Verification and validation (V&V) in finite element analysis (FEA) is becoming a major focus for organizations wanting to control the quality of their engineering solutions. Conversely, traditional FEA tools on the market today were not explicitly designed to support the rigorous standards of V&V. What is needed to bridge this gap between FEA codes and V&V in order to consistently deliver reliable solutions?
Verification guarantees that the answer converged for the finite element (FE) solution of the model and validation is needed to show that the model is truly representative of the problem at hand. The FE solution must be verified and the model must be validated in order to guarantee a reliable analysis.
An FEA tool must consistently separate the model definition from the FE solution, allowing the engineer to distinguish between the two possible sources of error(s). If the model definition cannot be separated from the FE solution during the analysis, then modeling and simulation errors are mixed, obfuscating where the problems lie. This obfuscation prevents the guarantee of reliability of the solution.
How do we guarantee reliability in FEA?
To avoid finite element modeling (mixed errors), and uncertain reliability, an FEA tool that is designed for V&V should feature:
- A hierarchic modeling framework which increases complexity of the model separate from meshing.
- A simple element library which is based only on shape, not on the model being solved.
- A single mesh; no need for iterative re-meshing. Higher order FE solutions would be added to the same mesh of the model to reduce numerical error.
- All higher order FE solutions would be kept for post-processing, and would demonstrate mesh independence through automatic convergence.
A Practical Example of V&V
We have an eye bolt in tension problem and want to check that the max 1st principal stress is less than an allowable value. We open up an FEA package and model it based on the problem data we have. We set up the model, mesh it, solve it and get an answer. How do we know the answer is reliable?
If we use an FEA tool explicitly designed to meet the requirements of V&V, we would get automatic convergence results to allow us to check our values as we increase from lowest to highest order. There is no change in the mesh to get an improved answer, as shown here:
The solution converges to a single value (96 ksi) demonstrating true mesh independence.
Once we’ve finished the verification step we can validate the model with confidence and make a well informed decision on how good it is.
We can confidently compare our design allowable with our FEA answer to make sure we are within an acceptable range. Now we have earned the right to say our model was good enough to represent the problem.
If the model wasn’t sufficient to represent the problem, we would now know that it wasn’t due to meshing errors. However, without separating the model from the FE solution, we would never know if the model is truly representative of the problem and therefore could not possibly meet the standards of V&V.
ESRD’s StressCheck® = Guaranteed Reliability
ESRD’s StressCheck®, available through the Altair Partner Alliance, is one of the rare commercially available FEA tools that was designed from the ground up to meet the requirements of V&V.
StressCheck® incorporates all of the features listed under the “How do we guarantee reliability in FEA?” section, and more, to make the FE solution verification step hassle-free and deliver reliable solutions.
Latest posts by Altair Partner Alliance (see all)
- Failing to Optimize the Human-in-the-Loop at the Earliest Stages of Design is More Expensive than You Think – Part 3 - March 13, 2019
- Failing to Optimize the Human-in-the-Loop at the Earliest Stages of Design is More Expensive than You Think – Part 2 - March 6, 2019
- Failing to Optimize the Human-in-the-Loop at the Earliest Stages of Design is More Expensive than You Think – Part 1 - February 27, 2019