小波神经网络的数据压缩技术在超声自动探伤系统中的应用  

The Application of Wavelet Neural Network in Data Compression of Automatic Ultrasonic Testing System

作　　者：赵秀粉[1] 殷国富[1] 苏真伟[1] 王应国[1] 马杰[1] 

机构地区：[1]四川大学制造科学与工程学院,成都610065

出　　处：《无损检测》2007年第12期716-719,共4页Nondestructive Testing

基　　金：四川省科研生产协作项目资助(06H020)

摘　　要：为了实现对大型回转体零件内部缺陷的检测与识别,研制了超声波自动检测系统。系统主要完成超声信号的采集和处理、数据的实时存储、缺陷的在线分析与识别等功能。要实现缺陷的在线检测与识别,必然需要大量的原始数据,为了减少数据的存储量,通过小波神经网络提取相应的权重因子,构成小波基的尺度参数和与之对应的平移参数,实现缺陷有用信息的压缩;在缺陷数据重构中,利用上述特性参数并结合信号的特征值,对信号进行拟合。解决了缺陷检测现场大量数据的保存问题,为缺陷的进一步识别提供了基础。An online and automatic ultrasonic testing system for detecting the inner flaws of lager cylindrical parts was developed.The main function of the system was to receive and process the ultrasonic signals,store testing data,and online analysis and inner flaws recognition.In order to reduce the storage of data,a technique of data compression and data reconstruction was introduced.Data compression was accomplished by abstracting the characteristic parameters such as the weight coefficients,scale parameters and move parameters.On the other side,signal reconstruction was realized by combining the above characteristic parameters and the characteristic value of signal.The problem of the save of huge data and the shortage of the specimen for the neural network learning of flaw-recognition model were solved.

关 键 词：超声波检测 小波神经网络 数据压缩 数据重构 

分 类 号：TG115.28[金属学及工艺—物理冶金]