How to get meaningful and correct results from your finite element model

This document gives guidelines to set up, run, and postprocess correct simulations with the finite element method. It is not an introduction to the method itself, but rather a list of things to check and possible mistakes to watch out for when doing a finite element simulation. The finite element method (FEM) is probably the most-used simulation technique in engineering. Modern finite-element software makes doing FE simulations easy – too easy, perhaps. Since you have a nice graphical user interface that guides you through the process of creating, solving, and postprocessing a finite element model, it may seem as if there is no need to know much about the inner workings of a finite element program or the underlying theory. However, creating a model without understanding finite elements is similar to flying an airplane without a pilot’s license. You may even land somewhere without crashing, but probably not where you intended to. This document is not a finite element introduction, see, for example, [3,7,10] for that. It is a guideline to give you some ideas how to correctly set up, solve and postprocess a finite element model. The techniques described here were developed working with the program Abaqus [9]; however, most of them should be easily transferable to other codes. I have not explained the theoretical basis for most of them; if you do not understand why a particular consideration is important, I recommend to study finite element theory to find out. 1 ar X iv :1 81 1. 05 75 3v 1 [ cs .N A ] 1 4 N ov 2 01 8 1 Setting up the model 1.1 General considerations These considerations are not restricted to finite element models, but are useful for any complex simulation method. 1.1-1. Even if you just need some number for your design – the main goal of an FEA is to understand the system. Always design your simulations so that you can at least qualitatively understand the results. Never believe the result of a simulation without thinking about its plausibility. 1.1-2. Define the goal of the simulation as precisely as possible. Which question is to be answered? Which quantities are to be calculated? Which conclusions are you going to draw from the simulation? Probably the most common error made in FE simulations is setting up a simulation without having a clear goal in mind. Be as specific as possible. Never set up a model “to see what happens” or “to see how stresses are distributed”. 1.1-3. Formulate your expectations for the simulation result beforehand and make an educated guess of what the results should be. If possible, estimate at least some quantities of your simulation using simplified assumptions. This will make it easier to spot problems later on and to improve your understanding of the system you are studying. 1.1-4. Based on the answer to the previous items, consider which effects you actually have to simulate. Keep the model as simple as possible. For example, if you only need to know whether a yield stress is exceeded somewhere in a metallic component, it is much easier to perform an elastic calculation and check the von Mises stress in the postprocessor (be wary of extrapolations, see 3.2-1) than to include plasticity in the model. 1.1-5. What is the required precision of your calculation? Do you need an estimate or a precise number? (See also 1.4-1 below.) 1.1-6. If your model is complex, create it in several steps. Start with simple materials, assume frictionless behaviour etc. Add complications step by step. Setting up the model in steps has two advantages: (i) if errors occur, it is much easier to find out what caused them; (ii) understanding the behaviour of the system is easier this way because you understand which addition caused which change in the model behaviour. Note, however, that checks you made in an early stage (for example on the mesh density) may have to be repeated later.