文档介绍：神经网络在电子元器件无铅微连接润湿特性研究中的应用
摘要
、80年代以来发展起来的一种处理复杂非线性问题十分有效的手段。它模拟人脑的特征,具有自组织、自学习、自适应、容错性等特点,被广泛应用于模式识别、预测及模糊控制等领域。焊接焊料润湿力的预测实质上是一种依据焊接工艺参数以及实验得出的实际润湿力建立一个网络然后再进行预测的问题,因此研究神经网络在电子元器件无铅微连接中的应用具有非常现实的意义。神经网络的结构类型很多,大致可以分为前向网络和反馈网络。在软测量中常用多层前向网络,其中最常用的两种是BP神经网络和RBF神经网络。
因此本文首先介绍人工神经网络的理论,主要阐述人工神经网络的结构和算法。然后再在MATLAB这一面向科学与工程计算的高级语言学习和掌握的基础上,利用其中的人工神经网络工具箱,依据实据实验测得的焊接工艺参数,分别建立BP和RBF网络模型。为了避免出现神经元饱和这一问题,在对网络进行训练之前必须对数据进行处理,以消除原始数据形式不同所带来的不利,因此对原始数据进行了归一化处理。然后利用处理过后的数据训练网络,在这过程中需要反反复复多次调节相关参数,进而确立最佳网络。最后用已建的人工神经网络模型,在已知焊接工艺参数的情况下,对焊料的润湿力进行预测来验证网络的正确性。
关键词:人工神经网络,BP网络,RBF网络,润湿力,预测
PLICATION OF WORK IN THE STUDY OF MICRO CONNECTION AND WETTING CHARACTERRISTICS OF LEAD-FREE PONENTS
ABSTRACT
The artificial work was generated in the 20th century, years, and developed to a very effective means to deal plex nonlinear problems since the 1980s. It simulates the characteristics of the human brain, with characteristics of anizing, self-learning, adaptive, fault tolerance, etc., and are widely used in pattern recognition, prediction and fuzzy control. The prediction of welding solder wetting force is essentially a basis welding process parameters and the actual experimental wetting force to establish work and then to predict, so the study of work in the lead-free micro-connection of ponents is of very realistic sensible. There are lots of kinds of work structure and can be divided into the work and work. In the soft monly used in works, including the two mon BP work and RBF work.
This article first introduces the theory of artificial works, mainly on the structures and algorithms of artificial works. Then on the basis of learning and mastering the high-level language of MATLAB-oriented scientific and puting, and establish the BP and work model respectively based on substantiated experimentally measured welding parameters using artificial work toolbox. In order to avoid the problem of neuron saturation, the data before training work must be processed to