This paper on "Steel Constructions with Thin-Walled Components - Problems and 'Pitfalls' in Effective Modelling and Assessment of FEA Results" was presented at the NAFEMS World Congress on The Evolution of Product Simulation From Established Methods to Virtual Testing & Prototyping - 24-28 April 2001, The Grand Hotel, Lake Como, Italy.

Summary

FE analysis plays an important role during the analysis and assessment phase of the design process. In all of the progress made in CAD-FEM correlation in recent years, the area of thin-walled structures has seen virtually none. This is due to the mathematical difficulties involved in the complicated deconstruction of three-dimensional structures into mid-surfaces, based partly on the assumption of analysts' experience. But also because most CAD developers are simply unaware of the fact that up to 80% of structures in the fields of lightweight design and production technologies, e.g. steel-frame construction, that undergo Finite Element Analysis (FEA) fall into the thin-walled category and therefore to a great amount shell elements are employed to conduct an FE analysis. On hand of examples from the industry, the authors wanted to stress problems and 'pitfalls' in effective modelling and assessment of FEA results of thin-walled components. Our findings, in addition to difficulties experienced by others with such structures, suggest that it is very successful to use shell elements for most parts of the structure, if one wants to cut analysis cost and have a good trade-off between good quality results and FE model size. Even in large numbers, linear 3D elements still exhibit very large errors in these cases and generally unfit for thin-walled structures. In the solution developed by FEMCOS for the purpose of automatic coincident mid-plane generation, presently limited to non-continuous changes in thickness (such as typical steel-frame structures), a serviceable solution is available, which greatly helps to improve the efficiency of mid-surface generation, therefore saving both development time and costs. The assessment of the results obtained from the FEA is essential for the safety and reliability of the structure and a basis for alteration in the structure towards an optimal design. We are going to show that this is not always a plain sailing. The authors are strongly supporters of the application of substructure/superelement technique even for the structures under consideration in this paper. Not only the tremendous reduction of computer time and memory need speaks for it, but also alterations in the design, optimisation, etc. can be carried out with astonishing less effort.