Enhancing Predictive Maintenance in Industrial IoT through Machine Learning Models
A. Jaya Mabel Rani, V. Prithivirajan, S. Arumai Shiney, Ch. Babaiah, Hemanth Swamy, G. Sivagamidevi
Abstract
Significant production losses and increased maintenance costs can arise from unplanned downtime in industrial settings caused by machine faults. In order to identify issues early and prevent significant breakdowns, predictive maintenance approaches make use of data collected from Internet of Things (IoT)-enabled devices installed in operating equipment. Here, we present a predictive maintenance system that employs machine learning methods, namely AdaBoost, to categorize various types of machine pauses in real-time, with a focus on knitting machines. Six different types of machine stops were classified using machine learning: idle stop, Lycra stop, full roll stop, gate stop, feeder stop, and needle stop. We did this by preprocessing and inputting machine learning model data, which included machine speeds and steps. After a rigorous training and optimization process that combines cross-validation and hyperparameter tuning, the model achieves an impressive accuracy rate of 92% on the test dataset. These results highlight how our suggested solution might help the textile sector become more productive and efficient overall by precisely detecting machine stoppage and enabling timely maintenance interventions.