Abstract:Ship transit forecasting can provide advanced movement information to a multidimensional wave motion compensation system. However, traditional deep learning models based on data-driven or physical data have difficulty in accurately predicting the rapid variations in ship motion attitudes that frequently occur in complex ocean environments. Therefore, a premonition-driven deep learning model is proposed in this paper to significantly improve the prediction accuracy of ship violent roll motion. This model applies precursor information to predict premonitory values of ship prospective motion attitudes based on an integrated statistical regression model and then employs these premonitory values to revise the outputs of the deep learning time series model. First, a hull attitude premonition mechanism is presented based on biological premonition cognitive behaviour, which captures the correlation relationships between extreme values by downgrading the globally continuous time series prediction problem to a time-independent regression problem. The corresponding methods for predicting local extreme values and deep feedforward neural network cluster are established based on these correlation relationships between the extreme values to achieve precise premonitory values. Second, the time and amplitude variables of the extreme angle values are transformed to countdown time variables and sequential sets of premonitory angle values. Two types of premonition units and their respective premonitory long short-term memory neural networks are constructed to upgrade the regression problem to a globally continuous time series prediction problem. Finally, a single-step output recurrent strategy is applied to forecast the short-term ship motion attitude for fifteen seconds by considering eleven initially predicted points across the two datasets. In sixty-six simulation experiments, the mean absolute error, root mean squared error, and mean absolute percentage error of the premonition-driven long short-term memory neural networks are considerably lower than those of traditional deep learning models. These simulation experiments verify that the premonition-driven deep learning model can effectively forecast short-term violent ship roll motion and provide safeguards for cargo transfer.

Nonparametric modeling of ship maneuvering motion based on self-designed fully connected neural network

Ship collision risk analysis: Modeling, visualization and prediction

Nonparametric modeling of ship maneuvering motions in calm water and regular waves based on R-LSTM hybrid method

On Neural Network Identification for Low-Speed Ship Maneuvering Model

Adaptive neural network asymptotic path-following control of underactuated ships with stochastic sea loads

Data-Driven system identification of 6-DoF ship motion in waves with neural networks

A Multi-Fidelity Prediction Model for Vertical Bending Moment and Total Longitudinal Stress of a Ship Based on Composite Neural Network

Application of the Improved BP Neural Network in Ship and Ocean Engineering

Prediction of wave-induced ship motions based on integrated neural network system and spatiotemporal wave-field data

A novel model-based parameters estimation combining local optimization and global optimization of nonlinear ship models with physical experiment dataset

A Data-Driven Method for Ship Motion Forecast

Dynamic nonparametric modeling of sail-assisted ship maneuvering motion based on GWO-KELM

Ship Motion Attitude Prediction Based on an Adaptive Dynamic Particle Swarm Optimization Algorithm and Bidirectional LSTM Neural Network

A hybrid ship-motion prediction model based on CNN–MRNN and IADPSO

Premonition-driven deep learning model for short-term ship violent roll motion prediction based on the hull attitude premonition mechanism

Adaptive Ensemble of Multi-kernel Gaussian Process Regressions based on Heuristic Model Screening for Nonparametric Modeling of Ship Maneuvering Motion

A Practical Estimation Method of Inland Ship Speed Under Complex and Changeful Navigation Environment

A data driven method for multi-step prediction of ship roll motion in high sea states

Online identification of a ship maneuvering model using a fast noisy input Gaussian process

A hybrid approach for forecasting ship motion using CNN–GRU–AM and GCWOA