塑封焊球阵列焊点三维形态预测及其"整体"近似优化设计

针对传统优化技术进行焊点三维形态优化时费时耗力且浪费计算资源等缺点，提出采用“整体”近似优化技术，结合线性最小二乘方法，BP（Back Propagaiton）神经网络，在整个设计变量空建立问题函数近似面，利用该函数近似面来求取目标函数最大值，并得到相应的设计变量值，达到爆点三维形态优化设计的目的，最后对线性最小二乘模型，BP神经网络拟合非线性函数的能力和近似优化能力进行了比较讨论，结果表明，利用“整体”近似优化技术来优化设计焊点形态是简便可行的，且BP网络模型拟合“整体”近似函数面比线性回归模型具有更高的精度，所得的近似优化结果更理想。

3-D Shape Prediction and Global Approximate Optimization of Plastic Ball Grid Array(PBGA) Solder Joints

The exact optimization of CAD system is computationally expensive and time wasted. In order to improve the efficiency, approximation techniques that reduce the number of function evaluations were used. The least square linear regression models and BP neural network models were developed to fit the approximate representation of the fatigue life as a function of the two design variables in the space of possible pad sizes and effective solder ball diameters. These approximate functions were used to determine the optimal solution. For the purposes of demonstrating the techniques, the PBGA225 assembly was used as an example to maximize the solder joint reliablility through optimal shape design. It is verified the global approximate optimization techniques are powerful and easy to use. The BP model provides a better function approximation and a more accurate optimal solution than the linear regression model.