A speed prediction method based on ship type elements of fat large ships

Abstract

The present invention discloses a speed prediction method based on ship type elements of fat ships, comprising: obtaining ship type elements including ship type main scale elements and hull line characteristic elements, using a multivariate high-order regression analysis method to obtain key ship type elements, and using them as a sample data set; constructing and training a speed prediction model based on an RBF neural network, whose input is the key ship type elements and output is the residual resistance coefficient and the self-propulsion factor; the trained speed prediction model is used for real-time speed prediction. On the basis of the ship type main scale elements, the present invention increases the influence of the hull line characteristic elements on the performance prediction, thereby improving the accuracy of the speed prediction.

Description

Technical Field

[0001] The invention relates to the technical field of ship and ocean engineering, and in particular to a speed prediction method based on ship type elements of a fat ship. Background Art

[0002] In recent years, with the globalization of the world market and trade, maritime transportation has flourished, and sea transportation has become the main mode of transportation for international logistics. In order to effectively cope with complex working environments, diverse cargo transportation, and reduce costs, higher requirements are put forward for ship type optimization. Ship speed, as one of the important performances of ships, determines the economic performance and performance of ships to a certain extent. Therefore, conducting research on ship speed performance based on intelligent decision-making of ship type solutions will play an important role in promoting the development of the ship field.

[0003] In the past few decades, most ship speed forecasts still rely on expert experience, and the accuracy and persuasiveness of these forecasts are still to be verified. The existing technologies have the following shortcomings:

[0004] (1) The existing speed prediction model is mainly based on the main macroscopic ship factors such as aspect ratio, width-to-draft ratio, and square coefficient, and does not reflect the impact of the hull line characteristic elements on performance;

[0005] (2) The existing forecast models are not accurate and efficient enough;

[0006] (3) Most existing design solutions use univariate regression analysis, which is a relatively simple method. Summary of the invention

[0007] In view of the deficiencies of the prior art, the present invention proposes a speed prediction method based on ship type elements of a fat ship.

[0008] The specific technical solutions are as follows:

[0009] A method for predicting speed based on ship type elements of a fat ship comprises the following steps:

[0010] S1: Obtain ship type elements and preprocess them, wherein the ship type elements include the main scale elements of the ship type and the characteristic elements of the hull line; use the multivariate high-order regression analysis method to obtain the optimal number of multiple regression analysis i, and the combination with the largest determination coefficient among all combinations of elements under the optimal number of elements, and all elements in the combination are key ship type elements, and use them as sample data sets;

[0011] S2: construct a speed prediction model based on RBF neural network and use the sample data set for training; the input of the speed prediction model is the key ship type elements, and the output is the residual resistance coefficient and the self-propulsion factor;

[0012] S3: After normalizing the key ship type elements obtained in real time, input them into the trained speed prediction model to obtain the predicted residual resistance coefficient and self-propulsion factor.

[0013] Furthermore, S1 is implemented by the following sub-steps:

[0014] (1.1) Obtain ship type elements and preprocess them;

[0015] (1.2) Set the number of variables i for the initial multiple regression analysis;

[0016] (1.3) Traverse all ship type elements and obtain all combinations of i elements;

[0017] (1.4) Calculate the coefficient of determination R of the quadratic regression of each combination i 2 and error;

[0018] (1.5) Determine whether all combinations meet the conditions, the condition is the determination coefficient R 2 If the value is greater than 0.95 and there is no overfitting, then the number of elements i = i + 1 is taken, and steps (1.3)-(1.5) are repeated until the condition is met. The number of elements at this time is the optimal number of elements.

[0019] (1.6) Output the optimal number of elements and determine the combination with the largest coefficient value among all combinations of elements. The ship type element corresponding to this combination is the key ship type element and constitutes the sample data set.

[0020] Furthermore, the preprocessing includes sequentially performing data cleaning and normalization.

[0021] Furthermore, S2 is implemented by the following sub-steps:

[0022] (2.1) Disordering the sample data set obtained in S1;

[0023] (2.2) Divide the sample data set obtained in step (2.1) into a training set and a test set;

[0024] (2.3) Normalize the training set and test set respectively;

[0025] (2.4) constructing a speed prediction model and using the training set for training, wherein the speed prediction model adopts an RBF neural network architecture, whose input is i columns of key ship type elements, and whose output is a residual resistance coefficient and a self-propulsion factor;

[0026] (2.5) Use the test set to calculate the error of the speed prediction model trained in step (2.4) to verify the effectiveness of the speed prediction model. If the error is less than the set threshold, the trained speed prediction model is output; otherwise, return to step (2.4).

[0027] The beneficial effects of the present invention are:

[0028] The present invention improves the accuracy of speed prediction. On the basis of the main scale elements of the ship type, the influence of the hull line characteristic elements on the performance prediction is added, so the speed can be predicted quickly and accurately. Moreover, through the relationship between the residual resistance coefficient and the self-propulsion factor and the ship type elements, quantitative data is provided as a reference for the ship type optimization design, thereby improving the ship type optimization efficiency and quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a flow chart of a method for predicting speed based on ship type elements of a fat ship in an embodiment of the present invention.

[0030] Figure 2 4 is a flow chart of multivariate high-order regression analysis in an embodiment of the present invention.

[0031] Figure 3 It is a training flow chart of the speed prediction model in an embodiment of the present invention. DETAILED DESCRIPTION

[0032] The present invention will be described in detail below according to the accompanying drawings and preferred embodiments, and the purpose and effect of the present invention will become more clear. The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0033] like Figure 1 As shown, a speed prediction method based on the ship type elements of fat ships is used. According to the ship type element data, a method combining multivariate high-order regression analysis and RBF neural network is used to obtain the intrinsic relationship between the ship type elements and the ship's residual resistance coefficient and self-propulsion factor. The method specifically includes the following steps:

[0034] S1: Figure 2 As shown in the figure, the key ship type factors that have an important impact on ship performance are obtained by using the multivariate high-order regression analysis method, which is specifically achieved through the following sub-steps:

[0035] (1.1) Obtain ship type elements and perform data cleaning and normalization on them in turn. Ship type elements include ship type main scale elements and hull line characteristic elements.

[0036] The main dimensional elements of the ship type are the main dimensions that represent the size of the hull. They are the basis for calculating various performance parameters of the ship, measuring the size of the ship, verifying various fees, and inspecting the passage of ships through restricted waterways such as canals and locks. These elements include: ship length-to-width ratio, ship breadth-to-draft ratio, slenderness coefficient, longitudinal position of the center of buoyancy, squareness coefficient, overall prism coefficient, overall vertical prism coefficient, and mid-transverse section coefficient.

[0037] The characteristic elements of the hull lines determine the shape of the main part of the ship, that is, the shape of the outer plate below the open continuous deck, including the underwater part and the above-water part; these elements include: cross-sectional area curve elements and designed waterplane elements, which can be specifically subdivided into the length of the bow section of the cross-sectional area curve, the length of the mid-body section parallel to the cross-sectional area curve, the length of the stern section of the cross-sectional area curve, the longitudinal position of the buoyancy center of the bow section, the longitudinal position of the buoyancy center of the stern section, the bow prism coefficient, the stern prism coefficient, the stern vertical prism coefficient, the bow vertical prism coefficient, the length of the designed waterline outflow section, the length of the designed waterline parallel to the mid-section, the length of the designed waterline inflow section, the longitudinal position of the outflow section centroid, the longitudinal position of the inflow section centroid, the longitudinal position of the buoyancy center, the overall waterplane coefficient, the stern waterplane coefficient, and the bow waterplane coefficient.

[0038] (1.2) Preliminarily set the number of variables i for multivariate regression analysis. The specific number can be set manually based on experience to reduce calculation time.

[0039] (1.3) Traverse all ship type elements and obtain all combinations of i elements.

[0040] (1.4) Calculate the coefficient of determination R of the quadratic regression of each combination i 2 and errors.

[0041] (1.5) Determine whether all combinations meet the conditions, the condition is the determination coefficient R 2 Close to 1 (in this embodiment, R 2 If the value is greater than 0.95 and there is no overfitting, if not, the number of variables of the multiple regression is increased (i.e., i=i+1) and the calculation is performed again, i.e., steps (1.3)-(1.5) are repeated until the condition is met, and the number of variables at this time is the optimal number of variables. In this embodiment, the amount of data contained in each ship type element is 300, and the optimal number of variables i=8 is finally determined for the multiple regression analysis.

[0042] (1.6) Output the optimal combination corresponding to the largest determination coefficient value in each combination under the optimal number of elements. The ship type element corresponding to the optimal combination is the key ship type element, which constitutes the sample data set.

[0043] S2: Figure 3 As shown in the figure, the key ship type factors selected by regression analysis are applied to the speed prediction model training based on RBF neural network, which is implemented through the following sub-steps:

[0044] (2.1) The sample data set obtained by S1 is adjusted in a disordered manner. The purpose is to allow the RBF neural network to perform random sampling during subsequent training, so that points of different degrees can be collected, making the data prediction effect better.

[0045] (2.2) Data division: Divide the sample data set into training set and test set.

[0046] (2.3) Normalize the training set and test set separately to achieve dimensionless processing.

[0047] (2.4) A speed prediction model is constructed and trained using the training set. The speed prediction model adopts the RBF neural network architecture, with i columns of key ship type elements as input and the residual resistance coefficient and self-propulsion factor as output.

[0048] (2.5) Use the test set to calculate the error of the speed prediction model trained in (2.4) to verify the effectiveness of the speed prediction model. If the error is less than the set threshold, it means that the speed prediction model training is completed; otherwise, return to step (2.4).

[0049] S3: After normalizing the key ship type elements obtained in real time, input them into the trained speed prediction model to obtain the predicted residual resistance coefficient and self-propulsion factor to achieve speed prediction.

[0050] The purpose of the present invention is to improve the speed forecasting method under the traditional mode, and to develop a new speed forecasting model based on the research means of big data analysis and on the basis of expert experience and knowledge, which is applied to the speed forecasting of fat and large ships which are currently popular.

[0051] Those skilled in the art can understand that the above are only preferred examples of the invention and are not intended to limit the invention. Although the invention is described in detail with reference to the above examples, those skilled in the art can still modify the technical solutions recorded in the above examples or replace some of the technical features therein with equivalents. Any modification, equivalent replacement, etc. made within the spirit and principle of the invention shall be included in the protection scope of the invention.

Claims

1. A method for predicting speed based on ship type elements of fat ships, characterized in that: The following steps are involved: S1: Obtain ship type elements and preprocess them, wherein the ship type elements include the main scale elements of the ship type and the characteristic elements of the hull line; use the multivariate high-order regression analysis method to obtain the optimal number of multiple regression analysis i, and the combination with the largest determination coefficient among all combinations of elements under the optimal number of elements, and all elements in the combination are key ship type elements, and use them as sample data sets; S2: construct a speed prediction model based on RBF neural network and use the sample data set for training; the input of the speed prediction model is the key ship type elements, and the output is the residual resistance coefficient and the self-propulsion factor; S3: After normalizing the key ship type elements obtained in real time, input them into the trained speed prediction model to obtain the predicted residual resistance coefficient and self-propulsion factor.

2. The method for predicting speed based on ship type elements of a fat ship according to claim 1, characterized in that: The S1 is implemented by the following sub-steps: (1.1) Obtain ship type elements and preprocess them; (1.2) Set the number of variables i for the initial multiple regression analysis; (1.3) Traverse all ship type elements and obtain all combinations of i elements; (1.4) Calculate the coefficient of determination R of the quadratic regression of each combination i 2 and error; (1.5) Determine whether all combinations meet the conditions, the condition is the determination coefficient R 2 If the value is greater than 0.95 and there is no overfitting, then the number of elements i = i + 1 is taken, and steps (1.3)-(1.5) are repeated until the condition is met. The number of elements at this time is the optimal number of elements. (1.6) Output the optimal number of elements and determine the combination with the largest coefficient value among all combinations of elements. The ship type element corresponding to this combination is the key ship type element and constitutes the sample data set.

3. The method for predicting speed based on ship type elements of fat ships according to claim 1, characterized in that: The preprocessing includes sequentially performing data cleaning and normalization.

4. The method for predicting speed based on ship type elements of a fat ship according to claim 1, characterized in that: The S2 is implemented by the following sub-steps: (2.1) Disordering the sample data set obtained in S1; (2.2) Divide the sample data set obtained in step (2.1) into a training set and a test set; (2.3) Normalize the training set and test set respectively; (2.4) constructing a speed prediction model and using the training set for training, wherein the speed prediction model adopts an RBF neural network architecture, whose input is i columns of key ship type elements, and whose output is a residual resistance coefficient and a self-propulsion factor; (2.5) Use the test set to calculate the error of the speed prediction model trained in step (2.4) to verify the effectiveness of the speed prediction model. If the error is less than the set threshold, the trained speed prediction model is output; otherwise, return to step (2.4).

Images