微电子封装中金丝球焊工艺质量预测模型设计与研究  

Design on Quality Prediction Model of Gold Wire Bonding Process in Microelectronics Packaging

作　　者：关晓丹 孙燕 赵鹏 

机构地区：[1]北华航天工业学院电子与控制工程学院,河北廊坊065000

出　　处：《北华航天工业学院学报》2018年第1期1-3,8,共4页Journal of North China Institute of Aerospace Engineering

基　　金：北华航天工业学院青年基金(KY-2014-15);河北省科技厅指导项目(15210206);河北省教育厅指导项目(ZC2016058)

摘　　要：微电子封装过程中焊接参数的选择直接影响金丝球焊接的质量,本文通过正交试验的方法将不同的参数值进行组合、试验并通过方差分析法从诸多参数中确定影响键合质量的主要参数。针对金丝球焊质量与其影响参数之间复杂的非线性耦合关系的特点,本文建立了BP神经网络,确定神经网络的各结构参数,建立了引线键合模型。通过对模型预测结果进行验证,表明所建模型具有较高的精度,能够准确地反映焊接质量综合指标的变化趋势。The selection of welding parameters directly affects the quality of gold wire welding in microelectronic packaging process. In this article, the parameters affecting the quality of wire bonding are analyzed by orthogonal testing with the methods of variance analysis and tests. Based on the results, parameters that have a major impact on the quality of wire bonding are optimized. Since the relationship is complicated and non-linear between the impacting parameters and bonding quality, this article introduces a neural network algorithm of BPNN to build a model describing it. The structural parameters of the neural network are identified and a quality predic- tion model of wire bonding is established. Results show that this proposed model has higher precision and it can accurately reflect the trends of the bonding quality indicators.

关 键 词：微组装 金丝球焊 正交试验 剪切力试验 神经网络 

分 类 号：TN405[电子电信—微电子学与固体电子学]