engleski

Numerical Experience with Generalized Convex Approximation for Design Optimization

A two-point generalized convex approximation method of Chickermane and Gea possess some extremely important features for design optimization: simplicity, cost effectiveness and accuracy. Unfortunately, the published paper contained a small set of five examples to prove its applicability for general design optimization purposes. Therefore, in this paper a more broaden numerical testing of the generalized convex approximation is presented. Generalized convex approximation is tested on a set of 18 real-life engineering design optimization problems. A full set of 97 functions and 10889 function approximations is tested and compared with analytical values. The results showed that majority of approximations performed exhibited excellent accuracy. Unfortunately, a small subset of performed approximations is extremely inaccurate and therefore unusable. However, if properly applied, generalized convex approximation method may be a valuable tool in design optimization.

design optimization; function approximation

nije evidentirano