Vessel Performance: Traditional Methods vs Machine Learning in the Era of Commercial Pressure

The shift: from a simple operational metric to a commercial and regulatory imperative

Driven by new market realities, vessel performance monitoring has evolved from a secondary operational task into a commercial and regulatory need. For a shipping company, the efficiency of a fleet is no longer just about bunker savings, but it is critical for regulatory compliance, particularly with the tightening of IMO’s regulations. Superior efficiency directly translates to lower operational costs by minimising exposure to emission control taxes and carbon pricing schemes like -one of shipping’s least favorite acronyms right now- the EU ETS.

Geopolitical disruptions and the volatility of bunker prices

This is further underscored by the impact of current geopolitical disruptions on global trade; the resulting rerouting and significantly longer transit distances, combined with increasingly volatile bunker prices, have heightened fuel costs and made performance monitoring a necessity to protect margins

Vessel performance: a competitive asset in the chartering market?

A well-performing vessel fleet not only meets regulatory requirements but also secures a premium in the chartering market. Charterers are increasingly selective, prioritising "green" ships that offer better energy efficiency ratings and lower fuel consumption, which, in turn, enhances their own profit margins and sustainability profiles. For a vessel to remain a competitive asset that generates cash, performance management becomes essential.

The traditional approach to vessel performance monitoring

Marine engineers have conventionally assessed and understood vessel performance using static performance curves. This involves filtering data for nominal weather and draft conditions, and then plotting key relationships like Power vs. Speed, Power vs. RPM, Fuel vs. Speed, and Speed vs. RPM. These plots allow engineers to visualise the deviation of the vessel's current state from baselines such as its sea trial results or a nominal propeller curve. This methodology was formalised by standards such as ISO 19030, which provides a framework for measuring changes in hull and propeller performance. As seen in Figure 1, raw data from different periods is averaged and compared against a reference baseline. By overlaying data from one period against another - for instance, comparing a clean hull period to a period six months later - engineers can estimate the degradation in performance caused by biofouling or paint degradation. These plots serve as the traditional marine engineering test, providing a visual representation of how much additional power (or fuel) is required to maintain a specific speed through the water (or rpm).

Figure 1: P-V comparison between two periods in time showcasing significant performance degradation

The Speed Through Water (STW) data is categorised into 1-knot increments, with the median power demand calculated for each discrete bin (Binned Avg). The baseline binned averages from Period 1 (P1 Overlap) are displayed on the Period 2 plot to highlight performance differences only in common operating speeds.

The limitations of traditional vessel performance methodologies

Despite their obvious utility, these traditional methodologies suffer from significant limitations. The primary challenge is the "normalisation" of data; it is exceptionally rare to find enough data points collected under identical environmental conditions (draft, trim, sea state, and wind) to make a truly "apples-to-apples" comparison. Perhaps the most restrictive drawback is the inability to view the vessel's life cycle holistically. These methods typically allow for the comparison of only two distinct periods at a time - often a "baseline" versus a "current" state. This creates a fragmented view of performance, making it difficult to pinpoint the exact moment a fouling event began or to evaluate the long-term effectiveness of different hull coatings across multiple years of operation.

What if we could transcend these snapshots and instead analyse the continuous, complex interactions between every variable affecting a vessel's efficiency?

Machine Learning: From static comparisons to continuous intelligence and true AI-powered efficiency

Imagine a technology capable of identifying deep patterns across thousands or even millions of data points, creating a multi-dimensional model that accounts for every environmental nuance. This is where Machine Learning (ML) revolutionises marine engineering. DeepSea's Cassandra Platform is the AI-driven vessel performance insights platform that makes this a reality. Rather than comparing two isolated periods, our ML models generate a continuous "Power Loss" or "Fouling Curve" over the entire lifespan of the vessel, providing the actionable insights you need to optimise, every nautical mile.

Figure 2: Performance degradation expressed as power loss (%) over an extended period of time.

As illustrated in the dynamic trendlines of Figure 2, Machine Learning can quantify daily performance deviations with unmatched precision. This is the power of true AI-driven performance. It provides a real-time, evolving narrative of your vessel's efficiency, delivering the actionable insights that allow owners to transition from reactive maintenance to a decisive, data-driven, proactive strategy that optimises every single nautical mile.

As regulatory pressure intensifies, fuel markets remain volatile, and charterers increasingly prioritise efficiency, vessel performance management is becoming a strategic capability rather than a technical afterthought. Traditional engineering approaches laid the foundation for understanding vessel behaviour, but the complexity of modern operations demands a more dynamic perspective.

Thanks to Machine Learning ability to analyse a vessel’s performance continuously, across its lifecycle, we can now gain a deeper understanding of inefficiencies and new opportunities for optimisation. In an environment where every percentage of efficiency matters, the ability to transform data into proactive decision-making will define the next generation of high-performing fleets.