Abstract
Electric motors serve as critical components in various industrial applications, contributing significantly to the overall efficiency and productivity of systems. However, the challenge of unexpected failures can lead to costly downtime and maintenance expenses. This study focuses on the use of machine learning techniques to implement a predictive maintenance strategy for electric motors, with the goal of increasing reliability and reducing operational disruptions without overspending on repairing or replacing items that still have a life expectancy. The initial section of this study provides a comprehensive overview of existing predictive maintenance methodologies, highlighting their limitations and the need for more advanced approaches. Traditional methods often fall short in capturing the intricate dynamics of electric motor degradation. In response, this research proposes a machine learning-based predictive maintenance model, leveraging historical data and key operational parameters. The focus in the case has been on the classifier approach with support vector Machines (SVM), K-nearest neighbour (KNN) and Naïve Bayes (NBE) being the focus; however, decision trees were further explored for reference. The core of the study involves the development and implementation of the predictive maintenance model using MATLAB. This powerful tool is widely employed in engineering research, and its code examples will be crucial for illustrating the model's practical implementation. The model utilises features such as temperature, vibration, and current consumption as inputs to machine learning algorithms. Specifically, we explore the effectiveness of Support Vector Machines (SVM), Random Forests (RF), and Neural Networks (NN), providing detailed MATLAB code examples for their implementation and optimisation. To ensure real-world applicability, the research delves into the integration of real-time monitoring systems with the predictive maintenance model. This enables continuous data acquisition, ensuring the model's adaptability to changing operating conditions. Case studies involving diverse types of electric motors in various industrial settings showcase the practicality and effectiveness of the proposed approach. In this case, for the purposes of the study, as the motor was not a smart motor, Raspberri-pi 3B was used to collect data from the motor with the corresponding sensors for vibration, voltage and temperature. Included in the data collection were environmental temperature and humidity (RH) factors whose effect on motor performance were being investigated. The anticipated outcomes of this research include improved reliability, reduced downtime, and significant cost savings for industries relying on electric motors. This work advances the existing state of predictive maintenance through the
iii
use of machine learning, contributing to the development of smarter and more efficient industrial systems. This study bridges the gap between standard maintenance procedures and the promise of machine learning for the predictive maintenance of electric motors. It also serves to highlight the different classifier methods that can be used if, for instance, the failure data is not available, a situation which was true for this case. The MATLAB-based examples provided throughout the study serve as a practical guide for engineers and practitioners seeking to implement similar predictive maintenance strategies in their respective domains, contributing to the evolution of maintenance practices in the realm of electric motors.
Keywords
Machine learning, Modelling, Binary classification, Data acquisition and analytics, decision tree, Predictive and prescriptive maintenance strategy.