Sandcat Optimized ANN-LSTM Framework for Advanced Fault Detection in Electric Vehicle Drive Motor

In order to guarantee the longevity, safety and
efficiency of drive motors, there is an urgent need for
sophisticated and dependable defect detection systems due to the
growing dependence on Electric Vehicles (EVs). Conventional
diagnostic techniques frequently struggle to handle intricate,
time-varying sensor data and are unable to adjust to changing
operational circumstances. A novel framework for EV motor
defect detection utilizing a Sandcat Optimized ANN-LSTM
model is proposed in this study to overcome these limitations.
To ensure high-quality inputs for learning, the system first
gathers raw sensor data from the motor and refines it using a
systematic preprocessing model that includes data cleaning,
one-hot encoding, and normalization. In order to standardize
input magnitudes and enhance the learning dynamics of the
model, feature engineering is utilized through feature scaling.
Accurate fault state classification is made possible by the use of
a hybrid ANN-LSTM network, which captures the motor
signals' temporal and spatial features. Sandcat Optimization,
which adjusts hyperparameters to attain the best accuracy and
efficiency, further improves the model's performance. The
designed framework provides a better performance analysis
from python software, which provides a reliable and clever way
to detect faults in EVs drive systems by successfully
differentiating
B. Parvathi Sangeetha
Department of Marine Engineering
AMET Deemed to be University,
Chennai, 603112, India,
parvathi@ametuniv.ac.in
N Janaki
Department of Electrical and Electronics
Engineering
Vels Institute of Science, Technology and
Advanced Studies,
Chennai, India
janaki.se@vistas.ac.in
mobility is the creation of precise and intelligent defect
detection systems [3].
B. Challenges in Existing topologies
The Conventional defect diagnostic methods, which
generally rely on rule-based systems, signal thresholding, or
custom feature extraction, are incapable of observing the non
linear and dynamic nature of motor function under different
conditions. Further, such methods tend to be inflexible and
not very scalable when faced with large scales of real-time
sensor data [4]. While some machine learning (ML) models
have been developed to detect faults automatically, they tend
to struggle with capturing sequential patterns and require
significant manual parameter tuning [5]. These limitations
highlight a need for a more intelligent and automated problem
detection services.
C. Role of Deep Learning and Optimization Model
between normal and defective motor
characteristics with higher accuracy of (97%)