Wang Bingchen, Omatu Sigeru, Abe Toshiro
Division of Computer and Systems Sciences, Graduate School of Engineering, Osaka Prefecture University, Sakai, Osaka 599-853, Japan.
IEEE Trans Pattern Anal Mach Intell. 2005 Jun;27(6):919-28. doi: 10.1109/TPAMI.2005.121.
In this paper, a system is described that uses the wavelet transform to automatically identify the particular failure mode of a known defective transmission device. The problem of identifying a particular failure mode within a costly failed assembly is of benefit in practical applications. In this system, external acoustic sensors, instead of intrusive vibrometers, are used to record the acoustic data of the operating transmission device. A skilled factory worker, who is unfamiliar with statistical classification, helps to determine the feature vector of the particular failure mode in the feature extraction process. In the automatic identification part, an improved learning vector quantization (LVQ) method with normalizing the inputting feature vectors is proposed to compensate for variations in practical data. Some acoustic data, which are collected from the manufacturing site, are utilized to test the effectiveness of the described identification system. The experimental results show that this system can identify the particular failure mode of a defective transmission device and find out the causes of failure successfully.
本文描述了一种系统，该系统使用小波变换来自动识别已知有缺陷的传输设备的特定故障模式。在成本高昂的故障组件中识别特定故障模式的问题在实际应用中是有益的。在该系统中，使用外部声学传感器而非侵入式振动计来记录运行中的传输设备的声学数据。在特征提取过程中，一名不熟悉统计分类的熟练工厂工人协助确定特定故障模式的特征向量。在自动识别部分，提出了一种对输入特征向量进行归一化的改进学习向量量化（LVQ）方法，以补偿实际数据中的变化。从制造现场收集的一些声学数据被用于测试所描述的识别系统的有效性。实验结果表明，该系统可以识别有缺陷的传输设备的特定故障模式，并成功找出故障原因。
Zotero 插件