Prediction of thermal error for feed system of machine tools based on random radial basis function neural network

Thermal errors affect the accuracy of computer numerical control machine tools and are produced by the thermal deformation of machine components due to temperature difference between heat source and ambient temperature of the machine tools. At present, most of the literature does not consider the randomness of the influencing factors of thermal error, leading to inaccurate predictions of machine tool thermal error. In this paper, a new inverse random model is proposed through the combination of the stochastic theory, genetic algorithm, and radial basis function neural network (RBFNN), to predict thermal error while considering the randomness of influencing factors. The randomness index of influencing factors can be identified using the inverse random RBFNN (IR-RBFNN). Furthermore, through the combination of the stochastic theory, RBFNN, and the improved exponential moving average method with abnormal data elimination, a new forward random radial basis function neural network (FR-RBFNN) is established according to the identified influencing factor random index. The models are verified through experimental results on a ball screw system. Compared with the traditional methods, the experimental data show that the proposed method provides a more accurate description of thermal errors while incorporating the randomness of factors affecting thermal error.