Abstract:
Induction motors are commonly used for different industrial work purposes. And
consider the most used machine on the planet due to its strong and rigid structure and
good-quality output. Therefore, due to their heavy use in industries, and commercial
and residential buildings they may experience various faults in the stator and rotor, such
as broken bar (BRB), full pole pitch (FPP), static eccentricity (SE), etc., and their early
detection is necessary to avoid further damage in the system. Fault detection in
induction motors is usually done by measuring noise, temperature, vibration, and
electrical quantities, including current, flux, and torque. There are many other
techniques used in the literature by the scientist for the fault detection induction motor.
In this work, a deep neural network (DNN) machine learning (ML) algorithm is
proposed and designed according to the simulated data. The designed DNN is then
compared with support vector machine (SVM) and random forest classifiers (RFC) for
the detection and classification of faults in induction motors using stator current and
stray flux. ANSYS Maxwell-based simulations are performed for four different loading
conditions (25%, 50%, 75%, and 100%) of the induction motor to obtain the stator
current and stray flux data on normal and faulty conditions (BRB1, BRB2, BRB3, FPP,
and SE). The results indicate that the proposed deep neural network algorithm has
shown better accuracy for stray flux compared with SVM and RFC on 100 % loading
conditions. Moreover, DNN performed well for other loading conditions in case of stray
flux. However, for stator current, the overall performance of all machine learning
algorithms was less efficient.
