The following list is a subjective assessment of the most common
finite element analysis errors encountered from working in several industry
sectors. If these sound familiar, you may want to review your analysis
procedures. Contact us; we can help you implement
best practice.
-
Doing analysis for the sake of it: Not being aware of
the end requirements of a finite element analysis - not all benefits of
analysis are quantifiable but an analysis specification is important and all
practitioners should be aware of it.
-
Lack of verification: Not having adequate verification
information to bridge the gap between benchmarking and ones' own finite element
analysis strategy. Test data sometimes exists but has been forgotten. Consider
the cost of tests to verify what the analysis team produces, compared with the
potential cost of believing the results when they are wrong.
-
Wrong elements: Using an inefficient finite element
type or model, e.g. a 3D model when 2D would do, or unreliable linear
triangular or tetrahedra elements.
-
Bad post-processing: Not post-processing results
correctly (especially stress) or consistently. Not checking unaveraged
stresses.
-
Assuming conservatism: Because one particular finite
element analysis is known to be conservative, a different analysis of a similar
structure under different conditions may not be so.
-
Attempting to predict contact stresses without modelling
contact: This might give sensible looking results but is seldom meaningful.
-
Not standardising finite element analysis procedures:
This has been a frequent cause of repeated or lost work. Any finite element
analysis team should have a documented standard modelling procedure for typical
analyses encountered within the organisation, and analysts should follow it
wherever possible. Non-standard analyses should be derived from the standard
procedures where possible. This is a quality issue - Procedures in Analysis
Reduce Errors (PARE).
-
Inadequate archiving: Another frequent cause of lost
work. Teams should have a master model store and documented instructions about
what and how to archive. Again, this is a quality related issue. For any kind
of analysis data, normal backup procedures are not sufficient - attention needs
to be paid to what information and file types are to be archived in order to
allow projects to be retraced, but without using excessive disk space.
-
Ignoring geometry or boundary condition approximations:
Try to understand how inappropriate restraint conditions in static or dynamic
analyses can affect results.
-
Ignoring errors associated with the mesh: Sometimes
these can cancel out errors associated with 9, which can confuse the user into
thinking that the model is more accurate than it is. A convergence test will
help.
