摘要
针对现代工业制造背景下的个性化机械零件,通常具有不规则和一定自相似性的分形特性,提出一种基于改进多重分形去趋势波动分析(MF-DFA)法的零件特征提取与缺陷识别方法。首先,选用三角形覆盖模块替代传统MF-DFA法中的正方形覆盖模块,解决传统MF-DFA法存在过度覆盖的问题,为零件图像缺陷识别提供更精准的数据;其次,利用改进MF-DFA法计算零件图像的多重分形谱;再利用核主成分分析(KPCA)方法提取零件图像的缺陷特征值;最后通过支持向量机(SVM)对零件缺陷进行识别。实验结果表明,三角覆盖二维MF-DFA算法能够准确提取零件特征,提高零件缺陷识别的准确率。
Aiming at personalized mechanical parts under the background of modern industrial manufacture,it usually have the fractal characteristics of irregularity and self-similarity,this paper presents a feature extraction and defect identification of parts method based on improved multifractality detrended fluctuation analysis(MF-DFA).First of all,the triangle covering module is used to replace the square covering module in the traditional MF-DFA method,addressing the problem of over-coverage in traditional MF-DFA method and provide more accurate data for defect identification of part images;Secondly,the modified MF-DFA method is used to calculate the multifractal spectrum of part images.Then,a method of kernelized principal component analysis(KPCA)is used to extract the value of defect features of part images;Finally,support vector machines(SVM)is adopted to identify parts defects.Experimental results show that the triangular cover two-dimensional MF-DFA algorithm can accurately extract part features and improve the accuracy of part defect recognition.
作者
何涛
王幸
王少东
王正家
盛文婷
HE Tao;WANG Xing;WANG Shao-dong;WANG Zheng-jia;SHENG Wen-ting(School of Mechanical Engineering,Hubei University of Technology,Wuhan 430068,China;Hubei Key Laboratory of Modern Manufacture Quality Engineering,Hubei University of Technology,Wuhan 430068,China)
出处
《组合机床与自动化加工技术》
北大核心
2021年第10期105-110,共6页
Modular Machine Tool & Automatic Manufacturing Technique
基金
国家自然科学基金项目(51275158)。
关键词
零件图像
去趋势波动分析
多重分形
特征提取
缺陷识别
part image
detrended fluctuation analysis
multifractality
feature extraction
defect identification
