Neural Network Based Nonlinear Autoregressive Models for CNC Machine Process

Abstract

Tool wear monitoring and control in the machining operations is necessary to enhance the productivity and decrease the operation cost. The identification of CNC machine process model and its control is so difficult due to its high non-linearity. Therefore, neural networks (NNs) one of the non-linear identification techniques, have been applied in addition to system identification field for the identification and control of nonlinear systems. In this paper, auto-regression recurrent neural network model structures NNARX and NNARMAX is proposed for CNC machine modeling with its cutting condition and vibration signals as input to obtain an accurate nonlinear system model for prediction of tool wear and surface roughness. Finally, the modeled neural network model structures for prediction of tool wear and surface roughness is validated with the observed tool wear and surface roughness for the accuracy analysis of modeled neural network model structures.