There are some questions that need to be answered before you start the analysis. For example, if you are interested mainly with the durability of a component then this will require a different approach than a linear static analysis. It is not usually cost effective to adopt a 'one size fits all' approach. You need to focus on what information you want to gain from the analysis and tailor your methods for the most cost effective solution.
The quickest and cheapest type of analysis is linear static because of faster solve times and there are less things to go wrong. You often find that non-linear analysis doesn’t work the first time around unless you put in extra effort up front when building the model, either way it takes much longer.
Some examples of questions that should be answered before starting the analysis are as follows;
1) Are you interested in a single component or an assembly of components? Try balancing the model with reaction forces at the interface of connecting parts rather than using contact analysis.
2) What is your primary concern, is it permanent deformation, strength, fatigue, thermal expansion, impact or noise and vibration? Try to find a solution using linear static analysis first; this can save you a lot of time and effort. Non-linear analysis can often be approximated with linear static analysis. Fatigue analysis requires a fine mesh but if you a primarily concerned with linear displacement then a course mesh will suffice.
3) Is the analysis for design or to correlate with a physical test? Analysis for design often requires a different approach to analysis for testing. For the latter it is important to precisely replicate the test rig whereas for design you want to be as close to reality as possible.
4) Does the analysis need to comply with a Design Standard code? FEA Analysis results are sometimes used in calculations from British Standard codes. These codes often work with nominal stress where as FEA models give peak stress.
5) Is your workstation and FEA software compatible with the requirements of the analysis? Large models and non-linear analysis usually require a high specification workstation. Know your limits before you get too far.