Cutting force prediction between different machine tool systems based on transfer learning method

Milling force can generally be predicted from orthogonal cutting data. However, the accuracy of the predicted result is largely affected by the dynamic properties of the machine tool system, and a large number of repeated milling tests are required for the different types of machine tool to reduce the prediction error. A data-driven method, which could predict the cutting force of different machine tools through only a small number of tests, is proposed in this paper based on the transfer learning method. First, the cutting force coefficients are obtained through orthogonal cutting test. Then, the influence of the dynamic properties of the machine tool system is considered by introducing a correction coefficient to improve the conversion accuracy from orthogonal cutting to helical milling process. After obtaining the cutting force of the source machine tool with sufficient cutting data, the regional adaptive method combining with neural network is utilized to predict the cutting force of the target machine tool with only a small amount cutting data. Finally, a series of milling tests are carried out on different machine tools to verify the accuracy of the proposed method. The results indicate that the developed method can predict the milling force with a high accuracy.