STRUCTURAL OPTIMIZATION USING MULTIQUADRIC RESPONSE SURFACE MODEL

摘要

In recent years, response surface methodology (RSM) has become a popular tool for Multi-disciplinary Design Optimization (MDO). RSM provides an overall perspective of system response within the design space and simplifies the process of integrating different mathematical models required in MDO. Traditional RSM is a global approach that uses polynomials to approximate the responses. The coefficients of the polynomial models are computed using least square method. The computed response surface model is the 'the best-fit' polynomial function from the available data. In general, the fitted function does not interpolate the available data. In this paper, we investigate the use of multiquadric function to build response surfaces (MQRS) for design optimization. Essentially MQ response surface models are multidimensional interpolation functions. One basis function is associated for each data point (hence MQ is a special cases of approximation using radial basis functions). Thus MQ models are very flexible. It can be built with very little data.