Predictive Maintenance of Industrial Machines using ML and IoT

This paper focuses on the development of a prototype, later evolving into a system that utilizes concepts of Machine Learning, Internet of Things, and Predictive Modeling to assess the compatibility of different algorithms with microcontrollers for predicting points of failure. The prototype is constructed around a network of sensors designed to measure physical aspects of the operational motor, seamlessly integrated with algorithms aimed at forecasting failure rates and time to failure in the testing environment. The DC motor connected to the microcontroller serves as the primary source of data collection, as parameters such as temperature, vibration, and current through the motor are essential for efficiently predicting failures. Readings are gathered for varying current values, which influence changes in the rotating speed of the DC motor. Analysis for safe levels and levels beyond critical thresholds is conducted by establishing a threshold current value for this purpose. Once parameters are documented for different current values over a one-minute operational time on a spreadsheet, algorithms analyze the readings and generate the final output. Datasets for both conditions are analyzed, and various combinations are tabulated with failure time and percentage alongside the microcontroller and algorithm used. With the assistance of this output, one can easily assess the accuracy of the algorithm with the microcontroller and the actual precision of the predictions. If implemented on a large scale, this work can be applied in the general aviation industry, within hydroelectric power plants, dam turbines, and heavy machinery, facilitating a gradual shift from reactive maintenance to preventive maintenance in society.