利用基因演算法和ANSYS workbench多目標遺傳演算法預測材料機械性質

Using Genetic Algorithms and ANSYS Workbench Multi-Objective Genetic Algorithm to Predict the Mechanical Properties

邱武耀(Wu-Yao Chiou) ； 邱評新(Ping-Shin Chiou) ； 蔡政諺(Jeng-Yan Tsai) ； 陳彥伊(Yan-Yi Chin)

《中華民國振動與噪音工程學會論文集》 2013 (2013/06) Pp. 142-148

摘要

本研究是利用振動量測物件的自然頻率和模態分析，並結合電腦計算模態分析來預測材料機械性質。此方法避免了以往所使用的破壞的方式來檢測材料的機械性質。基於傳統基因演算法因為群體同步搜尋，故會浪費電腦記憶體與演算速率，而使整體演算時間加長；因此，本研究嘗試使用兩種不同的方法：首先，利用MATLAB執行基因演算法（Genetic Algorithm，GA）並搭配有限元素分析軟體（ANSYS）；另一是利用ANSYS workbench多目標遺傳演算法（Multi-Objective Genetic Algorithm，MOGA）來反求機械性質。除了比較兩方法之計算精度，另探討計算效率。研究過程分別以矩形碳鋼、矩形鋁、高爾夫鐵桿球頭和木桿球頭為分析例。首先，利用實驗來得到頻率響應函數，再搭配模態分析軟體ME'scope進行模態振型曲線擬合而得到物件的模態數據，數據結果可用來印證數值分析模態振型的正確性；其次，設定材料機械性質之計算範圍，分別利用基因演算法和workbech多目標遺傳演算法來進行有限元素法機械性質的分析。數值結果顯示MOGA比GA更加準確和計算效率。所有數值與理論值相比較，楊氏係數除了GA計算碳鋼誤差7.25%較高外，其餘都低於5%；而蒲松比則誤差相對較大，MOGA誤差最大為16%，而GA誤差最大為20%，原因可能為蒲松比對頻率的敏感度較低所致。另外在計算效率上，本例MOGA的運算速度時間比GA快435倍以上。

關鍵字

實驗模態分析 ； 頻率響應函數 ； 模態參數 ； 基因演算法 ； 多目標遺傳演算法

並列摘要

This study is the use of the vibration measurement of the work piece natural frequency and modal analysis, combined with computational modal analysis to predict the mechanical properties of materials. This method avoids the destruction method to detect the mechanical properties of the material used in the normal. Based on the traditional genetic algorithm because the group synchronization searches, it will be a waste of computer memory space and calculus rate, resulting in a longer calculation time. Therefore, this study tries to use two different methods: First, the use of MATLAB genetic algorithm (GA), and with finite element analysis software (ANSYS). The other is the use of the ANSYS Workbench multi-objective genetic algorithm (MOGA) to reverse mechanical properties. In addition to compare the two methods of computational accuracy, and the other to investigate the computational efficiency.Research process using a rectangular carbon steel, rectangular aluminum, golf iron club and the wooden club for the analysis of cases. First, the use of experiments to get the frequency response function, and then with modal analysis software ME'scope the modal shape curve fitting object modal data, the data results can be used to prove the correctness of the numerical method mode shapes. Secondly, the calculation ranges of the set material mechanical properties of use the genetic algorithms and workbech of multi-objective genetic algorithm for the analysis of the mechanical properties of the finite element method.The numerical results show that MOGA is more accurate than the GA and computational efficiency. All data is compared with the theoretical value, the Young's modulus in addition to the GA to calculate the carbon steel error 7.25% higher, the rest are less than 5%, and the Poisson's ratio error is relatively large, the MOGA is the maximum error of 16%, while the GA error up to 20%, probably due to the Poisson's ratio with the lower sensitivity to frequency. In addition, the computational efficiency, this analysis cases the MOGA computing speed time more than 435 times faster than the GA.

並列關鍵字

experiment modal analysis ； frequency response function ； Genetic Algorithm ； Multi-Objective Genetic Algorithm