Ship initial stability height prediction method, device and system, and storage medium
Through gray correlation analysis and Bayesian optimization algorithm, the accuracy and real-time problems of ship initial stability high prediction in complex sea conditions are solved, and more efficient prediction results are achieved.
Patent Information
- Application Number
- CN202411305689.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-19
- Publication Date
- 2024-12-13
AI Technical Summary
The prior art is difficult to achieve fast and accurate prediction of ship's initial stability height (GM) under complex sea conditions. The traditional method has cumbersome calculations and slow real-time response, which fails to effectively balance real-time and accuracy.
Gray correlation analysis is used to select key factors, train RBF neural network, and hyperparameter optimization is used to optimize it through Bayesian optimization algorithm to achieve accurate prediction of the initial stability of the ship.
It effectively reduces prediction errors, improves prediction accuracy and real-time performance, and provides a reliable ship initial stability height prediction solution.
Smart Images
Figure CN119117207A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of ship safety, and in particular relates to a method and device, a system and a storage medium for predicting the initial stability height of a ship. Background Art
[0002] In the core link of port operations, the accurate execution of ship loading plan is directly related to transportation efficiency, cost savings and navigation safety. Among them, the accurate calculation and dynamic monitoring of ship's initial stability height (GM) constitute an indispensable key step. GM is a key indicator to measure the stability of a ship under a specific loading state. Its fluctuation not only profoundly affects the seaworthiness of the ship, but also directly relates to the port operation cycle and overall operational efficiency. Therefore, the rapid and accurate prediction of GM is of great significance to improving ship safety and operational efficiency.
[0003] Given that ship GM is affected by multiple nonlinear factors such as loading status and sea conditions, its dynamic changes are often difficult to predict, which poses a challenge to ship safety management. Traditional methods, such as relying on empirical formulas to calculate GM, can guide loading decisions to a certain extent, but they face the limitations of cumbersome calculation processes, slow real-time response, and difficulty in fully considering complex environmental factors. For example, although Wathne et al. considered GM factors in ro-ro ship loading planning, Zhang and Lee optimized loading through a multi-objective evolutionary algorithm combined with empirical formulas, and Puisa used MILP to optimize stability and loading. These methods have promoted the development of loading technology to varying degrees, but none of them has fundamentally solved the problem of balancing real-time performance and accuracy.
[0004] To overcome the above limitations, research in recent years has focused on smarter and more efficient GM prediction and evaluation methods. et al. tried to monitor GM through a specific system and evaluate stability using actual ship dynamic data; Terada et al. explored the prediction path of GM using autoregressive and self-organizing state space modeling techniques; Yun et al. explored the impact of GM from the perspective of ship maneuverability; Wawrzyński and Krata proposed an improved rolling cycle prediction model to achieve automatic real-time estimation of GM. However, although these methods have their own characteristics, they still face challenges in model complexity, data requirements and generalization ability in practical applications.
[0005] In summary, the development of an intelligent system that can efficiently integrate multivariate data and realize accurate GM prediction under complex sea conditions is of urgent need and important value for improving port operation efficiency and ensuring ship navigation safety. Summary of the invention
[0006] The technical problem to be solved by the present invention is to provide a method and device, a system and a storage medium for predicting the initial stability height of a ship.
[0007] To achieve the above object, the present invention adopts the following technical solution:
[0008] A method for predicting the initial stability height of a ship, comprising:
[0009] Step S1, obtaining historical ship type parameters;
[0010] Step S2: According to historical ship type parameters, the key factors related to GM are selected through grey correlation analysis, and the RBF neural network is trained;
[0011] Step S3, using the Bayesian optimization algorithm to optimize the hyperparameters of the trained RBF neural network;
[0012] Step S4: input the ship type parameters to be processed into the optimized and trained RBF neural network to predict the initial stability height of the ship.
[0013] Preferably, the pre-selected factors related to GM include: displacement, draft, beam ratio, center of gravity height, depth ratio, ship length, and ship width.
[0014] Preferably, step S2 comprises:
[0015] Step S21, selecting factors with strong correlation influencing the initial stability of the ship through grey correlation analysis according to historical ship type parameters;
[0016] Step S22: training the RBF neural network based on the factors that are highly correlated with the initial stability of the ship.
[0017] Preferably, in step S3, the hyperparameters of the trained RBF neural network are optimized by using a Bayesian optimization algorithm.
[0018] The present invention also provides a device for predicting the initial stability height of a ship, comprising:
[0019] Acquisition module, used to obtain historical ship type parameters;
[0020] The training module is used to select key factors related to GM through grey correlation analysis based on historical ship type parameters and train the RBF neural network;
[0021] The optimization module is used to optimize the hyperparameters of the trained RBF neural network using the Bayesian optimization algorithm;
[0022] The prediction module is used to input the ship type parameters to be processed into the optimized and trained RBF neural network to predict the initial stability height of the ship.
[0023] Preferably, the pre-selected factors related to GM include: displacement, draft, beam ratio, center of gravity height, depth ratio, ship length, and ship width.
[0024] Preferably, the training module includes:
[0025] A selection unit is used to select factors affecting the initial stability height of a ship through grey correlation analysis based on historical ship type parameters;
[0026] The training unit is used to train the RBF neural network according to the factors affecting the initial stability height of the ship.
[0027] As a preferred method, the optimization module optimizes the hyperparameters of the trained RBF neural network through the Bayesian optimization algorithm. The optimized RBF neural network can better adapt to the input ship type parameters and improve the prediction accuracy.
[0028] As a further preferred embodiment, the prediction module is used to input the ship type parameters to be processed into the RBF neural network optimized by the Bayesian optimization algorithm to realize the prediction of the initial stability height of the ship.
[0029] The present invention also provides a ship initial stability height prediction system, comprising: a memory and a processor, wherein the memory stores a computer program run by the processor, and the computer program executes a ship initial stability height prediction method when run by the processor.
[0030] The present invention also provides a storage medium, on which a computer program is stored, and the computer program executes the ship initial stability height prediction method when running.
[0031] The present invention has the following technical effects:
[0032] 1. The present invention adopts the Bayesian optimization algorithm to improve and optimize the RBF neural network, thereby effectively reducing the prediction error and achieving a more accurate prediction effect.
[0033] 2. In view of the complexity that the initial stability of ships is highly affected by many factors, grey correlation analysis is used to reduce the selection of input layers and improve the correlation between the input features and output values of the neural network.
[0034] 3. Aiming at the problem that the calculation steps of the ship's initial stability height are complicated and the real-time performance is poor, an improved RBF neural network is used to predict the initial stability height. The present invention has a lower error and provides a reliable solution for the prediction of the ship's initial stability height. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.
[0036] Figure 1 This is a flow chart of a method for predicting the initial stability height of a ship according to an embodiment of the present invention;
[0037] Figure 2 Schematic diagram of selecting factors affecting GM through grey relational analysis;
[0038] Figure 3 Schematic diagram of hyperparameter tuning of RBF neural network using Bayesian optimization algorithm. DETAILED DESCRIPTION
[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0040] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0041] Embodiment 1:
[0042] like Figure 1 As shown, an embodiment of the present invention provides a method for predicting the initial stability height of a ship, comprising:
[0043] Step S1, obtaining historical ship type parameters;
[0044] Step S2, according to the historical ship type parameters, select the factors related to GM to train the RBF neural network;
[0045] Step S3, performing hyperparameter optimization on the trained RBF neural network;
[0046] Step S4: input the ship type parameters to be processed into the optimized and trained RBF neural network to predict the initial stability height of the ship.
[0047] As an implementation method of an embodiment of the present invention, in step S1, the ship type parameters are the ship type data under the fully loaded state, and the pre-selected factors related to GM include: displacement, draft, beam ratio, center of gravity height, depth ratio, ship length, and ship width.
[0048] Furthermore, taking the passenger and vehicle ferry "Zijing 11" in the Qiongzhou Strait as an example, the relevant ship parameters under twelve load conditions, such as fully loaded departure and fully loaded arrival, actually measured between Xuwen Port and Haikou Port, are shown in Table 1:
[0049] Table 1
[0050] parameter describe value LOA captain 120.8m B Ship width 20.6m D Depth 6.3m T Full load draft 4.25m V Service Speed 12.5kn GT Gross Tonnage 8869ton Q Maximum passenger capacity 959
[0051] As an implementation of the embodiment of the present invention, step S2 includes:
[0052] Step S21, selecting factors affecting the initial stability height of the ship through grey correlation analysis according to historical ship type parameters;
[0053] Step S22: training the RBF neural network according to the highly correlated factors affecting the initial stability of the ship.
[0054] Furthermore, GM is closely related to the buoyancy center height KB, initial stability radius BM and center of gravity height KG, and considering the free liquid surface δGM f The impact is calculated using the following formula.
[0055] GM=KB+BM-KG-δGM f
[0056]
[0057] Where ρ is the liquid density; Δ is the displacement of the ship; i x is the moment of inertia of the free liquid surface, which can be divided into regular and irregular free liquid surfaces according to its shape.
[0058] Common regular free surface moments of inertia include: rectangle, isosceles triangle, circle, ellipse and isosceles trapezoid and other surface shapes of liquid surface moments of inertia, the calculation formula is as follows:
[0059]
[0060] Common irregular free surface moments of inertia include: the moments of inertia of the liquid surface with right triangle and right trapezoidal liquid surface shapes and the liquid surface with spherical, cylindrical or elliptical cylindrical shapes. The calculation formula is as follows:
[0061]
[0062] Among them, l is the maximum length of the liquid surface; b is the maximum width of the liquid surface; r is the radius of the circle; a and s are the semi-axis lengths along the length and width of the ship respectively; b1 and b2 are the widths of the liquid surface at the front and rear ends.
[0063] When the parent ship data is missing, the calculation of the initial stability height of the ship can be converted into the following formula based on the GM calculation formula:
[0064]
[0065]
[0066] Among them, B is the ship width; d is the design draft; D is the ship's molded depth; cw is the waterplane area coefficient; and cb is the block coefficient.
[0067] Further, in step S21, grey correlation analysis is a method for determining the degree of correlation between relevant factors based on the geometric similarity of the change curve. The input features and output values in the RBF neural network are not completely independent of each other, and the relationship between them is not clear, but in fact there is a certain correlation to form a grey system. Therefore, the grey comprehensive evaluation method can be used to establish an evaluation model, and the relationship between the reference sequence and each comparison sequence can be analyzed by comparing the statistical data related to the time series to evaluate their geometric relationship. The correlation between the input features and the output values is determined by calculating the grey correlation degree, and the following specific calculation steps are used:
[0068] 1) Determine the analysis order
[0069] To determine the analysis order, we must first establish an analysis series, where F is the reference sequence that reflects the system behavior characteristics, and Zi is the comparison sequence that affects the system behavior, that is,
[0070] F=F(k)
[0071] Zi=Zi(k)
[0072] Among them, k = 1, 2...n; i = 1, 2...m, n represents the length of the index, and m represents the number of indicators.
[0073] 2) Dimensionless processing of variables
[0074] Since the data dimensions of each factor column are different, the mean must be normalized for effective data processing, that is,
[0075]
[0076] 3) Calculate the grey correlation coefficient
[0077]
[0078] Among them, the resolution coefficient ρ∈(0,∞), and the smaller ρ is, the stronger the resolution ability is. The value range of ρ is (0,1), and ρ=0.5 is usually taken.
[0079] 4) Calculate the grey relational degree
[0080]
[0081] Taking the passenger and vehicle ferry "Zijing 11" in the Qiongzhou Strait as an example, the factors affecting the initial stability height of the ship include displacement, center of gravity height and other 7 parameters. In order to further determine the main factors affecting the initial stability height of the ship, the correlation ranking of each indicator is obtained through grey correlation analysis, as shown in Table 2:
[0082] Table 2
[0083] serial number Influencing factors Grey relational degree Sorting A1 Displacement 0.9806 1 A2 Draft 0.9788 2 A3 Width Ratio 0.9514 3 A4 Center of gravity height 0.7750 4 A5 Depth Ratio 0.0041 5 A6 captain 0.0011 6 A7 Ship width 0.0002 7
[0084] In order to further illustrate the strength of the correlation between variables, the grey relational analysis has certain classification standards: values between 0.8 and 1 indicate strong correlation, values between 0.6 and 0.8 indicate medium correlation, and values below 0.6 indicate weak correlation. Figure 2 shown.
[0085] according to Figure 2 It can be seen that the gray correlation of the depth ratio, length and width is less than 0.6, and the correlation is weak; the gray correlation of the center of gravity height is less than 0.8, and the correlation is medium; while the displacement, draft and width ratio show strong correlation. In order to simplify the selection of network input features, the combination of influencing factors with different correlations is selected as the network input data for predicting ship GM through gray correlation analysis, that is, the number of input layer nodes is trained according to the correlation. Therefore, the pre-selected factors affecting GM with strong correlation, namely displacement, draft and width ratio, are finally selected as the input features for training RBF neural network.
[0086] As an implementation of the embodiment of the present invention, in step S3, the hyperparameters of the trained RBF neural network are optimized by using a Bayesian optimization algorithm.
[0087] Furthermore, Bayesian optimization is a global optimization technique for optimizing functions with high computational cost and uncertainty. It selects the most promising sampling points by constructing a proxy model of the objective function, thereby finding the global optimal solution within a limited number of evaluations. The steps of hyperparameter optimization of the trained RBF neural network by Bayesian optimization algorithm include:
[0088] 1) Select the initial points: Select the initial points by Latin hypercube sampling method and evaluate the objective function values at these points.
[0089] 2) Constructing a proxy model: Use the evaluation results of the initial point to construct a proxy model of the objective function. This article constructs a proxy model using a Gaussian process suitable for processing complex and nonlinear relationships, and the calculation is as follows:
[0090]
[0091] 3) Select the next sampling point: For each candidate point x * , calculate its predictive distribution, mean μ and variance σ 2 Calculate as follows, and use the acquisition function to select α EI Select the next sampling point and calculate as follows:
[0092] μ(x * )=k(x * ,P) T K -1 L
[0093] σ 2 (x * )=k(x * ,x * )-k(x * ,P) T K -1 k(x * ,P)
[0094] α EI (d)=E[max(0,f(d + )-f(d))]
[0095] Among them, L is the corresponding objective function value, m is the mean function, K is the covariance matrix, represents multivariate Gaussian distribution, P is the set of all training data points, k is the kernel function, E represents the expected value, d represents the candidate point to be evaluated, and d + represents the currently known best point, f(d + ) is the current optimal objective function value
[0096] 4) Evaluate sampling points: Evaluate the newly selected sampling points on the true objective function.
[0097] 5) Update the proxy model: Add the new evaluation results to the dataset and update the proxy model.
[0098] 6) Iteration: Repeat steps 3)-5) until the stopping condition is met (such as reaching the maximum number of iterations or the improvement of the objective function value is insufficient).
[0099] Furthermore, after determining the input features of the RBF neural network through grey correlation analysis, the Bayesian optimization algorithm is used to tune the hyperparameters of the ship initial stability height prediction model. The results of the Bayesian optimization algorithm for the hyperparameter tuning of the RBF neural network are shown in the figure. Figure 3 shown.
[0100] In the present invention, Bayesian optimization optimizes the performance of the RBF neural network by iteratively sampling the two hyperparameter spaces of spread and hiddenNeurons, gradually exploring and using the feedback information (such as prediction error) of each sampling point to guide the selection of the next sampling. The optimization process defines the optimizable variables spread and hiddenNeurons, and uses the rbfObjective function as the target to evaluate the performance of the model under different hyperparameter combinations. After a maximum of 30 iterations, Bayesian reasoning can effectively search the hyperparameter space and finally find the optimal hyperparameter configuration that minimizes the error of the RBF neural network.
[0101] Embodiment 2:
[0102] The embodiment of the present invention further provides a device for predicting the initial stability height of a ship, comprising:
[0103] Acquisition module, used to obtain historical ship type parameters;
[0104] The training module is used to select key factors related to GM through grey correlation analysis based on historical ship type parameters and train the RBF neural network;
[0105] The optimization module is used to optimize the hyperparameters of the trained RBF neural network using the Bayesian optimization algorithm;
[0106] The prediction module is used to input the ship type parameters to be processed into the optimized and trained RBF neural network to predict the initial stability height of the ship.
[0107] As an implementation method of an embodiment of the present invention, the pre-selected factors related to GM include: displacement, draft, beam ratio, center of gravity height, depth ratio, ship length, and ship width.
[0108] As an implementation of an embodiment of the present invention, the training module includes:
[0109] A selection unit is used to select factors affecting the initial stability height of a ship through grey correlation analysis based on historical ship type parameters;
[0110] The training unit is used to train the RBF neural network according to the factors affecting the initial stability height of the ship.
[0111] As an implementation method of the present invention, the optimization module optimizes the hyperparameters of the trained RBF neural network through the Bayesian optimization algorithm. The optimized RBF neural network can improve the accuracy of prediction.
[0112] As a further implementation method, the prediction module inputs the ship type parameters to be processed into the RBF neural network optimized by the Bayesian optimization algorithm to achieve accurate prediction of the initial stability height of the ship.
[0113] Embodiment 3:
[0114] An embodiment of the present invention also provides a ship initial stability height prediction system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes a ship initial stability height prediction method when executed by the processor.
[0115] Embodiment 4:
[0116] An embodiment of the present invention further provides a storage medium, on which a computer program is stored, and the computer program executes a method for predicting the initial stability height of a ship when running.
[0117] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.
Claims
1. A method for predicting the initial stability height of a ship, characterized in that: include: Step S1, obtaining historical ship type parameters; Step S2: According to historical ship type parameters, the key factors related to GM are selected through grey correlation analysis, and the RBF neural network is trained; Step S3, using the Bayesian optimization algorithm to optimize the hyperparameters of the trained RBF neural network; Step S4: input the ship type parameters to be processed into the optimized and trained RBF neural network to predict the initial stability height of the ship.
2. The method for predicting the initial stability height of a ship according to claim 1, characterized in that: The pre-selection factors related to GM include: displacement, draft, beam ratio, center of gravity height, depth ratio, ship length, and ship width.
3. The method for predicting the initial stability height of a ship according to claim 2, characterized in that: Step S2 includes: Step S21, selecting factors with strong correlation influencing the initial stability of the ship through grey correlation analysis according to historical ship type parameters; Step S22: training the RBF neural network based on the factors that are highly correlated with the initial stability of the ship.
4. The method for predicting the initial stability height of a ship according to claim 3, characterized in that: In step S3, the hyperparameters of the trained RBF neural network are optimized using a Bayesian optimization algorithm.
5. A device for predicting the initial stability height of a ship, characterized in that: include: Acquisition module, used to obtain historical ship type parameters; The training module is used to select key factors related to GM through grey correlation analysis based on historical ship type parameters and train the RBF neural network; The optimization module is used to optimize the hyperparameters of the trained RBF neural network using the Bayesian optimization algorithm; The prediction module is used to input the ship type parameters to be processed into the optimized and trained RBF neural network to predict the initial stability height of the ship.
6. The device for predicting the initial stability height of a ship according to claim 5, characterized in that: The pre-selection factors related to GM include: displacement, draft, beam ratio, center of gravity height, depth ratio, ship length, and ship width.
7. The device for predicting the initial stability height of a ship according to claim 6, characterized in that: The training modules include: A selection unit is used to select factors affecting the initial stability height of a ship through grey correlation analysis based on historical ship type parameters; The training unit is used to train the RBF neural network according to the factors affecting the initial stability height of the ship.
8. The device for predicting the initial stability height of a ship according to claim 7, characterized in that: Also includes: The hyperparameters of the optimized RBF neural network have been optimized using the Bayesian optimization algorithm through the optimization module; The prediction module is used to input the ship type parameters to be processed into the optimized RBF neural network to output the prediction result of the initial stability height of the ship.
9. A ship initial stability height prediction system, characterized in that: include: A memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes the ship initial stability height prediction method as described in any one of claims 1 to 4 when running.
10. A storage medium, characterized in that: The storage medium stores a computer program, which, when running, executes the ship initial stability height prediction method as described in any one of claims 1 to 4.