Physics-informed experimental design for neural network-based cutting force model in end milling of carbon fiber reinforced polymer composites

Neural Network (NN)-based models showed robust capabilities while predicting cutting force during end milling of Carbon Fiber Reinforced Polymer (CFRP) Composites. NN models can effectively handle numerous factors and the inherent nonlinearities associated with the end milling of CFRP composites. However, a systematic approach is necessary to determine the quality and quantity of experimental cutting force data, which is critical for NN model training. This paper presents a physics-informed experimental design or Physics-Informed Advisor (PIA) that systematically determines cutting conditions for end milling experiments to generate training datasets. The PIA recommends the radial depth of cut and fiber orientation angle values to capture the process mechanics adequately and minimize experimental efforts. The cutting constant relationships of the Mechanistic force model are established using two independent Feed Forward Neural Networks (FFNN) trained using datasets from end milling experiments performed at cutting conditions determined using PIA. The predictions of the proposed approach are compared with the traditional Mechanistic force model presented in the literature and experimentally measured values. It is shown that integrating PIA with FFNNs enables an accurate estimation of cutting forces with reduced experimental efforts in generating training datasets.