A Deep Neural Network Approach for Fault Detection in Industrial Motors: Enhancing Reliability and Efficiency in Industry 4.0

In the era of Industry 4.0, the integration of sophisticated monitoring and control technologies is essential for the reliability and efficiency of industrial equipment. This paper presents a deep neural network (DNN)-based approach for fault detection and classification in industrial motors. The approach identifies common motor faults, including bent shaft, loose vibration, loosely mounted components, electrical faults, and bearing vibration. The proposed DNN model demonstrated robust performance, achieving an overall training accuracy of 83.49% ± 0.43% and a test accuracy of 81.91 % ± 0.82%. The model also achieved a mean precision of 82.05% ± 0.77%, mean recall of 81.91 % ± 0.82%, and mean F1-Score of81.86<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>± 0.78%. The 95% confidence intervals for accuracy, precision, recall, and F1-Score were (80.77%, 83.04%), (80.98%, 83.11 %), (80.77%, 83.04%), and (80.78<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>,82.95<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>), respectively. These results demonstrate the DNN model's ability to precisely classify various motor faults with minimal variability, as evidenced by its narrow confidence intervals and low standard deviations. The model's reliability and efficacy for real-world applications in proactive maintenance and fault detection are confirmed by the consistent performance that is observed across cross-validation folds.