Predictive Analytics on Engine Fault Code Repositories: Translating Fleet Telemetry into Actionable Maintenance Intelligence

Abstract

Predictive Analytics on Engine Fault Code Repositories: Translating Fleet Telemetry into Actionable Maintenance Intelligence demonstrates an objective, evidence-based examination of data-driven maintenance, with formal structure and scholarly tone. Research motivation, scope, and relevance of fleet maintenance and telematics are articulated, and aims, significance, and expected contributions to predictive maintenance practice are defined, followed by a review of existing fault-code analytics, fleet telemetry studies, and theoretical underpinnings of predictive maintenance. Discussion of repositories, data maturity, and addressed gaps precedes an overview of data provenance, quality, preprocessing, and integration of engine fault codes with, along with repository schema, data lineage, versioning, and access controls. Predictive maintenance focuses on anticipating failures before they occur. In fleet systems, patterns of engine fault codes derived from vehicle telemetry can indicate impending failures. Various repositories of engine fault code data have been assembled, but these have yet to be connected to fleet telemetry. A novel data source combines Engine Control Unit fault codes reported to the fleet service provider and decoded by the Original Equipment Manufacturer with over a billion rows from the fleet’s telematics database. Multiple methods exploit this integration to model future failures. A survival analysis approach predicts the time until each engine subsystem is likely to require service based on current operating conditions, while supervised-machine-learning classifiers assess fault-code occurrences, providing a foundational capability for planning engine maintenance for the next several thousand kilometers based on either current status or short-term future telemetry values.