Machine Learning enabled Condition Monitoring Models for Predictive Maintenance of Boilers

The fourth industrial revolution (Industry 4.0) encompasses three major technological trends driving this transformation: Connectivity, Intelligence and Flexible Automation. Predictive maintenance allows the organizations to identify potential problems in the production devices far before the failure occurs. This paper aims at implementing an interdisciplinary approach by studying key performance indicators of the boiler and modelling the data acquired, using machine learning algorithms.The efficiency of the boiler is mapped by analyzing the performance of its various components. It is observed that the performance of the oil heater and feedwater pump affect the efficiency of the boiler significantly. Predictions for the future are made by implementing a polynomial regression model on the Key Performance Indicators of the boiler. The critical point is defined as the point below which the performance of the boiler deteriorates significantly and affects productivity at a large scale.This study determines the number of days in which the oil heater and the water pump would reach their respective critical points. This eventually helps to gain valuable insights into the efficiency of the boiler and predict the consistency of the equipment so as to minimize the losses and make important decisions well ahead in time.