电触头钎焊接头超声检测信号的集成神经网络分类

研究电触头钎焊接头超声无损检测中的缺陷分类问题，提出了一种新的集成神经网络分类方法。该方法分四步：频率不变性预处理，多分辩分析，特征量预处理，集成 Ｂ Ｐ神经网络分类。使用不同中心频率探头检测得到的缺陷信号首先通过预处理变换到一个等效的参考频率上，然后利用离散小波变换提取特征量。特征量被预处理后，输入到集成 Ｂ Ｐ神经网络分类器中分类。本文用２１３ 个超声检测信号测试了集成神经网络的性能。实验结果表明了频率不变性技术和集成 Ｂ Ｐ神经网络分类技术的有效性。

Classification of Ultrasonic Electrical Contact Braze Joint Inspection Sig nals by Integrating Neural Networks

The flaw classification issue in ultrasonic nondestructive testing of electrical contact braze joint is researched in this paper.A new classification method based on integrating neural networks is presented.The overall approach consists of four major steps,namely,preprocessing for frequency invariance, multiresolution analysis,features processing and classfication by integrating multiplayer prerception(BP) neural networks.The data are first preprocessed whereby signals signals obtained by using different transducer center frequencies are transformed to an equivalent reference frequency signal.Discriminatory features are then extracted by using the discrete wavelet transform.Finally,The features are preprocessed,then classified by integrating BP neural networks. A set of 213 ultrasonic inspection signals has been used to test the performance of the neural network.The experimental results obtained by using this aproach demonstrate the effectiveness of the frequency invariance processing technique and Integrating BP neural network.