Abstract
The experimental results for silicon carbide (SiC) wheel with fine grit size grinding of silicon carbide (Si 3 N 4 ) revealed that the grinding parameters affect not only the ground silicon nitride surface roughness, but also the degree of surface damage. There exists complex non-linear relationship between the grinding parameters and surface quality. Better surface roughness doesn't surely mean less surface damage. A method of prediction of grinding quality based on support vector regression is then presented according to the condition of small samples. The result shows the prediction accuracy based on this method is obviously higher than neural network, which provides an effective way for optimizing the grinding parameters to ensure the grinding quality as well as grinding efficiency while grinding of silicon carbide using conventional abrasive.