A method, device, medium and product for constructing a three-degree-of-freedom mathematical model of ship power positioning

Abstract

The application discloses a ship power positioning three-degree-of-freedom mathematical model construction method, equipment, medium and product, relates to the field of ship automatic control and ocean engineering equipment, and comprises the following steps: establishing a ship three-degree-of-freedom dynamics benchmark model; constructing a neural network identifier; taking the motion state and thrust historical value of the ship as input, and training the neural network identifier by using a total loss function; the total loss function comprises a prediction error term and a physical constraint term; the trained neural network identifier is used to identify the to-be-identified physical parameters in the ship three-degree-of-freedom dynamics benchmark model, and a predicted physical parameter is obtained; and the predicted physical parameter is substituted into the ship three-degree-of-freedom dynamics benchmark model to construct a ship power positioning three-degree-of-freedom mathematical model. The application can solve the problem that a traditional method relies on physical tests and cannot adapt to parameter time variation and nonlinearity, so that the constructed ship power positioning three-degree-of-freedom mathematical model is always consistent with the actual ship dynamics.

Description

Technical Field

[0001] This application relates to the field of ship automatic control and marine engineering equipment, and in particular to a method, equipment, medium and product for constructing a three-degree-of-freedom mathematical model of ship dynamic positioning. Background Technology

[0002] Maintain or track a set position and heading. Its core lies in a precise mathematical model, mainly including: kinematic model, dynamic model, and environmental disturbance model.

[0003] Traditional dynamic positioning models often use a three-degree-of-freedom (Surge, Sway, Yaw) model. The model parameters are usually obtained in the following ways: 1) Model test: conducted in a towing tank or wind tunnel, which is costly; 2) Empirical formula: estimated based on ship type statistical formulas, which has limited accuracy; 3) Online identification: such as recursive least squares (RLS).

[0004] However, existing technologies have the following drawbacks: 1) Changes in ship load and draft cause time-varying model parameters, and fixed parameter models are prone to mismatch; 2) Methods such as RLS are sensitive to noise and converge slowly or even diverge under nonlinear conditions such as low speed and large drift angle; 3) Traditional methods are difficult to effectively handle the strong nonlinear characteristics of hydrodynamics. Summary of the Invention

[0005] The purpose of this application is to provide a method, device, medium, and product for constructing a three-degree-of-freedom mathematical model for ship dynamic positioning. This method solves the problems of traditional methods relying on physical experiments and being unable to adapt to time-varying and nonlinear parameters. It ensures that the constructed three-degree-of-freedom mathematical model for ship dynamic positioning always keeps in line with the actual ship dynamics, thereby improving the control accuracy and robustness of the ship dynamic positioning system.

[0006] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a method for constructing a three-degree-of-freedom mathematical model for ship dynamic positioning, including: Establish a three-degree-of-freedom dynamic benchmark model for a ship; Construct a neural network identifier; The neural network identifier is trained using the ship's motion state and thrust history as inputs and a total loss function; the total loss function includes a prediction error term and a physical constraint term. A trained neural network identifier is used to identify the physical parameters to be identified in the three-degree-of-freedom dynamic benchmark model of the ship, and the predicted physical parameters are obtained. Substitute the predicted physical parameters into the ship's three-degree-of-freedom dynamic benchmark model to construct a ship's dynamic positioning three-degree-of-freedom mathematical model.

[0007] Secondly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described method for constructing a three-degree-of-freedom mathematical model for ship dynamic positioning.

[0008] Thirdly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method for constructing a three-degree-of-freedom mathematical model for ship dynamic positioning.

[0009] Fourthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method for constructing a three-degree-of-freedom mathematical model for ship dynamic positioning.

[0010] According to the specific embodiments provided in this application, this application has the following technical effects: (1) By using a neural network identifier to identify physical parameters in real time and substitute them into the benchmark model, the constructed mathematical model can be updated in real time with the actual operating state of the ship, which solves the problem of dynamic mismatch between the model and the actual ship and ensures the effectiveness of the model under all operating conditions. (2) Introducing a “physical constraint term” into the loss function enables the neural network identifier to learn parameters that conform to physical meaning during the training process. This not only improves the interpretability and credibility of the physical parameters to be identified at the physical level, but also effectively prevents parameter divergence caused by sensor noise or abnormal data, thereby enhancing the robustness and safety of the dynamic positioning system. (3) By using the nonlinear mapping capability of neural networks to replace the traditional linear identification algorithm, the strong nonlinear hydrodynamic characteristics of ships under complex sea conditions can be captured more accurately. At the same time, by using "motion state and thrust history value" as input, the dynamic memory effect of the system is fully considered, thereby improving the parameter identification accuracy and algorithm convergence speed under complex nonlinear conditions such as low speed and large drift angle. (4) Online identification can be performed directly using actual ship operation data. There is no need to conduct expensive towing tank or wind tunnel tests, nor is it necessary to rely on empirical formulas with limited accuracy. This reduces the modeling cost and time cycle of the ship dynamic positioning system and has practical engineering value. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a flowchart illustrating a method for constructing a three-degree-of-freedom mathematical model for ship dynamic positioning, as provided in an embodiment of this application. Detailed Implementation

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, this application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0015] In one exemplary embodiment, such as Figure 1 As shown, a method for constructing a three-degree-of-freedom mathematical model for ship dynamic positioning is provided. This method is executed by a computer device, specifically by a computer device such as a terminal or a server alone, or by both a terminal and a server. In this embodiment, the method is described using a server as an example, and includes the following steps S1 to S5.

[0016] S1: Establish a three-degree-of-freedom dynamic benchmark model for the ship.

[0017] S2: Construct a neural network identifier.

[0018] S3: Using the ship's motion state and thrust history as input, train the neural network identifier using the total loss function; the total loss function includes a prediction error term and a physical constraint term.

[0019] S4: The trained neural network identifier is used to identify the physical parameters to be identified in the three-degree-of-freedom dynamic benchmark model of the ship, and the predicted physical parameters are obtained.

[0020] S5: Substitute the predicted physical parameters into the ship's three-degree-of-freedom dynamic benchmark model to construct a ship's dynamic positioning three-degree-of-freedom mathematical model.

[0021] By implementing steps S1 to S5 above, this application solves the problems of traditional methods relying on physical experiments and being unable to adapt to time-varying and nonlinear parameters, and improves the control accuracy and robustness of dynamic positioning systems. It is particularly suitable for dynamic positioning systems of offshore vessels such as engineering vessels, platform supply vessels, and offshore wind power installation vessels.

[0022] Furthermore, step S1 specifically includes: In the ship's coordinate system, a three-degree-of-freedom dynamic reference model of the ship is established, and the expression is as follows: in, The inertia matrix, The velocity in the ship's coordinate system. The acceleration in the ship's coordinate system. Here is the damping matrix. The control input torque generated by the thruster The disturbance torque of the environment (wind, waves, current).

[0023] , , , These represent the longitudinal velocity, lateral velocity, and bow roll rate in the ship's coordinate system, respectively.

[0024] Rewrite the model in linear regression form: In the formula: The regression matrix is ​​composed of measurable or estimable state variables. The physical parameters to be identified include those in the inertia matrix and the damping matrix.

[0025] Furthermore, step S2 specifically includes: The neural network identifier includes an input layer, a hidden layer, and an output layer; the input layer contains 12 nodes, and the hidden layer uses a Gaussian radial basis function as the activation function; the output layer contains 6 nodes, and the activation function of the output layer is a linear function.

[0026] More specifically, the neural network identifier constructed in this embodiment: (1) The neural network identifier outputs 6 nodes, and the target output is: ,in, To predict physical parameters, , , These are the estimated values ​​of inertial parameters in the longitudinal, lateral, and yaw directions, respectively. , , These are the estimated values ​​of the damping coefficient.

[0027] (2) Neural network identifiers learn without direct supervision signals, and the network output is trained through indirect learning. To predict thrust Approaching actual thrust .

[0028] (3) The neural network identifier has 12 nodes as input, containing the ship's current motion state and thrust history (used to capture the dynamic characteristics of the ship's motion). The input vector is: .in, Let be the horizontal position coordinate at time t. Let be the vertical position coordinate at time t. Let be the bow roll angle at time t. Let be the longitudinal velocity at time t. Let be the lateral velocity at time t. Let be the bow roll angular velocity at time t. Let be the longitudinal acceleration at time t. Let be the lateral acceleration at time t. Let be the bow roll acceleration at time t. The longitudinal thrust at time t-1 The lateral thrust at time t-1 Let t be the bow roll torque at time t-1.

[0029] (4) The output of the neural network identifier is bound to the ship dynamics model.

[0030] The k-th predicted physical parameter output by the neural network identifier is: in, For input samples The k-th predicted physical parameter, Let be the connection weight from the j-th node in the hidden layer to the k-th node in the output layer. Let be the Gaussian radial basis function of the j-th node in the hidden layer. Let be the center vector of the j-th node in the hidden layer. M is the bias of the k-th node in the output layer, and M is the number of hidden layer nodes.

[0031] Gaussian radial basis functions , Euclidean distance. is the width parameter of the j-th node in the hidden layer.

[0032] Will Vectorized representation: in, This is the output vector of the hidden layer. The weight matrix, This is the bias vector.

[0033] Network key mapping: The regression matrix is: In this embodiment, the training objective of the neural network identifier is to make the predicted thrust approximate the actual thrust. To this end, its total loss function includes two terms: prediction error term and physical constraint term.

[0034] (1) Prediction error term.

[0035] For a single sample, the squared error is , For input samples The predicted physical parameters.

[0036] (2) Physical constraints.

[0037] To ensure that the identified physical parameters have physical meaning, physical constraint terms are defined: in, For physical constraint terms, These are estimated values ​​for the inertial parameters. The first term is the estimated value of the damping coefficient. The second term is the penalty for negative inertia parameters. The damping parameter that penalizes negative values. This is the regularization coefficient.

[0038] (3) Total loss function.

[0039] For N samples, the total loss function is: in, For the total loss, For the prediction error term, The actual thrust of the i-th sample. Let be the regression matrix of the i-th sample. Let i be the predicted physical parameters corresponding to the i-th sample. For the physical constraint term of the i-th sample, The regularization coefficient is . The weight matrix, Describing the Frobenius norm, This is a weight decay term to prevent the neural network identifier from overfitting.

[0040] The gradient of the total loss function with respect to the weight matrix W and the bias vector b is calculated as follows: (1) Predict the gradient of the error term.

[0041] Let the error vector of a single sample be... ; Prediction error affects network output The gradient is ; According to the chain rule, the total loss function is... The gradient is: ; Vectorized representation: .

[0042] Similarly, the gradient with respect to the bias vector b is: ; Vectorized representation: .

[0043] (2) Gradient of physical constraint terms.

[0044] Define a function: The gradient vector is then : therefore: Vectorized representation: .

[0045] (3) Total gradient.

[0046] For the learning process of a neural network discriminator with N samples, the total gradient is: In this embodiment, the training of the neural network identifier includes hidden layer parameter learning and output layer parameter learning: the hidden layer parameter learning adopts an unsupervised learning method, which determines the center vector of the hidden layer nodes through K-Means clustering and determines the width parameter through adaptive width calculation; the output layer parameter learning adopts a supervised learning method, which updates the weight matrix and bias based on the error between the predicted thrust and the actual thrust.

[0047] (1) Hidden layer parameter learning (unsupervised learning).

[0048] Using input samples Perform K-Means clustering to determine cluster centers. (i.e., the center vector of the j-th node in the hidden layer), perform adaptive width parameter calculation: In the formula, This is the overlap coefficient (taken as 2 in this embodiment). This represents the number of sample sets belonging to the j-th cluster center, i.e., the total number of samples falling into the j-th cluster.

[0049] (2) Output weights and biases learning (supervised learning).

[0050] 1) Forward propagation.

[0051] Obtain the input sample at time t and actual thrust Calculate the hidden layer output: Calculate the network output: Calculate and predict thrust: 2) Error calculation: 2) Gradient calculation: The gradient of the weight matrix. The gradient of the bias vector.

[0052] (3) Improved parameter update with forgetting factor.

[0053] in, Let be the weight matrix at time t+1. Let be the weight matrix at time t. The learning rate at time t is non-linear. Forgetting factor, The gradient of the weight matrix. Let be the bias vector at time t+1. Let be the bias vector at time t. The gradient of the bias vector.

[0054] , It is the initial learning rate. It's about adjusting the sensitivity. It is to prevent division by zero (value) ), This is the forgetting factor (balancing the current gradient with historical weights).

[0055] After the weights are updated, the output parameters of the neural network need to be subject to hard constraints, namely: in, Let be the minimum physical parameter value of the i-th sample. This represents the maximum physical parameter value for the i-th sample.

[0056] Furthermore, step S4 specifically includes: The output of the neural network identifier A real-time three-degree-of-freedom ship dynamics benchmark model is used to form a three-degree-of-freedom mathematical model for ship dynamic positioning: This model can be directly used in the design of ship dynamic positioning control systems, such as traditional PID controllers.

[0057] In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments. The computer device can be a server or a terminal. The computer device includes a processor, a memory, an input / output interface (I / O), and a communication interface. The processor, memory, and I / O interface are connected via a system bus, and the communication interface is connected to the system bus via the I / O interface. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device stores data to be processed. The I / O interface of the computer device is used for exchanging information between the processor and external devices. The communication interface of the computer device is used for communicating with an external terminal via a network connection. When the computer program is executed by the processor, it implements the steps in the above-described method embodiments.

[0058] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0059] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0060] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0061] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0062] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchain. The processors involved in the embodiments provided in this application may be, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc.

[0063] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0064] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for constructing a three-degree-of-freedom mathematical model for ship dynamic positioning, characterized in that, include: Establish a three-degree-of-freedom dynamic benchmark model for a ship; Construct a neural network identifier; The neural network identifier is trained using the ship's motion state and thrust history as inputs and a total loss function; the total loss function includes a prediction error term and a physical constraint term. A trained neural network identifier is used to identify the physical parameters to be identified in the three-degree-of-freedom dynamic benchmark model of the ship, and the predicted physical parameters are obtained. Substitute the predicted physical parameters into the ship's three-degree-of-freedom dynamic benchmark model to construct a ship's dynamic positioning three-degree-of-freedom mathematical model.

2. The method for constructing a three-degree-of-freedom mathematical model for ship dynamic positioning according to claim 1, characterized in that, The expression for the three-degree-of-freedom dynamics benchmark model of the ship is: in, The inertia matrix, The velocity in the ship's coordinate system. The acceleration in the ship's coordinate system. Here is the damping matrix. The control input torque generated by the thruster This refers to the environmental disturbance torque.

3. The method for constructing a three-degree-of-freedom mathematical model for ship dynamic positioning according to claim 1, characterized in that, The neural network identifier includes an input layer, a hidden layer, and an output layer; the input layer contains 12 nodes, and the hidden layer uses a Gaussian radial basis function as the activation function; the output layer contains 6 nodes, and the activation function of the output layer is a linear function.

4. The method for constructing a three-degree-of-freedom mathematical model for ship dynamic positioning according to claim 1, characterized in that, The expression for the total loss function is: in, For the total loss, For the prediction error term, The actual thrust of the i-th sample. Let be the regression matrix of the i-th sample. Let i be the predicted physical parameters corresponding to the i-th sample. For the physical constraint term of the i-th sample, The regularization coefficient is . The weight matrix, This represents the Frobenius norm.

5. The method for constructing a three-degree-of-freedom mathematical model for ship dynamic positioning according to claim 1, characterized in that, The training of the neural network identifier includes hidden layer parameter learning and output layer parameter learning: The hidden layer parameter learning adopts an unsupervised learning method, which determines the center vector of the hidden layer node through K-Means clustering and determines the width parameter through adaptive width calculation. The output layer parameter learning adopts a supervised learning method, and the weight matrix and bias vector are updated based on the error between the predicted thrust and the actual thrust.

6. The method for constructing a three-degree-of-freedom mathematical model for ship dynamic positioning according to claim 5, characterized in that, The formula for calculating the width parameter is: in, Let be the width parameter of the j-th node in the hidden layer. Let be the center vector of the j-th node in the hidden layer, corresponding to the j-th cluster center. The overlap coefficient, For the input sample, Let be the set of samples belonging to the j-th cluster center.

7. The method for constructing a three-degree-of-freedom mathematical model for ship dynamic positioning according to claim 5, characterized in that, The update formulas for the weight matrix and bias vector are as follows: in, Let be the weight matrix at time t+1. Let be the weight matrix at time t. The learning rate at time t is non-linear. Forgetting factor, The gradient of the weight matrix. Let be the bias vector at time t+1. Let be the bias vector at time t. The gradient of the bias vector.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the method for constructing a three-degree-of-freedom mathematical model of ship dynamic positioning as described in any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the method for constructing a three-degree-of-freedom mathematical model of ship dynamic positioning as described in any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the method for constructing a three-degree-of-freedom mathematical model of ship dynamic positioning as described in any one of claims 1-7.

Images