This paper presents a predicted model of surface roughness of radial relief for resharpening end-mill. This model is constructed using a polynomial network. The major factors affecting grinding parameters are considered to be wheel spindle speed, feedrate, and grinding depth of cut. Experiments under specified conditions are deliberately designed and conducted to obtain the corresponding tested data for surface roughness that are used for training data of the proposed polynomial network. Consequently, a predicted model for surface roughness is established. Furthermore, a computer program in VB language is written based on this model. It can quickly calculate predicted values of surface roughness by simply inputting required cutting parameters. According to the experimental results, the developed polynomial network model shows high predicting capability on surface roughness of radial relief, and possesses promising potential in the application of predicting surface roughness in resharpening end-mill operation.