Abstract: Soft push processing based on the edge broken effect to drive crack is a novel processing technique. The surface texture of machined Si₃N₄ ceramic is collected, and the gray level co-occurrence matrix (GLCM) is used to analyze the relationship among contrast, entropy, correlation and machined surface texture. The radial basic network and competitive network are employed to predict and classify the texture characteristics in different processing parameters. The relative error value of the predicted results can be controlled within 5%. The larger the contrast and entropy are, the smaller the correlation is, and the greater the classification level is, the worse the surface roughness is. The processing quality is analyzed by exploring the effects of different process parameters on texture feature. With the increase in feed rate or groove depth, the machined surface quality is worse. With the increase in flange thickness, the machined surface quality is gradually poor. The machined surface quality can be improved when flange thickness is over 2.5 mm boundary point.

Key words: surface and interface of matterials, Si₃N₄ ceramics, texture characteristics, gray level co-occurrence matrix, neural network