基于BP神经网络的金属裂纹声发射信号特征参数的提取  

Collection on characteristic parameters of emitted signals of metal cracking sound based on BP neural network

作　　者：毛汉颖[1] 成建国[2] 黄振峰[2] 

机构地区：[1]广西工学院汽车工程系,广西柳州545006 [2]广西大学机械工程学院,广西南宁530004

出　　处：《机械设计》2010年第2期84-87,共4页Journal of Machine Design

基　　金：国家自然科学基金资助项目(50465002);广西自然科学基金资助项目(桂科基0448014)

摘　　要：金属裂纹声发射信号特征提取是根据其进行故障诊断的关键,提出了BP神经网络和模式识别相结合的提取金属材料疲劳声发射信号特征的新方法,并利用美国PAC公司SAMOS声发射检测系统采集到声发射的各种参数,应用该方法选择出一些对分类识别最有效的特征参数;并采用可分离性判据进一步验证其正确性。The collection on characteristic parameters of emitted signals of metal cracking sound is the key to carrying out accordingly the fault diagnosis. With the combination of BP neural network and mode recognition a new method for collecting the characteristic parameters of emitted signals of metal materials fatigue sound has been put forward, and using the SAMOS sound emission detection system of American PAC Company to collect various kinds of parameters of sound emission. Some most effective characteristic parameters for sort recognition were selected by the use of that method, and its correctness was further verified by adopting the separability criterion.

关 键 词：声发射 特征提取 BP神经网络 模式识别 

分 类 号：TB556[理学—物理]