CSP焊点焊后残余应力分析与预测

该文建立了芯片尺寸封装(CSP)焊点有限元分析模型并对其进行了再流焊焊后残余应力应变分析。以焊点直径、焊点高度、焊盘直径和焊点间距为输入参数，焊后残余应力为输出参数进行了灵敏度分析。选取灵敏度分析结果中对残余应力影响显著的因子作为输入，建立了带动量项神经网络预测模型，对CSP焊点焊后残余应力进行了预测。结果表明，置信度为95%时，焊点直径、焊盘直径和焊点间距对CSP焊点残余应力影响显著，灵敏度从大到小的排序为:焊点直径>焊盘直径>焊点间距。所建立的带动量项神经网络预测模型对CSP焊点焊后残余应力预测最大相对误差为7.93%，平均误差为3.19%，实现了对CSP焊点焊后残余应力准确预测。

Analysis and Prediction of Residual Stress after Reflow Soldering of CSP Solder Joints

A finite element model of chip?scale?package (CSP) solder joints is established, the paper also analyzes the residual stress and strain after reflow soldering. Put solder joint height, solder joint diameter, solder joint pad diameter and solder joints pitch as input parameters, and the maximum residual stress after reflow soldering as output parameter to analyze the sensitivity of the CSP solder joints. Select the factors that have significant influence on residual stress, establish neural network with momentum term prediction model and predict the residual stress of CSP solder joints. The results show that the significant factors affect the residual stress of the CSP solder joints are the solder joint diameter, solder joint pad diameter and solder joint pitch when the confidence is 95%. The order is solder joint diameter, solder joint pad diameter and solder joint pitch. The maximum relative error of the neural network with momentum term prediction model for prediction residual stress after reflow welding of CSP solder joints is 7.93%, the average error is 3.19%, which complete the accurate prediction of the residual stress.