you are here:   / News & Insights / Engineering Advantage Blog / The Importance of Validating Your FEA Process - Part 1

Engineering Advantage

The Importance of Validating Your FEA Process - Part 1

September 25, 2015 By: Nick Veikos

In an earlier blog, I discussed the importance of Verification and Validation in a simulation process, and I focused mostly on the Verification aspect. As a reminder for those not familiar with the terminology, Verification is the process by which we check that the simulation was conducted properly and Validation is the process to check whether results of the validated model reflect reality. In this discussion, I will focus on the importance of the Validation aspect, helping to answer the question: “Did I solve the right problem?”

Many years ago, when all computer graphics were green and computer screen degaussing was de-rigueur, I was working on a process to automatically correlate analysis and test data. FEA and test results for natural frequency and mode shape would be compared, and finite element model parameters would be adjusted to bring the results into agreement. At the time, I had a solid background on the mathematical and simulation aspect of the project, but very limited experience in the “real-world” of physical testing. As a naive analyst, I just considered that to be somebody else’s problem.

After months of coding sophisticated algorithms and “verifying” the software using simulated test data, my team and I were confident that we had developed something pretty impressive. We set up a test plan for the simplest structure we could think of – a cantilevered plate. This would be a slam dunk, we thought, ensuring we would receive additional funding for future work. Well, as you may have imagined, things did not work out so well – we did achieve correlation, but our “correlated” model looked nothing like the actual plate.

We had a verified model – the mesh was converged, the elements were well shaped, the material properties and geometry were correct, along with the mass and CG, and the results were well matched with hand calculations. So what went wrong? The experienced engineers out there know the answer – boundary conditions! We did not include support flexibility as one of the model parameters to be tweaked in order to achieve correlation. For the test, the plate had been periodically bolted along its edge and not welded to the support, as we assumed in our model. Only after including additional parameters to allow for some motion at the support, did we achieve success.

In this case, we had solved the analytical problem of a plate fully fixed at one end correctly, but it was the wrong problem to solve. Our assumptions were flawed from the beginning, giving us no chance of getting to the right answer.

This example highlights the importance of Validation as part of any simulation plan. While it is impractical and even counterproductive to test every component and assembly, some degree of testing must be implemented in order to verify the model assumptions and instill confidence in simulation as a predictive tool. Without this confidence, simulation-driven product development cannot become a reality.

Based on the testing results, modeling approaches, boundary conditions, and other simulation assumptions can be fine-tuned in order to reproduce the test data. This information should be captured for re-use on similar applications, thereby reducing the burden of testing every configuration and enabling simulation driven product development.

However, testing is not the only way to validate simulation assumptions. Some degree of validation can be achieved by starting off with a baseline simplified modeling approach, building complexity into the model in steps, and assessing the impact on the solution of each level of complexity. This will provide a good level of understanding for the amount and type of error being introduced by various modeling assumptions.

A typical example of this form of validation is first modeling a thin structure with shell elements, neglecting small features such as fillets, and re-modeling using 3-D solid elements, including all the features. The results will be different – the question becomes: is the difference important relative to your desired outcome?

In my next post I will discuss some aspects of a strong Validation plan for simulation.  In the meantime, I welcome you to share your stories about how failing to validate your analysis affected your professional life!  We can all learn from one another’s mistakes.