Experiment on diamond defect detection was conducted using ultra PAC system. A new idea of pattern recognition by means of wavelet packet neural network analysis was introduced. The idea was based on the feature extraction and analysis of defect signals, as well as on the network modeling of defect qualitative recognition. The results of experiment show that the wavelet packet analysis could make best use of the information in time-domain and in frequency-domain of the defect echo signals, partition further and analyze the high-frequency part which had not been subdivided by multi-resolution analysis, and choose the interrelated frequency bands to make it suited to signal spectrum. Thus, the time-frequency resolution was improved. The good local amplificatory property of the wavelet neural network and the study characteristic of multi-resolution analysis can help to achieve higher accuracy of the qualitative classification of synthetic diamond defects.%应用超声波探伤仪系统对合成大颗粒金刚石缺陷进行检测,针对缺陷信号特点提出利用小波包分析提取缺陷特征值,应用小波神经网络进行模式识别的方法,实现了从检测到的超声信号中提取出反映缺陷性质的相关信息,并通过这些信息对其进行分析,建立了网络模型以实现缺陷定性识别.实验结果表明,小波包分析能够挖掘利用缺陷回波信号时域和频域的信息,通过多层次划分频带,使在多分辨分析过程中未进行划分的高频区间再次分解,还可依据小被分析信号特征自适应挑选相对应的频带,达到和信号频谱相互配合,进而达到使时-频分辨率显著提高的效果,可见小波神经网络的良好局部放大特性和多分辨率学习特性,可使合成金刚石缺陷的定性分类获得较高的准确率.
展开▼
