Event overview
In ‘ML for Maritime: A robust and generalizable fuel prediction model for vessel performance,’ Cecile shared how Spinergie’s Vessel Performance team developed a robust machine learning model designed to predict vessel fuel consumption. She shared how our hybrid approach combines physics-based principles with data-driven techniques to create this key component of Smart Fleet Management.
About the Event
Key Insights
Multi-source data pipeline: we ingest GPS/AIS, onboard sensors, weather providers, and noon reports to build a rich picture of vessel operations.
Physics-informed ML for reliable, explainable results: our model is grounded in naval physics equations, making predictions interpretable, robust, and accurate even under unseen weather and navigation conditions.
One architecture, vessel-specific models: a single model structure is trained individually on each vessel’s own data, making it specific to that vessel’s hull and engine characteristics and easily deployable across any vessel type.
Transfer learning for sensor-less vessels: for vessels without onboard sensors, we fine-tune our models using aggregated fuel reports, extending coverage across the entire fleet.
A full product to understand and optimize fuel consumption: beyond prediction, the platform helps understand performance with trusted numbers, explain variances (weather vs. technical vs. operational), detect degradation or sub-optimal behaviours, optimize voyages, and benchmark fleet emissions.
Speakers
.jpg)
Cécile Daniel
Data Scientist
Topics discussed
EVENT
Paris Women in Machine Learning and Data Science
9/3/2026
GitGuardian, Paris
