Doraiswamy, S., Krishnamurty, S. and Grosse, I., “Decision Making in Finite Element Analysis”, 1999 ASME Computers in Engineering Conference, Las Vegas, Nevada, DETC99/CIE9058.
In the absence of analytical expressions for design performance parameters, designers must often resort to either statistical or computer-based numerical techniques for performance estimation. Computer-based techniques without adequate assumptions are often very expensive and thus are infeasible for a complex design. Statistical response surface methods are cheaper, but they have poor accuracy for complex design configurations.Moreover, an initial set of data points are needed for the response surface methodology, which can only be obtained by computer-based techniques or real world data. Finally, the selection of an appropriate model to evaluate a design should be based on designer's preferences. This paper addresses these issues through the development of a decision based modeling methodology that effectively unites the numerical and statistical approaches. In this work, concepts from decision analysis are employed to deal with trade-offs between modeling attributes from a maximum expected utility perspective to identify “overall best” finite element models based on performance attributes such as analysis accuracy, computational cost and model resolution. Its utilization in a design setup is illustrated with the aid of a windshield wiper case study for selecting the best overall finite element analysis model.
Keywords: Finite Element Analysis, Decision Making, Design of Experiments.