Truss optimization with frequency constraints using enhanced differential evolution based on adaptive directional mutation and nearest neighbor comparison. (December 2016)

• Record Type: Journal Article
• Authors: Pham, Hoang Anh
• Is Part Of: Advances in engineering software. Volume 102(2016)
• Page Start: 142
• Page End: 154
• Publication Date: 2016-12
• Subjects: Truss optimization -- Frequency constraint -- Differential evolution -- Adaptive p-best strategy -- Directional mutation -- Nearest neighbor comparison; Computer-aided engineering -- Periodicals; Engineering -- Computer programs -- Periodicals; Engineering -- Software -- Periodicals; Periodicals; 620.0028553
• DOI: 10.1016/j.advengsoft.2016.10.004
• Languages: English
• ISSNs: 0965-9978

Abstract

Highlights: A new differential evolution algorithm (ANDE) is proposed. ANDE is featured by adaptive directional mutation and nearest neighbor comparison. ANDE can well balance global exploration and local exploitation. ANDE can effectively reduce unnecessary function evaluations. ANDE is shown efficient for truss optimization with frequency constraints. Abstract: Truss optimization with dynamic constraints is well-known as challenging optimization problem and requires appropriate optimization techniques. In this article, a new differential evolution algorithm, named as ANDE, for solving shape and size truss optimization with frequency constraints is presented. Three modifications are introduced to conventional differential evolution (DE), including: 1) the adaptive p-best strategy to balance between global exploration and local exploitation; 2) the directional mutation rule to increase the possibility of creating improved solutions; 3) the nearest neighbor comparison method to prejudge a solution before evaluation and skip unpromising one. These modifications are relatively simple and do not require additional parameter setting for DE. Numerical results of five benchmark examples show that ANDE can provide good and stable results without violation of the frequency constraints. The optimal designs of ANDE in most cases are as good as or better than the results from some state-of-the-art metaheuristics. The benefit of ANDE is that it often uses fewer structural analyses than those required by the other methods.