A method for reconstructing the wake field of a bare ship hull based on finite indirect observation

Abstract

This application discloses a method for reconstructing the wake field of a bare ship hull based on finite indirect observation, relating to the field of ship flow field technology. This method integrates measured data from actual ship navigation and full-scale numerical simulation data of the ship. Using a flow field reconstruction model, it obtains the global reconstructed flow field of the stern region of the ship during actual navigation based on the observed local measured flow field. Then, using a viscopotential coupled self-propulsion solver combined with the simulated nominal wake field of the bare hull obtained from numerical simulation, the reconstructed nominal wake field of the bare hull can be obtained iteratively. This method enables the indirect inference and reconstruction of the nominal wake field of the complete region of the bare ship hull from the total velocity field observed from a finite local region of the stern with a propeller, providing technical support for obtaining the reconstructed nominal wake field of the bare hull and filling a gap in the industry.

Description

Technical Field

[0001] This application relates to the field of ship flow field technology, and in particular to a method for reconstructing the wake field of a real ship's bare hull based on limited indirect observation. Background Technology

[0002] As the International Maritime Organization further raises its requirements for ship energy efficiency, both regulations and market demands place higher demands on the energy-saving potential of fluid energy-saving technologies. Therefore, it is necessary to study the stern flow mechanism, explore the potential for fluid energy saving, and further optimize the design of combined configurations. Thus, it is urgent to obtain the detailed flow characteristics of real ships navigating in the real marine environment and to construct reliable flow field information that can be compared with numerical simulations.

[0003] However, the scale effect of ships, especially the scale effect caused by the complex interactions of the stern, energy-saving devices, propeller, and rudder in integrated energy-saving propulsion systems, makes it difficult to extrapolate the fine flow characteristics of a real ship navigating in a real marine environment from laboratory model tests. Furthermore, unlike laboratory model tests, the wake field of a bare hull cannot be directly measured in a real marine environment, and the measurement process cannot isolate the influence of the propeller, which hinders the analysis of drag composition and wake field characteristics. Simultaneously, due to limitations in space, location, observation angle, and laser power for installing flow field measurement equipment such as PIVs on actual ships, the observable flow field range during actual ship navigation is also relatively limited. These technical bottlenecks make it difficult to obtain the wake field of a bare hull when a real ship is navigating in a real marine environment, thus limiting the research and development of stern flow mechanisms. Summary of the Invention

[0004] To address the aforementioned problems and technical requirements, this applicant proposes a method for reconstructing the wake field of a bare ship hull based on finite indirect observation. The technical solution of this application is as follows:

[0005] A method for reconstructing the wake field of a bare ship hull based on finite indirect observation, the method comprising:

[0006] The local measured flow field of the stern observation area of ​​a ship during actual ship navigation was obtained using PIV technology. The stern observation area is a local area within the global stern area of ​​the ship.

[0007] The global reconstructed flow field of the stern region of a ship during actual navigation is obtained by using a flow field reconstruction model based on the local measured flow field. The flow field reconstruction model is used to reflect the mapping relationship between the characteristic parameters of the local region of the ship during actual navigation and the characteristic parameters of the global region of the stern region of the ship during actual navigation. The flow field reconstruction model is obtained by training a neural network using real-scale simulation data of the ship.

[0008] The nominal wake field of the bare hull is obtained by real-scale simulation data of the ship. The nominal wake field of the bare hull is iteratively solved using a viscous potential coupled self-propulsion solver that includes a RANS solver and a propeller potential flow solver. The iteration terminates when the output of the viscous potential coupled self-propulsion solver and the globally reconstructed flow field are determined. The nominal wake field of the bare hull of the ship is reconstructed using the output of the viscous potential coupled self-propulsion solver. The nominal wake field is used to characterize the effective wake velocity of the bare hull of the ship.

[0009] A further technical solution involves using a viscous potential coupled self-propulsion solver for iterative solutions, including:

[0010] Random perturbations are applied to the characteristic parameters of the simulated nominal wake field of the bare hull using perturbation parameters to generate a simulated perturbed nominal wake field.

[0011] The viscous potential coupled self-propulsion solver is used to solve the nominal wake field based on the simulated disturbance, and outputs the total velocity field at the propeller disk surface, the solved values ​​of the ship's macroscopic motion parameters, and the actual wake velocity.

[0012] Based on the error between the total velocity field at the propeller disk surface and the globally reconstructed flow field, as well as the error between the solved values ​​of the ship's macroscopic motion parameters and the actual values ​​of the collected ship's macroscopic motion parameters, the applied disturbance is updated using Bayesian filtering.

[0013] The steps of applying random perturbations to the characteristic parameters of the simulated nominal wake field of the bare hull using the perturbation parameters to generate a simulated perturbed nominal wake field are repeated using the updated perturbation parameters until the error between the total velocity field at the propeller disk and the global reconstructed flow field is within the corresponding error range, and the error between the solved value of the ship's macroscopic motion parameters and the actual value of the collected ship's macroscopic motion parameters is within the corresponding error range. At this point, the reconstructed nominal wake field of the bare hull, composed of the effective wake velocity output by the viscous potential coupled self-propulsion solver, is obtained.

[0014] A further technical solution involves using a viscous potential coupled self-propulsion solver to solve the problem based on the simulated disturbance nominal wake field, including:

[0015] Determine the propeller thrust and determine the corresponding volume force distribution based on the propeller thrust;

[0016] The volume force distribution is substituted into the blasting disk model inside the RANS solver, and the RANS solver is used to solve the nominal wake field based on the simulated disturbance to obtain the total velocity field at the propeller disk surface. The blasting disk model is used to represent the influence of propeller rotation on the flow.

[0017] The calculated circumferential average value of the total velocity field at the propeller disk surface is substituted into the propeller potential flow solver as the resultant velocity value of the propeller to obtain the propeller induced velocity.

[0018] The propeller thrust and volume force distribution are iteratively updated to solve the problem until the solution values ​​of the ship's macroscopic motion parameters converge. Then, the calculated total velocity field at the propeller disk surface, the solution values ​​of the ship's macroscopic motion parameters, and the effective wake velocity are output. The effective wake velocity is the difference between the total velocity field at the propeller disk surface and the propeller induced velocity.

[0019] The further technical solution involves solving for the ship's macroscopic motion parameters, including hull resistance, torque, and propeller thrust, and iteratively updating the propeller thrust and volume force distribution to obtain the solution, including:

[0020] If at least one of the hull resistance obtained by the solution and the torque obtained by the propeller potential flow solver fails to converge, the volume force distribution obtained by the propeller potential flow solver is used as the updated volume force distribution, and the step of substituting the volume force distribution into the pulsating disk model inside the RANS solver is executed again.

[0021] When the hull resistance and the torque obtained by the propeller potential flow solver both converge, the propeller thrust is updated by interpolation. When the propeller thrust also converges, the iterative solution ends. If the propeller thrust does not converge, the step of determining the corresponding volume force distribution based on the propeller thrust is executed again using the updated propeller thrust.

[0022] A further technical solution involves applying random perturbations to the characteristic parameters of the simulated nominal wake field of a bare ship hull, including:

[0023] A random perturbation of Δω(x)=F(G(x)) is applied to the characteristic parameters of the simulated nominal wake field of the bare hull, where G(x) is a Gaussian random process with mean μ and kernel function K(x,x′), F(G(x)) represents the result of mapping G(x) using the distribution constraint mapping operator, and x and x′ are the spatial position coordinates.

[0024] A further technical solution involves methods for training the flow field reconstruction model, including:

[0025] A real-scale numerical simulation of a ship under self-propelled conditions is performed to obtain the global simulated flow field of the stern region and extract the characteristic parameters of the global simulated flow field.

[0026] Local downsampling of the global simulated flow field under different sampling conditions was performed according to the observation area at the stern to obtain the low-frequency simulated flow field under different sampling conditions, and the characteristic parameters of each low-frequency simulated flow field were extracted respectively.

[0027] Using the characteristic parameters of the low-frequency simulated flow field as input and the characteristic parameters of the global simulated flow field as output, a flow field reconstruction model is obtained based on neural network training.

[0028] A further technical solution involves sampling conditions including the obstruction area and the sampling method. Different sampling conditions include at least one difference between the obstruction area and the sampling method. The method for locally downsampling the global simulated flow field under the sampling conditions according to the stern observation area includes:

[0029] The global simulated flow field is locally downsampled within the sampling area according to the sampling method indicated by the sampling conditions. The sampling area is the area other than the occlusion area indicated by the sampling conditions within the stern observation area. The occlusion area is used to simulate the observation blind zone within the stern observation area.

[0030] A further technical solution involves obtaining a global simulated flow field in the stern region of a ship, including:

[0031] We conducted a full-scale numerical simulation of the ship to obtain unsteady flow field data of the global stern region under different propeller rotation phases.

[0032] Phase averaging was performed on unsteady flow field data under multiple different propeller rotation phases to obtain the global simulated flow field of the stern region of the ship.

[0033] Further technical solutions include methods for obtaining the globally reconstructed flow field, such as:

[0034] The flow characteristics of the local measured flow field are extracted, and the global reconstructed flow field is determined based on the flow field reconstruction model by perturbation of the characteristic parameters. The approximation error between the characteristic parameters of the global reconstructed flow field and the flow characteristics of the local measured flow field is within the corresponding error range. Furthermore, the characteristic parameters of the local flow field obtained by mapping the characteristic parameters of the global reconstructed flow field through the flow field reconstruction model are also within the corresponding error range.

[0035] A further technical solution involves extracting characteristic parameters of the simulated flow field from any set of simulated flow fields, including the global simulated flow field and each set of low-frequency simulated flow fields.

[0036] Characteristic modes and characteristic decomposition coefficients are obtained by performing characteristic decomposition on the simulated flow field. The characteristic decomposition methods used include KL decomposition, POD decomposition and DMD decomposition.

[0037] The beneficial technical effects of this application are:

[0038] This application discloses a method for reconstructing the wake field of a bare ship hull based on limited indirect observation. This method integrates the local measured flow field of the stern observation area obtained from actual observations during the actual ship navigation process, as well as the actual-scale numerical simulation data of the ship. It realizes the analysis and reconstruction of the nominal wake field of the complete area of ​​the bare ship hull from the total velocity field observed in a limited local PIV region of the stern with propeller. This provides technical support for obtaining the nominal wake field of the bare ship hull and fills a gap in the industry.

[0039] Compared with traditional direct measurement and analysis techniques, the method of this application can obtain the nominal wake field of a ship or the flow field of other cross sections at similar locations through feature modeling and reconstruction analysis of the indirect measurement flow field within a limited area. Its application is scalable and can be applied to ship types of different sizes and different measurement locations to realize the reconstruction of the flow field in the area to be analyzed.

[0040] Compared with traditional analysis techniques based on a single data source, the method in this application can effectively integrate experimentally observed flow fields and numerically simulated flow fields under different conditions. Through feature decomposition, feature extraction, and feature fusion, it achieves effective utilization of key information from different data sources. For observed flow fields, it focuses on extracting information such as magnitude, distribution, and spatial correlation; for numerically simulated flow fields, it focuses on extracting information such as spatial modes and global distribution patterns. This enables the effective reconstruction of the total velocity field of the actual ship and the nominal wake field of the bare hull. Attached Figure Description

[0041] Figure 1 This is a flowchart of a method for reconstructing the wake field of a bare ship hull in one embodiment.

[0042] Figure 2 This is a flowchart of a method for constructing a flow field reconstruction model and obtaining a global reconstructed flow field from a local measured flow field in one embodiment.

[0043] Figure 3 This is a schematic diagram of the stern observation area and the blind spot within the stern observation area in an example.

[0044] Figure 4 This is a flowchart of a method for obtaining the reconstructed nominal wake field of a bare hull using a viscous potential coupling solver based on the globally reconstructed flow field and the simulated nominal wake field of the bare hull, as described in one embodiment.

[0045] Figure 5 This is a schematic diagram of an iterative solution method using a viscous potential coupling solver in one embodiment. Detailed Implementation

[0046] The specific embodiments of this application will be further described below with reference to the accompanying drawings.

[0047] This application discloses a method for reconstructing the wake field of a bare ship hull based on finite indirect observation. Please refer to [reference needed]. Figure 1 The flowchart shown illustrates that the method includes the following steps:

[0048] Step 1: Observe the local measured flow field of the stern observation area of ​​the ship during actual ship navigation using PIV technology. The stern observation area is the area formed by the observation range of PIV towards the stern of the ship. This stern observation area is a local area in the global stern area of ​​the ship. The global stern area is the entire stern flow field area of ​​the ship.

[0049] The obtained local measured flow field is the measured total velocity field within the stern observation area. The local measured flow field obtained through PIV technology has two limitations: firstly, it has spatial limitations and cannot cover the entire stern region; secondly, it is observed during actual ship navigation, therefore the observed result includes the total velocity field caused by the ship's propeller, and is not simply the wake field of the bare hull. Therefore, the reconstruction in this application includes two aspects: one is the reconstruction from the stern observation area to the entire stern region, and the other is the reconstruction of the wake field of the bare hull from the observed total velocity field.

[0050] Step 2: Using the flow field reconstruction model, the global reconstructed flow field of the stern region of the ship during actual navigation is obtained based on the local measured flow field. Please refer to [reference needed]. Figure 2 The flowchart shown below:

[0051] This step is used to reconstruct the flow field from the observation area at the stern to the global area at the stern. It uses a pre-trained flow field reconstruction model, which is trained on a neural network using real-scale simulation data of the ship. This flow field reconstruction model is used to reflect the mapping relationship between the characteristic parameters of the local area of ​​the ship during actual ship navigation and the characteristic parameters of the global area at the stern of the ship during actual ship navigation.

[0052] In one embodiment, when obtaining the global reconstructed flow field using the flow field reconstruction model, the flow characteristics of the measured local flow field are first extracted. In one embodiment, the flow characteristics include at least one of typical measuring point velocities, vortex structure characteristics, and large-scale characteristics of spatial correlation. Then, the global reconstructed flow field is determined based on the flow field reconstruction model by perturbation of characteristic parameters. The approximation error between the characteristic parameters of the global reconstructed flow field and the flow characteristics of the measured local flow field is within the corresponding error range, and the characteristic parameters of the local flow field obtained by mapping the characteristic parameters of the global reconstructed flow field through the flow field reconstruction model are also within the corresponding error range.

[0053] This step requires the use of a pre-trained flow field reconstruction model; therefore, in one embodiment, a method for pre-training the flow field reconstruction model is also included.

[0054] (1) Perform a real-scale numerical simulation of the ship under self-propulsion conditions to obtain the global simulated flow field of the global region of the stern of the ship, and extract the characteristic parameters of the global simulated flow field.

[0055] The global simulated flow field obtained through real-scale numerical simulation includes the influence of the propeller. During propeller rotation, the numerical simulation results of the propeller's forward measurement plane differ at different propeller rotation phases. Therefore, to compensate for the influence of the propeller rotation phase on the numerical simulation results, real-scale numerical simulations of the ship are performed to obtain unsteady flow field data of the stern global region under different propeller rotation phases. Then, phase averaging is performed on the unsteady flow field data under multiple different propeller rotation phases to finally obtain the global simulated flow field of the ship's stern global region.

[0056] When extracting characteristic parameters of the global simulated flow field, the global simulated flow field is decomposed to obtain characteristic modes and characteristic decomposition coefficients. The characteristic decomposition methods used include KL decomposition, POD decomposition and DMD decomposition. The appropriate characteristic decomposition method can be selected according to the actual situation.

[0057] (2) Local downsampling is performed on the global simulated flow field under different sampling conditions according to the observation area at the stern to obtain the low-frequency simulated flow field under different sampling conditions, and the characteristic parameters of each low-frequency simulated flow field are extracted respectively. The method for extracting the characteristic parameters of each low-frequency simulated flow field is the same as the method for extracting the characteristic parameters of the global simulated flow field described above, and will not be repeated in this embodiment.

[0058] As mentioned above, PIV observations are limited to the stern region and often fail to cover the entire stern area. Furthermore, due to the influence of the rotor hub, blade rotation, and the actual camera field of view, some areas will not be observed using PIV technology, forming observation blind spots. Therefore, the obtained local measured flow field often does not cover the entire stern observation area, but only a portion of the measured flow field. Please refer to [reference needed]. Figure 3 The diagram shows the observation area within the dashed box, which represents a portion of the observation blind zone.

[0059] Therefore, considering the above-mentioned characteristics of actual PIV observations on ships, the sampling conditions given when performing local downsampling include the occlusion area and the sampling method. The occlusion area is used to simulate the blind spot in the observation area at the stern. The occlusion area is related to the specifications and working process of the ship's propeller. The occlusion area can be a regular shape or an irregular shape, and can be set in advance according to the characteristics of the propeller.

[0060] The method for local downsampling of the global simulated flow field under sampling conditions according to the stern observation area includes: performing local downsampling of the global simulated flow field within the sampling area according to the sampling method indicated by the sampling conditions. The sampling area is the region within the stern observation area other than the obstructed area indicated by the sampling conditions. The obstructed area is used to simulate the observation blind zone within the stern observation area. Different sampling conditions include at least one difference between the obstructed area and the sampling method.

[0061] (3) The flow field reconstruction model is obtained by training a neural network based on the characteristic parameters of the low-frequency simulated flow field as input and the characteristic parameters of the global simulated flow field as output.

[0062] Step 3 completes the reconstruction from the stern observation area to the global stern region, and obtains the global reconstructed flow field based on the local measured flow field reconstruction. However, the global reconstructed flow field still contains the influence of the propeller. Therefore, further processing of the global reconstructed flow field is needed to remove the influence of the propeller, thereby obtaining the effective wake of the bare hull.

[0063] Starting from real-scale simulation data, this application first performs real-scale simulation on the ship to obtain the nominal wake field of the bare hull. The nominal wake field of the bare hull is a flow field obtained from real-scale simulation data and used to characterize the actual wake velocity experienced by the bare hull.

[0064] Then, the nominal wake field is simulated based on the bare hull using a viscous potential coupled self-propulsion solver that includes a RANS solver and a propeller potential flow solver. The iteration is continued until the output of the viscous potential coupled self-propulsion solver and the globally reconstructed flow field are determined to meet the iteration termination condition. The nominal wake field of the bare hull is then reconstructed using the output of the viscous potential coupled self-propulsion solver. The nominal wake field of the bare hull is used to characterize the effective wake velocity experienced by the bare hull during actual ship navigation.

[0065] The methods for iterative solutions using a viscous potential coupled self-propulsion solver include, please refer to... Figure 4 :

[0066] 1. By applying random perturbations to the characteristic parameters of the nominal wake field of the bare hull using the perturbation parameter ω, a simulated perturbed nominal wake field is generated. The perturbation parameter ω is the parameter used to apply the perturbation amount, and the amount of perturbation applied can be controlled by the perturbation parameter ω.

[0067] First, a spatial characteristic kernel function of the three-dimensional velocity field of the simulated nominal wake field of the bare ship hull is constructed based on kernel functions such as Gaussian kernel or exponential kernel. Then, singular value decomposition (SVD) is performed on the simulated nominal wake field of the bare ship hull to obtain the characteristic parameters of the simulated nominal wake field of the bare ship hull. Finally, random perturbation is applied to the characteristic parameters of the simulated nominal wake field of the bare ship hull.

[0068] In one embodiment, a random perturbation of Δω(x) = F(G(x)) is applied to the characteristic parameters of the simulated nominal wake field of the bare hull using a perturbation parameter ω. Here, G(x) is a Gaussian random process with mean μ and kernel function K(x,x′), F(G(x)) represents the result of mapping G(x) using a distribution-constrained mapping operator, and x and x′ represent two spatial coordinates. In one embodiment, the exponential kernel function used is:

[0069]

[0070] Where σ(x) is the characteristic distribution of a certain flow feature, including velocity distribution and vorticity distribution. l is the characteristic scale, which can be determined according to the key flow feature scale of a specific case, such as the radius of the propeller.

[0071] 2. The viscous potential coupled self-propulsion solver is used to solve the nominal wake field based on the simulated disturbance, and outputs the total velocity field U at the propeller disk surface, the solved values ​​of the ship's macroscopic motion parameters S, and the effective wake velocity U. effective Among these, the types of macroscopic motion parameters of a ship include propeller thrust, torque, hull resistance, and the power received by the propeller.

[0072] The viscopotential coupled self-propulsion solver includes a RANS solver and a propeller potential flow solver. The RANS solver is used to solve the total velocity field at the propeller disk surface and the hull drag of the bare hull under free surface conditions. The propeller potential flow solver can solve the propeller forces and the induced velocity u generated by the propeller under a given input wake. The propeller forces include propeller thrust and torque. The propeller potential flow solver can perform calculations using methods such as lift lines, lift surfaces, or the element method.

[0073] Regardless of the calculation method used by the propeller potential flow solver, the effect of propeller rotation on the flow is applied to the RANS solver through a throbbing disk model. By combining the propeller potential flow solver and the RANS solver, the viscous-potential coupled self-propulsion solver can effectively solve for the effective wake velocity U. effectiveIn other words, it is possible to distinguish the velocity induced by the propeller in the total velocity field U at the propeller disk surface of the ship when it is in propeller mode, that is, to distinguish the induced velocity u induced by the propeller. Then, by calculating the difference between the total velocity field U at the propeller disk surface and the induced velocity u, the effective wake velocity U can be calculated. effective =Uu.

[0074] In the solution process of the viscous potential coupled self-propulsion solver, the effective wake velocity U effective Since it cannot be explicitly solved, an iterative solution method is required. Methods for obtaining the effective wake velocity using a viscous-potential coupled self-propulsion solver include, please refer to... Figure 5 :

[0075] (1) Determine the propeller thrust. In the first loop, the propeller thrust is initialized. The initialized propeller thrust can be arbitrarily given or an empirical value can be used. During the iterative solution process, the propeller thrust will eventually converge to an accurate value, so the initialized propeller thrust does not need to be accurate. For example, the propeller thrust can be initialized as the hull resistance, similar to the common volume force distribution. After initializing the propeller thrust, the corresponding volume force distribution can be obtained.

[0076] (2) Determine the corresponding volume force distribution based on the propeller thrust. This step is a conventional technique and will not be described in detail in this embodiment.

[0077] (3) The volume force distribution is substituted into the blasting disk model inside the RANS solver. The RANS solver is then used to solve the nominal wake field based on the simulated disturbance to obtain the hull resistance and the total velocity field U at the propeller disk surface. The blasting disk model is used to represent the influence of propeller rotation on the flow. In practice, the RANS solver also needs to combine the corresponding navigation parameters when performing the solution, which will not be elaborated in this embodiment.

[0078] (4) Substitute the circumferential average value of the total velocity field U at the propeller disk surface as the resultant propeller velocity value into the propeller potential flow solver for solution. Under the condition of equal total thrust, calculate the radial circulation distribution of the propeller to obtain the propeller induced velocity, torque, and body force distribution. The difference between the calculated total velocity field U at the propeller disk surface and the propeller induced velocity u is the effective wake velocity U. effective The propeller potential flow solver also needs to consider the relevant navigation parameters and propeller geometry when performing the solution, which will not be elaborated on in this embodiment.

[0079] (5) When the calculated torque and hull resistance converge, the propeller thrust is interpolated. If the propeller thrust also converges, the iteration ends. If the propeller thrust does not converge, the propeller thrust is updated and the above steps (2)-(4) are re-executed using the updated propeller thrust.

[0080] (6) When at least one of the calculated torque and hull resistance fails to converge, the volume force distribution obtained by the propeller potential flow solver is used to update the volume force distribution, and the updated volume force distribution is used to perform the above steps (3) and (4).

[0081] 3. Based on the total velocity field U at the propeller disk surface and the globally reconstructed flow field U * The error between the calculated value S of the ship's macroscopic motion parameters and the actual value S collected from the data. * The error between the parameters is used to perform Bayesian updates on the parameters of the Gaussian random process G(x) through Kalman filtering.

[0082] 4. Based on the perturbation Δω(x) = F(G(x)) determined by the updated Gaussian stochastic process G(x), the perturbation parameters are used again to apply the perturbation Δω(x) = F(G(x)) to the characteristic parameters of the simulated nominal wake field of the bare hull to generate a simulated perturbed nominal wake field until the total velocity field U at the propeller disk surface matches the globally reconstructed flow field U. * The error between them is within the corresponding error range, and the solved value S of the ship's macroscopic motion parameter is consistent with the actual value S of the collected ship's macroscopic motion parameter. * When the error between the two sides is within the corresponding error range, the iteration termination condition is determined to be met.

[0083] At this point, the effective wake velocity U output by the viscous-potential coupled self-propulsion solver is obtained. effective The nominal accompanying flow field of the bare hull is reconstructed.

[0084] The above descriptions are merely preferred embodiments of this application, and this application is not limited to the above embodiments. It is understood that other improvements and variations that can be directly derived or conceived by those skilled in the art without departing from the spirit and concept of this application should be considered to be included within the protection scope of this application.

Claims

1. A method for reconstructing the wake field of a bare ship hull based on finite indirect observation, characterized in that, The method includes: The local measured flow field of the stern observation area of ​​a ship during actual ship navigation was obtained using PIV technology. The stern observation area is a local area within the global stern region of the ship. The global reconstructed flow field of the stern region of the ship during actual navigation is obtained by using the flow field reconstruction model based on the local measured flow field. The flow field reconstruction model is used to reflect the mapping relationship between the characteristic parameters of the local region of the ship during actual navigation and the characteristic parameters of the global region of the stern of the ship during actual navigation. The flow field reconstruction model is obtained by training a neural network using the actual scale simulation data of the ship. The nominal wake field of the bare hull is obtained by performing real-scale simulation data on the ship. The viscous potential coupled self-propulsion solver, which includes a RANS solver and a propeller potential flow solver, is used to iteratively solve the nominal wake field of the bare hull until the iteration termination condition is met based on the output of the viscous potential coupled self-propulsion solver and the global reconstructed flow field. At this point, the nominal wake field of the bare hull of the ship is reconstructed using the viscous potential coupled self-propulsion solver. The nominal wake field is used to characterize the actual wake velocity experienced by the bare hull of the ship. The method for iterative solution using a visco-potential coupled self-propulsion solver includes: applying random perturbations to the characteristic parameters of the simulated nominal wake field of the bare hull using perturbation parameters to generate a simulated perturbed nominal wake field; using the visco-potential coupled self-propulsion solver to solve based on the simulated perturbed nominal wake field; outputting the total velocity field at the propeller disk surface, the solved values ​​of the ship's macroscopic motion parameters, and the actual wake velocity; and applying Kalman filtering based on the error between the total velocity field at the propeller disk surface and the globally reconstructed flow field, and the error between the solved values ​​of the ship's macroscopic motion parameters and the actual values ​​of the collected ship's macroscopic motion parameters. The applied perturbation is updated using Bayesian methods. The updated perturbation parameters are then used to re-execute the step of applying random perturbations to the characteristic parameters of the simulated nominal wake field of the bare hull using the perturbation parameters to generate a simulated perturbed nominal wake field. This process continues until the error between the total velocity field at the propeller disk and the globally reconstructed flow field is within the corresponding error range, and the error between the solved value of the ship's macroscopic motion parameters and the actual value of the collected ship's macroscopic motion parameters is within the corresponding error range. At this point, the reconstructed nominal wake field of the bare hull, composed of the effective wake velocity output by the viscous potential coupled self-propulsion solver, is obtained.

2. The method for reconstructing the wake field of a bare ship hull according to claim 1, characterized in that, The method for solving the nominal wake field of the simulated disturbance using the viscous potential coupled self-propulsion solver includes: Determine the propeller thrust and determine the corresponding volume force distribution based on the propeller thrust; The volume force distribution is substituted into the blasting disk model inside the RANS solver, and the RANS solver is used to solve the nominal wake field based on the simulated disturbance to obtain the total velocity field at the propeller disk surface. The blasting disk model is used to represent the influence of propeller rotation on the flow. The calculated circumferential average value of the total velocity field at the propeller disk surface is substituted into the propeller potential flow solver as the resultant velocity value of the propeller to obtain the propeller induced velocity. The propeller thrust and volume force distribution are iteratively updated to solve the problem until the solution values ​​of the ship's macroscopic motion parameters converge. Then, the calculated total velocity field at the propeller disk surface, the solution values ​​of the ship's macroscopic motion parameters, and the effective wake velocity are output. The effective wake velocity is the difference between the total velocity field at the propeller disk surface and the propeller induced velocity.

3. The method for reconstructing the wake field of a bare ship hull according to claim 2, characterized in that, The solved values ​​of the ship's macroscopic motion parameters include hull resistance, torque, and propeller thrust. The iterative update of the propeller thrust and the volume force distribution includes the following: If at least one of the hull resistance obtained by the solution and the torque obtained by the propeller potential flow solver fails to converge, the volume force distribution obtained by the propeller potential flow solver is used as the updated volume force distribution, and the step of substituting the volume force distribution into the pulsating disk model inside the RANS solver is executed again. When the hull resistance and the torque obtained by the propeller potential flow solver both converge, the propeller thrust is updated by interpolation. When the propeller thrust also converges, the iterative solution ends. If the propeller thrust does not converge, the step of determining the corresponding volume force distribution based on the propeller thrust is executed again using the updated propeller thrust.

4. The method for reconstructing the wake field of a bare ship hull according to claim 1, characterized in that, The method for applying random perturbations to the characteristic parameters of the simulated nominal wake field of the bare hull includes: The perturbation amount applied to the characteristic parameters of the simulated nominal wake field of the bare hull is: random perturbations, of which It is a mean Kernel function is Gaussian random process, This indicates the use of distribution-constrained mapping operators to represent pairs The result after mapping and These are spatial location coordinates.

5. The method for reconstructing the wake field of a bare ship hull according to claim 1, characterized in that, The methods for training the flow field reconstruction model include: A real-scale numerical simulation of the ship under self-propelled conditions was performed to obtain the global simulated flow field of the global region at the stern of the ship, and the characteristic parameters of the global simulated flow field were extracted. According to the observation area at the stern, the global simulated flow field is locally downsampled under different sampling conditions to obtain the low-frequency simulated flow field under different sampling conditions, and the characteristic parameters of each low-frequency simulated flow field are extracted respectively. The flow field reconstruction model is obtained by training a neural network using the characteristic parameters of the low-frequency simulated flow field as input and the characteristic parameters of the global simulated flow field as output.

6. The method for reconstructing the wake field of a bare ship hull according to claim 5, characterized in that, Sampling conditions include the obstruction area and the sampling method. Different sampling conditions include at least one difference between the obstruction area and the sampling method. The method for local downsampling of the global simulated flow field according to the stern observation area under the sampling conditions includes: The global simulated flow field is locally downsampled within the sampling area according to the sampling method indicated by the sampling conditions. The sampling area is the area other than the occlusion area indicated by the sampling conditions within the stern observation area. The occlusion area is used to simulate the observation blind zone within the stern observation area.

7. The method for reconstructing the wake field of a bare ship hull according to claim 5, characterized in that, The method for obtaining the global simulated flow field of the stern region of the ship includes: A full-scale numerical simulation was performed on the ship to obtain unsteady flow field data of the global stern region under different propeller rotation phases; Phase averaging is performed on unsteady flow field data under multiple different propeller rotation phases to obtain the global simulated flow field of the stern region of the ship.

8. The method for reconstructing the wake field of a bare ship hull according to claim 1, characterized in that, The methods for obtaining the global reconstructed flow field include: The flow characteristics of the local measured flow field are extracted, and the global reconstructed flow field is determined based on the flow field reconstruction model by perturbation of the feature parameters. The approximation error between the feature parameters of the global reconstructed flow field and the flow characteristics of the local measured flow field is within the corresponding error range. Furthermore, the feature parameters of the local region flow field obtained by mapping the feature parameters of the global reconstructed flow field through the flow field reconstruction model are also within the corresponding error range of the approximation error between the feature parameters of the local measured flow field and the flow characteristics of the local measured flow field.

9. The method for reconstructing the wake field of a bare ship hull according to claim 5, characterized in that, For any set of simulated flow fields among the global simulated flow field and each set of low-frequency simulated flow fields, the methods for extracting the characteristic parameters of the simulated flow field include: The simulated flow field is subjected to eigendecomposition to obtain eigenmodes and eigendecomposition coefficients. The eigendecomposition methods used include KL decomposition, POD decomposition, and DMD decomposition.

Images