Abstract: In order to accurately predict the surface roughness and material removal rate of a multi-head screw rotor after simultaneous grinding with double abrasive belts, a prediction method based on a combined whale optimization algorithm-radial basis function (WOA-RBF) neural network is proposed. Comparison with RBF-based and convolutional neural networks (CNN)-based prediction models shows that the average relative error of prediction of the proposed model is lower than that of the RBF and CNN models, while the root mean square error and coefficient of determination are better than those of the comparators. The results of the one-factor prediction show that the surface roughness of the screw rotor increases with the increase of the main cylinder pressure and the abrasive belt grit size, decreases with the increase of the abrasive belt tension, and decreases and then increases with the increase of the abrasive belt linear speed. The material removal rate increases with the increase of main cylinder air pressure and belt line speed and belt grain size, and decreases with the increase of belt tension. The material removal rate of the ground workpiece is strongly influenced by device 1, while the surface roughness of the workpiece is strongly influenced by device 2. The method can provide a reference for the prediction of the grinding quality of other complex shaped workpieces.