The process parameters affecting the material removal rate in sand belt grinding were obtained through analyzing Preston equation.Based on experience in sand belt grinding and under certain experimental device conditions,the four main process parameters affecting the quality and efficiency of blade processing were identified:sand belt linear velocity,normal contact force between workpiece and sand belt,sand belt grinding particle size and workpiece feed speed.Then,through single-factor experiments and L18(34) orthogonal experiments,the influence law of each process parameter on the surface roughness and material removal rate of the workpiece was determined,and the reasonable range of each process parameter was comprehensively considered.A nonlinear prediction algorithm based on multilayer perceptron was used to train and fit the experimental data to obtain a prediction model for surface roughness and material removal rate.Finally,based on the validation experiment of sand belt grinding and the performance comparison between models,the predicted absolute percentage errors of the model for surface roughness and material removal rate are determined to be between 2.30% and 6.47%，1.00% and 6.67%,respectively,and the prediction computation time of this model is shorter,the robustness and rapidity of the model in predicting experimental results through process parameters are demonstrated.