Statistical modeling of Ship’s hydrodynamic performance indicator

Data Processing Framework for Ship Performance Analysis

“…There are several ways to establish (or obtain) a benchmarking standard, like model test experiments, full-scale sea trials, CFD analysis, etc. It may even be possible to establish a benchmarking standard using the in-service data recorded onboard a newly built ship, as suggested by Coraddu et al (2019) and Gupta et al (2021b). On the other hand, evaluating the current performance of a ship requires a good amount of data processing as the raw data collected during various voyages of a ship is susceptible to noise and errors.…”

“…The ship, of course, may have gone off-route as shown here, but referring to the GPS speed and heading of the ship at the corresponding time, shown in figure 4b, it is obvious that the navigation data is incorrect. Here, such an irrational position change can be detected through the two-stage steady-state (or stationarity) filter suggested by Gupta et al (2021b), based on the method developed by Dalheim and Steen (2020a). The first stage of the filter uses a sliding window to remove unsteady samples by performing a t-test on the slope of the data values, while the second stage performs an additional gradient check for the samples failing in the first stage to retain the misidentified samples.…”

“…Alternatively, it is suggested to observe the vertical shift of the calm-water speed-power curve, often termed as the change in power demand (adopted by Gupta et al (2021a) and Carchen and Atlar (2020)). Some researchers also formulated and used some indirect performance indicators like fuel consumption (Koboević et al (2019)), resistance (or fouling) coefficient (Munk (2016); Foteinos et al (2017); Carchen and Atlar (2020)), (generalized) admiralty coefficient (Ejdfors (2019); Gupta et al (2021b)), wake fraction (Carchen and Atlar (2020)), fuel efficiency (Kim et al (2021)), etc. In each of these cases, it is clearly seen (and most of the time acknowledged) that the results are quite sensitive to the quality of the data used to estimate the ship's performance.…”

Ship Performance Monitoring using Machine-learning

“…Ejdfors (2019) assumed admiralty coefficient to be constant for a given ship over a range of speed-power-displacement, therefore, it could be perceived as a summarized calm-water speed-power curve for a given displacement. Gupta et al (2021) analysed this assumption and concluded that this assumption only holds good for a generalized form of admiralty coefficient as the speed and displacement exponents in the original admiralty coefficient are not valid for modern hull forms. Gupta et al (2021), therefore, proposed to use the generalized admiralty coefficient (Δ ∕ ) as the statistical hydrodynamic performance indicator for a ship, with displacement and speed exponents, i.e., and , respectively, obtained statistical using the in-service data recorded onboard the ship.…”

“…Gupta et al (2021) analysed this assumption and concluded that this assumption only holds good for a generalized form of admiralty coefficient as the speed and displacement exponents in the original admiralty coefficient are not valid for modern hull forms. Gupta et al (2021), therefore, proposed to use the generalized admiralty coefficient (Δ ∕ ) as the statistical hydrodynamic performance indicator for a ship, with displacement and speed exponents, i.e., and , respectively, obtained statistical using the in-service data recorded onboard the ship. The current work uses the generalized admiralty coefficient to formulate the variables to account for the growth of marine fouling on the ship over time.…”