Full-space magnetic field multi-objective evolutionary prediction method and system for distributed degaussing system

Abstract

The application discloses a kind of distributed degaussing system full-space magnetic field multi-objective evolution prediction method and system, it is related to high altitude magnetic field prediction field, distributed degaussing system full-space magnetic field multi-objective evolution prediction method mainly includes: obtaining target ship structure parameters and near-field measurement data, obtains ellipsoid combination model according to target ship structure parameters and near-field measurement data;Optimize ellipsoid combination model using multi-objective evolution algorithm based on decomposition, obtain optimized ellipsoid combination model;The high altitude magnetic field of ship is predicted using the optimized ellipsoid combination model, and the prediction result is obtained.The distributed degaussing system full-space magnetic field multi-objective evolution prediction method and system provided by the application can improve the accuracy and adaptability of high altitude magnetic field prediction.

Description

Technical Field

[0001] The present invention relates to the field of high-altitude magnetic field prediction, and more specifically, to a method and system for multi-objective evolutionary prediction of full-space magnetic fields in a distributed degaussing system. Background Art

[0002] High-altitude magnetic field prediction is a key technology for magnetic field control. Existing ship-borne high-altitude magnetic field prediction technology still has certain limitations. For one thing, prediction accuracy needs to be improved. Traditional prediction methods are mostly based on simplified model assumptions, such as approximating the ship as a simple magnetic dipole, ellipsoid, or a regular magnetic charge surface distribution. For example, the traditional ellipsoidal ship magnetic field modeling method, while somewhat simplifying the complex ship structure modeling process, treats the ship as a regular ellipsoid and estimates the magnetic field distribution based on parameters such as its major axis, minor axis, and magnetization intensity. Real ships contain a vast array of diverse steel components, from heavy keels to sophisticated onboard equipment. Their individual magnetic characteristics and mutual magnetic coupling effects are extremely complex, making traditional models difficult to accurately simulate. Furthermore, traditional models have limited adaptability. Introducing a new parameter in mathematical calculations often triggers a series of chain reactions, requiring not only a reconsideration of the complex relationship between the new parameter and the existing parameters but also a reconstruction of the appropriate computational process. This undoubtedly greatly increases the complexity and workload of calculations, causing traditional calculation methods to often run into difficulties when faced with massive data and complex models, and reducing the adaptability of forecasting algorithms.

[0003] Current traditional algorithms used in ship magnetic field prediction include: Least squares parameter fitting: When simulating ship magnetic fields based on traditional models (such as magnetic dipole and ellipsoid models), the least squares method is often used to fit parameters. However, this method is sensitive to outliers and assumes an error distribution that doesn't match reality, which can easily lead to inaccurate parameter estimation and reduced model accuracy.

[0004] Fourier transform spectrum analysis: Some studies use Fourier transform to analyze ship magnetic field measurement data. This algorithm has certain effects on periodic magnetic field signals, but it has difficulty in processing non-periodic and nonlinear changes, and its accuracy is insufficient in high-altitude magnetic field forecasts.

[0005] Simple iterative algorithms for solving equations: Simple iterative algorithms are often used in forward models based on the integral equation of the simplified scalar magnetic potential and in ship magnetic field prediction methods based on the magnetic charge surface distribution. These algorithms have slow convergence and high computational complexity, making them difficult to meet real-time requirements for dynamic ship changes. They are also prone to falling into local optimal solutions, resulting in inaccurate magnetic field predictions.

[0006] Data processing based on fixed-rule algorithms: Early ship magnetic field data processing often used algorithms based on fixed rules, such as fixed-threshold filtering. However, in the complex and changing real-world navigation environment, these algorithms cannot adapt themselves and are prone to misjudging noise, affecting data processing performance and magnetic field prediction accuracy.

[0007] In summary, the existing ship high-altitude magnetic field prediction algorithm has shortcomings, and the ship degaussing technology needs to be improved in aspects such as the decoupling control of multiple degaussing power supply systems. Summary of the Invention

[0008] The purpose of the present invention is to provide a method and system for multi-objective evolutionary prediction of full-space magnetic field in a distributed degaussing system, which can improve the accuracy and adaptability of high-altitude magnetic field prediction.

[0009] The present invention provides a multi-objective evolutionary prediction method for the full-space magnetic field of a distributed degaussing system, comprising the following steps: S1: acquiring target ship structural parameters and near-field measurement data, and obtaining an ellipsoid combination model based on the target ship structural parameters and the near-field measurement data; S2: optimizing the ellipsoid combination model using a decomposition-based multi-objective evolutionary algorithm to obtain an optimized ellipsoid combination model; S3: predicting the ship's high-altitude magnetic field using the optimized ellipsoid combination model to obtain a prediction result.

[0010] Furthermore, step S2 specifically includes: S21: obtaining an initial population based on the ellipsoid combination model; S22: obtaining a weight vector of a preset weight number based on the ellipsoid combination model using a uniform distribution method; S23: constructing a neighborhood structure based on the weight vector; S24: performing iterative optimization based on the initial population, the weight vector and the neighborhood structure to obtain an updated population; S25: confirming that the number of iterations has reached the maximum number of iterations, terminating the iterative optimization, and obtaining the optimized ellipsoid combination model based on the current updated population.

[0011] Furthermore, the above-mentioned initial population is randomly generated and includes a preset number of individuals. Each individual represents a set of parameters based on the ellipsoid combination model, and the parameters include the number of ellipsoids, the major axis, the minor axis, and the position coordinates of each ellipsoid.

[0012] Furthermore, the above weight vector is used to construct a single-objective subproblem, which includes a prediction accuracy subproblem, a computational complexity subproblem, and a model stability subproblem.

[0013] Furthermore, step S24 specifically includes: S241: according to the initial population and the weight vector, using the tournament selection method to obtain the parent individual; S242: according to the parent individual, using the simulated binary crossover method to perform a crossover operation to obtain the offspring individual; S243: according to the offspring individual, using the polynomial mutation method to perform a mutation operation to obtain a new offspring individual; S244: according to the new offspring individual, obtaining a fitness value; S245: according to the fitness value and the neighborhood structure, updating the individuals in the neighborhood to obtain an updated population.

[0014] The present invention also provides a distributed degaussing system for predicting the full-space magnetic field using a multi-objective evolutionary prediction system. The system includes the following modules: an ellipsoid combination model construction module configured to obtain target ship structural parameters and near-field measurement data, and to obtain an ellipsoid combination model based on the target ship structural parameters and near-field measurement data; an ellipsoid combination model optimization module configured to optimize the ellipsoid combination model using a decomposition-based multi-objective evolutionary algorithm to obtain an optimized ellipsoid combination model; and a ship high-altitude magnetic field prediction module configured to predict the ship's high-altitude magnetic field using the optimized ellipsoid combination model to obtain a prediction result.

[0015] Furthermore, the above-mentioned ellipsoid combination model optimization module is specifically configured as follows: according to the ellipsoid combination model, an initial population is obtained; according to the ellipsoid combination model, a weight vector with a preset weight number is obtained using a uniform distribution method; according to the weight vector, a neighborhood structure is constructed; according to the initial population, the weight vector and the neighborhood structure, iterative optimization is performed to obtain an updated population; it is confirmed that the number of iterations reaches the maximum number of iterations, the iterative optimization is terminated, and the optimized ellipsoid combination model is obtained according to the current updated population.

[0016] Furthermore, the above-mentioned initial population is randomly generated and includes a preset number of individuals. Each individual represents a set of parameters based on the ellipsoid combination model, and the parameters include the number of ellipsoids, the major axis, the minor axis, and the position coordinates of each ellipsoid.

[0017] Furthermore, the above weight vector is used to construct a single-objective subproblem, which includes a prediction accuracy subproblem, a computational complexity subproblem, and a model stability subproblem.

[0018] Furthermore, the above-mentioned iterative optimization is performed based on the initial population, the weight vector and the neighborhood structure to obtain an updated population, specifically including: obtaining parent individuals based on the initial population and the weight vector using a tournament selection method; performing a crossover operation based on the parent individuals using a simulated binary crossover method to obtain offspring individuals; performing a mutation operation based on the offspring individuals using a polynomial mutation method to obtain new offspring individuals; obtaining a fitness value based on the new offspring individuals; and updating individuals in the neighborhood based on the fitness value and the neighborhood structure to obtain an updated population.

[0019] The implementation of the distributed degaussing system full-space magnetic field multi-objective evolution prediction method and system provided by the present invention has the following beneficial effects:

[0020] The present invention simulates the target ship structure and performs near-field measurements, and constructs an ellipsoid combination model based on the target ship structure parameters and near-field measurement data; uses a decomposition-based multi-objective evolutionary algorithm to optimize the ellipsoid combination model to obtain an optimized ellipsoid combination model; uses the optimized ellipsoid combination model to predict the high-altitude magnetic field of the ship to obtain a prediction result; the present invention solves the current problem of difficulty in optimizing multiple parameters in the process of establishing a high-precision magnetic field model, improves the accuracy and adaptability of high-altitude magnetic field prediction, and provides new ideas for improving prediction methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0022] Figure 1 This is a flow chart of the multi-objective evolutionary prediction method for the full-space magnetic field of a distributed degaussing system provided by the present invention;

[0023] Figure 2 Schematic diagram of the magnetic field measurement device for ship model laboratory provided by the present invention;

[0024] Figure 3 Schematic diagram of the arrangement of sensors for measuring the near-field magnetic field of a ship model provided by the present invention;

[0025] Figure 4 Schematic diagram of the arrangement of sensors for measuring magnetic field at high altitude on a ship model provided by the present invention;

[0026] Figure 5 Flowchart of the multi-objective evolutionary prediction method for the full-space magnetic field of a distributed degaussing system provided by the present invention. DETAILED DESCRIPTION

[0027] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.

[0028] Figure 1 A schematic diagram of a distributed degaussing system full-space magnetic field multi-objective evolutionary prediction method of this embodiment is shown. In this embodiment, the distributed degaussing system full-space magnetic field multi-objective evolutionary prediction method includes the following steps:

[0029] S1: Obtain target ship structural parameters and near-field measurement data, and obtain an ellipsoidal combined model based on the target ship structural parameters and near-field measurement data;

[0030] In an exemplary embodiment, the ellipsoid combination model includes ellipsoid number and magnetic moment combination parameters;

[0031] As an exemplary embodiment, in step S1, a combination of multiple rotating ellipsoids is used to simulate the magnetic field distribution of the ship. According to the actual structure and magnetic field characteristics of the ship, the ship is divided into multiple regions, and each region is approximated by one or more rotating ellipsoids. By reasonably adjusting the parameters of each rotating ellipsoid, such as the major axis and minor axis, which determine the shape and size of the ellipsoid, and the position coordinates determine its position in space, the combined model can more accurately fit the complex magnetic field distribution of the ship. The initial parameters of each rotating ellipsoid are determined by analyzing and calculating near-field measurement data.

[0032] In an exemplary embodiment, a 1:20 magnetic scaled ship model is selected, and the length L and width B of the ship model are recorded; Figure 2 and Figure 3 A circle of high-precision three-component magnetic field sensors is evenly spaced around the center of the ship model. Five sensors are arranged on each side with a distance of 0.5B. The ship model is dragged on the track from -L to L from the sensor arrangement position. During the dragging process, sensor data is collected every L / 20 distance. Repeat four groups to measure the three-directional induced magnetic field data and the fixed magnetic field data for near-field data modeling. After the measurement, the sensors are removed, as shown in the following example. Figure 4 , a row of three-component sensors are transferred to a position 20B above the center of the ship model and placed at equal intervals, so that the ship model is dragged on the track from -2L to 2L away from the sensor arrangement position. During the dragging process, sensor data is collected once every L / 10 distance. Four groups of three-directional induced magnetic field data and fixed magnetic field data are measured respectively for use in high-altitude data algorithm verification; the X component, Y component and Z component are used as control targets respectively to form a multi-target form, thereby constructing the objective function; first set the initial parameters, and use the rotating ellipsoid (magnetic dipole, surface magnetic charge, boundary element and other methods) to simulate the magnetic field calculation formula to calculate the initial magnetic moment combination of each ellipsoid; use the initial magnetic moment to calculate the magnetic field value of the simulation body and compare it with the imported ship model near-field measurement value;

[0033] S2: Optimize the ellipsoid combination model using a decomposition-based multi-objective evolutionary algorithm to obtain an optimized ellipsoid combination model;

[0034] As an exemplary embodiment, in step S2, the parameters are optimized by means of the MOEA / D algorithm to minimize the error between the model calculated magnetic field and the actual measured magnetic field, and to improve the accuracy of the model; a multi-objective evolutionary algorithm based on decomposition (MOEA / D) is introduced to optimize the ellipsoid combination model; multiple optimization objectives are set, including minimizing the error between the model calculated magnetic field and the near-field measured magnetic field to improve the prediction accuracy; minimizing the calculation complexity of the model to ensure fast operation in actual application; at the same time, the stability of the model is considered to ensure reliable performance of the model under different measurement data and environmental conditions; these objectives are interrelated and mutually constrained, and the MOEA / D algorithm decomposes the multi-objective problem into multiple single-objective sub-problems, and uses a weight vector to balance the importance of different objectives;

[0035] In an exemplary embodiment, step S2 specifically includes:

[0036] S21: obtaining an initial population according to the ellipsoid combination model;

[0037] In an exemplary embodiment, the initial population is randomly generated, and the initial population includes a preset number of individuals, each individual representing a set of parameters of the ellipsoid combination model, the parameters including the number of ellipsoids, the long axis, the short axis, and the position coordinates of each ellipsoid;

[0038] As an exemplary embodiment, the preset number of individuals is 10;

[0039] As an exemplary embodiment, in step S21, the population is first initialized, and each individual represents a set of parameters of the ellipsoid combination model; that is, an initial population of 120 individuals is randomly generated, and each individual represents a set of parameters of the ellipsoid combination model, the parameters including the number of rotating ellipsoids, the long axis, the short axis, and the position coordinates;

[0040] S22: obtaining a preset number of weight vectors by using a uniform distribution method according to the ellipsoid combination model;

[0041] In an exemplary embodiment, the weight vectors are used to construct single-objective sub-problems, and the single-objective sub-problems include a prediction accuracy sub-problem, a calculation complexity sub-problem, and a model stability sub-problem;

[0042] As an exemplary embodiment, the preset number of weights is 3;

[0043] As an exemplary embodiment, the single-objective sub-problems further include objective functions;

[0044] As an exemplary embodiment, in step S22, six different weight vectors are generated in a uniformly distributed manner. These weight vectors are used to construct a single-objective subproblem to balance the importance of different objectives (such as prediction accuracy, computational complexity, and model stability) in the optimization process.

[0045] S23: constructing a neighborhood structure based on the weight vector;

[0046] As an exemplary embodiment, in step S23, the neighborhood range of each individual is determined based on the distance between the weight vectors, and a neighborhood structure is constructed to prepare for subsequent information exchange between individuals;

[0047] S24: Perform iterative optimization based on the initial population, weight vector and neighborhood structure to obtain an updated population;

[0048] As an exemplary embodiment, in step S24, new individuals are generated through genetic operations such as crossover and mutation, and the parameter space is continuously explored. In each iteration, individuals are evaluated and selected based on the objective function values ​​of each sub-problem. At the same time, the neighborhood structure is utilized to enable information interaction and sharing between individuals, accelerating the convergence of the algorithm. After multiple iterations, the MOEA / D algorithm can find a set of model parameters that achieves a good balance between multiple objectives, thereby optimizing the ellipsoid combination model, enabling it to more accurately predict the ship's high-altitude magnetic field.

[0049] In an exemplary embodiment, step S24 specifically includes:

[0050] S241: Based on the initial population and weight vector, the parent individuals are obtained using the tournament selection method;

[0051] As an exemplary embodiment, in step S241, based on the objective function of each subproblem, several individuals are selected from the current population as parent individuals through methods such as tournament selection for subsequent crossover and mutation operations;

[0052] S242: performing a crossover operation on the parent individuals using a simulated binary crossover method to obtain offspring individuals;

[0053] As an exemplary embodiment, in step S242, a crossover operation is performed on the selected parent individuals using a method such as simulated binary crossover (SBX) with a crossover probability of 0.6 to generate offspring individuals. The crossover operation enables the offspring individuals to incorporate the excellent characteristics of the parent individuals, thereby increasing the diversity of the population.

[0054] S243: performing a mutation operation on the offspring individuals using a polynomial mutation method to obtain new offspring individuals;

[0055] As an exemplary embodiment, in step S243, the offspring individuals are mutated by using a polynomial mutation method with a mutation probability of 0.15; the mutation operation can introduce new genes into the population and avoid the algorithm from falling into a local optimal solution;

[0056] S244: Obtain the fitness value according to the new offspring individuals;

[0057] As an exemplary embodiment, in step S244, the newly generated offspring individuals are substituted into the ship high-altitude magnetic field prediction model based on the ellipsoid combination to calculate the fitness value under each sub-problem objective function, and the advantages and disadvantages of the offspring individuals are evaluated; the objective function includes the error between the calculated magnetic field and the measured magnetic field, the calculation complexity, and the model stability, etc.

[0058] S245: Update the individuals in the neighborhood according to the fitness value and the neighborhood structure to obtain an updated population;

[0059] As an exemplary embodiment, in step S245, the individuals in the neighborhood are updated according to the fitness value of the offspring individuals; if the offspring individuals perform better in a certain sub-problem, the offspring individuals replace the corresponding individuals in the neighborhood, realizing information sharing and cooperative evolution in the neighborhood;

[0060] S25: Confirm that the number of iterations reaches the maximum number of iterations, terminate the iterative optimization, and obtain the optimized ellipsoid combination model according to the current updated population;

[0061] As an exemplary embodiment, in step S25, it is checked whether the maximum number of iterations (set to 200) is reached; if so, the algorithm is terminated; if not, the step of "selecting operation" is returned to continue the iterative optimization;

[0062] S3: Use the optimized ellipsoid combination model to predict the ship high-altitude magnetic field to obtain a prediction result;

[0063] As an exemplary embodiment, in step S3, the number of ellipsoids and the magnetic moment combination parameters are obtained according to the optimization result, so that the magnetic field parameters on any height plane can be predicted;

[0064] As another exemplary embodiment, in step S3, the ellipsoid combination model optimized by the MOEA / D algorithm is used to predict the high-altitude magnetic field of the ship using the optimized parameters; the prediction results are compared and verified with the actual high-altitude magnetic field measurement data; a variety of evaluation indicators, such as the root mean square error (RMSE) and the mean absolute error (MAE), are used to measure the accuracy of the prediction; the performance of the model is evaluated and analyzed through a large amount of experimental data and actual application cases; if the prediction error exceeds the set threshold, the model parameters are further adjusted or the optimization algorithm is optimized to continuously improve the accuracy and reliability of the magnetic field prediction.

[0065] This embodiment provides a distributed degaussing system full-space magnetic field multi-objective evolutionary prediction system. The system includes the following modules: an ellipsoid combination model construction module, configured to obtain target ship structural parameters and near-field measurement data, and then generate an ellipsoid combination model based on the target ship structural parameters and near-field measurement data; an ellipsoid combination model optimization module, configured to optimize the ellipsoid combination model using a decomposition-based multi-objective evolutionary algorithm to obtain an optimized ellipsoid combination model; and a ship high-altitude magnetic field prediction module, configured to use the optimized ellipsoid combination model to predict the ship's high-altitude magnetic field and obtain a prediction result.

[0066] In an exemplary embodiment, the above-mentioned ellipsoid combination model optimization module is specifically configured as follows: according to the ellipsoid combination model, an initial population is obtained; according to the ellipsoid combination model, a weight vector with a preset weight number is obtained using a uniform distribution method; according to the weight vector, a neighborhood structure is constructed; according to the initial population, the weight vector and the neighborhood structure, iterative optimization is performed to obtain an updated population; it is confirmed that the number of iterations reaches the maximum number of iterations, the iterative optimization is terminated, and the optimized ellipsoid combination model is obtained according to the current updated population.

[0067] In an exemplary embodiment, the above-mentioned initial population is randomly generated, and the initial population includes a preset number of individuals, each individual represents a set of parameters based on the ellipsoid combination model, and the parameters include the number of ellipsoids, the major axis, the minor axis, and the position coordinates of each ellipsoid.

[0068] In an exemplary embodiment, the weight vector is used to construct a single-objective subproblem, which includes a prediction accuracy subproblem, a computational complexity subproblem, and a model stability subproblem.

[0069] In an exemplary embodiment, the above-mentioned iterative optimization is performed based on the initial population, weight vector and neighborhood structure to obtain an updated population, specifically including: obtaining parent individuals based on the initial population and weight vector using a tournament selection method; performing a crossover operation based on the parent individuals using a simulated binary crossover method to obtain offspring individuals; performing a mutation operation based on the offspring individuals using a polynomial mutation method to obtain new offspring individuals; obtaining a fitness value based on the new offspring individuals; and updating individuals in the neighborhood based on the fitness value and the neighborhood structure to obtain an updated population.

[0070] In some embodiments, the above-mentioned multi-objective evolutionary prediction method for the full-space magnetic field of the distributed degaussing system can also be implemented in the following manner. In this embodiment, the multi-objective evolutionary prediction method for the full-space magnetic field of the distributed degaussing system abandons the limitations of traditional single-objective optimization and incorporates multiple interrelated and mutually constrained objectives such as prediction accuracy, computational efficiency, and model stability into a unified optimization framework; through the MOEA / D algorithm, the complex multi-objective problem is cleverly decomposed into a series of single-objective sub-problems, and the weight vector is used to coordinate the priorities of different objectives;

[0071] When constructing a magnetic field prediction model based on a combination of rotating ellipsoids, the MOEA / D algorithm simultaneously optimizes the parameters of the rotating ellipsoids (such as number, major axis, minor axis, position coordinates, etc.) and the parameters required for the model to work together with other technologies (such as data processing flow, neural network, etc., if combined) to achieve collaborative optimization of multiple goals and improve the overall performance of the model; for example, in the optimization process, the relationship between the increase in calculation amount caused by improving forecast accuracy and ensuring calculation efficiency is balanced to find the best parameter combination so that the model has efficient computing power and good stability while meeting the accuracy requirements; using multiple rotating ellipsoid combinations To simulate the magnetic field distribution of the ship; according to the actual structure and magnetic field characteristics of the ship, the ship is divided into multiple regions, and each region is approximated by one or more rotating ellipsoids. By reasonably adjusting the parameters of each rotating ellipsoid, such as the major axis and minor axis, the shape and size of the ellipsoid are determined, and the position coordinates determine its position in space, so that the combined model can more accurately fit the complex magnetic field distribution of the ship; the initial parameters of each rotating ellipsoid are determined by near-field measurement data analysis and calculation; then, these parameters are optimized with the help of the MOEA / D algorithm to minimize the error between the model-calculated magnetic field and the actual measured magnetic field, thereby improving the accuracy of the model;

[0072] A decomposition-based multi-objective evolutionary algorithm (MOEA / D) is introduced to optimize the ellipsoid combination model. Multiple optimization objectives are set, including minimizing the error between the calculated magnetic field of the model and the measured near-field magnetic field to improve the prediction accuracy, minimizing the computational complexity of the model to ensure fast operation in practical applications, and considering the stability of the model to maintain reliable performance under different measurement data and environmental conditions. These objectives are interrelated and mutually constrained. The MOEA / D algorithm decomposes the multi-objective problem into multiple single-objective sub-problems and balances the importance of different objectives using a weight vector.

[0073] In the MOEA / D algorithm, the population is first initialized, with each individual representing a set of parameters for the ellipsoid combination model. New individuals are generated through genetic operations such as crossover and mutation to continuously explore the parameter space. In each iteration, the individuals are evaluated and selected based on the objective function values of each sub-problem. At the same time, the neighborhood structure is used to enable information exchange and sharing between individuals, accelerating the convergence speed of the algorithm. After multiple iterations, the MOEA / D algorithm can find a set of model parameters that achieve a good balance between multiple objectives, thereby optimizing the ellipsoid combination model to more accurately predict the ship's high-altitude magnetic field.

[0074] After optimization by the MOEA / D algorithm, the ellipsoid combination model uses the optimized parameters to predict the ship's high-altitude magnetic field. The prediction results are compared and verified with actual high-altitude magnetic field measurement data. Various evaluation indicators such as root mean square error (RMSE) and mean absolute error (MAE) are used to measure the accuracy of the prediction. Through a large amount of experimental data and actual application cases, the performance of the model is evaluated and analyzed. If the prediction error exceeds the set threshold, the model parameters or optimization algorithm are further adjusted to continuously improve the accuracy and reliability of the magnetic field prediction.

[0075] In some embodiments, the distributed degaussing system full-space magnetic field multi-objective evolutionary prediction method described above can also be implemented in the following manner. This embodiment takes a 1:20 magnetic scale ship model as the test object to verify the ship model high-altitude magnetic field prediction algorithm. In this embodiment, the distributed degaussing system full-space magnetic field multi-objective evolutionary prediction method includes:

[0076] Select a place far from large electromagnetic interference sources and with stable geomagnetic environment as the experimental site. A dedicated measurement platform is built in the laboratory for placing ship models and magnetic field measurement equipment.

[0077] Select a 1:20 magnetic scale ship model, record the ship model length L and width B. For example, Figure 2 and Figure 3A circle of high-precision three-component magnetic field sensors is evenly spaced around the center of the ship model. Five sensors are arranged on each side with a distance of 0.5B. The ship model is dragged on the track from -L to L from the sensor arrangement position. During the dragging process, sensor data is collected every L / 20 distance. Repeat four groups to measure the three-directional induced magnetic field data and the fixed magnetic field data for near-field data modeling. After the measurement, the sensors are removed, as shown in the following example. Figure 4 , a row of three-component sensors are transferred to a position 20B above the center of the ship model and placed at equal intervals. The ship model is dragged on the track from -2L to 2L away from the sensor arrangement position. During the dragging process, sensor data is collected once every L / 10 distance. Four sets of measurements are repeated to measure the three-directional induced magnetic field data and the fixed magnetic field data for high-altitude data verification. The X component, Y component, and Z component are respectively used as control targets to form a multi-target form, thereby constructing the objective function.

[0078] First, set the initial parameters and use the rotating ellipsoid (magnetic dipole, surface magnetic charge, boundary element method, etc.) to simulate the magnetic field calculation formula to calculate the initial magnetic moment combination of each ellipsoid;

[0079] The initial magnetic moment is used to calculate the magnetic field value of the simulation body and compared with the imported near-field measurement value of the ship model;

[0080] The MOEA / D algorithm was introduced to optimize the model parameters. The initial population size was set to 120 individuals, and the weight vectors were generated using a uniform distribution, resulting in a total of six different weight vectors. The crossover probability was set to 0.6, the mutation probability was set to 0.15, and the maximum number of iterations was set to 200. In each iteration, new individuals were generated through selection, crossover, and mutation operations, and individuals were evaluated and selected based on the set objective function (such as minimizing the error between the simulated and measured magnetic fields and reducing model complexity). The neighborhood structure was used to enable information interaction and sharing between individuals, accelerating the convergence of the algorithm. After 200 iterations, a set of optimized rotation ellipsoid parameters was obtained.

[0081] The optimized ellipsoid parameters are applied to the ship model high-altitude magnetic field prediction model. Based on the different position information of the ship model at a distance of -2L to 2L from the sensor layout, the high-altitude magnetic field prediction value at the corresponding position is calculated using the magnetic field simulation calculation formula. The simulated magnetic field value is compared with the near-field measurement value of the ship model to calculate the error between the two. The error calculation can use indicators such as root mean square error (RMSE) or mean absolute error (MAE) to measure the degree of deviation between the simulation results and the actual measurement. For example, the formula for calculating the root mean square error is:

[0082] ,

[0083] in, is the actual measured value, is the simulated calculated value, n is the number of measured data points;

[0084] Based on the error calculation results, the objective function of the MOEA / D algorithm was defined. The calculated RMSE was 25nT and the MAE was 12nT. Compared with the traditional modeling method based on single-objective optimization, the RMSE was reduced by 50% and the MAE was reduced by 40%, indicating that the technical solution of the present invention has higher accuracy and better performance in predicting the high-altitude magnetic field of ship models.

[0085] Figure 5 Flowchart showing a decomposed multi-objective evolutionary algorithm MOEA / D; the decomposed multi-objective evolutionary algorithm includes the following steps:

[0086] Step 1: Randomly generate an initial population of 120 individuals. Each individual represents a set of parameters based on the ellipsoid combination model, including the number of rotating ellipsoids, major axis, minor axis, position coordinates, etc.

[0087] Step 2: Generate six different weight vectors using a uniform distribution. These weight vectors are used to construct single-objective subproblems to balance the importance of different objectives (such as prediction accuracy, computational complexity, and model stability) in the optimization process.

[0088] Step 3: Determine the neighborhood range of each individual based on the distance between weight vectors and construct a neighborhood structure to prepare for subsequent information exchange between individuals;

[0089] Step 4: Based on the objective function of each subproblem, select several individuals from the current population as parent individuals through methods such as tournament selection for subsequent crossover and mutation operations.

[0090] For the selected parent individuals, a crossover operation is performed using methods such as simulated binary crossover (SBX) with a crossover probability of 0.6 to generate new offspring individuals. The crossover operation enables the offspring individuals to integrate the excellent characteristics of the parent individuals and increase the diversity of the population.

[0091] Mutation operation: The offspring individuals are mutated using methods such as polynomial mutation with a mutation probability of 0.15. Mutation operations can introduce new genes into the population and prevent the algorithm from falling into a local optimal solution. The newly generated offspring individuals are substituted into the ship high-altitude magnetic field prediction model based on ellipsoid combination, and their fitness values ​​under the objective functions of each sub-problem are calculated to evaluate the quality of the offspring individuals. The objective function includes the error between the model-calculated and measured magnetic fields, computational complexity, and model stability. Based on the fitness value of the offspring individual, the individuals in its neighborhood are updated. If the offspring individual performs better on a certain sub-problem, the corresponding individual in the neighborhood is replaced with the offspring individual to achieve information sharing and collaborative evolution within the neighborhood.

[0092] Step 5: Check whether the maximum number of iterations has been reached (set to 200); if so, terminate the algorithm; if not, return to the "Select Operation" step and continue iterative optimization;

[0093] According to the optimization results, the number of ellipsoids and their magnetic moment combination parameters are obtained, so that the magnetic field parameters on any height plane can be predicted.

[0094] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are protected by the present invention.

Claims

1. A multi-objective evolutionary prediction method for the full-space magnetic field of a distributed degaussing system, characterized by: The following steps are involved: S1: Acquire target ship structural parameters and near-field measurement data, and obtain an ellipsoidal combination model based on the target ship structural parameters and the near-field measurement data; The ellipsoid combination model uses a combination of multiple rotating ellipsoids to simulate the magnetic field distribution of the ship. According to the actual structure and magnetic field characteristics of the ship, the ship is divided into multiple regions, and each region is approximated by one or more rotating ellipsoids. By reasonably adjusting the parameters of each rotating ellipsoid, the major axis and minor axis determine the shape and size of the ellipsoid, and the position coordinates determine its position in space, so that the combined model can more accurately fit the complex magnetic field distribution of the ship. S2: optimizing the ellipsoid combination model using a decomposition-based multi-objective evolutionary algorithm to obtain an optimized ellipsoid combination model; S3: Use the optimized ellipsoid combination model to predict the high-altitude magnetic field of the ship and obtain the prediction results.

2. The multi-objective evolutionary prediction method for the full-space magnetic field of a distributed degaussing system according to claim 1 is characterized in that: Step S2 specifically includes: S21: obtaining an initial population according to the ellipsoid combination model; S22: obtaining a weight vector with a preset weight quantity using a uniform distribution method according to the ellipsoid combination model; S23: constructing a neighborhood structure according to the weight vector; S24: performing iterative optimization according to the initial population, the weight vector, and the neighborhood structure to obtain an updated population; S25: Confirm that the number of iterations reaches the maximum number of iterations, terminate the iterative optimization, and obtain the optimized ellipsoid combination model based on the current updated population.

3. The multi-objective evolutionary prediction method for the full-space magnetic field of a distributed degaussing system according to claim 2 is characterized in that: The initial population is randomly generated and includes a preset number of individuals. Each individual represents a set of parameters based on the ellipsoid combination model, and the parameters include the number of ellipsoids, the major axis, the minor axis, and the position coordinates of each ellipsoid.

4. The multi-objective evolutionary prediction method for the full-space magnetic field of a distributed degaussing system according to claim 2 is characterized in that: The weight vector is used to construct a single-objective subproblem, which includes a prediction accuracy subproblem, a computational complexity subproblem, and a model stability subproblem.

5. The multi-objective evolutionary prediction method for the full-space magnetic field of a distributed degaussing system according to claim 2 is characterized in that: Step S24 specifically includes: S241: Obtaining a parent individual using a tournament selection method according to the initial population and the weight vector; S242: performing a crossover operation on the parent individuals using a simulated binary crossover method to obtain offspring individuals; S243: performing a mutation operation on the offspring individual using a polynomial mutation method to obtain a new offspring individual; S244: Obtaining a fitness value according to the new offspring individual; S245: According to the fitness value and the neighborhood structure, the individuals in the neighborhood are updated to obtain an updated population.

6. A distributed degaussing system full-space magnetic field multi-objective evolution prediction system, characterized by: The system includes the following modules: The ellipsoid combination model construction module is configured to: obtain target ship structural parameters and near-field measurement data, and obtain an ellipsoid combination model based on the target ship structural parameters and the near-field measurement data; The ellipsoid combination model optimization module is configured to: optimize the ellipsoid combination model using a decomposition-based multi-objective evolutionary algorithm to obtain an optimized ellipsoid combination model; The ellipsoid combination model uses a combination of multiple rotating ellipsoids to simulate the magnetic field distribution of the ship. According to the actual structure and magnetic field characteristics of the ship, the ship is divided into multiple regions, and each region is approximated by one or more rotating ellipsoids. By reasonably adjusting the parameters of each rotating ellipsoid, the major axis and minor axis determine the shape and size of the ellipsoid, and the position coordinates determine its position in space, so that the combined model can more accurately fit the complex magnetic field distribution of the ship. The ship high-altitude magnetic field prediction module is configured as follows: using the optimized ellipsoid combination model to predict the ship high-altitude magnetic field and obtain the prediction result.

7. The distributed degaussing system full-space magnetic field multi-objective evolution prediction system according to claim 6 is characterized in that: The specific configuration of the ellipsoid combination model optimization module is: According to the ellipsoid combination model, an initial population is obtained; According to the ellipsoid combination model, a weight vector with a preset weight quantity is obtained by using a uniform distribution method; Constructing a neighborhood structure according to the weight vector; Performing iterative optimization based on the initial population, the weight vector, and the neighborhood structure to obtain an updated population; Confirm that the number of iterations reaches the maximum number of iterations, terminate the iterative optimization, and obtain the optimized ellipsoid combination model based on the current updated population.

8. The distributed degaussing system full-space magnetic field multi-objective evolution prediction system according to claim 7 is characterized in that: The initial population is randomly generated and includes a preset number of individuals. Each individual represents a set of parameters based on the ellipsoid combination model, and the parameters include the number of ellipsoids, the major axis, the minor axis, and the position coordinates of each ellipsoid.

9. The distributed degaussing system full-space magnetic field multi-objective evolution prediction system according to claim 7 is characterized in that: The weight vector is used to construct a single-objective subproblem, which includes a prediction accuracy subproblem, a computational complexity subproblem, and a model stability subproblem.

10. The distributed degaussing system full-space magnetic field multi-objective evolution prediction system according to claim 7 is characterized in that: The iterative optimization is performed according to the initial population, the weight vector and the neighborhood structure to obtain an updated population, specifically comprising: Obtaining a parent individual using a tournament selection method according to the initial population and the weight vector; Performing a crossover operation on the parent individuals using a simulated binary crossover method to obtain offspring individuals; Performing a mutation operation on the offspring individuals using a polynomial mutation method to obtain new offspring individuals; Obtaining a fitness value according to the new offspring individual; According to the fitness value and the neighborhood structure, individuals in the neighborhood are updated to obtain an updated population.

Images