Data-driven underwater navigation body pump-jet propeller multi-precision rapid optimization method

Abstract

The invention relates to a data-driven underwater navigation body pump-jet propeller multi-precision rapid optimization method, and relates to the field of fluid machinery design and optimization. The method comprises the following steps: firstly, carrying out parametric modeling on the rotor blade of the pump-jet propeller by adopting a B spline curve; then constructing an updatable low-precision hydrodynamic performance prediction model based on a modified leaf element momentum theory, wherein the model contains a correction coefficient; a low-precision model is utilized to quickly generate an initial sample, a Gaussian process agent model is constructed, candidate design points are selected through a Bayesian optimization collection function, and after high-precision CFD verification, a low-precision model coefficient is dynamically corrected through a simulation result, and the agent model is updated; and finally, outputting an optimal pump-jet propeller design scheme which meets thrust and torque constraints and has the maximum propelling efficiency. On the premise that the precision is not sacrificed, the calculation cost and the time period of design optimization of the pump-jet propeller are greatly reduced, and the problem that efficiency and precision cannot be considered in a traditional method is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of fluid machinery design and optimization, in particular to a rapid optimization design method for underwater vehicle propulsor, and more particularly to a multi-precision rapid optimization method for pump-jet propulsor. BACKGROUND

[0002] Underwater vehicles, such as submarines, are important ocean engineering equipment. The performance of its propulsor, especially the pump-jet propulsor (PJP), directly affects the hydrodynamic characteristics and noise level of the vehicle. Pump-jet propulsors usually work in the non-uniform wake formed by the vehicle tail, and this complex flow field environment poses great challenges to the design and optimization of the propulsor.

[0003] Currently, the design and optimization of pump-jet propulsors mainly rely on two types of methods: high-precision numerical simulation and low-precision rapid evaluation model. High-precision methods usually use computational fluid dynamics (CFD) technology based on Reynolds-averaged Navier-Stokes (RANS) equations. This type of method can accurately simulate complex flow field details and provide reliable basis for performance evaluation. However, its main defect is the extremely high computational cost. For example, for a fine transient flow field analysis of a pump-jet propulsor, the calculation time can be more than a month. Even a single steady-state CFD calculation requires about 10 hours on a high-performance workstation with 48 cores. This high cost of time and computational resources makes it difficult for CFD methods to adapt to the large number of iterative rounds required in the design process of the propulsor, severely restricting the design efficiency and exploration of the optimization space, and constituting the main bottleneck of the design cycle. Low-precision methods, such as blade element momentum theory (BEMT), are based on physical principles and simplified assumptions, and can achieve rapid evaluation of the hydrodynamic performance of the propulsor with extremely high computational efficiency. However, the prediction accuracy of this type of method is often insufficient to meet the requirements of engineering design. Standard BEMT models usually assume that the incoming flow is uniform, and cannot accurately account for the non-uniform wake effects caused by the hull, appendages, etc., and also ignore the complex interactions between the internal ducts, stators, etc. of the pump-jet propulsor. Data shows that when predicting the performance of a certain public pump-jet propulsor at the design condition using the standard BEMT method, the prediction error of the thrust coefficient is as high as 24.73%, and the prediction error of the efficiency is also as high as 8.11%, which is obviously insufficient to guide fine design.

[0004] Therefore, existing technologies face a dilemma: pursuing accuracy at the expense of efficiency, or pursuing efficiency at the expense of accuracy. There is an urgent need in the field for a novel optimization method that can balance computational efficiency and prediction accuracy, in order to overcome the bottlenecks of traditional design methods and achieve efficient and low-cost optimization design of pump-jet propulsion systems under real-world operating conditions. Summary of the Invention

[0005] In view of the shortcomings of existing technologies, this application provides a data-driven method for rapid optimization of pump-jet propulsion systems for underwater vehicles with high precision. This invention significantly reduces the computational cost and time cycle in the design and optimization process of pump-jet propulsion systems without sacrificing accuracy, solving the problem that traditional methods cannot balance efficiency and accuracy.

[0006] The technical means employed in this invention are as follows: A data-driven method for rapid multi-precision optimization of underwater vehicle pump-jet propulsion systems includes the following steps: Curve parametric technology is used to abstract the geometry of key components of pump-jet propulsion into finite design variables, forming an adjustable design space; A low-precision hydrodynamic performance prediction model containing at least one updatable correction coefficient is established, which is used to quickly evaluate design samples. Within the design space, a preliminary evaluation is conducted using a low-precision hydrodynamic performance prediction model, and a surrogate model is constructed. Candidate design points are selected through optimization algorithms. After verification by high-precision numerical simulation, the correction coefficients of the low-precision model and the surrogate model are updated with the high-precision results. This process is repeated until the convergence condition is met. Based on the corrected low-precision model after iterative convergence, and combined with the optimization algorithm, the optimal pump-jet propulsion design scheme that meets the performance objectives and constraints is searched in the design space and output.

[0007] Furthermore, curve parametric technology is used to abstract the geometry of key components of the pump-jet propulsion system into finite design variables, forming an adjustable design space, including: B-spline curves are used to parameterize the chord length distribution and pitch angle distribution of the rotor blades of the pump-jet propulsion unit. The beginning and end of the curve are fixed, and two design points are selected for each parameter between the beginning and end. The ordinates of the two design points are used as design variables to construct an adjustable design space.

[0008] Furthermore, the low-precision hydrodynamic performance prediction model is based on the modified blade element momentum theory. It compensates for the interference effect of non-uniform flow by introducing a correction coefficient. The correction coefficient is updated linearly in each iteration using high-precision numerical simulation results, thereby dynamically improving the prediction accuracy of the low-precision model.

[0009] Furthermore, the correction coefficients include a velocity correction coefficient and a thrust ratio correction coefficient.

[0010] Furthermore, a preliminary evaluation is conducted within the design space using a low-precision hydrodynamic performance prediction model, and a surrogate model is constructed. This includes: generating an initial sample set within the design space, evaluating the performance of the samples using a modifiable low-precision hydrodynamic performance prediction model, and constructing a surrogate model between the design variables and the performance target based on the evaluation results. The surrogate model is a Gaussian process regression model.

[0011] Furthermore, the selection of candidate design points through an optimization algorithm includes: based on the surrogate model, selecting one or more optimal candidate design points using the acquisition function in the optimization algorithm for high-precision evaluation, wherein the optimization algorithm is a Bayesian optimization algorithm and the acquisition function is a constrained expectation improved acquisition function.

[0012] Furthermore, after verification by high-precision numerical simulation, the correction coefficients of the low-precision model and the surrogate model are updated with the high-precision results, including: performing high-precision CFD numerical simulation on the selected candidate design points to obtain accurate hydrodynamic performance data; and using the accurate hydrodynamic performance data to update the correction coefficients in the correctable low-precision hydrodynamic performance prediction model to improve the prediction accuracy of the low-precision model in subsequent iterations.

[0013] Furthermore, using precise hydrodynamic performance data, the correction coefficients in the correctable low-precision hydrodynamic performance prediction model are updated, including: Calculate the speed correction factor using the following formula. and thrust ratio correction factor :

[0014]

[0015] in, This is the corrected MBEMT incoming flow rate. It is rotor thrust. It is the total thrust.

[0016] Furthermore, based on the iteratively converged modified low-precision model, and combined with optimization algorithms, the optimal pump-jet propulsion design scheme that meets the performance objectives and constraints is searched within the design space and output, including: The optimization objective is set to maximize the sum of the propulsion efficiency of the pump-jet propulsion system under two representative operating conditions; Set the constraint to the range of torque coefficient variation under two representative operating conditions.

[0017] Compared with the prior art, the present invention has the following advantages: 1. This invention explores a large-scale design space using a low-precision model and intelligently selects a very small number of the most valuable design points for time-consuming high-precision CFD simulation using a Bayesian optimization algorithm. This multi-precision collaborative strategy significantly reduces the reliance on high-cost CFD simulation, thereby reducing the total computational cost and time cycle of the optimization design by several orders of magnitude.

[0018] 2. This invention does not simply replace a high-precision model with a low-precision model, but rather creates a dynamic feedback correction loop. By continuously using high-precision CFD data to correct the low-precision model during the iteration process, it ensures that the optimization direction always converges towards the optimal solution in the real physical world, overcoming the fundamental defect of inaccurate predictions in traditional low-precision models, and guaranteeing the accuracy and engineering applicability of the final optimization results.

[0019] 3. Through efficient global optimization, this invention can find better designs in complex design spaces that are difficult to discover using traditional trial and error methods or local optimization.

[0020] The embodiments show that the pump-jet propulsion system optimized using the method of the present invention improves the average propulsion efficiency by 3.05% and the thrust by 1.03% while satisfying the torque constraint, and at the same time reduces the input power requirement by 1.94%, achieving significant performance gains. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart of a data-driven method for rapid multi-precision optimization of underwater vehicle pump-jet propulsion.

[0023] Figure 2 This is a specific execution flow of a data-driven multi-precision rapid optimization method for underwater vehicle pump-jet propulsion in an embodiment of the present invention.

[0024] Figure 3a This is a schematic diagram illustrating the parametric modeling of the chord length of the pump-jet propeller blade using B-spline curves in an embodiment of the present invention.

[0025] Figure 3b This is a schematic diagram illustrating the parametric modeling of the pitch angle distribution of the pump-jet propeller blades using B-spline curves in an embodiment of the present invention.

[0026] Figure 4This is a flowchart of the MBEMT model based on finite-point linear correction in an embodiment of the present invention.

[0027] Figure 5 This is a schematic diagram comparing the open water characteristics calculated by MBEMT after linear correction under the conditions of pump-jet propulsion experiment and rotor-stator interface flow in an embodiment of the present invention.

[0028] Figure 6a The geometric model of the front pump-jet propulsion unit is optimized for this embodiment of the invention.

[0029] Figure 6b This is the optimized geometric model of the pump-jet propulsion system in this embodiment of the invention.

[0030] Figure 7 This is a comparison chart of the performance curves of the pump-jet propulsion efficiency as a function of the advance rate coefficient before and after optimization in an embodiment of the present invention.

[0031] Figure 8a This is a comparison of absolute pressure cloud diagrams at different blade height sections for the optimized front pump-jet propulsion system in this embodiment of the invention.

[0032] Figure 8b This is a comparison of absolute pressure cloud diagrams of the optimized pump-jet propulsion unit at different blade height sections in an embodiment of the present invention. Detailed Implementation

[0033] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0034] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0035] like Figure 1As shown, the present invention provides a data-driven method for rapid multi-precision optimization of underwater vehicle pump-jet propulsion systems, comprising the following steps.

[0036] S1. By using curve parametric technology, the geometry of key components of the pump-jet propulsion system is abstracted into finite design variables, forming an adjustable design space.

[0037] This invention employs parametric modeling technology to define geometric design variables to be optimized. These design variables control the geometry of key components of the pump-jet propulsion system, such as the chord length distribution and pitch angle distribution of the rotor blades. Specifically, the parametric modeling technology uses B-spline curves, and the chord length and pitch angle distribution of the rotor blades are altered by controlling a few control points on the B-spline curves.

[0038] S2. Establish a low-precision hydrodynamic performance prediction model containing at least one updatable correction coefficient, which is used to quickly evaluate design samples.

[0039] Specifically, the correctable low-precision hydrodynamic performance prediction model is built based on the Corrected Modified Blade Element Momentum Theory (CMBEMT). This model introduces corrections that take into account non-uniform inflow and includes at least one velocity correction coefficient and one thrust percentage correction coefficient, which are dynamically corrected by high-precision CFD simulation results during the optimization iteration process.

[0040] S3. Within the design space, a preliminary evaluation is performed using a low-precision hydrodynamic performance prediction model, and a surrogate model is constructed. Candidate design points are selected through optimization algorithms. After verification by high-precision numerical simulation, the correction coefficients of the low-precision model and the surrogate model are updated with the high-precision results. This process is repeated until the convergence condition is met. Specifically, this includes the following steps: a. Low-precision evaluation and surrogate modeling: An initial sample set is generated within the design space. The performance of the samples is evaluated using the correctable low-precision model, and a surrogate model between the design variables and the performance objective is constructed based on the evaluation results. Preferably, the surrogate model is a Gaussian Process Regression (GPR) model.

[0041] b. Candidate Point Selection: Based on the surrogate model, one or more optimal candidate design points are selected using the acquisition function in the optimization algorithm for high-precision evaluation. This acquisition function aims to balance the exploration of the design space with the utilization of known optimal regions. Preferably, the optimization algorithm is a Bayesian optimization (BO) algorithm. The acquisition function is a Constrained Expected Improvement (CEI) acquisition function.

[0042] c. High-precision evaluation: Perform high-precision numerical simulations, such as CFD simulations, on the selected candidate design points to obtain accurate hydrodynamic performance data.

[0043] d. Low-precision model correction: Using the high-precision performance data obtained in step c, update the correction coefficients in the correctable low-precision model to improve the prediction accuracy of the low-precision model in subsequent iterations.

[0044] e. Proxy Model Update and Convergence Judgment: Add the high-precision evaluation results to the sample set, update the proxy model, and determine whether the optimization process has converged or reached the preset number of iterations. If not, return to step b to continue iterating; if yes, terminate the loop.

[0045] S4. Based on the corrected low-precision model after iterative convergence, and combined with the optimization algorithm, the optimal pump-jet propulsion design scheme that meets the performance target and constraints is searched in the design space and output.

[0046] In this application, the performance optimization objective is to maximize the propulsion efficiency of the pump-jet propulsion system under one or more operating conditions, and may include constraints on performance parameters such as torque.

[0047] The following specific application examples will further illustrate the solution and effects of the present invention.

[0048] like Figure 2As shown, the overall framework of this invention (CMBEMTCBO) is a multi-precision collaborative optimization process. This process first uses high-precision CFD results to initially correct a low-precision CMBEMT model. Then, based on a parameterized pump-jet propulsion model, initial samples are generated within the design space, and the corrected CMBEMT model is used for rapid hydrodynamic performance evaluation. Next, a surrogate model relationship between design variables and propulsion efficiency is constructed using a Gaussian process (GP) model. Then, the constraint expectation in Bayesian optimization (BO) is used to improve the acquisition function to intelligently select the next most valuable candidate point, and its performance is verified by high-precision CFD simulation. Finally, based on the CFD feedback results, the surrogate model is updated to guide the next round of optimization; furthermore, as a preferred method of this invention, the correction coefficients of the CMBEMT model are further updated using the high-precision data points, achieving online iterative correction of the low-precision model. This cycle continues until the convergence condition is met, ultimately outputting the optimal design.

[0049] The core idea of ​​this framework is to establish a dynamic feedback loop: using expensive but accurate high-precision information to continuously and strategically "teach" a cheap but fast low-precision model, making the entire optimization search process both intelligent and efficient.

[0050] First, parametric geometric modeling is performed. To optimize the geometry of the pump-jet propulsion system, a parametric model capable of describing its complex shape with a small number of design variables is needed. In this embodiment, the focus of optimization is on the rotor blades, whose key geometric parameters are the radial chord length distribution and pitch angle distribution.

[0051] Reference Figure 3a and Figure 3b In this embodiment, a third-order B-spline curve is used to parameterize the chord length and pitch angle distributions. By setting four control points, a smooth and continuous distribution curve can be generated. To ensure design flexibility while controlling the dimensionality of the optimization problem, the first and last two control points of the curve are fixed (i.e., keeping the geometric parameters of the blade root and blade tip unchanged), and only the ordinates of the two middle control points (i.e., chord length or pitch angle values) are used as design variables. Therefore, the chord length distribution is controlled by two variables, and the pitch angle distribution is controlled by two variables, forming a total four-dimensional design space. This method can effectively explore diverse leaf geometries while avoiding the "curse of dimensionality" caused by too many dependent variables.

[0052] Subsequently, a multi-precision hydrodynamic model is constructed, with the high-precision model serving as the "benchmark" or "truth" for performance evaluation. This embodiment employs CFD numerical simulation based on the RANS equations, and the turbulence model selected is k... ωSST model. Through fine mesh generation (e.g., approximately 4.8 × 10⁻⁶ meshes in the rotor domain). The stator domain is approximately 6.1 × The model employs rigorous solution settings to ensure the accuracy of the calculation results. The simulation results agree well with the experimental data, and the efficiency prediction error is within 2%. Although accurate, the model is computationally time-consuming; therefore, it is only used in the optimization loop to evaluate a few key points selected by the Bayesian optimization algorithm.

[0053] Low-precision models are the foundation for achieving rapid optimization. This invention makes two key improvements to the traditional leaf element momentum theory (BEMT) and constructs the CMBEMT model.

[0054] The first improvement was the development of the Modified Blade Element Momentum Theory (MBEMT), which allows it to account for non-uniform incoming flow. Standard BEMT assumes that the incoming flow velocity is uniform across the entire propeller disk surface. However, in reality, due to the hull shielding effect, the incoming flow velocity varies radially. MBEMT replaces the uniform incoming flow assumption by introducing radially varying axial and circumferential velocities, updating the velocity relationships on the blade element section as follows:

[0055]

[0056] in, Indicates the corrected radius The axial average velocity at that location Indicates the corrected radius Circumferential average velocity at that location This represents the extraction of the flow field at the radius. The axial average velocity at that location This represents the extraction of the flow field at the radius. Circumferential average velocity at that location Indicates rotational speed. Indicates the rotor radius. Indicates the axial induction factor. This represents the circumferential induction factor. This improvement allows the model to account for the major effects of non-uniform wake flow.

[0057] The second improvement involves introducing a linear correction mechanism, constructing the CMBEMT model. Even though MBEMT considers non-uniform inflow, it still neglects the complex influence of components such as ducts and stators on the flow field. To compensate for this deficiency, this invention introduces two linear correction coefficients: a velocity correction coefficient... and thrust ratio correction factor .

[0058]

[0059]

[0060] in, It is the incoming flow velocity extracted from high-precision model numerical simulation. The rotor thrust is obtained through high-precision model simulation. It is the total thrust obtained through high-precision model simulation. This refers to the corrected MBEMT inflow velocity. The scheme uses a given initial value h, continuously adjusts h to obtain vz', substitutes it into the MBEMT calculation, and compares it with the CFD results, gradually bringing the MBEMT results closer to the CFD results. The specific correction process is as follows: Figure 4 As shown: First, obtain the non-uniform inflow distribution based on CFD data, that is... The thrust and torque were calculated using the BEMT under the inflow conditions. An initial correction factor h was used to adjust the inflow velocity so that the torque calculated by BEMT matched the CFD results. Then, the torque and thrust predictions were updated, and the correction factor k was obtained by comparing the CFD thrust with the BEMT-calculated thrust. Linear fitting was performed on the correction factors obtained from a small amount of CFD data (two data points) to obtain the correction factors for varying operating conditions, thus establishing a finite-point linearly corrected MBEMT model. The resulting open-water performance characteristic curves of the pump-jet propulsion system are shown in the attached figure, using only the CFD results from the first and last two inflow rate points to determine the thrust ratio k and the velocity correction factor h, and performing linear correction on the MBEMT. Figure 5 As shown in the figure. The results indicate that the thrust coefficient and torque coefficient are basically consistent with the experimental results, and the trend of propulsion efficiency is also consistent with the experimental results, with only a certain deviation in numerical values, showing good overall agreement. The physical significance of these two coefficients lies in the fact that they quantify the overall deceleration effect of the duct / stator on the incoming flow and its contribution to the total thrust, respectively.

[0061] Before optimization begins, a small number (e.g., two) of CFD calculation points can be used to initially determine the optimal parameters. and With advance coefficient The linear relationship. Precession coefficient. The calculation formula is:

[0062] in, It is the far-field incoming flow velocity. It's the rotational speed. It is the rotor diameter.

[0063] More importantly, during the optimization iteration process, each new CFD calculation result is used to update and improve these two correction coefficients, thereby dynamically improving the prediction accuracy of the CMBEMT model.

[0064] As shown in Table 1 below, the effects of these improvements are significant. The thrust prediction error of the standard BEMT is as high as 24.73%, while the MBEMT, which takes into account the non-uniform inflow, reduces the error to 8.97%. Through further corrections by CMBEMT, the prediction results of the low-precision model can be made to highly match the prediction results of the high-precision CFD.

[0065] Table 1: Comparison of prediction accuracy of different low-precision models ( =0.8 operating condition)

[0066] Furthermore, combining the Bayesian optimization engine with iterative optimization of CMBEMT, Bayesian optimization (BO) serves as the core of intelligent decision-making, connecting low-precision exploration with high-precision verification. It primarily consists of two parts: a surrogate model and a data acquisition function. Surrogate Model: This embodiment employs a Gaussian process (GP) model. Based on existing sample points (design variables - performance), the GP model constructs a probabilistic response surface about the objective function (e.g., propulsion efficiency). It can not only predict the mean performance of any unknown point but also provide the uncertainty of the prediction. Data Acquisition Function: This embodiment uses a constrained expected improvement (CEI) data acquisition function. This function utilizes the mean and uncertainty information provided by the GP model to calculate the "value" of each point in the design space—that is, the expected improvement of that point compared to the current optimal value under the constraints. The optimization algorithm selects the point with the highest "value" as the next candidate point for high-precision CFD evaluation. This mechanism enables the algorithm to make intelligent trade-offs between regions with high uncertainty (exploration) and regions with good predictive performance (utilization).

[0067] Finally, the optimized CMBEMT was applied to a practical optimization problem. The optimization objective of this implementation was to maximize the pump-jet propulsion system under two representative operating conditions (advance coefficient). =0.7 and The sum of propulsion efficiencies under the condition of (=0.8). Meanwhile, to ensure that the optimized thruster input power does not change drastically, a constraint is set: the torque coefficient variation range under the two operating conditions must not exceed ±3% of the original design. The optimization problem can be expressed as:

[0068]

[0069]

[0070] When the optimization process begins, 80 initial sample points are generated first through Latin hypercube sampling. An initialized CMBEMT model is then used for rapid evaluation, and an initial GP surrogate model is constructed. Subsequently, the BO iterative loop begins. After multiple iterations, the optimization process converges, yielding the optimal design.

[0071] Figure 6a and Figure 6b The diagram shows a geometric comparison of the rotor blades before and after optimization. It can be observed that the optimized blade chord length distribution changes from the original "wider in the middle and narrower at both ends" to a monotonically decreasing trend from the blade root to the blade tip. The pitch angle distribution changes relatively little, mainly because torque is very sensitive to pitch angle changes, and the adjustment space is limited to meet strict torque constraints.

[0072] The optimized design improved the average thrust coefficient by 1.03% and reduced the average torque coefficient by 1.94% under both operating conditions, ultimately resulting in a 3.05% increase in average propulsion efficiency. This demonstrates that the method successfully improved the thrust and efficiency of the thruster while reducing the input power requirement.

[0073] Figure 7 The efficiency performance curves of the pump-jet propulsion system before and after optimization were compared over a wider operating range. It can be seen that the optimized design not only performs better at the two design points (=0.7, 0.8), but also shows a more significant efficiency improvement under off-design conditions with higher advance coefficients, indicating that the optimized design has good overall performance and robustness.

[0074] Figure 8a and Figure 8b By comparing the absolute pressure cloud maps of blade sections at different radial positions before and after optimization, the physical mechanism of the performance improvement was revealed. The optimized blade exhibited a significantly increased pressure difference between the pressure surface (high-pressure zone) and the suction surface (low-pressure zone). According to fluid dynamics principles, a larger pressure difference means the blade can generate greater thrust, which is entirely consistent with the improvement in macroscopic performance indicators (increased thrust), thus verifying the physical rationality of the optimization results.

[0075] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A data-driven method for rapid multi-precision optimization of underwater vehicle pump-jet propulsion systems, characterized in that, Includes the following steps: Curve parametric technology is used to abstract the geometry of key components of pump-jet propulsion into finite design variables, forming an adjustable design space; A low-precision hydrodynamic performance prediction model containing at least one updatable correction coefficient is established, which is used to quickly evaluate design samples. Within the design space, a preliminary evaluation is conducted using a low-precision hydrodynamic performance prediction model, and a surrogate model is constructed. Candidate design points are selected through optimization algorithms. After verification by high-precision numerical simulation, the correction coefficients of the low-precision model and the surrogate model are updated with the high-precision results. This process is repeated until the convergence condition is met. Based on the corrected low-precision model after iterative convergence, and combined with the optimization algorithm, the optimal pump-jet propulsion design scheme that meets the performance objectives and constraints is searched in the design space and output.

2. The data-driven multi-precision rapid optimization method for underwater vehicle pump-jet propulsion systems according to claim 1, characterized in that, Curve parametric technology is used to abstract the geometry of key components of the pump-jet propulsion system into finite design variables, forming an adjustable design space, including: B-spline curves are used to parameterize the chord length distribution and pitch angle distribution of the rotor blades of the pump-jet propulsion unit. The beginning and end of the curve are fixed, and two design points are selected for each parameter between the beginning and end. The ordinates of the two design points are used as design variables to construct an adjustable design space.

3. The data-driven multi-precision rapid optimization method for underwater vehicle pump-jet propulsion systems according to claim 1, characterized in that, The low-precision hydrodynamic performance prediction model is based on the modified blade element momentum theory. It compensates for the interference effect of non-uniform flow by introducing a correction coefficient. The correction coefficient is updated linearly in each iteration using high-precision numerical simulation results, thereby dynamically improving the prediction accuracy of the low-precision model.

4. The data-driven multi-precision rapid optimization method for underwater vehicle pump-jet propulsion systems according to claim 3, characterized in that, The correction factors include velocity correction factors and thrust ratio correction factors.

5. The data-driven multi-precision rapid optimization method for underwater vehicle pump-jet propulsion systems according to claim 1, characterized in that, The preliminary evaluation and surrogate model are carried out using a low-precision hydrodynamic performance prediction model within the design space, including: generating an initial sample set within the design space, evaluating the performance of the sample using a modifiable low-precision hydrodynamic performance prediction model, and constructing a surrogate model between the design variables and the performance target based on the evaluation results. The surrogate model is a Gaussian process regression model.

6. The data-driven multi-precision rapid optimization method for underwater vehicle pump-jet propulsion systems according to claim 5, characterized in that, The selection of candidate design points through an optimization algorithm includes: selecting one or more optimal candidate design points based on the surrogate model using a data acquisition function in the optimization algorithm for high-precision evaluation. The optimization algorithm is a Bayesian optimization algorithm, and the data acquisition function is a constrained expectation improved data acquisition function.

7. The data-driven multi-precision rapid optimization method for underwater vehicle pump-jet propulsion systems according to claim 6, characterized in that, After verification by high-precision numerical simulation, the correction coefficients of the low-precision model and the surrogate model are updated with the high-precision results. This includes: performing high-precision CFD numerical simulation on the selected candidate design points to obtain accurate hydrodynamic performance data; and using the accurate hydrodynamic performance data to update the correction coefficients in the correctable low-precision hydrodynamic performance prediction model to improve the prediction accuracy of the low-precision model in subsequent iterations.

8. The data-driven multi-precision rapid optimization method for underwater vehicle pump-jet propulsion systems according to claim 6, characterized in that, Using precise hydrodynamic performance data, update the correction coefficients in the correctable low-precision hydrodynamic performance prediction model, including: Calculate the speed correction factor using the following formula. and thrust ratio correction factor : in, This is the corrected MBEMT incoming flow rate. It is the incoming flow velocity extracted from high-precision model numerical simulation. The rotor thrust is obtained through high-precision model simulation. It is the total thrust obtained through high-precision model simulation.

9. A data-driven multi-precision rapid optimization method for underwater vehicle pump-jet propulsion systems according to claim 6, characterized in that, Based on the iteratively converged modified low-precision model, and combined with optimization algorithms, the optimal pump-jet propulsion design scheme that meets the performance objectives and constraints is searched within the design space and output, including: The optimization objective is set to maximize the sum of the propulsion efficiency of the pump-jet propulsion system under two representative operating conditions; Set the constraint to the range of torque coefficient variation under two representative operating conditions.

Images