基于相关性分析和SSA-BP神经网络的铝合金电阻点焊质量预测  

Quality prediction of aluminum alloy resistance spot welding based on correlation analysis and SSA-BP neural network

作　　者：董建伟 胡建明 罗震[1] DONG Jianwei;HU Jianming;LUO Zhen(School of Materials Science and Engineering,Tianjin University,Tianjin,300350,China)

机构地区：[1]天津大学,材料科学与工程学院,天津300350

出　　处：《焊接学报》2024年第2期13-18,32,I0003,I0004,共9页Transactions of The China Welding Institution

基　　金：国家自然科学基金资助项目(52075378)。

摘　　要：基于电阻点焊过程中工艺信号特征,在不同间距、不同间隙和不同间距与间隙3种条件下,引入相关性分析方法分析工艺信号与熔核直径之间的相关性,并建立基于麻雀搜索算法-BP神经网络(sparrow search algorithmback propagation neural network,SSA-BP)的电阻点焊质量预测模型,将功率、焊接电流、焊接电压和动态电阻作为预测模型输入特征.结果表明,经麻雀搜索算法优化后的BP神经网络在测试集上的决定系数R2、均方误差(meansquare error,MSE)、均方根误差(root mean square error,RMSE)和平均绝对误差(mean absolute error,MAE)分别为0.95,1.55,1.24和0.90,均优于BP模型.获得了功率、焊接电流、焊接电压和动态电阻与熔核直径的映射关系,可为焊接的工艺参数设计提供依据.Based on the characteristics of the process signals in the resistance spot welding process,three working conditions of different spacing,different gaps and different spacing and gaps are analyzed,and correlation analysis is introduced to extract the correlation between the process signals and the diameter of nugget.A resistance spot welding quality prediction model based on Sparrow Search Algorithm-Back Propagation Neural Network(SSA-BP)was established,and power,welding current,welding voltage and dynamic resistance are used as input features of the prediction model.The results show that the BP neural network optimized by the sparrow search algorithm outperforms the BP model on the test set with R2,MSE,RMSE and MAE of 0.95,1.55,1.24 and 0.90,respectively.It is also determined that there exists a mapping relationship between power,welding current,welding voltage and dynamic resistance and the diameter of the nugget,which provides a basis for the design of process parameters for welding.

关 键 词：电阻点焊 熔核直径 麻雀搜索算法 BP神经网络 相关性分析 

分 类 号：TG453.9[金属学及工艺—焊接]