A Comparison Meta-Modeling Methods Using Practical Industry Requirements
Liping Wang, Don Beeson, Gene Wiggs, Mahidhar Rayasam · 2006
In spite of the exponential growth of computing power, the enormous computational cost of complex and large -scale engineering design problems make it impractical to rely exclusively on original high fidelity simulation codes. Performing probabilistic and robust design on these long running applications is still a formidable task in any industry. The objective of this paper is to investigate the advantages and disad vantages of the several promising meta -modeling techniques under practical industry requirements. The comparisons are based on a set of performance criteria that are known to be important to industry. The meta -models include Multivariate Adaptive Regressio n Splines (MARS), Radial Basis Functions (RBF), Adaptive Weighted Least Squares (AWLS), Gaussian Process (GP), and quadratic response surface regression (RSM). Twenty benchmark problems representing different levels of dimensionality, non -linearity and noi se level are used for the evaluation.