基于GASA-BP的轧辊内部缺陷检测方法  

GASA-BP based method for detection of roller internal defects

作　　者：王蕾[1,2] 陶海然 郭钰瑶 张泽琳 夏绪辉 Wang Lei;Tao Hairan;Guo Yuyao;Zhang Zelin;Xia Xuhui(Key Laboratory of Metallurgical Equipment and Control of Ministry of Education,Wuhan University of Science and Technology,Wuhan 430081,China;Hubei Key Laboratory of Mechanical Transmission and Manufacturing Engineering,Wuhan University of Science and Technology,Wuhan 430081,China)

机构地区：[1]武汉科技大学冶金装备及其控制教育部重点实验室,湖北武汉430081 [2]武汉科技大学机械传动与制造工程湖北省重点实验室,湖北武汉430081

出　　处：《武汉科技大学学报》2023年第5期368-375,共8页Journal of Wuhan University of Science and Technology

基　　金：国家自然科学基金资助项目(52275503);湖北省杰出青年基金项目(2023AFA092);湖北省支持企业技术创新发展项目(2021BAB002,2022BAD102).

摘　　要：为了准确识别轧辊内部缺陷,采用超声波探伤仪对轧辊进行检测,并提出一种融合灰狼算法、遗传算法、模拟退火算法和BP神经网络的超声检测信号分类方法。对超声检测信号进行小波降噪,采用自适应灰狼优化阈值法确定最优的小波阈值;根据轧辊内部不同缺陷检测信号的时域和频域信息差异,从三层、四层和五层小波包分解中选择最优的分解层数,以实现对去噪检测信号的最佳处理,在此基础上构建检测信号的能量特征向量;采用经过遗传模拟退火算法优化的BP神经网络模型(GASA-BP)进行特征向量的识别,进而完成轧辊内部缺陷的分类。实验结果表明,该模型的识别准确率可达到97.41%,并且运算速度得到极大提高,能满足大部分企业对轧辊缺陷分类精度及效率的要求,为轧辊的进一步检测、修复及再制造提供了可靠信息。In order to accurately identify the internal defects of rollers,the ultrasonic flaw detector was used,and a classification method for ultrasonic detection signals was proposed by combining grey wolf algorithm,genetic algorithm,simulated annealing algorithm and BP neural network.Firstly,the ultrasonic signal was denoised via wavelet tranform,and the optimal wavelet threshold was determined by using an adaptive grey wolf optimized threshold method.According to the time-domain and frequency-domain information of different defect signals inside the rollers,the optimal layer number was selected from the wavelet packet decomposition of three,four and five layers to achieve the best processing effect.On this basis,the energy feature vector of the detection signal was constructed.Then BP neural network model optimized by genetic simulated annealing algorithm(GASA-BP)was used to identify the feature vector so as to classify the internal defects of the rollers.Experimental results show that the recognition accuracy of this model can reach 97.41%,and the computing speed is greatly improved,which can meet the requirements of most enterprises for the classification accuracy and efficiency of roller defects,and provide reliable information for further detection,repairment and remanufacturing of rollers.

关 键 词：轧辊内部缺陷 超声检测信号 小波降噪 自适应灰狼优化阈值法 GASA-BP 

分 类 号：TP391.7[自动化与计算机技术—计算机应用技术] TG333.17[自动化与计算机技术—计算机科学与技术]