激光焊接铝合金热裂纹预测数学模型研究  

Maths Model on Prediction with Hot-Crack in Aluminum-Alloy by Laser Welding

作　　者：吴爱民 张健 郭亮 张庆茂 夏琪[2,3] 

机构地区：[1]广东省微纳光子功能材料与器件重点实验室,广东广州510006 [2]上海市激光技术研究所,上海200233 [3]上海市激光束精细加工重点实验室,上海200233

出　　处：《应用激光》2013年第2期177-180,共4页Applied Laser

基　　金：中央财政支持地方高校专项资金项目激光先进制造创新培养基地资助项目(项目编号:510-C10293);广东省教育部产学研结合重点项目(项目编号:2010A090200048);广东省教育厅学科建设专项资助项目(项目编号:CXZD1139);国家863计划资助项目(项目编号:2012AA040210);上海市科学技术委员会科技攻关项目(项目编号:11DZ1110700)

摘　　要：在已有实验的基础上,利用已得实验数据,运用逐步回归模型和B.P神经网络模型,通过MATLAB平台进行理论计算。研究结果表明,理论计算与实验值具有良好的吻合性,其相对误差在8%以下,均方根误差在0.05以内;模型研究表明氧元素是促进铝合金焊接热裂纹增长的主要元素之一,镁等元素则抑制热裂纹的形成;通过采用改进的B.P神经网络模型,对多元逐步回归模型的预测能力进行了验证,其数值试验结果表明,热裂纹敏感系数同各元素含量呈现拟线性的多元函数关系。该逐步回归模型能够对激光焊接铝合金热裂纹的形成进行预测及控制,为激光焊接裂纹预测和控制的工程化应用提供一种比较可行的方法。On the basis of existed experiment, using the experimental data, employing the stepwise regression model and B.P neural network model, through the MATLAB platform of the theoretical calculation, the results of this study show that, the theoretical calculation and the experimental value has a good accord that the relative error below 8%, root mean square error within 0.05; meanwhile the research shows that oxygen is one of the main element to promote aluminum alloy welding hot crack growth , the element such as magnesium, inhibit the formation of hot crack; Using the improved B.P neural network model for verifying the prediction ability of multiple stepwise regression model, and the numerical computation results show that the relation between hot crack sensitive with each element content is presented quasi multivariate linear. The stepwise regression model on forecast and control of hot crack formation is a more feasible method in engineering application.

关 键 词：激光焊接 热裂纹 MATLAB 逐步回归 B P神经网络 

分 类 号：O141.4[理学—数学]