Advanced Sensor Fusion and Machine Learning Framework for Real-Time Health Monitoring of Electric Drives

Abstract

The importance of electric drives in the present industry automation, transportation and renewable energy systems is that they are highly efficient, controllable and operational reliable. Nevertheless, the extended working conditions in the harsh industrial environment might result in the wearing out of parts, ineffective working, and unpredictable failures of the system. Traditional methods of monitoring are usually based on the use of single sensor and diagnostic methods that use thresholds, but in the majority of cases, they are not capable of identifying early faults and system interactions. To address these shortcomings, this paper suggests a more developed sensor fusion and machine learning system to real-time health monitor electric drive systems. The suggested framework incorporates the data of various sensors such as vibration sensor, temperature sensor, current sensor and voltage sensor to record the whole operating qualities of the drive system. The method of sensor fusion in the form of a feature level is used to integrate the data of multiple sensors into an integrated proposal of system health. Then, machine learning models are implemented that help recognise the fused set of features that can accurately detect faults and the classify the health condition. The monitoring framework proposed has been tested on a system of simulation experiment running on the MATLAB/Simulink with a varied set of operating and fault conditions. The experimental findings suggest that the combined sensor fusion and machine learning process can considerably increase the accuracy of fault detection and minimise the time needed to make a diagnostic decision when compared to the traditional monitoring processes. The suggested structure can be successfully used as a predictive maintenance and intelligent monitoring of electric drives in high-tech industry settings.