how to exclude towing periods from AI vessel fuel baselines?

Fleet teams can keep AI fuel comparisons reliable by separating AI fuel consumption during towing events from ordinary sailing when building baselines.

How AI vessel fuel baselines Are Applied

To exclude towing periods from AI vessel fuel baselines, you need a consistent way to classify time and link that classification to the fuel and voyage consumption calculations. In practice, you treat towing as a distinct operating mode because towing resistance and tug assisted tow fuel profiles can distort normal performance trends, creating a voyage consumption exception that the AI should not learn as “baseline sailing.”

Create an operating-mode classification in voyage/engine data (e.g., Sailing, Towing, Manoeuvring, Port stay) and ensure towing periods are tagged at the same time granularity as fuel consumption records.
Use event-driven segmentation: define towing start and end using your operational logs (tug engagement, tow line connected, tow line released) and map those timestamps to the fuel time series before baseline training.
Apply a baseline exclusion rule: filter out all records where operating mode equals Towing (and optionally Manoeuvring if it includes tow-related positioning) so AI fuel baselines are trained only on comparable sailing conditions.
Validate data completeness and alignment: confirm that fuel measurement windows fully cover the tagged towing intervals and that there are no gaps or overlaps that would leak towing loads into “sailing” segments.
Document the exception logic: record the rationale and thresholds used to classify towing, including how you handle ambiguous cases (e.g., escorting with partial towing loads) and how you treat borderline towing resistance conditions.

Operational Impact

Improved cost attribution and budget visibility: excluding towing periods prevents inflated or deflated AI-derived fuel KPIs that would otherwise misstate OPEX for Fleet Managers and finance stakeholders.
Higher data confidence for model governance: Technical and AI readiness teams reduce the risk of training bias caused by towing loads, which are a known voyage consumption exception relative to normal sailing.
Better operational decision support: Technical Managers can compare like-for-like performance, making it easier to identify true efficiency issues rather than artifacts from tug assisted tow fuel conditions.

Important to know: If you cannot reliably tag towing start and end times, do not rely on “average over the voyage” filters. Instead, segment by the most trustworthy timestamps you have (tug engagement and tow line events) and then exclude those exact intervals from the baseline dataset so the AI fuel consumption during towing events cannot contaminate sailing baselines.