Finite Element Analysis – Evolving Acceptance
File: 3783, Original Post: June 7 2012, Last Updated: May 22 2014, By: Laurence Brundrett
Update Aug 2015
This article was originally written in 2012 because it became very difficult to use FEA to register fittings and vessels. The primary location not accepting FEA was Alberta, but with designs needing to be registered Canada wide, its usefulness became limited. We could only use FEA to prove the cycle life of vessels, not the structural integrity.
Three years later we note a counter change where FEA is being accepted for not just the proof of cycle life but ALSO proof of structural integrity in both vessels and fittings. The primary catalyst is the change in a number of personnel at ABSA, along with ABSAs increasing use of reciprocal registration, whether based on code calculations or FEA. Even without reciprocal registration, ABSA has been accepting FEA to prove items not covered by code rules. See individual updates below.
Original article last update May 2014
A product was successfully registered across Canada using FEA as the design basis. One year later FEA was again used as the design basis to register additional similar designs. FEA was used in both cases on the grounds that code rules do not cover its design. This time FEA was not accepted as a design basis in 2 provinces. What changed?
Subject: Section VIII, Division 1 (1998 Edition, 1999 Addenda); U-2(d), UG-19(b) and (c), UCI-16, and UCD-16
Date Issued: December 5, 2000
Question: If there are no applicable Code rules in Section VIII, Division 1 for the design of a vessel or vessel part, may a finite element analysis be performed in lieu of a proof test?
Reply: Yes, see U-2(g).
New emerging criteria: code rules do not cover the design as manufactured; however, code rules can be made to apply if we pretend that material is removed from the product. In the above example, the green shape is the product as built. If we pretend that material is removed, the red shape emerges and the design can be analysed using standard code rules. (Because code rules can be applied this way U-2(g) is no longer necessary. By this logic FEA and Burst testing is not allowed.) Appendix Y can now be used on the flange without reinforcement and UG-34 for the now flat unreinforced head. The FEA rating was used in most provinces except two where a much lower calculated rating was used.
Update Aug 2015: With changing personnel at ABSA, this requirement has changed. There will always be times when weird things have to be done to a design to get a CRN, but this case was extreme, and now it is becoming rare.
Back to ASME:
Subject: Section VIII, Division 1 (1998 Edition, 1999 Addenda); UG-34
Date Issued: June 13, 2000
File: BC00-l55 [BC00-267]
Question: Does Section VIII, Division 1 permit the use of stiffeners in the design of unstayed flat circular heads, bolted covers, or flanges?
Reply: Yes, provided the requirements of U-2(g) are satisfied
Code interpretations cannot always be used in Canada to support your design case. They did not work this time. See below for comments on one province that does not always accept interpretations.
Aug 2013 Addenda:
When can Finite Element Analysis (FEA) be used? In our experience FEA can always be used to validate the cycle life of a product – in many cases this is the only way possible. Depending primarily on the reviewer, FEA can sometimes be used to justify the design of vessel or fitting features that are not covered by code books. This original post covered a case where the reviewer did not allow it for features not in the ASME code books. We also encounter cases where FEA is allowed, but its use is restricted to one time use only – no repeat production of identical units are allowed. The rejection of the use of FEA is on the rise – it is worth considering whether your product can be redesigned for the Canadian market to avoid the need to use FEA for other than cycle life purposes. If you must use FEA for pressure boundary design, try to get your jurisdiction to sign off on its use in advance. Update Aug 2015: we find that increasingly FEA is allowed to be used to prove that designs are acceptable when code passages covering the design are not available.
Will validation be required? In our experience, validation is never required if the FEA is only used to predict the cycle life. Some reviewers require validation if pressure boundary components are designed using FEA. Ideally, a CRN is applied for and issued prior to fabrication, but in this case the actual issuing would not happen until the reviewer accepts the validation results – after construction is complete. Not all reviewers require validation.
Update Aug 2015: Reviewers are rarely asking for validation, but it still happens and occasionally extreme validation is required: For a range of closures a jurisdiction asked for each size to be run by FEA (usually only some sizes representative of the range need to be run). One burst test must be done to validate a matching FEA run, and the FEA must predict the burst test pressure! So the requirements for using FEA are still not always consistent, but it is better than the alternative offered by the jurisdiction: burst test every size of closure!
Validation of a FEA model is done by strain gauge or displacement measurements. Testing is usually done during the hydro test at regular hydro test pressures – there is no need to destroy a product as you would with UG-101 burst testing. Additional FEA run(s) are required to mimic the hydro test conditions and predict strain or displacement results under hydro test.
There will be some variation between the FEA prediction and the test results. Our experience shows that locations with simple geometry like centers of heads and straight shells should provide strain gauge results within a few percent of the FEA predictions. Areas with complex geometry are extremely hard to gauge well and cannot be expected to match. Unfortunately, these are the areas that we are most interested in. There are no acceptance criteria on test results – you have to work out what is acceptable with your reviewer.