Shipping is responsible for over 90% of global trade. Although it is generally considered a safe and clean mode of transportation, it still has a significant impact on the environment. Thus, state-of-the-art models that may contribute to the sustainable management of the life cycle of shipping operations without compromising safety standards are urgently needed. This chapter discusses the potential of artificial intelligence (AI) based digital twin models to monitor ship safety and efficiency. A paradigm shift is introduced in the form of a model that can predict ship motions and fuel consumption under real operational conditions using deep learning models. A bi-directional long short-term memory (LSTM) network with attention mechanisms is used to predict ship fuel consumption and a transformer neural network is employed to capture ship motions in realistic hydrometeorological conditions. By comparing the predicted results with available full scale measurement data, it is suggested that following further testing and validation, these models could perform satisfactorily in real conditions. Accordingly, they could be integrated into a framework for safe and sustainable ship operations.
TopIntroduction
Shipping plays a crucial role in global trade, accounting for over 90% of the transportation of goods worldwide (UNCTAD, 2022). While it is generally considered a safe and environmentally friendly mode of transport, it still has a significant impact on the environment (UNFCCC, 2022). As a result, there is a growing need to ensure the safety and sustainability of ship operations by embracing modern technology.
Originally developed by NASA, digital twins involve creating a virtual replica of a physical object or system to simulate and analyse its performance in real-time (Allen, 2021; Zhang, Hirdaris, & Tsoulakos, 2023). They can be described as machines or computer-based models that simulate, emulate, mirror, or “twin” the existence of a physical entity. For example, a ship digital twin refers to a virtual replica or representation of a real ship. This digital replica can be created by combining real-time data, simulation models, and advanced analytics, allowing for visualization, analysis, and optimization that aims to assure safe and sustainable operations in real operations (see Figure 1).
Figure 1. The ship digital twin for proactive optimization of ship operations
It is important to acknowledge that the accuracy of physics-based models may be limited to “as build” design or when it comes to dealing with complex operational conditions and extreme events (Kaur et al., 2020). Through deep learning algorithms offer the potential to improve the predictive capabilities of digital twin models (Huang et al., 2021; Lee et al., 2022; Nielsen et al., 2022). To address the challenges associated with predictive analytics and visualisation hybrid methods should be employed. For example, digital twin optimal deep learning models are designed to accurately predict ship motions (Nielsen et al., 2022; Zhang, Taimuri, Zhang et al, 2023) and fuel consumption (Chen, Lam, & Xiao, 2023; Uyanık et al., 2020; Zhang, Tsoulakos, Kujala et al, 2023). They may be combined by utilizing advanced AI methods training physics-based idealisations using comprehensive big data sets. Such predictions may enable ship operators to make informed decisions and minimize the ecological footprint of fuel-efficient shipping operations.
This chapter introduces an AI-based digital twin designed for enhancing the safety and sustainability of ship operations. In Section 2, an overview of digital twin models applicable to maritime operations is presented. The AI-based digital twin comprises of two layers, with the primary objectives to estimate ship fuel consumption and predict ship motions (Section 2 for further details). To evaluate and demonstrate the practical application of this approach, full scale measurement data is utilized, focusing specifically on a bulk carrier and a RoRo/Passenger ship (RoPax) (see Sections 4 and 5). Section 6, highlights the promising prospects of AI-based digital twins in the realm of developing intelligent decision support systems and for effectively monitoring ship safety and efficiency.
TopAn Overview Of Digital Twin Models For Use In Maritime Operations
Ship operations relate to seakeeping and ship performance in adverse conditions (Hirdaris & Mikkola, 2021). Models that may be used for the estimation of ship fuel consumption or seakeeping can be broadly classified into two categories namely (a) physics-based models and (b) Artificial Intelligence (AI) models.