A Data-Driven Model for Predicting Vehicle Component Failures: Insights from National Vehicle Inspection Records for Safety and Maintenance Policy

Abstract Although many countries are reviewing vehicle inspection policies to increase road safety, actual data on mechanical degradation is rarely reflected in current standards. The study analyzes more than 99 million inspection records collected in the country from 2014 to 2022, focusing on how these results can help policy-making. We have developed a model that tracks headlamp brightness as an important safety feature to estimate the probability of component failure according to cumulative vehicle mileage. Results show that more than 70% of vehicles exceed the legal brightness failure limit after traveling more than 130,000 km. This is a risk level that is not captured by the current inspection interval. These results highlight a significant gap between uniform inspection schedules and the lifespan or use of actual vehicle components. Based on these results, we recommend modifying inspection and maintenance policies based on empirically measured risks rather than at fixed time or distance intervals. To further support targeted supervision, we propose a real-time monitoring platform to identify groups of vehicles experiencing repetitive inspection failures. This data-driven approach provides a practical foundation to support resource allocation and transportation policy establishment, and shows that large-scale field data can lead to practical and rapid regulatory improvements.