论文部分内容阅读
超声波检测的脉冲回波信号在时频域中含有丰富信息,为有效提取缺陷特征相关的信息并对缺陷进行分类,文章提出一种Jmax-Fisher多特征优选的方法.首先对超声波脉冲回波信号在时域、频域及小波域中提取多维特征构成多特征提取技术框架;然后计算单维特征Fisher判据函数以获得不同特征维数下的最优特征组合;再进行Fisher降维,采用离散距离比作为指标获得最优特征维数,确定最优特征组合进行缺陷分类.实验证明,同常规的方法相比,该方法在缺陷分类识别上具有更高的准确率.
In order to effectively extract the information related to defect characteristics and classify the defects, the pulse echo signal of ultrasonic detection is rich in information in the time-frequency domain.This paper presents a Jmax-Fisher multi-feature optimization method.Firstly, the ultrasonic pulse echo signal In the time domain, frequency domain and wavelet domain to extract multi-dimensional features constitute a multi-feature extraction technology framework; and then calculate the one-dimensional feature Fisher criterion function to obtain the optimal feature combinations of different feature dimensions; then Fisher dimension, using discrete Distance ratio as the index to obtain the optimal feature dimension and determine the optimal combination of features for defect classification.Experiments show that compared with the conventional method, this method has higher accuracy in identifying the defect classification.