Predictive Vehicle Maintenance Using Machine Learning: Developing a Data-Driven Framework for Predicting Component Failure and Optimizing Maintenance Intervals in Modern Automobiles

Atinder Singh, Raghu Raja Mehra

Abstract: Most vehicles on Indian roads are still maintained in one of two ways: they are repaired after something breaks, or they are serviced at fixed intervals that take no account of how the vehicle has actually been driven. Both approaches waste money, the first through breakdowns and downtime and the second through parts replaced long before the end of their useful life. This study develops and evaluates a data-driven framework that uses machine learning to predict component failure in advance and to set maintenance intervals according to the real condition of each vehicle. Telemetry from on-board diagnostic (OBD-II) loggers, GPS units and eighteen months of workshop records were collected from 320 commercial vehicles operated by three fleets in Punjab, covering taxis, delivery vans and school buses. Over the study period the fleet recorded 1,146 unplanned failure events, with the battery, brake system and tyres together accounting for 58% of them. Five models were trained to predict whether a component would fail within the next 30 days. Gradient-boosted trees (XGBoost) achieved the best overall result, with an F1-score of 0.82 and an area under the ROC curve of 0.93, against 0.59 and 0.78 for a logistic regression baseline, while a long short-term memory (LSTM) network gave the longest median warning time of 21 days. Remaining-useful-life estimates were most accurate for gradual-wear components such as brake pads and batteries and least accurate for suspension parts. When the model was used to schedule servicing in a cost simulation, unplanned breakdowns fell by 56% and total annual maintenance cost per vehicle fell by 41% compared with a conventional fixed-interval schedule. The findings suggest that the real value of predictive maintenance lies not only in preventing failures but in allowing each component to be used for more of its safe working life.

Keywords: predictive maintenance; machine learning; remaining useful life; OBD-II telematics; XGBoost; LSTM; fleet management; maintenance scheduling

How to Cite?: Atinder Singh, Raghu Raja Mehra, "Predictive Vehicle Maintenance Using Machine Learning: Developing a Data-Driven Framework for Predicting Component Failure and Optimizing Maintenance Intervals in Modern Automobiles", Volume 15 Issue 9, September 2026, International Journal of Science and Research (IJSR), Pages: 1692-1699, https://www.ijsr.net/getabstract.php?paperid=SR26926091735, DOI: https://dx.doi.org/10.21275/SR26926091735

Download Citation: APA | MLA | BibTeX | EndNote | RefMan