Electric buses with new ally in fault prediction: Artificial Intelligence

Combining advanced techniques in machine learning, deep learning, and eXplainable AI, a new system is being developed to predict operational faults and generate understandable explanations for maintenance teams. FOX-PM is a project bringing together INESC TEC and CaetanoBus to create an innovative online and explainable predictive maintenance system for electric bus fleets. The goal is to increase system reliability, reduce downtime, and optimise resource use.

CaetanoBus currently operates a vast fleet of buses in several cities worldwide. In this context, the vehicles may operate under different conditions, influenced by external factors like temperature, usage patterns like driving style or route, or even natural wear and tear.

“These variations do not necessarily correspond to faults,” explained Rita Ribeiro, researcher at INESC TEC. “That’s why FOX-PM aims to develop models capable of automatically recognising and adapting to different normal operating regimes, helping to reduce false alarms and improve the accuracy of anomaly detection.”

More specifically, the project focuses on the pneumatic system of CaetanoBus’ electric vehicles – a critical component for braking and suspension systems and one responsible for a significant number of vehicle malfunctions.

“Using deep learning models, the project seeks to detect potential failures in advance, across different time horizons, and to apply eXplainable AI (XAI) techniques that can transparently explain the generated alerts,” added the researcher. Furthermore, to improve the interpretability of these alarms, the team will explore natural language models capable of translating technical explanations into accessible language for operational teams, fostering trust and transparency in AI-based decisions.

This approach will help increase system reliability, reduce downtime, and optimise resource use. “We expect the project to reduce unexpected stoppages and increase the reliability of electric vehicles, while extending equipment lifespan and improving energy efficiency. By strengthening the reliability of urban electric transportation, FOX-PM ultimately promotes sustainable e-mobility and supports the green transition of cities,” noted Rita Ribeiro.

The researcher also points out that, given the lack of a standard framework for evaluating predictive maintenance systems, the project will also study metrics and assessment strategies for alarms in a Just-in-Time (JiT) Predictive Maintenance context.

FOX-PM – Failure Online eXplanation for Predictive Maintenance – is a three-year project developed in collaboration with CaetanoBus and funded under Portugal 2030.

The researcher mentioned in this news piece INESC TEC and the Faculty of Sciences of the University of Porto