Predicting of cutting forces in a micromilling process based on frequency analysis of sensor signals and modified polynomial neural network algorithm

Recently, with increasing demand for precise micro-components productions, the importance of micro machining processes is increasing in many fields, including the automotive, aerospace engineering, medical instruments and computer industries. However, compared with macro machining processes, it is very difficult to observe the machining process due to its low MRR (material removal rate), very small tool size, high speed spindle, and low sensor signals levels, etc. Micro tool dynamometer can be a solution for this; however, its applications are limited due to the expense, sensitivity, robustness, and workpiece size. Thus, in the present study, a useful indirect cutting force measurement method involving an acceleration sensor and current hall sensor is proposed. A series of experiments were performed on a precise micro machining stage. Measured signals were analyzed in the frequency domain after FFT (Fast Fourier Transform), and the results were compared with the cutting force components measured via the acceleration sensor and current hall sensor, respectively. From the results, it could be verified that the proposed indirect cutting force measurement method is a useful way to monitor the micro end-milling processes. Finally, to predict the cutting forces in micromilling processes, the modified polynomial neural network (PNN) and the back-propagation neural network are compared.