各向异性自相关长度对三维计算机模拟表面参数的影响

Effects of Anisotropic Autocorrelation Length on Parameters of 3D Computer Simulation Surface

目的研究计算机模拟过程中不同自相关长度对重构后表面表征参数的影响。方法采用新FFT方法,保持特定参数(高度均方根、偏态、峰值)不变,对不同自相关长度(3、10、30、100)在两个方向上的不同组合,进行了共16次三维粗糙表面形貌计算机模拟实验,并从不同方向自相关长度组合的角度,对与形貌特征关联度较高的参数影响进行了分析。结果算数平均峰曲率、峰密度、均方根斜率总体上和两个方向上的自相关长度呈负相关,而最快下降自相关长度与两个方向上的自相关长度呈正相关,纹理高宽比与两个方向的自相关长度是否接近有关,其中峰密度、最快下降自相关长度都对不同自相关长度中的较小自相关长度比较敏感,而对较大自相关长度的变化反应较小,算数平均峰曲率、均方根斜率则表现出相同的较高敏感性。结论表面表征参数中的高度分布参数和自相关长度没有关系,而结构比例方面的参数和自相关长度关联显著。

The work aims to study the effects of autocorrelation length on reconstructed surface characterization parameters in the process of computer simulation, and lay a foundation for exploring the relationship among different surface characterization parameters, and also the relationship between autocorrelation function（ACF） and surface properties of usability. Computer simulation surface test of 3D rough surface morphology was carried out 16 times for various combinations of autocorrelation length 3, 10, 30, 100 in two directions using the same specific parameters（RMS, Skewness, Kurtosis） in the new FFT method proposed by the author. The effects of parameters which were correlated with morphological features were analyzed from theperspective of combinations of autocorrelation length in different direction. Mean summit curvature, summit density and RMS slope were negatively correlated with autocorrelation length in two different directions. However, the fastest descending autocorrelation length showed a positive correlation, and texture aspect ratio was related to proximity of autocorrelation length in two directions. Summit density and the fastest descending auto-correlation length were both sensitive to smaller autocorrelation length among different autocorrelation lengths, but less sensitive to changes in larger autocorrelation length. Mean summit curvature and RMS slope showed the same high sensitivity. Height distribution parameters among surface characterization parameters are not correlated with autocorrelation length on computer simulation surface, while structure proportion parameters are significantly correlated with autocorrelation length.