论文部分内容阅读
根据金属超声检测中缺陷脉冲回波为非稳态信号的特点，提出了一种基于小波变换和模式识别技术的缺陷定性分类方法.重点研究了利用小波变换提取反映缺陷性质的特征值以及运用模式识别技术对特征值进行缺陷定性识别的方法.为验证上述方法，设计了实验系统，同时对信号的采集、异常信号的剔除等问题进行了研究.利用实际焊接试样进行了实验，经小波变换提取缺陷特征值，然后采用BP(back propagation)神经网络，使缺陷的定性分类获得了较高的准确率.研究结果表明该方法可在一定程度上降低人为因素对缺陷定性识别的影响，获得较好的缺陷分类效果.
According to the character that the defect pulse echo is unsteady signal in metal ultrasonic testing, a method of qualitative classification of defects based on wavelet transform and pattern recognition is proposed, and the eigenvalues reflecting the nature of defect and the mode of operation Identification technology to characterize the value of the defect qualitative identification method.In order to verify the above method, the experimental system is designed, at the same time, the signal acquisition, abnormal signal rejection and other issues were studied using the actual welding samples were tested by the wavelet transform Then the BP neural network (BP) neural network was used to get the higher accuracy of the qualitative classification of the defects.The results show that this method can reduce the influence of human factors on the qualitative identification of defects to a certain extent, Good defect classification effect.