An image-based approach to predict instantaneous cutting forces using convolutional neural networks in end milling operation

Cutting force detection can contribute to predicting the productivity and quality of end milling operations. Instantaneous cutting force prediction of digital twins in end milling operations should be near real-time and accurate. This paper proposes an image-based approach that can contain more useful information due to a higher dimension and simplify the complexity of computing geometric data. The cutter frame image (CFI) is utilized as one of the inputs of a convolutional neural network (CNN) to predict instantaneous cutting forces. Considering the convenience of capturing massive data, the approach uses cutting forces generated from a mechanistic force model instead of experimental cutting forces to train the CNN. The correlation coefficient R² value between predicted results and simulated results is 0.9999 and the average time cost per image is 0.057 s in a cutting condition, which validates the possibility to use the image-based method to predict instantaneous cutting forces accurately and efficiently in the digital twin.