Health index degradation prediction of induction motor using deep neural network algorithm

• A deep neural network-based self-designed model predicts the health status of the motor. • Multiple sensors deployed on the motor provide continuous data for pre-processing and feature extraction. • The effectiveness of the model's accuracy and low mean square loss results show visible improvement over previously researched models. • The research offers greater scope in the industrial sector. • This research will help reduce the maintenance cost for uninterrupted machine operation through predictive modeling. Predictive and preventive methodologies are increasingly playing a role in improving the safety and reliability of the system. Early machine fault detection and health monitoring help avoid severe damage. The conventional methods, such as mathematical modeling, require extensive motor operation data and have become non-adaptive to the level of complications and randomness. Modern advancements in research have proved that Machine and Deep learning models are more effective for monitoring and detection systems offering reliable system accuracy. This paper explores a deep neural network-based self-designed model that not only detects and classifies the fault but also predicts the health status of the motor after the occurrence of the fault. Multiple sensors deployed on the motor provide continuous data for pre-processing and feature extraction during the operation. The non-linear Deep Neural Network regression model employed in this study performs well achieving 99.6 % accuracy. The effectiveness of the model's accuracy and low mean square loss results show visible improvement over previously researched models. Moreover, this model also provides helpful information regarding the motor's health in case of fault, ensuring the cycle rotation of the motor without causing permanent damage. The research offers greater scope in the industrial sector, where continuous health monitoring of a machine is a prime objective, and early detection of faults can prevent more significant losses. This research will also help reduce the maintenance cost for uninterrupted machine operation through predictive modeling.