• 
• 
• 
• 
• 

Conceived, organised and supported by Fluxus Technology Ltd

Supported by the European Commission, Research DG, Human Potential Programme, High-Level Scientific Conferences, HPCF-CT-2000-00118

In cooperation with the Society for Industrial and Applied Mathematics (SIAM)

About | ECITTT | Conference Programme | Abstracts (19.)

FEA AS CARRIED OUT BY DESIGNERS NOT ANALYSTS:
COMMON PROBLEMS AND PITFALLS

R P Johnson BSc MSc NRA MIMechE CEng
Director DAMT Limited, Editor of NAFEMS' Benchmark magazine
Newport Pagnell
Buckinghamshire
England MK16 9AT
Introduction

The continued presentation of the two NAFEMS courses, “The Basics of Stress Analysis” and “The Fundamentals of FEA for Design Engineers” has brought the author face to face with a steady supply of Designers and new FEA users. If we concede that these Designers and Analysts are fairly representative of new FE-users as a whole then the difficulties voiced should be of significant interest to those responsible for training FEA staff. The intention of this paper/presentation is to highlight the most common difficulties by following a number of examples, some of which are extracted from the aforementioned courses.

Common difficulties

For the less-experienced user the five most common difficulties with FEA seem to be as follows:

1. Time pressures require that analysis is performed quickly with minimum effort allocated to geometry and mesh generation. It is required that the CAD model developed prior to the FEA can be used with a minimal amount of effort. The requirement here is for the geometry export to be made successfully (via direct transfer, IGES, STEP or other) from the sending CAD system into the receiving FEA system. The CAD data is then used to build (say) a solid and, by the click of a button, a fully continuous and high-quality “tet” mesh is produced. In practice this is seldom the course of action and, more commonly, the user is plagued with a number of problems such as: Tolerance/accuracy problems; cutting the model to utilise symmetry; massive amounts of unwanted details; slender/sliver surfaces which create poor elements; cracks in the geometry and so on.



2. By way of a check-list the new user assumes that the following equation must be followed to the letter: Geometry + Mesh + Loads + Constraints + Material properties = Correct Results. Strict adherence to this list produces the common assumption that (applied) loads are put on one end of the structure and constraints/boundary conditions are put on the other end. This may lead the inexperienced analyst into the trap of over-constraining the model so that loads are partly carried by the structure and partly by the boundary conditions. Non-conservative displacements/strains/stresses may result and an otherwise incapable structure can be released into the outside world.



3. The new user is, perhaps, unaware that the operator “driving” the system one way or another can produce a vast array of different results. Also that the Finite Element method is an approximate technique and that one density of mesh will give one set of results, and some other density another set entirely. Differences in results can come from a number of sources and examples will be provided which illustrate the sensitivity that can occur in practice.



4. Most new users (perhaps even the majority of us) employ FEA to make relative comparisons between some control (or initial) run and a family of evolving runs which are designed to reduce the stresses in the component under investigation. Generally speaking this is a relatively safe use of FEA as the errors inherent in the second run will be of the same order as the first and errors therefore subtract from another. Relative improvements in the design can therefore be extracted from the results of a series of FEA runs. In direct contrast however, ABSOLUTE finite element analysis requires considerably more thought and the issues relating to a suitable failure criteria come to the fore. Further examples will be given to illustrate the difficulties to be faced by the less-experienced user.



5. Finally in results assessment there seems to be a tendency NOT to interrogate results thoroughly enough (perhaps we are all guilty at times). Most new users stop at peak stresses (from the whole model) and colour plots. More often required is a number of graphs or numeric plots in order to discover the true value of the displacement/strain/stress results presented by the package being used. Then even the most peculiar results can be explained (and should be) and the true path of the analysis understood fully. Again examples will be shown to illustrate that colour plots alone will not be sufficient to understand what the FEA system has undertaken.

Conclusion

Five main areas of difficulty for the Designer/less-experienced analyst have been identified. A number of examples will be given to highlight the significance of the actions taken by the analyst showing that a vast range “answers” can be obtained from a single FEA model.