This paper studies the optimization of grinding. Database of grinding technology is established in expert system in advance. Optimization is started from queries of grinding conditions which is produced through inference of expert system after entering the work conditions and completion of quality target as the initial parameters. At this stage only the suitable ranges of grinding parameters can be provided. Next, using the Taguchi method on basis of the completion of quality target gets the best values of grinding parameters. This paper studies the case of a rectangular sheet of low carbon steel in surface grinding. The controllable parameters include grinding in-feed depth, grinding depth, redressing depth, and flow rate of grinding fluid, the objective is to optimize the minimum surface roughness.