This research investigates the feasibility of designing a new force estimator for CNC machine tools. For the purpose of estimating the machining forces accurately, the dynamic behaviors of various CNC machine subsystems (such as feed drive, spindle drive), which are directly subjected to these periodic forces, are modeled in this study. Since the physical nonlinearities like friction are quite dominant in such systems; structured neural networks, which have remarkable nonlinear system modeling capabilities, are utilized as the fundamental design tool. All structured neural networks presented in this study take advantage of the harmonic nature of the machining forces and thus exclusively employ recursive discrete Fourier transform to model the effects of these forces efficiently. Furthermore, they are shown to outperform other estimation paradigms including Luenberger-style disturbance force/torque observers. Due to certain physical limitations imposed on spindle and feed drive systems, a model reference based approach is proposed in this research to design a general force estimator. The performance of this overall topology is evaluated under extreme conditions. Both its accuracy and bandwidth are found to be sufficient for most CNC machine tool applications including adaptive control, machine diagnostics, and process monitoring. |