An efficient coupling reduction model-based evolutionary topology optimization method. (February 2023)

• Record Type: Journal Article
• Authors: Yin, Jichao; Wang, Hu
• Is Part Of: Advances in engineering software. Volume 176(2023)
• Publication Date: 2023-02
• Subjects: Topology optimization -- Coupling reduction model -- Smooth boundary representation -- Large scale -- High efficiency; Computer-aided engineering -- Periodicals; Engineering -- Computer programs -- Periodicals; Engineering -- Software -- Periodicals; Periodicals; 620.0028553
• DOI: 10.1016/j.advengsoft.2022.103394
• Languages: English
• ISSNs: 0965-9978

Abstract

Highlights: Coupling reduction model is proposed to improve computational efficiency. Smooth boundary is obtained and extend method to handle 3D problems. "Relaxed updated" strategy is proposed to reduce the evaluations of models. The maximum speedup is up to 38.55 times in given numerical examples. Large-scale 3D problems can be solved easily due to storage reduction. Abstract: Expensive computational cost and zigzag structural topology hinder the applications of topology optimization in practical industry. In this paper, a coupling reduction model-based evolutionary topology (CRMETO) method is proposed to improve computational efficiency and obtain smooth boundary representation. In the proposed method, the coupling reduction model is based on the integration of high-fidelity multigrid conjugate gradient model and low-fidelity on-the-fly reduced order model. To further improve the efficiency, a "relaxed updated" strategy is put forward to reduce the number of evaluations of both low-fidelity and high-fidelity models. The proposed method is validated and compared with various 2D and 3D examples. With the increase in the scale of model, the advantage of the proposed method is much more obvious in terms of efficiency, the maximum speedup is up to 38.55 times in the given examples. Moreover, the memory also can be saved so that 3D problems with more than millions of degrees of freedom can be easily solved. Graphical abstract: Image, graphical abstract