ZeroMission & University College Cork: Harnessing Academic Excellence to Power AI-Driven Predictive EV Fleet Maintenance
ZeroMission is proud to present the results of a strategic research collaboration with University College Cork titled “Exploring Advanced AI-Driven Predictive Fleet Maintenance.” As electric vehicle (EV) fleets scale rapidly worldwide to meet sustainability and decarbonisation goals, conventional maintenance paradigms struggle to keep pace with the complexity and dynamics of EV systems. This partnership tapped into academic rigor and the brightest emerging minds to develop a four-layer predictive maintenance framework that transforms disparate fleet data into real-time, actionable intelligence, enabling proactive, efficient, and resilient fleet operations.
Background: The Maintenance Challenge in Electric Fleets
The electrification of fleets introduces new technical complexities. Unlike internal combustion engine vehicles, EVs rely heavily on interdependent subsystems such as high-voltage battery management, regenerative braking, and sophisticated power electronics. Traditional time- or usage-based maintenance approaches are insufficient; they either lead to unnecessary downtime or fail to detect early degradation. With ubiquitous telematics, IoT sensing, and cloud diagnostics, modern EVs generate rich, continuous data streams, yet many operators (including those supported on platforms like ZeroMission) face barriers in converting raw sensor data into operational insight due to fragmentation, inconsistent formats, and limited AI integration capacity.
Recognising this gap, ZeroMission and UCC embarked on a collaborative research initiative to build a scalable, practical solution: a conceptual, four-layer AI-driven framework tailored to the realities of EV fleet operation.
Research Collaboration: Industry Meets Academic Talent
The collaboration brought together ZeroMission’s operational domain expertise and University College Cork’s academic excellence. Under the mentorship of Alan Crowley, COO of ZeroMission, and the academic guidance of Lecturer Stephen Treacy, a team of student researchers, Yuan Chen, Zizheng Liu, Christopher Sabu, Qixuan Xie, and Priya Thomas Kutty, took on the challenge of designing and articulating a system that could turn complex, noisy, multi-source fleet data into predictive maintenance intelligence.
This partnership exemplifies how industry-academic collaboration, when anchored in mutual purpose and support, can accelerate innovation and yield solutions that are both technically rigorous and operationally relevant.
The Four-Layer Predictive Maintenance Framework
At the heart of the research is a four-layer AI framework structured to progressively refine and elevate raw fleet data into decision-ready outputs:
Data Collection
The foundational layer aggregates heterogeneous data streams including charging behaviour, driving patterns, vehicle telematics, and maintenance logs via vehicle-mounted remote information systems. The emphasis is on building a comprehensive and high-fidelity operational profile for each asset, enabling visibility into what, when, and how vehicles are being used and charged.Data Processing
Raw data is inherently noisy and inconsistent across sources. This layer applies structured preprocessing techniques such as outlier detection, normalization, and synthetic feature generation to standardize and enrich the dataset. The goal is to create a clean, analytically fertile basis for downstream modeling, addressing challenges like missing values, varying scales, and heterogeneous logging formats.AI Analysis
With preprocessed data in hand, advanced machine learning models are deployed to derive predictive insights. The research explores models including Long Short-Term Memory (LSTM) networks and Random Forests to forecast component health, remaining useful life (RUL), failure probabilities, and to contextualize charging behaviour against external variables like electricity pricing and infrastructure constraints. This layer is designed to assimilate multi-modal inputs (e.g., battery aging metrics, route load, charging patterns) into holistic assessments of fleet health and operational risk.Decision Support (Charging & Maintenance)
The final layer bridges model outputs to real-world action. Predictions are surfaced in a visual dashboard that translates probabilities and trends into maintenance warnings, optimized charging schedules, and actionable recommendations for fleet managers. The framework thus enables a shift from reactive troubleshooting to intelligent, schedule-aware, and cost-sensitive operational planning.
Operational and Strategic Impact
The proposed framework advances EV fleet management along multiple dimensions:
Reduced Downtime: By predicting failures ahead of time, fleets can schedule maintenance during low-impact windows, preserving availability.
Cost Efficiency: Targeted servicing avoids unnecessary part replacement and prevents cascading failures that incur higher repair costs.
Battery Longevity: Monitoring health and usage patterns through AI extends asset life by avoiding stress-inducing patterns and optimizing charge cycles.
Dynamic Charging: Integrating AI with external signals (e.g., time-of-use electricity pricing, grid load) enables smarter charging that balances operational need with cost and sustainability goals.
Scalability: The modular architecture allows adaptation to different vehicle types, geographies, and evolving sensor suites, making it future-resistant.
Why Talent-Driven Collaboration Matters
This initiative underscores a broader truth: solving complex, data-intensive operational problems requires not only tools and infrastructure but also fresh perspectives and intellectual investment. By engaging students directly in the design, modeling, and synthesis of the framework, with mentorship from industry leaders, ZeroMission gained access to creative problem-solving, while the researchers benefited from exposure to real-world constraints and impact-oriented thinking. The result is a bridge between academic insight and deployment-ready innovation.
Looking Ahead: From Research to Deployment
The framework developed through this collaboration provides a strategic blueprint. ZeroMission is now focused on operationalising these insights, integrating the predictive layers into live fleet environments, refining models with ongoing telemetry, and expanding dashboard capabilities for different user personas (e.g., operations managers, maintenance planners, and sustainability officers).
Further planned directions include:
Real-time adaptive learning to tune models as fleet usage evolves.
Incorporation of external data sources (weather, traffic, energy market signals) for richer context-aware prediction.
Fleet-level optimization that balances maintenance windows, vehicle availability, and route assignment.
Extension into vehicle-to-grid (V2G) scenarios to manage bidirectional energy flows in tandem with health-aware scheduling.
Acknowledgements
This work would not have been possible without the dedication and collaboration of many individuals and institutions. ZeroMission extends its sincere gratitude to Alan Crowley, COO of ZeroMission, whose mentorship and strategic insight guided the project’s vision; Lecturer Stephen Treacy of University College Cork for his academic stewardship; and the exceptional student researchers Yuan Chen, Zizheng Liu, Christopher Sabu, Qixuan Xie, and Priya Thomas Kutty for their technical contributions and commitment to translating complex theory into practical workflows.
Conclusion
The ZeroMission–UCC research programme demonstrates the multiplier effect of aligning industry needs with academic talent. By co-creating an AI-driven predictive maintenance framework tailored for modern EV fleets, this collaboration sets a new standard for data-to-decisions in sustainable transportation. As the project transitions from insight to implementation, ZeroMission remains committed to evolving smarter, more resilient fleet ecosystems, powered by collaboration, innovation, and the brightest minds.