ML-based failure detection approach for predictive maintenance in an industry 4.0 oriented web manufacturing control application

With the increasing reach of Industry 4.0, the Internet of Things (IoT) and Artificial Intelligence (AI) and many other technologies, have emerged as pivotal components in the era of information technology. Recent progress in AI, and its subfield machine learning (ML), have influenced numerous research domains inside the industry, achieving improvements that were not possible with conventional optimization techniques, notably illustrated in the case of predictive maintenance. A significant component of predictive maintenance is fault detection. The goal of predictive maintenance is to avoid future machine failure and consequent downtime by anticipating faults in machinery. As a result, the aim of this research is to employ ML models for fault identification process. A concrete dataset was employed with real sensors’ outputs allowing to make analysis of faults in machines mounted with various types of sensors. Four distinct supervised ML algorithms were applied in our work in which their hyper-parameters were tuned with both Grid Search and Random Search methods. A comparative study between the algorithms’ performances was made in order to select the best one with highest metrics percentages. A web-based application was also developed, to monitor data and send alerts when detecting anomalies either via SMS, email or inside the provided web interface.