Predictive maintenance algorithms based on anomaly detection make it possible to intervene just before an actual component failure occurs, reducing unnecessary spare parts replacements and optimising energy consumption in vehicles. This project combines efficiency in AI design (Green AI) with a direct positive impact on sustainability (AI for Green).
In traditional preventive maintenance models, interventions are scheduled based on a fixed calendar, regardless of the actual condition of the component. This leads to unnecessary waste and spare parts consumption, while undetected abnormal operations can cause excessive energy, fuel, or power usage.
The solution: anomaly detection and “real-time” maintenance
Small-scale algorithms were developed and optimised to run locally within vehicles with minimal resource consumption.
Main functions:
- Detect anomalies in operational data from multiple systems and components
- Learn behavioural patterns and predict imminent failures
- Enable maintenance scheduling just before actual failure, avoiding unnecessary replacements
Sustainability impact
- Waste reduction: only components that truly require replacement are changed
- Energy optimisation: identifies excessive fuel, battery, or power consumption caused by system malfunctions
- Operational availability: reduces unexpected downtime through predictive interventions
This approach generates a direct positive environmental impact by reducing emissions and resource consumption, while also extending component lifespan and improving overall fleet efficiency.
Last news
