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

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.