Ship docking hierarchical collision avoidance warning system based on physical-svr model

Abstract

The present application relates to the ship technology field, especially based on physical-SVR model's ship is docked grading collision avoidance early warning system, including the ship static parameter input, real-time state acquisition, input feature structure, physical model preliminary stress calculation, trained SVR model correction stress calculation, future trajectory prediction, ship-dock minimum distance calculation, grading early warning these module settings, through these module settings can play to improve the ship in the process of docking longitudinal force, lateral force and turning bow moment and other motion influencing factors prediction accuracy; improve the ship in the shallow water effect, bank wall effect and jamming effect under the influence of future trajectory prediction reliability; establish the quantization judgment basis of the safety distance between the ship body contour and the dock wall, dock bottom and dock end; form the grading collision avoidance early warning mechanism combining early warning prompt and emergency alarm, thereby providing technical support for the risk identification, advance adjustment and safety decision of the ship docking site operator.

Description

Technical Field

[0001] This invention relates to the field of marine technology, and in particular to a graded collision avoidance early warning system for ships entering dry dock based on a physical-SVR model. Background Technology

[0002] Existing collision avoidance and early warning methods for ships entering dry dock typically rely on the current ship position or human experience. This makes it difficult to identify potential collision risks between the ship and the dock walls or dock ends based on future trajectories, leading to delayed warnings, increased accident risk, and reduced dry dock production efficiency. This is primarily due to the extremely complex hydrodynamics during ship entry into dry dock, the poor accuracy of traditional physical models, and the insufficient generalization ability of purely data-driven models. Ship entry into dry dock is a low-speed, confined-water maneuvering process. The safety margin between the ship and the dock walls, dock bottom, and dock ends is small. Ship motion is easily affected by shallow water effects, bank effects, and blockage effects, resulting in complex nonlinear changes in longitudinal forces, lateral forces, and turning moments, making motion prediction difficult. While traditional physical models such as MMG have good physical interpretability, they mainly rely on empirical formulas and cannot accurately describe the complex hydrodynamic effects in highly confined waters such as dry docks. Purely data-driven models, while possessing nonlinear fitting capabilities, lack explicit physical constraints and have insufficient generalization ability for changes in ship type, draft, speed, bow angle, and lateral drift.

[0003] Based on the shortcomings of the existing technology, the technical problem to be solved by the present invention is: how to improve the accuracy of ship motion state prediction in the scenario of low-speed ship operation in restricted waters, and to predict the collision risk before the ship rubs or collides with the dock wall, dock bottom or dock end. Summary of the Invention

[0004] The purpose of this invention is to provide a graded collision avoidance early warning system for ship docking based on a physical-SVR model, so as to solve the problems existing in the prior art.

[0005] The above-mentioned technical objective of the present invention is achieved through the following technical solution: A graded collision avoidance early warning system for ships entering dry dock based on a physics-SVR model includes: The static parameters input is used to input the geometric parameters and safety threshold parameters of the ship and the dock before the operation begins. Real-time status data is collected by RTK equipment installed inside the ship. Input feature construction is used for data preprocessing and dimensionless transformation, performing dimensionless processing on the input and collected data; The physical model performs preliminary force calculations to calculate preliminary predicted values ​​of the ship's longitudinal force, lateral force, and bow turning moment based on the current ship status and dock boundary conditions. The trained SVR model corrects the force calculations and is used to revise the initial prediction results of the physical model; Future trajectory prediction is used to calculate the resultant force of the hull after the system obtains the hull correction force; the initial force of the physical model is added to the SVR error correction term to obtain the final corrected force used for ship motion prediction. The minimum distance calculation for dock is used to calculate the minimum distance between the ship's outer contour and the dock wall and dock end based on the ship's position and heading angle at future times. The tiered early warning system compares the minimum distance within the prediction time window with a safe distance threshold to assess potential future risks; it also compares the current distance with a safe distance threshold to assess current urgent risks.

[0006] In a further embodiment, the ship's static parameters and hull geometric parameters include length L, beam B, draft d, and displacement volume. The input parameters include: block coefficient Cb, ship mass m, moment of inertia about the vertical axis Iz, water density ρ, and dock entry reference velocity U; the input dock geometry parameters include the port dock wall position Y. L The location of the right dock wall is Y. R The dock location is X. E The water depth is h; the input safety threshold parameters include a lateral safety distance threshold of D. B,safe The longitudinal safety distance threshold is D E,safe ; where D B,safe Used to determine whether the ship's hull is close to the port and starboard dock walls, D E,safe Used to determine if the ship is close to the dock.

[0007] In a further embodiment, the RTK device is installed on the ship, and the real-time status collected by the RTK device includes the coordinates x0(t) and y0(t) of the ship's reference point in the fixed coordinate system of the dock, and the heading angle. (t), and the longitudinal velocity of the ship in the body coordinate system. (t), lateral velocity (t) and bow roll angular velocity (t); The X0 axis of the fixed coordinate system is along the direction of the ship entering the dock, and the Y0 axis is along the transverse direction of the dock; The X-axis of the body coordinate system points to the bow, and the Y-axis points to the starboard side of the ship.

[0008] In a further embodiment, the input feature construction module performs dimensionless processing on the ship's static parameters and real-time status acquisition data, dimensionlessly transforming the ship's beam, draft, and displacement volume into B', d', and d', respectively. The ship's longitudinal position, lateral position, longitudinal velocity, lateral velocity, and bow roll rate are dimensionlessly transformed into x', y', u', v', and r', respectively, forming the input feature vector for the current moment. .

[0009] In a further embodiment, the physical model preliminary force calculation module calculates the dimensionless longitudinal force X′ obtained from the physical model. phy Horizontal force Y′ phy and the turning torque N′ phy It is expressed as the sum of three parts: hull viscous hydrodynamics, external traction, and dock wall effect; among which, hull viscous hydrodynamics is calculated based on the derivative of ship maneuvering hydrodynamics, and dock wall effect is calculated based on the distance between the hull and the port and starboard dock walls, water depth h, ship width B, ship length L, and the current motion state as additional lateral force and turning moment.

[0010] In a further embodiment, the trained SVR model correction force calculation module calls three pre-trained offline SVR models to predict the longitudinal force correction term δX′. SVR (t), lateral force correction term δY′ SVR (t) and the turning moment correction term δN′ SVR (t); The SVR model is trained with the current input feature x(t) as input and the error between the preliminary prediction result of the physical model and the calculation result of CFD as output. After training, it is deployed to the trained SVR model correction force calculation module.

[0011] In a further embodiment, the future trajectory prediction module adds the initial forces of the physical model to the SVR error correction term to obtain the corrected dimensionless longitudinal force X'(t), lateral force Y'(t), and bow moment N'(t). Based on the corrected resultant forces, the ship's three-degree-of-freedom MMG equation of motion is adopted and solved using the fourth-order Runge-Kutta method to obtain the ship's position coordinates and heading angle at each future moment.

[0012] In a further embodiment, the dock minimum distance calculation module approximates the hull as a rectangular envelope in the horizontal plane, calculates the positions of four envelope feature points of the hull based on the ship's position coordinates and bow angle, and then calculates the shortest distance between the hull and the port and starboard dock walls and dock ends based on the positions of the four envelope feature points. and Furthermore, the minimum lateral distance within the prediction time window Tp is calculated. and minimum longitudinal distance .

[0013] In a further embodiment, the tiered early warning module predicts the minimum lateral distance within the prediction time window. Less than the lateral safety distance threshold D B,safe Or the minimum vertical distance within the prediction time window Less than the longitudinal safety distance threshold D E,safeWhen the current lateral distance is..., output a warning prompt; when the current lateral distance... Less than the lateral safety distance threshold D B,safe Or current vertical distance Less than the longitudinal safety distance threshold D E,safe When this happens, an emergency alarm will be output.

[0014] In a further embodiment, a warning information output module is also included, used to output the current hull outline, the predicted future trajectory, and the current distance between the hull and the port and starboard dock walls. Current distance between the hull and the dock Minimum lateral distance within the prediction time window Minimum vertical distance within the prediction time window Risk direction and risk level; the risk direction includes near the left dock wall, near the right dock wall and near the dock end; the risk level includes normal, early warning and emergency alarm.

[0015] In summary, the present invention has the following beneficial effects: 1. By setting up modules such as ship static parameter input, real-time status acquisition, input feature construction, preliminary force calculation of physical model, force calculation correction of trained SVR model, future trajectory prediction, ship-dock minimum distance calculation, and graded early warning, the system can improve the accuracy of predicting motion influencing factors such as longitudinal force, lateral force, and bow turning moment during ship docking; improve the reliability of future trajectory prediction under the influence of shallow water effect, quay wall effect, and blockage effect; establish a quantitative judgment basis for the safe distance between the ship's outline and the dock wall, dock bottom, and dock end; and form a graded collision avoidance early warning mechanism that combines early warning prompts and emergency alarms, thereby providing technical support for on-site docking personnel to identify risks, make advance adjustments, and make safety decisions.

[0016] 2. Compared with existing schemes that extrapolate ship trajectories based on GNSS historical position sequences, this invention does not only extrapolate future trajectories based on changes in the ship's position over a period of time. Instead, it first calculates the forces acting on the ship based on parameters such as the ship's main dimensions, draft, dock entry speed, bow angle, lateral offset, and dock wall boundaries. Then, it predicts the future motion state through the ship's three-degree-of-freedom motion equations. Therefore, this invention can incorporate the influence of longitudinal forces, lateral forces, and bow turning moments on the ship's subsequent motion during dock entry into the prediction process, making it more suitable for scenarios where the hydrodynamic effects are significant in the confined waters after the ship enters the dock.

[0017] 3. This invention adopts a composite modeling approach of "preliminary prediction by physical model + SVR error correction". The physical model can reflect the basic laws of ship maneuvering motion and hydrodynamic action, while the SVR model can learn the error between the physical model and the high-precision calculation results by using an offline CFD correction force database. The combination of the two avoids the problem of insufficient prediction accuracy of a purely physical model in highly restricted waters, and also avoids the problem of a purely data-driven model lacking physical basis and being overly dependent on samples. This helps to improve the reliability of ship docking motion prediction.

[0018] 4. This invention not only predicts the future trajectory of the ship's reference point or a few positioning points, but also generates a rectangular envelope of the ship's horizontal plane based on the predicted ship position and heading angle, and calculates the minimum distance between the ship's outer contour and the port wall, starboard wall and dock end. Therefore, the risk assessment object of this invention is not an abstract trajectory line, but the safety margin between the actual outer contour of the ship's hull and the dock boundary, which is more in line with the actual needs of ship docking collision avoidance early warning.

[0019] 5. This invention establishes an early warning judgment mechanism based on a predicted time window. At the current moment, the system not only calculates the current distance between the hull and the dock wall and dock end, but also predicts the minimum lateral distance and minimum longitudinal distance within a certain time window in the future. When the predicted minimum distance is lower than the safety threshold, the system can output an early warning prompt in advance. Therefore, compared with the method of alarming only based on the current ship position or current distance, this invention can identify potential risks before the ship actually enters the dangerous distance, providing advance time for on-site personnel to take measures such as deceleration, correction, and suspension of towing.

[0020] 6. This invention further distinguishes between "early warning" and "emergency alarm". When the predicted minimum distance is lower than the safety threshold, the system outputs an early warning; when the current distance is already lower than the safety threshold, the system outputs an emergency alarm. Through this hierarchical early warning method, on-site personnel can not only know whether there is a risk, but also judge the urgency of the risk, so as to arrange the operation and handling more reasonably.

[0021] 7. The input parameters of this invention include ship main dimensional parameters, docking motion state parameters, and dock boundary parameters, enabling calculations for different ship types, drafts, docking speeds, bow angles, lateral offsets, and dock conditions. Compared to extrapolation methods that rely solely on GNSS historical trajectories and empirical correction coefficients, this invention more fully utilizes parameters directly related to docking hydrodynamics and collision risks, making it more suitable for refined safety early warning in restricted docking waters.

[0022] In summary, this invention transforms the ship docking motion prediction results into risk warning information that can directly serve on-site operations through physical-SVR composite force prediction, three-degree-of-freedom motion trajectory prediction, minimum distance calculation of the ship's outer contour, and graded collision avoidance early warning judgment. This can improve the risk identification and decision support capabilities during the ship docking process. Attached Figure Description

[0023] Figure 1 This is a line graph showing the corrected force calculation results for working condition 1 of this invention; Figure 2 This is a line graph showing the corrected force calculation results for working condition 2 of this invention; Figure 3 This is a line graph showing the corrected force calculation results for working condition 3 of this invention; Figure 4 This is a line graph showing the corrected force calculation results for working condition 4 of this invention; Figure 5 This is a line graph showing the corrected force calculation results for working condition 5 of this invention; Figure 6 This is a line graph showing the corrected force calculation results for working condition 6 of this invention; Figure 7 This is a line graph showing the corrected force calculation results for working condition 7 of this invention; Figure 8 This is a line graph showing the corrected force calculation results for working condition 8 of this invention; Figure 9 This is a line graph showing the corrected force calculation results for working condition 9 of this invention; Figure 10 This is a line graph showing the corrected force calculation results for working condition 10 of the present invention; Figure 11 This is a line graph showing the corrected force calculation results for working condition 11 of the present invention; Figure 12 This is a line graph showing the corrected force calculation results for working condition 12 of the present invention; Figure 13 This is a line graph showing the corrected force calculation results for working condition 13 of the present invention; Figure 14 This is a line graph showing the corrected force calculation results for working condition 14 of the present invention; Figure 15 This is a line graph showing the corrected force calculation results for working condition 15 of the present invention; Figure 16 This is a line graph showing the corrected force calculation results for working condition 16 of the present invention; Figure 17 This is a line graph showing the corrected force calculation results for working condition 17 of the present invention; Figure 18 This is a line graph showing the corrected force calculation results for working condition 18 of the present invention; Figure 19This is a line graph showing the corrected force calculation results for working condition 19 of the present invention; Figure 20 This is a line graph showing the corrected force calculation results for working condition 20 of the present invention; Figure 21 This is a schematic diagram showing the orientation and position of the ship and the dock according to the present invention. Detailed Implementation

[0024] The present invention will be further described in detail below with reference to the accompanying drawings.

[0025] It should be noted that in the description of this invention, any descriptions of orientation, such as up, down, front, back, left, right, etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings. They are only for the purpose of facilitating the description of this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed or operated in a specific orientation, and should not be construed as a limitation of this invention.

[0026] Example 1: like Figures 1-20 As shown, the ship docking graded collision avoidance early warning system based on the physical-SVR model allows the ship static parameter input module to input the geometric parameters of the ship and dock, as well as safety threshold parameters, before operation. The ship static parameters and hull geometric parameters include the ship's length L, beam B, draft d, and displacement volume. The input parameters include: block coefficient Cb, ship mass m, moment of inertia about the vertical axis Iz, water density ρ, and dock entry reference velocity U; the input dock geometry parameters include the port dock wall position Y. L The right dock wall is located at Y. R The dock location is X. E The water depth is h; the input safety threshold parameters include a lateral safety distance threshold of D. B,safe The longitudinal safety distance threshold is D E,safe ; where D B,safe Used to determine whether the ship's hull is close to the port and starboard dock walls, D E,safe Used to determine if the ship is close to the dock.

[0027] The real-time status acquisition module, consisting of RTK equipment installed on the ship, collects data and transmits it to subsequent modules for further calculation of motion prediction and collision avoidance warning information. The real-time status (time t) acquired by the RTK equipment includes the ship's reference point's coordinates x0(t) and y0(t) in the dock's fixed coordinate system, as well as the bow angle. (t), and the longitudinal velocity of the ship in the body coordinate system. (t), lateral velocity (t) and bow roll angular velocity (t); the fixed coordinate system has its X0 axis along the direction of the ship entering the dock and its Y0 axis along the lateral direction of the dock; the body coordinate system has its X-axis pointing towards the bow and its Y-axis pointing towards the starboard side of the ship, with specific orientations as follows: Figure 21 As shown.

[0028] The position change in the fixed coordinate system and the velocity in the body coordinate system satisfy the following relationship: ; ; ; The above relationship is used to convert the ship's own motion state into position changes in the dock coordinate system.

[0029] The input feature construction module first preprocesses and dimensionlessly transforms the data. The above modules then input and collect data, performing dimensionless processing. The ship's beam, draft, and displacement volume are dimensionlessly transformed as follows: , , ; The superscript ' indicates a dimensionless quantity. The current ship position and motion state are also dimensionless, where the ship's longitudinal position, lateral position, longitudinal velocity, lateral velocity, and bow roll rate are respectively represented as: , , ; After the above processing, the system generates the input feature vector for the current moment. The input feature vector can be represented as: ; The preliminary force calculation module of the physical model is used to calculate the preliminary predicted values ​​of the ship's longitudinal force, lateral force, and turning moment based on the current ship state and dock boundary conditions. Let the dimensionless longitudinal force, lateral force, and turning moment obtained from the physical model be X′. phy 、Y′ phy and N′ phy Therefore, this invention represents it as the sum of three parts: hull viscous hydrodynamics, external traction, and dock wall effect: ; ; ; In this context, subscript H indicates viscous hydrodynamics of the hull, subscript T indicates external traction such as tugboats, towing trolleys, or cables, and subscript S indicates dock wall or shore wall effect. Longitudinal force... lateral force and turning torque Dimensionlessness is defined as: , , ; Under conditions of low-speed docking and small drift angle, the viscous hydrodynamics of the hull can be calculated based on the derivative of the ship's maneuvering hydrodynamics. Let R′0 be the straight-line resistance coefficient, and X′... vv 、X′ vr 、X′ rr Y′ is the longitudinal hydrodynamic derivative. v 、Y′ r 、Y′ vv 、Y′ vr 、Y′ rr For the transverse hydrodynamic derivative, N′ v 、N′ r 、N′ vvr 、N′ vrr For the hydrodynamic derivative of the bow turn, the viscous hydrodynamics of the hull can be expressed as: ; ; ; Straight-line drag coefficient With total drag coefficient The relationship is: ; S represents the wetted area of ​​the hull. The corresponding dimensionless parameters can be estimated using the following formula: ; Total drag coefficient According to the ship resistance law, we can obtain: ; Among them, C f C is the coefficient of frictional resistance. r ΔC is the residual drag coefficient. AR This represents the roughness compensation coefficient. Based on the existing 1957-ITTC formula, the Lambeau-Offenkel formula, and approximate empirical methods, the above formulas are expressed as follows: , , ; Where: Reynolds number , is the fluid kinematic viscosity coefficient.

[0030] Longitudinal hydrodynamic derivative , , It can be calculated using the empirical formula provided by Matsumoto: ; ; In the formula: , , The drafts are for the bow and stern respectively (dimensionless); ;in, satisfy: ; The transverse hydrodynamic derivative includes both linear and nonlinear components. The linear transverse hydrodynamic derivative can be calculated using approximate formulas found in wells: ; .

[0031] The nonlinear hydrodynamic derivative can be calculated using Zhou Zhaoming's regression formula for wellbore patterns: ; ; ; In the formula: For dimensionless draft, ; For aspect ratio, .

[0032] The linear bow-turning hydrodynamic derivative is calculated using an approximate formula from the well site: ; ; The nonlinear hydrodynamic derivative can be calculated using Zhou Zhaoming's regression formula for wellbore patterns: ; ; In the formula: .

[0033] The dock wall effect calculation unit calculates the additional lateral force and turning moment caused by the dock wall effect, i.e., Y′, based on the distance between the hull and the port and starboard dock walls, the water depth h, the ship's beam B, the ship's length L, and the current motion state. S and N′ S If the influence of the dock walls on the longitudinal resistance is considered, then X′ should also be calculated. S If we ignore the effect of the dock walls on the longitudinal force, then let X′ S =0: ; ; ; in, This is the distance from the ship to the shore.

[0034] The external traction calculation unit is used to handle the effects of tugboats, towing trolleys, or cables on the vessel. If the traction force or cable force can be obtained on-site, it is decomposed into longitudinal, transverse, and bow directions of the hull to obtain X′. T 、Y′ T and N′ T If the traction force cannot be directly measured on site, the docking speed U or the traction control command can be used as the equivalent input.

[0035] A trained SVR force correction calculation module can correct the initial prediction results of the physical model, obtaining more accurate hull force prediction results. This module calls three pre-trained offline SVR models to predict longitudinal force correction terms, lateral force correction terms, and bow moment correction terms, respectively. The system inputs the current input feature x(t) into the three SVR models, obtaining: ; ; ; Among them, f X f Y f N These are the longitudinal force correction model, lateral force correction model, and yaw moment correction model obtained through offline training, respectively; δX′ SVR (t), δY′ SVR (t), δN′ SVR (t) represents the error correction terms for the longitudinal force, lateral force, and steering torque of the physical model under the current operating conditions. The offline training method for this SVR error correction module will be introduced later.

[0036] The future trajectory prediction module calculates the resultant force of the hull after the system obtains the hull correction forces. The initial forces in the physical model are added to the SVR error correction term to obtain the final corrected forces used for ship motion prediction. Specifically, the system obtains the corrected resultant force according to the following formula, including dimensionless longitudinal force, lateral force, and turning moment: X'(t)=X' phy (t)+δX′ SVR (t); Y'(t)=Y' phy (t)+δY′ SVR (t); N'(t)=N' phy (t)+δN′ SVR (t); Where X′(t), Y′(t), and N′(t) are the dimensionless longitudinal force, lateral force, and turning torque after correction at the current moment, respectively.

[0037] Based on the calculated resultant forces X′(t), Y′(t), and N′(t) of the hull, the hull motion can be calculated. During the ship's docking process, the longitudinal motion, lateral motion, and bow-turning motion of the ship are mainly considered. Based on the MMG equations, the ship's motion in these three directions can be expressed in a dimensionless form. Let m′ be the dimensionless mass of the ship, and I′... z Let m′ be the dimensionless moment of inertia of the ship about its vertical axis. x and m′ y J′ represents the dimensionless mass added longitudinally and laterally, respectively. zAs an additional dimensionless moment of inertia, the equations of motion for the ship's three degrees of freedom can be written as: ; ; ; Wherein: additional mass and additional moment of inertia , , The method used by Zhou Zhaoming, which employs multiple regression on the three spectra of Yuan Liang, can be expressed as follows:

[0038]

[0039] ; Based on the above three-degree-of-freedom equations of motion for the ship, the fourth-order Runge-Kutta method can be used to obtain the ship's trajectory. .

[0040] The module for calculating the minimum distance between the ship and the dock is used to calculate the minimum distance between the hull's outer contour and the dock walls and dock ends based on the ship's position and heading angle at future times. This paper approximates the hull as a rectangular envelope in the horizontal plane to calculate the distance between the hull boundary and the dock walls and dock ends. The four envelope feature points of the hull can be represented as: ( =1,2,3,4); ; In the above formula, L is the length of the ship, and B is the width of the ship. For the heading angle, ( , () represents the ship position coordinates (dimensionless) predicted by the physical-SVR motion prediction model mentioned above.

[0041] Based on the positions of the four hull envelope points at a given time ( Given the array (e.g., 1, 2, 3, 4), calculate the shortest distances between the ship and the port and starboard walls and dock ends, which are called the shortest distances between the ship and the dock walls and dock ends. At time t, the shortest distances between the ship and the dock walls and dock ends are expressed as follows: , Therefore, the nearest distance within the prediction time window Tp (e.g., Tp = 120s) is defined as follows: ; The tiered early warning module compares the minimum distance within the prediction time window with the safe distance threshold to determine potential future risks; it also compares the current distance with the safe distance threshold to determine current urgent risks.

[0042] When the following conditions are met: < or < When this occurs, the system outputs a warning. This warning indicates that the vessel may not yet be approaching the dock wall or dock end, but based on future trajectory predictions, the vessel will enter a danger zone within the predicted time window. On-site operators need to take preemptive measures such as slowing down, correcting course, adjusting traction direction, or suspending traction. When the following conditions are met: < or < At this time, the system outputs an emergency alarm. This emergency alarm indicates that the vessel is approaching the dock wall or dock end and immediate safety measures should be taken. Specifically: and These represent the lateral safety distance threshold and the longitudinal safety distance threshold, respectively, and their values ​​are entered in the initial ship static parameter input module.

[0043] When both the predicted minimum distance and the current distance are not less than the corresponding safe distance threshold, the system outputs a normal state, i.e.: At that time, the system determined that the ship was in a normal docking state.

[0044] The early warning information output module displays the system's judgment results to on-site operators. The output includes the current hull outline, the predicted future trajectory, and the current distance D between the hull and the port and starboard dock walls. B (t), Current distance D between the hull and the dock E (t), Minimum lateral distance D within the prediction time window B,pre Minimum vertical distance D within the prediction time window E,pre Risk direction and risk level.

[0045] The risk directions include proximity to the left dock wall, proximity to the right dock wall, and proximity to the dock end; the risk levels include normal, early warning, and emergency alarm. The system can output the above information through a host computer interface, a dockside monitoring screen, an audible and visual alarm device, or a mobile terminal. When the system is in a normal state, it can be displayed in green; when an early warning is triggered, it can be displayed in yellow or orange; when an emergency alarm is triggered, it can be displayed in red and trigger an audible and visual alarm.

[0046] At this point, the online early warning process has completed one calculation. If the ship's docking process is not yet complete, the system returns to the ship status acquisition module to enter the next sampling cycle and repeat the above online early warning process.

[0047] The SVR error correction model used in the aforementioned online early warning process was not trained in real time at the ship docking site, but was obtained through offline training before the system was used.

[0048] First, determine the typical ship type and typical docking conditions. The typical ship type is determined by the ship's length L, beam B, draft d, displacement volume ∇, and block coefficient Cb; the typical docking conditions are determined by the docking speed U, bow angle, and other parameters. The lateral offset y0, dock entry position x0, water depth h, and distance between the hull and the dock wall are determined. For each typical working condition, the system constructs input features χi in a dimensionless manner according to the online process, where i represents the i-th training sample.

[0049] Then, CFD calculations were performed on the typical ship type and typical working condition to obtain the longitudinal force X of the hull under this working condition. CFD,i lateral force Y CFD,i and turning torque N CFD,i The above CFD calculation results are dimensionless according to the following formula: ; ; ; Subsequently, under the same ship type, same docking status, and same location conditions, the preliminary force calculation module of the physical model was invoked to obtain the preliminary calculation result X of the physical model. ′phy,i 、Y′ phy,i and N′ phy,i .

[0050] Next, the CFD calculation result is subtracted from the preliminary calculation result of the physical model to obtain the correction force of the i-th training sample: ; ; ; Therefore, the i-th training sample is represented as: Multiple training samples form the CFD correction force database.

[0051] Finally, the system uses x respectively i As input, with δX′ i ,δY′ i and δN′ i For output, the fitsvm toolbox in MATLAB is used to train three SVR models. After training, three correction functions f are obtained. X f Y and f N It was then deployed into the SVR error correction module of the online early warning system.

[0052] This patent, based on a dock and its typical docking conditions, uses StarCCM+ fluid simulation software to construct a corrected force database. In the training conditions, the actual dock dimensions are 280m × 40m × 13.6m, encompassing both KCS bulk carriers and KVLCC2 oil tankers, with a scale ratio of 1:80. The training conditions include 20 sets of conditions, and the relevant input parameters for these 20 sets are shown in Table 1 below: ; In the description of this invention, unless otherwise explicitly defined, terms such as "set up," "install," and "connect" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this invention in conjunction with the specific content of the technical solution.

[0053] This specific embodiment is merely an explanation of the present invention and is not intended to limit the invention. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they are within the scope of the claims of the present invention.

Claims

1. A graded collision avoidance early warning system for ship docking based on a physical-SVR model, characterized in that, include: The static parameters input is used to input the geometric parameters and safety threshold parameters of the ship and the dock before the operation begins. Real-time status data is collected by RTK equipment installed inside the ship. Input feature construction is used for data preprocessing and dimensionless transformation, performing dimensionless processing on the input and collected data; The physical model performs preliminary force calculations to calculate preliminary predicted values ​​of the ship's longitudinal force, lateral force, and bow turning moment based on the current ship status and dock boundary conditions. The trained SVR model corrects the force calculations and is used to revise the initial prediction results of the physical model; Future trajectory prediction is used to calculate the resultant force of the hull after the system obtains the hull correction force; The initial force of the physical model is added to the SVR error correction term to obtain the final corrected force used for ship motion prediction; The minimum distance calculation for dock is used to calculate the minimum distance between the ship's outer contour and the dock wall and dock end based on the ship's position and heading angle at future times. The tiered early warning system compares the minimum distance within the prediction time window with a safe distance threshold to assess potential future risks; it also compares the current distance with a safe distance threshold to assess current urgent risks.

2. The ship docking graded collision avoidance early warning system based on the physical-SVR model according to claim 1, characterized in that: The ship's static parameters and hull geometric parameters include length L, beam B, draft d, and displacement volume. The input parameters include: block coefficient Cb, ship mass m, moment of inertia about the vertical axis Iz, water density ρ, and dock entry reference velocity U; the input dock geometry parameters include the port dock wall position Y. L The location of the right dock wall is Y. R The dock location is X. E The water depth is h; the input safety threshold parameters include a lateral safety distance threshold of D. B,safe The longitudinal safety distance threshold is D E,safe ; Among them, D B,safe Used to determine whether the ship's hull is close to the port and starboard dock walls, D E,safe Used to determine if the ship is close to the dock.

3. The ship docking graded collision avoidance early warning system based on the physical-SVR model according to claim 1, characterized in that: The RTK equipment is installed on the ship, and the real-time status data collected by the RTK equipment includes the coordinates x0(t) and y0(t) of the ship's reference point in the fixed coordinate system of the dock, and the heading angle. (t), and the longitudinal velocity of the ship in the body coordinate system. (t), lateral velocity (t) and bow roll angular velocity (t); The X0 axis of the fixed coordinate system is along the direction of the ship entering the dock, and the Y0 axis is along the transverse direction of the dock; The X-axis of the body coordinate system points to the bow, and the Y-axis points to the starboard side of the ship.

4. The ship docking graded collision avoidance early warning system based on the physical-SVR model according to claim 1, characterized in that: The input feature construction module performs dimensionless processing on the ship's static parameters and real-time status acquisition data, dimensionlessly transforming the ship's beam, draft, and displacement volume into B', d', and d', respectively. The ship's longitudinal position, lateral position, longitudinal velocity, lateral velocity, and bow roll rate are dimensionlessly transformed into x', y', u', v', and r', respectively, forming the input feature vector for the current moment. .

5. The ship docking graded collision avoidance early warning system based on the physical-SVR model according to claim 1, characterized in that: The preliminary force calculation module of the physical model will calculate the dimensionless longitudinal force X′ obtained from the physical model. phy Horizontal force Y′ phy and the turning torque N′ phy It is expressed as the sum of three parts: hull viscous hydrodynamics, external traction, and dock wall effect; among which, hull viscous hydrodynamics is calculated based on the derivative of ship maneuvering hydrodynamics, and dock wall effect is calculated based on the distance between the hull and the port and starboard dock walls, water depth h, ship width B, ship length L, and the current motion state as additional lateral force and turning moment.

6. The ship docking graded collision avoidance early warning system based on the physical-SVR model according to claim 1, characterized in that: The trained SVR model correction force calculation module calls three pre-trained offline SVR models to predict the longitudinal force correction term δX′. SVR (t), lateral force correction term δY′ SVR (t) and the turning moment correction term δN′ SVR (t); The SVR model is trained by taking the current input feature x(t) as input and the error between the preliminary prediction result of the physical model and the calculation result of CFD as output. After training, it is deployed to the force calculation module of the trained SVR model.

7. The ship docking graded collision avoidance early warning system based on the physical-SVR model according to claim 1, characterized in that: The future trajectory prediction module adds the initial forces on the physical model to the SVR error correction term to obtain the corrected dimensionless longitudinal force X'(t), lateral force Y'(t), and bow moment N'(t). Based on the corrected resultant forces, the ship's three-degree-of-freedom MMG equations of motion are used and solved using the fourth-order Runge-Kutta method to obtain the ship's position coordinates and heading angle at each future moment.

8. The ship docking graded collision avoidance early warning system based on the physical-SVR model according to claim 1, characterized in that: The dock minimum distance calculation module approximates the hull as a rectangular envelope in the horizontal plane, calculates the positions of four envelope feature points of the hull based on the ship's position coordinates and bow angle, and then calculates the shortest distance between the hull and the port and starboard dock walls and dock ends based on the positions of the four envelope feature points. and Furthermore, the minimum lateral distance within the prediction time window Tp is calculated. and minimum longitudinal distance .

9. The ship docking graded collision avoidance early warning system based on the physical-SVR model according to claim 1, characterized in that: The tiered early warning module predicts the minimum horizontal distance within the prediction time window. Less than the lateral safety distance threshold D B,safe Or the minimum vertical distance within the prediction time window Less than the longitudinal safety distance threshold D E,safe When the current horizontal distance is... Less than the lateral safety distance threshold D B,safe Or current vertical distance Less than the longitudinal safety distance threshold D E,safe When this happens, an emergency alarm will be output.

10. The ship docking graded collision avoidance early warning system based on the physical-SVR model according to claim 1, characterized in that: It also includes a warning information output module, which outputs the current hull outline, the predicted future trajectory, and the current distance between the hull and the port and starboard dock walls. Current distance between the hull and the dock Minimum lateral distance within the prediction time window Minimum vertical distance within the prediction time window Risk direction and risk level; the risk direction includes near the left dock wall, near the right dock wall and near the dock end; the risk level includes normal, early warning and emergency alarm.

Images