Container ship wake flow field prediction method

Abstract

The invention discloses a container ship wake flow field prediction method, which comprises the steps of obtaining ship type sample data and corresponding wake flow field parameters, performing normalization processing and data cleaning, then performing correlation analysis, and performing screening to obtain parameters with correlation meeting requirements; for the parameters of which the correlation with the wake field parameters does not meet the requirement, acquiring data of which the correlation with the wake field parameters is higher than a set value in the parameters by adopting a principal component analysis method, and performing data dimension reduction on the data of which the correlation with the wake field parameters is higher than the set value; calculating the coefficient gain proportion of each characteristic parameter, and carrying out importance sorting on the stern molded line characteristic parameters to form a training data set; constructing and training a prediction model based on an adaptive enhancement algorithm; and normalizing stern molded line characteristic parameters detected in real time, and inputting the normalized stern molded line characteristic parameters into the trained prediction model to obtain predicted values of flow uniformity and wake fraction. According to the method, the calculated amount of wake field prediction is reduced, and the ship efficiency is improved.

Description

Technical Field

[0001] The invention relates to the field of ship and ocean engineering, and in particular to a method for predicting the wake field of a container ship. Background Art

[0002] In order to regulate the carbon emissions of ships, the International Maritime Organization (IMO) has formulated the International Shipping Carbon Intensity Rules as a short-term greenhouse gas emission reduction measure for ships, promoted the concept of "green ships", and clearly proposed two measures to evaluate the energy efficiency of ship emission reduction - the Energy Efficiency Existing-Ship Index (EEXI) and the Carbon Intensity Indicator (CII) rating program. Mandatory regulations have been made to reduce ship fuel consumption and greenhouse gas emissions, and ship emission reduction has become the mainstream direction of "green ships". Under this new regulation, most existing ships that cannot directly meet the standard can meet the standard by limiting the main engine power, reducing the speed and other measures. In addition, local line optimization, improved propellers, and installation of energy-saving devices are more efficient, easy to operate, and low-cost ways to meet the standards. Accurately and quickly predicting the ship's wake field and understanding its velocity distribution, flow field uniformity and other parameters have a guiding role in designing the structure of energy-saving devices. Rapidly predicting the wake field also has an important impact on increasing the propulsion efficiency when designing propellers.

[0003] With the development of computational fluid dynamics (CFD), researchers have developed many computational fluid dynamics methods by combining the governing equations of fluid motion with turbulence models. Since CFD technology takes up a lot of resources and time during model simulation and calculation, it is difficult for this technology to quickly calculate the results, and it will also greatly increase the computing cost. As the application field of machine learning continues to expand, introducing approximate models to estimate ship performance to reduce the burden of numerical calculations has become a new research direction.

[0004] However, the existing technology still has the following disadvantages: (1) Existing research on the prediction of container ship wake field mostly relies on ship tests and CFD methods, both of which require high costs. (2) Existing technology is difficult to cope with the calculation requirements of a large number of ship modification plans. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention proposes a method for predicting the wake field of a container ship.

[0006] The specific technical solutions are as follows:

[0007] A method for predicting the wake field of a container ship comprises the following steps:

[0008] S1: Acquire ship type sample data and corresponding wake field parameters; the ship type sample data includes stern line characteristic parameters, and the wake field data includes wake uniformity and wake fraction;

[0009] S2: normalize and clean the data obtained in S1;

[0010] S3: First, correlation analysis is performed between the characteristic parameters of the stern line and between the characteristic parameters of the stern line and the wake field parameters, and the parameters whose correlation meets the requirements are screened out; for those whose correlation with the wake field parameters does not meet the requirements, the principal component analysis method is used to obtain the data whose correlation with the wake field parameters is higher than the set value, and the data dimension reduction is performed on them; then the coefficient gain ratio of each characteristic parameter is calculated, and the importance of the characteristic parameters of the stern line is ranked to form a training data set;

[0011] S4: Build a prediction model based on the adaptive boosting algorithm and train the prediction model using the training data set;

[0012] S5: The characteristic parameters of the stern line detected in real time are normalized and input into the trained prediction model to obtain the predicted values ​​of the wake uniformity and the wake fraction.

[0013] Furthermore, in S2, data cleaning uses a box plot method to obtain outliers and remove them.

[0014] Furthermore, in S3, the correlation analysis uses a regression equation to calculate the Pearson correlation coefficient. If the Pearson correlation coefficient is greater than a set threshold, it is considered that the correlation requirement is met; otherwise, the requirement is not met.

[0015] Furthermore, in S4, a cross-training method is used to train the prediction model.

[0016] The beneficial effects of the present invention are:

[0017] The present invention can quickly solve the correlation problem between ship type elements and wake field, and reduce the workload required for ship testing and CFD calculation, save the time cost and calculation cost of the ship type optimization stage to the greatest extent, and obtain the wake field parameters that provide design reference for propeller design and energy-saving device structure. Studying the wake field characteristics behind the ship is of great significance for improving ship efficiency, ship type optimization and installing ship energy-saving appendages. The prediction model is used to achieve the purpose of evaluating the ship wake field using ship type data expressed based on the stern, laying the foundation for realizing intelligent ship type optimization design. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 4 is a flow chart of a method for predicting a container ship wake field in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] 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.

[0020] like Figure 1 As shown, a method for predicting the flow field of a container ship specifically includes the following steps:

[0021] S1: Obtain ship type sample data and corresponding wake field parameters. The ship type sample data includes stern line characteristic parameters, and the wake field parameters include wake uniformity and wake fraction.

[0022] S2: Preprocess the data obtained in S1, which is implemented through the following sub-steps:

[0023] (2.1) The data obtained by S1 is normalized to achieve dimensionlessness.

[0024] (2.2) Clean the data obtained in step (2.1). Since the collected data may contain data anomalies (i.e., data outside the normal range) due to data entry errors or unreasonable deformation, in order to ensure that the subsequent training of the prediction model is not affected by these values ​​and avoid wrong conclusions, it is necessary to obtain these abnormal data through the box plot method to solve the problems of abnormal items and invalid values; the principle of the box plot method is to represent the quartiles of the data set with a graph, so as to easily obtain abnormal values.

[0025] S3: Analyze the relationship between the characteristic parameters of the stern line and the wake field data, and construct a training set, which is specifically achieved through the following sub-steps:

[0026] (3.1) Correlation analysis: Machine learning can study the mapping between independent variables (i.e., stern line characteristic parameters) and dependent variables (i.e., wake field parameters), but before machine learning, it is necessary to know whether there is a certain connection between the independent variables and between the independent variables and the dependent variables, so as to measure the closeness between the variables. In this embodiment, the Pearson correlation coefficient is calculated using the regression equation. The larger the Pearson correlation coefficient, the stronger the correlation between the two variables. The independent variables with correlation coefficients greater than the set threshold are retained, and the remaining independent variables are screened out.

[0027] (3.2) Dimensionality reduction by principal component analysis (PCA): Since some independent variables in the previous correlation analysis have strong correlations with each other, but show weak correlations with the dependent variable when conducting correlation studies, these parameters are recorded as weak correlation parameters. The principal component analysis (PCA) method is used to analyze the relationship between the principal components of the data, and the parts of these weak correlation parameters that are more closely related to the dependent variable are mined as the main components for data dimensionality reduction.

[0028] (3.3) Importance analysis: Calculate the coefficient gain ratio of each feature parameter in the data set after dimensionality reduction processing, so as to obtain the importance ranking of each feature parameter and form a training data set.

[0029] In this embodiment, the ship type sample data and the corresponding flow field parameters are recorded as a group of sample data. First, 600 groups of sample data are obtained, and 4 groups of data with higher risk values ​​are removed after preprocessing. Through correlation analysis and dimensionality reduction processing, as well as the importance analysis of the parameters after dimensionality reduction, 596 groups of training data sets with higher importance and better accuracy are finally obtained. 20 groups are randomly selected as test groups, and a model with better training accuracy and credibility is obtained through cross-training method.

[0030] S4: Construct a prediction model based on the Adaboost (Adaptive Boosting) algorithm; the input of the prediction model is the characteristic parameters of the stern line, and the output is the wake uniformity and wake score. Use the training data set to train the prediction model; in order to improve the accuracy and credibility of the model training, the cross-training method is used in the training process.

[0031] S5: The characteristic parameters of the stern line detected in real time are normalized and input into the trained prediction model to obtain the predicted values ​​of the wake uniformity and the wake fraction, thereby realizing the wake field prediction.

[0032] 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 the flow field of a container ship, characterized in that: The following steps are involved: S1: Acquire ship type sample data and corresponding wake field parameters; the ship type sample data includes stern line characteristic parameters, and the wake field data includes wake uniformity and wake fraction; S2: normalize and clean the data obtained in S1; S3: First, correlation analysis is performed between the characteristic parameters of the stern line and between the characteristic parameters of the stern line and the wake field parameters, and the parameters whose correlation meets the requirements are screened out; for those whose correlation with the wake field parameters does not meet the requirements, the principal component analysis method is used to obtain the data whose correlation with the wake field parameters is higher than the set value, and the data dimension reduction is performed on them; then the coefficient gain ratio of each characteristic parameter is calculated, and the importance of the characteristic parameters of the stern line is ranked to form a training data set; S4: Build a prediction model based on the adaptive boosting algorithm and train the prediction model using the training data set; S5: The characteristic parameters of the stern line detected in real time are normalized and input into the trained prediction model to obtain the predicted values ​​of the wake uniformity and the wake fraction.

2. The method for predicting the wake field of a container ship according to claim 1, characterized in that: In S2, data cleaning uses a box plot method to obtain outliers and remove them.

3. The method for predicting the flow field of a container ship according to claim 1, characterized in that: In S3, the correlation analysis uses a regression equation to calculate the Pearson correlation coefficient. If the Pearson correlation coefficient is greater than a set threshold, it is considered that the correlation requirement is met; otherwise, the requirement is not met.

4. The method for predicting the flow field of a container ship according to claim 1, characterized in that: In S4, the prediction model is trained using a cross-training method.

Images