Search

This thesis provides a detailed analysis of the root cause of the acoustic noise and vibration generated in Switched Reluctance Machines (SRMs) and proposes control solutions to address the issues. Dynamic current profiling methods are proposed through multi-objective optimizations. The performance of the proposed controllers is verified through multi-physics simulations and experimental testing. The SRM under consideration in this study is modeled in the electromagnetic FEA simulation based on the given technical specification. The electrical and mechanical output data from the FEA are extracted for the SRM modeling for the dynamic simulations. The radial force and the torque are modeled as a function of the rotor position and phase currents using piecewise linear lookup tables in the dynamic simulations. The peak to peak force and torque ripples are predicted accurately during the process of off-line optimizations for the current profile generations. The electromagnetic, structural, and acoustic analyses are coupled in the multi-physics simulation software to predict the noise, vibration and harshness (NVH) performance of the SRM using the proposed current shapes. Two different optimization approaches are implemented to generate current profiles including control-based and current harmonic injections. The shape of the current or the injected current harmonics are optimized globally through an iterative multi-objective optimization algorithm. The control-based injections are either implemented on the actual phase reference currents or on the quadrature axis of the phase currents in the rotating reference frame. While the control based injections reduce the acoustic noise and vibration by flattening the radial force and torque productions, the methods perform poorly on some of the frequency ranges depending on the operating speeds. Then, the current harmonic injection algorithm is proposed to eliminate radial force harmonics at resonant natural mode frequencies, in addition to reducing the torque and radial force ripples. The dynamic simulation results provide a comparison of the conventional and proposed current profiles for the reduction of the radial force and torque ripple. The generated current waveforms from the optimizations are applied to FEA models to predict the vibration and acoustics noise through multi-physics simulation software. The current waveforms using three different optimization methods along with the conventional waveform demonstrated a significant comparative improvement for the proposed methodology. The experimental prototype SRM and the test platform have been developed to verify the performance of the proposed control algorithms. The dynamic and multi-physics FEA simulations and the experimental results demonstrated significant improvements in reducing the acoustic noise and vibration over a wide range of operating conditions compared to conventionally controlled SRMs.

The problem of accurately predicting the state-of-charge of a battery, and hence its lifetime - which is the duration of time over which a battery can supply energy to a load, has remained elusive. Because many modern applications are powered by a collection of batteries, new methods reported in the recent literature aim to extend the lifetime of a collection of batteries by using redundant batteries and scheduling a subset of the batteries for discharge. To compare the effectiveness of different strategies for scheduling batteries, we defined a new metric called discharge efficiency. We show that both the lifetime and the discharge efficiency can be improved by using redundant batteries and scheduling strategies. To cope with the uncertainties that come to pass during the discharge process, we formulated the problem of coordinating the discharge of a collection of batteries in the goal-seeking paradigm. This formulation allowed us to design an effective decision-maker to coordinate the discharge. The results in this thesis show that the decision-makers based on the goal-seeking paradigm improve lifetime of a collection of batteries between 12.7% (when the collection supplies constant power to a load) and 21.6% (when the collection supplies constant current to the load). Similarly depending on the discharge scenario, the discharge efficiency is improved by 11.7% to 13.4%. In addition, the new decision-maker utilizes a small number of redundant batteries more effectively than the scheduling methods reported in the literature.