E²G INDUSTRY

S UM M E R / FA L L 2 0 1 8 » VO LU M E 6

RECENT DEVELOPMENTS IN PIPING

VIBRATION SCREENING LIMITS
 E²G INDUSTRY INSIGHTS » VOL. 6

RECENT DEVELOPMENTS IN PIPING

VIBRATION SCREENING LIMITS

INTRODUCTION

Piping vibration limits are typically justified by experience and While design guidance is provided in standards such as
lack a well-documented technical basis. This article summarizes ASME OM-3 [2] for the nuclear industry and API 618 [3] for
the work performed by E²G and fellow industry collaborators reciprocating gas compressors, API/ASME standards do not exist
working as members of the API 579 Part 16 Task Group on the for the vibration fatigue assessments of in-service piping.
development of a new Fitness-for-Service method to evaluate
vibrations in process piping. The complete article can be found The proposed methodology addresses the following common
in the ASME PVP Proceedings PVP2018-84847. questions that arise when using existing methods:

Members of the API 579 Task Group developed the updated 1. In the case of multiple dominant spectrum peaks, which
vibration screening criteria by performing Finite Element frequency spectrum (FFT) peak should be assessed in
Analysis (FEA) on 20,000 randomly generated candidate piping the vibration frequency spectrum? How is a dominant
models, and evaluating the results using high-cycle welded spectral peak defined in situations with multi-modal
joint fatigue models for both constant amplitude and variable vibration?
amplitude loading.
2. What is the amount of safety margin embedded in the
The screening limits presented are applicable to situations of existing industry criteria?
mechanically-induced, two-phase flow-induced, turbulent-induced
vibration of single-phase process fluid, or wind-induced, 3. What is the difference between ‘correction’ and
vibration. Instead of using a single blanket limit, to reduce ‘danger,’ or ‘concern’ and ‘problem’ levels, when
conservatism, separate limits are now established for butt-welded multiple limits are defined in a single method?
and non-butt construction. Prior screening limits are mostly
overly-conservative and, when used, often leading to
unnecessary piping support alterations/additions, or costly
piping configuration changes and unit downtime.

E²G | The Equity Engineering Group, Inc.

REVIEW OF EXISTING SCREENING CRITERIA

Velocity RMS-Based Screening Peak-to-Peak Displacement-Based Screening

Most vibration screening criteria are either velocity-based The origins of the most commonly used displacement-based
or displacement-based. The Energy Institute (EI) Guidelines criteria are less known. These criteria are shown in Figure 2.
for Avoidance of Vibration Induced Fatigue Failure in Process Literature [5] credits Nimitz [6] in 1974 for the initial development
Pipework [1] presents the most commonly used velocity of the ‘haystack curves’; however, Reference [6] is not readily
limit criteria based on measurement of the root-mean-square available. E2G has identified the earliest publication of these
(RMS) velocity. limits in a 1979 Southwest Research Institute (SWRI) Report 160 [7].

The EI limits [1] are shown in Figure 1 and plotted against the The Nimitz/SWRI displacement criteria require measurement
most conservative ASME OM-3 [2] limit of 0.5 ips 0-pk or 0.352 of a peak-to-peak displacement and provide 5 choices of
ips RMS. This conversion of 0-pk amplitude units to RMS units limits. Reference [7] indicates these criteria are for average
assumes a crest factor (CF) of 1.414. piping systems constructed in accordance with good engineering
practices and that additional allowance for critical applications
The EI limits are based on an allowable dynamic stress or unreinforced branch connections should be made.
amplitude of 2.5 ksi, which corresponds to a BS7608 [4] Class These statements, along with the multiple choices for
F2 Welded Joint Fatigue curve, and a Safety Factor (SF) of 2.0. acceptability limits and lack of available technical basis, place an
The ‘concern’ limit forms the lower bound of a variety of FEA excessive amount of onus on the user.
models corresponding to the vibration velocity and frequency
that result in 2.5 ksi constant amplitude fatigue limit. The 2.5 ksi Figure 3 compares all of the displacement and velocity limits
fatigue limit is analogous to an endurance limit for flaw on a single plot by converting the displacement-based limits
initiation, and relates to a threshold stress intensity that does to RMS velocity per the fundamental relationship,
not result in crack growth.

The EI limits consist of 10 rigorously evaluated classes of 1 v = 2π ƒ · u

small-bore cantilever FEA models and one class of mainline
piping, per Appendix A of Reference [1]. The exact number of
geometry variations investigated for each class is not published.
Additionally, the basis of the upper ‘problem’ line is unknown, but 101
is approximately a factor of 4.9 greater than the ‘concern’ line.
A l l o w a b l e R M S Ve l o c i t y ( i p s )

Using these curves for random (variable amplitude) vibration

can be problematic, for two reasons. First, an assumed CF 100
of 1.414 was likely used to cast the FEA-generated velocities
from 0-pk units to RMS units, which is unconservative for most
random vibration. Secondly, a constant amplitude fatigue
limit was implemented which may not be satisfactory for
10-1
cases of Gaussian vibration or any level of high kurtosis
(non-Gaussian) vibration.

However, these are not necessarily oversights of the screening

10-2
limits. Experience has proven the EI criteria successful,
likely due to the conservative choice of a low 2.5 ksi 100 101 102
allowable stress amplitude. Additionally, based on the authors’ Ef f ect i v e F r equ en cy ( Hz )
experience, unless heavy slugflow or support impacting occurs,
kurtosis levels are rarely in excess of 3.5 for flow-induced piping EI Concern
vibration. EI Problem
Conservative ASME OM-3

Figure 1. Velocity-Based Screening Criteria

E²G | The Equity Engineering Group, Inc.

where ƒ is the frequency in Hz, and u is the displacement frequent, stress ranges previously below the threshold stress
amplitude. Figure 3 shows the EI ‘concern’ limit is nearly intensity will now produce stress intensities in excess of the
equivalent to the Nimitz/SWRI ‘marginal’ limit, and the threshold stress intensity. Therefore, random vibrations are less
EI ‘problem’ limit is consistent with the Nimitz/SWRI ‘danger’ likely to result in infinite life scenarios due to crack growth.
limit. This similarity is reassuring given the better-known
technical basis of the EI limits for small-bore piping and the If the highest measured vibration amplitude is less than 5x
historic use of the Nimitz/SWRI limits for mainline-routed piping. the signal RMS (CF=5.0), and if 5x RMS is less than half of
a non-propagating stress range, Soc, there is a substantial
It is important to note that the assumption of infinite life is probability that all stress ranges were likely below Soc for
embedded in both screening criteria. The existence of a the duration of the signal. Therefore, as long as the user is
constant amplitude fatigue limit for welded and non-welded confident that operational conditions were maintained steady
specimens is sometimes explained through the concept of over the duration of the vibration (system not subject to
the threshold stress intensity discussed in fracture mechanics transients), a non-propagating stress limit, Soc, can typically be
textbooks. However, for situations of random vibration, an used when coupled with rigorous NDE inspection to ensure
increasing probability exists that an initial flaw can be driven that no pre-existing flaws are present and the backsides of
to a depth or length such that all subsequent lower, but more single-sided pipe welds are free of excessive stress risers.

DISPLACEMENT-BASED SCREENING LIMITS SCREENING LIMITS CAST IN RMS VELOCITY

101 101
Allowable Displacement (pk-to-pk in)

Allowable RMS Velocity (ips)

100

10-2 10-1

10-2

10-3 10-3
100 101 102 100 101 102
Ef f e c tiv e F re q u e n c y (H z) Ef f ect i v e F r equ en cy ( Hz )

API 618 Design EI Concern

Nimitz / SWRI Threshold of Perception EI Problem
Nimitz / SWRI Design Conservative ASME OM-3
Nimitz / SWRI Marginal API 618 Design
Nimitz / SWRI Correction Nimitz/SWRI Threshold of Perception
Nimitz / SWRI Danger Nimitz/SWRI Design
Nimitz/SWRI Marginal

Nimitz/SWRI Correction
Nimitz/SWRI Danger

Figure 2. Commonly Used Displacement-Based Screening Criteria Figure 3. Compilation of All Commonly Used Screening Criteria
Cast into RMS Velocity Units

E²G | The Equity Engineering Group, Inc.

RMS vs. Peak-to-Peak Measurements

In the case of random vibration, an accurate signal RMS can with the RMS of the signal. In each of the 100 test cases,
be measured with substantially fewer data points than a true less than 10,000 samples are required for 10% accuracy of the
peak-to-peak signal for the same vibration time waveform. In signal RMS, as compared to over 175,000 samples for the same
addition, the signal RMS is tolerant to the occasional outliers in accuracy when searching for the true peak value (which was
data that can be sources of implementation error in identifying known ahead of time for these calculations since these were
a signal’s true peak-to-peak value. pre-generated datasets).

To illustrate these advantages of the signal RMS over the Using probability theory, it can be shown that a peak with
signal peak-to-peak, Figure 4 shows the results of 100 a CF greater than 5 (5x RMS) is only found with a 40%
pre-generated Gaussian narrow-band random signals probability if 2 million peaks are measured. Based on the
representative of vibration from two-phase flow-induced author’s experience, most Gaussian random piping vibration
turbulence. The signal was evaluated to determine the relative data have a CF of 4 to 4.5 for typically measured datasets.
error between the highest measured peak value and the actual This result alludes to the difficulty in obtaining a true peak value
pre-determined true peak in the dataset. The same was done for a random piping vibration.
with the RMS of the signal. In each of the 100 test cases,

100
Percent Error in Measured Pk Velocity

80

60

40

20

0
0 50K 100K 150K 200K 250K

100
Percent Error in Measured RMS Velocity

80

60

40

20

0
0 50K 100K 150K 200K 250K

Number of Samples

Figure 4. 100 Examples of the Relative Error in Finding the True Peak Value Versus the True Signal RMS in Pre-Generated Data as More Data Points are Sampled

E²G | The Equity Engineering Group, Inc.

Interpretation of Frequency when Screening Piping Vibration

Some practitioners interpret the screening limits as spectral In cases of narrow-band random vibration (i.e., approximately
comparisons by plotting the entire FFT spectrum against the one peak per zero-crossing), the frequency of occurrence of
screening criteria. While this works for vibration responses major peaks in the time waveform will almost exactly produce
with a dominant frequency, treatment of multi-modal the same result as shown in equation (2) and can usually be
piping vibration signals is often a source of confusion when performed through observation of the time waveform directly
viewing the amplitudes of an FFT. It is important to note that in the field.
FFT amplitudes do not describe how the phases of the signal
combine to form the overall stress-producing time waveform. Figure 5 shows the velocity FFT spectrum of a random vibration
Therefore, the better measure of fatigue damage is the time signal plotted against the EI screening limits. The entire
waveform itself, of which the overall energy is quantifiable spectrum is below the lower EI ‘concern’ limit, so many users
using the signal RMS. would consider the measured vibration to be acceptable.
However, the FFT amplitudes do not describe how the phases
In the case of constant amplitude vibration with a single of the signal combine to form the overall stress-producing time
dominant frequency or sinusoidal response, plotting of the signal. This is evident by plotting the overall RMS velocity at the
frequency for screening purposes is straightforward. However, effective frequency (red dot). The overall RMS value is more
if the vibration is broad-banded random, a single dominant representative of the total energy of the vibration. This example
frequency does not exist; instead, the energy is smeared over is representative of cases of random vibration from two-phase
a frequency range. In this circumstance, the highest amplitude flow-induced turbulence, where conventional screening methods
of one frequency bin does not represent the overall energy of can be unconservative.
the vibration. Assuming so will produce unconservative
results. Similar confusion can exist with multi-modal responses
typically found in reciprocating compressor systems. In either
case, prior screening methods do not provide clarification
into broad-band random, multi-modal and multi-peak
vibration analysis.

Instead of using the dominant spectral frequency, the evaluated 101

frequency should be the frequency of stress reversals 100

driving the fatigue damage. If multi-modal vibration exists in
the FFT, the effective frequency should be equal or closer to the 10-1
RMS Spectrum (ips RMS)

most dominant frequency in the FFT. For an effective frequency,

10-2
E2G recommends using the zero up-crossing frequency
calculated from the RMS-scaled frequency spectrum as [8], 10-3

10-4

N 10-5
1
Ʃ
2 2
ƒe = ƒi v RMS,i 2
V RMS 10-6
i
10-7
100 101 102
where V��� is the overall signal RMS velocity and νrms,i is Ef f ect i v e F r equ en cy ( Hz )
the ith amplitude of the RMS-scaled velocity spectrum at a
frequency ƒi for the N bins in the spectrum. In cases of pure EI Screening
sinusoidal vibration, equation (2) collapses to the frequency EI Screening
of the sinewave. Spectrum
Overall RMS and Effective Frequency

Figure 5. Example Showing Conventional Use of Existing EI

Screening Curves with Broadband Random Signal

E²G | The Equity Engineering Group, Inc.

Limitations of API 579-1/ASME FFS-1 Welded Joint Fatigue

While the API 579 Part 14 welded fatigue curve is a modern design fatigue curves were later modified for cycles
and successful method for evaluating fatigue in weldments greater than 1x10⁶ in post-1995 versions of ASME Section
below 10⁷ cycles, the methodology is not designed for use VIII-2 [13]. If that was the source for the non-welded endurance
in vibration service, where stress cycles can typically be limit, it was likely selected based on choosing the stress
above 10⁷ cycles. When developing the API 579 Part 14 corresponding to the right-most point on the elastic fatigue
welded SN curve, only 19 known weldments were tested to curve at 1x10⁶ cycles.
failure above 10⁷ cycles. This does not necessarily justify a
single sloped fatigue model (when viewed on a log-log plot). Historically, the most technically appropriate choice of
The Part 14 welded SN Curve simply does not have the allowable stress for screening becomes moot once
experimental backing to be used for cycles greater than 10⁷. conservative safety factors (SF) and stress concentration
Experimental data of near-threshold crack growth factors (SCFs) are selected. If a standard v-groove weld cap
behavior in weldments clearly show a non-propagating has an SCF of 2.0, and an SF of 2.0, the resulting non-welded
stress limit and have been documented by OHTA [9]. In the endurance limit would be (26,000 psi range)/[(2.0)(2.0)],
authors’ experience, use of a m ~ 3.0 SN curve past 10⁷ or 6,500 psi range. The result is close to the BS7608 design Soc
cycles commonly results in mean-life predictions that do not of 7,687 psi [4]. Fortunately, no additional SF should be
agree with the service life of most vibrating piping systems. required for the welded joint Soc since it is based on a mean-2σ
design level. Ultimately, use of the base metal allowable versus
Adjusted high-cycle fatigue models are currently under the welded fatigue allowable becomes a technicality when
development by E²G and are intended to be balloted in the considering both have been validated through experience.
upcoming edition of API 579 Part 14. For now, this article However, E²G’s goal is to reduce conservatism through the
uses the high-cycle fatigue models per European Standards refinement of such assumptions.
BS7608 [4] and PD5500 [10] due to their closer agreement with
near-threshold crack growth behavior. Random Vibration Allowable Stress

Random allowable stress often exhibits CFs greater than 1.5.

ALLOWABLE VIBRATION STRESSES
In this case, the method implements a bilinear version of a
log-log fatigue curve in the form of,
For the purposes of vibration screening, choice of an allowable
RMS fatigue stress is based on either the constant amplitude
fatigue limit (CAFL) per BS7608 [4] for cases of periodic m
vibration, or derived from spectral fatigue methods for 3 NS r = A
assessing Gaussian or non-Gaussian random vibration.
Results of both cases are used to establish screening limits and
are further discussed in the following two subsections. where N is the cycles to failure, A is the intercept parameter,
Sr is the stress range, and m is the negative inverse slope.
Periodic Vibration Allowable Stress The fatigue model consists of negative inverse slope parameters,
Periodic vibration often exhibits CFs of less than 1.5. To establish m = 3 for low cycle, and m +2 for high cycle behavior.
periodic vibration screening limits, σallow,��� is chosen from the
allowable stress of a mean – 2σ Design Class D welded joint Since narrow bandwidth Gaussian vibration has Rayleigh
non‐propagating stress range, adjusted for the assessment distributed amplitudes, it can be shown that Miner’s rule for
temperature. At 70°F, this limit has a range of Soc = 7687 psi damage accumulation is well-approximated by,
(53 MPa) and is converted to the RMS value of 2718 RMS psi
assuming a CF of 1.414.

∫
∞
ze -z² / 2
4 DG = ƒe Tdur dz
Wachel [11] [12] used an allowable stress range of 26,000 N (2zSRMS )
0
psi for carbon steel. This allowable range is consistent with
non-welded fatigue models originally found in pre-1995
editions of the ASME Section III and ASME Boiler and Pressure
Vessel Code Section VIII-2 [13]. However, the ASME non-welded

E²G | The Equity Engineering Group, Inc.

where fe is the average zero up-crossing frequency from Figure 6 is the result of using equation (4) to determine the
equation (2), and Tdur is the signal duration. The integration combination of RMS stress, frequency, and kurtosis that
variable z is defined as z = Sr/(2S���), and is equivalent to the results in 100% damage following 25 years duration of
signal CF. Implementation of this equation for the proposed narrow-band random vibration.
API 579 Level 2 and 3 vibration evaluations is further discussed
in Breaux, et al. [14]. For the purposes of determining random allowable vibration
stresses, ku is limited to 3.0 and a finite design life of 25 years
Equation (4) can be further augmented to calculate the is selected in equation (6). Further allowance is provided by
damage for non-Gaussian random vibration, D��, using the limiting the allowable RMS stress to the higher of the
following kurtosis damage amplification factor [8], 25-year random finite life allowable stress or 769 psi RMS,
which corresponds to Soc and a CF of 5.0, since most Gaussian
random vibration CFs follow 4.5 ≤ CF < 5.0.
1
g= m(m - 1)(ku - 3) +1 5
24 Equation (4) and use of Figure 6 offer substantial improvement
over the existing API 579 Part 14 Level 2 and 3 fatigue
evaluation methods for random vibration, since instead of
such that non-Gaussian damage, D�� = γD�.. tedious cycle counting, only the RMS velocity of the vibration is
needed to evaluate acceptability.
A non-Gaussian random vibration with the same RMS
velocity and a ku = 4.0 produces 1.83 times the fatigue
damage as compared to a Gaussian random vibration with the ALLOWABLE STRESSES FOR RANDOM VIBRATION
AND 25-YEAR LIFE PD5500 DESIGN CLASS D
same RMS velocity. In E²G’s experience, unless heavy slug flow
or support impacting occurs, kurtosis levels are rarely in excess
of 3.5 for typical flow-induced piping vibration. Due to the 1000
effect of the kurtosis damage amplification factor, in situations 900
with impacting, vibration screening is not recommended unless
Allowable RMS Velocity (ips)

800
signal kurtosis is explicitly calculated to verify the vibration is
Gaussian Random. 700

600

500

400

300

200

100
100 101 102 103
Ef f ect i v e F r equ en cy ( Hz )

Kurtosis = 3 (Gaussian)
Kurtosis = 4
Kurtosis = 5
Kurtosis = 6
Kurtosis = 7
Kurtosis = 8
Kurtosis = 9

Figure 6. Allowable RMS Vibration Stresses and Frequency

Combinations that Result in Design Damage less
than 1.0 for a 25-Year Duration

E²G | The Equity Engineering Group, Inc.

NUMERICAL INVESTIGATION OF CANDIDATE PIPING SYSTEMS

The allowable RMS vibration velocity, vallow,���, is determined Point mass and distributed weight correction factors K1 and
with choice of an allowable RMS stress, σallow,���; a stress K3 described by Wachel [11] [12] are included in the analysis
concentration factor, SCF; and the stress-to-velocity ratio, since the FEA models use an effective metal density to
Kv, that defines the dynamic stiffness of the piping system and capture the randomly selected product-specific gravities and
is frequency-dependent. insulation thickness. Choice of insulation thickness is based
on a randomly selected pipe OD and temperature using
the Equity Engineering Practices insulation table for use of
σ allow,RMS 15 lbf/ft³ insulation on process piping as a function of diameter
vallow,RMS = 6
Kv · SCF and temperature.

The effects of concentrated weights are captured by randomly

By random selection of the variables shown in Table 1, selecting a span leg and location to place a flanged gate valve
5,000 FEA piping models are developed for each of the 4 (with two mating flanges). The mass moments of inertia of
investigated geometries shown in Figure 7. The candidate the valves and mating flanges are included in the model with
geometries represent the majority of mainline-routed piping knowledge of the randomly selected valve class and pipe OD.
systems constructed in accordance with good engineering
practices. The FEA models used ABAQUS B32 beam elements Kv is derived using the ratio of the FEA-calculated maximum
and ELBOW31B elements to capture the flexibility of the long principal stress, σ1, at any location in the model to the velocity of
radius elbows. Short radius elbows are not included in the the natural frequency mode shape,
evaluation and are an area of further investigation. The chosen
geometries are intended to reflect those investigated by
Wachel, Morton, and Atkins in Reference [12]. s1 s1
7 Kv = =
The geometries shown in Figure 7 have translational and v 2p fU
rotation spring elements with stiffness randomly selected
per Table 1. Translational stiffness is randomly selected as
10³ lb/in, 10⁵ lb/in, or 10⁷ lb/in. Likewise, rotational stiffness is
selected as 10⁴ lb-in/rad, 10⁶ lb-in/rad, or 10⁸ lb-in/rad.
These values are intended to represent a free end/continuation,
A
moderately stiff, and rigid conditions, respectively. Because B
the piping supports are not explicitly included in the FEA
models, fatigue damage to the supports is not evaluated using A B C
this approach.

ELBOW31B elements are used for elbows, and therefore,

elbow stress intensity/SCFs are implicitly included in the Kv 3D in-Plane L-Bend 3D in-Plane Z-Bend

factors derived from the FEA. SCFs are factors on the nominal
far-field stress used to obtain the correct local
membrane-plus-bending stress at geometric discontinuities.
The product of the SCF and the nominal stress account for C
B
all geometric stress rising effect at a connection, except for B
the weld detail itself. SCFs are not explicitly modeled in the
Monte Carlo FEA simulations for items such as reducers, flanged A C A
elbows, tees/branch connections, stiffening rings, etc., that may
exist anywhere in the pipe spans. These sources of SCF are later
incorporated when defining a lower bound to the Monte Carlo
point cloud results. 3D in-Plane U-Bend 3D Bend

Figure 7. Piping Geometries Investigated

E²G | The Equity Engineering Group, Inc.

 TABLE 1: VARIED PARAMETERS AND PARAMETER RANGES FOR FEA PIPE MODELS

VARIABLE VARIABLE OPTIONS / RANGES ASSUMPTION

Pipe Outer Diameter NPS 2 to 48 per Pipe Tables Random Choice

Pipe Wall Thickness Schedule 40, STD, 80, 160, XXS Random Choice

Pipe Leg Lengths (ft) 0.0 through 20.0 Uniform Distribution

Translational Support Stiffness (lb/in) 1x10³, 1x10⁵, 1x10⁷ Random Choice

Rotational Support Stiffness (lb-in/rad) 1x10⁴, 1x10⁶, 1x10⁸ Random Choice

Presence of Flanged Gate Valve Leg A, Leg B, or Leg C Random Choice

Flanged Gate Valve Class 150 to 1,500 Classes Random Choice

Gate Valve and Flange Weights Based on Valve Classes for Sizes NPS 2 to 24 --

Temperature ( °F ) 70, 200, 400, 600, 800 Random Choice

Elastic Modulus Per Material and Temperature --

Insulation Thickness (15 lb/ft3 density) Per Insulation thickness Table in Equity EP 11-1-1 Random Choice

Material Low Chrome, Stainless, Carbon Steel Random Choice

Content Specific Gravity 0.7, 0.8, 0.9, 1.0 Random Choice

Elbow Type Long Radius (1.5 x OD) Constant

RESULTS

The presence of an unsupported mass (valve) appreciably curve is fit to the lowest allowable velocities found in 30 equally
increases the system Kv factors, regardless of mode shape. log-spaced frequency bins between 1 Hz and 150 Hz.
Additionally, on average, the first mode shape has substantially These values are plotted using red circles in Figure 9. The lower
less dynamic stress compared to higher order mode shapes. bound curves are subsequently shifted by an SCFs to account
These results are consistent with what was found by Wachel, for unmodeled structural discontinuities. Note that, since a
et al. [12]. welded fatigue model is used to define the allowable stresses,
the effect of the weld cap is already included in the result
Figure 9 shows the allowable periodic RMS velocity point cloud (i.e., a circumferential butt weld in a seamless pipe has an SCF = 1).
results, determined using equation 7, for the most limiting
geometry of a Z-Bend and a periodic allowable stress of 2718 In practice, SCFs can be computed using a variety of options,
psi RMS based on BS7608 (0.5Soc/1.414). Each point represents including ASME Section III Stress Indices, ASME B31.1 or B31.3
an FEA-calculated Kv factor (approximately 2,500 FEA cases Appendix D SIFs, ASME B31J (STP-PT-073 [15]), or FEA to
without valves and 2,500 FEA cases with a valve randomly placed determine the ratio of local membrane-plus-bending stress to
in leg A, B, or C). Points that lie above the existing EI guidance the far-field nominal stress for a geometric discontinuity.
suggest an unused margin. Points that lie below the screening
limit would suggest the EI screening limit is unconservative. There is a clear effect of the allowable velocity on frequency.
The periodic results suggest that mainline piping and piping Higher SCFs can be tolerated for frequencies below 10 Hz.
components with SCFs lower than 2 can be screened using the Substantial margin exists on the lower EI ‘concern’ limit, which
existing upper EI ‘problem’ line. would lead to overly conservative conclusions for SCFs > 8.0.

Final RMS velocity screening values are determined using Wachel, et al. [12] uses a non-welded endurance limit of 13,000
a lower bound to the point cloud results. The lower bound psi amplitude, a Kv factor of 318 psi/ips, and an SF of 2.0, and

E²G | The Equity Engineering Group, Inc.

concluded that a screening limit for piping without the effects The recommended mainline vibration screening values
of concentrated or distributed span weights would be 4.0 in/s have been divided into two groups: Group A and Group B.
0-pk. Wachel comments that the span weight correction factor The distinction between the groups is an SCF of 3.0. Group A
would be K3 = 1.5 for most systems. Casting this result into RMS (SCFs 1 to 3) coincides with mainline butt-weld connections
ips units results in a recommended screening limit (without the and other welded components with relatively low stress
effect of concentrated weights) of 4.0 ips/(1.5*1.414) = 1.88 concentration factors. Common examples may include
ips RMS. This aligns reasonably well with SCF = 2.0 at 10 Hz in butt-welded tees, sweep-o-lets, and piping systems that
Figure 9 A. terminate at certain vessel nozzles (e.g., a vessel nozzle
designed for cyclic service or with a large radius at the
When unsupported valves (point masses) are included in cases vessel-to-nozzle junction).
of periodic vibration, the increased Kv has a noticeable effect
on allowable velocities. Reference [12] estimates that a typical
concentrated weight-to-span weight ratio of 3.0 would result
in a concentrated span weight correction factor, K1 of 2.7.
This would further reduce the Reference [12] RMS screening
limit from 1.88 ips RMS to 0.7 ips RMS. A 0.7 ips RMS limit aligns
with the current SCF = 3.0 results in Figure 9B at 10 Hz.

Figure 10A and Figure 10B present the results for allowable
RMS velocities for cases of Gaussian random vibration
(ku = 3.0). Substantial margin is identified between the point
cloud and the lower EI ‘concern’ line for SCFs ≤ 3.0. The results
indicate that, for SCFs > 3.0, the lower EI ‘concern’ line is
capable of producing unconservative results for cases of
random vibration.

Comparison of the random to the periodic results shows

that the random results approximately align with the SCF = 3.0
on the periodic results. This is not surprising, since 3.0 times Group B (3 < SCFs ≤ 8) coincides with mainline non-butt weld
the random allowable limit of 769 RMS psi is 2307 RMS connections and other welded or threaded components with
psi, which is close to the periodic RMS allowable limit of higher stress concentration factors. Common examples may
0.5Soc/1.414 = 2717 psi. include fabricated tees, socket welds, weld-o-lets, threaded
connections, and fillet and partial penetration welds.
The results align well with a study by Sebastien, et al. [16],
who by performing a similar study on 3,708 candidate For frequencies where the FEA-calculated lower bounding
piping models, found that 181 of the models (5%) fail a 0.5 in/s curves are lower than the EI ‘concern’ line, primarily for Group
RMS allowable velocity. After checking the 181 failures against B (SCFs > 3.0) subjected to random vibration, the EI ‘concern’
French nuclear RCC-M seismic design requirements, 94% of line is selected as the default minimum. This is due to its
the 181 models were discarded. In the present study, the alignment with the Nimitz/SWRI ‘marginal’ line that has
results strongly support the 0.5 in/s RMS limit for a wide range shown historic success for most process piping applications.
of SCFs and frequencies and can be a substantial reduction While some over-conservatism may exist in the presented
in unnecessary conservatism over the 0.352 in/s RMS results, the authors present a rigorous technical basis to permit
(0.5 in/s 0-pk) lower bound limit in the current ASME OM-3 [2]. the user to understand the embedded margins.

E²G | The Equity Engineering Group, Inc.

 Allowable Velocities - Periodic Vibrations Allowable Velocities - Periodic Vibrations
Simulations without Valves - Z Bend Simulations Legs with Valves - Z Bend

10³ 10³
A B
Allowable RMS Velocity (ips)

10² 10²

10¹ 10¹

10⁰ 10⁰

10-1 10-1

10-2 10-2
10-2 10-1 10⁰ 10¹ 10² 10³ 10-2 10-1 10⁰ 10¹ 10² 10³
Effective Frequency (Hz)

Figure 9. Z-Bend FEA Results for Periodic Vibration Compared to Existing EI Screening Guidance Cases without Valves (A) and Cases with
Unsupported Valves (B)

25-Year Allowable Velocities - Random 25-Year Allowable Velocities - Random

Vibrations Simulations without Valves - Z Bend Vibrations Simulations Legs with Valves - Z Bend

10³ 10³
A B
Allowable RMS Velocity (ips)

10² 10²

10¹ 10¹

10⁰ 10⁰

10-1 10-1

10-2 10-2
10-2 10-1 10⁰ 10¹ 10² 10³ 10-2 10-1 10⁰ 10¹ 10² 10³
Effective Frequency (Hz)

Mode 5 Best Fit to Lower Bound (SCF = 1.0) Best Fit to Lower Bound (SCF = 7.0)
Mode 4 Best Fit to Lower Bound (SCF = 2.0) Best Fit to Lower Bound (SCF = 8.0)
Mode 3 Best Fit to Lower Bound (SCF = 3.0) Best Fit to Lower Bound (SCF = 9.0)
Mode 2 Best Fit to Lower Bound (SCF = 4.0) EI Concern
Mode 1 Best Fit to Lower Bound (SCF = 5.0) EI Problem
Lower Bound Points Best Fit to Lower Bound (SCF = 6.0)

Figure 10. Z-Bend FEA Results for Gaussian Random Vibration Compared to Existing EI Screening Guidance Cases without Valves (A) and Cases with
Unsupported Valves (B)

E²G | The Equity Engineering Group, Inc.

SCREENING LIMITS

Based on the use of rigorous FEA modeling, Figure 15 presents

reasonable mainline screening limits, which are also presented PROPOSED SCREENING LIMITS
in equation form in this section. Given the provided technical MAINLINE PIPING IN PERIODIC VIBRATION
basis, these limits will aid in the development of the future
proposed Level 1 FFS procedure. 10¹
A

Applicability and Limitations

Allowable RMS Velocity (ips)

All component types and welds are assumed free of
10⁰
defects or active damage mechanisms, and are consistent
with applicable ASME piping codes. The procedures are
limited to cases of piping vibration that do not involve
intermittent impacting. Impacting may lead to random
10-1
vibration with large peak amplitudes and high kurtosis
(non-Gaussian random vibration).

Intermittent impacting in a piping system can occur due to:

10-2
10⁰ 10¹ 102
1. Unstable two-phase flow regimes such as slug flow
Effective Frequency (Hz)
2. Loose pipe supports

3. Impact against pipe restraints

PROPOSED SCREENING LIMITS
MAINLINE PIPING IN RANDOM VIBRATION
The following are additional situations that do not apply with the
assessment procedures: 10¹
B

Shell‐mode piping vibration

Allowable RMS Velocity (ips)

Transient vibration, such as shock and impact 10⁰

(e.g., water hammer)

Metal temperatures in the creep regime, or outside

the temperature applicability of the weld joint fatigue 10-1
curves

Tubing and non‐metallic systems

10-2
Expansion joint bellows, flexible hoses 10⁰ 10¹ 102

Effective Frequency (Hz)

Large ductwork, square or rectangular

Group A Random
Group A Random w/Valves
Group B Random
Group B Random w/Valves

Figure 15. Proposed Vibration Screening Limits for Mainline Piping

Periodic Vibration Limits (A) and 25-Year Random
Vibration Limits (B)

E²G | The Equity Engineering Group, Inc.

Spectrum-Based Procedure

STEP 1 – Measure the velocity spectrum at the locations of

highest observed vibration on the piping system. The spectrum
must be based on the full effective bandwidth of the signal.

STEP 2 – Obtain the RMS velocity of the signal, VRMS.

STEP 3 – Obtain the effective frequency, ƒₑ, per equation (2).

Alternatively, the effective frequency may be inferred from the
waveform for narrow-band random signals by determining the
frequency at which the dominant waveform peaks occur.

STEP 4 – Compute the allowable RMS velocity, VRMSa,

per the following:

Group A Periodic without Unsupported Valves: Group B Periodic without Unsupported Valves:
0.352 in/s RMS, fe < 35Hz
a a
VRMS = 10.1 in/s RMS VRMS = RMS log₁₀ ( ƒₑ) + 0.48017
0.0394 10 2.127612 in/s RMS, fe ≥ 35Hz

Group A Periodic with Unsupported Valves: Group B Periodic with Unsupported Valves:

0.707 in/s RMS, fe < 150Hz 0.25 in/s RMS, fe < 16.5Hz
a a
VRMS = log₁₀ ( ƒₑ) + 0.48017 VRMS = log₁₀ ( ƒₑ) + 0.48017

0.0394 10 in/s RMS, fe ≥ 150Hz 0.0394 10 2.127612 in/s RMS, fe ≥ 16.5Hz

2.127612

Group A Random without Unsupported Valves: Group B Random without Unsupported Valves:
0.352 in/s RMS, fe < 35Hz 0.14 in/s RMS, fe < 4.8Hz
a a
VRMS = log₁₀ ( ƒₑ) + 0.48017
VRMS = log₁₀ ( ƒₑ) + 0.48017
0.0394 10 2.127612 in/s RMS, fe ≥ 35Hz 0.0394 10 2.127612 in/s RMS, fe ≥ 4.8Hz

Group A Random with Unsupported Valves: Group B Random with Unsupported Valves:
0.22 in/s RMS, fe < 13Hz 0.1 in/s RMS, fe < 2.4Hz
a a
VRMS = log₁₀ ( ƒₑ) + 0.48017 VRMS = log₁₀ ( ƒₑ) + 0.48017
0.0394 10 2.127612 in/s RMS, fe ≥ 13Hz 0.0394 10 2.127612 in/s RMS, fe ≥ 2.4Hz

STEP 5 – The acceptance criteria is satisfied if VRMS < VRMS a.

E²G | The Equity Engineering Group, Inc.

References
[1] Energy Institute, Guidelines for the Avoidance of Vibration
Induced Fatigue Failure in Process Pipework, 2nd Edition,
London: Energy Institute, 2008.
[2] Preoperational and Initial Startup Vibration Testing of Nuclear
Power Plant Systems, New York, NY: ANSI/ASME, 1990.
[3] Reciprocating Compressors for Petroleum, Chemical, and
Gas Industry Services, Washington, D.C.: American Petroleum
Institute , 2007.
[4] Guide to Fatigue Design and Assessment of Steel Products,
The British Standards Institution, 2014.
[5] L. Martins, M. Pereira da Silva, R. Noronha and E. Tinoco, "Discussion
of Industrial Piping Vibration Failure Criteria," Ouro Preto, 2005.
[6] W. W. Nimitz, "Reliability and Performance Assurance in the
Design of Reciprocating Compressor and Pump Installations," 1974.
[7] S. R. Institute, "Controlling the Effects of Pulsation and Fluid
Transients in Piping Systems - Report No. 160," SGA-PCRC,
San Antonio, TX, 1979.
[8] D. Benasciutti, Fatigue Analysis of Random Loadings,
University of Ferrara, 2004.
[9] A. M. Y. K. M. M. S. Y. H. OHTA, "Fatigue Crack Propagation Curve
for Design of Welded Structures," Transactions of the Japan
Welding Society, vol. 20, no. 1, 1989.
[10] Specification for Unfired Fusion Welded Pressure Vessels,
The British Standards Institute, 2015.
[11] J. Wachel, "Piping Vibration and Stress," in Proceedings of the
Machinery Vibration Monitoring and Analysis, 1981.
[12] J. C. A. K. a. M. S. Wachel, "Piping Vibration Analysis,"
in Proceedings of the Nineteenth Turbo-machinery Symposium,,
College Station, TX, 1990.
[13] Boiler and Pressure Vessel Code, ASME, 2017.
[14] L. Breaux, S. McNeill and G. Szasz, "Fitness-for-Service
Assessment of Piping Subject to Random Vibration,"
Vancouver, BC, 2016.
[15] A. Paulin, "Stress Intensity Factor and K-Factor Alignment for MICHAEL F.P.
Metallic Pipes," ASME, New York, NY, 2014.
BIFANO, Ph.D., P.E.
[16] C. Sebastien, P. Yannick, M. Pierre and G. Michael, "A 12 mm/s Senior Engineer II
RMS Screening Vibration Velocity for Pipes using ANSI-OM3
Standard and Regulatory Design Rules," in ASME Pressure 216.386.0496
Vessels and Piping Division Confrence, Denver, CO, 2005.
[email protected]
[17] API, 2016. API 579-1/ASME FFS-1 Fitness-For-Service, 3rd Edition.
The American Petroleum Institute

ANTHONY J.
FELLER
Staff Engineer I

216.658.4790

[email protected]

E²G | The Equity Engineering Group, Inc.