Application of Co-evolutionary Genetic Algorithm for Structural Optimization

協同進化遺傳算法在結構優化中的應用

顏弘錡(Hong-Chi Yan) ； 簡春娟(Chuen-Jiuan Jane)

《計量管理期刊》 18卷1期 (2021/05) Pp. 27-49

摘要

The paper studies the application of the co-evolutionary genetic algorithms in the structural optimization design. The optimization issues with constraints can be converted to those without constraints with the penalty techniques, making the general genetic algorithms applicable to find the optimal solution for the design problems. Firstly, three penalty parameters are defined to control the values of the penalty items. The first penalty parameter is related to the distance of the feasible region threshold, the second to the degree of the infringement against constraints, and the third to the number of infringement against constraints. Then two groups of parent populations are defined: the first group is the parent population of the penalty parameter and the other of the design parameters. By the co-evolutionary genetic algorithms, both groups of parent populations evolve simultaneously to get the optimal penalty parameters fitting for the design problem. Finally, this paper develops a structural optimization program combining the co-evolutionary procedure with commercial ANSYS FEA software. With this program, some testing examples and practical structural design issues are solved. The analysis results show that the method in this paper can get very satisfactory convergence for general structural optimization issues.

關鍵字

Coevolutionary genetic algorithms ； Genetic algorithms ； Structural optimization ； Penalty techniques

並列摘要

本文研究了協同進化遺傳算法在結構優化設計中的應用。可以使用懲罰函數將有約束的最佳化問題轉換為無約束的最佳化問題，從而使通用遺傳算法可用於找到設計問題的最佳解決方案。首先，定義了三個懲罰參數以控制懲罰項的值。第一個懲罰參數與可行區域閾值的距離有關，第二個與違反約束的程度有關，第三個與違反約束的數量有關。然後定義了兩組母體：第一組是懲罰參數的母體，另一組是設計參數。經由共同進化遺傳算法，兩組親代群體同時進化，獲得了適合設計問題的最佳罰分參數。最後，本文開發了一種將協同進化過程與商業ANSYS FEA程式相結合的結構優化程序。經由該程式，可以解決一些測試示例和實際的結構設計問題。分析結果顯示本文所提出的方法對於一般的結構最佳化問題具有很好的收斂性。

並列關鍵字

協同進化遺傳算法 ； 結構優化 ； 懲罰函數 ； ANSYS FEA