MACHINEDESIGNwww.machinedesign.com51• MARCH 21, 2002
Field variables, such asdisplacements and tem-peratures, are describedby lower-order polyno-mials, usually of thefirst and second order.Element order is fixedand does not changeduring the solution. Ele-ment shapes are re-stricted to primitives(tetrahedrons, wedges,and hexahedrons) andallow little deviationfrom ideal shapes.
Edited by Paul DvorakA common objective of FEM is finding displacements andstresses in a structure. This requires meshing (discretization) acontinuous mathematical model.
CONTINUOUS BODYDISCRETIZED BODY
Easy-to-make meshing errors can render finite-element analysis results misleading to dangerous.
Paul KurowskiDirector of Engineering DevelopmentGenexis Design Inc.London, Ontario, Canada
inite-element analysis, or the finite-element method (FEM) aswe will call it, hides plenty of traps for uninitiated users. Er-rors that come from idealization and meshing a part can bebad enough to render results either misleading or dangerous,depending on the importance of the analysis.Idealizing and defeaturing a 3D model eliminates small and unimpor-tant details. Sometimes the process replaces thin walls with surfaces, ordrops a dimension to work with a 2D representation of the part. Modelbuilding also uses simplified descriptions of material properties by, forinstance, considering them as linear-elastic materials (many are not),and assigning boundary conditions as rigid supports and time-inde-pendent loads. There are many other simplifying assumptions.The process eventuallyforms a mathematical de-scription of reality whichwe call a mathematicalmodel. To solve it with nu-merical techniques, themath model must be dis-cretized or meshed. (Dis-cretization and meshingare synonymous in theFEM). But mind you, cre-ating a mathematicalmodel is error prone.Here are a few of thethings that can go wrongwhen modeling, even be-fore meshing.
THE PROBLEMWITH MODELS
A mathematical modelcan be pictured as a con-tinuous domain with im-