New Zealand Engineering 1999 March

Manufacturing

Towards FEA Quality

In our articles, we have stressed that Finite Element Analysis (FEA) should not be used as a "black box", with CAD geometry as the input and stress contour plots as the output. The bottom line is that it is the responsibility of the engineer or designer using the software to determine if the results produced from FEA are valid.

As our previous two articles (New Zealand Engineering, April and September 1998) discussed, there are two aspects to this: the first is to ensure there are no model defects such as free edges, shell elements with reversed normals, etc; and the second to validate the results against those observed under test or operating conditions.

The issue of FEA validity is the primary concern of the NAFEMS organisation. This is based in Britain and was founded in the mid-80s with the aim "the promotion of quality and safety in analysing structures using finite element codes". Initially, the principal concern was developing benchmarks for FEA software, and NAFEMS still produces a collection of benchmarks ranging from sensitivity of results to element shape through to non-linear material benchmarks for plasticity and creep.

These benchmarks are invaluable for first time users of a particular analysis code or for users attempting analyses in a situation where the results are not obvious and are difficult to confirm with simple hand calculations. This particularly applies to non-linear problems where an analytic solution may not exist, and various FE packages vary widely in their ability to correctly model the behaviour or obtain a converged solution. Not all software packages are created equal and if the software is unable to match benchmark results for an extremely simple example, it is unlikely to produce valid results where the geometry and loading is far more complex.

The issue of software robustness also explains the apparent disparity in the cost of FEA software. Although advertising literature may suggest an "expensive" and "inexpensive" package have the same capabilities, the reality is that the less expensive software probably lacks the extensive testing across a wide range of application areas, and may give spurious results even for a well-posed problem. There is a maxim I frequently quote that "all FEA software eventually costs the same amount of money" – a software package with a greatly lower up-front cost will generally result in increased costs in model verification and "debugging" when using the product.

Although these benchmarks are still important, nowadays, the NAFEMS focus is more on FEA education and establishment of an accreditation scheme for finite element analysts. Towards the former goal, they hold regular seminars on various aspects of FEA (unfortunately all in their home country of England!) and produce a wide selection of books and booklets on interpretation of FE results for various analysis types (linear, non-linear, dynamic). These books are essential reading for the budding analyst.

Recently, they have developed a Registered Analyst Scheme. The scheme is based on accumulation of analysis and product experience, together with sufficient theoretical background. In particular, the requirements include:

• Degree equivalent in engineering, physics or applied mathematics

• A level of product experience based on the "importance" of the analysis work (code verification vs mission-critical work in the aerospace or nuclear industry). For the bulk of work done in NZ, a minimum of two years analysis in a particular product area (such as aircraft, dams, pressure vessels, etc.) is required to meet the minimum standard

• Coupled with the above is a requirement for numerical analysis which is broken up into categories such as linear static, dynamics, fracture mechanics, thermal, etc. The analysis experience is based on the complexity of the work, from simple linear static through to extensive validation against measured test results. Approximately a year’s experience in any numerical discipline is required for certification.

One of the aims of NAFEMS is to produce an ISO standard for analysis work. This was produced some time ago in draft form and is a valuable document on the checking that should take place during an analysis project. Because of procedural difficulties in producing such a standard, the registered analyst scheme is seen as the first step in driving analysts to seek peer review of their work and to build up a pool of personnel with expertise in specific areas.

Although the bulk of the NAFEMS membership (approximately 700 members) is based in Europe, there is now a rapid expansion in membership for the USA, Asia and Australia. Matrix is a full member of NAFEMS and Don Campbell is a registered analyst at the advanced level in a wide variety of product and analysis types.

If your company uses FEA as part of their design and analysis procedures, we would recommend looking at what NAFEMS offers both the analyst and the company. In particular, the registered analyst scheme provides an opportunity to undergo peer review of FEA work performed, a development path for the analyst and some assurance of quality standards for those relying on their results.

Don Campbell is a director of Matrix Applied Computing Ltd, Auckland