Nov-2022

Life or death of the PSA

Weld peaking or out-of-roundness affecting PSAs is evaluated, highlighting the value of conducting peaking measurements to determine a PSA’s inherent stress concentration.

Annette Karstensen
Becht Inc

Article Summary

Can the life of a pressure swing absorber (PSA) be extended beyond its intended design life, and how do you manage the PSA’s ‘midlife crisis’? PSAs are typically designed for a 20 to 30-year life with pressure cycling in a hydrogen environment. What can possibly go wrong? Quite a lot! But with appropriate integrity management and inspection strategy in place, a lot can go right too, and it is not uncommon that these vessels can survive well in advance of the original design life.

By the time the intended design life is approaching, many operators are challenged with having to make decisions regarding life extension or purchasing new vessels. While it is known that some PSA vessels do crack, the work described in this context highlights how it is possible to potentially extend life by developing safe inspection intervals based on a fracture mechanics approach.

In many cases, towards the end of design life or during midlife (50-60% of design life), operators elect to conduct a life extension study that includes a Level 3 Fitness for Service assessment and Finite Element Analysis (FEA) to evaluate the areas of high stress where fatigue cracks might initiate. The outcome from the FEA is used to develop an inspection strategy. In most cases, it will show that the areas of high stress are around the inlet and outlet nozzles. However, most PSA vessels are well fabricated with nozzles or manways having integral reinforcement and blended nozzle welds, making these areas less susceptible to crack initiation.

In Becht’s experience, the real integrity concern is weld peaking in the long seams and/or weld misalignment, and this will not be highlighted by an FEA made from design drawings. Other potential crack locations are at internal attachment welds, for example, where the inlet screen is welded to the bottom head. These locations have also been known to develop fatigue cracks in the past and need to be considered when developing an inspection plan for these vessels.

Local or global peaking
API 579¹ recommends measuring the peaking using a template across the seam weld. This will result in measuring the deformation locally across the longitudinal seam weld to obtain the local peaking. However, in Becht’s experience, the deformation often seen in the PSA can vary between local and global peaking, and calculating the associated stress will yield significantly different results. Local peaking is well captured by using a template across the seam weld; however, ‘global peaking’ or out-of-roundness will not be captured by this method.

The issue with peaking or out-of-roundness is the inherent bending stress that occurs due to the force misalignment. This bending stress needs to be considered in addition to the membrane stress when completing an integrity assessment. API 579 provides analytical closed-form solutions to calculate the bending component as a function of the peaking height based on local or global peaking assumptions. However, the difference in the calculated bending stress can be significantly different, depending on whether you are using a local or global peaking model, and consequently, the calculated fatigue life can be misleading. The difference in the local and global deformation as copied from API 579 is shown in Figure 1.

The deformation behaviour is most accurately captured by 3D laser scanning and imported into an FEA model to determine the associated through-wall stress for the fatigue analysis. The displacement of a PSA shell subjected to a laser scan and consequently analysed using an FEA is shown in Figure 2.

In this case, it is interesting to note that if a template had been used, the value of δ (see Figure 1) had been measured to approximately 3.5 mm. However, in this case, when looking at the laser scan data, the deformation more resembles the global ovality. This, in turn, resulted in stresses across the seam weld that were more severe than those measured if a template had been used. A comparison is shown in Table 1. Rb given in Table 1 is the membrane to bending ratio that needs to be considered in the fatigue assessment.

The deformation from the original fabrication shown in Figure 2 is only one example. There are examples where peaking occurs locally across the seam weld, which can be severe and cause failures due to fatigue cracking.⁵

In most cases, the issue with peaking is not taken directly into consideration during the design phase. In some cases, the peaking or deformation will be life-limiting for the PSA vessels. Therefore, it is recommended that a midlife assessment is carried out to evaluate the consumed life fraction based on laser scan data. Take, for example, the PSA subjected to laser scanning shown in Figure 2, with the bending stress of 69 MPa, assuming that the vessels were designed for 20 years (or 700,000 pressure cycles). Currently, it has been in operation for eight years and has been exposed to 280,000 pressure cycles. Therefore, the next step is to use an appropriate fatigue curve to determine the life fraction based on the stress defined by the FEA.

BS 7910: 2019 (Section 8.8.1) offers a fatigue assessment for welded joints with misalignment based on the stress concentration factor caused by the misalignment. A joint quality category equivalent to those given in BS 7608 is derived based on the additional bending stress from the misalignment. The results are summarised in Table 2. The quality curves in BS 7910 provide design life (mean plus two standard deviations). The fatigue life for mean plus 2 SD has been compared with that estimated using mean fatigue data. However, it should be noted that since the fatigue curves are developed for cyclic behaviour in air, it is recommended not to rely on mean properties as there is no account for the hydrogen environment.

In summary, for this case, the most severe stress range due to deformation would not have been detected if a template across the seam weld had been used.

Developing the life management strategy for PSA vessels using fracture mechanics
The fatigue performance and crack growth rate for a PSA depend on several factors:
- Hydrogen partial pressure
- Frequency
- Stress intensity factor range, ΔK, which is a function of crack size and stress level
- Load ratio (Min/Max)
- Material (base metal, weld or HAZ).

Several published papers have investigated the effects of hydrogen on fatigue crack growth. However, very few have managed to simulate the service conditions of PSA vessels. The effect of hydrogen is hugely important on the fatigue crack growth rate, and it is easy to become overly conservative when conducting the fatigue crack growth analysis. This can result in determining a remaining life that does not represent experience seen in service and unmanageable short inspection intervals.

Published fatigue crack growth rates⁴ were for tests conducted close to the operating conditions for many of the PSA vessels in operation in terms of hydrogen pressure, cycling frequency, and load ratio. Therefore, these crack growth rates have been used in this analysis.
The inspection and life management strategy should consider the potential that a fatigue crack can initiate due to excessive weld peaking. Therefore, it is recommended to establish an inspection interval based on the time it takes for the largest crack that could be missed during an inspection to grow to a limiting crack size.

While limiting crack sizes should be calculated employing procedures either in Part 9 of API 579 or BS 7910, in most cases for the seam welds, the limiting depth of the crack will be 80% through-wall, which is the limit depth for a surface breaking flaw in accordance with API 579. The reason the cracks will reach 80% is because the vessels are subjected to post-weld heat treatment, so welding residual stress does not contribute greatly to the fracture ratio.