多缺陷特征提取方式的金属棒材分析系统  

Metal bar analysis system based on multi-defect feature extraction method

作　　者：石涛[1] 王国富[1] 叶金才[1] 张法全[1] SHI Tao;WANG Guo-fu;YE Jin-cai;ZHANG Fa-quan(School of Information and Communication,Guilin University of Electronic Technology,Guilin 541004,China)

机构地区：[1]桂林电子科技大学信息与通信学院,广西桂林541004

出　　处：《桂林理工大学学报》2022年第3期742-748,共7页Journal of Guilin University of Technology

基　　金：国家自然科学基金项目(61861011)。

摘　　要：超声波法是无损检测领域常用方式之一,目前金属棒状材料的超声检测方法存在缺陷特征提取方式单一、质量检测精度受限的问题。为了增加缺陷特征提取方式的多样性,以提高金属棒材质量检测精度,设计了一种基于超声检测方式的金属棒状材料缺陷分析系统。首先搭建支持信号采集、波形存储、回波上传的硬件平台;然后,提出了一种基于神经网络的金属棒状材料缺陷分类方法和相应缺陷类型的特征提取方法;通过实验测试对试验模型和所提方法进行验证。结果表明:该系统对金属棒材缺陷特征的提取误差低于2%,具有较高的金属棒材缺陷分类准确度和金属棒材缺陷特征提取精度,提高了金属棒材质量检测的可靠性和准确性,可为金属棒材质量检测提供初步有用的参考信息。Ultrasonic testing is one of the commonly used methods in the field of non-destructive testing.At present,in the detection method of metal bars there is a single defect feature extraction method,which wsually leads to the problem of limited quality inspection accuracy.In order to increase the diversity of defect feature extraction and improve the accuracy of metal bar quality inspection,a metal bar defect analysis system based on ultrasonic testing was designed.First,a hardware platform was built,to support signal acquisition,waveform storage,and echo upload.Then,a neural network-based defect classification method for metal bars and a feature extraction method for corresponding defect types are proposed.Finally,the experimental model and the proposed method were verified in experimental tests.The results show that in the system there are errors less than 2%for the extraction of metal bar defect features.The high metal bar defect classification accuracy and metal bar defect feature extraction accuracy improve the reliability of metal bar quality inspection and accuracy.It can provide useful information for the quality inspection of metal bars.

关 键 词：超声波 无损检测 缺陷分析 神经网络 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]