Comparing Approaches to Design Fatigue Assessment

TWI Ltd,
Cambridge, UK

Proc of European Symposium on Pressure Equipment (ESOPE) 2013, 8-10 October, Paris, France

Abstract

The paper presents an assessment of the design fatigue limits of a pressure vessel that has been in service for hydrocarbon gas separation on an offshore installation. The pressure vessel was designed to the BS 1515 standard that is now superceded. The fatigue limits of welded regions of the vessel have been calculated using the modern design methods of EN 13445 and PD 5500 based on equivalent stress and maximum principal stress, considering both surface extrapolation and through wall linearisation to the fatigue hot spot.

The fatigue life of the different regions analysed are compared, and the effect of the treatment of stress on the fatigue life is illustrated. The paper discusses an approach that may be taken to extend the fatigue life of pressure vessels designed to old standards that remain in service.

Introduction

This paper presents an assessment of the fatigue life of a condensate flash pressure vessel that has been in service on an offshore installation. The vessel dates from the 1970’s and was designed to British Standard 1515. This standard pre-dates modern design assessment for fatigue as is found in the current pressure vessel construction standards PD 5500 [1], EN 13445 [2] and ASME VIIII [3].

The assessment provides an opportunity to make a comparison between the fatigue lives predicted using modern design methods. There are differences in approach and treatment of stresses between PD 5500 and EN 13445 yet the effect of these differences on the fatigue life for a given weld are not well understood. Since the condensate flash vessel contains design features that are typical for many pressure vessels it enables a comparison of the fatigue lives for different features to be made.

There is an increasing need to justify life extension beyond the specified design life of pressure vessels designed to older codes and standards within process plant on offshore installations. Many such vessels have been in service for over 25 years, yet are required for longer periods until economic field life is exhausted. Many of these vessels contain condensate liquefied hydrocarbon gases at low temperatures where the risk of corrosion is small, so the main constraint to life extension is the remaining fatigue life.

While replacement of vessels may be a possible option, this is more difficult and costly on an offshore installation where space is limited and logistics are more complex than would be on-shore. In some cases the length of future service required may not be long enough to justify the costs of replacement. Therefore offshore operators are seeking a methodology for justifying life extension for vessels designed to older codes through fatigue analysis and a campaign of risk based inspection.

Description of the vessel

The vessel that is the subject of the study is shown in Figure 1 is typical of many pressure vessels in service in the oil and gas sector. Its purpose is to separate methane from higher hydrocarbons that are in a condensed liquefied state. Pressurised liquefied condensate at low temperature flows through an inlet nozzle into the vessel where dissolved methane gas evaporates from the remaining liquefied condensate. Gas exits through an outlet nozzle at the top of the vessel while the remaining condensate liquid leaves the vessel through an outlet nozzle near its base.

The vessel comprises a cylindrical shell 1879mm internal diameter, 4.7m long and 25mm thick, with ellipsoidal heads at each end. One end contains a large manway that is set through the head, which is reinforced by an external reinforcement plate fillet welded to the manway and to the head. The condensate inlet nozzle and gas and condensate outlet nozzles are also set through the shell, while the safety valve and smaller instrumentation nozzles are of different designs.

The base of the vessel is supported on saddles with reinforcing plates. At the top of the heads are lifting lugs attached to reinforcing plates. There are also some external and internal attachments for securing insulation, impingement of the inlet flow, an anti-wave baffle and demister for removal of liquid from the outlet gas stream that are welded to the inside or outside of the pressure shell.

The vessel is constructed from pressure vessel carbon steel according to BS 1501-213-32A LT15 with a low temperature designation. Welding and manufacturing inspections were according to the techniques and standards of era of construction at the beginning of the 1970’s. Demonstrating a high quality of manufacture is a key element in building a case for life extension.

The vessel has a design pressure of 49bar and a minimum design temperature of -23°C. Cyclic stresses arise from pressure changes from atmospheric during start-up and shut down cycles and pressure transients during operation. In addition there is the potential for cyclic stresses arising from thermal expansion and contraction of attached pipework, thermal expansion and contraction of the vessel while fixed at the saddle supports, temperature differences in the inlet nozzle and reinforcing plate regions as cold condensate enters the vessel during start-up from ambient conditions, changes in weight of the volume of condensate held in the vessel, and flow induced vibration of the deflector plate and anti-wave baffle. 

In this paper, design fatigue limits for pressure cycling have been determined for the longitudinal seam welds in the shell, the welds connecting the heads to the shell near the knuckles and the toe of the manway reinforcing plate fillet weld. Results for other features will be reported at a later stage. Since the purpose of the paper is to illustrate and compare differences between approaches to fatigue design assessment, allowable cycles are determined as a function of the pressure range.

The operating history of such vessels is generally complex and often not known with any certainty. Operators normally have to make a conservative estimate of the previous duty of the basis of their experience of plant operations. It is therefore useful to have a parametric fatigue assessment as a function of loads of varying range from which operators can build up a Miners usage factor summation to estimate fatigue life.

Fatigue assessment in modern standards

Modern pressure vessel construction standards contain provisions for fatigue assessment of parent material and welds. PD 5500 (Annex C) and EN 13445 (Clause 18) contain these provisions for vessels designed to these standards. The principles of these methods are well established [4], but a summary of the approach for welds is given here to assist the reader in conjunction with this paper.

PD 5500 and EN 13445 are both based on a scheme of classifying welds of different geometric types and quality according to their fatigue behaviour. Fatigue design curves are provided for different weld classifications in terms of the nominal elastic stress range and number of allowable cycles. Users determine the number of allowable cycles for a given stress range, and compute a usage factor as a Miner’s law summation of the ratio of the actual and allowable cycles for different stress ranges.

Differences between PD 5500 and EN 13445 arise in the stress range that is to be used for fatigue assessment in a multi-axial stress system. PD 5500 is based on the range of the maximum principal stress only. In contrast, the main approach advocated in EN 13445 is the use of either a Tresca or von Mises equivalent stress range, according to established German practice; the use of maximum principal stress range is also permitted but specified in Annex P.

The effect of structural discontinuities such as a nozzle or reinforcement plate is taken into account by extrapolating the stress in the vicinity of the discontinuity to the fatigue hot-spot at the toe of the weld. A linear or quadratic surface extrapolation of the surface stress at two or three defined positions close to the weld toe is normally used. With the availability of modern three dimensional finite element analysis, other approaches for determining the structural hot-spot stress have developed by extrapolating the stress variation through thickness.

Stress analysis

A simplified, axisymmetric finite element model of the manway end of the vessel was generated in Abaqus/CAE version 6.12-1, and solved using Abaqus/Standard version 6.12‑1. The dimensions of the model were sourced from the original engineering drawing of the vessel where possible. However, as the drawing was 40 years old, not all of the dimensions were legible, as is commonly found, so some dimensions were taken from measurements made on the vessel directly.

The axisymmetric model can be seen in Figure 2. The manway has a thickness of 25mm and is set-through the semi-elliptical head. It has a compensation plate of 25mm thickness, which is welded to the head and to the manway itself. The head is welded to the main shell of the vessel close to the knuckle, which is illustrated in Figure 2.

Measurements on the vessel showed that the thickness at the knuckle varied from 25mm in the shell, to 29mm in the head near the knuckle, and 27mm in the head away from the knuckle. This was reproduced in the model. The end of the vessel was modelled up to the first saddle support where deformation is locally constrained. Non axisymmetric features such as the inlet nozzle, lifting lugs and other smaller nozzles were omitted in this simplified model.

The interface between the compensation plate and the vessel head is unfused. The unfused interface extends from the manway to head weld to the fillet weld at the end of the compensation plate, as shown in Figure 3. In order to define this region in the model, two parts were made (one above and one below this line), which were then tied together at the regions where the welds would be located. The unfused land was then defined as a contact region, where hard, frictionless contact was defined.

Pressure of 1bar was applied to the internal surface so that the results could be easily scaled to evaluate the stresses for different applied pressure ranges in a parametric study. This is applicable because linear geometry, small strain assumptions were implemented. Elastic material properties of Young’s modulus 207GPa and Poisson’s ratio 0.3 were used. The end of the vessel was fixed in the y-direction to simulate the constraint at the saddle supports. This was sufficiently far away from the knuckle region for there to be no end effects.

Stresses were extracted from the model at three welded areas of interest: the weld toe between the compensation plate and the head, the weld at the knuckle between the head and the shell, and the seam weld. Seven weld locations were investigated shown in Figure 2. The von Mises, Tresca and maximum principal stresses determined at these locations are summarised in Table 1.

The Tresca stresses are generally the highest and the von Mises stresses are the lowest. The maximum principal stresses are similar to the Tresca except in regions of high shear where Tresca dominates. Note that on the head side of the knuckle weld, the maximum principal stress is zero at the outside free surface because this is a region in compression. This can be seen in Figure 4.

The compensation plate weld toe shows the highest stress of the locations investigated. Here, structural hot spot stresses (SHSS) were determined using different methods. The SHSS is a measure of the stress concentration. This can be difficult to calculate using FEA because a singularity exists at sharp corners, such as the weld toe. Therefore SHSS is calculated from the stresses close to but not at the weld toe.

Table 2 compares the results from using three different methods for calculating SHSS: surface stress extrapolation (SSE) [5], through-thickness integration (TTI) and nodal force (NF). SSE linearly extrapolates the stresses at positions of 0.4t and 1.0t (where t is the thickness) from the weld toe. The TTI method uses through-thickness stresses to calculate a force per unit length of weld and moment per unit length of weld which can be used to calculate the structural stresses. The NF method [6] similarly calculates a force and moment per unit length of weld using nodal forces directly from the FEA, which are a more fundamental quantity. The TTI and NF stresses are calculated from stress and force components normal to a plane directly beneath the weld toe, as illustrated in Figure 3.

There is some variation in the SHSS stress at the weld toe depending which calculation method is used and stress (von Mises, Tresca or maximum principal) is used. The highest calculated SHSS is from TTI, which is similar to SSE based on Tresca or maximum principal. SHSS from Nodal force is approximately 10% lower than these results and SSE of von Mises the lowest of all. While these differences may not matter for high stress cycles, they could be significant for low cycle fatigue life.

The results were used to carry out a parametric study on the fatigue life at the welds locations due to cyclic changes in internal pressure.

Table 1 Stresses at 7 vessel locations (labelled in Figure 2) for 1 bar internal pressure

Table 2 Hot spot stresses at the compensation plate weld toe calculated using different methods for 1 bar internal pressure

Fatigue analysis

The fatigue lives of the three welded areas described above were assessed using the design methods of PD 5500: 2009 Annex C which represent a probability of failure of approximately 2.3% and BS EN 13445-3: 2009 Clause 18 and Annex P which represent a probability of failure of approximately 0.14%. The allowable fatigue cycles of each weld were determined as a function of pressure range using the Class of S-N curve appropriate to the weld detail indicated in Table C.2 of PD 5500 and Table 18.4 and the Table of Annex P of EN 13445. Assessments to EN 13445 assumed the vessel was constructed to Testing Group 1 and 2.

A thickness effect was taken into account for sections where the material thickness e > 22mm (PD 5500) and e > 25mm (EN 13445). The stress ranges obtained from the finite element model were divided by the factor (22/e)^1/4 and (25/e)^1/4 respectively. In practice this correction was very small.

Longitudinal seam weld

The weld detail was a full penetration butt weld made from both sides and free from significant defects. With respect to the seam welds in the shell, the lowest fatigue life would be from weld toe cracking at the longitudinal seam under the action of the hoop stress. The calculated stresses were higher on the inside of the vessel therefore the fatigue analysis was performed at this location. The effect of misalignment was not included in the fatigue analysis at this stage. The fatigue assessment was performed using both Class D and Class E curves according PD 5500 (as no information was available concerning the overfill profile) and the Class 80 curve according to EN 13445.

The design fatigue limits for the seam weld at the inside location according to PD 5500 and EN 13445 are shown in Figure 5 in terms of the number of allowable cycles as a function of the pressure range. Fatigue analysis performed using von Mises equivalent stress range (EN 13445) produced a higher fatigue limit than the use of Tresca equivalent stress range (EN 13445) and maximum principal stress range (PD 5500, EN13445). The design fatigue limit based on von Mises equivalent stress range with the Class 80 curve (EN 13445) exceeds that based on the Tresca equivalent stress range with Class 80 curve (EN 13445) and that based on the maximum principal stress range and the Class E curve (PD 5500) by a factor of 1.5 on cycles.

The design fatigue limits from using the Class D curve of PD 5500 and maximum principal stress range are very close to that from using the Class 80 curve of EN 13445 and von Mises equivalent stress. The fatigue limit based on the maximum principal stress range of PD 5500 and EN 13445 are almost identical due to the closeness of the Class D and Class 80 curves and the effect of the plate thickness correction used by each standard.

Knuckle weld

The weld detail was considered to be a full penetration butt weld made from both sides and free from significant defects. Determination of the stress in the hoop and axial direction revealed that the lowest fatigue limit would be from weld toe cracking at the knuckle under the action of the axial stress corresponding to a Class D fatigue strength according PD 5500 and Class 80 fatigue strength according EN 13445. The overfill profile was considered to be ≥ 150°. For cracking across the weld due to hoop stress, Class D and Class 80 fatigue curves would be assumed according to PD5500 and EN 13445. However, the hoop stress was found to be lower than the axial stress at this location.

The stress calculations revealed that the knuckle weld toe on the head side at the inside location of the vessel was the most susceptible region for fatigue cracking. The assessment was then performed at this location. As for the seam welds, the effect of misalignment was not included at this stage.

Figure 6 shows that the fatigue analysis performed using maximum principal stress range and the Class D curve (PD 5500) gives the highest fatigue limit and exceeds the lowest based on Tresca equivalent stress range and Class 80 curve by a factor of 1.5 on cycles.

Compensation plate weld

The weld detail considered was cracking from the toe of the fillet weld into the head. This detail has a nominal Class F fatigue strength according to PD 5500, and a Class 63 based on the maximum principal stress range and a Class 63 or 32 based on the equivalent stress range according EN 13445. The hot spot stress range at the weld toe was determined using surface extrapolation of maximum principal, Tresca and von Mises stresses. Since the hot spot stress incorporates the stress concentration due to the structural discontinuity, PD 5500 allows the use of the Class E curve with the maximum principal hot spot stress range. As no information was available on the size of the weld throat, both EN 13445 Class 63 and 32 curves were used for the fatigue assessment based on the equivalent stress range.

The results (Figure 7) showed that the fatigue analysis performed using the maximum principal stress range (Class E) according to PD 5500 gave the highest design fatigue limit. Use of the Class 32 curve and the Tresca equivalent stress range gave the lowest fatigue limit. The fatigue limit from using the Class E (PD 5500) curve and the maximum principal stress range exceeds that from using the Class 63 (EN 13445) curve and either the Tresca equivalent stress range or the maximum principal stress range by a factor of approximately 1.9 on cycles.

Comparison of hot spot stress approaches

Figure 8 compares the fatigue limits predicted using three different methods for calculating the structural hot spot stress at the compensation plate weld toe. The assessment was based on using the PD 5500 Class E curve. The results showed that the nodal force (NF) method gives better design fatigue life than the surface stress extrapolation (SSE) and through wall integration (TTI) which gave similar fatigue behaviour.

Comparison of the three weld details

The comparison of the design fatigue limits for the seam weld, knuckle weld and compensation plate weld toe using the maximum principal stress approach of EN 13445 shows that the compensation plate weld toe has the lowest fatigue limit. This is illustrated in Figure 9.

Comparing the fatigue analyses, we observe that PD 5500 can give a higher fatigue limit than EN 13445 for the same weld detail where it allows a different classification. We also observe that the fatigue analyses using Tresca equivalent stress range and maximum principal stress range are more conservative than using von Mises equivalent stress range for the same weld Class.

Life extension

Many pressure vessels have a nominal original design life of 25 years. This is the period which procurers often give to designers on which to make adequate allowance for corrosion and fatigue. It also reflects the typical service life and historical experience base of pressure vessels and can also be the period over which the financial investment is considered depreciated.

The achievement of the original design life does not mean that the vessel is no longer fit for service. The design assumptions and assessment may have been conservative with respect to the duty actually experienced and further safe service may be possible. It is, however, a moment to reassess the vessel, particularly if significant further service is required.

Operators seeking to extend the life of vessels designed to outdated standards are advised to identify and evaluate any features that would not be compliant with modern standards. The scope to extend life then depends largely on the quality of the original manufacture in terms of the number and size of fabrication flaws and material properties, the current condition and any service induced damage or degradation, and the service experienced. In short, fatigue life extension depends on showing:

• That the quality of manufacture is good in terms of the number and size of fabrication flaws,
• That any in-service degradation is not significant for the future life required,
• That any non-compliances with modern design standards are not significant for safety,
• That there is adequate toughness of the parent and weld materials for defect tolerance,
• That the calculated remaining fatigue life based on a crack growth/engineering critical assessment with an appropriate defect size is sufficient for the future service life required.

Conclusions

For the example of the pressure vessel analysed: it has been shown that

• The compensation plate fillet weld toe has a lower design fatigue limit than the full penetration seam or knuckle welds.
• The fatigue assessments based on Tresca equivalent stress and maximum principal stress are more conservative than based on von Mises equivalent stress for the same weld Class.
• The fatigue assessments based on the structural hot spot stress determined by through wall integration or surface extrapolation are more conservative than that based on nodal force equilibrium.
• Fatigue assessments made to EN 13445 and PD 5500 are similar except where the standards differ in the classification of a weld detail.

Acknowledgements

The authors wish to acknowledge the support of TWI’s Industrial Member’s Core Research Programme, Professors Donald McKenzie and David Nash of the Mechanics Research Group of the University of Strathclyde’s Department of Mechanical and Aerospace Engineering, and the oil company concerned in the production of this work.

References

• BS PD 5500: 2012, Specification for unfired fusion welded pressure vessels, British Standards Institute, 2012
• BS EN 13445: 2009: Specification for unfired pressure vessels, British Standards Institute, 2009.
• ASME VIII Division 2: 2010, Rules for the construction of pressure vessels, American Society of Mechanical Engineers, 2010.

• Wintle JB., Pressure Systems Casebook – Causes and avoidance of failures and defects, Published by Professional Engineering Publishing, ISBN 1 86058 421 7, 2004.
• Niemi E, Fricke W and Maddox S J, 2006: ‘Fatigue analysis of welded components - Designers guide to structural hot spot stress approach’. IIW, Woodhead Publishing.
• Dong P., 'Recommended structural stress procedure for fatigue assessment', OMAE Specialty Conference on Integrity of Floating Production, Storage& Offloading (FPSO) Systems, August 30 - September 2, 2004, paper OMAE-FPSO’04-0029