For most engineering problems there is no difference in accuracy between FEA codes, even CAD Embedded FEA vs dedicated FEA codes. In fact nearly all of the error is related to factors apart from the code itself. Far more relevant sources of error include:
- Geometry errors – how well does my CAD/FEM model match the real-world component/s?
- Load errors – how well are the loads (magnitude, distribution) really understood?
- Boundary conditions – It takes a good analyst to get realistic results near constraints and to achieve realistic deformations (and hence, stresses)
- Mesh errors – Is the mesh fine enough that ‘convergence’ is achieved and element quality checks are acceptable?
- Element and solver selection – Can the element and solver chosen capture the mechanics of materials present in the real world? For example a linear-static solver will not assess the tendency to buckle i.e. a 1m long, 5mm diameter rod will record exactly the same von Misses stress whether in compression or tension, and give no indication that the rod would buckle at a fraction of the tensile load when loaded in compression. This often trips people up when analysing sheetmetal parts.
- The human factor – while it is surprisingly easy to get a load, constraint, material or contact definition incorrect, interpretation of results is another area where there are some hidden traps and a good understanding of mechanics of materials, material science, and FEA go along way if realistic results are desired.