基于小波包-BP网络的超声检测缺陷类型识别

Faults identification based on wavelet packet and BP neural network for ultrasonic testing

针对A型反射超声波检测仪难以准确识别缺陷类型的问题,探讨了基于小波包和BP神经网络相结合的超声检测缺陷类型识别方法。对检测的多组超声缺陷信号分别进行3层小波包分解,提取小波包频谱能量特征,归一化后构造了各缺陷信号的特征向量,并分别组成训练样本集和测试样本集,用于3层BP神经网络的训练和网络识别效果检验。实验结果表明该方法能准确快速地识别出超声检测缺陷类型。

As A-mode pulse ultrasonic testing instrument cannot accurately identify faults,a method based on wavelet packet analysis and back propagation neural network was proposed to identify faults.Firstly groups of flaw signals were decomposed with wavelet packet analysis for three layers,the frequency spectrum energy characteristics were extracted from the reconstructed wavelet packet nodes coefficients,and the characteristics were normalized to compose the eigenvector.Then eigenvectors were classified to training the samples and the testing samples to trained and test the BP neural network.The experiment results show that the proposed method is effective and feasible.The trained BP neural network can accurately identify the ultrasonic faults.