Discrete-time modeling, control and signal processing for AC drives and motion servo system diagnostics

This thesis presents contributions made in the fundamental areas of discrete-time domain modeling, control, estimation and signal processing methods for AC drives and motion servo electromechanical actuator system diagnostics.;High fundamental frequency operation is becoming increasingly important as it allows for high density electrical and electromechanical power conversions. However, instability of the closed-loop current control has been one of the major obstacles in its use. In this research, a detailed discrete-time domain model is derived for general AC machines including the effects of the pulse-width-modulated-voltage-source-inverter (PWM-VSI) and the general fractional step control delay. Based on the model derived, a discrete-time domain AC machine current regulation method is developed and evaluated analytically and experimentally. The proposed regulator provides robust stability with superior dynamic performance at extremely high fundamental frequency operation or at extremely slow switching conditions where only a small number of control updates per fundamental cycle are allowed.;One of the emerging functional needs for variable speed AC drives is the on-line preventive diagnostics capability for the AC machine itself and for the whole electromechanical actuator system. This thesis presents contributions made in the areas of the state estimation and the signal processing methods for variable speed motion servo system diagnostics. As the first step for motion servo diagnostics, operating point independent fault signatures are developed. The problem of disturbance torque estimation is studied in detail and a state-observer-based disturbance estimator topology suitable for signal processing applications is developed. Specifically for gear tooth surface fault diagnostics, an observer-based kinematic error estimator was developed and the kinematic error estimate is suggested as the operating point independent gear surface fault signature. As the second step, signal processing methods suitable for real-time implementations are developed for the spatial domain re-sampling of regular time domain samples and are combined with the spatial domain synchronous averaging methods, which are used to filter out the asynchronous signatures efficiently. Lastly, a motion servo actuator with a gear surface failure is tested, and the proposed methods successfully extracted and recovered the kinematic error profiles of the defective gear during motion servo operations over a wide velocity range.