神经网络的焊球缺陷检测方法研究
时间:2019-09-15 15:17 来源:毕业论文 作者:毕业论文 点击:次
| [摘要]：倒装焊技术将芯片翻转，利用微凸点连接芯片与基底，由于该技术具有高自对准度、短互连等优点，可以大大提高芯片封装密度。倒装芯片逐渐成为一种主流封装技术，而得到广泛的应用。随着倒装焊技术向高单位密度、超细间距方向发展，芯片功率密度极大增加，散热更为困难，芯片热应力失配问题更为显著，应力集中更加明显，致使芯片焊球缺陷产生。本课题研究的目的在于利用神经网络对焊球缺陷进行检测方法。40861 神经网络( Neural Networks，简写为NNs）或称作连接模型（Connection Model），它是一种模仿动物神经网络行为特征，进行分布式并行信息处理的算法数学模型。这种网络依靠系统的复杂程度，通过调整内部大量节点之间相互连接的关系，从而达到处理信息的目的。 在现代工业生产中神经网络是运用的非常广泛的一种检测方法，有人工神经网络，生物神经网络，这里研究的是人工神经网络，提高产品的质量与可靠性它都有着重要的作用。首先对于神经网络检测技术本文进行深入的研究，说明神经网络（BP神经网络）检测技术的研究作用与发展情况,神经网络检测技术(BP神经网络)是目前非常先进的一种技术,应用领域更广,适用范围更宽。 利用超声波扫描（SAM）技术对倒装芯片进行检测，并对倒装芯片焊球SAM图像进行处理，主要包括：对图像进行预处理、图像的分割以及图像的特征提取，其中我们要利用到matlab软件进行执行，最后再对实验结果进行分析。matlab和神经网络检测是本设计的核心部分。 [毕业论文关键词]：焊球、倒装芯片、超声波检测、BP神经网络、图像特征提取 Research on detection method of solder bump defect based on the Neural Network Abstract: flip chip technology will flip chip, using micro convex point to connect the chip and the substrate, because the technology has high self alignment, short interconnect, etc., can greatly improve the density of the package of the chip. Flip chip is gradually becoming a mainstream packaging technology, which has been widely used. With flip chip technology to the high unit density, fine pitch direction, with the great increase in the density of the power of the chip, can dissipate heat more difficult, the chip thermal stress mismatch problem is more prominent, and stress concentration is more obvious, resulting in the chip solder ball defects produced. The purpose of this research is to use neural network detection method of defects in butt welding ball. Neural networks (neural networks, abbreviated as NNs) or is called the link connection model) model, it is an imitation of animal neural network behavior characteristics, of the distributed parallel processing algorithm of the mathematical model. This kind of network depends on the complexity of the system, through adjusting the relationship between the large number of nodes, so as to achieve the purpose of processing information. In neural networks in modern industrial production is very widely used as a detection method, artificial neural networks, biological neural networks and studied here is the artificial neural network, improve the quality and reliability of products it is a important role. First for neural network detection technology in this paper is in-depth study and description of neural network (BP neural network) detection technology research and development and neural network detection technology (BP neural network) is a technology is very advanced, wider application field, applicable scope wider. Using ultrasonic scanning (SAM) technique was used to detect the flip chip, and the flip chip solder ball Sam image processing, mainly including: the image pre processing, image segmentation and image feature extraction, which we should use MATLAB software to perform, the again on the experimental results were analyzed. MATLAB and neural network detection is the key part of the design. (责任编辑：qin) |