摘要倒装芯片技术是由对焊球的芯片和底板之间的互相连接，和往常的装配方式 对比下，更为的密集，所有才能渐渐的流行起来。但是在实际应用中，芯片和底 板之间热应力不同，有时会是芯片无法正常工作，两个焊球之间的距离也会缩小， 怎么样能够把芯片内的缺陷检验测量出来不是那么容易的，所以 FC 芯片缺陷的 检验测量方法的改进有着很大的价值。

超声波检测技术非接触无损检测方法，在我们日常生活实践中，在很多方面 都被使用到。这篇文章讲述了一种基于超声扫描图像的神经网络分辨问题的方 法，包含了焊球检验测量、焊球分离、信息获取和聚类讨论。但是因为有器材和 成本等方面的遗憾，相关系数方法的图像处理，然后提取的焊球和特征，最后进 行了聚类分析。72681

毕业论文关键词：倒装芯片焊球缺陷 超声波检测 相关系数 FCM 聚类

Defect Detection and Fuzzy Neural Network Classification of SAM Solder Ball

Abstract flip chip technology through the solder ball to the chip and the substrate to achieve interconnection, compared with the traditional packaging method, it has a higher density, and thus gradually become the mainstream。 But in the work, the chip and the substrate thermal expansion coefficient mismatch often result in failure of the solder joint fatigue and with the size of solder ball and bump pitches narrows further, how to realize the hidden between the chip and the substrate solder ball defects detection becomes more difficult, so the development and improvement of flip chip defects detection has important significance。

As a kind of nondestructive testing method for solder joint defects, ultrasonic testing technology has been widely used in the practical application。 This paper presents a method for automatic recognition of scanning acoustic images, which includes four phases: ball detection, ball segmentation, feature extraction and cluster analysis。 But due to the limitation of hardware conditions and the requirement for reducing the cost of, the correlation coefficient method image processing, then extract the feature of the solder ball, finally, cluster analysis was carried out using c-means algorithm the solder ball。

The experimental results show that the correlation coefficient method to deal with image, and then FCM clustering algorithm, the method is feasible。

key words: flip chip solder ball defects ultrasonic testing correlation coefficient FCM clustering

目 录

摘要Ⅰ

Abstract-Ⅱ

图清单-Ⅳ

表清单-Ⅳ

1 绪论 1

1。1 课题的研究背景及意义 1

1。2 倒装芯片及封装技术 1

1。3 倒装芯片缺陷检测方法 4

1。4 本论文的研究工作 5

2 SAM 缺陷检测系统 6

2。1 测试芯片 6

2。2 SAM 原理 6

2。3 小结 10

3 模糊神经网络 11

3。1 神经网络概述 11

3。2 FCM 聚类算法概述及原理 13

3。3 小结 15

4 信号处理 17

4。1 图像的获取及处理