Dynamic hydrodynamic modeling and intelligent parameter identification method for sailing of trailing suction dredger

Abstract

The invention discloses a dynamic hydrodynamic modeling and parameter intelligent identification method for sailing of a trailing suction dredger, and relates to the technical field of sailing of trailing suction dredgers. According to the hydrodynamic modeling and parameter intelligent identification method for the sailing dynamic state of the trailing suction hopper dredger, core data covering an operation manual, historical cases, ship design data and all-working-condition sailing data are systematically obtained, so that a comprehensive basis is provided for modeling; an MMG reference model parameter set is constructed in combination with a hydrostatic calculation book, all working condition data are fused to obtain an environment adaptive fusion model, model deviation is corrected through parameter compensation, real-time data are processed at high frequency and the model is dynamically optimized by means of an edge calculation unit, and a sliding window can be adjusted according to error feedback to accelerate convergence. It is ensured that the final model prediction error is smaller than a threshold value, reliable kinematics support is provided for intelligent operation such as ship autonomous collision avoidance and trajectory tracking, and dredging operation safety risks and energy losses are effectively reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of navigation technology of a trailing suction dredger, in particular to a trailing suction dredger navigation dynamic hydrodynamic force modeling and parameter intelligent identification method. BACKGROUND

[0002] Ship dynamic hydrodynamic force modeling is the bottom cornerstone of realizing the autonomous dredging of a trailing suction dredger. Its essence is to describe the interaction law between the ship and the environmental flow field through a mathematical model, which directly determines the precision, navigation safety and energy utilization efficiency of the dredging operation. The current international mainstream technology is the MMG (Maneuvering Modeling Group) framework. This framework establishes a linear equation through a decomposition method, decomposes the six degrees of freedom motion (surge, sway, heave, roll, pitch, yaw) of the ship into independent components such as static water resistance, propulsion, wave force, etc. for superposition, and can realize certain hydrodynamic force characterization under the condition of basic stability.

[0003] However, in actual dredging operations, the ship often faces complex working conditions such as dynamic changes in mud loading rate (30% to 100%), waves and time-varying flow rate. The existing technical solutions expose the following significant defects: First, the traditional MMG model is insufficient in modeling nonlinear coupling effects. The dynamic change of the mud loading rate in the dredging operation will cause the continuous migration of the ship's center of gravity, causing the coupling effect of multi-degree-of-freedom motion such as roll and yaw. The MMG model is based on the linear superposition assumption, and the static parameters cannot adapt to dynamic working conditions. Second, the real-time disturbance adaptation capability of existing data-driven technologies is weak. Although the SINDy (Sparse Identification) algorithm developed by the University of Cambridge can identify key nonlinear terms in the ship dynamics equation through navigation data, the North Sea real ship test in Norway also shows that it can reduce the track tracking error from ±4.3 meters to ±1.2 meters. However, such algorithms rely heavily on offline training data and are difficult to reflect time-varying environmental disturbances such as waves and flow rates in real time. Finally, the existing technology lacks a systematic solution to the fusion of multiple methods. The traditional MMG model has physical interpretability but weak nonlinear characterization ability, and the data-driven algorithm is good at capturing nonlinear characteristics but relies on offline data and lacks physical constraints, and the two have not formed a deep fusion. At the same time, for special working conditions such as sudden changes in mud tank loading rate, there is a lack of parameter dynamic calibration mechanism combining operating experience and historical cases, the model convergence speed is slow, and it cannot quickly adapt to changes in working conditions, further exacerbating the contradiction between model accuracy and operation safety.

[0004] Therefore, there is an urgent need for a trailing suction dredger navigation dynamic hydrodynamic force modeling and parameter intelligent identification method that can deeply integrate the physical interpretability of the MMG model and the nonlinear identification ability of the data-driven technology, and has real-time disturbance perception and parameter dynamic optimization functions, to solve the above technical bottlenecks and improve the autonomous operation capability and comprehensive performance of the ship in complex dredging environments. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a method for modeling and intelligently identifying parameters of dynamic hydrodynamic force of a trailing suction dredger, which solves the problems of weak model foundation caused by scattered core data acquisition, difficulty in adapting to complex dredging environment and dynamic working conditions due to reliance on a single model, loss of accuracy due to lack of targeted compensation of model parameters, lag in real-time data processing, and difficulty in ensuring prediction accuracy.

[0006] To achieve the above object, the present application is implemented by the following technical scheme: a method for modeling and intelligently identifying parameters of dynamic hydrodynamic force of a trailing suction dredger, comprising the following steps: acquiring core data; acquiring a static hydrodynamic force calculation book from the core data, and obtaining a MMG reference model parameter set based on the static hydrodynamic force calculation book; acquiring full-condition navigation data from the core data, and obtaining an environment-adaptive fusion model in combination with the MMG reference model parameter set; compensating the parameter set of the environment-adaptive fusion model to obtain a corrected parameter set; processing real-time data in the core database based on an edge computing unit, and obtaining a final environment-adaptive fusion model in combination with the corrected parameter set.

[0007] Further, the core data includes a dredger operation manual, a historical accident case library, ship main dimension data, a static hydrodynamic force calculation book, and full-condition navigation data.

[0008] Further, the process of acquiring core data is as follows: deploying perception and computing hardware, and connecting collected real-time data to an edge computing unit through a CAN bus, wherein the edge computing unit pre-stores a dredger operation manual and a historical accident case library; importing ship main dimension data and a static hydrodynamic force calculation book from ship design drawings, and collecting more than 1000 groups of full-condition navigation data, wherein the ship main dimension data includes ship length, type width, and draft, the static hydrodynamic force calculation book includes ship mass, mass matrix, additional mass matrix, static water restoring force coefficient, roll damping coefficient, and pitch damping coefficient, and the full-condition navigation data includes mud tank loading rate, speed, and rudder angle.

[0009] Further, the real-time data includes roll angle, pitch angle, loading rate, and torque value. The process of deploying perception and computing hardware is as follows: installing an attitude sensor at the ship body amidships or bow and stern for acquiring roll angle and pitch angle; installing a loading rate sensor on the top of the mud tank for acquiring loading rate; installing a torque sensor at the rudder or propeller for acquiring torque value.

[0010] Further, the process of obtaining the MMG reference model parameter set based on the static hydrodynamic calculation book is: Obtaining the inertia force coefficient set based on the ship mass indicated in the static hydrodynamic calculation book; Obtaining the viscous force coefficient set based on the ship main dimension data; Recording the inertia force coefficient set and the viscous force coefficient set as the MMG reference model parameter set.

[0011] Further, the process of obtaining the inertia force coefficient set based on the ship mass indicated in the static hydrodynamic calculation book is: Taking the ship mass indicated in the static hydrodynamic calculation book as the reference, setting the initial values of the inertia force coefficients of the longitudinal, lateral and vertical oscillations in the 6-DOF translational directions as m respectively; Taking the rotational inertia moments around the x-axis, y-axis and z-axis indicated in the static hydrodynamic calculation book as the initial values of the inertia force coefficients in the corresponding rotational directions, wherein the x-axis represents the roll, the y-axis represents the pitch and the z-axis represents the yaw; Extracting the added mass matrix from the static hydrodynamic calculation book, the added mass matrix including the longitudinal added mass, the lateral added mass, the vertical added mass and the roll added inertia moment, and superimposing the corresponding components of the added mass matrix to the set initial inertia force coefficients respectively: The longitudinal inertia force coefficient is corrected as the sum of the ship mass and the longitudinal added mass, the lateral inertia force coefficient is corrected as the sum of the ship mass and the lateral added mass, the vertical inertia force coefficient is corrected as the sum of the ship mass and the vertical added mass, and the roll inertia force coefficient is corrected as the sum of the x-axis rotational inertia moment and the roll added inertia moment, forming the inertia force coefficient set.

[0012] Further, the process of obtaining the viscous force coefficient set based on the ship main dimension data is: Recording the product of the ship length and the draft as the longitudinal characteristic area; Recording the product of the ship width and the draft as the lateral characteristic area; Reading the static hydrodynamic resistance coefficients from the static hydrodynamic calculation book, and inversely calculating the initial values of the longitudinal direction viscous force coefficient and the lateral direction viscous force coefficient respectively according to the fluid viscous resistance formula; Taking the roll damping coefficient and the pitch damping coefficient from the static hydrodynamic calculation book as the initial values of the viscous force coefficients in the corresponding rotational directions; Recording the initial values of the longitudinal direction viscous force coefficient, the lateral direction viscous force coefficient, the roll rotational direction viscous force coefficient and the pitch rotational direction viscous force coefficient as the viscous force coefficient set.

[0013] Further, the process of obtaining the MMG reference model parameter set based on the static hydrodynamic calculation book is: The SINDy algorithm is used for processing full-condition navigation data to obtain a candidate function library, and significant nonlinear terms are screened through L1 regularization sparse regression to obtain a set of nonlinear correction factors. The set of nonlinear correction factors is coupled with a set of MMG reference parameters to construct a fused 6-DOF motion equation, denoted as a fusion model. The wave height data collected by the attitude sensor is calculated using the JONSWAP wave spectrum to obtain wave excitation force, and the damping matrix of the fusion model is corrected to obtain an environment-adaptive fusion model.

[0014] Further, the process of compensating the parameter set of the environment-adaptive fusion model to obtain the modified parameter set is as follows: The trigger condition-parameter compensation rule chain is constructed based on the dredger operation manual and the historical accident case library to form a rule library. The edge computing unit collects the current state in real time, matches with the rule library, compensates the parameter set of the environment-adaptive fusion model, and obtains the modified parameter set. When the compensation amount of the modified parameter set exceeds the threshold of ±20%, the historical optimal parameter is extracted from the full-condition data to smoothly transition to the modified parameter set.

[0015] Further, the process of processing real-time data in the core database based on the edge computing unit and combining the modified parameter set to obtain the final environment-adaptive fusion model is as follows: The edge computing unit collects real-time data at a frequency of 10Hz, and uses a sliding window to remove 3σ outliers to obtain a preprocessed data set. Based on the preprocessed data set, the modified parameter set is iteratively optimized using the least squares method, updated every 60s to obtain an iterative parameter set, and the environment-adaptive fusion model is updated synchronously to obtain the final environment-adaptive fusion model. The root mean square error between the predicted motion state of the final environment-adaptive fusion model and the real-time data is calculated, and if the root mean square error is greater than the set threshold, the sliding window is reduced to N=50 groups to accelerate convergence, so that the prediction error of the final environment-adaptive fusion model is less than the error threshold.

[0016] The present application has the following advantages: By systematically acquiring core data covering operation manuals, historical cases, ship design data, and full-working-condition navigation data, a comprehensive foundation is provided for modeling. Then, by combining the static hydrodynamic calculation book, the MMG benchmark model parameter set is constructed, the environment-adaptive fusion model is obtained by fusing full-working-condition data, the model deviation is corrected through parameter compensation, and finally, relying on the edge computing unit, real-time data is processed at high frequency and the model is dynamically optimized. In addition, according to error feedback, the sliding window can be adjusted to accelerate convergence, ensuring that the final model prediction error is less than the threshold value. This provides reliable kinematic support for intelligent operations such as ship autonomous collision avoidance and trajectory tracking, effectively reduces the safety risk and energy loss of dredging operations, and solves the problems of weak model foundation caused by scattered core data acquisition in traditional trailing suction dredger hydrodynamic modeling, difficulty in adapting to complex dredging environments and dynamic working conditions relying on a single model, and lack of targeted compensation for model parameters, which may lead to inaccurate results, real-time data processing lag, and difficulty in ensuring prediction accuracy.

[0017] Of course, implementing any product of the present application does not necessarily require all the advantages described above. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 Flow chart of the method for modeling and intelligent identification of parameters of the trailing suction dredger ship's navigation dynamic hydrodynamics. DETAILED DESCRIPTION

[0019] Please refer to Figure 1 The embodiment of the present application provides a technical solution: a method for modeling and intelligent identification of parameters of the trailing suction dredger ship's navigation dynamic hydrodynamics, comprising the following steps: Acquire core data.

[0020] The core data includes dredging ship operation manual, historical accident case library, ship main dimension data, static hydrodynamic calculation book, and full-working-condition navigation data.

[0021] Traditional modeling often relies only on static hydrodynamic calculation books or a small amount of navigation data, resulting in models that lack consideration of operation logic (such as rudder operation specifications), risk scenarios (such as sudden instability of loading rate), ship physical characteristics (such as fluid resistance characteristics determined by main dimensions), and full-working-condition changes (such as 30%-100% loading rate of mud tanks, different navigation speeds / rudder angles), which may lead to problems such as "model parameters deviating from actual operations and poor multi-working-condition adaptability".

[0022] By clearly defining the complete composition of core data, we provide "operational experience-risk case-physical characteristics-actual working conditions" four-dimensional data support from the source for subsequent modeling: The operation manual and accident case library provide the basis for the construction of the rule chain of the subsequent expert system, avoiding parameter compensation without experience basis; The main scale and static hydrodynamic calculation book ensures that the initial parameters of the MMG benchmark model conform to the physical laws of the ship, avoiding the initial parameters from deviating from the characteristics of the ship; The full working condition navigation data provides sufficient samples for the SINDy algorithm to select nonlinear terms, avoiding weak model generalization ability due to incomplete working condition coverage, ultimately ensuring that the model has the basis to adapt to multiple dynamic working conditions of dredging operations from the data level, reducing the prediction deviation caused by incomplete data.

[0023] The process of obtaining core data is: Deploy perception and computing hardware, and access the collected real-time data to the edge computing unit through the CAN bus. The edge computing unit pre-stores the operation manual and historical accident case library of the dredging ship. Import the ship main scale data in the ship design drawings, static hydrodynamic calculation book, and collect more than 1000 sets of full working condition navigation data. The ship main scale data includes ship length, type width and draft. The static hydrodynamic calculation book includes ship mass, mass matrix, additional mass matrix, static water restoring force coefficient, roll damping coefficient and pitch damping coefficient. The full working condition navigation data includes mud tank loading rate, speed and rudder angle.

[0024] Traditional data collection often relies on offline sensors or decentralized storage, which has the problem of "high data transmission delay, mismatch between hardware and modeling requirements", such as the inability to collect dynamic parameters such as loading rate and roll angle in real time, or the need to transmit data to a remote server, resulting in real-time modeling lag. By deploying special sensors at key positions on the ship (attitude sensors at the center of the ship, mud tank loading rate sensors, and torque sensors for rudders and propellers), combined with CAN bus access to edge computing units, real-time synchronous collection of ship motion state (roll angle, pitch angle), loading state (mud tank loading rate), and power state (torque) is achieved, providing hardware support for subsequent 10Hz high-frequency data processing, avoiding model iteration lag caused by real-time data missing; At the same time, the edge computing unit pre-stores the operation manual and accident case library, which can quickly call basic data, avoiding the impact of data retrieval delay on working condition response speed during subsequent parameter compensation, and improving the real-time adaptation ability of the model.

[0025] The traditional modeling is prone to the problems of "design data deviating from actual working conditions and insufficient sample quantity", such as using only the static water force data in the design stage, ignoring the working condition changes in different loading rates and sailing speeds in actual operation, or the model cannot cover extreme working conditions such as light load and heavy load due to insufficient sample quantity; the main dimension data (ship length, type width, draft) in the ship design drawing and the static water force calculation book (including key parameters such as ship mass, additional mass matrix and roll damping coefficient) are clearly imported, ensuring that the inertia force coefficient and viscous force coefficient of the MMG reference model are calculated with accurate physical parameter input, avoiding the deviation caused by the initial parameters due to the ambiguity of design data; at the same time, more than 1000 groups of full working condition sailing data (including mud tank loading rate 0-100%, sailing speed 0-15kn, rudder angle-35° to +35°) are collected, covering the full scene of "light load transfer-heavy load dredging-low speed operation-high speed sailing", providing sufficient samples for the SINDy algorithm to construct a candidate function library of 42 base functions, avoiding the omission of nonlinear term screening due to incomplete working condition coverage, and improving the generalization ability of the model to dynamic working conditions; and the specific indicators of each data (such as the additional mass matrix in the static water force calculation book and the loading rate in the full working condition data) are clearly defined, ensuring that the data accuracy meets the modeling requirements, and avoiding parameter calculation errors caused by ambiguous data indicators.

[0026] The real-time data includes roll angle, pitch angle, loading rate and torque value; The process of deploying the perception and computing hardware is: Install attitude sensors at the ship body amidships or bow and stern to obtain roll angle and pitch angle; Install a loading rate sensor on the top of the mud tank to obtain the loading rate; Install a torque sensor at the rudder or propeller to obtain the torque value.

[0027] When deploying the perception and computing hardware, by installing attitude sensors at the center of the ship hull / bow and stern, loading rate sensors at the top of the mud tank, torque sensors at the rudder / propeller, and accessing the edge computing unit pre-stored with the manual of dredger operation and the historical accident case library through the CAN bus, the ship motion state (roll angle, pitch angle), loading state (loading rate), and power state (torque) data can be collected in real time, avoiding the data transmission delay and key parameter missing caused by traditional offline collection or dispersed storage. Meanwhile, the pre-stored basic data of the edge computing unit can be quickly called to provide immediate data support for subsequent parameter compensation. The main dimension data (ship length, beam, draft) in the ship design drawings and the static hydrodynamic calculation book (including ship mass, additional mass matrix, etc.) can ensure that the inertial force coefficients and viscous force coefficients of the subsequent MMG reference model are calculated based on accurate physical parameters, avoiding the initial parameter deviation caused by traditional reliance on fuzzy design data. The collection of more than 1000 groups of full working condition navigation data including mud tank loading rate, speed, and rudder angle can cover all scenarios such as light load, heavy load, low speed, and high speed in dredging operation, solving the problem of weak model generalization caused by insufficient sample size and incomplete working condition coverage. Moreover, the specific indicators of each data are clear, ensuring data accuracy and providing sufficient and reliable sample basis for subsequent construction of fusion model and selection of nonlinear terms.

[0028] Obtaining a static hydrodynamic calculation book from the core data, and obtaining a MMG reference model parameter set based on the static hydrodynamic calculation book.

[0029] Obtaining an inertial force coefficient set based on the ship mass indicated in the static hydrodynamic calculation book; Obtaining a viscous force coefficient set based on the main dimension data of the ship; Recording the inertial force coefficient set and the viscous force coefficient set as the MMG reference model parameter set.

[0030] The process of obtaining the inertial force coefficient set based on the ship mass indicated in the static hydrodynamic calculation book is as follows: Taking the ship mass indicated in the static hydrodynamic calculation book as a reference, the initial values of the inertial force coefficients of the longitudinal, transverse, and vertical oscillations in the 6-DOF translational directions are set as m, respectively; The rotational inertia moments around the x, y, and z axes indicated in the static hydrodynamic calculation book are directly taken as the initial values of the inertial force coefficients in the corresponding rotational directions, where the x axis represents roll, the y axis represents pitch, and the z axis represents bow; Extracting the added mass matrix from the static hydrodynamic calculation book, which includes the longitudinal added mass, transverse added mass, vertical added mass, and transverse added inertia moment, and superimposing the corresponding components of the added mass matrix to the set initial inertial force coefficients, respectively; The surge inertia force coefficient is modified as the sum of the ship mass and the surge added mass, the sway inertia force coefficient is modified as the sum of the ship mass and the sway added mass, the heave inertia force coefficient is modified as the sum of the ship mass and the heave added mass, and the roll inertia force coefficient is modified as the sum of the x-axis rotational inertia moment and the roll added inertia moment, to form the inertia force coefficient set.

[0031] It should be noted that the ship mass in the hydrostatic calculation book is a precise physical parameter obtained through professional hydrostatic analysis. The initial values of the inertia force coefficients of the surge, sway and heave are set as the mass, ensuring that the translational direction inertia parameters are completely matched with the actual mass properties of the ship, and avoiding the prediction deviation of the model for the linear motion of the ship caused by the deviation of the translational inertia parameters from the true value; the rotational inertia moments around the x (roll), y (pitch) and z (yaw) axes directly reflect the inertia characteristics of the ship rotating around different axes, and the initial values can make the rotational direction inertia parameters conform to the actual rotation law of the ship, and avoid the prediction distortion of the model for the rotational motion such as roll and pitch caused by inaccurate rotational inertia parameters.

[0032] When the ship moves in water, the surrounding fluid will generate additional inertia that hinders its motion. The added mass matrix (including the surge added mass, the sway added mass, the heave added mass and the roll added inertia moment) in the hydrostatic calculation book can accurately quantify this additional effect, and through superposition and modification, the final inertia force coefficients (such as surge = ship mass + surge added mass) can accurately reflect the comprehensive inertia of “ship mass + fluid added mass”, ensuring that the inertia force coefficient set is consistent with the actual motion inertia characteristics of the ship in water, providing high-precision inertia parameter basis for the subsequent MMG reference model, and avoiding the chain deviation of subsequent fusion model construction and parameter compensation caused by inertia parameter error.

[0033] The process of obtaining the viscous force coefficient set based on the main dimension data of the ship is as follows: The product of the ship length and the draft is denoted as the surge characteristic area; The product of the ship width and the draft is denoted as the sway characteristic area; The hydrostatic resistance coefficients of the hydrostatic calculation book are read, and the initial values of the surge direction viscous force coefficient and the initial values of the sway direction viscous force coefficient are inversely calculated according to the fluid viscous resistance formula; The roll damping coefficient and the pitch damping coefficient of the hydrostatic calculation book are directly used as the initial values of the viscous force coefficients in the corresponding rotational direction; The initial values of the surge direction viscous force coefficient, the initial values of the sway direction viscous force coefficient, the initial values of the roll rotational direction viscous force coefficient and the initial values of the pitch rotational direction viscous force coefficient are denoted as the viscous force coefficient set.

[0034] The ship main dimension data (ship length, type width, draft) is derived from design drawings, which is a precise parameter reflecting the underwater profile of the ship. The longitudinal and transverse characteristic areas calculated therefrom can accurately quantify the effective area of the ship in contact with the water flow in the longitudinal and transverse directions, providing basic data that conforms to the actual structure of the ship for subsequent viscous force coefficient calculation, avoiding distortion of resistance calculation due to deviation of characteristic area. The static water resistance coefficient in the static water force calculation book is a key parameter obtained through professional fluid mechanics analysis. Combined with the fluid viscous resistance formula (including water density, design speed, etc. Variables) to back calculate the coefficient, the longitudinal and transverse viscous force coefficients can accurately match the actual water flow resistance law of the ship at different speeds, avoiding deviation of the model in judging the straight sailing state of the ship due to inaccurate prediction of the translation direction viscous force.

[0035] The roll and pitch damping coefficients are the quantitative results of the fluid damping effect of the ship's rotational motion in the static water force calculation. Directly as the initial value can ensure that the rotational direction viscous force coefficient conforms to the actual fluid resistance characteristics when the ship rolls and pitches, avoiding distortion of the model in judging the rolling motion of the ship due to deviation of the rotational damping prediction; Form a structured viscous force coefficient set, which can be directly used as a core component of the MMG reference model parameter set, ensuring that the MMG reference model can completely and accurately depict the viscous resistance influence of the ship in water, reducing the overall prediction deviation of the model caused by viscous coefficient error from the parameter source.

[0036] Obtain full working condition sailing data from core data, and obtain environment adaptive fusion model combined with MMG reference model parameter set.

[0037] The SINDy algorithm is used to process the full working condition sailing data to obtain a candidate function library. Significant nonlinear terms are selected through L1 regularization sparse regression to obtain a set of nonlinear correction factors. Specifically: Extract the ship motion state parameters in the full working condition sailing data, which at least include the linear velocity components (longitudinal velocity u, transverse velocity v, vertical velocity w) and angular velocity components (roll angle velocity p, pitch angle velocity q, yaw angle velocity r) of the ship's 6-DOF motion. Perform 3σ criterion outlier rejection and Z-score standardization processing on the motion state parameters to obtain a standardized motion state data set D.

[0038] Based on the SINDy algorithm (sparse factor is set to 0.01), the linear velocity components and angular velocity components in the standardized motion state data set D are used as basic variables to construct a candidate function library L containing basic terms, quadratic terms and cross terms: Basic terms: including longitudinal velocity u, transverse velocity v, vertical velocity w, roll angle velocity p, pitch angle velocity q, and yaw angle velocity r, a total of 6 base functions; Quadratic terms: including u 2 , v2 , w 2 , p 2 , q 2 , r 2 , uv, uw, up, uq, ur, vw, vp, vq, vr, wp, wq, wr, pq, pr, qr, a total of 21 base functions; Cross terms: including up, uq, ur, vp, vq, vr, wp, wq, wr, u 2 p, u 2 q, v 2 p, v 2 q, w 2 p, w 2 q, a total of 15 base functions; The total number of base functions of the candidate function library L is 42, and each base function is constructed based on the variables of the standardized motion state data set D, ensuring that the function value range is adapted to the data characteristics; The central difference method is used to calculate the time derivative of each parameter for the surge speed u, the sway speed v, the heave speed w, the roll angular speed p, the pitch angular speed q, and the yaw angular speed r in the standardized motion state data set D, to obtain the motion state derivative vector.

[0039] The candidate function library L is used as the input matrix, and the motion state derivative vector is used as the output vector to construct a linear regression model: motion state derivative vector = candidate function library * base function coefficient vector + error term. The L1 regularization method is used to solve the linear regression model, and by minimizing the objective function, the coefficients corresponding to the redundant base functions are suppressed, and the base function coefficients that significantly contribute to the motion state derivative are retained.

[0040] The coefficient threshold is set, and the base functions corresponding to the coefficients in the solved coefficient vector whose absolute value is greater than the coefficient threshold are determined as significant nonlinear terms; the significant nonlinear terms and their corresponding coefficients are extracted to form a nonlinear correction factor set, wherein each element of the nonlinear correction factor set includes a function expression of the significant nonlinear term and a coefficient value corresponding to the function, and the coefficient value satisfies the physical meaning constraint (such as the non-negative value of the resistance-related term coefficient).

[0041] The nonlinear correction factor set is coupled with the MMG reference parameter set to construct a fused 6-DOF motion equation, denoted as the fusion model, in detail: First, the MMG reference parameter set is extracted, and the nonlinear correction factor set obtained by the SINDy algorithm is extracted, and based on the 6-DOF motion equation of the classic MMG model, the nonlinear correction factor set is integrated into the core matrix of the equation: For the inertia matrix, the main diagonal components of the inertia force coefficient set in the MMG reference parameter set are retained, and the nonlinear correction (nonlinear compensation of the focused damping term in the initial coupling stage) is not introduced at present; For the damping matrix, on the basis of the viscous force coefficient set in the MMG reference parameter set, the significant nonlinear terms and their coefficients in the nonlinear correction factor set are superimposed to realize the nonlinear characterization of the damping characteristics; For the Coriolis force matrix and the hydrostatic restoring force matrix, the MMG reference parameter set is temporarily set; Finally, the fused 6-DOF motion equation is formed, denoted as the fusion model.

[0042] The wave height data collected by the attitude sensor is calculated using the JONSWAP wave spectrum to obtain the wave excitation force, and the damping matrix of the fusion model M1 is corrected to obtain an environment-adaptive fusion model, specifically: The attitude sensor deployed on the ship's amidships or bow and stern is called to collect the wave height data (unit: m) of the ship's working water area in real time, and the collection frequency is synchronized with the data processing frequency of the edge computing unit (10 Hz); The collected wave height data is preprocessed: abnormal values (such as out-of-range data caused by instantaneous sensor failure) are removed by the 3σ criterion, and the average wave height is calculated using a sliding window (the window size is consistent with the real-time data processing window) to obtain effective wave height data for wave spectrum analysis.

[0043] According to the preprocessed effective wave height, combined with the common sea conditions in the dredging working water area (such as nearshore or estuary), the core parameters of the JONSWAP spectrum are set: spectral peak period, spectral peak factor, spectral peak factor. Substitute the JONSWAP wave spectrum formula to calculate the wave power spectral density at different circular frequencies. Based on the calculated wave power spectral density at different circular frequencies, combined with the ship's main dimension data, the wave excitation force in 6-DOF direction of the ship is calculated using linear potential flow theory, including longitudinal excitation force, transverse excitation force, vertical excitation force, roll excitation moment, pitch excitation moment, and bow excitation moment. During the calculation, the wave power spectral density and the corresponding wave excitation force transfer function (derived based on ship lines) are convolved and integrated to obtain the time-domain wave excitation force in each direction.

[0044] The wave excitation force will increase the additional damping of the ship's motion (such as the damping effect of waves on roll), so the wave excitation force in each direction is converted into the corresponding damping correction, and the roll additional damping coefficient is calculated according to the ratio of the roll excitation moment to the roll angular velocity; The above additional damping coefficient is superimposed into the damping matrix of the fusion model to obtain the corrected damping matrix; Replace the original damping matrix in the fusion model to finally obtain the environment-adaptive fusion model, which can adapt to the influence of the wave environment on the ship's hydrodynamic characteristics in real time and reduce the prediction error caused by wave disturbance.

[0045] The SINDy algorithm is based on full-condition navigation data (including mud tank loading rate, speed, etc. Full scene data) to construct a candidate function library containing speed terms, quadratic terms, and cross terms, which can comprehensively cover the nonlinear characteristics of ship motion; L1 regularization sparse regression can accurately select the nonlinear terms that significantly affect the motion state, eliminate redundant terms, and form a nonlinear correction factor set F that not only retains the accuracy of data-driven but also avoids overfitting, providing a reliable nonlinear correction basis for subsequent model fusion.

[0046] The MMG benchmark parameter set (including inertial force and viscous force coefficient) has clear physical meaning, ensuring that the model meets the basic physical laws of ship motion, and the nonlinear correction factor set F can supplement the nonlinear effects not covered by the MMG model. The fusion model coupled by the two can avoid the problem of insufficient nonlinear description of pure physical models, and solve the problem of poor physical interpretability and no actual physical meaning of parameters in pure data models, and can more accurately reflect the actual motion characteristics of ships. The JONSWAP wave spectrum can accurately calculate the wave excitation force based on the real wave height data collected by the attitude sensor, and by modifying the damping matrix of the fusion model, the model can adapt to the wave environment of the operation water area in real time, effectively offsetting the interference of waves on the hydrodynamic characteristics of the ship, greatly reducing the model prediction error caused by environmental factors, and ensuring the high reliability of the model in complex sea conditions.

[0047] Compensate the parameter set of the environment-adaptive fusion model to obtain a modified parameter set.

[0048] Based on the dredger operation manual and the historical accident case library, a trigger condition-parameter compensation rule chain is constructed to form a rule library, specifically: The dredger operation manual is structurally analyzed to extract the core content of "working condition limit-operation parameter requirement", such as the recommended speed range corresponding to different mud tank loading rates (0-30% light load, 30%-70% medium load, 70%-100% heavy load) (light load ≤12kn, heavy load ≤8kn), rudder angle operation threshold (normal working condition ≤25°, low speed working condition ≤30°), and water dynamic parameter adjustment specification under special working conditions (such as berthing, dredging operation), a total of 120+ basic operation rule elements are extracted; The historical accident case library (including 30+ instability events caused by sudden changes in loading rate) is analyzed to filter out the correlation between "working condition characteristics-parameter anomaly-instability consequences", such as the key features of typical instability cases such as "mud tank loading rate increased by ≥20% within 10s + speed ≤3kn → insufficient roll damping coefficient → ship roll exceeds threshold" and "large rudder angle (δ ≥30°) + wave period ≥8s → rudder factor attenuation → heading deviation exceeds 5°", a total of 30+ risk correlation elements are extracted.

[0049] Based on the extracted information, four key state parameters are identified: mud tank loading rate λ (real-time value and 10s / 30s change rate), speed u (real-time value), rudder angle δ (real-time value), and wave period T (calculated from wave height data collected by the attitude sensor). Each parameter is assigned a quantization interval (e.g., λ change rate: [0, 5%], [5%, 20%], [20%, +∞]). For the core hydrodynamic parameter set of the environment-adaptive fusion model (including 12 key parameters such as turning damping coefficient, rudder efficiency factor, longitudinal viscous force coefficient, and roll damping coefficient), the compensation direction (increase / decrease) and quantization range of each parameter are clearly defined (e.g., turning damping coefficient compensation range: ±5%-±20%).

[0050] The "IF-THEN" production rule form is used to construct the trigger condition-parameter compensation rule chain, for example: Rule 1: IF (λ10s change rate ≥ 20%) AND (u ≤ 3kn) THEN (turning damping coefficient +15%, rudder efficiency factor -8%); Rule 2: IF (δ ≥ 30°) AND (T ≥ 8s) THEN (rudder efficiency factor +10%, longitudinal viscous force coefficient +5%); Rule 3: IF (λ ≥ 80%) AND (u ≥ 10kn) THEN (roll damping coefficient +12%, vertical viscous force coefficient +7%); A total of 150+ rule chains are constructed, and all rule chains are stored in a structured rule library according to "working condition type (light load / medium load / heavy load, calm water / wave water)". Each rule is assigned a matching priority (risk-related rules have higher priority than basic operation rules) to facilitate subsequent rapid matching.

[0051] The edge computing unit collects the current state in real time, matches with the rule library, compensates the parameter set of the environment-adaptive fusion model, and obtains the modified parameter set. Specifically: The edge computing unit synchronously collects the following current state parameters at a frequency of 10Hz: Mud tank loading rate λ: real-time value collected by the mud tank top loading rate sensor, 10s change rate (Δλ 10 = (λ 10 current-λ 10 previous) / λ Speed u, rudder angle δ: real-time values obtained from the ship navigation system and the rudder control system, respectively; Wave period T: calculated based on real-time wave height data collected by the attitude sensor using spectral analysis method; The collected parameters are subjected to 3σ outlier rejection (e.g., λ out-of-range value caused by sensor instantaneous failure), and the standardized current state parameter vector (λ, Δλ 10,u,δ,T).

[0052] Given the current state parameter vector, iterate through all rule chains in the rule base for triggering conditions, and calculate the matching degree (e.g., the rate of change Δλ) for each triggering condition dimension. 10 =22% and "Δλ" in Rule 1 10 The matching degree of "≥20%" is 100%, and the matching degree of u=2.8kn and "u≤3kn" is 100%; calculate the comprehensive matching degree of a single rule (the weighted average of the matching degrees of each dimension, with risk dimensions such as Δλ). 10 With weights set to 0.4, u weights set to 0.3, δ weights set to 0.2, and T weights set to 0.1, rules with a comprehensive matching degree ≥ 90% are selected as valid matching rules. If there are multiple valid matching rules, they are sorted by rule priority (risk-related rules > basic operation rules) and comprehensive matching degree (from high to low), and the rule with the highest priority and the highest matching degree is selected as the final execution rule (if the priorities are the same, the rule with the most compensation parameters is selected).

[0053] Based on the parameter compensation instructions of the final execution rule, the parameter set of the environment-adaptive fusion model is compensated: Extract the current parameter set P1' of the environment-adaptive fusion model (including 12 parameters such as the cyclonic damping coefficient D_r and the rudder effect factor K_δ); Calculate the compensation value of each parameter according to the compensation direction and compensation amount specified in the rules (e.g., in rule 1, D_r compensation value = D_r current × 15%, K_δ compensation value = K_δ current × (-8%)). The compensation value is superimposed on the original parameters to obtain the corrected parameter set P2 (P2=P1'+compensation value), and then synchronized to the model calculation module via the ship's local area network to replace the original parameter set.

[0054] When the compensation amount in the corrected parameter set exceeds the ±20% threshold, historically optimal parameters are extracted from the full-condition data for a smooth transition to the corrected parameter set. Specifically: Calculate the compensation percentage for each parameter in the corrected parameter set P2: Compensation percentage = (parameter value in P2 - corresponding parameter value in P1') / corresponding parameter value in P1' × 100%; Iterate through the compensation percentages of all parameters in P2 and determine whether any parameter's compensation percentage exceeds the ±20% threshold (e.g., yaw damping coefficient compensation percentage = 25% > 20%, or rudder effect factor compensation percentage = -22% < -20%). If there are parameters that exceed the threshold, the "historical best parameter smooth transition" mechanism is triggered; if the compensation amount of all parameters is within ±20%, then P2 is directly confirmed as the final corrected parameter set.

[0055] Retrieve the collected 1000 sets of full-condition navigation data, and extract the historical optimal parameters according to the following steps: Condition similarity matching: Take the current state parameter vector (λ, Δλ 10 , u, δ, T) as the reference, and select "similar condition data" from the full-condition data, with the deviation of each parameter from the current value ≤10% (for example, if the current λ = 85%, select records with λ ∈ [76.5%, 93.5%] from the full-condition data; if the current u = 9kn, select records with u ∈ [8.1kn, 9.9kn]); Optimal parameter selection: For the historical model parameter set (including D_r, K_δ, etc. 12 parameters) corresponding to the similar condition data, calculate the "model prediction error" (the root mean square error RMSE of the model prediction value corresponding to the historical parameters and the actual navigation data) of each parameter in this condition, and select the parameter group with the smallest RMSE as the "historical optimal parameter set P_opt" (for example, the parameter group with RMSE ≤0.08m / s in the similar condition).

[0056] Use "linear interpolation method" to realize the smooth transition of P2 and P_opt: Set the transition time window t = 10s (i.e. 10 data collection periods, corresponding to the edge computing unit 10Hz collection frequency), calculate the parameter interpolation increment of each transition period: ΔP_t = (P_opt - P2) / t; In each transition period, update the current corrected parameter set to P2_t = P2 + ΔP_t × t_current (t_current is the current transition period number, t_current ∈ [1, 10]); When t_current = 10, the parameter set is completely transitioned to P_opt, and the parameter set at this time is determined as the final corrected parameter set P2_final, ensuring continuous and smooth parameter change, and avoiding model dynamic instability caused by excessive compensation (such as sudden change of roll angular velocity, fluctuation of heading).

[0057] It should be noted that the rule chain is constructed based on the "condition-operation specification" of the operation manual (such as the speed and rudder angle limits corresponding to different load rates) and the "instability characteristics-parameter abnormal association" of the historical accident case library (such as sudden increase of load rate leading to insufficient turning damping), which not only conforms to the physical law of ship operation, but also covers multiple conditions such as "light load / heavy load, calm / wavey waters", avoiding the problem of no compensation basis in special conditions due to missing rules, providing accurate and interpretable execution standards for subsequent parameter compensation.

[0058] The edge computing unit synchronously collects current state parameters such as loading rate, speed, rudder angle at a frequency of 10 Hz, and can quickly capture changes in working conditions; through multi-dimensional fuzzy matching algorithm and real-time comparison with rule base (matching degree ≥ 90% triggers compensation), it can adjust model parameters (such as turning damping coefficient, rudder efficiency factor) in the moment of working condition change (such as loading rate increasing by ≥ 20% within 10s), avoiding the expansion of the deviation between model prediction and actual movement due to lagging compensation, and improving the real-time adaptability of the model to dynamic working conditions.

[0059] The ±20% threshold can effectively constrain the compensation magnitude, avoiding parameter abnormalities caused by "overfitting"; when the trigger threshold is reached, the historical optimal parameters (the parameter set with the smallest model prediction error) under similar working conditions are extracted from the collected 1000 or more groups of full working condition data, and linear interpolation is used for smooth transition, ensuring continuous and smooth parameter changes, suppressing the negative effects of excessive compensation, and ensuring the reliability of the corrected parameter set, improving the overall robustness of the model.

[0060] The edge computing unit processes real-time data in the core database and obtains the final environment-adaptive fusion model based on the corrected parameter set.

[0061] The edge computing unit collects real-time data at a frequency of 10 Hz, uses a sliding window to remove 3σ outliers, and obtains a preprocessed data set. Specifically: Based on the deployed perception hardware architecture, the edge computing unit synchronously collects real-time data through the CAN bus, and the collected objects are the core running state parameters of the ship: Motion state data: roll angle (unit: °) and pitch angle (unit: °) output by the attitude sensor at the ship's amidships / bow and stern, and roll angular velocity (unit: ° / s) and pitch angular velocity (unit: ° / s) obtained by differential calculation; Loading state data: real-time mud tank loading rate λ (unit: %) output by the mud tank top loading rate sensor; Power-related data: torque value (unit: ) output by the torque sensor at the rudder / pusher, and real-time speed u (unit: kn) and rudder angle δ (unit: °) transmitted by the ship navigation system; The collection frequency is strictly controlled at 10 Hz (i.e., 10 data samples and transmissions per second), ensuring that the data timestamps are synchronized and avoiding subsequent processing errors caused by timing deviations.

[0062] The edge computing unit initializes the sliding window, with the initial window size set to N = 100 groups (corresponding to 10 seconds of collected data, forming an adaptive logic with the subsequent "reduced to N = 50 groups"), and the window uses a "first-in, first-out" caching mechanism: with each new data collection, the oldest data in the window is removed, and the window always maintains the latest N groups of data, ensuring data timeliness and working condition relevance.

[0063] For each group of parameters (such as roll angle, load rate, speed) in the sliding window, 3σ outlier rejection is performed: First, calculate the mean μ (such as the mean of the roll angle) and the standard deviation σ of a parameter in the window; Second, determine the outlier threshold: the lower limit is μ-3σ, and the upper limit is μ+3σ; Third, remove the abnormal data (such as roll angle out-of-range values caused by sensor instantaneous failure) in the window that exceeds the range [μ-3σ, μ+3σ], if the window data is insufficient after removal N / 2, then supplement the normal data of the previous window to maintain the integrity of the window; Fourth, store the window data after removing outliers in the "timestamp-motion state-load state-dynamic data" structure to form a pre-processed data set, providing high-quality input for subsequent parameter optimization.

[0064] Based on the pre-processed data set, the least squares method is used to iteratively optimize the corrected parameter set, updated every 60s, to obtain the iterative parameter set, and the environment adaptive fusion model is updated synchronously to obtain the final environment adaptive fusion model, specifically: The optimization object is the corrected parameter set P2 (including 12 core hydrodynamic parameters such as the inertia force coefficient, the viscous force coefficient, and the damping matrix component of the environment adaptive fusion model); based on the least squares method, the objective function is constructed: taking the actual motion state (such as the actual roll angular velocity p_real and the actual surge velocity u_real) in the pre-processed data set as the benchmark, minimizing the residual sum of squares of the predicted motion state and the actual value of the environment adaptive fusion model.

[0065] The edge computing unit sets a 60s timing trigger mechanism: every 60s of pre-processed data (corresponding to 600 groups of 10Hz acquisition data, about 6 sliding window data) is accumulated, and the parameter optimization is started once: First, substitute the actual motion state parameters in the pre-processed data set into the objective function J, and solve the parameter correction amount ΔP (such as the inertia force coefficient correction amount ΔI and the viscous force coefficient correction amount ΔD) that minimizes J through matrix differentiation; Second, calculate the iterative parameter set P3=P2+ΔP, where ΔP needs to satisfy the physical constraints (such as non-negative damping coefficient and inertia coefficient consistent with ship mass distribution law), if ΔP exceeds the constraint range, it is truncated to the boundary value; Third, synchronize the iterative parameter set P3 to the environment adaptive fusion model through the ship local area network, replace the original parameter set P2, complete the model parameter update, and obtain the preliminary "final environment adaptive fusion model".

[0066] The root mean square error of the predicted motion state of the final environment-adaptive fusion model and the real-time data is calculated. If the root mean square error is greater than a set threshold, the sliding window is reduced to N=50 groups for accelerated convergence, so that the prediction error of the final environment-adaptive fusion model is less than the error threshold. Specifically: The edge computing unit collects the latest 10 groups of synchronous data (corresponding to 1 second of collection) in real time as "real-time verification data", and calculates the error with the contemporaneous predicted motion state (such as predicted roll angle, predicted speed) output by the final environment-adaptive fusion model: For each motion state parameter, the RMSE is calculated according to the formula as the overall model prediction error index.

[0067] Set the model prediction error threshold (such as RMSE≤0.1m / s, which meets the accuracy requirement of ship autonomous collision avoidance): If the calculated RMSE is less than or equal to the set threshold, it is determined that the current "final environment-adaptive fusion model" meets the accuracy requirement, and the sliding window size (N=100 groups) is maintained, and the next 60s iteration optimization cycle is entered; If the RMSE is greater than the set threshold, it is determined that the model convergence speed is insufficient, and the sliding window size is immediately reduced from N=100 groups to N=50 groups: after the window is reduced, the data update frequency is improved (the window data is replaced every 5 seconds), which can capture the working condition changes (such as sudden change of mud tank loading rate) faster, so that the least square method optimization can correct the parameters more frequently, and accelerate the model convergence.

[0068] After the sliding window is reduced, the "data collection-exception value elimination-parameter optimization-RMSE calculation" process is repeated: the RMSE is recalculated every 30s (corresponding to 6 data quantities of the reduced window), until the RMSE is less than or equal to the set threshold, at which point it is confirmed that the prediction error of the "final environment-adaptive fusion model" meets the requirement, and the window adjustment is stopped, and the current window size and optimization frequency are maintained, providing a stable kinematic basis for ship autonomous collision avoidance and trajectory tracking.

[0069] An electronic device, comprising: a processor; and a memory having computer program instructions stored therein, which, when executed by the processor, cause the processor to perform the trailing suction dredger ship motion dynamic hydrodynamic modeling and parameter intelligent identification method as described above.

[0070] A computer-readable storage medium for storing a program, which, when executed by a processor, implements the trailing suction dredger ship motion dynamic hydrodynamic modeling and parameter intelligent identification method as described above.

[0071] Those skilled in the art will appreciate that embodiments of the present application can be devised for a variety of applications. It is intended that the present application be limited only by the scope of the appended claims, and it is intended that various modifications and alterations made by those skilled in the art be considered as within the scope of the present application. The embodiments of the present application will be described with reference to the attached drawings, wherein:

[0072] The present application is described in reference to the drawings using a flowchart and / or a block diagram of an embodiment of the system, apparatus (system), and computer program product according to the present application. It will be understood that each block of the flowchart and / or block diagram, and combinations of blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processing device, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0073] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0074] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0075] While the preferred embodiments of the application have been described, additional variations and modifications can be made to the embodiments by those skilled in the art once they learn of the basic inventive concepts. Therefore, the appended claims are intended to cover all such modifications and alterations as fall within the true spirit and scope of the present application.

[0076] Obviously, many modifications and variations of the present application are possible in light of the above teachings. It is, therefore, to be understood that within the scope of the appended claims and their equivalents, the application can be practiced otherwise than as specifically described.

Claims

1. A method for dynamic hydrodynamic modeling and intelligent parameter identification of a trailing suction hopper dredger, characterized in that, Includes the following steps: Obtain core data; Obtain hydrostatic calculation sheets from core data, and obtain the MMG baseline model parameter set based on the hydrostatic calculation sheets; Full-condition navigation data is obtained from the core data and combined with the MMG baseline model parameter set to obtain an environment-adaptive fusion model. The parameter set of the environment-adaptive fusion model is compensated to obtain the corrected parameter set; The edge computing unit processes real-time data from the core database and combines it with the corrected parameter set to obtain the final environment-adaptive fusion model.

2. The method for dynamic hydrodynamic modeling and intelligent parameter identification of a trailing suction hopper dredger according to claim 1, characterized in that, The core data includes dredging vessel operation manuals, historical accident case databases, main dimensions of the vessel, hydrostatic calculation sheets, and navigation data under all operating conditions.

3. The method for dynamic hydrodynamic modeling and intelligent parameter identification of a trailing suction hopper dredger according to claim 2, characterized in that, The process of acquiring core data is as follows: Deploy sensing and computing hardware to connect the collected real-time data to the edge computing unit via CAN bus. The edge computing unit pre-stores the dredging vessel operation manual and a historical accident case library. Import the ship's main dimensions data and hydrostatic calculation sheets from the ship design drawings, and collect more than 1,000 sets of full-condition navigation data. The ship's main dimensions data include the ship's length, beam, and draft. The hydrostatic calculation sheets include the ship's mass, mass matrix, added mass matrix, hydrostatic restoring force coefficient, roll damping coefficient, and pitch damping coefficient. The full-condition navigation data includes the mud hull loading rate, speed, and rudder angle.

4. The method for dynamic hydrodynamic modeling and intelligent parameter identification of a trailing suction hopper dredger according to claim 3, characterized in that, The real-time data includes roll angle, pitch angle, load factor, and torque value; The process of deploying the sensing and computing hardware is as follows: Attitude sensors are installed amidships or at the bow and stern to obtain roll and pitch angles. A loading rate sensor is installed on the top of the mud tank to obtain the loading rate; A torque sensor is installed on the servo or thruster to obtain the torque value.

5. The method for dynamic hydrodynamic modeling and intelligent parameter identification of a trailing suction hopper dredger according to claim 3, characterized in that, The process of obtaining the MMG baseline model parameter set based on hydrostatic calculations is as follows: The set of inertial force coefficients is obtained based on the ship mass specified in the hydrostatic calculation book; A set of viscous force coefficients was obtained based on the ship's principal dimension data; The set of inertial force coefficients and the set of viscous force coefficients are denoted as the MMG reference model parameter set.

6. The method for dynamic hydrodynamic modeling and intelligent parameter identification of a trailing suction hopper dredger according to claim 5, characterized in that, The process of obtaining the set of inertial force coefficients based on the ship's mass as stated in the hydrostatic calculation book is as follows: Based on the ship mass stated in the hydrostatic calculation book, the initial values ​​of the inertial force coefficients for the sway, roll, and heave in the six-degree-of-freedom translational direction are set to m respectively; The rotational moments of inertia about the x-axis, y-axis, and z-axis as stated in the hydrostatic calculation book are directly used as the initial values ​​of the inertial force coefficients in the corresponding rotational directions, where the x-axis represents roll, the y-axis represents pitch, and the z-axis represents bow roll. Extract the additional mass matrix from the hydrostatic calculation report. The additional mass matrix includes the surge additional mass, sway additional mass, heave additional mass, and roll additional moment of inertia. Then, superimpose the corresponding components of the additional mass matrix onto the predetermined initial inertial force coefficients. The sway inertial force coefficient is corrected to the sum of the ship's mass and the sway-added mass; the yaw inertial force coefficient is corrected to the sum of the ship's mass and the yaw-added mass; the heave inertial force coefficient is corrected to the sum of the ship's mass and the heave-added mass; and the roll inertial force coefficient is corrected to the sum of the x-axis rotational moment of inertia and the roll-added moment of inertia, thus forming a set of inertial force coefficients.

7. The method for dynamic hydrodynamic modeling and intelligent parameter identification of a trailing suction hopper dredger according to claim 5, characterized in that, The process of obtaining the viscous force coefficient set based on the ship's principal dimension data is as follows: The product of the ship's length and draft is denoted as the sway characteristic area; The product of the width and the draft is denoted as the sway characteristic area; Read the hydrostatic resistance coefficient from the hydrostatic calculation book, and deduce the initial values ​​of the viscous force coefficient in the longitudinal and transverse directions respectively based on the fluid viscous resistance formula. The roll damping coefficient and pitch damping coefficient from the hydrostatic calculation sheet are directly used as the initial values ​​of the viscous force coefficient in the corresponding rotation direction. The initial values ​​of the viscous force coefficients in the sway direction, the sway direction, the roll rotation direction, and the pitch rotation direction are denoted as the viscous force coefficient set.

8. The method for dynamic hydrodynamic modeling and intelligent parameter identification of a trailing suction hopper dredger according to claim 4, characterized in that, The process of obtaining full-condition navigation data from core data and combining it with the MMG baseline model parameter set to obtain an environment-adaptive fusion model is as follows: The SINDy algorithm was used to process the navigation data under all operating conditions to obtain a candidate function library. Significant nonlinear terms were screened by L1 regularized sparse regression to obtain a set of nonlinear correction factors. The nonlinear correction factor set is coupled with the MMG reference parameter set to construct the fused 6-DOF motion equations, denoted as the fused model. The wave height data collected by the attitude sensor was calculated using the JONSWAP wave spectrum to obtain the wave excitation force. The damping matrix of the fusion model was then corrected to obtain an environment-adaptive fusion model.

9. The method for dynamic hydrodynamic modeling and intelligent parameter identification of a trailing suction hopper dredger according to claim 3, characterized in that, The process of compensating the parameter set of the environment-adaptive fusion model to obtain the corrected parameter set is as follows: Based on the dredging vessel operation manual and historical accident case database, a trigger condition-parameter compensation rule chain is constructed to form a rule base; The edge computing unit collects the current state in real time, matches it with the rule base, and compensates the parameter set of the environment-adaptive fusion model to obtain the corrected parameter set. When the compensation amount in the corrected parameter set exceeds the ±20% threshold, the historical best parameters are extracted from the full-condition data and smoothly transitioned to the corrected parameter set.

10. The method for dynamic hydrodynamic modeling and intelligent parameter identification of a trailing suction hopper dredger according to claim 3, characterized in that, The process of processing real-time data in the core database based on edge computing units and combining it with the corrected parameter set to obtain the final environment-adaptive fusion model is as follows: The edge computing unit collects real-time data at a frequency of 10Hz and uses a sliding window to remove 3σ outliers to obtain a preprocessed dataset. Based on the preprocessed dataset, the least squares method is used to iteratively optimize the corrected parameter set, which is updated every 60 seconds to obtain the iterative parameter set. The environment-adaptive fusion model is updated synchronously to obtain the final environment-adaptive fusion model. Calculate the root mean square error between the motion state predicted by the final environment-adaptive fusion model and the real-time data. If the root mean square error is greater than a set threshold, reduce the sliding window to N=50 groups to accelerate convergence, so that the prediction error of the final environment-adaptive fusion model is less than the error threshold.

Images