Enhancing School Bus Engine Performance: Predictive Maintenance and Analytics for Sustainable Fleet Operations

Abstract

A smart solution for preventive information, diagnostics, and prognostics for durable engine management on school buses will improve vehicle safety and performance, reduce maintenance costs and time, and extend vehicle life while minimizing environmental impacts. Additionally, the solution provides knowledge to the school district on effective vehicle maintenance; it enhances students’ welfare and educational outcomes. School transportation is the only major public transit system for children, which operates two trips a day each school day. 480,000 yellow school buses transport 25 million children 3.3 million miles to and from school. Researchers have shown exacerbated environmental and health impacts from considerable emissions and idle time. Schools are increasingly choosing diesel engine vehicles over hybrid or electric power vehicles. The higher acquisition and maintenance costs, the reliance on battery power based on outside temperature, and the requirements of off-route and after-school activities have limited the investment in these alternative power school buses. The interest in this paper is primarily based on diesel-powered school buses that constitute 90% of school buses and trips. High emission durations caused by aging and/or poor maintenance of school buses are a result of the non-usage of technologies and management practices to achieve low emissions. Its growth is hindered by the inability of some school districts to detect engine damage early, poor excuses regarding the price for effective maintenance, and uncertainty about maintenance cost savings and equipment longevity.

The proposed diagnostic and prognostic maintenance solution employs an open-source machine learning algorithm to train bus engine and emission rate models and minimize idle time and wear parts using vehicles’ sub-minute real-time GPS locations, vehicle-activated event logs, and real-time diagnostics. The training database manages a variety of engine models by replacing training data with diagnostic and prognostic information in a model training feedback loop with engine manufacturers. The environmental algorithm for real-time emissions rate measures the most discriminant temperature, pressure, and emissions for idle time and proposed torque sub-ranges. The algorithms are portable to passenger buses, fire trucks, police vehicles, snow plows, street sweepers, traffic management vehicles, and construction vehicles that often perform daily short, low-speed, and stop-and-go cycles and are driven by student drivers. Pilot implementations performed for school buses in Los Angeles have shown promising results. The full proposed solution can be implemented using existing resources in the transportation community.