An improved scheme for continuum structural topology optimization based on bi-directional evolutionary structural optimization

Abstract

Bidirectional Evolutionary Structural Optimization (BESO) has recently become a prominent research focus for continuum structural topology optimization. When using the conventional BESO method, oscillation in the iterative process and relatively low optimization performance are observed. To address these challenges, an improved scheme is proposed in this article. This scheme incorporates three key innovations: adaptive evolution rate, nonlinear filtering function, and historical sensitivity weighting factor. Firstly, an adaptive strategy with an exponential evolution rate is introduced. A dynamic adjustment mechanism is utilized to overcome the directional bias inherent in fixed evolution rate. Secondly, a nonlinear sensitivity filtering method based on the arcsine function is designed. This method alleviates the phenomenon of over-averaging by increasing the weight of the central element, contributing to a superior filtering effect. Furthermore, an improved weighted geometric mean method is employed to assign different weights to historical sensitivities, which mitigates oscillation in the iterative process effectively. Finally, several representative numerical examples demonstrate the superior computational efficiency, the reduction in iterative oscillation, and the better optimization performance of the scheme.