This paper was produced for the 2019 NAFEMS World Congress in Quebec Canada

Resource Abstract

When setting up numerical models there are trade-offs between model complexity, solve time, and accuracy. Recent trends are driving usage of simulation earlier in the design process, and recent software releases such as ANSYS Discovery Live seek to support analysis of early design concepts in a rapid and efficient manner. It is important to acknowledge that each simulation has varying needs with regards to accuracy and speed of solution, but the analyst needs to have a reasonable understanding of these trade-offs so that they can appropriately choose how to build and solve the model. Work will be presented comparing results from models which have been solved with varying degrees of complexity. High fidelity models which include non-linear interactions with mating parts, well refined meshes, etc. and which have been validated against test data will be compared to simplified models. In each case two simplified models are run. The first is solved in the original solver and preserves the original mesh, but uses simplified constraints and assumptions to remove non-linearities such that the model could have been solved in Discovery Live. The second simplified model is then solved in Discovery Live. The resulting comparisons allow a determination of how much the results change due to modifications in the applied constraints/physics vs how much change is due to meshing decisions or other coding variances inherent to Discovery Live. The models being presented all come from industrial control valves, but will include different solutions focused on pressure retaining stresses, natural frequency, and heat transfer. One observed outcome is that in some applications a model can be simplified with a minimal loss of accuracy, while in other applications the higher fidelity was quite important, but experience based predictions were not always a good indicator of when the fidelity was critical. Specifics of the methodology in making the comparisons will be shared, along with the observed outcomes. Potential considerations with regards to selecting simplified models for early concept evaluation versus high fidelity models for final product validation will also be discussed.