigue crack growth rates has little influence on the total crack
growth per cycle provided the latter includes an explicit calculation of creep
crack growth (i.e. equation (4)). In such circumstances, there is no creep-fatigue
interaction and no requirement to enhance fatigue crack growth rates. It is only
necessary to consider a creep-fatigue interaction when the effect of cyclic
loading on creep is shown to be significant despite fatigue crack growth having
been estimated to be only a small fraction of the total crack growth per cycle.
For such conditions, the constants in equation (3) should be determined from
tests with hold times relevant to the service application being assessed (e.g. [3]).
Similarly, in cases where cracks are propagated by fatigue through material
heavily damaged by prior creep, propagation rates are likely to be increased. In
these circumstances, a factor should be applied to the fatigue crack growth
113
constant to account for the amount of prior creep damage. This should be
determined experimentally [13].
a > rp
a < rp
da
(Δε ,T ,υ, th )
da
(ΔK eq,T ,υ,th ) + 1 . da (C *,T )
dN υ dt
dN
Figure 7. Crack growth regimes requiring Method-I and Method-II assessment

approaches.
4.3 Assessment methods
The type of cyclic crack growth data required for high temperature defect
assessment depends on the size of the defect relative to the cyclic plastic zone at
the surface of the component. For cracks larger than the cyclic plastic zone
(Figure 7), a power law modified for crack closure is used (e.g. equation (3) in
conjunction with equation (4)). This approach is referred to as Method-I in [12].
For small defects embedded in the cyclic plastic zone (Figure 7), a strain based
method for the calculation of crack growth is adopted. This approach is referred
to as Method-II [12].
The Method-I approach is applied when the crack length exceeds the cyclic
plastic zone size at the surface of the component (i.e. a > rp). In this regime, the
total crack growth per cycle is determined using equation (4), Figure 7. The
fatigue crack growth rate is given by equation (3) with the constants modified
for hold time effects if creep-fatigue interactions are shown to be significant. If
fatigue crack growth is shown to be insignificant, the high temperature defect
assessment procedure focuses on the use of equation (4) without (da/dN)F [12].
Method-II is applied to cracks growing inside the cyclic plastic zone at the
component surface when creep is significant (Figure 7). In this regime, a high
strain creep-fatigue crack growth law such as equation (7) is used. is the
total surface creep damage (taking account of stress state if necessary)
accumulated in the appropriate time from every cycle, and is:
(10)
114
where is the creep damage accumulated in the jth cycle and the summation
is carried out over the appropriate time. is evaluated at the surface of the
uncracked component, for example by a ductility exhaustion calculation, i.e.
(11)
where is the instantaneous equivalent creep strain rate during the hold time
and is the creep ductility at that strain rate, accounting for stress. is
evaluated at the instantaneous stress during the hold time obtained from stress
relaxation data.
When , equation (7) predicts an infinite crack growth rate. However,
this should not be interpreted as predicting failure of the component. This
corresponds to the exhaustion of creep ductility at the surface of the component
and the instantaneous crack depth, a, should be set to the depth of the cyclic
plastic zone, rp. If rp is greater than the crack depth that the structure can safely
tolerate under service and overload conditions then remedial action should be
taken. Cracks deeper than rp are subjected to nominally cyclic elastic
deformation and the Method-I approach should be followed.
5. Concluding remarks
The crack growth rate data used for the defect assessment of high temperature
structures has been reviewed.
The material property parameters necessary for the defect assessment of such
components subject to creep-fatigue loading can be strongly dependent on the
specific operating conditions relating to the practical application under
consideration. The rigorous creep-fatigue defect assessment of a component can
therefore require a significant investment in the determination of appropriate
material property data.
References
1. Holdsworth, S. R. High temperature fatigue crack growth. In: Marriott, J. B. (Ed.) High
Temperature Crack Growth in Steam Turbine Materials. CEC COST Monograph
EUR 14678EN, 1994. Pp. 129–176.
2. Dowling, N. E. Crack growth during low cycle fatigue. In: Cyclic Stress-Strain and
Plastic Deformation Aspects of Fatigue Crack Growth. ASTM STP 637, 1977.
Pp. 97–121.
115
3. Holdsworth, S. R. Factors influencing high temperature HSF crack growth rates in

turbine casting steels. In: Ainsworth, R. A. & Skelton, R. P. (Eds.). Behaviour of
Defects at High Temperatures. ESIS-15. MEP, London, 1993. Pp. 327–349.
4. Holdsworth, S. R. Material data requirements for assessing defect integrity at high

temperatures. In: Strang, A. (Ed.). Materials for High Temperature Power
Generation and Process Plant Applications. IOM, 2000. Pp. 177–197.
5. Skelton, R. P. & Haigh, J. R. Fatigue crack growth rates and thresholds in steels under
oxidising conditions. Mat. Sci. Engng., 1978. Vol. 36, pp. 133–137.
6. Skelton, R. P., Beech, S. M., Holdsworth, S. R., Neate, G. J., Miller, D. A. & Priest, R.
H. Round robin tests on creep-fatigue crack growth in a ferritic steel at 550°C.
In: Ainsworth, R. A. & Skelton, R. P. (Eds.). Behaviour of Defects at High
Temperatures. ESIS-15. MEP, London, 1993. Pp. 299–325.
7. Holdsworth, S. R. Creep-fatigue crack growth from a stress concentration. Materials at

High Temperatures, 1998. Vol. 15, No. 2, pp. 111–116.
8. Skelton, R. P. Growth of short cracks during high strain fatigue cycling. In: Low Cycle Fatigue
and Life Prediction. ASTM STP 770. ASTM, Philadelphia, 1982. Pp. 337–381.
9. Levaillant, C. & Pineau, A. Assessment of high temperature low cycle fatigue life of
austenitic stainless steels using intergranular damage as a correlating
parameter. In: Ainsworth, R. A. & Skelton, R. P. (Eds.). Behaviour of Defects at
High Temperatures. ESIS-15. MEP, London, 1982. Pp. 169–193.
10. Skelton, R. P. Application of small specimen crack growth data to engineering

components at high temperatures: A Review. In: Low Cycle Fatigue. ASTM STP
942. ASTM, 1988. Pp. 209–235.
11. Skelton, R. P. Damage factors during high temperature fatigue crack growth. In:
Ainsworth, R. A. & Skelton, R. P. (Eds.). Behaviour of Defects at High
Temperatures. ESIS-15. MEP, London, 1993. Pp. 191–218.
12. R5, 2003. Assessment procedures for the high temperature response of structures.
Procedure R5, Issue 3. British Energy, Gloucester, UK.
13. BS-7910. 1999. Guidance on methods for assessing the acceptability of flaws in
metallic structures. British Standards Institution.
116
Life assessment and extension for gas turbine blades
Life assessment and extension for

gas turbine blades
Pertti Auerkari, Jorma Salonen, Juhani Rantala

Sari Mäkinen
Ilkka Karvonen, Heikki Tanttari
Lappeenrannan Energia, Lappeenranta, Finland
Pekka Kangas
Neste Oil, Kilpilahti, Finland
Jukka Taattola
FIAF, Tampere, Finland
Abstract
Gas turbine blades operate under severe loading conditions, and for satisfactory
service must be designed to withstand multiple damage mechanisms. On the
other hand, if the assumptions on the future service are sufficiently conservative
or actual service conditions more benign than foreseen, there is potential for
blade life extension. This generally does not justify fully omitting intermediate
inspections, as intervening damage by e.g. overheating or foreign objects is not
excluded, and restoring coatings or other repairs could be needed. Nevertheless,
a significant extension to the time of final replacement can be possible. In this
work, example cases are used to highlight the technical challenges and results of
blade life extension. Such aspects include assessment of in-service stress and
temperature histories, interpretation of condition assessment, and materials
performance. The example cases refer to ex-service blade samples extracted at
the time approximately corresponding to the nominal blade life as defined by the
engine supplier. Both land based and aircraft turbines are considered, and in
117
spite of the differences in design and service profiles, similarities can be pointed
out. The results for the example cases generally suggest significant potential for
life extension, for reasons related to the case-specific service histories.
1. Introduction
For the assumed service conditions, the design of gas turbines largely relies on
the available information of type of service, fuels and customer expectations. As
the actual service history will only be roughly known and the turbine will be in
reality designed for a range of possible service histories, the machine and its
major components could have considerable life extension potential if used under
less severe conditions than assumed. On the other hand, new turbines tend to be
“prototypes” in the sense of including new features even when compared to
nominally similar “sister” machines. Also, during long term service some
changes can be expected from the new opportunities created by the technical
development during the life cycle of the turbine. Therefore, without additional
effort the true remaining life of a particular component is not well established
when approaching its (nominal) replacement limits. The first (hottest) rows of
turbine blades will usually have the shortest cycles of reconditioning and
renewal, with frequent priority in life assessment and maintenance. The last
blade rows require less frequent attention but will also have a finite life.
Although less intricate in geometry than in the first rows, blades of later rows are
also not particularly cheap to replace as they are much larger in size than those
in the first rows. There may be limited information on design, material and
service history, and multiple damage mechanisms can affect the blade life.
However, at this stage there is accumulated information on the service history,
and also the blade condition can be evaluated in detail [1–3]. This work will
review selected example cases on life assessment of ex-service gas turbine
blades, particularly when approaching the nominal or suggested life limits. The
examples cover different machine types, materials and design traditions, to
explore the life extension potential.
2. Ex-service turbine blades

Principal features of the four example blade types are shown in Table 1.
118
Table 1. Features of example blades; eoh = equivalent operating hours.
Case/feature 1 2 3 4
Turbine type W251 V64.3 F6A (Aircraft)
Operation CHP CHP CHP Flights
Time in service 88 000 eoh 109 000 eoh 90 000 eoh 2 300 eoh
Row/no. of rows 2/3 4/4 3/3 1/2
1)
Blade material Udimet 500 Udimet 720 GTD 111 Nimonic
Coating None None Chromized Aluminized
2)
Cooling No No No Air
1) contradicting OEM documentation; 2) air cooling in blade root only
One to two blades from each turbine were sacrificed for detailed assessment. For
reasonably conservative conclusions, sampling aimed to select blades that were
dimensionally typical but otherwise by visual and other non-destructive
inspection among the worst blades within each blade set.
3. Methods of assessment
The sample blades were subjected to non-destructive testing (NDT) including
visual (VT), stereo microscopy, fluorescent penetrant inspection (FPI) and eddy
current (ET) methods, to detect surface cracks or comparable defects. Surface
roughness was measured with contacting line profiling for blade cases 1−3. The
blades were sectioned to assess the internal condition, typically at the blade root,
at 55−60% of the airfoil length, i.e. in the hottest region estimated from external
colouring, and close to the tip of the blade. The sections were inspected by light
optical and scanning electron microscopy (SEM). X-ray spectrum analysis and
EDS were used for chemical analysis. The materials properties, like the as-new
mechanical strength at the operating temperatures, were obtained for the case 1
blade (U500) from [4] and for case 2 blade (U720) from the materials database
of TU Darmstadt [5]. For case 3, GTD 111 properties were assumed to comply
with those included in [1]. For case 4, short term creep testing at 950°C and a
selected test stress was used to support the assessment through comparison with
the performance of new material. The blade and material condition was assumed
to be sufficiently described by the service history, observations in the non-
destructive and microstructural assessments, and other available information on
119
the blades and their damage. Life assessment was mainly based on simple life
fraction (creep and fatigue) analyses on the results, although the likely impact of
other potential damage mechanisms was also assessed.
4. Results
An overview of the results of non-destructive inspections and other surface
assessment of the blades is shown in Table 2.
Comparison of ex-service and as-new (when available) blades showed that the
differences between individual blades were sufficiently large to mask any
possible systematic in-service change in the blade dimensions. Nevertheless, no
obvious strain or distortion was observed in any of the blades. The chemical
analysis of the blade materials confirmed the expected compositions. In general,
most extensive service-related changes in the aerofoil cross-sections appear at or
close to the leading and trailing edges. The observed features reflect the
influence of the flue gas temperature, and in cool regions possibly also of the
heat treatments.
Table 2. Observations in NDT and surface assessment of the blade cases.
Feature/case Case 1 Case 2 Case 3 Case 4

Surface colour Dark grey Light greenish Light grey Blackish
Cracks 1) None (FPI, ET) None (FPI, ET) None (FPI, ET) None (VT)
2)
Roughness Ra < 6 µm Ra < 5 µm Ra < 4 µm -
3)
Impact size Max 1 mm Max 0.5 mm Max 0.5 mm -
Tip condition Modest wear Small dent Modest wear Modest wear
1) As indicated in NDT; 2) surface at leading edge; 3) impacted particles
Where more than one blade was sectioned (cases 2 and 3), there was very little
difference between the sample blades. No clearly defective features related to
e.g. fabrication were observed in the blades or their inspected sections.
Examples of the cross-section microstructures at the hot section of the case 3
blade airfoil are shown in Figures 2 and 3. The evaluated thickness of degraded
surface and the mean size of the gamma’ particles are summarised in Table 3.
120
Table 3. Features observed from the hot airfoil and root sections.
Feature Case 1 Case 2 Case 3 Case 4

1)
Surface layer ≤ 0.1 / 0.3 mm ≤ 0.1 / 0.2 mm ≤ 0.1 / 0.2 mm < 0.05 mm
2) 3) 3)
Size of γ‘ 0.25 / 0.20 μm 0.3–0.4 μm Double peak -
1) Oxidised + transformed material on leading/trailing edge, hot sections
2) Mean size, hot section / fir tree (root section)
3) As in the root section or otherwise in practically as-new state
Using the classical model for thermally activated coarsening of second phase
particles (gamma’) with a radius growing from r0 to r:
B ⋅t Q
r 3 − r03 = ⋅ exp( − ) (1)
T RT
where t is time, T absolute temperature, Q the apparent activation energy of
coarsening, R the gas constant and B a material-dependent constant [1, 6]. From
known time of exposure t, characteristic (mean) particle radius r and material
specific constants Q and B, the effective temperature T can be solved, provided
that there is sufficient change in r.
Comparison of the microstructures at the hot airfoil cross-section and at the
blade root regions shows the extent of thermal coarsening that has taken place in
the airfoil (Table 3). Applying the equation (1) on the gamma’ particles with the
mean size from the aerofoil section as r and that from the root section as r0, the
resulting estimate of the effective material temperature at the hot section is
obtained and shown in Table 4. For the blades of case 2, the microstructure at
the 60% cross-section of the aerofoil showed no significant change between the
airfoil and root section materials, indicating that the low service temperature has
rendered the approach based on particle coarsening inapplicable. Therefore, the
estimate of the blade material temperature was based on the measured flue gas
inlet and outlet temperatures that must limit the temperatures of uncooled blades.
For the blades of case 3, the microstructure of the airfoil cross-section showed
only small differences between blade samples at equivalent sections, and in all
airfoil sections the original two-peak size distribution and other as-new features
of the gamma’ particles were still clearly observable. Again, the differences in
particle coarsening were judged inadequate for assessing the effective material
temperatures, the estimates of the blade material temperatures were based on the
flue gas inlet and outlet temperatures.
121
Figure 1. Cross-section of the airfoil leading edge (case 3).
Figure 2. Blade microstructure, hot cross-section of the airfoil (case 3).
122
Table 4. Estimated maximum temperatures at the blade hot sections.

Location 60% lead edge 60% lead edge 55% lead edge (Hot airfoil)
Max temp 765°C ≤ 600°C ≤ 700°C < 900°C
Est. from γ‘ coarsening Flue gas temp Flue gas temp User info
Under steady base load service, the static tensile stress from the centrifugal mass
loads of the blade may dominate. This stress at a blade radius r [2, 4]
ρ ⋅ω 2 ra
σz = ⋅ ∫ A ⋅ r ⋅ dr (2)
A r
where ra is the tip radius of the blade, A the airfoil cross-section area at radius r,
ω the angular velocity and ρ density of the blade material. As the level of this
stress increases towards the root of the aerofoil, and material becomes weaker
towards the location of the maximum temperature, it is the combined influence
of mechanical loading and temperature-dependent material strength that
determines the location of critical position along the airfoil. To account for the
impact of the additional stress components and other uncertainties, a safe side
approach could use a stress assessed for a cross-section towards the root of the
airfoil and a temperature assessed for a location closer to the observed
maximum. The estimates of the static stress levels are summarised in Table 5.
Table 5. Estimated static stress levels at the blade cross-sections of interest.

Rpm 4 918 5 400 5 150 -
Location 60%/root 60%/root 55%/root Min. section
1)
Stress 138/184 MPa - /321 MPa 216/357 MPa 94 MPa 2)
1) static, at the selected hot section / root of the airfoil; 2) test stress
123
5. Discussion
When evaluating the results from inspections, other measurements and
assessments, it is important to consider all potential damage mechanisms that
could significantly influence the blade life. Such an overview of the blade
condition is shown in Table 6 for the example cases.
Table 6. Estimated impact of the potential damage mechanisms on the example blades.
Mechanism Case 1 Case 2 Case 3 Case 4

Overheating Minor None None Minor
Overloading None None None None
Wear/erosion Minor Minor Minor Minor
FOD 1) Minor Minor Minor Minor
Corrosion None None None None
Oxidation Modest Minor Modest Modest
Embrittlement Minor None None Minor
2) 2) 2)
Fatigue Minor Minor Minor Modest
3) 3)
Creep Life limiting Minimal Minimal Life limiting
1) FOD = Foreign object damage
2) Potential lead mechanism for cases 2 and 3
3) Hot sections / high temperatures only
In general, only few indications of degradation were observed, mainly some

modest oxidation, wear and impact traces on the blade surfaces, and some
coarsening of the gamma’ microstructure at the hot section, particularly in case 1
blade. Based on the evidence, it was judged likely that creep could be the life
limiting damage mechanism for case 1 and 4 blades, and fatigue for case 2 and
case 3 blades, although there was as yet no evidence of progressing fatigue
damage or cracking. This would only apply if no short term mechanisms like
foreign object damage (FOD), overheating or overloading, did not intervene. For
the blades of case 2, it is also possible that attachment clearances and fretting
can require mitigating action. The outcome of the resulting life assessment is
summarised in Table 7, and for cases 1−3 in Figure 3.
124
Table 7. Comparison of the blade cases including the predicted safe additional life.
Issue/prediction Case 1 Case 2 Case 3 Case 4

Location 60% / root 60% / root 55% / root Airfoil
Lead mechanism Creep Fatigue? Fatigue? Creep
Added life 25 000 h > 100 000 h 90 000 h ~2 000 h
The predicted residual life appears considerable in all cases, suggesting that the
example blades can allow further operation while replacement had been
recommended by OEM or the maintenance service provider. Particularly in
cases where the operating temperature is low, the impact from most potential
damage mechanisms seems to be modest or minimal after reaching the nominal
(recommended) service life. Inversely, increasing temperatures will boost the
rate of life consumption through several damage mechanisms. It is also to be
remembered that the operator, control system and the maintenance procedures
must remain alert to short term events and mechanisms that could intervene and
in the worst case terminate the blade life abruptly. For example, overheating,
verloading, foreign or internal objects, or control failure could result in dramatic
loss of the residual life.
Nevertheless, the results from the example cases suggest significant potential
for life extension and related savings, when the service conditions are not too
severe. The benefits from base load operation, clean natural gas firing, and
proper maintenance are likely to guarantee such benefits. The opposite may be
true for new cheap fuels, increased cycling in service, and less vigilant attitudes
towards operation and maintenance.
6. Summary
Four example cases of ex-service gas turbine blades have been subjected to life
assessment at or close to the nominal or suggested life limit. The examples cover
a range of machine types, materials, designs and service conditions with
foreseen interest and potential in life extension.
125
100
80
Failure
Creep life fraction (%)
60
Micro-
40 1 cracking
20
Internal 3
damage
2
0
0 20 40 60 80 100
Fatigue life fraction (%)
Figure 3. Approximate position of the blade cases 1–3 on a creep-fatigue life fraction map.
The results suggest that the life of the example blades can be significantly
extended from the nominal or recommended limits in all cases. For land-based
turbines, the likely reasons for the positive outcome include base load operation,
clean natural gas fuel, modest operating temperatures, and relatively
conservative design for these blades. In case of the aircraft turbine example,
additional experimental support was needed to confirm the life extension
potential. However, life extension will also require that to avoid intervening
damage or cracking, proper condition monitoring, inspection and maintenance
programs must be continued. Some vigilance is needed to monitor the progress
of multiple damage mechanisms and their combinations that can threaten the
blade life. This would be particularly important if the operational conditions
were to be modified, e.g. by becoming more cyclic in future.
126
Acknowledgement
This work was a part of the European COST 538 project, and the national
project LC-Power/LifePower. The authors wish to acknowledge the financial
and technical support by the Finnish Funding Agency for Technology and
Innovation (Tekes), the national industry group of the projects, and VTT. The
expert technical help by Alfred Scholz (TU Darmstadt, Germany) and Erica
Vacchieri (Ansaldo Richerche, Italy) in the COST 538 project is also much
appreciated.
References
1. Viswanathan, R. Damage mechanisms and life assessment of high-temperature
components. ASM International, 1989. 497 p.
2. Reed, R. C. Superalloys – fundamentals and applications. Cambridge University Press,

2006. 372 p.
3. Bohl, W. Strömungsmaschinen 2 – Berechnung und Kalkulation. Würzburg, Vogel

Buchverlag, 2005. 286 p.
4. NRIM Creep Data Sheet 34B. NRIM, Tokyo, 1993. 62 p.
5. Auerkari, P., Salonen, J., Scholz, A. & Vacchieri, E. Intermediate report of COST 538
project, 2007. 8 p.
6. Auerkari, P., Salonen, J., Gustafsson, T. E., Karvonen, I., Tanttari, H. & Mäkinen, S.
Evaluation of the operating conditions of gas turbine blades and vanes after long
term service. Baltica VII – Life Management and Maintenance for Power Plants,
Helsinki–Stockholm 12–14 June, 2007. VTT Symposium 247, Vol. 2, pp. 25–34.
127
Fire incidents in underground coal storage
Juha Sipilä and Pertti Auerkari*

*VTT Technical Research Centre of Finland
Espoo, Finland
Abstract
Autoignition of coal in underground storage facilities is reviewed in the light of

recent incidents of smouldering fires. The opportunities to prevent and
extinguish closed-space fires are also considered considering the efficiency of
the available alternatives. The challenges of such fires are highlighted by an
example case, describing the observations and outcome of a smouldering coal
storage fire. The principles of self-heating and most critical factors in
spontaneous combustion such as the condition and quality of coal are fairly well
known, but usually only provide partial help in fire prevention. The experience
suggests that nitrogen injection can be useful for extinguishing controllable fires,
while more extensive measures may be needed against uncontrolled fires. In
general, efficient fire extinction during plant operation can be challenging, as
any air ingress tends to feed the fire. Methods and indicators for predicting and
detecting the ignition are discussed, and improvements are suggested to enhance
the storage and plant availability. Also, new experience on dealing with frozen
coal is discussed in brief.
1. Introduction: coal storage fires

Likelihood of autoignition, or self-initiated fire of coal without external heat
source, depends partly on the coal properties (Tables 1 and 2). Also, the
boundaries of coal from different origins tend to show increased sensitivity to
autoignition in storage. This will not happen without oxygen, but air ingress is
128
often difficult to exclude. The heat from oxidation increases the local
temperature and the reaction rate. Warm coal when entering storage will reduce
the heating step to autoignition. Also, fractionation in transport and entrance to
the storage, and leakage water may create air channels for easier oxidation.
Mainly bituminous coals are considered here (Table 1).
Coal oxidation is slow, when temperature is below about 40–50°C. From
about 50–80°C the reaction rate can remain relatively steady but is able to
remove humidity. At still higher temperatures the reaction rate progressively
increases, and finally at about 200–250°C, self-sustaining combustion can be
expected. However, a layer of coal can be an effective obstacle for heat transfer,
and then the critical temperature for self-sustained reaction can be 70–80°C for
bituminous coal [1–3]. The factors affecting autoignition are described in brief
in Table 2. Depending on coal type, the likelihood of such ignition is highest
about 3–6 months after entering storage. Uncompacted coal containing coarse
particles can ignite within 1 to 1.5 months. Complete oxidation of coal carbon to
carbon dioxide will produce 252–504 kJ/mol of thermal energy. The oxidation
rate of coal increases approximately by a factor of ten when temperature rises
from 30°C to 100°C. The rate is reduced during oxidation that will form a partly
protective (absorbed) layer on the coal particles. The final gaseous oxidation
products are typically released as carbon dioxide, carbon monoxide and water or
steam [4, 5].
Table 1. Typical ranges of properties for coal used in pulverised coal firing.
Property Typical range Notes

Water content 8–12% -
Ash content 10–20% Residual minerals etc. (unburnt fraction)
Volatiles 27–36% E.g. hydrocarbons
Sulphur content 0.2–1.2% Depends on local acceptance & cleaning system
Particle size 0–50 mm As-delivered and in storage before coal mills
Heating value ~ 25 000 kJ/kg Approximate basis level
Hardgrove index 0–90 Describes ease of crushing e.g. in coal mills
129
Table 2. Factors affecting coal autoignition; V/A = volume to surface area ratio.
Factor Effect on autoignition Notes

Coal rank Lower rank – easier autoignition E.g. safer with anthracite
Petrography High liptinite – easier autoignition High inertite – less autoignition
Volatiles More volatiles – easier ignition Usually lower rank = more volatiles
Methane More methane – reduced oxidation No ambient oxidation at > 8 m3/t
Porosity High porosity – easier autoignition Lower thermal conductivity
Particle size Small size – easier autoignition Same V/A → similar ignition temp.
Water content High content – easier autoignition Easiest for large particles, young coal
Ash/minerals High alkalis – easier oxidation Ash, borates, CaCl2 – less oxidation
Pyrite content High content – easier autoignition Significant only when > 2% pyrite
The oxidation process of solid coal can be divided into partial reactions:
C (s) + O2 (g) → chemically adsorbed “oxycoal” (1)
Oxycoal → CO2 + H2O + heat (main reaction) (2)
Oxycoal → CO + hydrocarbons + heat (side reaction) (3)
The reaction (1) takes place at low temperatures, below 80°C. The reactions (2)
and (3) are providing heat (33.4–35.1 MJ/kg) for autoignition, if sufficient
amount of air is available and heating can be maintained to sustain the reactions.
In addition to the fire risk, the adverse effects of oxidation include reduction of
the heating value of the stored coal. With sustained heating the oxidising
reaction can be intensive enough to maintain the fire, while opening and
spreading the burning coal may result in cooling and self-extinction (Figure 1).
130
Figure 1. Digging out an open storage coal fire to spreading it for self-extinction.
The underground coal storage at the Salmisaari CHP power plant has been
operational since 2004, and replaces the earlier above-ground storage. The
storage has a total capacity of 250 000 tons, which corresponds to about half of
the yearly fuel consumption of the plant. The storage consists of four silos, ∅ 40
m × 65 m each, with the silo bottom at a depth of -120 m (Figure 2). During the
relatively short operational history, smouldering fires have occurred in the silos.
Figure 2. The underground coal storage at Salmisaari power plant.
131
Coal is discharged from the silos to the boiler plant on horizontal and vertical
conveyors. These are critical components with maximum allowable operating
temperature of 70°C; the delivery time of a new vertical conveyor is about 8
months. During yearly maintenance periods the plant may be out of service for
the summer months. During this time the stored coal cannot be simply burned in
the boiler, which is otherwise common way to discard heated batches of coal.
2. Detecting coal ignition and fires

There are several (non-standard) methods to predict how likely self-heating and
autoignition are for a given coal type [6, 7]. These methods typically take into
account e.g. dry ash-free oxygen content, oxygen-carbon ratio, time of adiabatic
heating to a given limit temperature, and content of inherent water and volatiles.
The ignition can be detected from burning odours (coal tar and phenols), brown
smoke, emerging gases and increasing temperature. A reasonable goal is an
alarm of the elevated fire risk before attaining the critical temperature of 70–
80°C. A fully safe storage requires that even local temperatures are not rising
above 40°C. Symptoms of coal bed heating have been listed in Table 3. Self-
heating of coal can be indicated by monitoring the gaseous reaction products that
will emerge sequentially, with general stages of autoignition as shown in Figure
3. Normal carbon dioxide content in air (about 390 ppm) can be high enough to
disturb early detection of coal oxidation, but carbon monoxide concentration is a
more sensitive indicator. Carbon monoxide is a combustible gas that can spread
evenly in air, and dangerous to humans at 30–300 ppm concentration depending
on time of exposure. Significant amounts of CO are already formed at about
50°C, with rapid increase when approaching the critical regime of 70–80°C.
Other gaseous products like methane and hydrogen are only produced in notable
amounts above 100–150°C. These gases together with ethane, propylene and
coal dust are also potential sources of explosions.
132
Table 3. Symptoms and indications of heating in a coal storage
Temperature Indications Notes

40°C Limit temperature for autoignition Safe storage below this limit
50°C Acceleration of self-heating Still relatively slow heating
60°C Steam formation, burnt coal odour −
> 60°C Development of autoignition Heating to ignition point expected
70–80°C Alarm of autoignition Elevated fire risk
250
Carbon monoxide
200 Hydrogen
Concentration (ppm)
Ethylene
Propylene
150
100
50
0
50 100 150 200 250 300
Temperature (C)
a) b)
Figure 3. a) Stages in autoignition of coal; b) development of gaseous reaction products
at low temperature oxidation of coal (adapted from [2, 3]).
For wide free areas of coal surface, thermal cameras can be used to determine
the surface temperature gradients, and then to apply thermocouple sensors for more
detailed information. Continuous temperature measurement of coal entering storage
can be arranged from an even coal flow on belt conveyors. Any heated coal can
be directed either to combustion or to a short-term storage with enhanced
monitoring and prevention systems. On coal entering storage directly from ship,
differences up to about 20°C have been shown in the same coal batch [1].
Measuring highest local temperatures for hot spots in the main storage is more
difficult, although they are most likely to appear within 1–3 m from the surface
providing air access. A sufficiently dense network of sensors can be a challenge in
underground silos, and surface measurements may be too late to indicate deep hot
133
spots. Somewhat deeper measurements can be made using thermocouples attached

to lances that are pushed through the surface layers (1–3 m) of the stored coal.
For analysing gas compositions to detect fires early, sufficient sensitivity to
carbon monoxide and oxygen content must be available, with supporting
information on the CO2 content. Measurements of methane and hydrogen
concentrations do not provide early information on coal heating, but should be
available for assessment of the risk of dust (or hybrid) explosion. Increasing
methane and hydrogen concentrations indicate a further extended stage of the
fire as these gases appear at higher temperatures. Odour analysers could indicate
an initiating fire even earlier than traditional CO and smoke sensors.
3. Approaches to coal fire prevention

At early stages a fire can be prevented by cutting it off from access to oxygen.
This is only possible in storage facilities that can be closed and sealed to
interrupt the air flow. On the other hand, adding air flow by ventilation can
remove the heat that is being produced by the oxidation process. In underground
rock storage such advantages are unlikely, but efficient ventilation is needed for
human access. Evaporation of water from the coal surface requires heat and this
is taken from the coal. This will cool the coal bed, and increasing air flow for
ventilation will lead to coal cooling and drying, in the best case increasing the
heating value of coal.
The oxidation reaction can be stopped by preventing oxygen access so that air
oxygen is replaced by an inert gas, such as nitrogen or carbon dioxide. For fire
extinction, in practice the oxygen content must not be above 4%. The auto-
ignition risk can be in principle reduced by ensuring that there are no other fire
or heat sources in the vicinity, and that the storage is dry and clean. Emergency
transfer of coal must be made possible, and no already heated coal should be
brought to the storage. Coals with highly differing qualities and grain sizes should
be stored separately, and strongly self-heating coals should be used first [9].
An explosion can occur in flammable gases like methane, CO, hydrogen and
light hydrocarbons, in coal dust, or both together (hybrid explosion). Coal dust
can ignite as a suspended dust bed in air, or as a precipitated dust layer, with
differences in the igniting reactions. The igniting energy can be provided by a
spark or even human static discharge.
Methane originates from coal deposits and can also be formed as a
consequence of self-heating. Methane can explode at concentrations of 5–15% in
134
the presence of oxygen content higher than 12.4%. Typical coal dust with a
mean particle size of 60 μm can explode at concentrations of 40–5 000 g/m3,
when oxygen content is at least 12%, at ignition energy below 10 mJ. For
prevention, formation of coal dust can be controlled by appropriate process
design. With typical water content of crushed coal of about 8–12% the amount
of dust remains small. When necessary, water content can be increased e.g. on
the conveyor belt in transport, and the effect can be enhanced by additives that
lower surface tension. The explosion risk from methane can be much reduced by
making the storage space inert immediately when the methane content increases
to above 1%. The explosion hazard can also be reduced by monitoring CO
content and starting the inert gas injection before significant methane formation.
This approach could stop the autoignition process at the early stages before
temperature can rise to the critical regime [2, 4].
4. Fire control and extinction

The process and spreading of fire requires fuel, oxygen and heat. Removal of
any of these elements will end the fire. Extinction of coal fires in underground
(or other closed space) storage is based nearly without exception on replacement
of air (oxygen) by an inert gas [8, 10].
Carbon dioxide is not widely used for extinguishing coal fires because of the
Boudouard reaction where carbon (coal) at high temperatures will reduce a part
of carbon dioxide to flammable and toxic carbon monoxide. However, carbon
dioxide is otherwise an efficient agent to extinguish fires and is applied as an
extinguishing (flue) gas component. Nitrogen is easily available in liquid and
gaseous form, and can be transported to fire sites in large quantities. If needed,
nitrogen can also be produced locally using a nitrogen generator (pressure swing
absorption or PSA equipment). Figure 4 shows nitrogen application at the
Salmisaari power plant. Nitrogen is cold, dry and practically non-reactive in fire
extinction, generally acceptable from safety point of view, and it can be
evaporated from liquid phase up to 300 m3/min. Nitrogen is somewhat inconvenient
to store for long periods, and can only be led to the silos with purpose-built
pipelines. Nitrogen is cheaper (and more easily available) than CO2 but more
expensive than GCP [8].
Gaseous combustion products (GCP) or flue gases mostly consist of nitrogen
together with carbon dioxide. GCPs generally have low oxygen content (0–3%),
and they are quite cost effective as GCPs can be made at the site at rates up to
135
1 800 m3/min. One potential limitation is the included steam that could produce
flammable CO and hydrogen gas. Table 4 is comparing different inert gas
systems for fire extinction. Sealing off can also be an effective approach for
preventing or reducing oxygen access to a coal fire. Fly ash of coal fired plants
mixed with water (≥ 1:1 by weight) is a cost effective, easily available and well
performing sealing material against oxygen ingress.
Table 4. Characteristics of selected inert gas fire extinguishing systems [1].
System Flue gas generator1) Jet blast2) N2 generator3)

Output 0.5 m3/s 7–30 m3/s 0.001–5.0 m3/s
Exit temperature 54°C 85°C Ambient
Oxygen 2% 0.5–1.6% 0–3%
Nitrogen 82–84% 80–85% 97–100%
CO2 14–15% 13–16% -
CO 0–15 ppm 3 ppm-0.4% -
Steam 1.2% 1–2% -
Water droplets n/a Can be abundant -
1) Tomlinson boiler 2) e.g. Steamexfire or GAG 3) e.g. evaporators or pressure swing absorption
Figure 4. Nitrogen injection at the Salmisaari power station.
136
5. Case 1: smouldering fire in 2008

The case involves a smouldering fire in the Salmisaari underground coal storage
in 2008, lasting for four months. The fire was first indicated by elevated CO
levels 15th September in the silo no. 4, suggesting autoignition. This silo mainly
contained Russian low sulphur coal, with an about one meter layer of old Polish
coal at the bottom. Extinguishing measures were started by using nitrogen, but
the effort was not fully successful before discharging the silo completely.
Thermal images showed two hot spots on the silo surfaces, one next to a
maintenance door and another on nearly opposite side of the silo. At the time of
thermal imaging the fire status had already improved and the coal surface level
was only at a height of 6.3 m. At discharging, relatively strong steam flow was
observed next to the bottom maintenance door and the concrete wall [1].
Complaints from the inhabitants living in the vicinity of the plant were
received, mainly about flue gas fan noise and smell of hydrogen sulphide. The
inhabitants were asked to close doors and windows and wait for further
instructions. At the early stages of the fire, the fire alarm was followed by a
national level radio notification that turned out to be unnecessary. The
complaints and general alarm were naturally not considered good for the public
image of the plant.
No production limits were introduced during the incident, although personnel
traffic in the storage was only allowed with oxygen breathing support. The
completed fire extinguishing by alternate nitrogen filling and coal discharge took
about 4 months. At the time of discharge coal was often heated but a thin layer
of cool coal from another silo was used to protect the conveyor belt. Also, the
belt was watered if hot coal was detected coming out of the silo. The water flow
was sufficient to prevent significant formation of explosive gas mixtures.
The provider of the belt conveyor had noted that the rubber belt starts to
harden at temperatures exceeding 70°C. Visual inspection of the belt conveyor
showed no obvious damage. Rubber hardening is expected to shorten belt life,
but although hardening cannot be excluded, it has not been confirmed.
Cracking and spalling of the sprayed concrete in the silo wall was observed on
an area of about 2 500 m2, which is about one third of the silo wall area in contact
with coal when the silo is full. An estimated temperature difference of about
120°C in the adjoining layers is enough to spall the porous sprayed concrete of the
rock surface. Such a temperature difference is thought to arise already if the heated
flue gas from the fire finds a channel along the silo walls or drains, i.e. the fire
137
does not need to touch the wall. The observed 20–30 mm wide cracks in a 120 mm
thick fibre reinforced concrete suggest heavy thermal loads. The damage at the
maintenance door at the silo bottom indicated that the smouldering fire occurred
close to it (Figure 5). Partly burnt coal, ash and hard slag around the door opening
suggest that this door provided air ingress during the fire. Also, several hollow
areas with ash inside, indicated local fire against the silo wall. During fire the
measured surface temperature was 50°C at the 1.1 m thick concrete wall, and
350°C on the steel door. This was hot enough to damage the fire seal of the door.
The concrete surfaces near the silo also suffered damage near the discharge cone,
at the nitrogen injection tubes and in the ceiling of the silo. Furthermore, the
bellows of the discharge cone had nearly completely burnt away, resulting in
additional ingress of air during the fire. The future improvement should aim to
detect similar ignition earlier. There are studies suggesting that the smouldering
fires (with indicative levels of gaseous reaction products) may extend quite slowly
towards the upper surfaces and therefore real fires in silos may become extensive
before being detected. The silos should be maintained in a condition that allows
full discharging. More importantly, air tightness of the silos must be improved,
and such action has been taken e.g. by sealing the maintenance door at the bottom
of the silos. In addition, the density of nitrogen tubing should be increased in the
silos, and an improved nitrogen system appears preferable at the discharge cones.
Proper selection of coal that will be stored in the silos remains important. The
causes and consequences of autoignition and fires in the underground coal storage,
with simplified aspects of mitigation, are outlined in Figure 6.
Figure 5. Post-fire inside appearance of the bottom maintenance door opening [1].
138
Low rank coal First in, last out

Long outage
Small grains,
boundary layer
Self-heating Conveyor damage
Tight packing,
long storage
Autoignition N2 injection
Air ingress
Low control Reduced fuel value

Poor monitoring
Energy losses
Poor access
Figure 6. Causes and consequences of autoignition in underground coal storage, with

mitigation (simplified); note that reduced access is also limiting efficiency of mitigation.
6. Case 2: challenge by cold coal

The first case was about problems that may occur with heated coal. An
unexpected new challenge has been too cold frozen coal, when in February 2010
low sulphur Russian coal was received as 17 000 ton shiploads and stored into
the silos 1 and 2. This coal was mined in Siberia (perhaps at -50°C) and
transported first by rail to an intermediate storage in Vysotsk harbour (near
Viborg), then by ship to Helsinki (Salmisaari) and finally (after crushing) by
conveyors to the storage silos. At the time of offloading from ship, the
temperature in Helsinki was about -15°C. Because of relatively cold winter, coal
was apparently frozen from the mine to the silos, and did not thaw even in the
intermediate storage. The offloading was otherwise as usual but the deep frozen
coal formed abundant dust that was colouring the nearby sea ice black.
Offloading was also slow, because the coal batch in the ship was frozed solid
and contained abundant impurities like wood and steel that frequently stopped
the crusher conveyor.
When discharging the same coal from the silos, it was found to include ice
blocks with variable amounts of coal in them. As such blocks were not
introduced into the silos after coal crushing, this ice was apparently forming
139
from the groundwater seeping into the silos. During discharging from the silos,
attempts were made to thaw the frozen coal from below by using electric
heaters. In case of ice blockage at the discharge cone, automatic weighing
control reacts by increasing the rate of coal discharge to maximum, so that
uneven and at times a large amount of coal and ice was falling to the conveyor
belt. A part of this fell on the floor, but the icy chunks on the conveyor belt
could also cause yet another blockage at the discharge station of the next silo
when the chunks were too large to pass through. Such a blockage sweeped the
coal from the conveyor belt to the floor, so that both the blockages by ice chunks
and the fallen coal required frequent manual clearing. As a result of blockages,
not enough coal could be discharged from the silos 1 and 2, and manual clearing
and cleaning work was required 24 h a day.
If frozen water can expand within the silo walls (leak water drains), freezing
damage of the silo walls is possible although such a risk was considered small in
this case. Thawing of frozen coal in the storage silos will take a long time, as the
temperature in the silos is relatively low (+10…14°C), and the total mass of coal
very large, with low thermal conductivity. External heating by e.g. steam has
been considered but must be introduced with caution to avoid coal ignition.
Of the other silos at the time, no. 3 was out of service due to maintenance
work that was estimated to take several weeks, and silo no 4 only contained a
small amount of Polish coal corresponding to about 1–2 days consumption of the
plant. Reserve fuel of the plant is heavy fuel oil that could not be pumped out of
its storage tank due to low temperature (high viscosity). This was because the
water based heating circulation did not work properly, apparently because its
channels were blocked by falling rocks. To nevertheless operate the plant, four
trucks were continuously transporting additional coal from Hanasaari (about 5
km away) to Salmisaari, so that the Salmisaari plant could be operated at more
or less full rating. The situation was relieved at the end of February, after
receiving a shipload of unfrozen Polish coal to the silo no. 4.
This already suggests future measures to avoid the problem with too cold coal:
frozen coal batches should not be introduced to the storage, especially not to all
operating silos in the same time, and the groundwater drainage of the storage
could be better. Assuming gradually warming climate, the frozen coal problem
may in any case fade away in future, if harsh winters like the one in 2010 will
become increasingly uncommon.
140
7. Conclusions and recommendations

The closed space character of the underground silo storage has brought added
emerging risk that is not comparable to that for above-ground open storage.
There is no unique standard to predict the autoignition potential of given type of
coal, even for open storage. The example case of the underground storage fire
had one of the root causes in air leakage through the maintenance door frame at
the silo bottom. As the fire was limited to the vicinity of this door, it is believed
that similar events are reduced or prevented by gas tight sealing.
Some self-heating in the storage will probably occasionally occur, with
sufficiently high temperatures for re-ignition. To reduce or avoid autoignition
events, selecting and maintaining proper grade, grain size and temperature of the
stored coal remains important. Complete air tightness may not be possible when
discharging the silos but should be aimed at towards the lower parts of the silos.
The condition of the storage will be continuously monitored at multiple sites
using thermocouples and gas indicators. Nitrogen has been successfully used for
extinguishing smouldering fires, but this can take a long time to complete when
the silo is being used. From the consequences point of view, in the worst case
the fire will damage the main conveyors, disabling the storage and the power
plant during the critical winter months. The future monitoring systems must be
compatible with the process control and sufficiently tolerant to the operational
environment, to withstand corrosion, dust and humidity without requiring
regular maintenance in the storage.
Unlike heated coal that in principle is only an enhanced problem in closed
space storage, unexpected new trouble has occurred with too cold coal in the
same storage. Compounded in this case by groundwater seepage to the storage,
unavailability of coal from adjacent silos and unavailability of the alternative
reserve fuel, the resulting trouble concentrated to the time of year when the
power from the Salmisaari plant was most needed. With some difficulty,
however, plant operation was secured by continuous manual effort at the silo
discharge and conveyor systems and additional coal transport from the open
storage of a neighbouring plant. Although the challenge by frozen coal is new,
prevention and mitigation might be more straightforward than for heated coal.
141
References
1. Sipilä, J. Preventing spontaneous combustion in an underground coal storage. Espoo:
Helsinki University of Technology, 2009. Master’s Thesis. 128 p. (In Finnish).
2. Anon. Understanding self-ignition of coal. Innovative Technologies for Exploration,

Extinction and Monitoring of Coal Fires in North China. A literature study
elaborated in co-operation of Deutsche Montan Technologie GmbH (DMT) and
Federal Institute for Materials Research and Testing (BAM), 2000. 32 p.
3. Walker, S. Uncontrolled fires in coal and coal wastes. IEA Coal Research – The Clean
Coal Centre. London, 1999
4. Grossman, S. L., Davidi, S. & Cohen, H. Explosion risks during the confined storage of
bituminous coals. Fuel, 1995. Vol. 74, No. 12, pp. 1772–1775.
5. Wang, H., Dlugogorski, B. Z. & Kennedy, E. M. Coal oxidation at low temperatures:

oxygen consumption, oxidation products, reaction mechanism and kinetic
modelling. Progress in Energy and Combustion Science, 2003. Vol. 29, No. 6,
pp. 487–513.
6. Smith, A. C. & Lazzara, C. P. Spontaneous combustion of U.S. coals. R19079,

Washington, DC, USA, Department of the Interior, Bureau of Mines, 1987.
7. Smith, M. A. & Glasser, D. Spontaneous combustion of carbonaceous stockpiles. Part

II. Factors affecting the rate of the low-temperature oxidation reaction. Fuel,
2005. Vol. 84, No. 9, pp. 1161–1170.
8. Ray, S. K., Zutshi, A., Bhowmick, B. C., Sahay, N. & Singh, R. P. Fighting mine fires
using gases with particular reference to nitrogen. The Journal of The South
African Institute of Mining and Metallurgy, 2000. Vol. 100, No. 4, pp. 265–272.
9. Fierro, V., Miranda, J. L., Romero, C., Andres, J. M., Arriaga, A. & Schmal, D. Model
predictions and experimental results on self-heating prevention of stockpiled
coal. Fuel, 2001. Vol. 80, pp. 125–134.
10. Tuomisaari, M., Baroudi, D. & Latva, R. Extinguishing smouldering fires in silos.
Brandforsk project 745-961. Espoo: VTT, 1998. VTT Publications 339. 74 p.
142
Development and Implementation of i-LAS (indentation based turbine remaining life
assessment system)
Development and Implementation of i-LAS

(indentation based turbine remaining
life assessment system)
Woosung Choi*, Geewook Song and Jaeraeyang Koo

Korea Electric Power Research Institute, KEPCO
Daejeon, Republic of Korea
Abstract
Steam turbine rotors are the most critical and highly stressed components of the
steam power plant and have the highest capital cost among all the components. It
requires not only a long outage for replacement and repairing, and also be major
guideline to determine outage interval. Thus it is necessary to establish plans
with which plants can be operated and managed effectively against damage
problems. In this study, remaining life assessment tool i-LAS(indentation based
turbine remaining life assessment system) has been developed for turbine rotor
components. i-LAS is capable of evaluating remnant rotor life quickly
considering actual in-service conditions. A modified damage model was
established and successfully implemented to i-LAS system. A case study has
been carried out to validate effectiveness and reliability of i-LAS for turbine
rotor of 500 MW fossil power plants in Korea.
1. Introduction
The number of unit of 500 MW Korean standard fossil power plant is 27 in
Korea. Most of these facilities are facing a requirement for retrofit and expected
to perform a feasibility study. Feasibility assessment is to determine replacement
time and scope of facilities with consideration of the economic effects. Component
life assessment and reliability evaluation of critical facilities are performed for
143
assessment system)
feasibility assessment. Thus, it is necessary to develop advanced technology to

improve accuracy of life assessment method for more reliable feasibility study.
In this regard, a lot of methods have been developed for this reason: accelerated
degradation test, computer aided analytic method, and non-destructive technique.
Some limitations, however, have been repeatedly reported in the listed methods.
For example, accelerated degradation tests cannot be done in-situ condition
because the test method is destructive and extraction of specimens from
industrial facility in service is difficult. Computer aided analytic method such as
finite element method may reduce the cost and testing time, however, there are
uncertainties caused by various assumptions such as simplified geometry and
operation condition. Indentation test method is a non-destructive technique and
is useful to determine material properties of a component. This method can
measure various mechanical properties such as elastic modulus, tensile strength,
and residual stress by analyzing the indentation load-depth curve (Oliver &
Pharr 1992, Lee & Kwon 2006).
Mechanical properties of low alloy ferrite steel for turbine facilities are
degraded in long-term service under high temperature and rapid rotation
operation condition. In order to estimate degradation status of a component, a
portable indentation tester can be utilized for on-site measurements. Then,
remaining life can be readily evaluated using i-LAS.
In this study, remaining life assessment tool i-LAS (indentation based turbine
remaining life assessment system) has been developed for turbine rotor
components. A modified damage calculation approach has been implemented to
i-LAS making it capable of quickly evaluating rotor remnant life with actual in-
service conditions considered. Furthermore, a case study has been carried out to
validate the reliability of i-LAS for turbine rotor of 500 MW fossil power plants
of Korea.
2. Indentation based turbine remaining life

assessment system (i-LAS)
Figure 1 describes how remaining life of a component can be evaluated using
i-LAS. To obtain essential material parameters, measurements are carried out
using portable indentation tester and hardness tester. Two testers measure
parameters at two different locations: virgin area and aged area. Measured data
are then transferred to calculation module and component damage consumption
is calculated using the modified damage model. Ultimately, remaining life is
144
assessment system)
estimated based on previously calculated damage and actual operation history

information.
Figure 1. Architecture for simplified life prediction using i-LAS.
Figure 2. Introduction page of the i-LAS.
2.1 System structure
The i-LAS is constituted of information, measure, data, and life prediction

menus as shown in Figure 3. All menu and functions are developed on Windows
API basis. It is possible to import previous result from database after selecting
plant information for database saving.
145
assessment system)
Figure 3. Overview of ATLAS Menu.
2.2 Detailed description of System
2.2.1 Information
Information module basically acquires plant specific data and material data.
Basic information includes plant name, facility characteristics, and material
information. In addition, plant specific operation history such as the number of
start-up and operation hour is used in information menu. Recently, it has become
relatively easy to obtain operation condition because most of the power plants
have their independent plant information system. Obtained operation history
information is relayed to the Life module for remaining life. Information menu
window is shown in Figure 4.
Figure 4. Information menu window.
146
assessment system)
2.2.2 Measure
Measure module describes locations to be evaluated and mechanical parameters

to be obtained and displays such information on screen. A portable indentation
tester is used to achieve such data. Highly aged position and lowly aged position
are indicated in this module. The locations can be found in the reference [5–6].
Figure 5. Measure menu window.
2.2.3 Data
In this menu, test results by portable indentation and hardness tester are
imported. Elastic modulus, yield stress, and tensile strength are obtained from
indentation tester and hardness is measures by hardness tester according to the
aging states. Reference (or undamaged) values can be imported from indentation
result for in-situ virgin states or for uni-axial tensile test where virgin material is
used. Also, each value can be revised and used for stress-strain curve plotting.
Using chart composition function, undamaged and damaged properties are
compared on the chart at same scale.
147
assessment system)
Figure 6. Measured value displayed in the Data menu.
2.2.4 Life
Life menu is comprised of damage consumption part and remaining life

consumption part as shown in Figure 7. Generally, creep and fatigue damage
should be considered in accurate remnant life assessment. i-LAS procedure is
based on quantitative relationship between hardness and mechanical properties
about damage. Detailed explanations on damage evaluation algorithm are
described in reference [1–5]. In information menu, users are prompted to input
plant data: operation hour and number of start-up considering cold, warm, hot
states. In damage consumption part, total damage rates are calculated by
embedded mathematical functions which use measured mechanical properties.
According to the operation history, finally, remaining life is determined in the
form of hour or cycle. Left on Figure 7 shows a HIP (High-Intermediate Pressure)
turbine facility of 500 MW Korea standard fossil.
148
assessment system)
Figure 7. Windows of Life menu.
3. Case study
3.1 Basic information
As of 2010, 27 standard fossil power plants of 500 MW are in operation in

Korea and total power generated by the plants add up to 14 240 MW. Most of
these facilities are faced with the need for improved performance and expected
to perform a feasibility assessment. A retired 500 MW turbine rotor is selected
for the case study and the remaining life is estimated by i-LAS. Table 1 shows a
status of measured turbine facilities. This turbine facility begun its operation in
1983, and marked 141 000 hour operation and the recorded number of start-up is
191. Table 2 shows chemical composition of the turbine rotor material. Low-alloy
ferritic steel, 1Cr-1Mo-1/4V, is widely used in high temperature applications.
149
assessment system)
Table 1. Turbine facility status.
Operating time The number of Capacity

Completion date
(Hour) start-up (cycle) (MW)
1983. 12 141 000 191 500
Table 2. Chemical composition (wt.%) of 1Cr-1Mo-1/4V.
C Si Mn S P Ni Cr Mo V Sn
0.29 0.01 0.74 0.004 0.007 0.06 1.29 1.24 0.25 0.0047
Figure 8 shows stress-strain curve measured in the undamaged and damaged

location of the rotor respectively using portable indentation tester. Measure-
ments are done in different locations but in the same rotor. Detailed information
is described in reference [5–6].
1000
900
800
700
600
Stress (MPa)
500
400
300
200
100 No.1
0
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20
Strain
(a) Undamaged location
150
assessment system)
1000
900
800
700
600
Stress (MPa)
500
400
300
200
100 No.3
0
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20
Strain
(b) Damaged location
Figure 8. Stress-strain curves by portable indentation tester.
3.2 Assessment result
The damage rate is obtained by measured tensile properties and hardness for
turbine rotor of 500MW Power plant. Figure 9 shows damage rates at lowly
aged and highly aged locations of turbine rotor. Remaining life is calculated
from damage rate and operation history as shown in Table 3. According to the
assessment results, the remaining lifetime of the 16-year-old turbine rotor is
about 198 365 hour (22.6 year). When considering that the design lifetime of
large turbine facility is approximately 30 years, this result (38.6 year) appears to
be reasonable.
151
assessment system)
Figure 9. Damage rates from indentation test.
Table 3. Summary of result.
Remaining Life
Lowly aged location Highly aged location
Modified damage model 384 420 hour (43.9 year) 198 365 hour (22.6 year)
4. Conclusion
i-LAS (indentation based turbine remaining life assessment system) has been
developed to calculate remaining lifetime of major turbine facilities such as rotor
and casing. This system, called i-LAS, focuses on quickness, simplicity and
accuracy based on material characteristic, operation condition. User can easily
evaluate remaining life of turbine facilities and all information can be revised or
saved in database. A modified damage model was successfully implemented to
i-LAS system. Using developed program, a reliable life assessment result can be
easily obtained. i-LAS is applied to one of 500MW fossil power plant in Korea.
The damager rate of the 16-year-old turbine rotor is calculated by the modified
damage model. Reasonable remaining lifetime when compared to general
service time, is obtained by modified damage model.
152
assessment system)
References
1. Lemaitre, J. & Desmorat, R. Engineering Damage Mechanics. Springer, 2004.
2. Tasan, C. C., Hoefnagels, J. P. M. & Geers, M. G. D. A critical assessment of indentation-

based ductile damage quantification. Acta Materialia, 2009. Vol. 57, pp. 4957–4966.
3. Seok, C. & Koo, J. Evaluation of material degradation 1Cr-1Mo-0.25V steel by non-

destructive method. Materials Science and Engineering A, 2005. Vol. 395,
pp. 141–147.
4. Lee, K. et al. Safety assessment based on mapping of degraded mechanical

properties of materials for power plant pipeline using instrumented indentation
technique. Journal of Loss Prevention in the Process Industries, 2009. Vol. 22,
pp. 868–872.
5. Song, G., Choi, W., Koo, J. & Jinseong. Investment reduction through the life extention
of aged turbine. KEPCO TDR Report, 2009.
6. Choi, W. & Song, G. Life prediction system of turbine facilities for fossil power plant
using portable indentation tester. Patent No.1020090087051, 2009.
7. Choi, W., Kim, B., Hyun, J. & Song, G. Development and Implementation of advanced
turbine life assessment system. Power Gen Asia Track 4, 2008.
8. Choi, W. S., Fleury, E., Kim, B.-S. & Hyun, J.-S. A Life assessment for steam turbine
components based on viscoplastic analysis. Journal of Solid Mechanics and
Materials Engineering, 2008. Vol. 2, No. 4, pp. 478–486.
9. Allen, D.-J. & Fenton, S. T. A hardness based creep rupture model for new and service
aged P91 steel. Baltica VII, 2004, pp. 156– 170.
153
Creep damage and long term life of steam line components: case X20
Creep damage and long term life of steam

line components: case X20
Stefan Holmström, Pertti Auerkari, Jorma Salonen

Anssi Laukkanen, Juhani Rantala
Reino Nikkarila
Inspecta Oy, Espoo, Finland
Jarmo Junttanen
Stora Enso Oyj, Varkaus, Finland
Abstract
Hot steam lines operate at high temperatures and are designed against creep. In
service they will consume the creep life at the highest rate in locations with the
most adverse combination of material strength (weakness) and high stress (e.g.
system loads). Adverse effects in stress state are promoted by features of
geometry and discontinuities in the materials properties. In practice it is essential
to find the areas of maximum damage, as these will determine the locations and
timing to inspect and finally to repair or replace before failures or unplanned
outages. The inspection experience from power plant steam systems have
provided the established views on the expected locations of the early creep
damage, and rules on timing the next inspection. The experience has also shown that
optimal timing is materials dependent, and that the steel grade X20CrMoV11-1
(X20, 11% Cr steel) performs particularly well in the inspection statistics. This
paper describes a case with observed creep cavitation and cracking in a branch
weld of a X20 steam header that was replaced after 135 000 service hours. The
characteristics of the observed damage and its evolution are discussed.
154
1. Introduction
During creep of pipes and tubes under pressure, stresses will redistribute to
retain the maximum equivalent stresses and strains at the inner surface, while the
maximum principal stress will shift to outer surface. This is the classical
explanation for long term creep cavitation and cracking starting from the
external surface, together with damage models where the maximum tensile
principal stress will largely dominate the formation of long term creep cavitation
damage on the grain boundaries of polycrystalline metals.
Modified by details in geometry and variation in creep properties of welds, the
established principle is to inspect only the likely weakest spots (“mechanical
fuses”) of large structures from outside for condition and life assessment. In this
context life assessment really means setting the optimal time for the next
inspection, after defining possible needs (and timing) for repairs or replacements.
It has turned out that the hot steam lines of conventional thermal power plants
are designed in such a similar way that the observed damage level together with
the expired time in service is a good (albeit somewhat rough) indicator of life or
the suitable and safe time to the next inspection. This naturally assumes that the
plant and the steam line is operated and maintained to retain its preceding
characteristics, or that at least any future changes can be sufficiently quantified.
Simple guidelines exist for inspectors to classify the damage and make conclusions
based on the observations. The original guidelines for this purpose [1] were
material independent and quite conservative, but after collating more evidence
from inspection statistics, the later guidelines have been made to include
material-dependent recommendations [2]. Such guidelines that are based on long
term inspection statistics suffer from lack of similarly established rules and
recommendations for steels that only recently entered the accepted range of
standard materials and service in plant.
However, even well established steels can be considered new for long term
inspection experience. A good example is the steel X20CrMoV11-1, or X20 for
short, which is an 11%Cr steel that has been used in high temperature
components in Europe since 1960’s. The available inspection statistics suggests
that this steel and its welds suffer less creep damage after e.g. 100 000 service
hours than common low alloy steels. The reasons for this are not entirely
obvious, as the design to the same code and nominal life should be able to
compensate for the differences in creep strength between the alloys.
155
The good service and inspection experience with X20 is also of interest and
possible significance for the martensitic steels such as P91 (X10CrMoVNb9-1)
and P92 (X10CrWMoVNb9-2) that are used in new steam plants and
share many basic microstructural features with X20. For these steels no similar
extensive operational and inspection experience is yet available, but instead
some reported surprises on early creep damage [3]. A comparison of selected
steels for hot steam lines is shown in Table 1.
Table 1. Comparison of steel X20 with older low-alloy and newer 9%Cr steels [4].
Alloy Cr Other alloying Rp0.2 RT Ru/100 kh, MPa

(EN 10216-2) wt % elements MPa min 550/600°C
10CrMo9-10 2.0-2.5 Mo 280 53 / 20
14MoV6-3 0.3-0.6 Mo,V 320 87 / 41
X20CrMoV11-1 10-12.5 Mo,V 490 128 / 59
X10CrMoVNb9-1 8.0-9.5 Mo,V,Nb,N 450 166 / 94
X10CrWMoVNb9-2 8.0-9.5 W,Mo,V,Nb,N,B 440 187 / 113
Below a case is described on a X20 header that was replaced after 135 000 h of
service due to creep cracking at a branch weld. As creep cracking so early is not
common in steam line components made of this material, a fairly detailed study
was launched to clarify the case. This paper will report results on characterising
the damage originally detected in the in-service inspections of the steam line.
2. Header case: operation and inspection history

The header was designed in 1989 to the Finnish standard SFS 2618 that is in
practice compatible with e.g. the former German (TRD 300-series) design rules.
The dimensions of the main body of the header are ∅508 × 32 mm and those of
the branch to the boiler ∅310 × 40 mm. The nominal design creep life was
200 000 h, i.e. against mean stress corresponding to 200 000 h creep rupture at
540°C (129 MPa). The allowable stress was 102.3 MPa for the branch to the
boiler, at the design pressure of 12 MPa. The relative safety margin in time was
3.1 (base material mean strength). The observed maximum creep damage was in
the weld of the branch to the boiler, in the saddle point position (Figure 1).
156
a) b) c)
Figure 1. a) Section of the steam header with the main boiler branch, b) a 3D tracking
image of the header section; and c) cross-section of the damaged saddle point location
(arrow shows the location of maximum damage).
The in-service inspections were performed in 2002−2007 by the same person of

Inspecta Testing, mainly with fluorescent magnetic particle and replica testing.
The damage indications and follow-up actions are summarised in Table 2.
Table 2. Inspection history of the branch weld of the header (saddle point, fine grained
heat affected zone of the side of the header body side).
Year Observed HAZ indications Action

2002 Cavitation & microcracking (class 4) None
2003 Macroscopic cracking (50 mm max) Repair by grinding 2 mm 1)
2004 Microcracking (class 4) None
2006 Macroscopic cracking (4 mm max) None
2007 Macroscopic crack (10 mm max) Header replacement
1) class 3 damage still present after grinding
The first indications of creep cavitation and microcracking were found in the
weld (HAZ) of interest after 103 000 service hours. The end of the header
service after 135 000 service hours corresponds to 68% of the nominal design
157
(creep) life. The maximum total grinding depth at the cracked site was about
3 mm on the base material side of the header body.
3. Predicted creep strain

Of creep properties, standards such as EN 10216-2 only list mean stress to creep
rupture for a range of temperatures. To assess expected damage, creep strain can
be of more interest, to reflect the likely events well before failure. Although
creep strain is not the same as (cavitation) damage, it is thought to be generally
increasing with it. Therefore creep curves were constructed using X20 materials
data from [5–9] for the X20 parent material. The strain model used was the
logistic creep strain prediction tool (LCSP) based on time to rupture and creep
curve shape functions [10, 11]. At the design limiting stress and temperature
(about 100 MPa/540°C) the predicted base material creep behaviour is shown as
the lower curve in Figure 2. It is (as usual) assumed that the weld is 20% weaker
(in stress terms) than the parent material, and the resulting predicted creep
behaviour of the weak part of the weld (HAZ) is shown as the upper curve in
Figure 2. It is seen that at the design limit of loading, by the time of the 2002
inspection or after 103 000 h of service, the predicted creep strain at the weak
HAZ is still below 1.5%, and that this level of strain would require some
300 000 h of service for the parent material. However, the general experience
from other materials suggests that such a low strain level would not be enough
for the observed damage (microcracking or class 4 creep damage).
Therefore, either the material has been clearly weaker than average X20, or
loads (and/or temperatures) significantly higher than those assumed in design.
Since the design stress levels already include the usual safety factors, the actual
deviation from the usual operational conditions or material state have been more
extensive than those usually covered by the design and manufacturing practices.
Further insight into the indicated damage was therefore sought by a more
detailed characterisation of the cracked weld region.
158
2.0 %
1.5 %
Strain (%)
1.0 %
0.5 %
0.0 %
0 100000 200000 300000
Time (h)
Figure 2. Predicted creep strain history for X20 parent material (lower curve) and outer
heat affected zone (Type IV position, upper curve) at the design limit of the branch to the
boiler; the vertical line shows time of first inspection.
4. Features of the saddle point damage

The cross-section of the saddle point crack location is shown in Figure 3. The
corresponding location and the crack from the weld surface side is shown in
Figure 4. It is seen that the crack is here clearly a discontinuous “toilet paper”
crack, with actual short cracks and inter-crack ligaments on roughly the same
plane along the HAZ. Note that at this stage the original pipe surface of the
cracked location has been ground to a depth of more than 2 mm due to
preparation for inspections and earlier repair grinding.
Metallography of a cross-section of the main crack (Figures 5−8) shows the
appearance of the crack and cavities in front of it. The cavitation is extending at
least to a depth of 6 mm from the crack tip. All cavities are multidirectional and
jagged in shape (Figures 7 and 8). The crack tip is not strongly blunted and
appears to grow by joining creep cavities that surround the tip.
159
Figure 3. Cross-section of the saddle point crack location.
Figure 4. Appearance of the “toilet paper” crack on the header surface (axial direction of
the header body horizontal).
160
Figure 5. Cross-section of the surface crack.
Figure 6. Creep cavitation in front of the crack tip.
161
Figure 7. Creep cavitation ahead of the crack tip.
Figure 8. Creep cavities near the crack tip.
162
Such appearance of creep cavitation has been reported previously in welds of

ferritic-martensitic 9%Cr steels in addition to X20; however, most such reports
refer to observations from shorter term studies in the laboratory.
At least one reason why X20 components show relatively better performance
in resisting creep damage than low alloy steels is the intrinsic complexity of the
microstructure in the grain interior. The lath, block and packet boundaries are
likely to distribute strain and damage more effectively than simple grain
boundaries of conventional low alloy steels, even after suffering depletion of the
strengthening particles in the HAZ of welds. The intrinsic strength appears as
jagged, multidirectional shape of the cavities and as intermittent or bridging
cracks (“toilet paper cracking”), where ligaments of metal are holding the
material together at the site that was in the surface inspection interpreted as a
single unified long crack.
The shape of the crack together with the inspection history and the already
relatively long time in service indicate that the crack has been growing quite
slowly and would not have accelerated quickly if left in place for about one year.
However, the fact remains that the weld did show genuine creep cracking after
only 135 000 hours of service. The review suggests that this is not directly
because of the design of the header vessel or the welded branch. In principle
reasons for the early damage could be related to operational conditions, e.g.
additional system stresses (moments) from the thermal movement of the piping.
As no similar cracking was found on the corresponding opposite saddle point
position, asymmetry in such thermal loading is possible. However, 90° from the
saddle point position, both sides of the same HAZ showed some initial surface
cracking (about 2 mm single cracks only). The remaining possibility is that the
material was not typical average welded X20. Some deviation in heat treatments
of the base material or the weld is possible but not proven. Further studies may
clarify the remaining aspects of uncertainty, but the case can nevertheless be
used as a benchmark for interpretation of observed creep damage in welded
components after long term service. Factors that could reduce the creep life of
the component and an assessment on their significance are shown in Table 3.
Apart from the effects of single factors, also combined effects of e.g. system
stresses and material weakening are possible. This would reduce the deviation
needed in individual factors for the same impact.
Although the observed creep damage was relatively early for X20, it can be
considered long-term damage so that its growth rate can be used to support the
decisions on safe condition-based timing of inspections.
163
Table 3. Summary of potential factors resulting in relatively early header cracking
Cause Assessment results Notes

1. Design error None in the review Not common
2. Fabrication None found in dimensions Not excluded in heat treatment
3. High temp. No evidence Uncommon far from boiler
4. High pressure Unlikely Uncommon in steam lines
5. System loads Possible, extent unknown Thermal piping movement
6. Weak steel Possible but not to the extreme Evidence in microstructure
Combined Possible (e.g. 2, 5, 6) -
5. Conclusions
A X20CrMoV11-1 (X20) steam header has been retired after 135 000 hours of
service due to creep cracking of a branch weld. The expired time to cracking
(68% of the nominal design life) was considered relatively short in comparison
with the general inspection experience with X20 steam pipe welds.
The observed creep damage occurred in the saddle point HAZ (type IV) and
showed an appearance thought to be characteristic to 9–11%Cr steels with
tempered martensitic microstructure. The shape of cavities and crack bridging
suggest residual microstructural resistance to damage growth that would not
proceed similarly in e.g. low alloy steels. These features are thought to be due to
the microstructure where multiple small angle boundaries of the laths and blocks
can distribute strain and damage wider than when only grain boundaries offer
opportunities for cavity nucleation and growth. The observed relatively early
cracking is probably due system stresses and/or weaker than average material
rather than design, dimensions or operating temperature/pressure. In spite of
relatively early creep damage for X20, the observed damage growth rate can be
used to support recommendations for safe condition-based timing of inspections.
Acknowledgement
The authors wish to acknowledge the financial support of the Academy of
Finland, Decision no. 117700 (by Prof. Kim Wallin), the Värmeforsk project on
164
Creep damage development in welded X20 and P91 components, and the industry
financed Finnish project LifePower.
References
1. Neubauer, B. & Wedel, U. Restlife estimation of creeping components by means of
replicas. In: Advances in Life Prediction Methods. New York: ASME, 1983.
P. 307.
2. Auerkari, P., Holmström, S., Salonen, J., Borggreen, K., Storesund, J. & Rui, Wu.
Evaluation of creep damage from replica inspection results. Nordtest Project No.
1306-96. Espoo: VTT, 1997. Report VTT VALB211. 33 p.
3. Brett, S. J. The creep strength of weak thick section modified 9Cr forgings. Baltica V –
Condition and Life Management for Power Plants, Vol1. VTT Symposium 211,
2001. Pp. 35–44.
4. EN 10216-2, European Standard, Seamless steel tubes for pressure purposes. Part 2:
Non-alloy and alloy steel tubes with specified elevated temperature properties,
2007.
5. Storesund, J., Borggreen, K. & Weilin, Z. Creep behaviour and lifetime of large welds in
X20 CrMoV 12-1 – results based on simulation and inspection. Int. Journal of
Pressure Vessels and Piping, 2006. Vol. 83, pp. 875–883.
6. Hua-Tang, Yao et al. A review of creep analysis and design under multiaxial stress
states. Nuclear Engineering and Design, 2007. Vol. 237, pp. 1969–1986.
7. Bürgel, R. et al. Schädigungsentwicklung an X20CrMoV121 nach Langzeit-

beanspruchung und Zuordnung von Restlebensdauer von Komponenten im
Kriechbereich. VDEh und Arbeitsgemeinshaft für Hochtemperaturwerkstoffe,
13. Vortragsveranstaltung, Düsseldorf , 1990. Pp. 94–107.
8. Standard Data Tables, Vol. 2, Ferritic steels. SPRINT Specific project SP-249,
Implementation of Power Plant Component Life Assessment Technology using
a knowledge-based system, 1995.9. Entwicklung von Zeitstandschäden in den
Stählen X10CrMoVNb9-1 (P/T 91) und X20CrMoV 12-1. Untersuchungsbericht
35/1999, Bericht zum Forschungsvorhaben VGB-Nr. 160.14.
165
10 Holmström, S. & Auerkari, P. Robust prediction of full creep curves from minimal data
and time to rupture. Energy Materials; Materials Science and Engineering for
Energy Systems, 2006. Vol. 1, No. 4, pp. 249–255.
11. Holmström, S., Laukkanen, A., Rantala, J., Kolari, K., Keinänen, H. & Lehtinen, O.
Modeling and verification of creep strain and exhaustion in a welded steam
header. Journal of Pressure Vessel Technology, 2009. Vol. 131, 5 p.
166
Remaining lifetime assessment of power plant steam boilers
Remaining lifetime assessment of

power plant steam boilers
Vaclav Liska1 and Vaclav Mentl 2

Skoda Research Ltd. 1, Accredited Labs., Plzen, Czech Republic
University of West Bohemia2, Dept. Material Science and Technology,
Plzen, Czech Republic
Abstract
The energy producing power plants are designed for operational period of 20, 30
years. During this period, inspections are realized to investigate the operational
capability of the respective components and the plant as a whole, and when the
designed time is approaching its limit, the crucial questions are raised with
respect to the following possible operation, its safety and risks that stem from
the fact that the continuous degradation of material properties occurred during
the longtime service as a result of service conditions, e.g. high temperatures,
fatigue loading etc.
In opposite to the non-destructive techniques, accelerated creep to rupture tests
of high temperature boiler components, e.g. high temperature headers, can give
quantitative results as far as the remaining lifetime of the component is concerned.
Several steam turbine boilers were inspected according to the customer´s
demand to evaluate the remaining lifetime of the boilers that were operated more
than 160 000 and 200 000 hours respectively. The evaluation was based on an
extensive NDT expection and the measurement of mechanical properties
(including creep test data) of high temperature components. Making use of the
Larson-Miller parameter in comparison with replica testing made it possible to
evaluate quantitatively the lifetime exhaustion, to make an assessment of the
remaining lifetime and to make a recommendation as far as the future inspection
intervals of the boilers are concerned.
167
On the basis of accelerated creep test data performed on the degraded materials,
the remaining lifetime hours were calculated for the three „safety“ situations :
1. „ZERO SAFETY“ ( neither recommended k = 1,5 safety coefficient for
working stress nor +70°C increase of working temperature were taken into
consideration ).
2. „STRESS SAFETY“ (1,5 safety coefficient for working stress and real
working temperature were taken into consideration).
3. „FULL SAFETY“ (both 1,5 safety coefficient for working stress and
working temperature + 70°C were taken into consideration).
At the Baltica VII conference, an example of this procedure was presented [1].
This paper summarizes the results of several more boilers and the results of the
remaining lifetime calculation are discussed.
Keywords: service degradation, remaining lifetime assessment, steam

turbine boilers
1. Introduction
The creep degradation was again taken into consideration as the decisive
damaging factor in case of primary and secondary superheaters. Metallographic
investigation and basic mechanical tests were performed.
To receive quantitative results, accelerated short-term creep tests were carried
out of the creep samples machined of the during the in-site inspection selected
super-heater tubes. The creep test conditions (stresses and temperatures) were
chosen in order that the resulting times to rupture were reached within the
relatively short time to be able to evaluate the results in time according to the
requirement of the customer.
Two more real cases of creep test results and remaining lifetime evaluation in
addition to that one discussed during Baltica VII are presented. For the CASE
„1“, see [1].
168
2. Case „2“
2.1 Boiler description and technical details
• Steam output 350 t/h

• Steam pressure 6.2 MPa
• Steam temperature 485°C
• Fuel Natural Gas/Oil
• Year of Production 1976
• Service Hours 204 871
Boiler was made of common steels used in boilers produced 30–40 years ago in
1970s. The St35.8III steel was used for the tubes of water walls, economizer and
inlet primary superheater, 15Mo3 steel was used for inlet secondary superheater
tubes, 13CrMo44 for outlet primary and secondary superheater tubes.
Table 1. Creep test results.
Specimen Temp. Stress Hours to

Steel Result Location
designation [oC] [MPa] Rupture
10.1 590 110 373 Rupture Primary
10.2 590 90 817 Rupture SH
14.1 590 160 2,5 Rupture
14.2 590 110 92 Rupture
13CrMo44
14.3 590 70 1 754 Rupture Secondary
15.1 590 90 543 Rupture SH
15.2 590 110 135 Rupture
15.3 590 80 1 317 Rupture
12.1 580 140 2 Rupture
12.2 580 90 187 Rupture
12.3 580 60 1 559 Rupture Secondary
15Mo3
13.1 580 80 755 Rupture SH
13.2 580 100 60 Rupture
13.3 580 70 716 Rupture
St35.8III 11.1 500 110 984 Rupture Primary
11.2 500 140 197 Rupture SH
169
St 35.8
1000
Stress [MPa]
100
Experimental results
Creep to Rupture for St 35.8
10
13000,0 14000,0 15000,0 16000,0 17000,0 18000,0 19000,0 20000,0 21000,0
Parameter Larson-Miller P = T (20 + log t)
Figure 1. Comparison of tube creep results of the St 35.8 III steel after 205 000 hrs of
service and creep-to-rupture properties of this steel. (For the creep-to-rupture properties
of this steel see Taschenbuch : Stahl und Rohrenwerk Reisholz GmbH, einschliesslich
23. Nachtragslieferung, abgeschlossen im November 1971, Teil II).
13CrMo 44
1000
Napětí [Mpa]
100
LM pro 13CrMo44
Výsledky SH-2
Výsledky SH -1
10
15000 16000 17000 18000 19000 20000 21000 22000 23000
P = T (20 + log t)
Figure 2. Comparison of tube creep results of the 13CrMo44 steel after 205 000 hrs of
of this steel see "Warmfeste Werkstoffe für Kraftwerksanlagen“, VEB WTZ
Kraftwerksanlagenbau Pirna 1965, Autorenkollektiv: H. Seidel (Federfuhrung), G. Arndt,
R. Elze, R. Hennig, B. Melzer, K.-H.Patzer, W. Rossner, H.-K. Shiller).
170
15Mo 3
1000
Napětí [M pa]
100
y = -569,68Ln(x) + 5700,5
2
R = 0,9767
10
15000 16000 17000 18000 19000 20000 21000 22000 23000
P = T (20 + log t)
Figure 3. Comparison of tube creep results of the 15Mo3 steel after 205 000 hrs of
of this steel see "Warmfeste Werkstoffe für Kraftwerksanlagen“, VEB WTZ Kraftwerks-
anlagenbau Pirna 1965, Autoren: H. Seidel (Federfuhrung), G. Arndt, R. Elze, R. Hennig,
B. Melzer, K.-H.Patzer, W. Rossner, H.-K. Shiller).
For the purposes of this evaluation, the regression lines were constructed
through the experimental points. These regression equations were then used to
calculate the respective values of the Larson-Miller Parameters for the three
abovementioned „safety“ situations what resulted in the remaining operational
hours given in the Table 2.
Table 2. Remaining lifetime of the tested tubes resulted in the following results (only
„FULL SAFETY“ is given).
Working Remaining
Material
Component temperature, lifetime,
designation
°C years
Primary superheater (inlet) St 35.8 III 385 51,9
Primary superheater (outlet) 13CrMo44 440 15 867,5
Secondary superheater (inlet) 15Mo3 450 261,7
Secondary superheater (outlet) 13CrMo44 485 311,2
171
2.2 Conclusions (Case „2“)
Taking into consideration even the full operational safety of the boiler, i.e. the
abovementioned item No.3 „Full safety“, the remaining lifetimes of the tubes
tested is sufficiently long in order that the components are operated safely till the
next inspection. It can be thus assumed that all other high temperature
components of this boiler in question made of these steels possess the similar
quality.
3. CASE „3“
3.1 Boiler description and technical details
- Nominal steam output: 730 t/hr

- Steam temperature: 515°C
- Steam pressure: 9.2 MPa
- Operational hours (approx.): 177 000
Table 3. Results of the creep tests – steel 15 CD2.05 (A213T2), HTSH outlet.
Larson-Miller Stress Temperature Hours to Test

Result
Parameter [MPa] [°C] rupture sample
20001 75 580 3400 26 rupture
19849 90 580 1098 26F *) rupture
20001 75 580 3218 66 rupture
19849 90 580 1159 66F *) rupture
*) tube flame side
Table 4. Remaining lifetime calculation – 15 CD 2.05 (T2) steel, tubes 26,66. *) maximum
radius and minimum thickness measured during testing included.
Pressure Remaining
Tube radius Tube thickness, Stress Temp.
[MPa] lifetime
[mm/*] [mm/*] [MPa] [oC]
[years]
9,2 24,4 5 44,9 515 1 506,15
13,8 24,4 5 67,3 515 79,33
13,8 24,4 5 67,3 585 0,618
172
Steel 15 CD2.05
1000
Stress [MPa]
100
y = -0,0358x + 794,32
R2 = 0,9893
10
17000 18000 19000 20000 21000
LMP = T (20 + log t)
Figure 4. Creep to Rupture Curve and accelerated Creep Test Results, steel 15 CD2.05
(A213 T2).
Table 5. Remaining lifetime calculation – 15 CD 2.05 (T2) steel, tube 26. (Temperature
determined according to the scale thickness after cleaning).
Tube Remaining
Pressure Tube radius Stress Temperatu
thickness lifetime
[MPa] [mm/*] [MPa] re [oC]
[mm/*] [years]
9,2 24,4 5 44,896 511 2 071
13,8 24,4 5 67,344 511 107
13,8 24,4 5 67,344 581 0,799
173
Table 6. Remaining lifetime calculation – 15 CD 2.05 (T2) steel, tube 66. (Temperature
determined according to the scale thickness after cleaning).
Tube Remaining
Pressure Tube radius Stress Temperature
thickness lifetime
[MPa] [mm/*] [MPa] [oC]
[mm/*] [years]
9,2 24,4 5,1 44,017 509 2 810
13,8 24,4 5,1 66,024 509 144
13,8 24,4 5,1 66,024 579 1,038
3.2 Conclusions (Case „3“)
The HTSH tubes creep tests and remaining lifetime calculations revealed a
considerable exhaustion of the remaining lifetime capacity, so that when taking
into consideration the full operational safety of the boiler, i.e. the
abovementioned item No.3 „Full safety“, the remaining lifetime has been
(nearly) exhausted and the respective components should be replaced as soon as
possible.
Till that time, the boiler will be operated with limited safety what should
result in the precise keeping the service pressures and temperatures within the
designed limits. The present situation of the HTSH tubes is also the result of the
steel grade and tube dimensions used in the construction.
4. Final summary and conclusions

Two further examples (in addition to [1]) of boiler HTSH components remaining
lifetime evaluation were presented. Making use of Larson-Miller parameter
made it possible to perform accelerated creep tests of the tube material and
quantify the remaining lifetime of the structure. The acceleration of the creep
tests can substantially shorten the outage time and improve the economy.
Together with the other common metallographic, chemical and mechanical tests,
this procedure makes it possible to improve the overall assessment of the
remaining lifetime of structures. Nevertheless, the number of service hours and
the year of production should be taken into consideration when making decisions
on the frequency of the subsequent inspections. The creep phenomena are of
cumulative nature and accelerate close to the end of the lifetime exhaustion. It
should be also taken into consideration, that only creep (as survailing phenomenon
174
in case of high temperature components) was considered in the remaining life

calculation. Cyclic loading causing material fatigue and corrosion that can cause
the geometry changes of the components and thus increase the stresses can thus
contribute to the material damage and accelerate the process of material
degradation.
References
1. Mentl, V. & Liška, V. Remaining life assessment of steam turbine boilers. In: Veivo, J.
& Auerkari, P. (Eds.). Proc. of International Conference on Life Management
and Maintenance for Power Plants, BALTICA VII, Helsinki–Stockholm–Helsinki,
2007. VTT Symposium 246, Vol. 1. Pp. 106–112. ISBN 978-951-38-6315-9.
2. Liška, V., Koc, J., Chocholoušek, M. & Mentl, V. Hodnocení stupně degradace
vysokoteplotních komponent, Sb. konf. „Kotle, energetická zařízení a kogenerace
2008. Praha, 3/2008.
3. Mentl, V. & Liška, V. Failure risk assessment and remaining lifetime evaluation of power
plant steam boilers. 3rd Int. Conf. on Engineering Failure Analysis ICEFA III.
Sitges, Spain, 7/2008.
4. Liška, V., Koc, J., Chocholoušek, M. & Mentl, V. Evaluation of degradation extent and
remaining lifetime assessment of steam turbine boilers. Int. Conf. “Optimizing
Operational Outages”. Budapest, Hungary, 9/2008.
175
Creep strain, damage and life prediction for welded 0.5CMV steel
Creep strain, damage and life prediction

for welded 0.5CMV steel
Juhani Rantala, Pertti Auerkari, Stefan Holmström,

Jorma Salonen and Anssi Laukkanen
Espoo, Finland
Abstract
Creep damage as cavitation and cracking has historically occurred somewhat

earlier in the inspections of steam lines made of 0.5CMV steel, when compared
to other low alloy steels. The difference has been attributed to modest creep
ductility of 0.5CMV steel, and has resulted in attention paid to inspection and
maintenance. As the steel is still being used in vintage plants, it is of interest for
users while also serving as a model material that can show some essential
features of long term creep in much shorter laboratory testing. Particularly
challenging service can combine welded thick-wall components with pre-
existing fabrication defects, resulting in an unfavorable combination of high
stresses, multi-axial loading, and locally weak material. Examples are shown of
the observed damage evolution in welded 0.5CMV steam line material in front
of a crack-like defect. Long-term multi-axial loading does not necessarily result
in type IV failure of welds, but can also induce creep damage and cracking close
to the fusion line. This particularly applies to welds with undermatching weld
metal such as those made with consumables corresponding to 2.25Cr-1Mo steel.
With the LICON approach, it has been shown that this IIIa type of damage can
be reproduced in about 5 000 h of multi-axial creep testing, while in plant such
damage may require more than 100 000 h of service. The microstructural
features of the observed damage and a comparison of uniaxial and CT test data
suggest that the LICON approach of life prediction could be applied also for
176
welded components where the location of maximum damage deviates from the
type IV position.
1. Introduction
Ageing power plants make use of some vintage materials that are no more
applied in new plant. However, even for old materials new issues can arise, for
example on the need, timing and procedures of repairs of aged components with
indicated creep damage. As the on-condition maintenance according to the
observed state is often based on earlier experience, it may not be self-evident
whether such practices are prudent or overly conservative, e.g. for the reason of
the experience based on worst cases, suboptimal technical practices or on
conservative attitudes towards early damage [1, 2].
A well known vintage material in power plant applications is 0.5CMV steel,
also known as 14MoV6-3 by its European standard designation, for example in
the most recent edition of EN 10216-2 [3] on seamless tubes and pipes. This
steel type has been widely used from 1960’s to 1970’s in hot live and reheat
steam lines of thermal power plants. Considering its low alloy content
(nominally 0.5% Cr, 0.5% Mo and 0.25% V) it has high creep strength due to
MX-type precipitation strengthening, when properly heat treated. However, it
can have relatively modest creep ductility, and also susceptibility to reduced
cross-weld creep strength similar to other ferritic steels. Ductility issues and the
emergence of competitive newer steels (such as X20CrMoV11-1 and
X10CrMoVNb9-1) made the steel redundant in new plants from late 1970’s to
early 1980’s onwards. Nevertheless, limited ductility and weld behaviour make
this material of interest for those plants that are still operating with pipings made
of it, and also as a model material for exploring the evolution of creep damage.
This paper aims to contribute to the high temperature life assessment
knowledge on creep damage evolution and creep of welded 0.5CMV steel
(14MoV6-3). The work aims to provide a qualitative and quantitative description
of creep damage and crack initiation in a weld prepared with undermatching
consumables, as a test case for the LICON approach of accelerated creep life
assessment [4], and creep strain and rupture modeling for a COST case study
material (also 0.5CMV).
177
2. Materials and methods
2.1 Creep (CT) testing of 0.5CMV steel
Cross-weld CT specimens (W = 25 mm) were extracted from an ex-service

0.5CMV steam line material that had been re-heat treated to correspond as well
as possible to as-new steel, and then welded with an undermatching consumable
corresponding to that for welding 2¼Cr-1Mo steel (10CrMo9-10). The welding
was performed to a special groove to provide a straight HAZ for subsequent
testing. The welded 0.5CMV material and its pedigree data was provided by
British Energy (UK). All CT creep testing for crack initiation was performed
under constant load with monitoring of temperature, mouth opening at load line
and crack growth (by DCPD). The testing conditions were set so that the longest
testing time to crack initiation (to at least 0.5 mm) would be about 5000 h at
540°C. A narrow (0.14 mm) EDM tip was machined into the middle of the HAZ
of each CT specimen notch for creep crack initiation from the HAZ.
2.2 Creep strain and rupture modeling
Creep strain was modeled using the logistic creep strain prediction (LCSP)
methodology, described in detail elsewhere [5]. The model requires a time to
rupture model as base and two shape functions. The two presented rupture
models were fitted to average strength values from standard rupture data tables
of EN 10216-2 and from ISO data (1978) evaluated in the SPRINT 249 project
[6]. The strain model is based on SPRINT 249 / ISO data time to strain tables
enhanced with rather high temperature creep curves of a COST case data. The
COST case data consisted of 6 tests in the stress range of 40-55 MPa at
temperatures of 600-650°C. For modeling the procedures were in accordance to
the ECCC Recommendations [7].
The COST and the EN/SPRINT/ISO rupture data values overlap at 600°C / 40 MPa
with a time factor of about 2.7 calculated as SPRINT over COST data rupture
times. The corresponding stress reduction factor is 0.63 in comparison to models
based on data from both [3] and [6].
The temperature range for the EN 10216-2 based rupture model is from 490 to
560°C, SPRINT/ISO data (both rupture and time to strain tables) have a
temperature range of 480–600°C. The stress ranges are 59–279 MPa (EN
standard rupture values) and 8–299 MPa for the SPRINT/ISO data. Since the
178
bulk of the data is describing times to rupture (or strain) at durations from
10 000 h to 250 000 h, applicability of the rupture and strain models may not be
well established for short term predictions.
The logistic creep strain prediction (LCSP) approach [5] was used to model
time to strain using the average stress to given strains of 0.1% to 5% as shown
for the 550°C isotherm in Figure 1.
250000
250000
200000
Time to strain (h)

200000
Time to strain (h)
150000
150000
100000
100000
50000
50000
0
0
50
60 0.01
m)
70 m
80 m/
Stre
ss ( 90 0 (m
M 10 0 0 .0 0 ain
Pa) 11 0 S tr
12 1
Figure 1. The 550°C isotherm for time to specified strain (SPRINT/ISO data).
3. Results
The CT specimen testing results with the observed time to creep crack initiation
are shown in Table 1. The longest test has practically achieved the intended
maximum duration of 5 000 h at the time of interruption. The crack appearance
after interruption is shown in Figures 3 and 4. The emerging creep crack was
growing intergranularly in the coarse grained heat affected zone (CGHAZ), but
in the shortest tests also straining in the main tensile stress direction, at least as
179
crack tip opening (Figure 5). When extensive enough, all tests resulted in type
III cracking along the fusion line. More details of the crack in the longest test
(CT5) are shown in Figures 6–9.
Table 1. Results of CT specimen testing (time to crack initiation; here Sref1 = plane strain
Mises stress; Sref2 = plane stress Tresca stress.
Specimen K (MPa√m) Sref1 (MPa) Sref2 (MPa) Time (h)

CT1 15.0 85.1 135.2 198
CT2 13.5 76.5 121.6 434
CT3 12.0 67.8 107.8 888
CT4 11.0 62.3 99.0 2705
CT5 9.0 51.0 81.1 4673
CT6 8.0 45.6 72.4 11453
A good description of the rupture data was found using the DESA software of
TU Darmstadt. The COST data required a stress correction factor of 0.63 for
minimal residuals in comparison with the models based on values given in [3]
and [6]. The time to rupture at specified stress and temperature can be described
by the Manson-Brown master curve
log10 (tu ) − log10 (ta ) N

= ∑ Bi f (σ )
i
PMB = R
⎡ T + 273 − Ta ⎤ i =0 (1)
⎢⎣ 1000 ⎥⎦
where tu is the time to rupture, T the temperature in °C and log(ta), Ta, R and Bi
are fitting parameters. Fitting resulted in a scatter factor [7] of Z = 1.06 for the
EN 10216-2 values, and Z = 1.04 for the SPRINT values. For the SPRINT
model the stress dependence was better described with a stress function in the
form of σn instead of log(σ).
180
The LCSP model for the whole creep curve is defined as [7]
1/ p
⎛ log(t u ) + C ⎞
log(ε t ) = ⎜⎜ − 1⎟⎟ ⋅ x0 (2)
⎝ log(t ε ) + C ⎠
where ε t is the predicted strain at time t ε , and t u is the rupture time that can be
taken from the master curve for creep rupture. The p and x0 parameters are
temperature and stress dependent, and C is a constant.
-4 12
-4.05 11
-4.1 -4 10 12
-4.15 -4.05 9 11
-4.2 8
-4.1 10
x0
-4.25 7
p
-4.15 9
-4.3 6
-4.2 8
-4.35 5
x0
-4.25 7
p
-4.4 4
-4.45 -4.3 6
3
-4.35 5
510 510
Te
Te
-4.4 4
mp
mp
520 520
er
era
530 -4.45 530 3

tua
tu
540
re
540
re
50
50
75
75
(°C
(°C
10
10
550 550
12
12
0
15
0
)
15
)
5
17
5
17
(MPa)
0
(M Pa)
0
Stress
5
Stress
5
Figure 2. Temperature and stress dependence of the shape parameters for 0.5CMV steel.
A comparison of predicted vs. measured/table values of time to specified strain

is shown in Figure 10. The predicted evolution of creep strains and strain rates as
a function of life fraction at 550°C and 70 MPa (= 100 000 h life from SPRINT/
ISO data) using the LCSP model is shown in Figure 11.
181
Figure 3. The crack of specimens CT1-CT4: crack has grown to CGHAZ close to the
fusion line. Scale as in Figure 4.
182
Figure 4. The crack of specimen CT5: crack has grown through CGHAZ to region close to
the fusion line (type III crack).
Figure 5. The crack tip of specimen CT2: note opening of intergranular cracks in the
vertical (loading) direction.
183
Figure 6. The end section of the crack of specimen CT5 close to the fusion line.
Figure 7. A detail of Figure 4; crack of specimen CT5 extending to weld metal.
Figure 8. A detail of Figure 7; crack of specimen CT5 in weld metal.
184
Figure 9. A detail of Figure 6; crack tip of specimen CT5.
Figure 10. Specimen/test CT6: the crack is growing towards the CG-HAZ.
185
5
Predicted time to strain log(h)
0
0 1 2 3 4 5 6
Time to strain log(h)
Figure 11. Predicted vs. measured/table values of time to specified strain (continuous
curves from tests, vertically aligned data from [6]; Z = 2.1)
3
0.1 1 .10
4
1 .10
Strain rate
5
1 .10
Strain
0.05
6
1 .10
7
1 .10
8
1 .10
0.5 1 0 0.5 1
Life fraction Life fraction
(a) (b)
Figure 12. a) Normalised creep strain and b) creep strain rate at 550°C / 70 MPa
(continuous line, 100 000 h life from [6] and at 600°C/58 MPa (dashed line, a simulated
creep test).
186
4. Discussion
The accuracy of the creep strain model mainly relies on the mean stress values to
given strain as reported in [6]. The model has been validated by calculating
minimum creep rates from the LCSP model and comparing these to values from
[8] at 565°C for both short and long term tests (Figure 13). The LCSP shape
functions from the data predict minimum creep rates with a scatter factor Z = 4.5,
and with a very satisfactory slope. The results suggest that the LCSP approach is
not only robust and easy to apply in the sense of requiring minimal data, but that
the model predicts creep strains and strain rates surprisingly accurately (Figures
11 and 13).
1.E-01
1.E-02
Predicted creep rates by LCSP
1.E-03
1.E-04
1.E-05
1.E-06
1.E-07
1.E-08
1.E-08 1.E-07 1.E-06 1.E-05 1.E-04 1.E-03 1.E-02 1.E-01
Creep rates (1/h) in [8]
Figure 13. LCSP predicted minimum creep rates for 0.5CMV steel in comparison to rates
given in [8], with time to rupture calculated using SRF = 0.7 in relation to EN-10216 model;
for strain rate Z = 4.5.
187
In the LICON approach of life prediction, the stress multi-axiality at the tip of a
notch or crack of a CT specimen is applied to accelerate creep crack initiation
[4]. The approach aims to reduce the risk of unwanted mechanism changes that
could result in non-conservative predicted life (Table 2). By definition, the
approach requires that intergranular creep cavitation appears at the initiating
crack to indicate classical long term damage. For welds of ferritic steels, this was
typically interpreted as distributed dense creep cavitation at the fine grained or
intercritical HAZ.
Table 2. Comparison of methods to accelerate creep and creep failure
Elevated quantity Advantages Limitations/disadvantages
Temperature Simple, strong acceleration Mechanism change possible 1)
(Effective) stress Simple, strong acceleration Mechanism change possible 1)
Stress multi-axiality Damage enhancement Control of stress level 2)

1) with little or no enhancement of creep cavitation damage
2) complex testing geometry, requires finite element analysis (FEA)
However, in the present example the observed damage is not at this type IV
position, but rather lies on the inner side of HAZ, either within the CGHAZ
(specimens CT1 to CT3) or close to the fusion line (CT5). There is no indication
of the damage deviating towards the type IV position, although the initial notch
position would easily allow this as the crack first initiates close to the central
region of HAZ (Figures 4–9). It also seems that towards longer times approaching
5 000 h of testing, the crack stays very close to the fusion line (Figures 6–9). The
distance from the apparent fusion line is up to few grain boundaries only, and in
one location the crack even branches to the weld metal side (Figures 7 and 8).
The crack position appears roughly consistent of type IIIa damage, where carbon
diffusion to the higher chromium weld metal creates a carbon depleted weaker
zone in the HAZ next to the fusion line. Under typical plant conditions this may
typically take some 200 000 h at similar or higher temperatures [9].
Another suggested indicator of long term damage is gb cavitation on the
grain boundaries orientated against the highest tensile stress (Figure 14). While
similar densities of creep cavitation damage as in type IV damage are not likely,
188
it is concluded that the observed mode of creep damage is expected to remain

within the applicability limits of the LICON approach. Comparison of CT test
results to uniaxial standard rupture stress also suggests potential for life
prediction when using the LICON principle (Figure 15). Note that here a better –
though still not perfect – compatibility of the slopes is obtained by taking the
uniaxial correspondence to the weld metal (2.25Cr-1Mo) than to parent metal
(0.5CMV). This may make sense as the crack is following the fusion boundary
and also explores the weld metal side (Figures 7 and 8).
Further confirmation is sought from additional FEA of the case.
Figure 14. Grain boundary cavitation parallel to crack growth in CGHAZ of specimen CT5.
189
1000
0.5CMV-CT P22 uniax 0.5CMV uniax

y = 297.24x-0.1603 y = 260.36x-0.1403 y = 1516.4x-0.2359
R2 = 0.9605 R2 = 0.9925 R2 = 0.9987
Stress (MPa)
100
BM - P22 BM - 0.5CMV
CT-time to initiation Power (BM - P22)
Power (BM - 0.5CMV) Power (CT-time to initiation)
10
1E+02 1E+03 1E+04 1E+05 1E+06
Time (h)
Figure 15. Comparison of CT (multi-axial) testing results and uniaxial data for 0.5CMV
and 2.25Cr-1Mo steel (weld metal) using the LICON approach.
There remain some questions of interest in the performance of the LICON approach.
When isothermally applied, it is not immediately clear how well does the
method account for the presumably differing kinetics of long term thermal
degradation (e.g. growth, coarsening and dissolution of strengthening precipitates)
and stress activated damage such as transport and coalescence of vacancies and
grain boundary cavities. In determining the long term standard creep properties
by uniaxial testing the problem is circumvented by testing up to the time-
temperature-stress range of the expected applications. When this is too slow and
accelerated creep testing must be used, increasing the test temperature rather that
stress is often recommended to avoid non-conservative life prediction [7]. While
increased stress multi-axiality is not very similar to increased stress level in this
respect, disregarding the possible additional impact of thermal degradation is not
automatically justified. Against this background it appears of some consolation
that multi-axial (CT) creep testing of 0.5CMV welds has been shown to
reproduce the expected type of long term damage in a surprisingly short time.
190
5. Summary
To characterize creep crack initiation in the HAZ of 0.5CMV steel welded with
an undermatching P22 consumable, creep testing of CT specimens with the
notch tip in the HAZ has been performed up to 4700 h at 540°C. The observed
creep crack tip in the longest test resided on the inner side of HAZ, close to the
fusion line (type III or IIIa damage). The damage was characterized by
intergranular crack growth preceded by creep cavitation mainly on the same and
parallel grain boundaries approximately perpendicular to the principal loading
direction. It was concluded that in spite of relatively low overall creep cavitation
density, the observed damage is of the true long term type for the case and is
expected to remain within the applicability limits of the LICON approach.
Approximately 5000 h of testing time seems therefore sufficient to achieve the
desired stable characteristics of long term damage.
A Manson-Brown type of expression was optimized to describe creep rupture
of 0.5CMV steel. The reference material from an ex-service component required
a stress correction factor of 0.63 in comparison with the standard mean values of
EN 10216-2. The creep strain model mainly relies on the data from the European
SPRINT 249 project. The model has been validated by calculating minimum
creep rates from the LCSP model and comparing these with values from
literature. Good agreement was observed with a scatter factor of Z = 4.5, and
also with a very satisfactory slope. The results suggest that the LCSP approach is
not only robust and easy to apply in the sense of requiring minimal data, but that
the resulting models can accurately predict creep strains and strain rates.
Acknowledgement
David Dean from British Energy is acknowledged for providing the welded
CMV test material and related background information. The Academy of
Finland is acknowledged for funding the work under Decision no. 117700 (prof.
Kim Wallin).
References
1. Salonen, J., Auerkari, P., Lehtinen, O. & Pihkakoski, M. Experience on in-service
damage in power plant components. 2nd Int. Conf. on Engineering Failure
Analysis (ICEFA-II), Paper O8.1. Toronto, Canada, 12–15 Sept 2006.
191
2. Auerkari, P., Salonen, J., McNiven, U. & Jovanovic, A. Classification of damage and
prediction of its development in high temperature components of the power
industry. Failures '96. A. A. Balkema (1996), pp. 269–280.
3. EN 10216, 2002. Seamless steel tubes for pressure purposes. Technical delivery
conditions. Part 2: Non-alloy and alloy steel tubes with specified elevated
temperature properties. CEN, Brussels.
4. Auerkari, P., Bendick, W., Holdsworth, S., Rantala, J. H., Hurst, R., Coussement, C. &
Hack, R. Predicting long-term creep behaviour using the LICON methodology.
Proc. of the Conf Advances in Materials Technology for Fossil Power Plants.
University of Wales, Swansea, 1–6 April 2001.
5. Holmström, S. & Auerkari, P. Robust prediction of full creep strain curves from minimal
data and time to rupture. Materials Science and Engineering for Energy
Systems 1 (2006), pp. 249–255.
6. Standard Data Tables, Vol. 2, Ferritic steels. SPRINT SP 249, 1995.
7. ECCC Recommendations, Vol. 5: Guidance for the assessment of uniaxial creep data,
2005.
8. Evans, R. W. & Wilshire, B. Introduction to Creep. Institute of Materials, London, 1993,

p. 91.
9. Brett, S. (RWE npower, UK), 2006. Personal communication.
10. Holdsworth, S. & Mazza, E. Exploring the applicability of the LICON methodology for
a 1%CrMoV steel. Materials at High Temperatures, Vol. 25, No. 4, pp. 267–276.
11. Hales, R. The role of cavity growth mechanisms in determining creep rupture under
multiaxial stresses. Fat. Fract. Eng. Mater. Struct., 1994. Vol. 17, 579–591.
192
Different types of cracking of P91 steel weld joints after long-term creep tests
Different types of cracking of P91 steel weld

joints after long-term creep tests
Dagmar Jandová, Josef Kasl and Eva Chvostová

ŠKODA VÝZKUM s.r.o.
Plzeň, Czech Republic
Abstract
The paper deals with creep testing and microstructural investigation of trial weld
joints prepared of wrought and cast 9Cr-1Mo-V steels using GTAW & SMAW
method. Creep testing was carried out at temperature range from 525°C to 625°C,
the longest time to rupture of 45 811 hrs was achieved. The creep strengths of weld
joints for 100 000 hrs were calculated. Different types of cracking were observed in
dependency on conditions of creep test and the type of weld joint. Type I and Type
II fractures occurred at high applied stresses at relatively low temperatures in the
tube weld joint and also in two specimens of the cast plate weld joint after creep test
at the lowest temperature and the highest temperature. All other fractures were of the
Type IV. Causes of different fracture location in tested weld joints were elucidated
on the base of substructure evolution in individual zones – the weld metal, the heat
affected zone and the base material. Two processes occur simultaneously, which
result in the creep damage: i) softening of solid solution as a result of Laves phase
precipitation and ii) formation and coalescence of cavities in the soft fine grained
parts of heat affected zone.
1. Introduction
Grade 91 steel is widely used for production of components for fossil fuel power
plants, which operate under severe service conditions for many years. A great
attention was paid to investigation of creep properties of this grade and a lot of
creep data were collected. Nevertheless experience has shown that some difficulties
193
persist in production and operating of components made of P91 steel. A number of

unexpected in-service failures of grade P91 components correctly fabricated with
desirable tempered martensitic structure demonstrates that study of long-term
properties and failure development is still of a great importance. Type IV
cracking of welds, which occurs in the fine grained region towards the parent side
edge of the heat affected zone, belongs to the most often observed failure [1].
The paper deals with a study of trial weld joints prepared in conditions of real
production of piping and cast components of steam turbine. Fracture mechanism
after long-term creep exposures is correlated with microstructural processes
taking place in individual parts of weldments. Partial results of performed
experiments have already been presented in several conferences [2, 3, 4].
2. Experimental procedures
Two trial weld joints were fabricated from wrought or cast P91 steel using
GTAW & SMAW methods. The first weld joint (C) was produced by joining
cast plates 500 × 150 × 25 mm in size and the second one (C1) by joining tube
segments with an outer diameter of 325 mm, a wall thickness of 25 mm and a
length of 400 mm. The plates and the tube segments were joined in PA and in
PC position respectively. The same conditions of welding process were kept for
both weld joints [2]. The chemical composition of the base materials and the
weld metals was verified using XRF method (Table 1). Both base materials were
austenitized at 1 050°C for 1.5 hrs, then oil quenched and tempered at 750°C for
3.5 hrs. The post-weld heat treatment (PWHT) was applied as follows:
(740÷750)°C/2.5 hrs for the C weld joint and 760°C/2.5 hrs for the C1 weld
joint. Smooth cross-weld specimens underwent long-term creep testing to
rupture. Fractographic analysis was carried out using scanning electron
microscopy (SEM). Metallographic samples were prepared in longitudinal
section, macro and microstructures were revealed using Vilella´s reagent,
location of fracture was specified and structure was observed using light
microscopy and SEM. Cross-weld Vickers hardness profiles of as received weld
joints in conditions after PWHT and after creep tests were measured.
Transmission electron microscopy (TEM) was used for study of microstructural
changes taking place during creep exposures. Carbon extraction replicas were
prepared from the weld metals and the base materials. Thin foils were prepared
from the weld metal (WM), the coarse prior austenitic grain heat affected zone
(CG HAZ), the fine prior austenitic grain heat affected zone (FG HAZ) and the
194
base material (BM). Foils were jet electropolished in a 6% solution of perchloric

acid in methanol at a temperature range from -60°C to -40°C. Secondary phases
were identified using energy dispersive X-ray microanalysis (EDX) and selected
area electron diffraction (SEAD).
Table 1. Chemical composition of base materials and weld metals (wt%).
C Mn Si Cr Mo V Ni Nb Al N P S
BM-t 0.12 0.49 0.29 8.6 0.96 0.21 0.30 0.07 0.011 0.06 0.011 0.004
BM-p 0.12 0.41 0.21 8.8 0.92 0.22 0.10 0.09 0.004 0.06 0.014 0.008
WM-t 0.12 0.70 0.21 9.1 1.05 0.21 0.73 0.05 0.003 0.04 0.010 0.006
WM-p 0.12 0.71 0.23 9.1 1.05 0.20 0.68 0.05 0.005 0.04 0.012 0.009
t ...tube weld joint, p…plate weld joint
3. Results
3.1 Creep testing

The long-term creep rupture testing was carried out at the air at stresses in a
range from 40 MPa to 240 MPa and at temperatures ranged from 525°C to
625°C. Results are summarized in Table 2 and graphically represented in
Figure1 and 2. The stress is plotted vs. the Larson-Miller parameter P = T · [C +
log τ], where T represents temperature given in kelvin, C = 25 for P91 steel and
τ means time to the rupture in hours. Full lines in graphs represent creep strength
of the base materials for 100 000 hrs according to ECCC Data Sheets
(X10CrMoVNb 9-1 steel for the tube weld joint and GX12 CrMoVNbN 9-1
steel for the plate weld joint). Fractures occurred in different parts of weldments
in dependency on conditions of the creep test: in WM (Type I), in WM and HAZ
(Type II), in FG HAZ (Type IV) or in BM unaffected by welding. Fracture
location is specified in Table 2 and in graphical representations (Figures 1 and 2).
The creep strength of the weld joint for a definite temperature and time to
rupture was calculated using interpolation and extrapolation and two different
equations: (1) the standard regression equation [5] and (2) Larson-Miller equation.
1 1 1 1
log τ = A1 + A2 ⋅ log + + A3 ⋅ log + ⋅ log sinh ( A6 ⋅ σ ⋅ T ) +
T A5 T A5
+ A4 ⋅ log sinh ( A6 ⋅ σ ⋅ T ) (1)
195
Symbol σ represents applied stress, T is temperature in kelvins and Ai are

materials constants, which were calculated using the least square method. Then
creep strengths for 100 000 hrs and selected temperatures were calculated. The
second relation (2) was obtained if Larson-Miller parameter P substitute τ in
equation (1) and T = 873. New constants Ai were calculated and the creep
strengths were determined. The creep strengths of weld joints for 100 000 hrs
and temperatures from 500°C up to 600°C were compared with the creep
strength of the base materials according to ECCC Data Sheets (Table 3 and 4).
Table 2. Results of creep tests.
Temp. Stress Specimen Time [hrs] A [%] Z [%] Fracture type

[°C] [MPa] tube plate tube plate tube plate tube plate tube plate
525 240 2C1 C16 3,772 9,309 14.6 11.9 72.4 51.0 II IV,BM
525 220 3C1 C1 11,546 49,611 9.6 6.9 29.7 14.4 I II
550 200 6C1 - 1,183 - 14.3 - 72.4 - II -
550 180 7C1 C4 9,853 31,920 6.3 2.5 14.2 14.4 I IV∗
550 160 - C5 - 33,189 - 1.6 - 4.9 - IV∗
575 200 11C1 - 134 - 20.2 - 84.9 - II -
575 180 25C1 C6 960 2,853 15.8 7.6 84.9 19.2 BM IV
575 160 12C1 C15 4,704 3,793 7.8 4.3 37.8 12.1 IV IV
575 140 8C1 C7 9,608 10,031 3,0 3.0 18.8 4.9 IV IV
575 120 9C1 C8 12,624 19,289 2.1 1.1 19.0 7.4 IV IV
600 140 16C1 C3 981 1,797 20.4 4.8 59.3 12.3 IV IV
600 120 15C1 C12 2,242 2,610 5.0 1.8 4.9 14.7 IV IV
600 100 13C1 - 6,080 - 2.4 - 12.1 - IV -
600 90 26C1 - 8,165 - 0.5 - 11.9 - IV -
600 80 14C1 C10 10,181 25,818 2.1 1.1 18.8 9.8 IV IV∗
600 70 24C1 C11 27,471 45,811 1.6 1.4 4.9 4.9 IV∗ IV∗
625 100 18C1 C2 1,777 1,061 3.6 3.2 18.8 16.7 IV IV
625 80 17C1 C9 3,970 2,291 2.0 1.1 16.7 5.2 IV IV
625 60 19C1 C13 13,673 19,210 1.8 1.2 2.2 7.4 IV IV∗
625 50 20C1 C14 29,962 29,312 3.4 9.9 9.5 35.2 IV∗ I∗
A…elongation, Z…area reduction, I, II or IV...type of fracture, ∗…cavities not only in HAZ, but also in WM
and/or BM.
196
Table 3. Creep strength of tube weld joint for 100 000 hrs in comparison with that of the
base material.
Tube weld joint X10CrMoVNb 9 1

Difference
Creep strength Creep strength
Temp.
Regres. Lars.-Miller. Average ECCC Data Sheet
A B C=(A+B)/2 D (C-D)/D
[°C] [MPa] [MPa] [MPa] [MPa] [%]
500 216 211 214 213 0,4
510 193 193 193 195 -0,9
520 172 175 174 177 -1,9
530 152 158 155 162 -4,2
540 134 141 137 148 -7,1
550 117 124 121 136 -11,2
560 101 109 105 124 -15,2
570 87 94 90 113 -20,0
580 74 80 77 102 -24,6
590 61 67 64 92 -30,7
600 49 56 53 82 -35,7
Table 4. Creep strength of plate weld joint for 100 000 hrs in comparison with that of the
base material.
Plate weld joint GX12CrMoVNbN9-1

Difference
Creep strength Creep strength
Temp.
Regres. Lars.-Miller. Average ECCC Data Sheet
A B C=(A+B)/2 D (C-D)/D
[°C] [MPa] [MPa] [MPa] [MPa] [%]
500 229 246 238 223 6,5
510 208 221 215 206 4,1
520 187 198 193 190 1,3
530 167 175 171 175 -2,3
540 147 153 150 160 -6,3
550 130 134 132 146 -9,6
560 113 115 114 132 -13,6
570 98 98 98 119 -17,6
580 84 83 84 106 -21,2
590 72 70 71 95 -25,3
600 61 57 59 86 -31,4
197
The creep strengths of both, tube and plate weld joint, fall into the usually
permitted ±20% scatter band of the creep strength of the corresponding base
material at temperatures up to 570°C. At higher temperatures it decreases below
the bottom of that scatter band (Figures 1 and 2). Nevertheless creep strength of
the plate weld joint is marginally higher than that of the tube weld joint within
the whole tested temperature range.
Figure 1. Stress vs. Larson-Miller parameter for the tube weld joint in comparison with a
creep strength for 100 000 hrs of X10CrMoVNb steel. Type of fracture is specified.
Figure 2. Stress vs. Larson-Miller parameter for the plate weld joint in comparison with a
creep strength for 100 000 hrs of GX12CrMoVNbN steel. Type of fracture is specified.
198
3.2 Fractographic analysis

Fracture analysis of cross-weld crept specimens revealed some differences in
creep failure between both trial weld joints exposed at the same conditions.
Fractures of the tube weld joint were located in WM at high stresses and
relatively low temperatures and in BM and HAZ at higher temperatures and
lower stresses. Cracks propagated in the weld metal close the fusion line at
highest stress for a given temperature in a temperature range from 525°C to
575°C (Type II) and in the central part of the weld metal at lower stresses at
525°C and 550°C (Type I). A decrease in applied stresses for specimens tested at
575°C resulted in the crack propagation in the base material near the edge of
HAZ. Elongations and area reductions were relatively high; elongation of
specimens of Type II fracture and also the specimen, which ruptured in BM,
ranged from 14% to 20% and their area reduction from 72% to 85%. The
fractures occurred by ductile transcrystalline mechanism and fracture surfaces
revealed a dimple morphology (Figure 3). The elongation and area reduction of
specimens of Type I fracture ranged from 6% to 10% and from 14% to 30%
respectively.
2 mm 50 μm
a b
Figure 3. Fracture surface of the tube weld joint crept at 525°C/240 MPa/3 772 hrs:
a) general view, b) detail of the central part. Specimen 2C1, SEM micrographs.
The transgranular creep fractures of the tube weld joint occurred in the fine prior
austenite grain heat affected zone (Type IV). Elongation and area reduction were
199
usually low; elongation of a few percent and area reduction below 20%. Only
one specimen (16C1), which was exposed at the highest stress at 600°C, showed
remarkable macro-plastic deformation. Fracture surfaces covered with oxide
layers showed some evidence of dimple morphology (Figure 2). Cavities were
observed at fracture surfaces after long exposures at high temperature (Figure 4b).
1 mm 100 μm
a b
Figure 4. Fracture surface of the tube weld joint crept at 600°C/70 MPa/27 471 hrs:
a) general view, b) detail of the central part. Specimen 24C1, SEM micrographs.
2 mm 500 μm
a b
Figure 5. Fracture surface of the plate weld joint crept at 575°C/160 MPa/3 793 hrs:
a) general view, b) detail of the central part. Specimen C15, SEM micrographs.
200
The crept test-pieces of the plate weld joint ruptured generally in the low
temperature HAZ region, near to the base material (Type IV); exceptions were
detected after creep tests at the lowest and the highest temperatures. Surfaces of
Type IV fractures revealed the dimple morphology similar to the tube weld joint
and, in addition, some intercrystalline facets were detected in C6 and C15
specimen (Figure 5). The fracture of C16 specimen loaded with the highest
stress at 525°C was initiated in FG HAZ and occurred partly in HAZ and in
unaffected BM, while fracture of C1 specimen exposed at lower stress at the
same temperature partly occurred in WM and partly in the HAZ (Type II). The
lowest loading at the highest temperature caused the fracture in the central part
of WM (Type I). Some cavities were observed on the fracture surface of this
specimen, especially along previous austenite grain boundaries (Figure 6).
Elongation of specimens with Type IV cracking was usually a few percent
and area reduction did not exceed 20%. Only C16 specimen showed significant
macro-plastic deformation after test at the highest stress and the lowest
temperature. Plastic deformation was also evident in C14 specimen of Type I
fracture after test at the lowest stress and the highest temperature.
1 mm 100 μm
a b
Figure 6. Fracture surface of the plate weld joint crept at 625°C/50 MPa/29 312 hrs:
a) general view, b) detail of the central part. Specimen C14, SEM micrographs.
3.3 Hardness measurement

Before creep testing both weld joints showed typical crossweld hardness profile
with higher hardness in the weld metal than in the unaffected base material, with
local maxima in the coarse prior austenite grain heat affected zones close the
fusion line and local minima in the fine prior austenite grain heat affected zones.
201
After creep tests some changes in hardness were found out in dependency on the
creep conditions.
Hardness across the tube weld joint before creep testing varied from about 180
to 232 HV10. Hardness of unaffected base material was 200 HV10, hardness of
the weld metal ranged from 207 to 224 HV10. Creep exposures at temperatures
ranging from 525°C to 575°C did not significantly influence the hardness.
Changes were found only in the plate weld joint after the exposure of almost
50 000 hrs at 525°C – hardness of WM and regions along the fusion line decreased,
but still exceeded the hardness of BM. On the contrary long-term exposure at
600°C resulted in apparent increase by about 10 HV10 in the weld metal and
also in the base material. Exposures at 625°C resulted in hardness decrease in
the weld metal by about 20 HV 10. Cross-weld hardnes profiles after PWHT and
after creep tests at 600°C and 625°C are shown in Figure 7a.
a)
b)
Figure 7. Cros-weld hardness profiles: a) the tube weld joint, b) the plate weld joint.
202
Hardness of the plate weld joint before creep testing was rather higher than that
of the pipe weld joint as a result of lower temperature of post-weld heat
treatment. Hardness of the base material and the weld metal were 210 HV10 and
226 HV10 respectively. During creep tests at temperatures from 525°C up to
600°C hardness slightly increases in the base material and also in the weld
metal, while after tests at 625°C hardness of the base material increased and
hardness the weld of metal remarkable decreased, especially after test at
625°C/50 MPa/ 29 962 hrs (Figure 7b).
3.4 Microstructural analyses
Microstructure of both the tube and the plate weld joints fully consists of
tempered martensite. Neither polygonal ferrite nor fresh martensite was
observed after PWHT. No defects were detected near the fusion line or in the
weld metal, which could cause a premature fracture. Characteristic acicular
structure was evident in WM, BM and CG HAZ, while fine featureless structure
was present in the FG HAZ.
Substructure before creep exposure was lath-like tempered martensite, laths
were divided into subgrains. Coarse M23C6 carbides occurred at boundaries of
prior austenite grains, ferritic laths and subgrains. Non-uniform distribution of
fine vanadium carbonitride was observed within laths. Globular silicon oxide
particles were present in the weld metal. Some differences were observed in size
and distribution of secondary phases in individual zones of weld joint. Density
of coarse particles was relatively low in WM and in CG HAZ and high in FG
HAZ. On the contrary density of fine vanadium carbonitride was higher in the
CG HAZ and lower in FG HAZ. A high dislocation density occurred in
subgrains with relatively high density of fine precipitate. Subgrains in the FG
HAZ were almost free of dislocations.
During creep tests cavities appeared in FG HAZ on the both sides of weld
joint. Size and density of cavities increased with increasing temperature and time
to rupture. Coalescence of cavities results in Type IV cracking. Cavities were
observed only exceptionally in specimens of the tube weld joint, which revealed
Type I and Type II cracking and in specimen ruptured in the base material.
Individual cavities were also observed in other parts of weld joint – in the base
material and/or in the weld metal after creep exposures longer than 30 000 hrs at
550°C and 20 000 hrs at temperatures above 600°C.
Precipitation reaction taking place during creep exposures strongly depended
203
on conditions of the creep loading [6]. Precipitation of Laves phase occurred at

550°C and higher and also several Z-phase particles were observed in both weld
joins after long creep exposures at
625°C. Laves phase corresponded to
the largest particles in the areas in-
vestigated. Sporadically observed Z-
phase were a plate-like shape with
dimensions of about 100 nm.
In the tube weld joint no important
changes in size and distribution of
secondary phases was observed after
creep test at 625°C/60 MPa/13 673 hrs,
nevertheless a growth of subgrains was
evident in FG HAZ. After creep
exposure at 625°C/50 MPa/29 962 hrs
200 nm
subgrains in FG HAZ were significant-
ly coarser than those in BM and WM.
Laves phase was not identified while Figure 8. Cluster of particles in WM of
some particles of Z-phase were the plate weld, 625°C/50 MPa/29,312 hrs,
observed on replicas prepared from C14 specimen. TEM micrograph.
WM and HAZ.
Significant changes in substructure were observed in the plate weld joint
after exposures at 550°C and 625°C for about 30 000 hrs and at 625°C for about
14 000 hrs. In WM coarse particles of Laves phase were identified after test at
550°C/160 MPa/33 189 hrs using SAED and EDX microanalysis. Iron and
molybdenum were prevailing in Laves phase, while chromium predominated in
M23C6 carbides. Changes in density of vanadium carbonitrides were not found
out. Laves phase was not observed in other regions of the plate weld joint.
After creep exposure at 625°C/50 MPa/29 312 hrs massive particles of Laves
phase (exceeded 1μm in size) were observed in the weld metal, apparently
smaller ones in other regions of the weldment (Figure 8). A significant growth of
subgrains occurred in WM. Massive particles often formed clusters with
carbides and inclusions and drew atoms of molybdenum from ferritic matrix.
Depletion of solid solution about molybdenum resulted in increase in size of
subgrain and remarkable softening of the weld metal. Substructures of the weld
metal in the tube weld joint and the plate one are shown in Figure 9.
204
500 nm 500 nm
a b
Figure 9. Substructure of WM after creep tests at 625°C and 50 MPa: a) the tube weld
joint, 20C1 specimen, b) the plate one, C14 specimen. TEM micrographs.
4. Discussion
Creep testing demonstrated that creep strength of the plate weld joint is rather
higher that that of the tube weld joint, although the welding process and geo-
metry of weld joint were identical. The attention has been payed to the (Mn +Ni)
content because it can influence Ac1 temperature. The welded tube segments and
cast plates contained 0.8 and 0.5 wt.% of (Mn +Ni) respectively, while (Mn +Ni)
concentration was 1.4 wt.% in weld metal. It has been shown that increasing
(Mn +Ni) content in P91 base material between 0.60 and 1.25 wt.% resulted in a
decrease in the Ac1 temperature of about 45°C [7]. In spite of this fact the
temperature of applied PWHT of the tube weld joint was 760°C; it means from
10°C to 20°C higher than that of the plate weld joint and simultaneuously 10°C
higher than the temperature of tempering of the base material. Although Ac1
temperature was not apparently exceeded during PWHT, a temperature of 760°C
seams to be too high and caused drop in creep strength of the tube weld point in
205
comparison to the plate one. The PWHT resulted in decrease in hardness of the
tube weldment and Type I and Type II fracture, if relatively low temperatures ad
high stresses were applied. Fracture occurred by ductile transcrystalline mechanism
similar to that at room temperature. At these conditions the cast material is more
susceptible to fracture than the wrought one.
At relatively high temperatures and low stresses cavitation failure in FG HAZ
causes Type IV fracture of both weld joints. In these zones lath-like structure is
replaced by soft equiaxed grains and subgrains with low dislocation density and
with coarser secondary phase particles at boundaries. Dislocation creep results in
formation of steps at grain/subgrain boundaries and nucleation of cavities.
Cavities can be also formed as a result of interaction between dislocations and
coarse particles at grain boundaries. The growth of cavities is promoted by the
grain boundary diffusion and coalescence of cavities results in crack
propagation. Subgrain boundaries form also chanels of relatively high speed
diffusion similar to high angle boundaries, therefore intercrystalline facets are
observed on Type IV fracture surfaces only exceptionally.
Precipitation of Laves phase occurs predominately in the weld metal during
long exposures at a temperature of 550°C and higher and also in other parts of
weldments. Occurrence of fine Laves phase particles has a positive influence on
strength; hardness increases after creep exposures at 600°C up to about 30 000
hrs. However with increasing time and temperature of the creep exposure growth
of Laves phase particles causes softening of matrix as a result of depletion of
solid solution about molybdenum. Relatively soft weld metal can be easily
deformed and at massive Laves phase particles and their clusters with silicon
inclusions and chromium carbides cavities often nucleate. It is interesting that a
higher temperature of PWHT probably retarded the Laves phase precipitation in
the tube weld joint.
At temperatures above 550°C both processes occur: cavitation failure in FG
HAZ and softening of solid solution mainly in the weld metal. Usually Type IV
cracking occurs, however at 625°C softening of the weld metal can predominate
and then Type I fracture is realized.
Precipitation of Z-phase seams to be of a secondary importance in comparison
to precipitation of Laves phase. It can be expected, that detrimental influence of
Z-phase, which causes dissolution of vanadium nitrides and decrease in
precipitation strengthening, could be shown after creep exposures exceeding
significantly 30 000 hrs at temperatures above 600°C.
206
5. Conclusions
The submitted study of weld joints prepared in conditions used in the industrial
practice showed that small differences in chemical composition and PWHT can
result in different kinetics of precipitation processes in individual zones of
weldments. At high temperatures failure is concentrated in FG HAZ and Type
IV fracture occurs, however at specific conditions failure in the weld metal can
be dominant and then Type I fracture is realized.
Acknowledgements
This work was supported by Grant projects SMS 4771868401 and OC09041
from the Ministry of Education, Youth and Sports of the Czech Republic.
References
1. Brett, S. J. UK experience with modified 9Cr (grade 91) steel. In: Proc. BALTICA VII –
Life Management and maintenance for Power Plants, Vol. 1. J. Veivo &
P. Auerkari (Eds.). Espoo: VTT Technical Research Centre of Finland, 2007.
Pp. 48–60.
2. Jandová, D., Kasl, J. & Kanta, V. Long-term creep testing and microstructure evaluation
of P91 steel weld joints. In: Proc. BALTICA VII – Life Management and
maintenance for Power Plants, Vol. 1. J. Veivo & P. Auerkari (Eds.). Espoo: VTT
Technical Research Centre of Finland, 2007. Pp. 143–155.
3. Jandová, D., Kasl, J. & Kanta, V. Influence of substructure on creep failure of P91 steel
welds. In: Proc. 2nd ECCC Creep Conference Creep & Fracture in High
Temperature Components – Design & Life Assessment Issues. I. A. Shibli &
S. R. Holdsworth (Eds.). Lancaster, PA, USA: DEStech Publications, 2009.
Pp. 177–188.
4. Jandová, D., Kasl, J. & Kanta, V. Creep failure in welds of cast and wrought P91 steel
after long-term creep testing. In: Proc. IIV International Conference on Advances
in Welding and Allied Technologies. Z. Sun Eds. Singapore Welding Society,
Singapore, 2009. Pp. 149–154.
207
5. Bina, V. & Hakl , J. Relation between creep strength and strength for specific creep
strain at temperature up to 1 200°C. Mater. Sci. Engn., 1997. Vol. A234–236,
pp. 583–586.
6. Jandová, D., Kasl, J. & Chvostová, E. Influence of precipitation on dislocation sub-

structure and creep properties of P91 steel weld joints. Materials at High
Temperatures (in print).
7. Seifert, J., Alexandrov, B., Lippold, J., Sanders, J., Tanzosh, J. & King, B. Examination
of phase transformation during PWHT of steel P91. In: Proc. IIW International
Conference Safety and Reliability of Welded Components in Energy and Processing
Industry. P. Mayr, G. Posch & H. Cerjak (Eds.). Graz: Graz University of
Technology, 2008. Pp. 75–80.
208
New mechanical testing equipment for testing in real and simulated service environments
(HIPS)
New mechanical testing equipment for

testing in real and simulated service
environments (HIPS)
Pekka Moilanen and Stefan Holmström

Espoo, Finland
Abstract
This paper aims to introduce the application areas of the multifunctional high
precision pneumatic loading system (HIPS), based on VTT developed and
patented technology. The HIPS has been utilized in a number of different
challenging environments, among others laboratory simulated nuclear power
plant water environments such as the BWR, PWR and SCW, high temperature
gas environment and actual nuclear reactor inpile testing. The test palette is also
large, i.e. tensile testing, stress corrosion cracking, fatigue, creep fatigue,
radiation impact on tensile and fatigue, and many variants with combined
environmental impact. The HIPS has also successfully been used as mover
system in a version of the contact electrode resistance measurement. The wide
spectrum of tests enabled by the flexible concept is shortly described together
with some test results from the different types of tests. The technological
development path from single bellows tensile loading devices towards the more
demanding double bellows fatigue and combined tension/compression/internal
pressure system is described.
1. Introduction
The primary goal of this paper is to present the multitude of possibilities offered
by the high precision pneumatic loading system (HIPS). The first successful
209
(HIPS)
testing devices were published in the dissertation work of the first author [1] of
this paper. Since then the HIPS has been developed from the basic 3-point
bending type loading frame to a multipurpose testing platform capable of
performing a multitude of materials testing types in demanding simulated and
actual service environments. The HIPS variants covered to date is presented in
Figure 1.
Bellows Tensile Specimen Linear Variable Differential Transformer
In-core reactor fatigue testing devices
Diameter of the modulus = 50 mm

Length = 220 mm
Weight = 900 g
Figure 1. HIPS testing device variants, including test environments Flow diagram of a
“basic” HIPS test- rig (one directional loading).
210
(HIPS)
The patented basic control setup presented is the heart of the HIPS and is the
same for all the applications give a few small modifications demanded by the
selected environments. The main benefit of the loading system is that there are
no moving loading lead throughs, only pressure line and electrical feedback
connections. The testing device itself can be far away from the control units and
servo valves (over 30 m and a ∅1.6 mm has been successfully implemented).
This feature is a must for instance in nuclear reactor in-pile testing. A list of the
testing environments and testing types performed to date is presented in Table 1.
Table 1. HIPS test environments, testing types and references.
Environment Test types Test specification Reference

Tensile test constant 2 sample rig, axial loading,
In-pile
displacement rate stagnant water [2]
(BR2 test reactor)
10e-7 1/s at 90°C
In-pile 2 sample rig, creep fatigue
(BR-2 test Creep fatigue loading (holding 10 and 100 s), [3]
reactor) stagnant water 90°C
Multiaxial creep
internal pressure +
In-pile
tensile/compressive under construction [4]
(Osiris reactor)
load with diametral
strain measurement
BWR water 6 spec. autoclaves
chemistry Stress corrosion
displ.rate 2×10-8 mm/s, [5, 6]
(autoclave) cracking (SCC)
288°C, 90 bar
PWR water Fatigue Sinus wave 0.1 Hz
chemistry (HLWR environment) [7, 8]
(autoclave) 350°C / 160 bar
PWR water Crack initiation and
chemistry growth (K) testing Miniature autoclave
ongoing
(miniature (∅ 64 mm)
autoclave)
Crack initiation and Miniature autoclave
SCW environment growth (K) testing (∅ 64 mm) double bellows ongoing
650°C / 350 bar
loading unit
High temperature Tension and compression with
Creep, creep fatigue ongoing
(air) 2 bellows - 600°C
211
(HIPS)
One of the main features of the HIPS is the level of accuracy both in
displacement (strain) and force (stress) in the various testing environments. The
calibration of each system requires naturally thorough test runs with
instrumented pressure, load and temperature feedback. The sensitivity of control
realized by the pneumatic servo control enables excellent accuracy of the
measured displacement, as the displacement fluctuation (< ±2 μm) easily surpassed
the limits set by the ASTM-1152-87 standard. An example of a rising load test
(load as a function of displacement) is presented in Figure 2. The load accuracy
also easily fulfils the same ASTM standard requirements.
4500 0.00039
Specimen type: 3PB
Specimen size: 10x10x55 mm3
4000 Material: pressure vessel steel Load
Displacement rate: 2.4 10-5 mm/min
0.00037
Test environment: BWR
3500
Environment pressure: 93 bar
Temperature: 288 oC PD 0.00035
3000
2500 0.00033
Load [N]
PD [V]
3900
2000 3850
0.00031
3800
3750
1500 3700
3650
0.00029
3600 ESIS deviation limits
1000 3550
+/- 1 % from measured value
3500
3450
0.00027
500 3400
0.5 0.55 0.6 0.65 0.7
0 0.00025
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Displacement [mm]
Figure 2. Displacement accuracy of HIPS in a rising load test
The HIPS control configuration enables loading using any of the available test
parameters as inputs, i.e. pressure, force, displacements, strain and of course
time. The usual modes are pressure (or load) as a function of time for creep and
displacement as a function of time for tensile and fatigue testing.
The HIPS technology can also be utilized for applications where movement in
itself is the essential feature. A good example of this is the HIPS powered
contact electrode resistance (CER) [9] measuring device. In the CER prototype
tests, an accuracy of ± 0.1 μm can be accomplished for the electrode movement
in air [9].
212
(HIPS)
2. Power plant water environments

For the challenging water environments of the nuclear light water reactors, i.e.
boiling water reactors (BWR), pressurized water reactors (PWR) and the future
envisaged supercritical water reactors (SCWR), the simulated test environments
are crucial for validation of the structural materials. Testing in these conditions
has traditionally been conducted with hydraulic or electrical step motor systems
with lead throughs for the loading rod. The HIPS technology is improving the
integrity and the reliability of the tests by introducing a system without parts
moving through the pressure boundaries. As shown in Table 1, the wide range of
nuclear water chemistries, pressures and temperatures have been covered, as
well as different test types including fatigue, tensile and crack initiation/growth
testing. An example of a fatigue test is given in Figure 3.
Half life 300

loop at4
specimen
s t re s s , M P a
+/-N0,4=%
250
1500
200
150
Crack closure 100

visible at
50
N = 2550
strain, %
0
-0,5 -0,3 -0,1 0,1 0,3 0,5
-50
-100
-150
5 cycles
-200 200 cycles
1500 cycles
-250 2550 cycles
3000 cycles
-300
Figure 3. Fatigue cycles (0.1 Hz) for AISI 316 at 350°C / 160 bar (PWR environment).
An example of a typical load-displacement curve of Inconel 182 with the effects

of sulphate intrusions is shown in Figure 4. The load starts to decrease quickly
after sulphate is injected into the autoclave inlet water because of crack growth.
When the sulphate intrusion is finished, the load decrease rate begins to slow
down immediately. The lowest crack growth rates were measured in pure water
213
(HIPS)
prior to the first sulphate transient. The injection of a small amount of sulphate,
producing a minor increase in the conductivity of the outlet water, affected the
measured crack growth rate considerably in this alloy.
3000
SO42- to
SO42- to
Pre-loading 10 ppb
2- 30 ppb
level (0 ppb SO4 )
2500
2-
2000 SO4 to
10 ppb
Load, N
1500 2-
SO4 to
0 ppb
1000 SO42- to
0 ppb
500 2-
SO4 to
100 ppb
0
0.00 0.05 0.10 0.15 0.20 0.25
Δ, mm
Figure 4. Example of a typical load-displacement curve of Inconel 182 with the effects of
sulphate intrusions [1].
3. Nuclear reactor in-pile environments

It has been demonstrated that it is technically feasible to carry out well defined
and controlled in-reactor dynamic tensile tests. This makes it possible to
investigate the intrinsic role of the applied stress and the displacement damage
acting concurrently in determining the global deformation behaviour of the
material under dynamic irradiation conditions. The HIPS device has been used
to determine stress and strain curves for pure Cu and Cu-alloy specimens in the
BR-2 reactor in Belgium. The reactor pool water temperature was 90°C, neutron
flux ~0.3*1014 n cm-2s-1s (E > 1 MeV) and damage rate ~2*10-4 h-1. The test was
a constant displacement rate test with a strain rate ~10-7 1/s. The most significant
feature of the results was that during the dynamic in-reactor test, the material
deforms uniformly and in a homogeneous fashion without showing any signs of
yield drop and plastic instability (i.e. low temperature embrittlement) as
214
(HIPS)
commonly demonstrated by post-irradiation experiments. However, the fact that

the material under the present irradiation and test conditions only shows a
uniform elongation of about 12% is still a matter of concern, particularly since it
remains unknown as to what causes this reduction in the uniform elongation.
Further investigations are necessary to determine and understand the factors
responsible for the reduction in the uniform elongation due to irradiation. The
same pure copper tensile tested at 100oC in the unirradiated condition yields, for
example, a uniform elongation of about 56%. New experiments have been done
addressing this issue. An example of the test series with OFHC copper is
presented in Figure 5.
350
(a)
OFHC - COPPER
300 dose = 0.047 dpa
-3 -1
δε/δt = 1.2x10 s
250 Post-irradiation
STRESS (MPa)
200
In-reactor
150 Test No.2 Test No.1
-3 -1
δε/δt = 1.2x10 s
-7 -1
100 δε/δt = 1.3x10 s
Unirradiated
50
Tirr = Ttest = 363K
0
0 5 10 15 20 25 30
STRAIN (%)
Figure 5. In-pile and post irradiation tensile tests of copper in BR2-reactor
The copper alloys components in ITER are expected to be exposed to thermal

and mechanical loads (both static and cyclic) at the same time as being
bombarded with a neutron flux. At the present time it is not known what the
dynamic effect of the applied stress will be on the damage accumulation and
thus on the mechanical performance of the copper alloy. It has already been
shown that copper alloys are sensitive to the dynamic interaction between
fatigue and creep. These effects could significantly affect the lifetime of
components. The next step in the in-pile testing of these materials includes creep
fatigue testing. An example of the creep fatigue HIPS is shown in Figure 6, and
215
(HIPS)
the creep fatigue strain controlled signal with stress response is presented in
Figure 7.
Figure 6. HIPS probe for in-pile fatigue testing
Figure 7. In-situ fatigue stress and strain curves and selected loops for a Cu-alloy
216
(HIPS)
4. High temperature environments

For the high temperature applications where creep is an active damage
mechanism the HIPS technology opens opportunities for most of the creep
related testing types. To date the experience of the high temperature HIPS has
been collected from constant load tests performed with a prototype single
bellows (tensile load only) system and from the recently developed second
generation push/pull bellows version where creep-fatigue tests have been
performed. An example of a 600°C creep-fatigue test data set is shown in Figure
8. The loading frame with the bellows is shown in Figure 9.
500
400
300
200
10
100
Stress (MPa)
55
0
-0.60 % -0.40 % -0.20 % 0.00 % 0.20 % 0.40 % 0.60 % 100
-100
562
-200
680
-300
-400
-500
Strain (%)
Figure 8. Creep fatigue loops for P24 steel at 600°C.
217
(HIPS)
Figure 9. The high temperature HIPS in front of the furnace.
5. Discussion and future plans

It is evident that the HIPS technology enables many different options for
materials testing, also in difficult environments. The HIPS technology is also
being developed towards smaller dimensions. The best example for this
development is the miniature autoclave concept under construction both at VTT
and JRC Petten (Netherlands). This testing system is enabling effective adaptation
to the space limitations in supercritical conditions. The compact unification of
the testing / loading frame with the autoclave itself will make the testing in flowing
media (like difficult Gen IV coolants) within reach. One of the more advanced
developments towards miniature autoclaves for the SCW ( 350 bar / 650°C)
conditions is shown in Figure 10 (left side), together with another in-situ device
or the MELODIE testing system (right side) for multiaxial loading developed in
cooperation with CEA (France). The multiaxiality is here controlled by
combined tensile loading and internal pressure of the specimen.
218
(HIPS)
Figure 10. Double bellows SCW (left) and MELODIE (right) testing devices.
References
1. Moilanen, P. Pneumatic servo-controlled material testing device capable of operating
at high temperature water and irradiation conditions. Espoo: VTT, 2004. VTT
Publications 532. 154 p. ISBN 951-38-6384-0; 951-38-6385-9.
2. Singh, B. N., Tähtinen, S., Moilanen, P., Jacquet, P. & Dekeyser, J. In-reactor uniaxial
tensile testing of pure copper at 90 ºC. Journal of Nuclear Materials, 2003. Vol.
320, No. 3, pp. 299–304. (TENSILE IN PILE).
3. Singh, B. N., Johansen, B. S., Tähtinen, S., Moilanen, P., Saarela, S., Jacquet, P.,
Dekeyser, J. & Stubbins, J. F. Final report on in-reactor creep-fatigue deformation
behaviour of a CuCrZr alloy: COFAT 2 2008. Risø National Laboratory for
Sustainable Energy, Roskilde, Denmark. Riso-R-1629. 41 p.
4. Guimbal, Ph., Auclair, M., Carassou, S., Moilanen, P., Tähtinen, S. & Villard, J.-F.
MELODIE, an advanced device for study of the irradiation creep of LWR
cladding with online biaxial control of stress and strain. IGORR International
Group on Research Reactors. 12th Conference, Beijing, China, 28–30 October
2009. 5 p.
219
(HIPS)
5. Karjalainen-Roikonen, P., Moilanen, P., Toivonen, A. & Aaltonen, P. Method develop-

ment for studies of environmentally assisted cracking (EAC). In: Solin, J. et al.
(Eds.). RATU2, The Finnish Research Programme on the Structural Integrity of
Nuclear Power Plants: Synthesis of Achievements 1995–1998. Espoo: VTT,
1998. VTT Symposium 190. Pp. 105−123.
6. Toivonen, A., Moilanen P., Pyykkönen M., Tähtinen S., Rintamaa R. & Saario T. The
feasibility of small size specimens for testing of environmentally assisted
cracking of irradiated and materials under irradiation in reactor core. Nuclear
Engineering and Design, 1999. Vol. 193, pp. 309–316. (PSCFM device)
7. Solin, J., Karjalainen-Roikonen, P., Moilanen, P. & Marquis, G. Fatigue testing in

reactor environments for quantitative plant life management. 2nd International
Conference on Fatigue of Reactor Components. Snowbird, UT, 29–31 July
2002. EPRI OECD USNRC. 16 p. (FATIGUE)
8. Solin, J. Fatigue of stabilized ss and 316 NG alloy in PWR environment. Proc.

PVP2006-ICPVT-11. ASME Pressure Vessels and Piping Division Conference.
Vancouver, BC, Canada, 23–27 July 2006. American Society of Mechanical
Engineers, ASME. 11 p.
9. Moilanen, P., Arilahti, E., Bojinov, M., Laitinen, T., Mäkelä, K., Mäkelä, M., Mäkinen, R.,
Saario, T., Sirkiä, P. & Toivonen, A. Pneumatic servo-controlled fracture resistance
measuring device (PSFM-Device) and contact electric resistance measuring
device (CER Device). Enlarged Halden Programme Group Meeting, Loen, Norway,
24−29 May 1999. 16 p.
220
Characterization of true stress-strain behavior using optical monitoring
Characterization of true stress-strain

behavior using optical monitoring
Tom Andersson, Lauri Elers, Anssi Laukkanen and Pekka Nevasmaa

Espoo, Finland
Abstract
The present paper discusses the development and validation of miniature

specimen techniques for the characterization of inherent strength properties of
local microstructural regions of the weldment. The technique is based in in-situ
optical monitoring of the specimen deformation behavior, particularly the
localization of deformation, during the entire tensile test event using optical
monitoring system. Also post-processing techniques are developed and documented
for the system output. Examples of weldment characterization using 1 × 2 × 8 mm3
size specimens and determination of the associated true-stress-strain-curves are
given.
1. Introduction
Availability of true-stress-strain data and quantitative depiction of local state of
deformation is vital in structural and fitness-for-services (FFS) analysis [1], in
order to reliably assess fracture mechanism and failure behavior in externally
loaded welded structures. This is particularly important in the case of new
thermomechanically processed and mechanically strain hardened steel grades,
since welding inevitably results in local softening and/or hardening phenomena
in their ‘metastable’ microstructure. As a result, the mechanical properties can
often vary substantially between the parent steel, the heat-affected zone (HAZ)
and the weld metal. Moreover, a material property gradient also exists across the
HAZ microstructure with the distance from the fusion boundary. Due to the
221
possibility to apply very small-size specimens, optical monitoring gives an

opportunity to capture and study the tensile behavior of the different local,
narrow microstructural regions within the weldment HAZ. This way, the inherent
stress-strain characteristics of these microstructures can be quantitatively
determined.
The optical monitoring system studied here can be used to test several
different types of materials ranging from structural steels and stainless steels to
copper alloys. The challenges relating to the use of specimens of different size
and geometry (e.g., flat-bar, round-bar), however, are no yet satisfactorily
known. This requires systematic study of the monitoring event and possible
monitoring errors arising from the detection failures caused by specimen
geometry related aspects.
2. Objective and Materials

The aim was to develop and validate tensile testing technique that uses miniature-
size, i.e., 1 × 2 × 8 mm3 flat-bar tensile specimen geometry for the characterization
of different metallic materials and weldments both in parallel and transversal
directions in relation to the weld. In conjunction with continuous optical
monitoring of the specimen during the entire test event, the particular intention
was to measure and monitor the localization of deformation in a specimen, with
material’s true-stress-strain curve and the associated strain-hardening parameters
as awaited outcome of the test. The work also aimed at defining the capabilities,
potency and limitations of the optical monitoring system in this respect.
As candidate material, multiple-pass GMAW weldments were made onto
austenitic stainless steel grade EN 1.4318 (AISI 301 LN) in 5 mm plate
thickness and mechanically strain-hardened by cold rolling into C850 delivery
condition (i.e., the ultimate tensile strength of 850 MPa). The applied filler metal
was grade 308 LSi. Welding was made using three different levels of heat input:
2.4, 3.3 and 4.8 kJ/cm; consequently, three series of specimens: M3, M4 and M6
were tested, respectively. An example of macrosection of the investigated
GMAW weldment is presented in Figure 1, together with the marked locations
from where the 1 × 2 mm2 size miniature flat-bar tensile specimens were
extracted within the narrow HAZ region of the welded joint [2]. The intention
here was to sample both the grain-coarsened HAZ microstructure at the
immediate vicinity of the weld fusion boundary, and the fine-grained/inter-
222
critical HAZ microstructure further off the fusion boundary. In the former case,
sampling also some amount of weld metal could not be avoided, see Figure 1.
As reference material, direct-quenched & tempered (DQT) bainitic-
martensitic extra-high strength structural steel plate HT80 manufactured
by Sumitomo Metal Industries, Japan, was adopted [3]. As only base material
was investigated here, round-bar tensile specimen geometry of Ø 6 mm diameter
size was used instead of flat-bar specimen type.
Figure 1. Example of the macrosection of the GMAW multipass welded austenitic EN

1.4318 (AISI 301 LN) steel, 5 mm plate thickness. Extraction of the 1 × 2 mm2 size flat-
bar specimens (in realistic scale): inner rectangular lines: specimen’s measuring gauge
dimensions; outer rectangular lines: specimen-end dimensions. Right-hand side:
specimen sampling the grain-coarsened HAZ immediately adjacent to the fusion
boundary (with some weld metal included); left-hand side: specimen sampling the outer
HAZ (without any weld metal involved) [2].
223
3. Methods
The optical “Aramis” monitoring system applied in the present study includes
two cameras, a tensile testing machine and a computer in which the software for
data acquisition and processing is installed. The tensile testing machine is used
as in traditional tensile tests. The system takes pictures of the specimen at fixed
intervals and it uses these pictures to measure strain components (in x-, y-, and
z-directions). Test specimen is painted gray with a plain mat color.
The system adds a grid on top of the specimen in processing of the pictures
(4–6 squares in vertical direction in miniature specimen). The position of each
square in the grid is determined by tracing certain pixels inside the square. The
system specifies the pixels by a color variation. At the beginning of the analysis
squares are partially on top of each other so that there are no gaps between them
at the end of the analysis. The system determines the strain components at each
increment by calculation of the displacement at each square in the grid and
numerically differentiating the resulting field.
As the “Aramis” system employs two cameras in angle to each other, all three
strain components can be determined. This is especially helpful in tests that are
made using round-bar specimens. In standard tensile test only the mean value of
strain can be attained, whereas the present optical system produces local values
of strain. The system also produces a video of all the pictures it has taken.
Flawed, or otherwise disturbed, data can be cut down during the analysis. Such
flawed data can occur due to fluctuation in lighting conditions especially at the
boundaries of the tensile specimen. Displacement velocity used in the present
system is usually slower than in standard tensile tests to aid image processing.
The actual tensile test being completed, the data produced by the system is run
through a computer script, see Figure 2. This post-processing phase extracts
flawed stages and pin points the necking specimen section.
4. Results
The following figures (Figures 3–5) present the true stress-strain curves
determined for the EN 1.4318 steel multi-pass GMAW weldments using the flat-
bar specimens extracted from the weld metal and the HAZ c.f. Figure 2.
Comparison between true and engineering stress-strain curves are shown in
Figure 6. Figure 7 presents the true stress-strain curve for the DQT HT80 steel
base material determined using round-bar specimens.
224
Figure 2. Flow chart of the post processing procedure [2].
225
Figure 3. True stress-strain curves for the weld metal and the HAZ – specimen M3 [2].
226
Figure 6. True and engineering stress-strain curves for the HAZ – specimen M4 [2].
227
Figure 7.True stress-strain curve for the DQT HT80 steel: round-bar specimens. [4]
Figure 8. Images from “Aramis” video output: 1 at the beginning of test, 2 at the end of
the elastic area, and 3 at end of the test.
228
Figure 8 shows three contour plots of specimen geometry. First contour plot is
from the beginning of tensile test, second in the end of elastic region and third in
the end of tensile test when reduction of area is the largest before the final
specimen breakage.
5. Discussion
Fracture mechanics based integrity assessment of mismatching welds and their
interfaces are well presented in the FITNET procedure [1]. Estimation
procedures for welds with mismatching mechanical properties are one of the
most complicated areas of fracture mechanical fitness-for-purpose evaluation.
Thus, often various approximating steps need to be undertaken in an effort to
perform and complete a specific analysis. Numerical modeling can thereby be
based on detailed finite element analyses (FEA) of the structure in question. The
experimental characterization needs to be accomplished with the necessary
accuracy in order to sample those particular weldment regions displaying the
most significant mechanical property mismatch in terms of strength and
toughness. Care should be undertaken in the selection of an appropriate testing
approach; these typically rely on sc. ”small specimen” testing techniques both
with respect to mechanical (tensile) properties and fracture mechanical
toughness characterization. In evaluation of the crack driving force and in
selection of applicable plasticity corrections (failure assessment diagrams),
however, one typically needs to utilize pre-existing solutions for the stress
intensity factor (SIF), limit load, failure assessment diagram (FAD) etc., or take
the long road and derive these parameters on the basis of a FEA. In both cases,
the decision on how material properties are treated needs to be made upon and is
one of the crucial stages in determining the analysis outcome and the quality of
the attained results [5].
The FITNET structural integrity assessment method requires principal
material tensile and fracture mechanical properties for carrying out the analysis.
Particularly, some form of true stress-strain curve must be specified and fracture
mechanical properties need to be specified by one of various available means.
Option 1, the standard option, is recommended for use when the mismatch in
tensile properties is less than 10%. Option 2, the mismatch option, is able to treat
situations where both base and weld metal have an unspecified degree of
mismatch. Option 3 is also able to handle mismatch via an equivalent stress-
strain curve concept. In FITNET terms, the numerical analyses are option 4
229
analyses. The FAD is determined on a case-by-case basis for the combination of

structure and material over a range of applied loading from linear-elasticity to
plastic collapse [5].
Tensile test results obtained for GMAW welds on strain-hardened austenitic
stainless steel EN 1.4318 (AISI 301 LN) using miniature flat-bar specimens with
the cross-section geometry of 1 × 2 mm2, are presented in Figures 3–6. The
results demonstrate that in all cases the weld metal exhibits undermatching
strength, in relation to the tensile properties of the outer HAZs and, finally, the
base material (in C850 delivery condition). According to a parallel study [2],
local softening in the EN 1.4318 steel (C850) as a result of welding occurred,
and progressively advanced, via diffusion controlled mechanism at peak
temperatures exceeding 900 °C. Bearing in mind that welding was made using
AISI 308 LSi type filler metal with considerably lower strength (Rp0.2 = 370
MPa; Rm = 620 MPa) than the strain-hardened parent steel, it can be concluded
that the true-stress strain curves according to Figures 3–6 provide a realistic
description of the strength and stress-strain characteristics of the different
regions of the GMAW weldments in question. Thus, the local, narrow weld
regions with mismatching strength were successfully hit and sampled using
miniature size 1 × 2 mm2 flat-bar specimens (c.f. Figure 2).
Figure 7 shows that it is also possible to use round-bar specimens with
geometry of Ø 6 mm diameter size for the determination of the true-stress strain
characteristics. The results are comparable with those of miniature flat-bar
tensile specimens [4].
One advantage of the present technique is that the initial set up procedure is
fast since the test specimen only needs to be painted and no strain gauges are
needed. As the present optical monitoring system uses two cameras to record the
tensile test this allows one to extract strain data from different parts of the
specimen even after the test has already ended. Recorded data can be used to
create visually superior videos.
With respect to the present technique, some challenges still exist. These
include for instance the necessity for the paint layer to be smooth. In addition,
the post-processing procedures lack in some sophisticated features that are
needed in advanced fracture mechanical analyses; thus VTT has developed its
own code to extract specific data from “Aramis” system output.
230
6. Conclusions
The present “Aramis” equipment enables continuous optical monitoring and
complete mapping of localization of deformation in a specimen during the
tensile test. Thus, true stress-strain curves with strain hardening characteristics
can be successfully obtained using flat-bar specimens as small as 1 × 2 × 8 mm3.
As shown in Figure 7, the system still has some difficulties of tracking,
especially, the boundary grid squares. Fortunately this flaw is simply avoided in
the post-processing by removing from the data the stages that differ too much
from the previous one.
True-stress-strain curves provide more accurate information than their
engineering equivalents for the use in computational fracture mechanics.
Determining the inherent properties of various mismatching weldment regions
using miniature specimen techniques enables to gain full advantage of
sophisticated numerical non-linear modeling techniques that allow the use of
continuous material property descriptions. Especially regarding the FITNET
procedure, FADs produced from the true stress-strain curves are more accurate;
hence safety margins can be smaller when applying FADs to actual structural
design.
7. Future works
In the following works, the present optical monitoring system is planned to be
used in measure the strain rates in a specimen that consist more than one weld
zone (i.e., transverse in relation to the weld). These tests would be compared
with ones that are done to specimens consisting only one zone (i.e., parallel to
the weld). It is considered extremely beneficial to develop analyzes procedures
capable of successfully characterizing material properties across multiple zones
with only one single specimen. Experiments to analyze specimens that include
cracks are also planned.
8. Acknowledgements
The present paper belongs to the ‘FRAS’ project within the national “SAFIR
2010” nuclear research programme. Provision of the primary data from the
TEKES ‘MIS-MATCH’ project is appreciated.
231
References
1. ’FITNET’ Fitness-for-Service (FFS) Procedure. Vol. 1. Eds. M. Kocak, S. Webster, J. J.
Janosch, R. A. Ainsworth, R. Koers. European Fitness-for-service Network.
CEN 2006.
2. Nevasmaa, P., Sirén, M., Alhainen, J., Andersson, T., Elers, L., Muukkonen, T.,
Somani, M. C. & Karjalainen, L. P. Paikallisen alilujuuden (mis-match) vaikutus
muokkaamalla lujitettujen austeniittisten ruostumattomien terästen hitsausliitosten
ominaisuuksiin. Espoo: VTT, 2009. VTT Tutkimusraportti VTT-R-06034-09. 50 s.
(In Finnish).
3. Slater, S., Priest, A. H., Álvares, J. A., Gallo, C., Polanco, J. A., Saiz, A., Gutiérrez-
Solana, F., Ponsot, A., Wallin, K., Laukkanen, A., Demofonti, G., Di Biaggio, M.
& Mecozzi, E. An energy balance approach for crack arrest – final report.
Luxembourg, 2004. 171 p.
4. Elers, L. The validity of fracture resistance curve characterization methods for

extended ductile crack growth. Espoo: VTT, 2008. VTT Research Report VTT-
R-10197-08. 58 p.
5. Laukkanen, A., Karjalainen-Roikonen, P., Nevasmaa, P., Andersson, T. & Elers, L.

Fracture mechanical fitness-for-purpose assessment of undermatching high
strength steels welds. Espoo: VTT, 2009. VTT Research Report VTT-R-01115-
09. 45 p.
232
Non-contact measurement of oxide thickness with a laser ultrasonic technique
Non-contact measurement of oxide

thickness with a laser ultrasonic technique
Jonne Haapalainen1, Joona Eskelinen1, Tarja Jäppinen2,

Stefan Sandlin2 and Edward Hæggström1
1
Electronics Research Laboratory, Dept of Physics,
University of Helsinki, Helsinki, Finland
2
Abstract
Non-contact laser ultrasound (LU) method is a potential tool for structural health
evaluation in power plants. In this study a LU surface wave method was
employed to determine the oxide layer thickness in zircaloy tubes used as
cladding of fuel rods in nuclear reactors. Ultrasound excitation was done with a
pulsed laser whereas a laser doppler interferometer was used to remotely receive
the propagating ultrasonic signals. To show the method’s capability results from
a 8 µm zirconium oxide layer measurement are presented.
1. Introduction
Ultrasonic non-destructive evaluation is widely used in structural health
evaluation. It can be used to detect e.g. cracks and corrosion, but it requires
typically contact or immersion between the transducer and the sample. Non-
contact laser ultrasound (LU) provides means to overcome these limitations e.g.
in high temperature or radiative environments. Compared to traditional
ultrasound testing, LU provides better spatial resolution and wider frequency
bandwidth. We present principles of the LU method and give an example of its
use for oxide layer thickness measurement.
233
1.1 Ultrasound
Ultrasound is a propagating mechanical disturbance. The velocity and attenuation

of ultrasonic waves are affected by the mechanical properties of the sample.
Ultrasound can propagate in multiple modes, cfr Table 1.
Table 1. Different wavemodes employed in ultrasonic testing.
Bulk waves Surface waves Guided waves
Propagates in bulk Propagates along the Propagates along thin

material. Reflects from surface of the material. Can surfaces, such as tubes.
defects in the bulk material. be used for surface Velocity depends on the
characterization. sample geometry and
frequency
1.2 Laser Ultrasound
Laser ultrasound [1] is used for non-destructive evaluation (NDE) in this study.
A short (few nanoseconds) laser pulse is absorbed close to the surface of the
sample. Thermal expansion due to heating excites an ultrasonic wave which
probes the sample (Figure 1). Broadband, high frequency, (1 ns pulse generates
~ 100 MHz waves) ultrasonic waves can be excited. Moreover, the efficiency to
generate different wave modes is good. In metals the normal skin depth of the
excitation laser penetration into the sample is a few nm, which provides local
excitation (the beam diameter can be as small as 10 µm). Propagating ultrasonic
waves can be detected by measuring the surface displacement (out-of-plane as
well as in-plane) of the sample with an interferometric receiver providing a non-
contact and remote measurement.
234
Laser Thermal Ultrasonic

excitation Absorption Heating expansion wave Detection
Figure 1. Principle of laser ultrasound testing.
1.3 Thickness profiling
Surface waves can be used for thickness profiling of a thin layers [2–4].
Typically surface waves probe a depth which equals their wavelength. If the
material parameters changes as a function of depth (for example different
layers), varying the wavelength allows probing the layer structure. The material
gradient is seen as dispersion (frequency dependent velocity), Figure 2.
Figure 2. Dispersion of surface wave in a zircaloy tube as a function of covering oxide

layer thickness.
235
2. Experiments
We determined the thickness of an oxide layer covering a zircaloy tube sample.
The oxide layer was grown in an oven in O2 + Ar atmosphere. The layer
thickness was also measured destructively with an optical microscope (Figure 3).
Oxide
Cladding
Figure 3. Optical image of the sample showing the cladding, oxide and backing epoxy.
Excitation was done with a Nd:YAG laser emitting a 1 mJ, 1 ns long pulse with
a 100 Hz repetition frequency. The excitation beam was scanned with a mirror
that was attached to a translation stage. Forty 100 µm long steps were taken for a
total of 4 mm measurement distance. Non contact receiving was done with a
laser doppler vibrometer (LDV). Each measurement point was averaged 100
times to improve the signal-to-noise ratio (SNR).
A 2D-FFT was next performed on the measured data to determine the phase
velocity of the propagating surface waves in frequencies between 9 MHz
(limitation of method due to finite thickness of base cladding) and 20 MHz
(upper frequency of the receiver). A theoretical fit to the measured dispersion
curve was used to determine the thickness of the oxide layer.
236
3. Results
Results of the oxide thickness measurement obtained with both LU and
destructive optical methods are presented in Table 2. The root mean square error
(RMSE) between the surface wave velocity measurement and the theoretical
curve was less than 1 m/s (< 1% of velocity range), which indicates good
agreement between the theory and measurement.
Table 2. LU vs. optical measurement of the oxide thickness.
Oven time Optical measurement LU measurement

0h 0 µm 0 µm
6h 8 µm 6 µm
4. Conclusions
Principles of LU method for NDE was presented. The proposed LU oxide
thickness measurement correlated well with the reference optical measurement
for thin samples. The difference between the optical and the LU thickness
estimate is probably due to a difference in measurement position and unequal
oxide thickness. When the measurement was done on thicker (> 10 µm) layers,
the RMSE between the theory and the measurement exceeded 10 m/s, which is
assumed to be due to problems in layer integrity.
Laser ultrasound holds potential for multiple other measurements in the area
of NDT in power plants. Possible application areas include:
• Crack detection in power plants [5]
• Changes in elasticity of fuel tubes [6]
• Wall thinning measurements [7].
References
1. Scruby, C. B. & Drain, L. E. Laser ultrasonics: techniques and applications. Inst of Physics
Pub Inc., 1990.
237
2. Every, A. G. Measurement of the near-surface elastic properties of solids and thin

supported films. Measurement Science and Technology, Vol. 13 (2002) 5,
pp. 21–39.
3. Windels, F. W., Vanaverbeke, S. & Leroy, O. Thin coating characterization by Rayleigh

waves: An analytical model based on normal-mode theory. The Journal of the
Acoustical Society of America, Vol. 110 (2001) 3, pp. 1349–1359.
4. Hurley, D. C., Tewary, V. K. & Richards, A. J. Surface acoustic wave methods to

determine the anisotropic elastic properties of thin films. Measurement Science
and Technology, Vol. 12 (2001) 9, pp. 1486–1494.
5. Ochiai, I. Development and Applications of Laser-ultrasonic Testing in Nuclear Industry.

In 1st International Symposium on Laser Ultrasonics: Science, Technology and
Applications. Montreal, Canada, 2008.
6. Liu, I. & Yang, C. Laser Ultrasound Technique for Nondestructive Characterization of

Material Properties in Hydrogen-Charged Zircaloy Tubes. In 1st International
Symposium on Laser Ultrasonics: Science, Technology and Applications.
Montreal, Canada, 2008.
7. Murfin, A. S. & Dewhurst, R. J. Estimation of wall thinning in mild steel using laser
ultrasound Lamb waves and a non-steady-state photo-emf detector, Ultrasonics,
Vol. 40 (2002) 1–8, pp. 777–781.
238
On-line flaw growth monitoring in high temperature plant
On-line flaw growth monitoring in high

temperature plant
Ian Atkinson, KANDE International Ltd, Knutsford, UK

Stephen Kelly, Doosan Babcock, Renfrew, UK
Philip Lay, Doosan Babcock, Renfrew, UK
Colin Brett, E.ON Engineering, Nottingham, UK
Abstract
The operating conditions experienced by power plants place high demands on

plant components, and ensuring plant availability without compromising safety
or plant integrity can be a significant challenge to operators. On-line
measurement of known flaws in plant has the potential to provide plant operators
with substantial savings in overall operating costs through the better
management of inspection and repair activities. This is particularly true of high-
temperature plant, where the preferred inspection techniques for measuring
crack growth are unsuitable for on-line application and must be applied during
plant shutdown, which can lead to low accuracy and substantial inspection costs,
as well as loss of production.
To address such problems, we have been working on the development of in-situ,
high-temperature ultrasonic crack growth measurement techniques for several
years, and will this year proceed to a trial installation at a UK fossil-fuel plant.
This paper reports the results of a laboratory trial to demonstrate the
measurement of fatigue crack growth in a thick-walled ferritic steel pipe section
at 615°C. This was part of a programme of tests performed to verify the viability
of the transducers and crack-growth measurement techniques, prior to the trial
on operational plant.
239
1. Introduction
The operational demands placed on current generation power plants, particularly
those in countries where commercial markets in power generation operate, can be
severe and ultimately have an adverse effect on plant integrity. For example, thermal
cycling associated with load following can lead to fatigue crack growth, while
excursions above the design operating temperature can accelerate creep damage.
Because such damage can be expected to occur, NDE inspections of vulnerable
components are performed periodically with the aim of detecting new flaws.
When in-service flaws are discovered in plant components, it is not always
feasible or necessary to replace them immediately. However, in such situations it
is important to determine if the flaw is growing and if so, at what rate, so that
planning for its management and eventual possible repair or replacement can be
undertaken. In the case of cracks in steel components, the preferred way of
measuring flaw growth is by periodic ultrasonic inspection during plant
shutdown. The cost of such inspections, including NDE services, site preparation
and reinstatement, plus the cost of lost production, can be substantial, especially
if it becomes necessary to have plant outages specifically to accommodate the
flaw monitoring requirement.
Using periodic inspection to measure flaw growth accurately can also be
problematic. The variations arising from small differences in measurement
location and inspection set-up, combined with random errors in data analysis,
can make it difficult to be certain if measured growth is real, or merely a
reflection of the uncertainty between repeated measurements. This can lead to
the situation where flaws apparently grow rapidly over one operational cycle and
shrink over the next. The avoidance of such problems requires careful inspection
design and implementation to minimise both systematic and random errors. In
ultrasonic inspection, such improvements may be achieved by the use of
permanently installed probes.
For high temperature plant such as power stations, the above problems are
likely to be exacerbated for a number of reasons, for example:
• Outage duration may be extended because of the time required for the plant
to cool to a temperature at which the work can be performed without risking
harm to personnel or damage to inspection equipment.
• Surface preparation to remove high temperature oxidation products that
adversely affect ultrasonic coupling can also remove reference marks and
240
change the surface form, making it more difficult to obtain accurate repeat
measurements.
• In-service cracks may be subjected to compressive stresses at ambient
temperatures, making them more difficult to detect and measure accurately.
High temperature in-situ monitoring with permanently attached probes would
avoid these problems:
• Continuous access to accurate information on flaw size makes shutdown
specifically for inspection unnecessary. In addition, it enables the operator to
build up detailed information on flaw growth rates, which will help to
inform decisions such as when to repair or replace affected components.
• Once probes have been installed, further surface preparation is unnecessary
and hence uncertainties of probe position and component shape are
eliminated and measurement accuracy is increased.
• Measurements are made under the load conditions that are driving the flaw
growth and hence crack closure is unlikely to be an issue.
Until recently however, it has not been practical to perform long-term in-situ
ultrasonic measurements on high-temperature plant due to the lack of
transducers with the required sensitivity and the robustness necessary for long-
term application in an operational environment subject to both thermal cycling
and temperature extremes. In this paper we report on work performed by
KANDE International, Doosan Babcock and E.ON to develop a capability for
measuring fatigue crack growth in ferritic steel, thick-walled pipes at
temperatures around 600°C.
2. Background
The work reported in this paper is a continuation of work started in an earlier
collaborative project known as ULTRASMART (Ultrasonic Scanning and
Monitoring at Running Plant Temperatures) (1, 1). Under the ULTRASMART
project, KANDE developed a robust, high temperature ultrasonic probe capable
of working at temperatures greater than 600°C for extended periods of time. In
September 2007 three prototype transducers were installed on a steam-pipe at
E.ON’s Ratcliffe-on-Soar power station (Figure 1), one measuring wall
241
thickness and the other pair in a Time-of-Flight Diffraction (TOFD)

configuration suitable for crack growth measurement.
Figure 1. High temperature probe installation at Ratcliffe in 2007.
Due to the exploratory nature of the study, these probes were installed well away
from any weld or known flaws, so as to avoid any problems arising from adverse
effects associated with the installation procedure. At the time of writing these
probes are still working after operating for more than 11 500 hr at 565°C and
experiencing more than 70 plant starts.
Because of the encouraging results obtained, it was decided to proceed to a
trial installation on a known flaw in 2010. It is planned that this installation will
be made on a defective weld that is scheduled to be replaced in 2012, as this will
provide an opportunity to obtain confirmation of the measured flaw size by
destructive examination.
The planned trial is a considerable commitment for all the parties involved
and in order to reduce the risks and uncertainties associated with it, laboratory
studies have been performed to demonstrate and verify the performance of the
transducers, the welding procedures for transducer attachment and the
measurement technique itself. The results of the latter study are reported below.
242
3. Experimental Work
3.1 Specimens
The experimental work reported here was performed on fully representative

2¼Cr steel pipe sections removed from power plant service. To minimise
interface losses and any materials compatibility issues, the transducer delay lines
and the welding consumables used were also made of 2¼Cr steel.
The flaw type selected for the trial is a steam-line circumferential bore crack
(Figure 2). These are fatigue cracks that initiate at the inner surface, close to the
circumferential welds that join the pipe sections, and propagate in the through-
wall direction as a result of the stresses generated by thermal cycling as the plant
is stopped and started, according to the demand for power.
Figure 2. Section through a steam-line circumferential bore crack.
243
For the laboratory trial of the proposed inspection technique, the aim was to
simulate the case of a steam-line bore crack by measuring the growth of a
fatigue crack in a specimen subjected to cyclic loading at high temperature. The
specimen used was 464 mm long by 100 mm wide with a wall thickness of
approximately 62 mm. A full-width starter notch was machined across the
middle of the inner surface of the specimen and was approximately 3 mm deep
at the specimen centreline.
3.2 Inspection technique
The technique selected to monitor the growth of these cracks is a variant of the two-
probe TOFD technique, in which a beam of ultrasound from a transmitter probe
impinges on the crack and forward diffracted signals from the crack tip are detected
by a receiver probe symmetrically positioned on the opposite side of the weld.
In conventional TOFD inspection, a pair of probes is scanned on the
component surface, and the movement of the probes relative to the flaw gives
rise to characteristic arc-shaped signals that are used to identify the flaw and
determine the flaw extent. In the case of in-situ monitoring, probe movement is
not possible and therefore the characteristic arc-shaped signal shapes cannot be
obtained. Instead it is necessary to monitor the signals obtained over time, in
order to discriminate between the signal from a growing crack, which will
change over time and signals from other sources such a backwall reflection or
gas pores in the weld, which will not. To collect the required data, KANDE has
adapted its USBox UT system to collect data periodically over an extended time
interval and to display it as a two dimensional image in which one axis
represents flaw depth, and the other axis represents the time elapsed since the
start of data collection.
3.3 Transducers
The transducers used for the trial reported here had 6.3 mm diameter circular
elements generating longitudinal mode ultrasonic beams with a nominal resonant
frequency of 5 MHz. The transducer delay lines were approximately 40 mm long
by 20 mm diameter and made of 2¼Cr steel. In order to give 45° angled
longitudinal beams in the test specimen and in preparation for attachment by
welding, the delay lines were machined to 45° knife-edges. Mineral insulated
semi-rigid coaxial cables were used to carry signals to and from the probes.
244
3.4 Transducer attachment
In preparation for transducer attachment, pads of 2¼Cr weld metal were

deposited on the outer surface of the specimen at the intended probe locations
and heat treated to control material hardness and prevent embrittlement close to
the flaw location.
The transducers were subsequently welded onto these pads (Figure 3) using a
welding procedure developed by Doosan Babcock. A probe centre separation
(PCS) of 104 mm was selected to give maximum sensitivity approximately
10 mm from the bore surface of the 62 mm (nominal) wall thickness pipe and to
provide good coverage of the inner half of the pipe wall at the crack position.
Figure 3. High temperature probes welded onto specimen in TOFD arrangement.
245
4. Experimental Setup and Results

The specimen was mounted in a 600 kN materials testing machine for cycling
under compressive load in a four-point bend arrangement. It was heated using a
resistance heating element and insulated using multiple layers of ceramic blanket
insulation (Figure 4). Thermocouples welded to the side of the block, were used
for temperature control and monitoring.
Figure 4. Specimen in 4-point bend rig at completion of test with insulation partly removed
to show heating element arrangement.
The specimen and transducers were heated from ambient temperature to 615°C
before fatigue cycling was started but UT data was recorded every 30 s
throughout the full duration of the experiment. The resulting record is shown in
Figure 5 and explained in detail below.
The specimen was initially set to reach a temperature of 585°C and allowed to
stabilise at this temperature overnight. The reduction in ultrasonic velocity with
increasing temperature can be seen in the increased arrival time of the ultrasonic
246
signals at the left-hand side of the image and the horizontal lines seen in the
central portion of the image correspond to the period at 585°C. The temperature
was then increased to 615°C, which can be seen as a further small increase in the
arrival time of all the visible signals.
After the temperature had stabilised at 615°C, cyclic loading was commenced,
shortly after which crack growth was observed (at approximately 20 hours
elapsed time in Figure 5). Load cycling then continued for approximately
80 minutes, with a reduction in applied load after 60 minutes to slow the crack
growth rate. After cycling was halted the specimen was held at 615°C for 30
minutes before data recording was stopped.
0 2 4 6 8 10 12 14 16 18 20 22
Elapsed time (hours)
Figure 5. UT data from monitoring trial.
The signals from the tip of the growing fatigue crack are more clearly seen in
Figure 6. This detail extracted from Figure 5 covers a period of approximately
90 minutes and shows the weaker signal from the crack tip diverging from the
higher amplitude starter notch and backwall reflection signals, giving a steeply
inclined line rising across a background of nearly horizontal lines.
247
Crack-tip signal
Notch signal
Backwall signal
0 10 20 30 40 50 60 70 80 90
Elapsed time (minutes)
Figure 6. Crack growth detail.
In addition to the inclined line representing the crack tip signal, Figure 6 shows
several other points of interest:
• The ‘inclined line’ signal increases in amplitude as the crack tip grows
towards the axis of the ultrasonic beams.
• The ‘inclined line’ curves upwards, due to the crack growth rate
increasing as the load is applied to a reducing ligament.
• The visible decrease in the slope of the curve after approximately 60 min.
indicates where the peak load was reduced to slow the growth rate.
• As the fatigue crack starts to grow, the starter notch signal does not
disappear completely but reduces in amplitude and curves increasingly
248
‘downwards’. This indicates that the notch is wide enough to provide a

target that reflects some portion of the beam back to the receiving probe until
all available paths are interrupted by the increasing large fatigue crack.
• As the crack height increases, the backwall reflected signals are also
increasingly interrupted and hence the backwall signal reduces in
amplitude and tends to slightly longer range.
• The horizontal signals at the right-hand side of the image show where
cyclic loading was stopped and the specimen held at 615°C before
recording was stopped.
Due to a number of uncertainties, of which ultrasonic velocity at the measurement
temperature, probe separation and wall thickness at the point of measurement are
the most significant, absolute crack size is difficult to determine accurately.
However, measurement of the crack growth, which is a differential measurement
where both initial and final depths are subject to the same systematic errors,
should be reasonably accurate provided the total growth is not too large a
proportion of the overall wall thickness. Hence, by assuming that:
• The longitudinal velocity in 2¼Cr steel at 615°C is 5.45 ±0.1 mm/µs (2).
• The specimen is 62 ±2 mm thick at the position of measurement.
• The effective PCS is 98 ±5 mm.
It can be estimated that the observed growth under cyclic loading, measured
from the tip of the starter notch, in the through-wall direction, was 16 ±1 mm.
The actual crack depth has not yet been confirmed by destructive examination
but Figure 7 shows the result of a magnetic particle inspection of the fatigue
crack at the edge of the specimen, performed after completion of the test. This
indicates crack growth of approximately 20 mm at the edge of the block, which
is reasonably consistent with the result obtained from the UT measurement made
on the specimen centreline.
249
Figure 7. Fatigue crack detail after completion of test.
5. Discussion and Conclusions

The experimental trials undertaken clearly demonstrate the feasibility of
detecting and measuring crack growth at 615°C using a two probe TOFD
ultrasonic technique.
The diffracted signals obtained were substantial and had a good S/N ratio.
This gives us confidence that accelerated ageing of the transducer, which is an
unavoidable result of high temperature operation, should not degrade the
performance to the point where the diffracted signals are too weak to be
detected, in the expected lifetime of the installation. This is supported by earlier
experimental work and by the previous Ratcliffe trial where the probes have
operated for more than two years without significant reduction in output.
Similarly, the signal levels obtained suggest that the higher temperature
operation is also likely to be feasible. Results obtained in earlier work lead us to
believe that performance should remain good up to at least 650°C and that there
may be some capability at temperatures as high as 730°C, with the caveat that
both sensitivity and lifetime are expected to reduce increasingly rapidly as the
operating temperature increases towards the circa 800°C Curie Temperature of
the transducer material.
250
Whilst it is clear that accurate crack growth measurement is feasible under the
constant temperature conditions used in this trial. Under plant operating
conditions, where thermal cycling is driving flaw growth and the cycling rate is
both slow and variable, accurate monitoring is likely to be more difficult.
Probably the most significant difficulty will arise from the sensitivity of the
ultrasonic velocity to relatively small changes in temperature. This may make it
difficult to differentiate between changes due to crack growth and changes due
to temperature variation. One way of addressing this problem would be by
processing the data to correct for the effects of temperature variation, with
temperature either being measured directly or estimated from the backwall echo
position. While the processing required for this is similar to that used in tools
already available for TOFD data enhancement, its application would not
necessarily be straightforward in practice, and it may be better to avoid the issue
altogether, by collecting data each time the plant passes a set temperature, rather
than collecting data at fixed intervals. This approach has the further advantage
that the recorded data could then be scaled directly in terms of the number of
thermal cycles, rather than in elapsed time, and hence any flaw growth recorded
would be more likely to appear as a linear feature, of the type seen in the trial
data. KANDE is currently adapting its monitoring systems to implement this
mode of operation.
It is also worth noting that although small amounts of crack growth can be
detected and measured with accuracy using fixed probes once the flaw signal has
been identified in the B-scan image, there will be some minimum amount of
growth that has to occur before the signal can be unambiguously identified.
Accurate positioning of the crack tip by conventional inspection techniques prior
to the installation of the fixed probes should help to reduce the uncertainty
associated with signal identification, and correlation based signal processing
techniques could also be applied, to help identify small changes between
consecutive signals before they become apparent to the naked eye.
References
1. Atkinson, I., Gregory, C., Kelly, S. P. & Kirk, K. J. Ultrasmart: Developments in ultra-
sonic flaw detection and monitoring for high temperature plant applications.
Proceedings of CREEP8 Eighth International Conference on Creep and Fatigue
at Elevated Temperatures. San Antonio, Texas, 22–26 July, 2007.
251
2. Kelly, S. P., Atkinson, I., Gregory, C. & Kirk, K. J. On-line ultrasonic inspections at
elevated temperatures. IEEE International Ultrasonics Symposium, UFFC. New
York, September 2007.
3. Andrews, K. W. Testing at high temperatures. In: Szilard, J. (Ed.) Ultrasonic Testing –

Non-conventional testing techniques. Wiley, 1982.
252
Experience on the periodic wet bar diagnosis of water-cooled generator stator windings
Experience on the periodic wet bar

diagnosis of water-cooled generator
stator windings
Hee-Soo Kim and Yong-Chae Bae

Korea Electric Power Research Institute
Daejeon, Republic of Korea
Abstract
In this paper, of various integrity evaluation methods for water-cooled stator

windings, it is verified that a wet bar diagnosis method is most sensitive.
Generally, there are vibration characteristics test, leak test, and water absorption
test, as mechanical integrity evaluation methods for water-cooled stator
windings. When a generator-rotor is extracted, mechanical integrity evaluation
for water-cooled stator windings are carried out by using all of them. In spite of
our continuous efforts, critical accidents frequently occurred and tremendous
economic cost is paid for stator winding repair. Fortunately, it is confirmed that
integrity is evaluated more clearly by analyzing capacitance change and voltage
applied into windings among various integrity evaluation methods for water-
cooled stator windings.
1. Introduction
It is expected that the mean life of a generator is approximately 20 years. The
unique design of the water-cooled generator produces more power per unit size
than a conventionally cooled generator. As a result, increased electromagnetic
forces act on the stator windings and the bar restraining system in the slot and
end-winding are designed to accommodate these increased forces. Deterioration
of any part in the bar restraining system may lead to accelerated wear of stator
253
winding insulation and damage to cooling water circuits. It is important that slot
and end-winding restraining systems be well maintained in accordance with
standard recommendations.
Figure 1 shows a structure of generator stator windings and Figure 2 shows
the flowchart for mechanical integrity evaluation of generator stator windings.
As mechanical integrity evaluation methods, there are vibration characteristic
test, leak test, and water absorption test. As based on the test results, i) vibration
characteristics of windings are changed so the resonance is prevented or ii)
leakage parts are found so they are repaired or iii) wet bars are repaired or
replaced within 1 year.
4 5
3 6 7
8
2
1
① STATOR CORE ⑥ STATOR FRAME 9

② RETAINING RING ⑦ HYDROGEN COOLER
③ ROTOR ⑧ STATOR BAR
④ BEARING ⑨ TERMINAL BOX 10
⑤ END SHIELD ⑩ HIGH VOLTAGE BUSHING
Figure 1. Structure of generator stator windings.
There have been a number of instances of stator winding cooling system water
leaks. Generally, operation of the generator has not been impacted, and it has
been possible to correct the condition without major maintenance. However,
there have been instances of stator winding insulation failures where water
penetration of the stator bar insulation, either at the series loop or in the actual
groundwall, has been a factor in the failures. In some instances, evidence
suggests that mechanical damage to the groundwall may have contributed to the
hipot failure. There is, however, no doubt that the presence of water was a
significant factor in each of these failures. As a result of these experiences, it is
evident that more attention is required to eliminate water leaks which affect the
stator bar groundwall insulation, particularly on the stator bar end arms. Even
extremely small water leaks can be detrimental if allowed to persist. This is
254
particularly the case if the stator water pumping unit is left in operation during
maintenance outages when the generator is degassed. Under these circumstances
the pressure differential provides the greatest impetus for water to be forced
through a leak site into the groundwall. Regardless, capillary action may cause
water to leak from a bar even though hydrogen pressure is maintained higher
than the cooling water pressure. If water leaks occur, they generally develop in
the region of the series or phase connections. Stator bars are stringently tested
for leaks prior to assembly, and the entire winding is tested twice after assembly
to further ensure winding integrity. Leaks which occur in service develop as a
result of normal operating conditions, and often take a long time to occur.
The diagnosis methods which are applied in wet bar diagnosis are similar to
GE’s capacitance mapping method but have some different characteristics. The
capacitance diagram method can display the distribution of capacitance of each
bar while other stochastic diagnosis methods such as normal probability plot and
box plot can diagnose the wet bar among many bars. As well, the new developed
diagnosis methods are applied to analyze wet bars for 500 MW generator stator
windings.
Figure 3 summarized results of field tests reported to GE. The data represents
approximately 50% of the tests done from January 1991 through July 1993. As
shown in Figure 3, it is confirmed that the recommended tests are enough to
detect leaks. Also, it is confirmed that mechanical testing is more conservative
than electrical testing. Similarly, Figure 4 shows the cooling water leakage
accidents by EPRI and water leakage accidents are increased rapidly in power
plants which is operated more than 14 years.
It is known that water absorption test is an effective diagnosis method to
distinguish wet bars by measuring the capacitance of insulator. Basically, the
difference of dielectric constant between water and insulator is used. However, it
is truly difficult to test and diagnose the condition of an insulator because of the
structure of stator windings and complexity of the insulator. Actually, the
capacitance between copper and insulator is very low and it is not enough to
diagnose wet bars by the stochastic process.
Now, South Korea has more 30 units which are operated and constructed in
the mid-90’s. Of those units, if some units are outaged suddenly by these
mechanical problems, it is expected that a ripple effect on the electric power
system has a severe shock.
255
Testing Setup
Air Puffing Mode
Vacuum Drying Mode
Leak Test
Vacuum Decay Test

No
Yes
Pressure Drop Test

No
Yes
Helium Tracer Test

No
Yes
Water Absorption Test

No
Yes
Vibration Test
No
Yes
Repair &
Testing End
Replacement
Figure 2. Mechanical integrity evaluation test for generator stator windings.
Figure 3. 1991 through 1993 testing summary.
256
Figure 4. Cooling water leakage accident due to operation time
In this paper, the mechanical integrity evaluation methods are described.

Especially, it is shown that the water absorption test is most sensitive of them.
But the capacitance value has an order pico-Farad(pF). For this reason, it is easy
to make a serious mistake, for example, if you simply decide the wet bar by
comparing the amplitude of 5 pF with 8 pF algebraically. Accordingly, it is
needed to raise the reliability of wet bar diagnosis by developing several
stochastic diagnosis methods.
2. Water absorption test

The principle of the water absorption test for stator bars is that it is to measure
the capacitance between sensor(+) and copper(-) to press the sensor to the mica
surface. A condenser is used to store the electricity. If voltage is applied to the
two parallel metal plates, an amount of electricity is stored in this plate, which is
called capacitance. If the electric charges are charged in the storage battery as
shown in Figure 5, the charges are distributed within the plates. The electric flux
is perpendicular to the plates and a uniform magnetic field is formed to a parallel
shape. The capacitance of the parallel plate storage battery is equal to equation
(1) where C is capacitance, ε r is permittivity of a dielectric substance, ε 0 is
permittivity constant in vacuum condition, A is area of sensor and d is distance
between copper and sensor.
A
C =ε 0 εr [F] (1)
d
257
Capacitance can be increased in generator stator windings in only two cases.

(1) When the area of sensor(A) is wide or the distance of poles(d) is short,
(2) When a material which permittivity is bigger between a sensor and copper
is added.
In case (2), a new material for which permittivity is big cannot be added. Also,
the area of sensor and the distance between the poles are constant. Accordingly,
the capacitance can be increased if only the water which is leaked inside bars
exists. With a reference, the permittivity of water is 20 times the permittivity of
mica which composes the bar. The water absorption test is to use the above
mentioned physical phenomena. In evaluating whether wet bars are present or
not, the stochastic diagnosis methods are used with capacitance of windings. It
should not be determined through quantitative comparison of the capacitance of
bars simply because of the small quantity in the water absorption test. Therefore,
as long as the comparable data are numerous, the judgment errors for wet bar
can be heightened.
Figure 5. Capacitance for parallel plate storage battery.
Figure 6 shows the structure of stator end winding. The measurable positions are
the end winding parts of the turbine end side and collector end side. Among
them, until now, they could only be measured in the position which is bent from
the core part. The capacitance cannot be measured as other zones are covered by
insulation paint or is tied by binders.
If all the windings are absorbed uniformly due to leak of coolant water, it will
be diagnosed that they are normal in the stochastic methods. But, it is going to
be a long shot. Accordingly, in this paper, these stochastic diagnostic methods
are applied under the assumption that the water absorption of the water-cooled
stator windings is generated under manufacture defects or continuous operation.
258
For the water absorption test, the rotor must be removed and the surface of
bars is cleaned, and then 3-phases are grounded. Finally, the capacitance of
winding can be measured. The upper, inner, and outer sides can be measured in
the top bar, and the inner and outer sides can be measured in the bottom bar as
shown in Figure 6.
Upper
Measurable points
Outer
Inner
Figure 6. Only measurable points for capacitance measurement & section of winding.
3. Wet bar diagnosis
3.1 Case I
This generator is operated over more 15 years and is 500 MW class generator.
Figure 7 shows the water absorption test results which are measured in 2006,
2007, and 2008 year respectively where CET-TOP is an upper part of top bar of
collector end side and CET-OUT is an outer part of top bar of collector end side.
2006 and 2007 data are measured before sudden outage and 2008 is measured
after sudden accident. It can be known that phase voltage is 12.7 kV and 10.8 kV
is applied in No. 23 winding and the capacitance of No. 23 winding of stator has
been increased gradually. Figure 8 shows the stochastic wet bar diagnosis results
in 2007, and Figure 9 shows the stochastic wet bar diagnosis results in 2008 year
respectively. It is known that abnormal data don’t exist in 2006 year but the
capacitance of 23 winding of stator is higher than other windings in 2007 and
2008 year. Especially, it is increased extraordinarily in 2008 year. That is to say,
it is beyond the 3 sigma limit in 2007 year. But after insulation breakdown
accident, it approaches 5 sigma limit in 2008 year. In stochastic wet bar
diagnosis, if it is more than 3 sigma limit, it means suspected wet bar and if it is
more than 5 sigma limit, it means wet bar.
259
CET-TOP
22
2006
20 2007
2008
Capacitance (pF)
18
16
14
12
10
0 5 10 15 20 25 30 35 40
Stator winding No.
(a) CET-TOP
CET-OUT
26 2006
2007
24 2008
22
Capacitance (pF)
20
18
16
14
12
0 5 10 15 20 25 30 35 40
Stator winding No.
(b) CET-OUT
Figure 7. Capacitance trend change due to operation time.
260
(a) CET-TOP
(b) CET-OUT
Figure 8. Wet bar diagnosis results of 2007 year.
261
(a) CET-TOP
(b) CET-OUT
Figure 9. Wet bar diagnosis results of 2008 year.
262
3.2 Case II
This generator is also operated over 25 years and is 500 MW class generator.
Figure 10 shows the water absorption test results which are measured in 2006
and 2009 year respectively where TET-TOP is an upper part of top bar of
turbine end side and TET-OUT is an outer part of top bar of turbine end side.
Actually, capacitance of No. 2 winding of collector end side was higher than No.
2 winding of turbine end side in 2006. In those days, spare bars were absent so
winding was not replaced. As a stopgap measure, they were repaired by using
vacuum pressure impregnation method. In spite of that, as shown in Figure 10, it
can be confirmed that capacitance of No. 2 winding of turbine end side in 2009
year has been increased in comparison with capacitance of 2006 year. From the
above results, it can be predicted that cooling water was leaked in No. 2 winding
of turbine end side which was repaired in 2006 year. Figure 11 shows the
stochastic wet bar diagnosis results for TET-TOP and TET-OUT in 2006 year
respectively. Also, Figure 12 shows the stochastic wet bar diagnosis results for
TET-TOP and TET-OUT in 2009 year respectively. In the stochastic diagnosis,
capacitance of No. 2 winding of turbine end side isn’t higher than 3 sigma limit.
But it is confirmed that capacitance of No. 2 winding was increased remarkably
irrespective of other bars in 2009 year. Unlike No. 23 winding as mentioned
before, equal phase voltage are applied but as sharing voltage of No. 2 winding
is 2.7 kV, it is still lower than No. 23 winding. And sharing voltage of No. 40
winding is 4.5 kV.
Figure 13 shows the precise wet bar diagnosis test. Generally, cooling water is
leaked in clip-to-strand which water is supplied. Accordingly, water leaks begin
from clip-to-strand, that is to say, as shown in Figure 13(a), No. of 26, 27, 28,
29, 30 are start points of water leaks. Therefore, if No. 2 and No. 40 windings
are wet bars, capacitances of No. 3, 8, 13, 18, 23, 28 of Figure 13(b) and 13(c)
must be heightened in comparison with other numbers. To my surprise, it can be
known that the above results agree to our thought well. After all, in order to
confirm if they are wet bars, insulation materials are removed and bubble test is
carried out as shown in Figure 14. It can be confirmed that No. 2 and No. 40
windings of turbine end side must be wet bars from Figure 14. As spare bars are
not prepared until now, vacuum pressure impregnation method is applied for
water absorption suppression. Because this generator has been operated for
about 30 years, spare bars will be prepared or going to be rewound soon.
263
TET_TOP
16
2009.12
2006.03
15
14
Capacitance(pF)
13
12
11
10
0 5 10 15 20 25 30 35 40
Winding No.
(a) TET-TOP
TET_OUT
17
2009.12
16 2006.03
15
Capacitance(pF)
14
13
12
11
10
9
0 5 10 15 20 25 30 35 40
Winding No.
(b) TET-OUT
Figure 10. Capacitance trend change due to operation time.
264
(a) TET-TOP
(b) TET-OUT
Figure 11. Wet bar diagnosis results for turbine end side in 2006 year.
265
(a) TET-TOP
(b) TET-OUT
Figure 12. Wet bar diagnosis results for turbine end side in 2009 year.
266
Core
1 6
Bar #42 11
2 7 16
Bar #1
3 8 12 21
Bar #2 13 17
22 26 End-Winding
4 9 18
Bar #3 14
5 10 23 27
Bar #4 19
15 24 28
20
25 29
30
(a) capacitance measurement position for precise diagnosis
TE side - #42-1-2-3-4
9
28
8 23
3 13 18
Capacitance (pF)
7
8
4
0 5 10 15 20 25 30
Measurement position
(b) wet bar diagnosis results for No. 42, 1, 2, 3 and 4
TE side - #38-39-40-41-42
10
9 18 23
13 28
3
8
Capacitance (pF)
4
0 5 10 15 20 25 30
Measurement position
(c) wet bar diagnosis results for No. 38, 39, 40, 41 and 42
Figure 13. Precise capacitance measurement and wet bar diagnosis.
267
Figure 14. Bubble test for leak point detection.
4. Conclusions
During overhaul period, leak test and vibration characteristic test are carried out
for generator stator windings. In addition, water absorption test which was
developed independently by KEPRI is carried out periodically.
Especially, winding repair due to vibration characteristic change and leak test
result is not a great difficult work. But the countermeasures due to wet bar
diagnosis have no choice but to replace winding unfortunately. In addition, it is
so difficult to determine replacement of winding through the stochastic wet bar
diagnosis too. That is because capacitance of bar insulation has very small
value[pF] and there may be errors of stochastic diagnosis. Accordingly, precise
water absorption test and more reliable stochastic wet bar diagnosis are needed.
In this paper, 2 cases for wet bar diagnosis are described. It is verified that all
the diagnosis indicate wet bars well. From this paper, the importance of periodic
wet bar diagnosis can be confirmed.
268
References
1. Stanton, D. J. Diagnosing and repairing water leaks in stator windings. GE company review.
Schenectady, New York, 1992.
2. Inoue, Y. et al. Technology for detecting wet bars in water-cooled stator winding of
turbine generators. IEEE IEMDC, June 2003, pp. 1337–1343.
3. Kim, H. S. & Bae, Y. C. Wet bar detection by using water absorption detector. J. of
Mechanical Science and Technology, 2008. Vol. 22, pp. 1163–1173.
4. Kim, H. S. & Bae, Y. C. An experimental study on the dryness judgement criterion for
water-cooled generator stator windings. Trans. ASME-B, 2010. Vol. 34, No. 2,
pp. 137–143.
5. Kim H. S. et al. A study on the water absorption diagnosis method through capacitance
measurement for generator stator windings. 2006’s KSPE Spring Annual
Meeting, pp. 285–286.
269
A breakthrough in ultrasonic detection and sizing of partially closed cracks?
A breakthrough in ultrasonic detection and

sizing of partially closed cracks?
Stefan Sandlin
Espoo, Finland
Abstract
Ultrasonic testing is one of the most efficient methods available for finding and
sizing of cracks in structural components such as for example nuclear reactor
circuits. However, some fatigue or stress corrosion cracks (SCC) may be closed
or partially closed due to the presence of residual stresses pressing the crack
surfaces against each other. In these cases the ultrasound may pass through the
closed crack almost without any reflection or scattering. Water or oxide in the
crack may also influence the transparency of the crack. This means that the
crack may be invisible to ultrasound and, hence, the crack may not be found at
all or its size may be seriously underestimated. This may lead to catastrophic
failure of the component. Tohoku University in Japan has developed a method in
which cracks are monitored both at the insonifying frequency f and at the
subharmonic frequency f/2 using a phased array for focused reception of
ultrasound and a LiNbO3 single-crystal transmitter for generation of intense
ultrasound. The subharmonic ultrasound is generated only at the closed parts of
the crack. The method therefore combines normal phased array testing with
phased array testing at subharmonic frequency. A commercial version of the
equipment for research is under development by a Japanese company. This
imaging method is called subharmonic phased array for crack evaluation
(SPACE). Different nonlinear ultrasonic methods for evaluation of closed cracks
have been proposed for decades (based on superharmonics), however, SPACE
seems to include several new innovative ideas and it seems to have the best
potential for becoming a standard method for inspection of nuclear reactor
270
circuits for closed cracks. Subharmonics have a better signal to noise ratio than
superharmonics because subharmonics are generated only at closed cracks while
superharmonics are generated also in transducers, liquid couplers and
electronics. We therefore mainly restrict ourselves to the description of the
subharmonic SPACE method in this work. A more traditional sizing method
using phased arrays is, however, briefly presented as reference.
1. Introduction
Ultrasonic testing is one of the most efficient methods available for finding and
sizing of cracks in structural components such as for example nuclear reactor
circuits. However, some fatigue or stress corrosion cracks (SCC) may be closed
or partially closed due to the presence of residual stresses pressing the crack
surfaces against each other. Water or oxide in the crack may also influence the
transparency of the crack. In these cases the ultrasound may pass through the
closed crack almost without any reflection or scattering. Defects in diffusion
joining and friction stir welding may also exhibit poor ultrasonic reflection or
scattering properties. This means that the crack or joint defect may be invisible
to ultrasound and, hence, the crack may not be found at all or its size may be
seriously underestimated. This may lead to catastrophic failure of the component.
Traditional ultrasonic systems operate in linear mode; i.e. the received
scattered ultrasound from defects lies in the same frequency range as the
insonifying ultrasound. The amplitudes of waves diffracted or reflected at a
crack tip or a crack surface are determined by the acoustic impedance mismatch
between the material and the air gap. For tight cracks the acoustic impedance
mismatch may be negligible. In nonlinear testing, defects behave as active
radiation sources of new frequency components rather than as passive scatterers
in conventional ultrasonic inspection (Solodov & Busse, 2007). The nonlinear
techniques are therefore sensitive to closed fatigue or SCC cracks. In non-classic
nonlinearity cracks may be observed by acoustically activated contact clapping
in cracks. The nonlinear modulation frequencies caused by clapping of the crack
interfaces often unveil flaws with higher contrast and more reliably than higher
harmonics (Pfleiderer, 2006). A minimum amplitude of the insonifying
ultrasound is needed for generation of subharmonics; this is illustrated by
Yamanaka et al. (2004) by saying “if we find subharmonic waves when
increasing the input wave amplitude, there is a partly closed crack, hidden in the
object”.
271
Tohoku University in Japan has developed a method in which cracks are

monitored both at the insonifying (or fundamental) frequency f and at the
subharmonic frequency f/2 using a phased array for focused reception of
ultrasound and a LiNbO3 single-crystal transmitter for generation of intense
ultrasound. The method therefore combines normal phased array testing with
phased array testing at subharmonic frequency (Ohara et al. 2007). A
commercial version of the equipment for research is under development by a
Japanese company (Yamanaka, 2009). This imaging method is called
subharmonic phased array for crack evaluation (SPACE). Different nonlinear
ultrasonic methods for evaluation of closed cracks have been proposed for
decades (based on superharmonics), however, SPACE seems to include several
new innovative ideas and it seems to have the best potential for becoming a
standard method for inspection of nuclear reactor circuits for closed cracks.
Subharmonics have a better signal-to-noise ratio than superharmonics because
subharmonics are generated only at closed cracks while superharmonics are
generated also in transducers, liquid couplers and electronics. Further, the
attenuation of the subharmonic waves is much lower than that of the
fundamental wave and the superharmonic waves. At best the amplitude of the
second harmonic wave is of the order of a few percent of the fundamental one
(Kawashima et al. 2006). We therefore mainly restrict ourselves to the
description of the subharmonic SPACE method in this work. The method is
illustrated by imaging of closed fatigue cracks in an aluminium specimen
(A7075) and an austenitic stainless steel specimen (SUS316L). Further, the
method is also illustrated by imaging of SCC cracks in two other stainless steel
specimens (SUS304). According to Ohara et al. (2008) SPACE can measure
crack depths with an accuracy of approximately 1 mm. As will be seen, the
combination of subharmonic wave detection and phased array technique seems
to lead to substantial improvement beyond the present state-of-the-art in sizing
and detection of closed cracks. The Japanese scientists have also tested a phased
array based on laser interferometry; unfortunately the paper dealing with this
topic is mainly in Japanese language. On the application side, localization and
sizing of closed or partially closed fatigue and SCC crack in nuclear power plant
components was already mentioned. Another important application could be
detection of flaws in dissimilar metal joints. These joints can be made by for
example diffusion bonding, explosion welding, clad welding, brazing etc. A flaw
at the interface between two different metals may be difficult to observe using
ultrasonic techniques as there is a strong reflection also from a perfect interface.
272
In nonlinear ultrasonic technique this kind of flaws become active radiation

sources of subharmonic ultrasound and the SPACE method is therefore expected
to be effective in quality evaluation of these kinds of joints.
2. Goal
The goal of this work was to summarize the published work on the use of
subharmonic ultrasound for detecting and sizing of closed cracks or other tight
defects. The work is intended to form the basis for a possible further detailed
experimental and theoretical evaluation of the technique.
3. Fundamental principles
Detection of ultrasonic diffraction from the tip of a partially closed crack has
been the “traditional” technique for sizing of this kind of cracks. The diffraction
is illustrated in Figure 1. Parts of the crack which are in contact will also cause
diffraction and these signals have been detected by special phased array
transducers. However, the detection of the crack tip diffraction is often difficult
Figure 1. The generation of diffracted waves at crack tips (Gilmour et al. 2009).
(Dupond et al. 2004). The use of subharmonic ultrasound offers a mean for
further improvement of the sizing of closed cracks. In this chapter the basic
principles of detecting and sizing of closed cracks using subharmonic phased
array technique are outlined. The properties of subharmonic ultrasound is the
main topic of this chapter, but as the phased array technique plays an important
273
role in the measurement technique using subharmonic ultrasound we also briefly

present the phased array technique.
3.1 Nonlinear ultrasound
Nonlinear ultrasound has been expected to become a primary technique of

detecting and evaluating closed cracks since the publishing of the pioneering
work by Buck, Morris and Richardson in 1978. Nonlinear ultrasound is based on
the detection of superharmonic waves (2f, 3f,…), subharmonic waves (f/2,
f/3,…), ultra-subharmonic waves, frequency pairs etc. generated by the
interaction of large amplitude ultrasound with closed cracks, where f is the
central frequency of the insonifying wave. The theory of nonlinear ultrasound is
quite complicated; therefore only a brief phenomenological presentation will be
given below. The reader who is interested in a deeper and more theoretical
description of nonlinear ultrasound in non-destructive evaluation is referred to
for example the doctoral thesis by Pfleiderer (2006). The thesis is in German
language and written at University of Stuttgart.
3.2 Subharmonic ultrasound in non-destructive

evaluation
Although subharmonic waves have better selectivity for closed cracks than
superharmonic waves, the measurement of closed crack depths in the thickness
direction has not been realized before the SPACE-system (Subharmonic Phased
Array for Crack Evaluation) was developed at Tohoku University in Japan
(Ohara et al. 2008). Further, the signal-to-noise for superharmonics is not very
high because these waves are also generated in for example in piezoelectric
transducers and liquid media causing background noise. Subharmonics, on the
other hand, are generated only at closed crack surfaces. The intensity of
subharmonic waves is higher than that of superharmonic waves (Yamanaka et al.
2004). Subharmonic waves are also less attenuated as they travel through the
material. Because of these reasons we only concentrate on the subharmonic
technique in the following. For the following brief and simplified description we
assume that ωI is the frequency of the intense insonifying ultrasound
(fundamental frequency), ωR is the resonance frequency of the crack and ω is the
frequency of the subharmonic ultrasound. According to Ohara et al. (2008) the
subharmonic wave is generated through a resonance phenomenon, called
274
subharmonic resonance (SR), between the fundamental wave ωI and Rayleigh

waves travelling along the crack surfaces as illustrated in Figure 2. The
generated subharmonic wave radiates out mainly through the crack tips. The
amplitude of the fundamental wave needs to reach a minimum value at the
closed crack before subharmonic waves can be generated. Figure 3 illustrates
that there is an optimal relation between the input frequency ωI and the
resonance frequency of the crack ωR, i.e. when ωR = ωI/2 the displacement
amplitude of the generated subharmonic wave will have a maximum (Figure 3
a). If this condition is not fulfilled the displacement amplitude will drop (Figure
3 b). However, according to Yamanaka (2006) the resonance peak is broad
enough to facilitate practical inspections using subharmonic ultrasound. Cracks
of different lengths can therefore be evaluated using a single input frequency.
Nevertheless, a variable frequency transmitter is highly desired to cover a wide
range of crack lengths. The SR is a resonance phenomenon in the entire crack.
However, SPACE images indicate that subharmonic waves are generated not
only at the crack tip but also at other parts of the crack. To explain this feature
Ohara et al. (2008) propose a new concept of localized subharmonic resonances
(LSRs). The SPACE image is therefore built up of several LSR sources along
the crack as schematically shown in Figure 4. Ohara et al. further conclude that
LSR will take place with large probability at the edge of a long, closed crack.
This is a practical advantage of SPACE.
Figure 2. The input wave periodically opens the closed crack and subharmonic waves are
generated by subharmonic resonance (Ohara et al. 2008).
275
Figure 3. Generation of subharmonic ultrasound at a crack. The displacement amplitude

of the subharmonic wave will have a maximum when the input frequency ωI divided by
two is equal to the resonance frequency ωR of the crack as in (a). In (b) the input
frequency is larger and the generation of subharmonics is less efficient (Ohara et al.
2008).
Figure 4. Localized subharmonic resonances (LSRs) will produce subharmonic waves at

the closed parts of the crack (Ohara et al. 2008).
276
3.3 Phased array
Phased array probes use an array of probe elements, all individually wired,
pulsed and time-shifted on both pulsing and receiving. The elements are usually
pulsed are groups of 8 to 32 elements during an inspection. Using this technique
it is possible to scan without moving the probe, to change the probe angle and to
change the focus. In Figure 5 these techniques are called electronic scan (E-
scan), sectorial scan (S-scan) and depth focusing (dynamic depth focusing).
Combinations of these are also possible (Davis & Moles, 2006). The use of time-
delays mimics for example wedges and lenses in traditional ultrasonic testing.
Figure 5. The principles of an ultrasonic phased array probe. Scanning, probe angle
sweep and choice of focal depth can be done by selecting individual time-delays for the
elements as shown in the upper part of the picture (Davis & Moles, 2006).
Dupond et al. (2004) in France have used a special large aperture phased array
transducer (diameter 100 mm and 121 elements) for partially closed crack
detection based of diffraction at the crack tip or other diffracting points along the
crack. They used 45° shear waves at a frequency of 4.5 MHz. The ultrasonic
beam is swept over the crack from the back wall to the tip by scanning and by
changing the focal length of the transducer as shown in Figure 14. Dupond et al.
did not utilize subharmonic waves. A short comparison of this technique with
the SPACE method will be given in the discussion.
277
3.4 The SPACE equipment
A schematic view of the SPACE equipment is seen in Figure 6. Large amplitude

ultrasound is generated by a LiNbO3 single-crystal transmitter and the scattered
fundamental and subharmonic waves are received by a phased array using a
digital filtering technique. A sectorial scan is also shown in Figure 6 (Yamanaka
et al. 2006). Figure 7 is from a brochure by Tohoku University and illustrates the
inspection principle of a welded tubular part.
Figure 6. A schematic view of phased array method for imaging of a closed crack tip
using subharmonic ultrasound. A sectorial scan of the crack area can be seen to the right
(Yamanaka et al. 2006).
278
Figure 7. The array sensor and the LiNbO3 single crystal transmitter. To the right a
schematic picture showing SPACE inspection of a welded part (the pictures are from a
brochure by Tohoku University).
4. Detection and sizing results

Ohara et al. (2007) used an aluminium specimen with a fatigue crack and two
stainless steel specimens, one with a fatigue crack and the other with a SCC
crack to evaluate performance of the SPACE equipment. In the fatigue crack
specimens the crack opening and closure was controlled by three point bending
as shown in Figure 8. All specimens had approximately the same dimensions. A
7 MHz three-cycle burst was used as input signal from the LiNbO3 transmitter.
The transmitter voltage was set to 139 V (peak-to-peak) for the aluminium
specimen and to 262 V (peak-to-peak) for the stainless steel SUS316 specimens.
This corresponds to wave amplitudes at the crack of 17 nm and 7 nm
respectively. These values were selected to be above the subharmonic generation
threshold. The corresponding voltage for the stainless steel specimens SUS304
was 220 V (peak-to-peak) giving a displacement amplitude of 10.3 nm (peak-to-
peak) at the crack. In another article Ohara et al. (2009) give a brief description
of the crack sizing process.
279
Figure 8. Three point bending was used to control the opening and closure of fatigue
cracks. The dimensions are in mm (Yamanaka, 2007).
4.1 Imaging of a fatigue crack in aluminium (A7075)
Figure 9 shows crack sizing results for an aluminium specimen. At fundamental

frequency the crack tip was observed only when it was (primarily) open (a). On
the other hand, at subharmonic frequency the crack tip is clearly observed when
the crack tip is closed (e). In case (b) when the crack is primarily open a very
weak subharmonic signal can still be seen just below the crack tip, this seems to
indicate that some small parts of the crack are still closed and thus favouring
some generation of subharmonic waves. The actual location of the crack tip is
illustrated by photographs (c and f) in Figure 9.
280
Figure 9. Phased array images of a crack tip in an aluminium specimen (A7075). The
crack tip is open in a) and b) while it is closed in d) and e). Images a) and d) are taken at
fundamental frequency while b) and e) are taken at subharmonic frequency. The closed
crack tip cannot be observed at fundamental frequency d) but it is clearly observed at
subharmonic frequency e) (Ohara et al. 2007).
4.2 Imaging of a fatigue crack in stainless steel

(SUS316L)
In Figure 10 the sectorial scans (a, c and e) are taken at fundamental frequency,
while b, d and f are taken at subharmonic frequency. The crack opening increase
from almost closed for the topmost scans to almost open for the bottom scans,
due to the increase of bending stress from 19 MPa to 84 MPa and finally to
112 MPa.
Some details from Figure 10:
In pictures (a) and (b) the bending stress was 19 MPa.
(a) Sectorial scan at fundamental frequency f. Only the notch is clearly visible,
this indicates that the crack is mainly closed
281
(b) Sectorial scan at subharmonic frequency f/2. Both the crack tip and the
middle part are visible indicating that the subharmonic wave can image
closed cracks.
In pictures (c) and (d) the bending stress was 84 MPa.
(c) Sectorial scan at fundamental frequency f. The middle parts become visible
indicating that the crack is more open and that the middle part is the
boundary between the open lower part and the closed upper part.
(d) Sectorial scan at subharmonic frequency f/2. Both the crack tip and the
middle parts are still visible although with decreased intensity. This indicates
that these parts are still partially closed.
In pictures (e) and (f) the bending stress was 112 MPa.
(e) Sectorial scan at fundamental frequency f. No part of the crack is visible
indicating that the crack tip is still closed. It is surprising that there is no
diffracted signal from the border between the open part and the closed tip. A
possible explanation might be that this border is zigzaging and thereby
giving very weak scattering of the fundamental wave in the direction to the
phased array. As the experimental setup is that of Figure 8 the incident
fundamental wave is reflected to the left from the open crack and nothing
will reach the phased array.
(f) Sectorial scan at subharmonic frequency f/2. The middle part has
disappeared (i.e. this part is open), but the tip is visible indicating the
uppermost part of the crack is still closed and generates subharmonic waves.
282
Figure 10. Images of a fatigue crack in stainless steel (SUS316L). Fundamental and sub-
harmonic images (a, b) at nominal bending stress of 19 MPa. Corresponding images (c,
d) and (e, f) for nominal bending stresses of 84 MPa and 112 MPa respectively (Ohara et
al. 2008).
4.3 Imaging of SCC-cracks in stainless steel (SUS304),

specimen 1
For the two stainless steel specimens SUS304 the transmitter voltage was set to
220 V (peak-to-peak) giving a displacement amplitude of 10.3 nm (peak-to-
peak) at the crack. The central frequency of the fundamental wave was 7 MHz.
In the specimen in Figure 11 (c) branched SCC cracks have grown from the
initial fatigue crack. In the fundamental image (a), branched SCCs, B and C are
observed on both sides of the tip of the fatigue crack. This suggests that B and C
were primary open. Another SCC at A was observed above the tip of the fatigue
crack in the subharmonic image, although A was slightly visible also in the
fundamental image, this shows that A was primarily closed. The crack depth was
measured to be 13.8 mm from Figure 11. According to metallurgical
measurements on etched cross-sections the deepest SCC was 13 mm.
283
Figure 11. Fundamental (a) and subharmonic (b) images of stainless steel specimen 1
containing SCC cracks. (c) a picture of the etched cross-section (Ohara et al. 2008).
4.4 Imaging of SCC-cracks in stainless steel (SUS304),

specimen 2
The measurement settings for this specimen were the same as for specimen 1
above. In Figure 12 the tip of the branched SCC labelled A was observed in the
fundamental image (a) suggesting that A was open. On the other hand the root of
the SCC labelled A and two other branched SCCs labelled B and C were
observed in the subharmonic image (b). This shows that these parts were
primarily closed. The crack depth was measured to be 13.5 mm from the
fundamental image and 15.2 mm from the subharmonic image. Metallurgical
measurement from etched cross-sections showed that the deepest part of the
crack was 16 mm.
284
Figure 12. Fundamental (a) and subharmonic (b) images of stainless steel specimen 2
containing SCC cracks. (c) a picture of the etched cross-section (Ohara et al. 2008).
4.5 The measurement accuracy
Ohara et al. (2008) examined the measurement accuracy of SPACE by comparing

the true and measured depths of closed fatigue and stress corrosion cracks as
shown in Figure 13. They demonstrated that the subharmonic images indicated
the correct crack depth with an accuracy of approximately 1 mm. Particularly in
the fatigue cracks, the subharmonic images significantly reduced the
measurement error compared to the fundamental images. The subharmonic and
fundamental images indicated almost the same crack depth for stress corrosion
cracks in specimen 1, and the reduction in measurement error was less than 2
mm in specimen 2.
285
Figure 13. Comparison of measured and true crack depths for fatigue and SCC cracks.
The black markers refer to results obtained using subharmonic frequency f/2. In two
cases the crack tip was not observed at all (measured crack depth zero) at fundamental
frequency f (Ohara et al. 2008).
4.6 Sizing results using more conventional methods
In this section we briefly present a more traditional closed crack evaluation done
in France (Dupond et al. 2004) as reference for the above SPACE results. The
French used pulse-echo technique in immersion mode as illustrated in Figure 14.
The specimen was stainless steel (304L). They used a standard focused
transducer (SFT) and a 121 element phased array (PA); the main transducer
characteristics are summarized in Table 1. The closure stress applied (by four
point bending) to the crack caused an underestimation of the crack height for
both transducers as shown in Figure 15. For the standard focused transducer the
error is about 4 mm while the error for the phased array is 1.5–2.0 mm.
286
Figure 14. The principle of inspection according to Dupond et al. (2004). The crack
closure was controlled by four point bending.
Table 1. The main transducer characteristics (Dupond et al. 2004).
Satyanarayan et al. (2007) tested the performance of a 64 element phased array

at 5 MHz on mild-steel pipe with wall thickness of 10 mm and three
circumferential notches of depths 3-, 5-, and 7 mm. The notches simulated inner
surface breaking cracks; no realistic cracks were used in this work. They used
the phased array to mimic conventional ultrasonic angular transducers with
different angles (i.e. they used the phased array with pre-set angles of 30, 35, 40,
45, 50, 55 and 60 degrees) and some of their experimental sizing results are
summarized in Tab. 2 together with the percent errors. The percent errors are
about 3% for the 7 mm notch and they will of course increase for smaller
notches reaching over 13% for the 3 mm notch at transducer angles 55 and 60
degrees (these two are not shown in Table 2).
287
Figure 15. Destructively (actual profile) and ultrasonically obtained crack height for a
partially closed crack. The large aperture phased array (PA) gives a better result than the
standard focused transducer (Dupond et al. 2004).
Table 2. Sizing errors using different transducer angles on a mild-steel pipe with a wall
thickness of 10 mm. Notches of depths 3-, 5-, and 7 mm mimic inner surface breaking
cracks (Satyanarayan et al. 2007).
288
5. Summary and discussion

In this report we concentrated mainly on the use of subharmonic ultrasound for
detecting and sizing of closed or partially closed cracks (such as stress corrosion
cracks and fatigue cracks) in nuclear power plant components or in other critical
engineering structures. The subharmonic NDE-technique has been developed at
Tohoku University (Japan) by professor Yamanaka and his group during the last
years and it is called SPACE (Subharmonic Phased Array for Crack Evaluation).
The technique is based on generating an intense ultrasonic wave with central
frequency f using a LiNbO3 single-crystal transmitter. When this insonifying (or
fundamental) wave interacts with a closed or partially closed crack scattered
waves of central frequencies f/2 (subharmonic) and f (fundamental) are
generated. These waves are detected using a phased array together with a
filtering technique. Other kinds of waves, such as superharmonics, are also
generated, but these are generally of low intensity and tend to disappear in the
noise. The intensity of the subharmonic wave may be comparable to that of the
fundamental wave, further; the subharmonic wave is less attenuated compared
with the fundamental wave and superharmonics because of the lower frequency.
According to the articles cited in this report subharmonic waves are only
generated at closed parts of a crack; they are therefore very useful in detecting
and sizing of these. For subharmonic generation the amplitude of the
fundamental wave should exceed a certain material dependent threshold and the
frequency should fit into a certain frequency range. However, according to the
cited articles, this frequency range is not very critical. Based on these reported
observations it seems that the Japanese scientists have developed a method that
substantially improves the detectability and sizing of closed cracks. In a
comparison between ordinary phased array and SPACE the main advantage of
the latter is that some closed crack tips which are not observed at fundamental
frequency are quite clearly observed at subharmonic frequency. According to the
Japanese researchers the equipment for making subharmonic measurements can
be integrated into ordinary phased array instrumentations. Further, the Japanese
researchers have also developed a scanning laser interferometric detection
system which can replace the piezoelectric phased array. The laser detection
technique will expand the frequency bandwidth of the system and make the
array configuration more flexible: The laser array will be especially useful in
imaging of cracks with complex shapes, such as stress corrosion cracks
(Yamamoto et al. 2008). Unfortunately this paper is mainly in Japanese
289
language. Professor Solodov at University of Stuttgart was also contacted during

this work to get his opinion about the potential of nonlinear ultrasonics in NDE
(see Chapter 6).
6. Conclusions and recommendations

Based on the studied articles about the SPACE method it seems that the
Japanese scientist have presented a method that makes an important progress
beyond the state-of-the-art in detecting and sizing of closed or partially closed
cracks. This regardless of the fact that the sizing process itself could have been
described more in detail and compared to the sizing process used in normal
phased array technique. The reasons for the progress beyond the state-of-the-art
are mainly:
1. Closed crack tips that cannot be observed at all at fundamental frequency are
observed at subharmonic frequency.
2. According to the articles cited in this work subharmonics are generated only
at closed cracks whereas superharmonics are generated in transducers, liquid
couplers, electronics etc. Therefore the subharmonic wave provides increased
selectivity in identification of closed cracks.
3. Sizing using the subharmonic wave gives crack depths closer to the
metallurgical measurements from etched cross-sections than sizing using the
fundamental wave.
4. By comparing phased array images produced at subharmonic and fundamental
frequency it is possible to identify closed and open parts of a crack.
5. The intensity of the subharmonic wave is much higher than that of the
superharmonic waves, thus giving a better signal-to-noise ratio.
6. The attenuation of the subharmonic wave is less than that of the fundamental
wave due to the lower frequency.
Because the SPACE method is new some problems may be encountered,
especially in future field applications. Here we try to identify some of these
potential problems:
1. Transfer of intense ultrasound to the inspected component and choice of the
needed amplitude.
290
2. Choice of the optimal fundamental frequency for optimal generation of

subharmonic waves.
3. The influence of oxide and/or water in the crack on the detection and sizing
performance.
4. The mechanism of subharmonic generation is not “crystal clear” yet
(Solodov, 2009).
5. The influence of possible unbroken ligaments on the ultrasonic signal.
6. Spurious indications due to leakage of the fundamental wave into the
measured subharmonic signal.
Besides these concerns regarding practical applications we conclude that the
performance of SPACE in sizing closed cracks (fatigue and SCC) in stainless
steel and aluminium specimens has been demonstrated. Detecting tight flaws in
electron beam welds, friction stir welds, diffusion bonds, explosion joints etc
might also be interesting applications. Detection of flaws in dissimilar metal
joints may be added to the list. University of Stuttgart also has quite an
experience in working with nonlinear ultrasonics and with subharmonics (and
ultra-subharmonics) in particular (Solodov, 2009). Solodov recommends further
investigations of this emerging field of ultrasonic NDE. According to Solodov
the ultrasonic sensitivity to small tight cracks may be further increased by
including ultra-subharmonics into the analysis. Because of this and the obvious
potential of the SPACE method it is recommended that the real capability of this
method is experimentally evaluated on specimens with suitable tight cracks.
Information about specimens with realistic artificial flaws can be found in for
example the doctoral thesis by Kemppainen (2006).
Acknowledgements
This work was done within the project “Monitoring of the structural integrity of
reactor circuit” (RAKEMON), which is a part of the program SAFIR2010,
Finnish national research program on NPP safety 2007−2010. Some helpful
e-mail discussions with Professor Kazushi Yamanaka (Tohoku University) and
Professor Igor Solodov (Stuttgart University) are gratefully acknowledged.
291
References
Buck, O., Morris, W. L. & Richardson, J. M. 1978. Acoustic harmonic generation at
unbonded and fatigue cracks. Appl. Phys. Lett. Vol. 33, No. 5, pp. 371–373.
Davis, J. M. & Moles, M. 2006. Resolving capabilities of phased array sectorial scans
(S-cans) on diffracted tip signals. Insight, Vol. 48, No. 4. 7 p.
Dupond, O., Bredif, P., Poidevin, C. & De Nathasn, N. 2004. Advanced phased array
transducer for detection of closed crack tip diffraction. Internet Adress, 2009-08-
31: http://www-civa.cea.fr/home/liblocal/docs/PubliOff/Dupond_etal_ICNDE_
2004.pdf.
Gilmour, T. H. (Ed.) 2005. In-service non-destructive estimation of the remaining fatigue

life of welded joints. Internet Address 2009-08-31: http://www.shipstructure.org/
pdf/444.pdf.
Kawashima, K., Murase, M., Yamada, R., Matsushima, M., Uematsu, M. & Fujita, F.
2006. Nonlinear Ultrasonic Imaging of Imperfectly Bonded Interfaces. Ultrasonics,
44. Pp. 1329–1333.
Kemppainen, M. 2006. Realistic artificial flaws for NDE qualification – a novel manu-
facturing method based on thermal fatigue. Helsinki University of Technology,
Laboratory of Engineering Materials. 93 p. Internet Address 2009-11-27:
http://lib.tkk.fi/Diss/2006/isbn9512282631.
Ohara, Y., Mihara, T., Sasaki, R., Ogata, T., Yamamoto, S., Kishimoto, Y. & Yamanaka, K.
2007. Imaging of closed cracks using nonlinear response of elastic waves at
subharmonic frequency. Appl. Phys. Lett. 90.
Ohara, Y., Yamamoto, S., Mihara, T. & Yamanaka, K. 2008. Ultrasonic evaluation of
closed cracks using subharmonic phased array. Jpn. J. Appl. Phys., Vol. 47.
No. 5. Pp. 3908–3915.
Ohara, Y., Endo, H., Mihara, T. & Yamanaka, K. 2009. Ultrasonic measurement of closed
stress corrosion crack depth using subharmonic phased array. Jpn. J. Appl.
Phys., Vol 48. 6 p.
292
Pfleiderer, K. Frequenzkonversion aufgrund nichtlinearer akustischer Phänomene: Grundlagen

und Anwendung zur defektselektiven zerstörungsfreien Prüfung. Institut für
Kunststoffprüfung und Kunststoffkunde der Universität Stuttgart. 120 p.
Satyanarayan, L., Sridhar, C., Krishnamurthy, C. V. & Balasubramaniam, K. 2007.

Simulation of ultrasonic phased array technique for imaging and sizing of
defects using longitudinal waves. International Journal of Pressure Vessels and
Piping 84. Pp. 716–729.
Solodov, I. & Busse, G. 2007. Elastic wave nonlinearity for monitoring of localized damage in
engineering materials. 19th International Congress on Acoustics, Madrid, 2–7
September 2007. 6 p.
Solodov, I. 2009. Private communication.
Yamanaka, K., Mihara, T. & Tsuij, T. 2004. Evaluation of closed cracks by analysis of
subharmonic ultrasound. Insight, Vol. 46, No. 11, pp. 666–670.
Yamanaka, K., Sasaki, R., Ogata, T., Ohara, Y. & Mihara, T. 2006. Time domain analysis
of subharmonic ultrasound for practical crack sizing. Review of Quantitative non-
destructive Evaluation, Vol. 25. pp. 283–290.
Yamanaka, K. 2009. Private communication.
Yamamoto, S., Ohara, Y., Mihara, T. & Yamanaka, K. 2008. Application of laser
interferometer to subharmonic phased array for crack evaluation (SPACE).
Tohoku University. Pp. 198–203.
293
Development of nondestructive techniques for fuel cladding evaluation in hot cell
Development of nondestructive techniques

for fuel cladding evaluation in hot cell
Tarja Jäppinen and Stefan Sandlin

Espoo, Finland
Abstract
In nuclear light water reactors (LWR), an effective heat transfer from the fuel
through the fuel cladding to the water in the reactor is essential. Furthermore, the
cladding should prevent the radionuclides in the fuel from contaminating the
reactor circuit. During service the cladding suffers from damage by corrosion,
microstructural changes, hydrogen pick-up, creep etc.
Regular nondestructive evaluation in a hot cell is a relatively new complement
to laborious destructive evaluation of in-service degradation of irradiated
material. In this work the possibilities of laser ultrasonics and eddy current
techniques in hot cells are presented. The main emphasis is on oxide thickness
measurements and on detection of cracks and delaminations for oxide layers of
tubular or planar fuel cladding materials. The detectability of hydrogen
concentration in the cladding wall is also discussed.
1. Introduction
A brief review and applications of ultrasonics and eddy current techniques for
nondestructive evaluation (NDE) of materials degradation due to irradiation are
presented in this work. One important advantage of nondestructive examinations
is that the same fuel rod can be reinserted in the reactor for further irradiation
and re-examined again
There is a limited space available in a hot cell. Also the components to be
examined are radioactive. Therefore the nondestructive evaluation measurements
294
should be completely remotely controlled and the amount of material such as

sensors, scanners, cables inside the cell should be minimal.
Ultrasonic Rayleigh waves are widely applied for measuring the thickness of
oxide layers on metals. In the case of oxide on zirconium alloy fuel cladding, the
thicknesses of interest range from few micrometers to few tens of micrometers.
Oxide thicknesses can also be measured using the eddy current lift-off effect.
Radiation damage and hydrogen pick-up will change the density and the elastic
modulus of the alloy and therefore ultrasonics should have sensitivity to these
changes. Furthermore, the electric conductivity of the alloy changes due to
radiation damage and hydrogen pick-up. It should therefore be possible to
estimate the damage by eddy current measurements [1]. Edsinger et al. have
issued a patent on measurement of hydrogen concentration in zirconium alloys
in the fuel pool of nuclear power plants based on eddy current technique [2].
2. Fuel cladding damage

This section briefly considers those damaging processes that may be of interest
for hot cell NDE measurements or otherwise influence the performance of these
measurements. Figure 1 illustrates some damaging processes that occur in fuel
rods during service [3]. In a fuel rod cladding tube the wall thickness is usually
about 0.6 mm and the tube diameter is about 10 mm. The outer surface of the
cladding undergoes corrosion which changes the heat transfer properties and
reduces the wall thickness and leads to pick-up of hydrogen in the zirconium
alloy. It is generally said that thickness of the oxide should stay under 100 µm.
Hydrogen concentrations in excess of the solubility limit lead to precipitation
of hydrides which degrade the strength and ductility of the cladding. Further, as
the burnup increases there will be an increasing amount of pellet-cladding
mechanical interaction due to gap closure by fuel swelling, cladding creep down
(due to difference between internal gas and external coolant pressure) and
differential thermal expansion between pellets and cladding. The complete or
partial disappearance of the gap between the fuel pellet and the cladding will
have influence on especially the performance of ultrasonic guided wave
inspection as these wave modes are plate waves. According to Edsinger et al. [2]
an average concentration across the cladding wall of the order of 600 ppm leads
to measurable hydrogen embrittlement. However, the hydrogen concentration
across the wall is not uniform. The concentration at the outer surface of the
cladding may be ten times higher than at the inner surface. At the end of the fuel
295
rod life the corrosion rate as well as the hydrogen uptake accelerates rapidly [4].
However, zirconium alloys are said to be protected against hydrogen ingress by
the surface oxide film. Also the presence or absence of cracks or pores passing
through the oxide film into the metal/oxide interface is a key factor in
determining the hydrogen uptake. [4] These statements may have important
consequences for the detailed NDE assessment of cladding integrity in hot cells.
Further, because of the pellet-cladding interaction the geometry of the tube
will change and this will have consequences for both eddy current and laser
ultrasonic inspection. In laser ultrasonic inspection these changes in geometry
may influence on the efficiency of ultrasonic generation and detection as well as
on the distance travelled by the ultrasound. The geometry variation of the sample
surface causes tilting of the eddy current probe and probe alignment is needed.
The severity of these phenomena should be assessed and required precautions
should be taken.
Figure 1. Schematic illustration of the different processes occurring in a fuel rod during
irradiation [3]
296
Figure 2 illustrates the oxide growth along a fuel rod with Zircaloy-4 cladding
after three irradiation cycles and after an estimated burn-up of 65 GWd/tU. The
oxide thickness was measured axially along four trajectories on the rod separated
circumferentially by 90 degrees. The average was calculated from trajectories
taken every 30 degree [5]. Figure 3 from the same reference shows the hydride
morphology on a cross-section of the same fuel rod as in the previous figure.
The hydrides are mostly just below the oxide layer. The hydrogen content was
measured by hot vacuum extraction and ranged from 50 ppm to 1200 ppm.
Some cracks can be seen in the oxide layer.
Figure 2. Oxide thickness on a Zircaloy-4 fuel cladding tube after a fuel burn-up of about
65 GWd/tU using eddy current measurement [5]
297
Figure 3. Hydrides can be seen in the Zircaloy-4 cladding just below the oxide layer.
Some vertical cracks can be seen in the oxide layer [5]
3. Challenges in the hot cell environment

The radioactive environment in the hot cell gives an extra challenge to the
nondestructive measurements. In common, it is recommended that no electronic
equipments shall be placed inside the hot cell or they must be shielded with lead.
Since the walls of the hot cells are thick, approximately one meter in thickness,
the cable distance from the measuring probe to the device is long. Most
nondestructive measuring methods (e.g. high frequency ultrasound using
electrical transducers or eddy current) need a preamplifier near the transducer. In
most of the cases also the preamplifier needs to be located outside the radiation
protective walls of hot cell.
The radioactive radiation from the samples is impacting also on the transducers.
The gamma dose level needs to be taken into account in designing the probes,
cables and optical fibres and assure that the probes are resistant to the radiation.
The operator cannot go inside the hot cell. The samples and equipments inside
the hot cells are therefore handled with remotely controlled arms. The movement
of these arms is limiting the usability of the probes and samples and causes
restrictions in the testing methods. For example the measuring probes need to be
298
well implemented in a cassette that will just be inserted to the manipulator inside
the hot cell and be ready to use.
4. NDE methods for hot cell applications

Two nondestructive methods to evaluate the oxide thickness on fuel claddings
are discussed more in this section. The methods are laser ultrasonics (LU) and
eddy current (EC) technique. Some estimates of the possibility of using these
methods for measuring hydrogen concentration and hydride distribution are also
given. The eddy current and laser ultrasonic techniques are expected to be
complementary to each other. Laser ultrasonics has the advantage of being
completely non-contact since both ultrasonic wave generation and detection are
done using laser beams. These can be brought into the hot cell using optical
fibres. For both laser ultrasonics and eddy current techniques a scanning mechanism
is needed to scan the EC probe or optical fibres over the inspected specimen.
4.1 Laser ultrasonics
Figure 4 shows a schematic view of a laser ultrasonic inspection system for

testing tubes [6]. Ultrasonic waves are excited with one pulsed laser beam and
the propagating surface (Rayleigh or Lamb) or bulk waves are detected by
another laser beam using interferometric techniques. The technique is fully
noncontact (it can therefore be applied to hot and moving components, velocity
up to 5 m/s [6]) and the measurement beams can be scanned along and around
the tube. Further, the ease of changing the generation and detection spot
geometries according to wave type to be measured, the large bandwidth and the
absence of coupling medium give advantages over traditional ultrasonic
techniques. By analysing the dispersive (the frequency dependence of ultrasonic
velocity) properties of surface waves and/or guided waves (Lamb) the properties
of surface layers and the integrity of the tube wall can be assessed.
299
Figure 4. Inspection of a tube by a laser ultrasonic system [6]. Ultrasonic waves are
excited and received in a fully noncontact manner using laser beams. The measurement
heads are scanned over the tube surface.
4.1.1 Current laser ultrasonic applications
Kruger et al. [7] have applied the LU technique to simultaneously obtain the
elastic properties, density and thickness of thermal spray coatings (WC-Co).
They also used the technique to detect cracks (perpendicular to the coating) in
the coating and coating delamination. The thickness of the thermal spray
coatings (> 200 µm) was much larger than thickness range of zirconia on fuel
cladding, but the properties of the thermal spray coatings, like their special
lamellar structure, inhomogeneous composition and the presence of
discontinuities like pores and cracks makes the work of Kruger et al. interesting
also for the development of the application of LU to characterization of
zirconium alloy fuel claddings. Kruger et al. determined the coating thickness,
density and elastic modulus by comparing measured ultrasonic dispersion curves
to theoretical curves and their results are shown in Figure 5.
300
Figure 5. Determination of coating thickness (h), density (ρ) and elastic modulus (E) from
measured and theoretical dispersion curves (a) and a comparison of the obtained
thickness values to thickness values measured with optical microscopy [7].
Kruger et al. scanned a steel specimen with WC-Co coating with laser
generation and detection spots separated by 0.5 mm [7]. The presence of cracks
will block the propagation of surface acoustic waves. The measurement of a
waveform parameter like the amplitude of the Rayleigh wave can easily identify
zones with and without cracks. In Figure 5 (left) the distance between peaks is
found to be the same as the distance between the vertical cracks measured by
metallography from the sample cross-section.
If there is complete delamination between the coating and the substrate, the
coating will vibrate like a membrane when excited by the generation laser. This
vibration can be readily detected with a system responding at sufficient low
frequencies. Signals obtained from a WC-Co coating in a region where there is
good adhesion show a clear difference from signals from a region where the
coating is detached as shown in Figure 6. In the tested sample, there were large
oscillations at about 200 kHz in the signals from the detached area.
301
Figure 6. Laser ultrasonic detection of cracks in a WC-Co coating layer. Left, upper
picture, cross-sectional view of the sample with the generation/detection scheme and
(lower left) the variation of a waveform parameter along the sample. Right, the variation of
the vibrational amplitude in a region with partial delamination between the coating and the
substrate [8].
Figure 7. Laser ultrasonic detection of delamination. The signal from the delaminated
region is clearly different from the signal from a well bonded zone [8].
302
4.1.2 Characterization of hydrogen content
Hydrogen embrittlement is one of the major mechanisms responsible for the

degradation of ductility of Zircaloy cladding materials. Currently the characterization
of hydrogen concentration (HC) very often relies on destructive methods that are
time-consuming and costly. Yang and Huang [9] present an ultrasound based
non-destructive method for determination of HC in Zircaloy claddings. This
method is based on low frequency acoustic microscopy (AM) rather than laser
ultrasonics as shown in Figure 8. However, the same authors have ongoing work
for determining HC also using laser ultrasound, if successful, this new work will
make the technique more suited for hot cell applications as laser ultrasonics is a
non-contact method and the need for immersion will therefore be eliminated.
Figure 8. Use of acoustic microscopy (AM) for characterization of hydrogen content in zirconium
alloy claddings. The central frequency of the transducer is slightly less than 4 MHz [9].
The underlying principle of the AM technique for HC assessment is to detect the

change in ultrasonic wave speed caused by hydrogen in the Zircaloy cladding.
303
The ultrasonic waves propagating in the cladding are dispersive, meaning that
the ultrasonic waves have wave speeds depending on their frequency. Two
acoustic paths are important in Figure 8. The first path is the downward acoustic
ray N; this ray is reflected from the specimen as the upward ray N. In the second
path, an oblique acoustic ray S from the transducer excites a guided wave G
travelling along the sample. This wave radiates back to the transducer through
the ray S. The phase velocity of the guided wave G carries the information of the
mechanical properties that is related to the HC. Due to different ultrasonic wave
speeds of waves propagating along NN and SGS, an alternating constructive and
destructive interference signal can be detected when the transducer is scanned
vertically (z-direction). By analysing the alternating period in the vertical scan,
the phase velocity of the guided wave G can be determined. An elaborate
Fourier-based signal processing scheme can be used to analyse the periodic
phenomena and extract the dispersion spectra of the guided waves. A total
scanning distance of 5 mm with 200 steps is used to obtain the dispersion
spectra. The phase velocity of the fundamental anti-symmetric Lamb mode A0 is
used for a correlation with HC in the cladding.
4.2 Eddy current techniques
Single frequency eddy current measurements are routinely used for oxide
thickness measurements on zirconium fuel cladding tubes in hot cells. Small
diameter pencil probes are usually applied. The frequency of the eddy current
driving current is usually between 1 and 3 MHz. At these frequencies the
electromagnetic wave does not penetrate through the wall of the zirconium tube.
The probe response is mainly dependent on the lift-off, i.e. the oxide induced
separation between the tip of a probe and the metal surface. This method usually
gives reasonably good results if the oxide layer is non-magnetic. The EC lift-off
measurement technique is illustrated in Figure 9 where the EC probe is a coil
driven by an AC current. The electromagnetic field generated by the coil will
create eddy currents in the metal nearby the probe tip and these currents in turn
induces a new AC current in the coil thereby changing the impedance of the coil.
The coil impedance will depend on the electrical conductivity and permeability
of metal under the probe, on the thickness of the metal, on the frequency of the
driving AC current and on the distance between the probe tip (lift-off) and the
metal surface. Therefore the measured impedance can be correlated to the
thickness of the oxide layer and to the wall thickness.
304
Figure 9. The principle of the eddy current oxide thickness measurements (a) Cladding
tube with white oxide layer only (b) Cladding tube with white oxide layer and grey CRUD
layer [10]
In the Figure 9 the left hand side shows a cross-section of the zirconium tube in
as-fabricated condition. The right hand side shows a cross-section of the tube
after irradiation. The thickness of metallic tube wall has been decreased and an
oxide layer has been built up on the outer surface of the tube. In the lower right
part of the picture there is a deposit (CRUD) layer on the surface of the oxide
layer. The deposit not only causes an extra lift-off, it may also significantly
change the EC response due to its ferromagnetic nature [10].
305
A reference tube is needed to enable coating thickness measurement. A thin

isolating tape or a plastic foil is usually passed around the tube to simulate oxide
layers of different thickness.
For the evaluation of the cladding health a lower frequency should be applied.
The lower frequency should be of the order of 500 kHz to achieve 0.6 mm
penetration depth which is the wall thickness of the tube. The low frequency
eddy current technique can be used for wall thickness measurement but it is also
sensitive to the inner surface defects (cracks) and major inclusions in the
zirconium wall.
Also measurements with two frequencies can be applied to achieve the results
from the tube oxide and the cladding health at the same time. Also when the tube
wall thickness is not known a two frequency technique should be applied
The future reactor types bring new geometries to be inspected. The geometry
of the sample is giving a probe manipulator a variety of different challenges. In
the Figures 10 and 11 there are two different eddy current measuring set ups, one
for flat sample and the other for tubular sample.
Figure 10. A schematic setup for planar sample eddy current measurements.
306
Figure 11. An eddy current measuring setup for tubular sample.
5. Conclusions
Eddy current techniques are widely used in the hot cell environments, as their
tolerance to radiation has been tested in real service. However, the accuracy of
the oxide thickness measured with EC should be improved.
Laser ultrasonics is much more rarely used in the field of nuclear NDE
applications. The recent results in the developing laser ultrasonics suggest that
the method is promising also for hot cell environments.
6. Summary
NDE techniques for nondestructive evaluation of materials degradation due to
irradiation are presented. Two methods, laser ultrasound and eddy current, are
more specifically discussed.
There is a limited space available in a hot cell and the components to be
examined are radioactive. Therefore the NDE measurements should be
completely remotely controlled and the number of apparatuses inside the hot cell
should be minimal. This will restrict or limit the handling of the equipment.
Eddy current measurements based on the lift-off technique and laser
ultrasonics are proposed as promising future techniques for oxide measurements
of the irradiated fuel cladding.
307
References
1. De, P. K. John, J. T., Banerjee, S., Jayakumar, T., Thavaimuthu, M. & Raj, B. Assessment
of hydrogen levels in Zircaloy-2 by non-destructive testing. Journal of Nuclear
Materials, Vol. 252 (1997), pp. 43–54.
2. Edsinger, K., Yagnik, S. & Johnson, D. P. Method and apparatus for measuring hydrogen
concentration in zirconium alloy components in the fuel pool of a nuclear power
plant. Patent No.: US 7,388,369 B2, 2008.
3. Rudling, P. & Patterson, C. Fuel material technology report. Volume IV. ANT International,
Skultuna, Sweden, 2009.
4. IAEA 1998. Waterside corrosion of zirconium alloy in nuclear power plants. IAEA-
TECDOC-996. 313 p.
5. Kim, S.-K., Bang, J.-G., Kim, D.-H., Yang, Y.-S., Song, K.-W. & Kim, D.-S. Mechanical
property evaluation of high burn-up nuclear fuel cladding using the ring tensile
test. Met. Mater. Int., Vol. 15 (2009) 4, pp. 547–553.
6. IOS Intelligent Optical Systems. Inc. Brochure.
7. Kruger, S., Lévesque, D. Bescond, C., Lima, R., Maple, B., Campagne, B., Blouin, A. &
Monchalin, J.-P. Laser-ultrasonic evaluation of thermal spray coatings. 1st
International Symposium on Laser Ultrasonics: Science, Technology and
Applications. 16–18 July, 2008. Montreal. Canada.
8. Kruger, S., Blouin, A., Lévesque, D. & Monchalin, J.-P. Laser-ultrasonic evaluation of
thermal spray coatings. Presentation.
9. Yang, C.-H. & Huang, M.-F. Characterization of hydrogen concentration in Zircaloy

claddings using low-frequency acoustic microscope with a PVDF/LFB transducer.
Journal of Nuclear Materials, Vol. 335 (2004), pp. 359–365.
9. Yagnik, S. K., Johnson, D. P. & Kervinen, J. A. Eddy Current Measurements of Corrosion

and Metal Loss in Zircaloy Cladding with Ferromagnetic CRUD. Nuclear
Technology, Vol. 147 (2004), pp. 291–300.
308
Series title, number and report
code of publication
VTT Symposium 265

VTT-SYMP-265
Author(s)
Pertti Auerkari & Juha Veivo (eds.)
Title
Baltica VIII
Life Management and Maintenance for Power Plants. Vol. 2
Abstract
BALTICA VIII. International Conference on Life Management and Maintenance
for Power Plants, Helsinki–Stockholm–Helsinki, May 18–20, 2010.
This is one of the two volumes of the proceedings of the BALTICA VIII
Conference. The Conference aims to review recent experience, new technology
and views on future development for supporting successful, safe and productive
operation of power plants.
The triannual BALTICA Conference offers an updated state-of-the-art review on
the life management and maintenance of thermal power plants. The themes in
BALTICA VIII include e.g. consequences of nuclear revival with issues of the life
cycle, new insights into tools and methods for life management and
improvements in managing the assets under constraints such as aging, cycling,
new fuels and increasing fraction of low to zero emission systems, and recent
experience on example cases from new and ageing plants.
ISBN
978-951-38-7593-2 (soft back ed.)
978-951-38-7594-7 (URL: http://www.vtt.fi/publications/index.jsp)
Series title and ISSN Project number
VTT Symposium 71111
0357-9387 (soft back ed.)
1455-0873 (URL: http://www.vtt.fi/publications/index.jsp)
Date Language Pages
May 2010 English 308 p.
Name of project Commissioned by
Baltica VIII
Keywords Publisher
thermal power plant, nuclear plant, boiler, VTT Technical Research Centre of Finland
turbine, condition, life, fuel, maintenance, P.O. Box 1000, FI-02044 VTT, Finland
inspection, monitoring, risk, reliability, material, Phone internat. +358 20 722 4520
damage, corrosion, fatigue, creep, degradation Fax +358 20 722 4374

Life Management and Maintenance For Power Plants Vol. 2010

Uploaded by

Copyright:

Available Formats

Life Management and Maintenance For Power Plants Vol. 2010

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

Life Management and Maintenance For Power Plants Vol. 2010

Uploaded by

Copyright:

Available Formats

What are some of the themes discussed in the conference?

What are some of the themes discussed in the conference?

What are some of the challenges faced in managing power plants?

What are some of the challenges faced in managing power plants?

Keywords: thermal power plant, nuclear plant,

VTT SYMPOSIUM 265 boiler, turbine, condition, life, fuel, maintenance,

Edited by Pertti Auerkari & Juha Veivo

JULKAISIJA – UTGIVARE – PUBLISHER

VTT, Vuorimiehentie 5, PL 1000, 02044 VTT

VTT, Bergsmansvägen 5, PB 1000, 02044 VTT

VTT Technical Research Centre of Finland

Edita Prima Oy, Helsinki 2010

Espoo, 18 May 2010

Pertti Auerkari & Juha Veivo

Benchmarking for plant maintenance

Kari Komonen, Toni Ahonen and Susanna Kunttu

According to the management of many industrial enterprises ‘benchmarking’

2. Web-tool for benchmarking

Continual statistical Determination Statistical

Pareto grouping of Statistical

Figure 1. Schematic description of benchmarking system.

3. Features of benchmarking system

3.1 Data input

Figure 2. The illustration of the data input page.

The above mentioned indicators are affected by several external influencing

Figure 3. The conclusion page of the positioning.

3.3 Performance indicators

The simplified system of the key performance indicators can be classified in a

objectives such as MTTF or MTTR are performance indicators, which should be

Also other key performance indicators seem to be a function of exogenous

Log (Plant replacement value)

Figure 4. Example of causalities between key performance indicators: ‘Breakdown hours

In the case of the food industry as an empirical evidence, multivariate regression

Processing of food and liquid (N = 34)

Processing of food and liquids

Figure 5. The predicting power of independent variables on relative maintenance costs in

3.4 Locating differences

Applying positioning to all the cases in the sample it is possible to compare

Figure 6. The conclusion page of locating differences.

3.5 Best practices and planning

modes of operations determined by action variables such as preventive maintenance

Komonen, K. 2002b. Views on performance indicators and benchmarking in industrial

Komonen, K. 2005. E-FAMEMAIN Benchmarking tool for industrial maintenance – a web

Komonen, K. 2006. E-FAMEMAIN: A new kind of web-based benchmarking tool.

Komonen K. 2006b. Availability guarantees and physical assets management: Empirical

Managing corrosion in biomass boilers:

Satu Tuurna1, Tommi Varis1, Kimmo Ruusuvuori1, Stefan Holmström1,

Oxidation and corrosion of materials in highly demanding applications is critical

2. In-service boiler testing

Figure 1. An as-coated tube attached to a rotating fixture.

Figure 2. A coated tube installed into an economiser in boiler 2.

Table 1. Summary of produced coatings for different boilers. AM = amorphous.

Boiler Powder Tube

3. Results of field testing

5. Benefits and limitations of coatings

3. Nielsen, H. P., Frandsen, F. J., Dam-Johansen, K. & Baxter, L. L. The implications of

5. Montgomery, M., Vilhelmsen, T. & Jensen, S. A. Potential high temperature corrosion

Selecting the right material for recovery

Esa K. Vakkilainen and Pekka Pohjanne*

John M. Brear, Paul Jarvisand Stephen Scully+