基于改进灰狼优化算法应力强度因子预测模型  

Stress Intensity Factor Prediction Model Based on Improved Grey Wolf Optimization Algorithm

作　　者：潘海珠[1,2] 葛海淼 苏小红[2] 袁琪[1] PAN Haizhu;GE Haimiao;SU Xiaohong;YUAN Qi(College of Computer and Control Engineering,Qiqihar University,Qiqihar 161006,Helongjiang,China;School of Computer Science and Technology,Harbin Institute of Technology,Harbin 150001,China)

机构地区：[1]齐齐哈尔大学计算机与控制工程学院,黑龙江齐齐哈尔161006 [2]哈尔滨工业大学计算机科学与技术学院,哈尔滨150001

出　　处：《实验室研究与探索》2021年第10期6-11,共6页Research and Exploration In Laboratory

基　　金：国家自然科学基金项目(61672191);黑龙江省自然科学基金项目(LH2020F050);黑龙江省省属本科高校基本科研业务费面上项目(135509116)。

摘　　要：含裂纹功能梯度材料的应力强度因子是表征材料断裂的重要参数,由于功能梯度材料属性的梯度变化,使得很难求得理论解,为此提出了运用概率变异的灰狼优化算法。结合分层指数模型求得的混合型应力强度因子值作为数据样本,训练一个支持向量回归(Support Vector Regression, SVR)模型,实现对含任意方向裂纹的一般属性功能梯度材料应力强度因子的预测。实验结果表明,与传统灰狼优化算法寻参的SVR预测模型相比,提出的概率变异灰狼优化算法SVR模型预测的应力强度因子值与理论解的拟合度更好,在优化SVR参数、提高预测精度、减小预测误差等方面表现更优。The stress intensity factor(SIF) of functionally graded material(FGM) with the crack is an important parameter to characterize the fracture of FGMs.The properties of functionally graded materials vary spatially so thatit is difficult to solve the crack problem analytically.The probability mutation grey wolf optimization algorithmis applied to optimize the parameters of support vector regression(SVR).And a predictionmodel is trained with sample data from SIFssolved analytically by the multi-layered model.The prediction model can predict the mixed-mode SIFs of FGMs with arbitrary orientation crack and arbitrary properties.Experimental results show that the predictions of SIFswith the model based on SVR inprobability mutation grey wolfoptimization algorithmfit more excellently with the results from analytical solutions than SVR model based on the grey wolf optimization algorithm. And this method has better performance in optimizing parameters, improving prediction accuracy, and reducing prediction error than SVR modelbased on the traditional grey wolf algorithm.

关 键 词：概率变异灰狼优化算法 支持向量回归 预测 应力强度因子 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]