Fault Diagnosis and Condition Monitoring of Brushless DC Motor Drive for Electric Vehicle Applications

Abstract

With the increase in Electric Vehicles (EVs) adoption, the need for reliable, efficient, and fault-tolerant motor drive systems also increases. Brushless Direct Current (BLDC) motors are standard in EVs applications because they exhibit high efficiency, small size, and low maintenance. However, like any electric machine, BLDC motors can incur faults in various components that could affect motor behaviour (performance, safety, operating life). Therefore, this thesis includes an in-depth study of developing and detecting faults and real-time condition monitoring of BLDC motor drives for EVs. The research starts with comprehensively classifying BLDC motors based on faults occurring during regular use. Common faults occurring in BLDC motors include inter-turn, coil-to-coil, phase-to-phase, open-circuit, and external motor-inverter connection faults. These external faults consist of single-phase, double phase, and ground faults. Perform scanning of MATLAB/Simulink-based mathematical models of healthy and faulty BLDC motor operation; the research scan will be successful by simulating the system input-output and ascertaining the impact on the system's faulty operation. After that, a Machine learning model is used to classify the healthy and faulty conditions of the motor. For that, the next step is the significance of diagnostic methods, where signal-based (monitoring data characterizations), model-based (simulated motor faults), and data-based (using machine learning algorithms) diagnostic methods fundamentally apply features to separate data characterizations of faults. Once this was done, the researcher trained and tested the BLDC dataset through multiple machine learning algorithms: support vector machine (SVM), k-nearest neighbors (KNN), decision trees, and neural networks, primarily to demonstrate clarity to the BLDC dataset with faults, specifically the labelled faults classification. Complementary to the above, using the ANSYS Maxwell model captures and models magnetic flux distribution for perimeter validation of scientific honesty and duplicated accuracy for further experimentation or distinguishing different coils in experimentation. Experimentation includes a built-in health monitoring setup across the STM32 NUCLEO H743ZI2 microcontroller for monitoring BLDC systems health and processing time to complete tasks of real-time data acquisition from various sensor inputs (voltage, current, speed, vibration, and temperature). A fusing of the internal and external sensors where real-time data acquisition acts separately in task order of voltage and current, speed and vibration, temperature, and provides a processing buffer for separate data task acquisition. At the end, iv Opal-RT platform is utilized to study and analyze the characteristics of the motor, for the known healthy and faulty conditions, at a high degree of fidelity. The proposed research work has been implemented using MATLAB/Simulink software, Ansys, Opal-RT, and further, experimentally validated over a small-scale laboratory set-up of rating 1 kW, 610 rpm BLDC motor drive. The main objectives behind the present work have been the following: 1. 2. 3. 4. To the mathematical modelling of a healthy and faulty BLDC motor drive. To analyze and diagnose faults of the BLDC motor drive. To simulate a healthy and faulty BLDC motor drive in the MATLAB/Simulink, Ansys, and Opal-RT environment. To validate simulation results through an experimental setup.