Tracking error reduction in CNC machining by reshaping the kinematic trajectory

In this paper, a method of reducing the tracking error in CNC machining is proposed. The structured neural network is used to approximate the discontinuous friction in CNC machining, which has jump points and uncertainties. With the estimated nonlinear friction function, the reshaped trajectory can be computed from the desired one by solving a second order ODE such that when the reshaped trajectory is fed into the CNC controller, the output is the desired trajectory and the tracking error is eliminated in certain sense. The proposed reshape method is also shown to be robust with respect to certain parameters of the dynamic system.