Analysis of Support Vector Regression for Approximation of Complex Engineering Analyses

TLDR

A variety of metamodeling techniques have been developed to reduce the computational expense of computer-based analysis and simulation codes, including response surface methodology, kriging, radial basis functions, and multivariate adaptive regression splines. The paper presents Support Vector Regression as an alternative technique for approximating complex engineering analyses. The authors compare SVR approximations to four existing metamodeling techniques using a testbed of 22 engineering analysis functions, highlighting its computational efficiency. SVR achieves more accurate and robust function approximations than the four techniques, showing great promise for future metamodeling applications.

Abstract

A variety of metamodeling techniques have been developed in the past decade to reduce the computational expense of computer-based analysis and simulation codes. Metamodeling is the process of building a “model of a model” that provides a fast surrogate for a computationally expensive computer code. Common metamodeling techniques include response surface methodology, kriging, radial basis functions, and multivariate adaptive regression splines. In this paper, we present Support Vector Regression (SVR) as an alternative technique for approximating complex engineering analyses. The computationally efficient theory behind SVR is presented, and SVR approximations are compared against the aforementioned four metamodeling techniques using a testbed of 22 engineering analysis functions. SVR achieves more accurate and more robust function approximations than these four metamodeling techniques and shows great promise for future metamodeling applications.