China Mechanical Engineering

Previous Articles     Next Articles

Logarithmic Wavelet Spectrum Method for Computing Fractal Dimension of Machined Surface Profiles

Zhang Xueliang;Wang Yusong;Wen Shuhua;Fan Shirong;Chen Yonghui;Lan Guosheng   

  1. Taiyuan University of Science and Technology, Taiyuan, 030024
  • Online:2016-12-10 Published:2016-12-15
  • Supported by:
     

机械加工表面轮廓分形维数对数小波谱计算方法

张学良;王余松;温淑花;范世荣;陈永会;兰国生   

  1. 太原科技大学,太原,030024
  • 基金资助:
    国家自然科学基金资助项目(51275328);山西省自然科学基金资助项目(201601D011062)

Abstract: A multi-scale wavelet decomposition for surface profiles was achieved herein by the multi-scale analysis capability of wavelet; and a logarithmic wavelet spectrum method for calculating the fractal dimensions of machined surface profiles and the effective decomposition scale concept were proposed. Fractal profiles with different fractal dimensions and different sampling intervals were generated by M-B function. The fractal dimensions of simulated profiles whose dimensions were known were calculated by logarithmic wavelet spectrum method, which was compared with the calculation results of other five kinds of methods such as the power spectral density (PSD) method. The results show that logarithmic wavelet spectrum method may process multi-scale fractal features nicely, and it has the highest accuracy with errors less than 0.15% when sym4 wavelet is adopted. Finally, logarithmic wavelet spectrum method was applied to calculate the fractal dimensions of actual machined surface profiles and the results illustrate its practicability.

Key words: machined surface profile, fractal dimension, wavelet decomposition, effective decomposition level, logarithmic wavelet spectrum

摘要: 为了提高接触表面的建模精度,利用小波的多尺度分析能力,对表面轮廓进行多尺度小波分解,提出了计算机械加工表面轮廓分形维数的对数小波谱法以及有效分解尺度概念,并认为轮廓只在有效分解尺度上具有分形特征;通过M-B函数模拟生成不同分形维数、不同采样区间的分形轮廓;应用对数小波谱法计算了模拟轮廓的分形维数,进而与功率谱密度法(PSD法)等5种方法的计算结果进行了分析比较,结果表明:对数小波谱法能很好地处理分形的多尺度特征,并且选用sym4小波时计算精度最高,误差在0.15%以内;最后应用对数小波谱法对一实际机械加工表面轮廓分形维数进行了计算,说明了其实用性。

关键词: 机械加工表面轮廓, 分形维数, 小波分解, 有效分解尺度, 对数小波谱

CLC Number: