Professional Documents
Culture Documents
Determining Allowable Weld Flaw Sizes in Pulsation Bottles
Determining Allowable Weld Flaw Sizes in Pulsation Bottles
in pulsation bottles
By
The welds of pulsation bottles are commonly inspected to detect flaws that could lead to fatigue failures. A weld flaw is a site
where a fatigue crack will initiate in the weld. The dynamic loading from the normal compressor operation will cause a fatigue crack
to grow due to the cyclic loading. The crack may continue to grow since the dynamic loading is always present. The rate of crack
growth depends on many factors, of which the main one is the magnitude of the local dynamic stress. It may be the case that the weld
flaw size is small enough or the dynamic load magnitude is low enough for the crack growth threshold to be exceeded, and crack
growth leading to failure will never occur even when the flaw exists in an infinite loading cycle environment. However, if the dynamic
stress is high enough or the initial weld flaw is large enough, a fatigue crack will propagate and lead to a failure.
This paper will present a new approach to determine critical weld flaw sizes for head and shell nozzles of pulsation bottles.
The approach uses a combination of finite element analysis for determining dynamic stress magnitudes, and engineering critical
assessment methodologies to determine allowable stress magnitudes for different flaw geometries. The result of this work has led
to development of a criteria to determine if remedial actions are needed when weld flaws are detected for new or in-service
pulsation bottles.
This paper details an undertaking by Wood to develop weld flaw acceptability criteria for Williams, for a pulsation bottle weld
inspection program. Detailed analysis was performed on a variety of pulsation bottle nozzles in which the maximum allowable flaw
sizes were determined. Williams could then use these results as benchmarks for other nozzles with similar geometries. The objective
was to allow Williams to make more informed decisions on the necessity of weld flaw repair during its weld inspection program.
The data shown in this paper is only applicable to the specific nozzles analyzed, but the general approach can serve as a basis for
those designing their own weld flaw acceptability criteria.
Background
Williams had experienced fatigue failures in over 15 pulsation bottles associated with the reciprocating compressors at their facilities.
The failures have occurred at the welds between the nozzles and the shell of these bottles. An example of a typical failure is shown
in Figure 1. The bottles have a 30-year service life and had been in operation for a range of between 2 and 9 years when the failures
occurred. The shortened life of these bottles is a serious issue as it impacts the profitability and safety of these facilities.
An investigation into the causes of these failures was undertaken.
A review of the failed bottles determined that although the bottles were properly built and inspected to the proper industry standards,
autopsies of the failures revealed flaws in the welds that were not detected during normal inspections performed at the time of
construction. These weld flaws became the location where fatigue cracks initiate and propagate due to normal dynamic loading. Weld
flaws are relatively easy or reliable to detect in butt welds using a conventional x-ray technique. The problem with many pulsation
bottle nozzles is the overlay of welds at the nozzle, shell and repad that make weld flaw detection difficult.
The objective of the FEA was to input bottle geometry and vibration levels and output predicted dynamic stress levels at each weld.
The stress levels would then be used in the ECA to predict maximum allowable weld flaw sizes.
For each nozzle, a 3D finite element model was built. A modal analysis was conducted to determine the mechanical natural
frequencies for the primary vibratory mode shapes. The modal results include a normalized displacement as well as a normalized
stress for each mechanical natural frequency. These modal displacements (vibration) and stress results can be used to determine
a relationship between vibration and stress. For example, a dynamic stress of “A” psi in the nozzle is the result of “B” mils pk-pk
displacement at the center of the bottle shell. The modal results can be used to determine a normalized dynamic stress per unit
vibration (psi/mil displacement) of each weld. Assuming that the bottle vibration is equal to the Williams vibration guidelines as shown
in Table 2, the dynamic stress (psi pk) can then be calculated.
Assumptions
The finite element models were simplified to include a single nozzle and a partial model of the bottle shell as shown in Figure 4.
The model assumes the repads are in full contact with the shell. The repads are only welded to the shell at the inner and outer
diameter of the repad, and there may be a small gap between the repad and shell. The effect of the gap between the shell and
repad has been shown to have a minimal impact on the overall dynamic response of the bottle and nozzle from other work; therefore,
the simplified model used in this analysis is appropriate.
The bottle model will calculate several vibratory modes. The dynamic stress depends on which vibratory mode is being evaluated.
Each mode shape has a specific vibration measurement location used in the analysis. This location for extracting the normalized
vibration from the finite element model must remain consistent when measuring vibrations in the field during the weld inspection
program.
Typical mode shapes are shown in Figure 4. Mode 1 is commonly described as the shell rocking mode. It is commonly seen on
single-nozzle bottles, but is rare on multi-nozzle bottles. Mode 2 is the shell cantilever mode which is common on single and
multi-nozzle bottles. Mode 3 is the shell axial mode. It is rarely a problem in field applications as the frequency tends to be very high
where low excitation forces exist. Mode 4 is the head-bending mode where the nozzle flexes relative to the head. This mode requires
including a span of attached piping in the analysis to better estimate the frequency. The pipe span length is based on typical support
spacing requirements.
Mode 1 Mode 2
Mode 3 Mode 4
The locations where vibration results are extracted from the model are shown in Figure 5. Field vibration measurements must be
recorded in the same locations. The FEA assumes the nozzle has zero vibration for Modes 1 to 3, and that the end of the shell has zero
vibration for Mode 4. It may be necessary to measure relative vibration between the shell and nozzles, and then subtract the two
vibration measurements to obtain the relative measurement for comparison to the model results.
Stress calculation
The dynamic stress is extracted from the finite element model using the hot spot stress method as described in
BS EN 13445-3:2009 [1]. The hot spot stress method is an extrapolation technique used to better estimate the stress at a singularity,
that is, an artificial stress riser in the model due to the simplified geometry. The hot spot stress method requires an element mesh size
relative to the material thickness in the area where stress is of interest, to extract stress results at specific points along a stress path.
For example, the stress at 0.4e and 0.6e as shown in Figure 6, are used to estimate the hot spot stress.
Thickness decreases
beyond repad
There are many variables to consider in this analysis, such as bottle geometries, weld locations, flaw locations, vibration levels,
and vibratory mode shapes. It was impractical to model every permutation of these variables. As the FEA progressed, it became
clear that modeling certain scenarios was not necessary:
On shell nozzles, the nozzle axial mode is rarely observed in the field. Therefore, normalized stresses were calculated only for
the rocking and cantilever modes
Cantilever modes had consistently higher stresses than rocking modes. For shell nozzles, only cantilever stress results were
considered for the ECA. Head nozzles only have one mode due to the axial symmetry of the model.
The inner-nozzle weld consistently had the lowest stresses
Analysis of the Williams’ Marginal and Correction vibration guidelines were found to be most appropriate. There is no benefit
in analyzing the Danger guideline since immediate action is indicated by the Correction criterion.
Typical normalized stress extrapolation results are shown in Figure 8 to Figure 10 for nozzle 1, as per Table 1. Each figure shows the
stress varying along a path from the weld flaw location at 0” to a distance away from the flow that is required to extract the stress
results. The blue bars represent the stress calculated along the path by the finite element program. The purple vertical bars represent
the stress amplitudes used to calculate the hot spot stress. This hot spot stress and the vibration is used to calculate the normalized
stress. The normalized stresses were used to calculate the predicted stresses at each vibration level as shown in Table 3. The key stress
results from the FEA that are used in the ECA are highlighted in Table 3.
Analysis of welds to determine the allowable flaw sizes via a fracture-mechanics approach is termed an ECA. An ECA considers
fatigue-crack growth under fatigue loading, and fracture and/or plastic collapse check under maximum principal stresses. However,
based on the type of fatigue loading (with number of cycles >10^10) experienced by these welds, fatigue-crack growth is the
expected governing mode of failure rather than the failure under high tensile loading. Thus, it will be sufficient to perform a
fatigue-only assessment via a fracture-mechanics approach.
Assessment Approach
A general ECA approach includes a combined fracture and fatigue assessment to account for fatigue-crack growth and failure of crack
due to fracture stress. The approach is shown in the flowchart in Figure 11. However, for this study, a fatigue-only assessment is
appropriate to provide the required results in terms of critical stress range threshold. The flowchart for the fatigue-only assessment
is shown in Figure 12.
1) The ECA is a fatigue-only ECA. The In-Air Fatigue Crack Growth Rate Curves provided in BS 7910:2019 [2] are used to
calculate crack growth.
2) A cycle count of 10^10 cycles/year is considered for the stress range threshold. With this high cycle count, any stress
range above the threshold causes the initial flaw to grow to failure within a small fraction of a year. Similarly, any stress
range below the threshold does not grow the flaw and predicts infinite life.
3) Appropriate stress intensity factor solutions are considered for each flaw based on its location on a particular weld at
the nozzle-shell interface. The effect of stress-range magnification due to the presence of weld toe is also considered
in the assessment.
4) The geometric stress concentration effects are not considered in this ECA to determine the threshold. The effects are
considered in the actual stress ranges obtained from the finite element analysis.
A range of preselected flaw sizes are considered as an input for this ECA to perform the fatigue crack growth assessment. The flaw
sizes selected for this study are provided in Table 4. It should be noted that these flaw sizes do not account for the non-destructive
evaluation (NDE) equipment sizing errors. The flaw sizes are the raw flaw sizes based on the results from this analysis. These results
need to be modified or interpreted by considering the applicable NDE sizing error. For example, the threshold stress range may result
in a critical flaw size of a 3mm x 9mm raw flaw size based on the raw flaw size of 2mm x 9mm determined by this analysis,
plus 1mm in flaw height due to the NDE equipment error.
Case # Flaw length (mm) Flaw height (mm) Aspect ratio (height/length)
1 30 5 1/6
2 25 5 1/5
3 20 5 1/4
4 15 5 1/3
5 24 4 1/6
6 20 4 1/5
7 16 4 1/4
8 12 4 1/3
9 18 3 1/6
10 15 3 1/5
11 12 3 1/4
12 9 3 1/3
13 12 2 1/6
14 10 2 1/5
15 8 2 1/4
16 6 2 1/3
17 6 1 1/6
18 5 1 1/5
19 4 1 1/4
20 3 1 1/3
Results
The combined results from the FEA and ECA are shown in the example chart, Figure 13. The flaw sizes are shown as the blue and
red vertical bars. The ratio of the stress range versus the fatigue limit is represented by the green and purple area of the chart.
The boundary between these two areas is the maximum allowable stress from the ECA. The horizontal red and orange lines represent
the stress ratio determined by the FEA based on the Marginal and Correction vibration levels. The intersection boundary between the
green and purple areas from the ECA with the horizontal red and orange lines represents the boundary where a flaw size will be
acceptable and unacceptable. As shown by the example of Figure 13, the results from the ECA and FEA indicate that flaw configuration
17 to 20 will be acceptable based on the bottle having vibration equal to the Correction criterion. Figure 13 shows that the stress
range based on the Marginal vibration criterion is less than the allowable stress range. This means that all weld flaw configuration
considered in this analysis will be acceptable if the bottle vibration is not higher than the Marginal criterion.
Maximum allowable
Stress range
stress range from ECA
from FEA
Charts similar to Figure 13 were created for all nozzles and weld locations. Example charts for one nozzle configuration are shown in
Figure 14 to Figure 17. The charts show there are significant differences between the critical weld flaw size and vibration levels for the
different weld and flaw locations.
Figure 15: Allowable flaw sizes for surface flaw in nozzle wall at Weld 1
Figure 17: Allowable flaw sizes for surface flaw in reinforcing pad to weld interface at Weld 1
Phased array testing was done for other bottles in service. Flaws were found in some bottles and tracked over time. The size of the
potential flaws did not increase over an extended time. Ultimately, a basin-wide design survey, combined with effective remediation of
outside forces has been shown to be a reliable means of preventing failures and managing the risk.
One shortcoming of the analysis approach described in this project is that the critical weld flaw size is based on assumed vibration
levels. Vibration levels will change with compressor operating conditions, speed, and load step. The system may experience higher
vibration than the levels assumed in this analysis, which could result in fatigue-crack growth. Also, other factors such as misalignment
and excessive pipe strain will cause increased vibration levels as well as higher static stress. Vibration must be monitored on a regular
basis and pipe strain must be reduced as much as possible, for this approach for analysis of critical weld flaw size to be an effective
method of managing the risk of fatigue failures.
Conclusions
A combination of remediating known design deficiencies through repairs and targeted design standards, paired with effective
remediation of outside forces has been shown to be a cost-effective and reliable means of preventing failures and managing overall
risk. The methods described in this paper are an effective means of analyzing at-risk bottles with known design or fabrication
deficiencies. The potential for phased array testing to provide a false-positive flaw detection, combined with the high cost of the
testing itself, may prevent this from becoming a cost-effective solution to managing the risks for a large fleet. The use of ECA, FEA and
phased array testing may be an effective approach to specific component failure investigations.
References
1. BS EN 13445-3:2009, BSI British Standards, Unfired pressure vessels – Part 3: Design.
2. BS 7910:2019, BSI British Standards, Guide to methods for assessing the acceptability of flaws in metallic structures.