Study on the Computing Method for Surface Roughness Coefficient Based on a BP Neural Network

This paper established a computing model of surface roughness coefficient (JRC). To begin with, fractional Brownian functions were adopted to generate random fractal curve, which compared with standard Barton’s curve. Slope root mean square and curvature root mean square were calculated for test samples. Based on neural network algorithm and particle swarm algorithm, BP-PSO calculation program were constructed. Test samples were used to train the neural network and JRC value under two-dimensional condition were calculated. Since random error and systematic error are considered, the method in this paper is easier to acquire accurate result compared with traditional explicit quantitative relationship between geometrical statistical parameters and JRC value. By scanning three-dimensional joint surface, point cloud data were acquired. Spatial distribution of effective contacting area of joint surface was analyzed according to different shearing direction, and the roughness coefficient computing method was reformed according to three-dimensional condition. The calculating model in this paper offers a new way to evaluate JRC value.