Artificial Intelligence-Based Predictive Maintenance Framework for Industrial Machinery Using IoT Sensor Data

Authors

  • Javier Monroy Associate Professor, Department of Systems Engineering and Automation, Universidad de Mála Author
  • Caroline L. König Associate Professor, Department of Computer Science, Universitat Politècnica de Catalunya, Barcelona, Spain Author

Keywords:

Predictive Maintenance, Industrial Internet Of Things, Gradient Boosting, Condition Monitoring, Industrial Machinery

Abstract

Predictive maintenance remains difficult to implement reliably when industrial sensor data are incomplete, failures are rare, and maintenance decisions require sufficient warning time. This paper develops an artificial intelligence-based framework that converts Internet of Things sensor measurements into maintenance alerts for rotating industrial machinery. The framework uses a single predictive method, gradient-boosted decision trees, supported by causal feature extraction, explicit data-quality checks, probability calibration, and a persistent-alert policy. Vibration, temperature, motor current, rotational speed, and operating-load measurements provide the information needed to distinguish deterioration from ordinary changes in operating conditions. The prediction task estimates the probability of a defined failure occurring within a specified planning horizon. Evaluation is organized around chronological separation, independent equipment histories, event-level detection, warning lead time, and false-alert burden. Because no industrial dataset accompanies this study, the results comprise reproducible analytical calculations and engineering deductions rather than measured predictive performance. The analysis demonstrates that a classifier with assumed sensitivity of 90% and a false-positive rate of 1% would achieve precision of approximately 47.6% at a failure prevalence of 1%, but only 8.3% at a prevalence of 0.1%. Further calculations show how intervention effectiveness, response delay, and monitoring availability constrain practical value. These results establish requirements for evaluating the proposed framework without claiming that its predictive capability has been validated. The central contribution is a coherent connection between sensor observations, model outputs, and maintenance actions. Reliable implementation depends on defensible labels, operationally meaningful evaluation, and disciplined management of uncertainty as much as on the learning algorithm.

Downloads

Published

2026-03-30

Issue

Section

Articles

How to Cite

Artificial Intelligence-Based Predictive Maintenance Framework for Industrial Machinery Using IoT Sensor Data. (2026). International Journal of Engineering Innovation and Computational Science, 1(1), 17-32. https://journovapublishing.com/journals/ijeics/article/view/5

Similar Articles

You may also start an advanced similarity search for this article.