一种基于人工神经网络模型的无铅无卤回流焊优化工艺

Process optimization of Pb-free and halogen-free reflow soldering based on neural network

为优化大规模工业化应用的无铅无卤回流焊工艺参数,构建无铅无卤回流焊工艺优化神经网络模型,其中输入层参数为合金牌号、预热起始温度、预热终止温度、预热时间、焊接温度和焊接时间,输出层参数为焊点推拉力和表面绝缘电阻,进行模型的预测验证和工艺优化应用,并对优化后的合金焊点进行金相切片、推拉力试验、表面绝缘电阻等测试。结果表明：模型具有较高预测精度和较强工业实用价值,焊点推拉力相对训练误差为1.1%-2.6%、表面绝缘电阻相对训练误差为1.3%-3.5%;在神经网络模型预测的最佳工艺参数下,合金焊点的推拉力达30 N、表面绝缘电阻值达1012Ω。

In order to optimize Pb-free and halogen-free reflow soldering process parameters for large-scale industrial application,the neural network model for process optimization of Pb-free and halogen-free reflow soldering was built,with alloy grade,preheating starting temperature,preheating final temperature,preheating time,soldering temperature and soldering time as input parameters,and with push-pull force and surface insulating resistance of joints as output parameters. And the model was verified and used for process optimization. Moreover,the microsection,push-pull force and surface insulating resistance of joints were tested. The results show that the neural network model is of high precision and practicability,and the prediction error of push-pull force is between 1.1% and 2.6%,and the prediction error of surface insulating resistance is between 1.3% and 3.5%.By the optimum process parameters,the push-pull force of joint reachs 30 N and surface insulating resistance reaches 1012Ω.