Electric drives, as adequately described over years of development and research, are the current solution to the question of automatic, precise control when it comes to the operation of electric motors. Comprised of the electric motor itself, power modulator, power source, control unit, sensor unit, load and more; an electric drive is an apt practical demonstration of a closed-loop system. Due to its highly controllable nature which has proven to be reliable and efficient, the electric drive has found itself being used in numerous applications such as - industrial automation, electric vehicles, renewable energy systems, robotics and smart manufacturing. However, the number of components present within the electric drive bring up the possibility of loss of function/efficiency of drive, in case of fault in said components, leading to unexpected shutdowns and unaccounted for maintenance costs. Hence, an important aspect of the operation of electric drives is the fault detection of, and within its components, ensuring safe and timely working of the equipment. Traditional fault diagnosis techniques are largely based on signal processing, analytical models and rule- based monitoring methods; but these approaches often face limitations when put under complex operating conditions and varying load environments. This review paper attempts to address the limitations of the traditional fault diagnosis techniques, while suggesting adoption of machine learning based fault diagnosis covering different models and their effectiveness at dealing with such faults in electric drives. Various faults are specifically addressed in this review including - stator winding faults, rotor faults, bearing defects, inverter switch faults and sensor failures; discussed along with commonly used feature extraction methods such as wavelet transform, Fourier analysis and statistical indica- tors. Furthermore, supervised machine learning algorithms such as Artificial Neural Networks (ANNs), Support Vector Machines (SVMs), Decision Trees, Random Forest and Deep Learning models are reviewed for fault detection and classification. Finally, a comparative analysis of existing methods, advantages, and limitations is presented. The study concludes that integrating model-based approaches with machine learning techniques offers a capable manner of handling modern electric drives while providing useful predictive maintenance solutions and smart monitoring systems to maintain efficiency of any electric drive.

Index Terms—Electric Drives, Fault Diagnosis, Machine Learning, Predictive Maintenance, Inverter Faults, Induction Motor, Artificial Neural Network, Support Vector Machine, Deep Learning, Condition Monitoring