An unmanned ship real voyage monitoring and forecasting system based on a CFD numerical proxy model

Abstract

The application discloses an unmanned ship real voyage monitoring and forecasting system based on a CFD numerical agent model, which comprises an unmanned ship, an unmanned ship twin, a CFD numerical agent model, a remote visualization module and a digital twin interaction module. The CFD numerical agent model comprises a wave parameter inversion analysis module and a wave force forecasting module, which are used for constructing a nonlinear mapping relationship between the wave sailing mechanics and the motion state response characteristics of the unmanned ship and the controlled wave spectrum, inputting the real voyage unmanned ship motion response data sequence and the ship surrounding wave height monitoring data transmitted by the digital twin interconnection module, analyzing the wave environment characteristic parameters around the unmanned ship and the evolution state of the unmanned ship in the future Δt range, and further obtaining the force state of the unmanned ship. The system realizes real-time and efficient forecasting of the wave force suffered by the unmanned ship, and provides reliable safety analysis for the sailing safety of the unmanned ship.

Description

Technical Field

[0001] This invention relates to the field of unmanned surface vessel (USV) navigation, and in particular to a real-world monitoring and forecasting system for USVs based on a CFD numerical proxy model. Background Technology

[0002] While the autonomous navigation capabilities of unmanned surface vessels (USVs) have made significant progress, the corresponding monitoring and safety forecasting technologies are still far from perfect. USVs face complex and uncertain sea conditions in actual maritime environments, with surges dynamically changing over a wide area. Large ships experience significant rolling when encountering waves, and small and medium-sized USVs respond even more significantly to surges. USVs need the ability to analyze the wave composition in their immediate vicinity to select effective wave-crossing strategies, which relies on advancements in monitoring and safety forecasting technologies. USV execution platforms require a synchronous digital twin system for navigation support. This digital twin system needs to establish a mathematical connection with the actual navigation waters; a mere 3D model of a virtual scene cannot provide real-time guidance for actual operations. The digital twin system needs to calculate the sea conditions in the area where the USV is located, calculate the disturbances experienced by the USV, and assess the navigation safety of the USV. Existing virtual simulation platforms for unmanned surface vessels (USVs) mainly use image data transmitted from front-end camera sensors and the latitude and longitude coordinates of the USV as information for monitoring the USV's position and environment. They cannot perform numerical analysis of the wave environment in which the USV is located, cannot effectively calculate the mechanical effects of wave generation on the USV during its current navigation, and cannot provide reference information on environmental disturbances for selecting targeted propulsion strategies for the USV. This may result in wave damage and unnecessary power loss during the USV's navigation.

[0003] The research case of virtual simulation platform proposed in Chinese patent application No. CN202111214855.0 creates a three-dimensional virtual environment that can be used for autonomous navigation testing, but it lacks real-time visualization updates of the ship's hull status.

[0004] The virtual simulation system proposed in Chinese patent application CN202110198914.3 can monitor ship speed and main engine speed and calculate energy consumption ratio, but it does not detect the wave characteristics of the navigation area and fails to build a safety forecasting system.

[0005] The ship attitude prediction method proposed in Chinese patent application No. CN202110261215.9 directly predicts the future attitude of the ship according to the time series analysis method. However, it lacks the analysis of the real-time wave load on the hull. The time series analysis method does not consider the influence of ship maneuvering dynamics during actual navigation and fails to determine the wave level of the current sea area based on the knowledge base. Summary of the Invention

[0006] In order to overcome the above-mentioned shortcomings and deficiencies of the prior art, the purpose of this invention is to provide an unmanned surface vessel real-time monitoring and forecasting system based on a CFD numerical proxy model.

[0007] The objective of this invention is achieved through the following technical solution:

[0008] A real-world monitoring and forecasting system for unmanned surface vessels based on a CFD numerical surrogate model includes:

[0009] Unmanned surface vessel (USV): Equipped with left and right thrusters and an inertial navigation sensor array, with wave height monitoring points on the hull;

[0010] Digital twin interconnection module: used to receive encoder speed of the left and right thrusters of the unmanned surface vessel (USV), data from the USV's inertial navigation sensor array, and data from the USV's frequency wave height monitoring, to realize dynamic information interaction between the USV twin and the actual USV.

[0011] The CFD numerical proxy model, including a wave parameter inversion and analysis module and a wave force prediction module, is used to construct the nonlinear mapping relationship between the wave navigation mechanics and motion state response characteristics of the unmanned surface vessel and the controlled wave spectrum. It inputs the actual unmanned surface vessel motion response data sequence and the wave height monitoring data around the vessel transmitted by the digital twin interconnection module, analyzes the wave environment characteristic parameters around the unmanned surface vessel and their evolution state in the future Δt range, and further obtains the force state of the unmanned surface vessel.

[0012] Unmanned surface vessel twin: includes a motion attitude calculation module and an unmanned surface vessel virtual geometric model; used to receive data information from the digital twin interconnection module and the CFD numerical proxy model, drive the unmanned surface vessel virtual geometric model to follow the real motion of the unmanned surface vessel in the ocean wave scene, and display the motion state at the future time Δt in the remote visualization module;

[0013] The remote visualization module includes a programmable twin ocean interface and a fusion display interface for unmanned surface vessel (USV) capsizing safety assessment. The programmable twin ocean interface is used to visualize the dynamic perspective changes during the USV's wave-following motion. It changes the ocean view based on the USV's attitude and provides visual wave feedback based on wave morphology, simulating the ocean environment surrounding the real USV. The fusion display interface for USV capsizing safety assessment is used to visualize the results of the USV capsizing safety assessment.

[0014] Furthermore, the capsizing safety assessment of the unmanned surface vessel (USV) utilizes the rate of change of angular velocity to evaluate the attitude safety of the USV during stable navigation. Based on the wave resistance and torque experienced by the USV at this moment, calculated by the wave force prediction module, the rate of change of angular velocity (φ) within the future time interval Δt is calculated. x ,φ y ,φ zThe three-axis norm threshold m is defined by combining the magnitude of the three-axis inertia. The system monitors whether the value of m exceeds the danger threshold and displays the attitude and angular velocity change rate of the unmanned surface vessel at the future time Δt on the interface.

[0015] Furthermore, the unmanned surface vessel inertial navigation sensor array specifically comprises:

[0016] A pair of identical inertial navigation sensors are arranged at equal intervals from the center of mass on a horizontal plane passing through the center of mass;

[0017] A cluster of identical inertial navigation sensors are arranged at equal intervals from the center of mass on a longitudinal section passing through the center of mass.

[0018] Furthermore, the unmanned surface vessel inertial navigation sensor array employs a symmetric plane inertial navigation matrix optimization method for attitude measurement.

[0019] Furthermore, the wave parameter inversion and analysis module can analyze the six-degree-of-freedom attitude response characteristics and wave height monitoring data of the unmanned surface vessel under unknown wave conditions, and construct the composition of the real-time wave spectrum encountered by the unmanned surface vessel in actual operation. The wave parameters obtained by analysis include the main wave direction and the encountered wave speed, the main wave period, and the significant wave height, i.e., the maximum one-third peak spectrum.

[0020] The wave parameter inversion analysis module is a multi-input multi-output deep neural network model with twelve hidden layers. The training dataset of the model is taken from several pre-conducted CFD simulation cases under known superimposed wave conditions. The CFD simulation cases calculate the six-degree-of-freedom attitude response and wave force sequence of the unmanned surface vessel, as well as the wave height sequence at the monitoring points around the unmanned surface vessel.

[0021] The training process of the wave parameter inversion analytical module is as follows:

[0022] The roll, pitch, heave, and sway responses from the unmanned surface vessel's attitude response time series, along with the wave height sequences from all four monitoring points around the vessel, are transformed into the frequency domain. The entire frequency band is divided into N sub-segments. The characteristic amplitudes of each sub-segment are normalized and used as input to the wave parameter inversion analysis module. The data depth of its encoding range is M. The total input consists of equivalent grayscale image matrix blocks of size 8×N×M. The superimposed wave parameters from the CFD simulation case are used as the output. The superimposed wave parameters include the main wave direction, encounter wave speed, main wave period, and meaningful wave height, forming a nonlinear input-output dataset.

[0023] The nonlinear input-output dataset is input into the deep neural network model for training, and the parameters of each hidden layer node are constructed to form a wave parameter inversion analysis module.

[0024] Furthermore, the CFD simulation case uses an orthogonal parameter table to design its boundary conditions. The parameter table involves the equivalent encounter velocity and meaningful wave height in the heading direction. The wave frequency combination is selected according to the wave spectrum with a set step size. The wave frequency combination includes the main direction of incoming waves and the main period of waves.

[0025] Furthermore, the wave force prediction module, based on the characteristic wave field information of the current unmanned surface vessel, can predict the bidirectional wave drift force on the current vessel under known wave characteristic parameters and the characteristic state of the vessel in the waves, thus avoiding the need for lengthy CFD calculations.

[0026] Furthermore, the wave force prediction module is specifically a multi-input multi-output residual neural network;

[0027] During training, the input parameters are the attitude response data within the historical window in the CFD simulation case dataset, and the initial wave parameter matrix is ​​used for convolution processing. The output parameters are the bidirectional wave drift forces experienced by the vessel at each time point within the historical window.

[0028] Furthermore, the programmable twin ocean interface is specifically described using the PM wave spectrum.

[0029] Furthermore, the assessment of the capsizing safety of the unmanned surface vessel during navigation is based on the following formula:

[0030]

[0031] Using the calculated total wave drag torque, the thruster thrust, and the measured hull angular velocity at this moment, the auxiliary parameter of angular velocity (φ) is solved. x ,φ y ,φ z The ratio of the angular velocity of the hull around each axis at two adjacent moments is used to comprehensively monitor the auxiliary parameters of angular velocity. The angular velocity should not exceed the safety threshold. The hull attitude of the unmanned surface vessel is displayed in an orthographic axonometric view on the digital sea surface. The initial coordinates of each point on the hull and the current attitude coordinates in the axonometric view satisfy the following relationship, providing the operator with a convenient observation view to understand the hull attitude status.

[0032]

[0033] Because the three axes of inertia are different, a three-axis norm threshold metric m is defined to comprehensively examine the capsizing safety of the unmanned surface vessel (USV) along each axis, where m = max{(φ x ,φ y ,φ z )2,ξ x φ x ,ξ y φ y ,ξ z φ zWhen this value exceeds the danger threshold, a warning is issued to the unmanned surface vessel indicating that the maneuvering angle is too large.

[0034] Furthermore, the boundary conditions are specifically defined as follows: the interval is divided according to the equivalent encounter velocity of the unmanned surface vessel in the bow direction, and then the range of the significant wave height expected to be encountered by the vessel is 1 / 16h. m ~h m According to Δh=1 / 16h m The wave height level intervals are divided according to the Fourier decomposition principle. Each level wave train is represented by the superposition of several wavelets. Under each equivalent encounter velocity, the six wavelets of four different frequency wave height combinations are superimposed in the frequency domain. Each wavelet corresponds to a main wave direction and directional distribution function spectrum.

[0035] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0036] This invention enables unmanned surface vessels (USVs) to achieve deep mathematical perception of the actual sea conditions during navigation, realizes real-time and efficient prediction of wave forces acting on USVs, provides reliable safety analysis for USV navigation safety, facilitates efficient and precise control of USVs when navigating in actual sea conditions, and enables remote monitoring of USV status. Attached Figure Description

[0037] Figure 1 This is a system structure diagram of the present invention;

[0038] Figure 2 This is a schematic diagram of the CFD pre-simulation case design of the present invention;

[0039] Figure 3 This is a schematic diagram of the topology of the unmanned surface inertial navigation matrix array structure of the present invention;

[0040] Figure 4 This is a schematic diagram of the interface of the visualization module of the present invention. Detailed Implementation

[0041] The present invention will be further described in detail below with reference to the embodiments, but the implementation of the present invention is not limited thereto.

[0042] like Figures 1-4 The unmanned surface vessel (USV) real-time monitoring and forecasting system based on a CFD numerical proxy model consists of modular components, including: an USV, a digital twin interconnection module, an USV twin, a CFD numerical proxy model, and a remote visualization module.

[0043] The unmanned surface vessel is equipped with left and right thrusters and an inertial navigation sensor array, and wave height monitoring points are set on the hull.

[0044] The digital twin interconnection module is used to receive encoder speed data from the left and right thrusters of the unmanned surface vessel (USV), data from the USV's inertial navigation sensor array, and data from the USV's circumferential wave height monitoring, thereby enabling dynamic information interaction between the digital twin USV and the actual USV.

[0045] The digital twin interconnection module receives and publishes the rotational speeds of the UAV's left and right thruster encoders, data from the UAV's inertial navigation sensor array, and data from the UAV's frequency height monitoring. Servers and clients are deployed on the UAV and shore-based digital twin backend, respectively. Raw sensor data from the UAV is transmitted using message packets with specified data flags, input to the UAV's industrial control computer via RS232-USB serial communication, and then encapsulated into messages. The industrial control computer system receives the data, creates message handles, publishes the messages via a data radio, and repeats the message twice. The data radio message is then uploaded to the server's web interface according to the Socket protocol. The shore-based digital twin backend receives the messages from the cloud, further refactors them into a Unity-compatible data format, and interprets the inertial navigation sensor information using the same method.

[0046] The CFD numerical proxy model, including a wave parameter inversion and analysis module and a wave force prediction module, is used to construct the nonlinear mapping relationship between the wave navigation mechanics and motion state response characteristics of the unmanned surface vessel and the controlled wave spectrum. It inputs the actual unmanned surface vessel motion response data sequence and the wave height monitoring data around the vessel transmitted by the digital twin interconnection module, analyzes the wave environment characteristic parameters around the unmanned surface vessel and their evolution state in the future Δt range, and further obtains the force state of the unmanned surface vessel.

[0047] The wave parameter inversion and analysis module analyzes the six-degree-of-freedom attitude response characteristics and wave height monitoring data of the unmanned surface vessel (USV) under unknown wave conditions. It constructs the composition of the real-time wave spectrum encountered by the USV in actual operation. The obtained wave parameters include the main wave direction and the encountered wave speed, the main wave period, and the significant wave height, i.e., the maximum one-third peak spectrum.

[0048] The inversion analytical model is a multi-input multi-output (MIMO) deep neural network (DNN) model with twelve hidden layers. The training dataset for this model is taken from several pre-conducted CFD simulation cases under known superimposed wave conditions. These simulation cases calculate the six-DOF attitude response and wave force sequences of the unmanned surface vessel (USV), as well as the wave height sequences at monitoring points around the USV. Segments with ergodic characteristics are selected, and the roll, pitch, heave, and sway responses from the USV's attitude response time series, along with the wave height sequences at all four monitoring points around the USV, are transformed into the frequency domain. The entire frequency band is divided into N sub-segments, and the characteristic amplitudes of each sub-segment are normalized and used as input. The data depth of the encoding range is M, and the total input consists of an 8×N×M equivalent grayscale image matrix block. The superimposed wave parameters from the CFD wave cases are used as output, including the wave principal direction, encounter wave speed, wave principal period, and meaningful wave height. This constitutes a nonlinear input-output dataset, which is then input into the deep neural network model for training, constructing the parameters of each hidden layer node, and forming the wave parameter inversion module.

[0049] The CFD case set employs an orthogonalized parameter table to design its boundary conditions. This parameter table includes the equivalent encounter velocity in the heading direction, semantic wave height, and frequency combinations. The frequency combinations implicitly contain two wave information parameters: the dominant direction of the incoming wave and the dominant wave period. The PM wave power spectrum S under the combined influence of multiple parameters is then used. w (ω,H 1 / 3 ,ψ m and the directional extension function R w (ω,H 1 / 3 ), thus obtaining the wave direction spectrum S w (ω,H 1 / 3 ,ψ m ) = R w (ω,H 1 / 3 )×S w (ω). Where H 1 / 3 ,ψ m The statistical three-dimensional wave height and main wave direction are defined for the sea waves in the preset sea state. The equivalent encounter current velocity span is set to 1.5-9 m / s (3-18 knots) with a step size of 1.5 m / s for the bow direction; the meaningful wave height span is 0.025-0.4 m with an inter-group step size of 0.025 m; under each equivalent encounter current velocity, six sub-waves of four different frequency wave height combinations are superimposed in the frequency domain; this method is used to simulate the encounter wave field under various unmanned surface vessel (USV) navigation environments, and to calculate the motion and mechanical responses of the USV over a long time span.

[0050] The characteristic parameters of wave conditions refer to the initial conditions of the CFD case, namely the equivalent encounter current velocity, meaningful wave height, dominant direction of incoming waves, and dominant wave period. After setting these initial conditions in the CFD case, numerical calculations are performed on the motion of the unmanned surface vessel (USV) in the wave field. Wave height monitoring points around the USV are set up, and the wave height at these locations is recorded, using the initial centroid coordinate system of the USV as the reference coordinate system. After the calculation is completed, the time series of the USV's six-degree-of-freedom attitude response and the wave force sequence, as well as the time sequence of the wave height values ​​at the monitoring points around the USV, can be obtained.

[0051] The wave force prediction module, based on the characteristic wave field information of the current unmanned surface vessel (USV), predicts the bidirectional wave drift force experienced by the USV under known wave characteristic parameters and the vessel's characteristic state within the waves, avoiding lengthy CFD calculations. A multi-input multi-output residual neural network is pre-trained. The input parameters are the attitude response data within a historical window from the simulation case dataset. Convolution processing is performed using the initial wave parameter matrix, and the final output data is the bidirectional wave drift force experienced by the USV at each time point within the historical window (including that moment). The initial wave parameters include the equivalent encounter wave speed, meaningful wave height, main wave direction, and main wave period. The attitude response data includes the vessel's roll, pitch, sway, and heave attitude responses. The network has twenty hidden layers. Each hidden neuron successively performs different order difference quotient operations on the historical attitude response data and performs cross-correlation encoding operations with the initial wave parameter matrix, constructing several residual modules for inter-layer output coupling, and passing the output to the next layer through activation functions at each level. The training data consists of a CFD case set. Each case has a 5-second lead time for the first sample, and subsequent cases are divided into several training samples with a time step of 0.02 seconds. The wave force load currently experienced by the unmanned surface vessel is calculated in the forward algorithm and output to a remote visualization module to further predict the short-term wave force change trend in the capsizing safety assessment.

[0052] In this embodiment, the centroid refers to the equivalent center of action of the total weight of the unmanned surface vessel.

[0053] The unmanned surface vessel twin includes a motion attitude calculation module and an unmanned surface vessel virtual geometric model. It is used to receive data information from the digital twin interconnection module and the CFD numerical proxy model, drive the unmanned surface vessel virtual geometric model to follow the real motion of the unmanned surface vessel in the ocean wave scene, and display the motion state at the future time Δt in the remote visualization module.

[0054] The motion attitude calculation module receives data from the digital twin interconnection module, specifically by reading and calculating the equivalent thrust of the left and right thrusters according to a predetermined protocol. During unmanned surface vessel (USV) navigation, the continuous high-frequency action of waves causes significant disturbance to the inertial navigation sensor data, making it unreliable to characterize the overall attitude state solely based on single-sensor data. This invention introduces a symmetrical inertial navigation matrix to optimize attitude measurement. A pair of identical inertial navigation systems are arranged equidistantly from the center of mass on a horizontal plane passing through it, and the same symmetrical arrangement is used on a longitudinal section passing through the center of mass. Based on the principle of rigid body motion, the angular velocities measured at each point are consistent, allowing for a weighted average in the same direction to obtain the angular velocity of the center of mass. The weights are adaptively adjusted based on multiple experiments. Since each point spins around the Earth's coordinate system, it generates an acceleration component around the center of mass. By symmetrically arranging a pair of sensors equidistant from the center of mass axis, it is ensured that the axial components of the additional rotational acceleration have specific numerical relationships. By assigning adaptive weights and performing calculations, the absolute acceleration of the center of mass can be obtained. The shipboard industrial control computer receives and identifies data from each individual unit of the inertial navigation sensor array, performs joint calibration using the following mathematical method to obtain the current precise acceleration, combines angular velocity and other sensor data to perform iterative calculations for dead reckoning, solves for the current model state parameters, and pre-writes this data processing program as a C# script. In the formula, a... ex ,a ey ,a ez The axial components representing the central acceleration. α 1x ,α 1y ,α 1z This represents the weighting coefficient of the local acceleration of one of the inertial navigation systems in a pair of symmetrically arranged inertial navigation systems in each axis component. The specific value is calculated and selected based on the calibration test results before actual use. It is not fixed as a constant in order to accommodate actual installation errors. These represent the tangential and radial spin acceleration components in the geodetic coordinate system relative to the centroid of the sensor's installation location, respectively.

[0055]

[0056]

[0057]

[0058] After identifying the thrust of the unmanned surface vessel's (USV) propulsion system and its current calibrated attitude, these are written into the center-of-mass state interface C of the USV twin geometric model. t (α,β,θ) represents the attitude angles of the unmanned surface vessel along each axis. It defines the attitude orientation of the vessel in the NE-T coordinate system fixed at the origin. Considering the continuity and smoothness of the unmanned surface vessel's motion and to reduce the impact of the data sampling rate R limitation, the following smoothing factor μ is introduced for state updates.

[0059]

[0060] In this embodiment, the sensor is composed of a quaternary attitude sensor, and the thruster is a virtual thruster.

[0061] The remote visualization module includes a programmable twin ocean interface and a fusion display interface for unmanned surface vessel (USV) capsizing safety assessment. The programmable twin ocean interface is used to visualize the dynamic perspective changes during the USV's wave-following motion. It changes the ocean view based on the USV's attitude and provides visual wave feedback based on wave morphology, simulating the ocean environment surrounding the real USV. The fusion display interface for USV capsizing safety assessment is used to visualize the results of the USV capsizing safety assessment.

[0062] The unmanned surface vessel (USV) capsizing safety assessment utilizes the rate of change of angular velocity to evaluate the attitude safety of the USV during stable navigation. Based on the wave resistance and torque experienced by the USV at the current moment calculated by the wave force prediction module, the rate of change of angular velocity (φ) within the future time interval Δt is calculated. x ,φ y ,φ z The three-axis norm threshold m is defined by combining the magnitude of the three-axis inertia. The system monitors whether the value of m exceeds the danger threshold and displays the attitude and angular velocity change rate of the unmanned surface vessel at the future time Δt on the interface.

[0063] The perspective from a real-world unmanned surface vessel (USV) is simulated on the display device, including the dynamic changes of real-time ocean waves. The programmable ocean interface is described using the PM wave spectrum. Wave parameters obtained from the USV attitude inversion module drive the direction function and power spectrum to digitally adjust the wave surface, which is then displayed in a background visualization. Additionally, the capsizing safety of the USV during navigation is assessed, with the control equation as follows: [Equation omitted for brevity]. The left side of the equation contains ω... x0 ω y0 ω z0 The right side of the equation represents the axial components of the angular velocity at the current moment; F Rx ,F Rz The two-axis metered thrust component represents the right thruster. This represents the equivalent center of action of the right thruster in a geodetic coordinate system fixed at the center of mass. It represents the equivalent wave torque.

[0064]

[0065] Using the calculated total wave drag torque, the thruster thrust, and the measured hull angular velocity at this moment, the auxiliary parameter of angular velocity (φ) is solved. x ,φ y ,φ zThis refers to the ratio of the angular velocities of the hull around each axis at two adjacent moments. The auxiliary parameters for comprehensive monitoring of angular velocity should not exceed the safety threshold. The attitude of the unmanned surface vessel is then displayed in an orthographic axonometric view on the digitized sea surface. The initial coordinates of each point on the hull in the axonometric view and the current attitude coordinates satisfy the following relationship, providing the operator with a convenient observation view to understand the attitude status of the hull.

[0066]

[0067] Because the inertia of the three axes differs, a three-axis norm threshold metric *m* is defined to comprehensively assess the capsizing safety of the unmanned surface vessel (USV) along each axis. *m* = max{(φ... x ,φ y ,φ z )2,ξ x φ x ,ξ y φ y ,ξ z φ z When this value exceeds the danger threshold, a warning is issued to the unmanned surface vessel indicating that the maneuvering angle is too large.

[0068] This invention proposes a digital twin system for unmanned surface vessel (USV) status monitoring and safety forecasting, which implements forecasting functionality based on a pre-constructed CFD numerical proxy model. The CFD numerical proxy model is built upon a pre-constructed CFD simulation case set, serving as a pre-trained deep learning knowledge base to describe the nonlinear mapping relationship between fluid boundary condition parameters and ship motion response, mechanical response, and other data. In the CFD case set, fluid boundary conditions control the frequency and number of superimposed wavelets in a specified random wave field, equivalent to meaningful wave height, dominant wave direction, and current velocity, quantifying sea state conditions as a discrete combination of wave spectrum and wave height. During actual navigation, the system receives real-time feedback ship motion attitude data sequences from the actual USV via a digital twin interface, performs wave parameter inversion conforming to the mapping relationship, outputs the meaningful wave height, main frequency components, and dominant wave direction, and provides wave force forecasts for the future motion state of the USV. The CFD numerical proxy model provides a scientific and efficient means for analyzing the variable sea conditions encountered by USVs during navigation. Furthermore, based on the current attitude, propulsion status, and predicted wave forces of the USV's twin, a capsizing safety assessment is performed. This invention includes a remote visualization module that enables real-time inversion simulation of the sea conditions surrounding the unmanned surface vessel (USV). It also integrates and displays a capsizing safety assessment of the USV, showing a two-dimensional mathematical isometric plot reflecting the overall attitude of the USV in the corner of the view, clearly revealing quantitative data on the USV's safety assessment to shore-based observers. The visualization backend of the digital twin model provides highly immersive remote safety monitoring capabilities enhanced by physical information fusion.

[0069] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the embodiments described above. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A real-time monitoring and forecasting system for unmanned surface vessels based on a CFD numerical surrogate model, characterized in that, include: Unmanned surface vessel (USV): Equipped with left and right thrusters and an inertial navigation sensor array, with wave height monitoring points on the hull; Digital twin interconnection module: used to receive encoder speed of the left and right thrusters of the unmanned surface vessel (USV), data from the USV's inertial navigation sensor array, and data from the vessel's circumference height monitoring, to realize dynamic information interaction between the USV twin and the actual USV. The CFD numerical proxy model, including a wave parameter inversion and analysis module and a wave force prediction module, is used to construct the nonlinear mapping relationship between the wave navigation mechanics and motion state response characteristics of the unmanned surface vessel and the controlled wave spectrum. It inputs the actual unmanned surface vessel motion response data sequence and the wave height monitoring data around the vessel transmitted by the digital twin interconnection module, analyzes the wave environment characteristic parameters around the unmanned surface vessel and their evolution state in the future Δt range, and further obtains the force state of the unmanned surface vessel. Unmanned surface vessel twin: includes a motion attitude calculation module and an unmanned surface vessel virtual geometric model; used to receive data information from the digital twin interconnection module and the CFD numerical proxy model, drive the unmanned surface vessel virtual geometric model to follow the real motion of the unmanned surface vessel in the ocean wave scene, and display the motion state at the future time Δt in the remote visualization module; The remote visualization module includes a programmable twin ocean interface and a fusion display interface for unmanned surface vessel (USV) capsizing safety assessment. The programmable twin ocean interface is used to visualize the dynamic perspective changes during the USV's wave-following motion. It changes the ocean view based on the USV's attitude and provides visual wave feedback based on wave morphology, simulating the ocean environment surrounding the real USV. The fusion display interface for USV capsizing safety assessment is used to visualize the results of the USV capsizing safety assessment. The wave parameter inversion and analysis module analyzes the six-degree-of-freedom attitude response characteristics and wave height monitoring data of the unmanned surface vessel under unknown wave conditions, and constructs the composition of the real-time wave spectrum encountered by the unmanned surface vessel. The wave parameters obtained by the analysis include the main wave direction and the encountered wave speed, the main wave period, and the significant wave height, i.e., the maximum one-third peak spectrum. The wave parameter inversion and analysis module is a multi-input multi-output deep neural network model with twelve hidden layers. The training dataset of the model is taken from several pre-conducted CFD simulation cases under known superimposed wave conditions. The CFD simulation cases calculate the six-degree-of-freedom attitude response and wave force sequence of the unmanned surface vessel, as well as the wave height sequence at the monitoring points around the unmanned surface vessel. The training process of the wave parameter inversion analytical module is as follows: The roll, pitch, heave, and sway responses from the unmanned surface vessel's attitude response time series, along with the wave height sequences from all four monitoring points around the vessel, are transformed into the frequency domain. The entire frequency band is divided into N sub-segments. The characteristic amplitudes of each sub-segment are normalized and used as input to the wave parameter inversion analysis module. The data depth of its encoding range is M. The total input consists of equivalent grayscale image matrix blocks of size 8×N×M. The superimposed wave parameters from the CFD simulation case are used as the output. The superimposed wave parameters include the main wave direction, encounter wave speed, main wave period, and meaningful wave height, forming a nonlinear input-output dataset. The nonlinear input-output dataset is input into the deep neural network model for training, and the parameters of each hidden layer node are constructed to form a wave parameter inversion analysis module.

2. The unmanned surface vessel real-time monitoring and forecasting system according to claim 1, characterized in that, The capsizing safety assessment of the unmanned surface vessel (USV) utilizes the rate of change of angular velocity to evaluate the attitude safety of the USV during stable navigation. Based on the wave drag and torque experienced by the USV at the current moment, calculated by the wave force prediction module, the rate of change of angular velocity in the future time interval Δt is calculated. The three-axis norm threshold m is defined by combining the magnitude of the three-axis inertia. The system monitors whether the value of m exceeds the danger threshold and displays the attitude and angular velocity change rate of the unmanned surface vessel at the future time Δt on the interface.

3. The unmanned surface vessel real-time monitoring and forecasting system according to claim 1, characterized in that, The unmanned surface vessel inertial navigation sensor array is specifically: A pair of identical inertial navigation sensors are arranged at equal intervals from the center of mass on a horizontal plane passing through the center of mass; A cluster of identical inertial navigation sensors are arranged at equal intervals from the center of mass on a longitudinal section passing through the center of mass.

4. The unmanned surface vessel real-time monitoring and forecasting system according to claim 1, characterized in that, The CFD simulation case uses an orthogonal parameter table to design its boundary conditions. The parameter table involves the equivalent encounter velocity and meaningful wave height in the heading direction. The wave frequency combination is selected according to the wave spectrum with a set step size. The wave frequency combination includes the main direction of incoming waves and the main period of waves.

5. The unmanned surface vessel real-time monitoring and forecasting system according to claim 1, characterized in that, The wave force prediction module, based on the characteristic wave field information of the current unmanned surface vessel, can predict the bidirectional wave drift force on the current vessel under known wave characteristic parameters and the characteristic state of the vessel in the waves, thus avoiding the need for lengthy CFD calculations.

6. The unmanned surface vessel real-time monitoring and forecasting system according to claim 5, characterized in that, The wave force prediction module is specifically a multi-input multi-output residual neural network; During training, the input parameters are the attitude response data within the historical window in the CFD simulation case dataset, and the initial wave parameter matrix is ​​used for convolution processing. The output parameters are the bidirectional wave drift forces experienced by the vessel at each time point within the historical window.

7. The unmanned surface vessel real-time monitoring and forecasting system according to claim 1, characterized in that, The programmable twin ocean interface is specifically described using the PM wave spectrum.

8. The unmanned surface vessel real-time monitoring and forecasting system according to any one of claims 1-7, characterized in that, The assessment of the capsizing safety of the unmanned surface vessel during navigation is based on the following formula: Using the calculated total wave drag torque, the thruster thrust, and the measured hull angular velocity at this moment, the auxiliary parameters of angular velocity are solved. That is, the ratio of the angular velocity of the hull around each axis at two adjacent moments. The auxiliary parameters of the angular velocity should not exceed the safety threshold. The attitude of the unmanned surface vessel is displayed by fusion of orthographic projection on the digital sea surface. The initial coordinates of each point of the hull in the isometric view and the current attitude coordinates satisfy the following relationship, providing the operator with a convenient observation view to understand the attitude status of the hull. Because the magnitudes of the three-axis inertia are different, a three-axis norm threshold metric is defined. To comprehensively assess the capsizing safety of the unmanned surface vessel across all axes, When this value exceeds the danger threshold, a warning is issued to the unmanned surface vessel indicating that the maneuvering angle is too large.

9. The unmanned surface vessel real-time monitoring and forecasting system according to claim 4, characterized in that, The boundary conditions are specifically defined by dividing the interval according to the equivalent encounter velocity of the unmanned surface vessel (USV) in the bow direction, and then based on the expected meaningful wave height range of 1 / 16h. m ~h m According to Δh=1 / 16h m The wave height level intervals are divided according to the Fourier decomposition principle. Each level wave train is represented by the superposition of several wavelets. Under each equivalent encounter velocity, the six wavelets of four different frequency wave height combinations are superimposed in the frequency domain. Each wavelet corresponds to a main wave direction and directional distribution function spectrum.

Images