Modeling and Parameter Identification for Condition Monitoring of Permanent Magnet Synchronous Machine Drive Systems

Abstract

Electric machines can be found in a variety of industrial applications, including renewable energy, transportation, and military systems. Among electric machines, Permanent Magnet Synchronous Machines (PMSMs), such as Surface-Mount Permanent Magnet (SMPM) and Interior Permanent Magnet (IPM) machines, have been preferred for high-performance applications due to their high torque density, high power density, and potential for precise control and high efficiency over a wide operating range. However, PMSMs can experience magnetic, electrical, and mechanical faults, compromising the system performance and safety. Fault diagnosis and condition monitoring techniques aim to identify abnormal conditions and track the health status during operation. In particular, if the machine’s health condition is continuously monitored, faults can be recognized at early stages, and corrective actions can be implemented. Model-based techniques use unusual changes in state variables, parameters, and outputs to monitor the machine’s health and determine whether the machine is experiencing a fault. Online parameter identification offers continuous monitoring of health conditions through parameter variation during operation. Moreover, if the parameters have physical meaning, health conditions can be tracked and diagnosed more straightforwardly. However, in some cases, faults and operating conditions might have similar effects on the parameters. Additionally, each fault causes different imbalances in the PMSM dynamics that standard models do not capture. Furthermore, parameter identification has an inherent implementation challenge, since accurate estimation requires persistently exciting inputs which may conflict with control objectives and compromise control performance. This dissertation presents research that seeks to address open issues regarding the application of parameter identification to fault diagnosis and condition monitoring of SMPM machines.

The first part of this dissertation addresses the incorporation of operational constraints into the Simultaneous Identification and Control (SIC) formulation for SMPM machines. Specifically, a SIC methodology that explicitly considers the voltage and current inverter limits for SMPM machines is presented. The current and voltage constraints are derived by mapping three-phase voltage and current constraints into their two-phase equivalents. These constraints are incorporated into a SIC formulation that consists of an adaptive current regulator and a Receding Horizon Adaptive Input Design (RHAID). The SIC formulation utilizes the quadrature-axis current for torque production, while the direct-axis current is used to inject the excitation required for accurate convergence. The inverter constraints are incorporated in the RHAID, which minimizes losses while maximizing the excitation characteristics of the reference direct current. Accurate torque regulation is performed through the adaptive current regulator. Simulations demonstrate the effectiveness of the SIC formulation on constraint enforcement at different operating conditions.

The rest of this dissertation studies the modeling and parameter identification for fault diagnosis and condition monitoring of SMPM machines. First, lumped-parameter models are formulated to capture the distinctive dynamic features of SMPM machines under demagnetization, eccentricity, and inter-turn short conditions. In addition to the standard model parameters, these parameterizations incorporate parameters that capture specific oscillations produced by the different faults. Based on these models, parameter identification strategies are formulated for detecting demagnetization, eccentricity, and inter-turn short. The inputs are designed to guarantee sufficient conditions for accurate parameter convergence while avoiding control perturbations. Afterward, a parameter identification strategy for comprehensive fault detection is formulated by incorporating the estimators for demagnetization, eccentricity, and inter-turn short into a sole strategy. Simulation and co-simulation results demonstrate the effectiveness of the proposed parameter estimators for recognizing the different fault conditions.