Optimization Design for Buckling of Composite tubes

• Main Authors: Jhih-Long Chen, 陳致融
• Other Authors: Shun-Fa Hwung
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/p88sp7

碩士 === 國立雲林科技大學 === 機械工程系 === 103 === The composite of this research is made from prepreg material, and the strength of composite tube mainly depends on the fiber angle of each layer. In the optimization of fiber angles for composite laminates, the fiber angles of some layers have major effects on the objective, while some just have minor effects. To find optimal solution and accelerate convergence, an elite comparison mode has been included in a genetic algorithm. In order to validate this proposed algorithm, comparison with the results from the literature has been done. The results showed that the improved algorithm could fast search the optimal solution. For complex problems, the time to find the optimal solution is long because it takes a lot time in finite element analysis (FEA), even though the elite comparison genetic algorithm is used. Therefore, in this study, we attempt to replace FEA by artificial neural network to reduce the execution time. We use back-propagation neural network and the train algorithm is Bayesian regularization. Also, we validate results by comparing with the literature. The results indicate that the proposed method can accelerate the convergence speed and reduce the optimization time. As the optimization for the buckling of composite tubes is considered, the optimal solution is [90/0/45/-45]s, if finite element prediction is used. analyze value is 147063 N, and predicted value is 148333.7 N. When artificial neural network is included, the optimal solution is [45/90/0/45] s, analyze value is 160628 N, and predicted value is 116491 N. Although the obtained solution is not the optimal one, it is reasonable, and especially the optimization time is significantly reduced. Hence, it may be applicable in industry problems.