Linking Testing and Analysis

Editor’s Note: Tony Abbey teaches live NAFEMS FEA classes in the US, Europe and Asia. He also teaches NAFEMS e-learning classes globally. Contact [email protected] for details.

I wrote an article in DE a while back discussing the background to verification and validation. There was some criticism of this article from those who thought I had not defined the process correctly. Verification and validation are attempts to describe a process that is evolving from a basis of well-understood checking procedures. It is not a hard science as such, so there are no rules to break!

I’m going to brace myself for further feedback; this month’s article discusses validation, focusing on the link between test and analysis. In a nutshell, can we prove a FEA (finite element analysis) simulation is correct by validating it against a physical test or, perhaps better still, real-world evidence.

Limitations of Validation

Should we fully validate every analysis? We cannot, because we don’t have the time or resource to set up an independent test for every structure we are considering. Validation really becomes a broad description of any data, information or thought process that helps confirm at least some of the behavior of the FEA simulation. This validation evidence must be external to the FEA process. We are bringing something fresh to the table, other than the mathematics of the simulation.

If we could carry out a test to validate our analysis, is the test the definitive answer? There is a tendency to believe so; test is real, analysis is virtual. However, the most challenging aspects of simulation are loads and boundary conditions. In many cases, test setups also struggle to model real-world representations of these.

Working Together

Validation is a cross-discipline approach. The analysis and test teams should work closely together. For example, if part of a structure consists of a beam supported at each end, then there may be questions about how rigidly each end is supported. The analysis may have assumed a fully built-in end. Test evidence may show some flexibility at each support. Strain gaging close to the supported ends can show the actual bending moment that the test structure is seeing. From this, engineers can make an estimate of the test edge flexibility and re-run the analysis. At this stage, everybody can sit back and question whether the correlated test and analysis results now match the real-world structure!

The most important aspect of the modern verification and validation process is to ensure that the analysis and test planning are coordinated. In the example, this means the strain gaging is deliberately planned to validate the boundary condition assumptions. Other strain gaging can be in place to validate any redundant load path assumptions around the supporting fixture. The overarching validation can be based on mid-span deflection. Traditionally, testing has been associated mainly with the big requirements, such as the latter. Clearly, a deeper and more integrated approach can give much more valuable design information.

Informal and Formal Approaches

In practice, the modern approach has been used in many industries for years. I saw this firsthand in the ’70s, on combat aircraft structures. However, much of the cross-discipline coordination was done on an ad-hoc basis, working within overall plans. Today that process is formally defined.

This early experience taught me the importance of discussing and observing as much test evidence as possible informally. Valuable data can sometimes be gleaned from the informal approach. For example, an anti-tank weapon was designed with an electronic “brain” at the rear. Sufficient time had to elapse after impact, so that the brain could decide how to respond. The intermediate structure was a set of long spindly legs. The FEA simulation attempted to show that the legs would survive long enough for the message to get through. However, no specific validation of my results was available. Chatting about this with a test range engineer months later, I learned they had collected some fragments over the test series! Enough of one legs had survived to show the failure mode was almost exactly what the analysis had predicted. The survival time predictions were properly validated and could be used to assess structural design changes.

Throughout my career, I was lucky enough to interact with many test groups. However, one regret is that I did not spend significant time working within such a group. For any young engineer planning their career path, I would strongly recommend they spend time on the testing side. Like analysis, testing is an art as well as a science; the only way to learn is hands-on.