Assessment of Artificial Neural Network-based Induction Motor Fault Classifier Using Continuous Wavelet Transform

Abstract

Induction motors are widely used in industrial, commercial, and residential applications due to their significant advantages over other types of electric motors. These motors are subjected to a variety of operating stresses that can result in faults. Bearing faults, stator interturn faults, and cracked rotor bars are the most common recurrent faults in induction motors. Early detection of induction motor faults is critical for reliable and cost-effective operation. Faults and failures of induction motor can lead to excessive downtimes and generate large losses in terms of maintenance and revenues. The purchasing and installation cost of equipment usually cost less than half of total expenditure over the life of machine for maintenance. Maintenance cost is 15% to 40% of the total cost and it can go up to 80% of the total cost. In many cases, the failure of a critically loaded machine can bring an entire industry process to a halt. The growing demand for high-quality and low-cost production has increased the need for automated manufacturing systems with effective monitoring and control capabilities. Condition monitoring and fault diagnosis of an induction motor are critical in the manufacturing process. It can reduce maintenance costs and the risk of unexpected failures by allowing for the early detection of catastrophic failures. There are many condition monitoring methods, including vibration monitoring, thermal monitoring, chemical monitoring, acoustic emission monitoring, but all these methods require expensive sensors or specialized tools. Whereas the current monitoring does not require additional costly sensors as basic electrical quantities voltage and current are readily measured by voltage and current transformers that are always installed as a part of the protection system. As a result current monitoring is non-intrusive and may be implemented even if the motor is at the remote end from the motor control center. Thus MCSA proves to be a low cost online nondestructive fault diagnosis and detection system to provide accurate assessment of motor faults. This chapter presents experimental results for multiple fault detection in induction motors using signal processing and artificial neural network approaches. The continuous wavelet transform was used to analyse motor line currents recorded under various fault conditions. A feedforward neural network was used for fault characterization based on fault features extracted using continuous wavelet transform.