Ship turbulence model constant correction and display method and system

Abstract

The invention provides a ship turbulence model constant correction and display method and system, and the method comprises the steps: carrying out the numerical simulation of a k-epsilon turbulence model, generating a virtual flow field of a ship, and extracting a plurality of discrete speed observation data from the virtual flow field; a Monte Carlo sampling method is adopted to generate a plurality of set members of model constants of the k-omega turbulence model, based on the initial value of the state parameter vector of each set member, the k-omega turbulence model is adopted to calculate the predicted value of the state parameter vector of each set member, and then a set covariance matrix is calculated; and a Kalman gain matrix is calculated according to the set covariance matrix, then data assimilation is performed on the state parameter vector of each set member to obtain the corrected state parameter vector of each set member, and the average value of all the corrected state parameter vectors is calculated to obtain the optimal model constant vector, so that the cost is effectively saved.

Description

Technical Field

[0001] The present invention relates to the technical field of numerical simulation of ships and ocean engineering, and in particular to a method and system for correcting and displaying constants of a ship turbulence model. Background Art

[0002] The wake field of a ship, especially the wake field in the propeller working area, is directly related to the ship's resistance and propulsion performance, the vibration excitation characteristics of the stern area, and the comfort of the crew. Therefore, accurate prediction of the stern flow and wake field distribution based on the Computational Fluid Dynamics (CFD) method is the basis for ship design and performance optimization. However, due to the significant geometric changes in the stern area of ​​the ship and the complexity of the ship's wake field flowing through the stern, including the turbulent characteristics of the wake boundary layer, the 3D separation characteristics of the hull flow, and the interference characteristics of the free surface waveform on the stern flow, the existing CFD software and methods have problems such as insufficient prediction accuracy and failure to meet practical requirements. Taking the Reynolds Average Navier-Stokes (RANS) equation as an example, the accuracy of RANS turbulence model calculation is highly dependent on model constants, and the default values ​​of these constants are calibrated by classical flows such as flat boundary layers and free shear flows. For actual complex flow fields such as ship stern flow, the use of default values ​​often leads to significant errors in flow field calculation results compared with experimental measurement results. Many experiments have also proved that there is a certain deviation between the model constant values ​​corresponding to the observed data and the default values. However, it is extremely difficult to select appropriate model constants for specific calculation objects. Currently, there is no very effective method to accurately identify the various empirical constants of the turbulence model equations. In the simulation of ship flow fields, a suitable turbulence model constant is generally selected through a large number of trial calculations. This process is time-consuming and labor-intensive, and cannot meet the increasingly in-depth research needs.

[0003] With the development of artificial intelligence technologies such as deep neural networks, data-driven algorithms have also begun to be applied to the quantification of uncertainty in turbulence models and the improvement and construction of models. For example, machine learning is used to improve the prediction accuracy of turbulence separation on the wing surface of the SA turbulence model, and the network reconstructed model is embedded in the SA model. After adopting this model, the lift coefficient and pressure prediction accuracy of the airfoil surface are greatly improved. For example, using a supervised learning algorithm to establish a mathematical model of the closed term of the turbulence model, and then embedding the model trained by machine learning into the CFD numerical simulation to improve the model performance.

[0004] In addition, the optimization of turbulence model constants by using experimental data-driven algorithms is a common optimization strategy. For example, the parameters in the k-ε turbulence model were optimized by using random optimization and Monte Carlo sampling techniques. After parameter optimization, the prediction error of the reattachment length on the leeward side of the building was significantly reduced. For example, the adjoint-driven inversion process combined with artificial neural networks and Gaussian process regression is used to correct the parameters and other closure coefficients of the turbulence model to better predict the flow field of the turbulent boundary layer. For example, the Latin hypercube sampling (LHS) method is used to study the uncertainty contained in the k-ε turbulence model constants. By studying the parameter uncertainty in the reverse flow, it is found that the model parameters have a great influence on the calculation results such as the average velocity of the fluid, turbulence intensity, reattachment point position and wall pressure. For example, a data-driven method for obtaining the optimal model constants is proposed. By applying small perturbations, the minimum variance minimum between the experiment and the prediction is obtained. This method cannot obtain the global optimal model constants, but can only obtain the local optimal constants. The final prediction results may still have a large error with the observed values.

[0005] For the correction of turbulence model constants, the commonly used method is the optimal constant estimation based on data assimilation. Data assimilation is the integration of experimental observations and model predictions. Its three major elements are prediction models, observation data and assimilation algorithms. This method incorporates observation data into model predictions to change the model's running trajectory, ultimately achieving the purpose of optimizing model performance and improving prediction accuracy. Data assimilation was first used in meteorological forecasting, and then expanded to geology, hydrology, system monitoring and other fields. In recent years, data assimilation has been introduced into the field of computational fluid dynamics and has become an important method for optimizing turbulence model constants. For example, the uncertainty of constants in different RANS models is studied based on the data assimilation method. For example, data assimilation is used to compare and study the prediction effects of different RANS models on free jets after optimizing the turbulence model constants. The study shows that after the model constants are optimized, the prediction accuracy of the model can be improved. For example, a hybrid data assimilation method, namely the ensemble-variational method (EnVar), is studied. The k-ωSST turbulence model constants are recalibrated by this method. The results show that it effectively improves the accuracy of the flow characteristics prediction model. Summary of the invention

[0006] In order to solve the problem that it is difficult to select suitable model constants in the current RANS numerical simulation of ships, the present invention provides a ship turbulence model constant correction and display method, which uses the k-ε turbulence model in the RANS turbulence model for numerical simulation, combines discrete velocity observation data and uses the ensemble Kalman filter algorithm to calculate the Kalman gain matrix, and corrects the model constants of the k-ω turbulence model in the RANS turbulence model through data assimilation, without requiring a large amount of training data and expensive computing resources, effectively saving costs, and having high flexibility and versatility. The present invention also relates to a ship turbulence model constant correction and display system.

[0007] The technical solution of the present invention is as follows:

[0008] A method for correcting and displaying constants of a ship turbulence model, characterized in that it comprises the following steps:

[0009] Virtual flow field generation and velocity observation data extraction steps: setting the boundary conditions and initial conditions of a ship flow field, based on the boundary conditions and initial conditions and using the k-ε turbulence model in the RANS turbulence model to perform numerical simulation, generate the virtual flow field of the ship, and extract multiple discrete velocity observation data from the virtual flow field;

[0010] The steps of calculating the predicted value of the state parameter vector are as follows: using the Monte Carlo sampling method to randomly sample in the model constant space of the k-ω turbulence model in the RANS turbulence model, generating multiple set members of the model constants of the k-ω turbulence model, and calculating the predicted value of the state parameter vector of each set member based on the initial value of the state parameter vector of each set member and using the k-ω turbulence model;

[0011] State parameter vector correction step: Based on the ensemble Kalman filter algorithm, the ensemble covariance matrix is ​​calculated according to the number of ensemble members, the predicted value of the state parameter vector and the average value of the predicted values ​​of all state parameter vectors, and the Kalman gain matrix is ​​calculated according to the ensemble covariance matrix. The state parameter vector of each ensemble member is data assimilated according to the Kalman gain matrix, the speed observation data and the predicted value of the state parameter vector to obtain the corrected state parameter vector of each ensemble member, and the average value of all corrected state parameter vectors is calculated to obtain the optimal model constant vector;

[0012] Velocity field calculation and display steps: Input the optimal model constant vector into the k-ω turbulence model, calculate the optimized velocity field, compare the optimized velocity field with the initial velocity field in the initial conditions, and display the comparison results.

[0013] Preferably, in the step of calculating the predicted value of the state parameter vector, the Monte Carlo sampling method is used to perform random sampling in the model constant space of the k-ω turbulence model in the RANS turbulence model, including: the model constant of the k-ω turbulence model is a default value, the default value is taken as the average value, and the Monte Carlo sampling method is used to perform random sampling in the average value to generate an initial set with multiple set members.

[0014] Preferably, in the step of calculating the predicted value of the state parameter vector, the state parameter vector in each set member will be brought into the k-ω turbulence model for calculation starting from the initial value until the turbulence numerical simulation calculation converges, thereby calculating the predicted value of the state parameter vector of each set member.

[0015] Preferably, in the state parameter vector correction step, data assimilation is performed on the state parameter vector of each set member, a data assimilation algorithm is used to parse the mapping relationship, the inverse problem is solved by integrating relevant speed observation data, and the Kalman filter algorithm is used to iterate the analysis process to obtain the optimal model constant vector.

[0016] Preferably, the state parameter vector includes velocity, pressure and temperature.

[0017] Preferably, in the virtual flow field generation and velocity observation data extraction steps, the boundary conditions include inlet boundary conditions, outlet boundary conditions and hull surface boundary conditions; and / or, the initial conditions include initial velocity field, initial pressure field and initial turbulence field.

[0018] A ship turbulence model constant correction and display system, characterized in that it comprises a virtual flow field generation and velocity observation data extraction module, a state parameter vector prediction value calculation module, a state parameter vector correction module and a velocity field calculation and display module which are connected in sequence.

[0019] The virtual flow field generation and velocity observation data extraction module sets the boundary conditions and initial conditions of a ship flow field, performs numerical simulation based on the boundary conditions and initial conditions and adopts the k-ε turbulence model in the RANS turbulence model to generate a virtual flow field of the ship, and extracts a plurality of discrete velocity observation data from the virtual flow field;

[0020] The state parameter vector prediction value calculation module adopts a Monte Carlo sampling method to perform random sampling in the model constant space of the k-ω turbulence model in the RANS turbulence model, generates multiple set members of the model constant of the k-ω turbulence model, and calculates the prediction value of the state parameter vector of each set member based on the initial value of the state parameter vector of each set member and adopts the k-ω turbulence model;

[0021] The state parameter vector correction module, based on the ensemble Kalman filter algorithm, calculates the ensemble covariance matrix according to the number of ensemble members, the predicted value of the state parameter vector and the average value of the predicted values ​​of all the state parameter vectors, and calculates the Kalman gain matrix according to the ensemble covariance matrix, performs data assimilation on the state parameter vector of each ensemble member according to the Kalman gain matrix, the speed observation data and the predicted value of the state parameter vector to obtain the corrected state parameter vector of each ensemble member, and calculates the average value of all the corrected state parameter vectors to obtain the optimal model constant vector;

[0022] The velocity field calculation and display module inputs the optimal model constant vector into the k-ω turbulence model, calculates the optimized velocity field, compares the optimized velocity field with the initial velocity field in the initial condition, and displays the comparison result.

[0023] Preferably, in the state parameter vector prediction value calculation module, the Monte Carlo sampling method is used to perform random sampling in the model constant space of the k-ω turbulence model in the RANS turbulence model, including: the model constant of the k-ω turbulence model is a default value, the default value is taken as the average value, and the Monte Carlo sampling method is used to perform random sampling in the average value to generate an initial set with multiple set members.

[0024] Preferably, in the state parameter vector correction module, data assimilation is performed on the state parameter vector of each set member, a data assimilation algorithm is used to parse the mapping relationship, the inverse problem is solved by integrating relevant speed observation data, and the Kalman filter algorithm is used to iterate the analysis process to obtain the optimal model constant vector.

[0025] Preferably, the state parameter vector includes velocity, pressure and temperature;

[0026] And / or, the boundary conditions include inlet boundary conditions, outlet boundary conditions and hull surface boundary conditions;

[0027] And / or, the initial conditions include an initial velocity field, an initial pressure field and an initial turbulence field.

[0028] The beneficial effects of the present invention are:

[0029] The present invention provides a method for correcting and displaying constants of a ship turbulence model. The method first sets boundary conditions and initial conditions of a ship flow field, performs numerical simulation based on the boundary conditions and initial conditions and using a k-ε turbulence model in a RANS turbulence model, generates a virtual flow field of the ship using the k-ε turbulence model, extracts multiple discrete velocity observation data from the virtual flow field, and then uses a Monte Carlo sampling method to perform random sampling in a model constant space of a k-ω turbulence model (Baseline example) in the RANS turbulence model to generate multiple set members of the ship turbulence model constants. Based on the initial value of a state parameter vector of each set member and using the k-ω turbulence model, a predicted value of a state parameter vector of each set member is calculated. Then, based on an ensemble Kalman filtering algorithm, a predicted value of a state parameter vector of each set member is calculated according to the number of set members, the predicted value of the state parameter vector, and the average predicted value of all state parameter vectors. The ensemble covariance matrix is ​​calculated by the mean, and the Kalman gain matrix is ​​calculated based on the ensemble covariance matrix. The state parameter vector of each ensemble member is assimilated according to the Kalman gain matrix, the velocity observation data and the predicted value of the state parameter vector to obtain the corrected state parameter vector of each ensemble member, and the average value of all corrected state parameter vectors is calculated to obtain the optimal model constant vector. The state parameter vectors after the ensemble Kalman filter data assimilation are averaged to obtain the optimized model constant matrix. The optimal model constant vector is substituted into the original calculation model for confirmation and verification, that is, the optimal model constant vector is finally input into the k-ω turbulence model to calculate the optimized velocity field, and the optimized velocity field is compared with the initial velocity field in the initial condition, and the comparison results are displayed, which effectively improves the optimization speed, saves costs, and has high flexibility and versatility.

[0030] The present invention adopts the Ensemble Kalman filter (EnKF) algorithm to have advantages in dealing with small samples, high dimensions and highly nonlinear problems. Compared with machine learning methods, such as neural networks, the EnKF algorithm can correct the turbulence model constants by using less discrete observation data without a large amount of training data and expensive computing resources, especially when the ship model experiment measurement cost is high and the observation data is scarce, this advantage is very obvious. This makes the EnKF algorithm a more efficient and economical choice, and the EnKF algorithm performs data assimilation (also known as the EnKF assimilation algorithm) to become an effective method for solving complex fluid dynamics problems in actual ship flow fields. In addition, the EnKF assimilation algorithm is a non-invasive data assimilation algorithm that does not require explicit modification or adjustment of the original prediction model, so it is relatively simple to implement, can be coupled with different CFD software (Fluent, OpenFOAM, etc.), and is suitable for different types of prediction models, with high flexibility and versatility.

[0031] The present invention also relates to a ship turbulence model constant correction and display system, which corresponds to the above-mentioned ship turbulence model constant correction and display method, and can be understood as a system for realizing the above-mentioned ship turbulence model constant correction and display method, including a virtual flow field generation and velocity observation data extraction module, a state parameter vector prediction value calculation module, a state parameter vector correction module and a velocity field calculation and display module connected in sequence, and each module works in coordination with each other. Based on experimental observation data, numerical simulation is performed by adopting the k-ε turbulence model in the RANS turbulence model, and the Kalman gain matrix is ​​calculated by combining discrete velocity observation data and adopting the ensemble Kalman filter algorithm. The model constants of the k-ω turbulence model in the RANS turbulence model are corrected by a non-invasive data assimilation algorithm. No large amount of training data and expensive computing resources are required, which effectively saves costs and has high flexibility and versatility. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a flow chart for solving the direct problem and the inverse problem through the RANS turbulence model.

[0033] Figure 2 It is a flow chart of the ship turbulence model constant correction and display method of the present invention.

[0034] Figure 3 It is a preferred flow chart of the ship turbulence model constant correction and display method of the present invention.

[0035] Figure 4 It is a schematic diagram of the comparison results of the velocity field before and after optimization of the ship turbulence model constant correction and display method of the present invention. DETAILED DESCRIPTION

[0036] The present invention will be described below in conjunction with the accompanying drawings.

[0037] The RANS model constant optimization method based on data assimilation is an effective method to improve the prediction accuracy of the ship flow field. In order to study the feasibility and effectiveness of this method in the ship flow field, the present invention proposes a ship turbulence model constant correction and display method, which can also be called a ship turbulence model constant correction (optimization) and display method based on data assimilation. The 14000TEU Panamax container ship is taken as the research object, and the k-ω model numerical simulation is adopted. The data assimilation is combined with discrete velocity observation data to correct the model constants of the k-ω turbulence model. The process of optimizing the turbulence model constants through data assimilation is essentially to solve the inverse problem, such as Figure 1The forward problem and inverse problem are shown in Figure 1. The prediction model (i.e., RANS turbulence model) establishes the mapping relationship between the prediction state parameters (velocity, pressure, and temperature) and the model constants, while data assimilation uses the assimilation algorithm to analyze the mapping relationship and integrates the relevant observation data to solve the inverse problem, that is, using the output value of the prediction model to optimize the input value, thereby inferring and calibrating the model constants. The flowchart of this method is shown in Figure 1. Figure 2 As shown, the following steps are included in sequence:

[0038] 1. Steps for generating virtual flow field and extracting velocity observation data: set the boundary conditions and initial conditions of a ship's flow field, perform numerical simulation based on the boundary conditions and initial conditions and use the k-ε turbulence model in the RANS turbulence model to generate a virtual flow field of the ship, and extract multiple discrete velocity observation data from the virtual flow field.

[0039] Specifically, Figure 3 As shown in the figure, firstly, the boundary conditions and initial conditions of a ship flow field are set. The boundary conditions include the inlet boundary conditions, the outlet boundary conditions and the hull surface boundary conditions. The inlet boundary conditions include the flow velocity, flow direction, turbulence intensity, etc. at the inlet, which are set according to the actual situation or test data. The outlet boundary conditions are usually set as the free flow outlet, and the pressure or velocity of the fluid can also be set as needed. The hull surface boundary conditions are usually set as the no-slip wall, that is, it is assumed that the velocity of the fluid is equal to that of the hull surface; the initial conditions include the initial velocity field, the initial pressure field and the initial temperature field. Then, through numerical simulation, the k-ε turbulence model will calculate the distribution of parameters such as the velocity field and turbulence intensity in the entire flow field according to the set boundary conditions and initial conditions, and then generate the virtual flow field of the ship, assume the virtual flow field as the real flow field, and extract 100 velocity observation data y from the virtual flow field. exp , extract the velocity observation data y exp As the observation information input of the Ensemble Kalman filter (EnKF).

[0040] 2. Steps for calculating the predicted value of the state parameter vector: Use the Monte Carlo sampling method to perform random sampling in the model constant space of the k-ω turbulence model (Baseline example) in the RANS turbulence model to generate multiple set members of the model constants of the k-ω turbulence model, based on the initial value of the state parameter vector of each set member, and use the k-ω turbulence model to calculate the predicted value of the state parameter vector of each set member.

[0041] Specifically, statistical information is obtained: the model constants in the k-ω turbulence model (Baseline example) are the default values, and the default values ​​are taken as the mean values. The model constant space of the Baseline example is randomly sampled by the Monte Carlo sampling method to generate N set members of the model constants of the k-ω turbulence model. Each set member contains a state parameter vector (velocity vector and k-ω turbulence model constant vector). Then, based on the initial value of the state parameter vector of each set member, the k-ω turbulence model in the RANS turbulence model is used to calculate the predicted value of the state parameter vector of each set member, and the calculation is performed according to the following formula:

[0042] x f =F(x 0 ,v) (1)

[0043] In the above formula, x f is the predicted value of the state parameter vector, x 0 is the initial value of the state parameter vector, F is the k-ω turbulence model in the RANS turbulence model, and ν is the system noise.

[0044] Set the initial value x of the state parameter vector of each member of the set 0 Substitute them into the RANS turbulence model in turn and calculate until convergence to obtain the corresponding N groups of predicted values ​​of state parameter vectors x f , as the prediction information input of the Ensemble Kalman filter (EnKF). The state parameter vector of each ensemble member is in the form of x i =(U i ,θ i ) T , where U is the velocity vector, θ is the turbulence model constant vector, and i is the geometric member index.

[0045] Experimental data acquisition: Extract discrete velocity observation data y from the flow field exp Since the verification ship used in the present invention has no experimental data, a flow field is recalculated using the k-ε turbulence model, and the flow field is assumed to be a real flow field, from which 100 velocity observation points can be extracted. The extracted observation data y exp As observation information input to the ensemble Kalman filter algorithm.

[0046] 3. State parameter vector correction step: Based on the ensemble Kalman filter algorithm, the ensemble covariance matrix is ​​calculated according to the number of ensemble members, the predicted value of the state parameter vector and the average value of the predicted values ​​of all state parameter vectors, and the Kalman gain matrix is ​​calculated according to the ensemble covariance matrix. The state parameter vector of each ensemble member is data assimilated according to the Kalman gain matrix, the speed observation data and the predicted value of the state parameter vector to obtain the corrected state parameter vector of each ensemble member, and the average value of all corrected state parameter vectors is calculated to obtain the optimal model constant vector. Among the many data assimilation algorithms, the present invention selects the ensemble Kalman filter (EnKF) algorithm proposed by Evensen in 1994. The EnKF algorithm is developed from the classical Kalman filter and the extended Kalman filter algorithm, and is a kalman filter based on the Monte Carlo theory. The principle can be simply understood as artificially creating a set with N set members, assuming a normal distribution, and correcting the N set members separately through filtering methods so that the state parameters corresponding to the N combined members are close to the true values ​​of the system under study. It is especially suitable for numerical model optimization of high-order nonlinear systems such as turbulence problems.

[0047] This step integrates the uncertainty information of the speed observation data and the statistical information of the set members to determine the Kalman gain matrix K and complete the update of the set members. Specifically, based on the set Kalman filter algorithm, the set covariance matrix P is calculated according to the number of set members, the predicted value of each state parameter vector, and the average value of the predicted values ​​of all state parameter vectors, and is calculated according to the following formula:

[0048]

[0049] In the above formula, P is the set covariance matrix, N is the number of set members, is the average value of the predicted values ​​of all state parameter vectors, i is the set member index, T is the transposed matrix, and then the covariance matrix R of the set measurement disturbance is obtained based on the velocity observation data.

[0050] Then the Kalman gain matrix is ​​calculated based on the collective covariance matrix P, and is calculated according to the following formula:

[0051] K=PH T (HPH T +R) -1 (3)

[0052] In the above formula, P is the collective covariance matrix, R is the covariance matrix of the collective measurement disturbance, H is the observation function, and the covariance matrix of the collective measurement disturbance is W=(ω 1 ,ω 2,…,ω n ), where ω is the observation error, which is an artificially given constant, and n is the number of velocity observation data.

[0053] Finally, the state parameter vector of each ensemble member is assimilated according to the Kalman gain matrix K, the velocity observation data and the predicted value of the state parameter vector (i.e., the model constant of the k-ω turbulence model is corrected), and the correction is performed according to the following formula:

[0054]

[0055] In the above formula, ω represents the observation error, y exp represents the speed observation data, Represents the predicted value of the state parameter vector of the i-th set member. According to the above formula, the corrected state parameter vector of each set member can be obtained. Then the average value of all corrected state parameter vectors is calculated to obtain the optimal model constant vector.

[0056] This step is to ensemble Kalman filter algorithm filtering correction, and the statistical information (prediction value x f ) and observation information (velocity observation data y exp ) is used as the input of the ensemble Kalman filter algorithm, and the Python program is used to implement the assimilation process shown in equations (2) to (4) to modify the model constant values.

[0057] 4. Velocity field calculation and display step: This step recalculates the flow field, brings the corrected model constants back into the k-ω turbulence model of the RANS model to calculate, obtains the optimized velocity field, and compares the velocity field results before and after optimization. In other words, the optimal model constant vector is input into the k-ω turbulence model to calculate the optimized velocity field, and the optimized velocity field is compared with the initial velocity field in the initial condition, and the comparison results are displayed. The comparison results are as follows: Figure 4 As shown, the velocity distribution along the y-axis at different cross sections is shown, the solid line is the result before optimization (Baseline, or the result before assimilation, that is, the initial velocity field in the initial condition), the hollow origin is the true value (Truth), and the dotted line is the result after optimization (Da, or the result after assimilation, that is, the optimized velocity field). By comparison, it can be concluded that the result after assimilation is closer to the true value, thereby verifying the effect and feasibility of the ship turbulence model constant correction scheme based on data assimilation of the present invention.

[0058] The present invention also relates to a ship turbulence model constant correction and display system, which corresponds to the above-mentioned ship turbulence model constant correction and display method, and can be understood as a system for implementing the above-mentioned method. The system includes a virtual flow field generation and velocity observation data extraction module, a state parameter vector prediction value calculation module, a state parameter vector correction module and a velocity field calculation and display module connected in sequence. Specifically,

[0059] The virtual flow field generation and velocity observation data extraction module sets the boundary conditions and initial conditions of a ship flow field, performs numerical simulation based on the boundary conditions and initial conditions and adopts the k-ε turbulence model to generate a virtual flow field of the ship, and extracts a plurality of discrete velocity observation data from the virtual flow field;

[0060] The state parameter vector prediction value calculation module adopts a Monte Carlo sampling method to perform random sampling in the model constant space of the k-ω turbulence model in the RANS turbulence model, generates multiple set members of the model constant of the k-ω turbulence model, and calculates the prediction value of the state parameter vector of each set member based on the initial value of the state parameter vector of each set member and adopts the k-ω turbulence model;

[0061] The state parameter vector correction module, based on the ensemble Kalman filter algorithm, calculates the ensemble covariance matrix according to the number of ensemble members, the predicted value of the state parameter vector and the average value of the predicted values ​​of all the state parameter vectors, and calculates the Kalman gain matrix according to the ensemble covariance matrix, performs data assimilation on the state parameter vector of each ensemble member according to the Kalman gain matrix, the speed observation data and the predicted value of the state parameter vector to obtain the corrected state parameter vector of each ensemble member, and calculates the average value of all the corrected state parameter vectors to obtain the optimal model constant vector;

[0062] The velocity field calculation and display module inputs the optimal model constant vector into the k-ω turbulence model, calculates the optimized velocity field, compares the optimized velocity field with the initial velocity field in the initial condition, and displays the comparison result.

[0063] Preferably, in the state parameter vector prediction value calculation module, the Monte Carlo sampling method is used to perform random sampling in the model constant space of the k-ω turbulence model in the RANS turbulence model, including: the model constant of the k-ω turbulence model is a default value, the default value is taken as the average value, and the Monte Carlo sampling method is used to perform random sampling in the average value to generate an initial set with multiple set members.

[0064] Preferably, in the state parameter vector correction module, data assimilation is performed on the state parameter vector of each set member, a data assimilation algorithm is used to parse the mapping relationship, the inverse problem is solved by integrating relevant speed observation data, and the Kalman filter algorithm is used to iterate the analysis process to obtain the optimal model constant vector.

[0065] Preferably, the state parameter vector includes velocity, pressure and temperature.

[0066] Preferably, the boundary conditions include inlet boundary conditions, outlet boundary conditions and hull surface boundary conditions.

[0067] Preferably, the initial conditions include an initial velocity field, an initial pressure field and an initial turbulence field.

[0068] The present invention provides an objective and scientific ship turbulence model constant correction and display method and system, which uses the k-ε turbulence model in the RANS prediction model for numerical simulation, combines discrete velocity observation data and uses the ensemble Kalman filter algorithm to calculate the Kalman gain matrix, and corrects the model constants of the k-ω turbulence model in the RANS prediction model. It does not require a large amount of training data and expensive computing resources, effectively saves costs, and has high flexibility and versatility.

[0069] It should be noted that the above-described specific implementations can enable those skilled in the art to more fully understand the invention, but do not limit the invention in any way. Therefore, although this specification has described the invention in detail with reference to the drawings and embodiments, those skilled in the art should understand that the invention can still be modified or replaced by equivalents. In short, all technical solutions and improvements that do not deviate from the spirit and scope of the invention should be included in the protection scope of the patent for the invention.

Claims

1. A method for correcting and displaying constants of a ship turbulence model, characterized in that: The following steps are involved: Virtual flow field generation and velocity observation data extraction steps: setting the boundary conditions and initial conditions of a ship flow field, based on the boundary conditions and initial conditions and using the k-ε turbulence model in the RANS turbulence model to perform numerical simulation, generate the virtual flow field of the ship, and extract multiple discrete velocity observation data from the virtual flow field; The steps of calculating the predicted value of the state parameter vector are as follows: using the Monte Carlo sampling method to randomly sample in the model constant space of the k-ω turbulence model in the RANS turbulence model, generating multiple set members of the model constants of the k-ω turbulence model, and calculating the predicted value of the state parameter vector of each set member based on the initial value of the state parameter vector of each set member and using the k-ω turbulence model; State parameter vector correction step: Based on the ensemble Kalman filter algorithm, the ensemble covariance matrix is ​​calculated according to the number of ensemble members, the predicted value of the state parameter vector and the average value of the predicted values ​​of all state parameter vectors, and the Kalman gain matrix is ​​calculated according to the ensemble covariance matrix. The state parameter vector of each ensemble member is data assimilated according to the Kalman gain matrix, the speed observation data and the predicted value of the state parameter vector to obtain the corrected state parameter vector of each ensemble member, and the average value of all corrected state parameter vectors is calculated to obtain the optimal model constant vector; Velocity field calculation and display steps: Input the optimal model constant vector into the k-ω turbulence model, calculate the optimized velocity field, compare the optimized velocity field with the initial velocity field in the initial conditions, and display the comparison results.

2. The ship turbulence model constant correction and display method according to claim 1, characterized in that: In the step of calculating the predicted value of the state parameter vector, the Monte Carlo sampling method is used to perform random sampling in the model constant space of the k-ω turbulence model in the RANS turbulence model, including: the model constant of the k-ω turbulence model is a default value, the default value is used as the average value, and the Monte Carlo sampling method is used to perform random sampling in the average value to generate an initial set with multiple set members.

3. The ship turbulence model constant correction and display method according to claim 1 or 2, characterized in that: In the step of calculating the predicted value of the state parameter vector, the state parameter vector in each set member will be brought into the k-ω turbulence model for calculation starting from the initial value until the turbulence numerical simulation calculation converges, thereby calculating the predicted value of the state parameter vector of each set member.

4. The ship turbulence model constant correction and display method according to claim 3 is characterized in that: In the state parameter vector correction step, data assimilation is performed on the state parameter vector of each set member, a data assimilation algorithm is used to analyze the mapping relationship, the inverse problem is solved by integrating relevant speed observation data, and the Kalman filter algorithm is used to analyze the process iteration to obtain the optimal model constant vector.

5. The ship turbulence model constant correction and display method according to claim 1, characterized in that: The state parameter vector includes velocity, pressure and temperature.

6. The ship turbulence model constant correction and display method according to claim 1, characterized in that: In the virtual flow field generation and velocity observation data extraction steps, the boundary conditions include inlet boundary conditions, outlet boundary conditions and hull surface boundary conditions; and / or, the initial conditions include initial velocity field, initial pressure field and initial turbulence field.

7. A ship turbulence model constant correction and display system, characterized in that: It includes a virtual flow field generation and velocity observation data extraction module, a state parameter vector prediction value calculation module, a state parameter vector correction module and a velocity field calculation and display module, which are connected in sequence. The virtual flow field generation and velocity observation data extraction module sets the boundary conditions and initial conditions of a ship flow field, performs numerical simulation based on the boundary conditions and initial conditions and adopts the k-ε turbulence model in the RANS turbulence model to generate a virtual flow field of the ship, and extracts a plurality of discrete velocity observation data from the virtual flow field; The state parameter vector prediction value calculation module adopts a Monte Carlo sampling method to perform random sampling in the model constant space of the k-ω turbulence model in the RANS turbulence model, generates multiple set members of the model constant of the k-ω turbulence model, and calculates the prediction value of the state parameter vector of each set member based on the initial value of the state parameter vector of each set member and adopts the k-ω turbulence model; The state parameter vector correction module, based on the ensemble Kalman filter algorithm, calculates the ensemble covariance matrix according to the number of ensemble members, the predicted value of the state parameter vector and the average value of the predicted values ​​of all the state parameter vectors, and calculates the Kalman gain matrix according to the ensemble covariance matrix, performs data assimilation on the state parameter vector of each ensemble member according to the Kalman gain matrix, the speed observation data and the predicted value of the state parameter vector to obtain the corrected state parameter vector of each ensemble member, and calculates the average value of all the corrected state parameter vectors to obtain the optimal model constant vector; The velocity field calculation and display module inputs the optimal model constant vector into the k-ω turbulence model, calculates the optimized velocity field, compares the optimized velocity field with the initial velocity field in the initial condition, and displays the comparison result.

8. The ship turbulence model constant correction and display system according to claim 7, characterized in that: In the state parameter vector prediction value calculation module, the Monte Carlo sampling method is used to perform random sampling in the model constant space of the k-ω turbulence model in the RANS turbulence model, including: the model constant of the k-ω turbulence model is a default value, the default value is used as an average value, and the Monte Carlo sampling method is used to perform random sampling in the average value to generate an initial set with multiple set members.

9. The ship turbulence model constant correction and display system according to claim 7 or 8, characterized in that: In the state parameter vector correction module, data assimilation is performed on the state parameter vector of each set member, a data assimilation algorithm is used to analyze the mapping relationship, the inverse problem is solved by integrating relevant speed observation data, and the Kalman filter algorithm is used to analyze the process iteration to obtain the optimal model constant vector.

10. The ship turbulence model constant correction and display system according to claim 7 or 8, characterized in that: The state parameter vector includes velocity, pressure and temperature; And / or, the boundary conditions include inlet boundary conditions, outlet boundary conditions and hull surface boundary conditions; And / or, the initial conditions include an initial velocity field, an initial pressure field and an initial turbulence field.

Images