Robustness Analysis of Gaussian Filters for Surface Texture of Additive Manufacturing Products

Recently, there has been increased interest in research in the direction of measuring and analyzing surface textures. This is due to the emergence of new processing methods that form anisotropic surface structures of products, primarily in additive manufacturing. One of the important and complex tasks is the areal filtering of the measurement results. The results obtained should be robust and do not take into account features of the texture of the multifunctional surfaces. Now the linear and robust regression Gaussian filters have been finally approved and practically used. This article analyzes the robustness of Gaussian filters for the height parameters of the texture. For this, four types of surfaces are modeled after turning, milling, grinding and polishing. Four types of texture features (slope, step, ledge, and valley) are modeled for each of the surfaces. Errors depending on the filtering method are calculated for seven height parameters for each of the four surfaces with four features. It has been found that the robust regression filter performs better than the linear Gaussian filter with all types of texture features. In this case, the error is 6.5% for the robust Gaussian filter and 24.2% for the linear Gaussian filter. The largest error was shown by the parameter Sz for the step. Therefore, research aimed at developing new robust filtering methods is relevant.