AI-Based Diagnosis and Predictive Maintenance in Vehicles
Vehicles produce a continuous stream of operating information through On-Board Diagnostics (OBD) systems and onboard sensors. Diagnostic trouble codes and fixed thresholds are valuable tools, but they do not always explain faults that emerge gradually or influence several signals together. This review considers how sensor analytics and artificial-intelligence methods can be applied to vehicle fault detection and classification, failure-risk estimation, predictive maintenance and vehicle-health assessment. It discusses OBD data and its quality, rule-based diagnosis, supervised and unsupervised learning, anomaly detection, root-cause analysis and explainable artificial intelligence (XAI). The review also considers practical issues such as noisy or missing readings, scarce fault labels, class imbalance, changing driving conditions and the distinction between an unusual reading and a verified fault. The studies examined often concentrate on individual diagnostic tasks. In a practical system, their outputs therefore need to be brought together carefully: an anomaly score, a fault category, a likely cause and an estimate of future risk describe different things and should not be treated as interchangeable. In response, this paper outlines AutoSense AI as a proposed framework at the design stage. Its intended workflow includes data simulation and preprocessing, feature engineering, anomaly detection, fault classification, probable root-cause analysis, failure-risk prediction, vehicle-health scoring, maintenance planning, XAI and diagnostic reporting. The framework has not been implemented or tested on vehicles, so no measured accuracy or maintenance improvement is claimed. A staged evaluation using simulated data followed by real-vehicle data is proposed for future work.
Authors
- Shruti Santosh Kadam (ORCID: https://orcid.org/0009-0002-9026-0019)
- Mrunalinee vikas Kanse
- Tanushka Vitthal Pawar (ORCID: https://orcid.org/0009-0000-6120-8129)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24416424
- Primary Topic
- Fault Detection and Control Systems
- Type
- article
- Field-Weighted Citation Impact
- 0.00