Voyage performance evaluation based on a digital twin model

Abstract

Using digital twins in voyage performance evaluation is becoming critical for ocean vessels to reduce GHG emissions. A novel GBM approach is proposed in this paper to establish a digital twin model for voyage performance prediction. The weather hindcast data are introduced to enrich noon reports (NR) and automatic identification system (AIS) datasets, which are split into training and validation sets to develop GBM. The NR and AIS datasets collected from a 57000DWT bulk carrier are used to demonstrate the fidelity and capability of the proposed GBM. The voyage performance prediction from the GBM shows better accuracy than those from pure WBM or pure BBMs. An arrival time forecast and a weather routing showcase are also presented to demonstrate the application effects of GBM. The proposed GBM provides a satisfying prediction of ship speed and fuel consumption without mandatory sensor-collected data, thus applicable for a varity of vessels. In those cases where more sensors are available onboard, the proposed approach can incorporate sensor data to improve the model accuracy further.

Ship’s Digital Twin—A Review of Modelling Challenges and Applications

“…Modelling methods used for DT development in the reviewed literature vary based on scope of application, starting from conventional mathematical modelling and statistical analysis used in [13,15,17,18,21,22], the finite elements method (FEM) used in [11,12,16,20,21,26], data-driven modelling (DDM) used in [8,[12][13][14], artificial neural networks (ANN) used in [8][9][10][11], grey-box modelling (GBM) in [10,27], cycle mean value (CMV) in [15,19], boundary element method (BEM) used in [24], model based systems engineering (MBSE) used in [23], and model-based design (MBD) as used in [25].…”

“…BBM represents the input-output relation based on acquired data, thus being based entirely on empirical data, while WBM represents purely theoretical modelling based on system equations, as stated in [27]. In between the two, GBM combines acquired data with the theoretical model, and its use has been proposed in [10] to develop a DT predicting fuel consumption and ship speed. Moreover, a GBM method has been proposed in [27] to develop a DT model of a boost converter as part of a shipboard power system.…”

“…For data processing purposes, various methods have been used, such as filtering outliers, as in [8,13], filtering noisy data, as in [18], interpolation and calculation of wave spectra, as in [16], data fusion, as in [10,12], calculating root-mean-square (RMS) values from raw data, as in [13,20], convex optimisation, as in [21], skewness and kurtosis analysis, as in [13], and Fourier transformation, as in [26].…”

“…The SDT concept relies on data it would expect for the advanced data modelling methods, such as artificial intelligence, used in [8][9][10][11][12], and data-driven modelling (DDM), used in [12][13][14], to take prime in this area. However, conventional mathematical models have often been used for the purposes of creating an SDT, as in [15][16][17][18][19][20][21][22].…”

“…SDT applications found in papers [8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27] were analysed according to the standards of the DT formulation process set in this paper. For every case study, methods used to form a DT, applications, and steps of the DT formulation process were analysed.…”