An artificial intelligence and multi-objective optimization-based spoiler structure design method and system, a terminal, and a storage medium

Abstract

The application discloses a kind of based on artificial intelligence and multi-objective optimization's spoiler structure design method, system, terminal and storage medium, the method includes: selecting water drop shape spoiler column as optimization starting point, defines control point and describes the profile of spoiler column, forms design variable according to control point, and sets upper and lower limit for each design variable;Three artificial neural network proxy models are constructed, and after design variable is respectively input into three artificial neural network proxy models and is respectively predicted, the pressure drop, average temperature and temperature uniformity coefficient of double-elliptical spoiler structure are output;A multi-objective optimization model is established, a target function is determined, a Pareto optimal solution set is generated based on NSGA-II algorithm, and the most optimal asymmetric double-elliptical structure is screened from the Pareto optimal solution set;Judgment is carried out by introducing comprehensive performance evaluation criteria, and the asymmetric double-elliptical structure meeting the performance requirements is symmetrized to obtain a double-elliptical spoiler structure.The application realizes the high-performance design of spoiler structure.

Description

Technical Field

[0001] This invention relates to the field of turbulence structure design technology, and in particular to a turbulence structure design method, system, terminal, and computer-readable storage medium based on artificial intelligence and multi-objective optimization. Background Technology

[0002] Turbulence structures (such as turbulence columns and fins) are widely used in enhancing heat transfer. Their core function is to improve the heat exchange efficiency between the fluid and the solid wall by disrupting the fluid boundary layer, inducing secondary flow, or generating turbulence. These structures have important applications in microchannel heat sinks, electronic device cooling, and aerospace thermal management.

[0003] With the development of technologies such as high-performance computing and electric vehicles, the demand for efficient heat dissipation is increasing, making the optimized design of turbulence structures a research hotspot. Existing research mainly focuses on turbulence structures with regular geometric shapes, such as circles, triangles, squares, and hexagons. Studies comparing turbulence columns with different cross-sectional shapes (triangular, circular, hexagonal, etc.) have found that circular and hexagonal structures perform better in terms of pressure drop and thermal resistance; teardrop-shaped structures, due to their streamlined design, have also been shown to have good overall performance. However, all of the above-mentioned regular-shaped turbulence structures have significant drawbacks: enhanced heat transfer is often accompanied by a high pressure drop, leading to increased energy consumption; the fixed geometry makes it difficult to achieve an optimal balance between heat transfer and flow resistance through simple adjustments; and they have poor adaptability to complex flow scenarios (such as non-uniform heat sources), with localized overheating problems still existing.

[0004] Therefore, existing technologies still need to be improved and developed. Summary of the Invention

[0005] The main objective of this invention is to provide a design method, system, terminal, and computer-readable storage medium for turbulence structures based on artificial intelligence and multi-objective optimization, aiming to solve the problems of difficulty in balancing heat transfer enhancement and flow resistance reduction, low optimization efficiency, and poor adaptability of existing turbulence structures.

[0006] To achieve the above objectives, the present invention provides a method for designing a turbulence structure based on artificial intelligence and multi-objective optimization. This method includes the following steps: The teardrop-shaped baffle column is selected as the starting point for optimization. A first preset number of control points are defined to describe the outline of the baffle column. A second preset number of design variables are formed based on the first preset number of control points, and upper and lower limits are set for each design variable. Three artificial neural network proxy models are constructed. The second preset number of design variables are input into the three artificial neural network proxy models respectively for prediction. The pressure drop, average temperature and temperature uniformity coefficient of the double elliptical turbulence structure are output. A multi-objective optimization model was established, and the objective function was determined to minimize the pressure drop, average temperature, and temperature uniformity coefficient. After forming the dataset, the Pareto optimal solution set was generated based on the NSGA-II algorithm, and the asymmetric double elliptic structure with the best comprehensive performance was selected from the Pareto optimal solution set. A comprehensive performance evaluation criterion is introduced, and the asymmetric double elliptical structure is evaluated based on the ratio of enhanced heat transfer to increased flow resistance. The asymmetric double elliptical structure that meets the performance requirements is then symmetricalized to obtain a symmetrical double elliptical turbulence structure.

[0007] Optionally, in the aforementioned turbulence structure design method based on artificial intelligence and multi-objective optimization, the first preset number of control points is 8 control points, and the second preset number of design variables is 16 design variables.

[0008] Optionally, the aforementioned method for designing a turbulence structure based on artificial intelligence and multi-objective optimization, wherein selecting a teardrop-shaped turbulence column as the optimization starting point, defining a first preset number of control points to describe the outline of the turbulence column, forming a second preset number of design variables based on the first preset number of control points, and setting upper and lower limits for each design variable, specifically includes: A teardrop-shaped turbulence column whose comprehensive performance meets the preset requirements was selected as the starting point for optimization. The outline of the spoiler column is described by defining the x and y coordinates of 8 control points, and a total of 16 independent design variables are formed based on the 8 control points. Based on geometric feasibility and flow rationality, upper and lower limits are set for each design variable.

[0009] Optionally, in the aforementioned artificial intelligence and multi-objective optimization-based turbulence structure design method, the artificial neural network proxy model includes an input layer, a hidden layer, and an output layer, and the training of the artificial neural network proxy model employs the Adam optimizer and an early stopping mechanism.

[0010] Optionally, the aforementioned method for designing a turbulence structure based on artificial intelligence and multi-objective optimization, wherein the construction of three artificial neural network surrogate models, inputting a second preset number of design variables into the three artificial neural network surrogate models for prediction, and outputting the pressure drop, average temperature, and temperature uniformity coefficient of the double-elliptical turbulence structure, specifically includes: The second preset number of design variables are normalized and preprocessed before being passed into each three-layer fully connected artificial neural network proxy model; The design variables are received through the input layer, the two hidden layers complete the nonlinear feature mapping through the activation function, and the output layer outputs the predicted values ​​of three performance indicators through the linear activation function, respectively obtaining the pressure drop, average temperature and temperature uniformity coefficient of the double elliptical turbulence structure.

[0011] Optionally, the aforementioned method for designing turbulence structures based on artificial intelligence and multi-objective optimization includes establishing a multi-objective optimization model, determining the objective function with the goal of minimizing pressure drop, average temperature, and temperature uniformity coefficient, generating a Pareto optimal solution set based on the NSGA-II algorithm after forming a dataset, and selecting the asymmetric double-elliptic structure with the best overall performance from the Pareto optimal solution set. Specifically, this includes: A multi-objective optimization model was established, with the objectives being to minimize pressure drop, average temperature, and temperature uniformity coefficient, and the objective function was determined. After the dataset is generated, a global search is performed in the 16-dimensional design space based on the NSGA-II algorithm. After iterative optimization, a Pareto optimal solution set is generated through non-dominated sorting, crowding calculation and elite retention strategy. The asymmetric double elliptic structure with the best overall performance is then selected from the Pareto optimal solution set.

[0012] Optionally, the aforementioned turbulence structure design method based on artificial intelligence and multi-objective optimization, wherein the introduction of a comprehensive performance evaluation criterion, based on the ratio of enhanced heat transfer to increased flow resistance, evaluates the asymmetric double-elliptical structure, and symmetrizes the asymmetric double-elliptical structure that meets the performance requirements to obtain a symmetrical double-elliptical turbulence structure, specifically includes: Representative points are selected from the frontier of the Pareto optimal solution set for detailed analysis. A comprehensive performance evaluation criterion is introduced, and the asymmetric double elliptical structure is finally evaluated based on the ratio of enhanced heat transfer to increased flow resistance. The selected optimized structure was verified by CFD simulation. The asymmetric double elliptical structure that met the performance requirements was symmetricized to obtain a symmetrical double elliptical turbulence structure.

[0013] Furthermore, to achieve the above objectives, the present invention also provides a disturbance structure design system based on artificial intelligence and multi-objective optimization, wherein the disturbance structure design system based on artificial intelligence and multi-objective optimization includes: The parametric modeling module is used to select the teardrop-shaped baffle column as the optimization starting point, define a first preset number of control points to describe the outline of the baffle column, form a second preset number of design variables based on the first preset number of control points, and set upper and lower limits for each design variable. The surrogate model prediction module is used to construct three artificial neural network surrogate models. After inputting the second preset number of design variables into the three artificial neural network surrogate models for prediction, the module outputs the pressure drop, average temperature, and temperature uniformity coefficient of the double elliptical turbulence structure. The multi-objective optimization module is used to establish a multi-objective optimization model, determine the objective function with the goal of minimizing pressure drop, average temperature and temperature uniformity coefficient, generate Pareto optimal solution set based on NSGA-II algorithm after forming the dataset, and select the asymmetric double elliptic structure with the best comprehensive performance from the Pareto optimal solution set. The structural verification and improvement module is used to introduce comprehensive performance evaluation criteria. Based on the ratio of enhanced heat transfer to increased flow resistance, the asymmetric double elliptical structure is evaluated. The asymmetric double elliptical structure that meets the performance requirements is symmetrically processed to obtain a symmetrical double elliptical turbulence structure.

[0014] Furthermore, to achieve the above objectives, the present invention also provides a terminal, wherein the terminal includes: a memory, a processor, and a turbulence structure design program based on artificial intelligence and multi-objective optimization stored in the memory and executable on the processor, wherein when the turbulence structure design program based on artificial intelligence and multi-objective optimization is executed by the processor, it implements the steps of the turbulence structure design method based on artificial intelligence and multi-objective optimization as described above.

[0015] Furthermore, to achieve the above objectives, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a turbulence structure design program based on artificial intelligence and multi-objective optimization, and when the turbulence structure design program based on artificial intelligence and multi-objective optimization is executed by a processor, it implements the steps of the turbulence structure design method based on artificial intelligence and multi-objective optimization as described above.

[0016] In this invention, a teardrop-shaped turbulence column is selected as the optimization starting point. A first preset number of control points are defined to describe the outline of the turbulence column. A second preset number of design variables are formed based on the first preset number of control points, and upper and lower limits are set for each design variable. Three artificial neural network surrogate models are constructed. The second preset number of design variables are input into the three artificial neural network surrogate models respectively for prediction, and the pressure drop, average temperature, and temperature uniformity coefficient of the double elliptical turbulence structure are output. A multi-objective optimization model is established, and the objective function is determined to minimize the pressure drop, average temperature, and temperature uniformity coefficient. After forming a dataset, the Pareto optimal solution set is generated based on the NSGA-II algorithm, and the asymmetric double elliptical structure with the best comprehensive performance is selected from the Pareto optimal solution set. A comprehensive performance evaluation criterion is introduced, and the asymmetric double elliptical structure is evaluated based on the ratio of heat transfer enhancement to flow resistance increase. The asymmetric double elliptical structure that meets the performance requirements is symmetricized to obtain a symmetrical double elliptical turbulence structure. This invention uses a teardrop-shaped turbulence column as the initial structure, defines its geometric profile through parametric modeling, and uses an artificial neural network to construct a surrogate model to quickly predict the heat transfer performance (such as average temperature and temperature uniformity) and flow characteristics (such as pressure drop) under different structural parameters. Based on this, the NSGA-II multi-objective optimization algorithm is combined to efficiently search the Pareto optimal solution set in the global design space with the objectives of minimizing pressure drop, minimizing average temperature, and optimizing temperature uniformity, thereby achieving high-performance design of the turbulence structure. Attached Figure Description

[0017] Figure 1 This is a flowchart of a preferred embodiment of the turbulence structure design method based on artificial intelligence and multi-objective optimization of the present invention; Figure 2 This is a diagram showing the initial shape and the movement range of movable points in a preferred embodiment of the turbulence structure design method based on artificial intelligence and multi-objective optimization of the present invention. Figure 3 This is a schematic diagram of the artificial neural network proxy model in a preferred embodiment of the turbulence structure design method based on artificial intelligence and multi-objective optimization of the present invention; Figure 4 This is a Pareto front solution set diagram in a preferred embodiment of the perturbation structure design method based on artificial intelligence and multi-objective optimization of the present invention; Figure 5 This is a comparison diagram of structure D, structure D*, and structure E in a preferred embodiment of the turbulence structure design method based on artificial intelligence and multi-objective optimization of the present invention. Figure 6 This is a structural diagram of a preferred embodiment of the turbulence structure design system based on artificial intelligence and multi-objective optimization of the present invention; Figure 7This is a structural diagram of a preferred embodiment of the terminal of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0019] The preferred embodiment of the present invention describes a perturbation structure design method based on artificial intelligence and multi-objective optimization, such as... Figure 1 As shown, the perturbation structure design method based on artificial intelligence and multi-objective optimization includes the following steps: Step S10: Select the teardrop-shaped baffle column as the optimization starting point, define the first preset number control points to describe the outline of the baffle column, form the second preset number of design variables based on the first preset number control points, and set upper and lower limits for each design variable.

[0020] Specifically, the first preset number of control points is 8 control points, and the second preset number of design variables is 16 design variables. Baseline structure: A teardrop-shaped baffle column whose comprehensive performance meets the preset requirements is selected as the optimization starting point; Parametric method: such as... Figure 2 As shown, the outline of the spoiler column is described by defining the x and y coordinates of 8 control points (A, B, C, D, E, F, G, H). Based on the 8 control points, a total of 16 independent design variables (a1 to a) are formed. 16 Design space constraints: To ensure geometric feasibility and flow rationality, upper and lower limits are set for each design variable based on geometric feasibility and flow rationality, as shown in Table 1. The constraint principle is to avoid overlapping design variables that could lead to modeling failure, and to prevent the structure from being too close to the flow channel wall, which could cause flow blockage.

[0021] Table 1: Design Space and Initial Shape

[0022] Step S20: Construct three artificial neural network proxy models. Input the second preset number of design variables into the three artificial neural network proxy models respectively for prediction, and output the pressure drop, average temperature and temperature uniformity coefficient of the double elliptical turbulence structure.

[0023] Specifically, the artificial neural network proxy model includes an input layer, a hidden layer, and an output layer. The training of the artificial neural network proxy model employs the Adam optimizer and an early stopping mechanism. A second preset number of design variables are preprocessed by normalization and then input into each three-layer fully connected artificial neural network proxy model. The input layer receives the design variables, the two hidden layers complete nonlinear feature mapping through activation functions, and the output layer outputs predicted values ​​of three performance indicators through a linear activation function, yielding the pressure drop, average temperature, and temperature uniformity coefficient of the double-elliptical perturbation structure.

[0024] like Figure 3 As shown, three independent ANN (Artificial Neural Network) surrogate models are constructed to predict pressure drop (ΔP), average temperature (Tave), and temperature uniformity coefficient (Tσ), respectively. Each ANN surrogate model uses the same input layer (16 neurons, corresponding to the design variables), the hidden layer structure is determined through tuning, and the output layer is a single target value. The pressure drop (ΔP), average temperature (Tave), and temperature uniformity coefficient (Tσ) of the double-elliptical perturbation structure were predicted using 16 independent geometric design variables corresponding to 8 key control points as inputs. After normalization preprocessing, these variables were fed into a three-layer fully connected ANN surrogate model. Through a forward propagation process of "the input layer receiving variables, two hidden layers completing nonlinear feature mapping through activation functions, and then the output layer outputting predicted values ​​through linear activation functions," the prediction results of the three performance indicators were obtained. The model training adopted the Adam optimizer and early stopping mechanism. Finally, the determination coefficients (R²) of ΔP, Tave, and Tσ predictions reached 0.968, 0.911, and 0.944, respectively, with relative errors of less than 3%. This model can efficiently replace the time-consuming CFD (Computational Fluid Dynamics) simulation and provide rapid performance evaluation support for subsequent NSGA-II (Non-dominated Sorting Genetic Algorithm II) multi-objective optimization.

[0025] Data preparation: Latin Hypercube Sampling (LHS) was used to generate 4000 sets of sample points within the design space. For each set of geometric parameters, CFD (Computational Fluid Dynamics) simulations were performed to obtain the corresponding ΔP, Tave, and Tσ values, which constituted the training dataset.

[0026] Training and Validation: Three independent three-layer fully connected neural networks were constructed. The input layer receives standardized design variables, the hidden layer performs nonlinear feature mapping using the tanh activation function, and the output layer outputs the predicted values ​​of the three performance indicators using linear activation functions. Validation showed that the coefficient of determination (R²) was greater than 0.91 and the relative error was less than 3%, ultimately achieving second-level accurate prediction from geometric parameters to performance indicators, providing efficient support for subsequent NSGA-II multi-objective optimization. The dataset was divided into training and validation sets in an 8:2 ratio. After training, the surrogate model exhibited extremely high prediction accuracy (R² > 0.91) and extremely low root mean square error (RMSE), completely replacing time-consuming CFD (Computational Fluid Dynamics) simulations for rapid performance evaluation.

[0027] The ANN surrogate model reduces the time for a single performance evaluation from hours of CFD simulation to seconds, making global optimization of high-dimensional complex design spaces possible.

[0028] Step S30: Establish a multi-objective optimization model, determine the objective function with the goal of minimizing pressure drop, average temperature and temperature uniformity coefficient, generate the Pareto optimal solution set based on the NSGA-II algorithm after forming the dataset, and select the asymmetric double elliptic structure with the best comprehensive performance from the Pareto optimal solution set.

[0029] Specifically, a multi-objective optimization model is established, with the objectives being to minimize pressure drop, average temperature, and temperature uniformity coefficient, and the objective function is determined. After forming the dataset, a global search is performed in the 16-dimensional design space based on the NSGA-II algorithm. After iterative optimization, a Pareto optimal solution set is generated through non-dominated sorting, crowding calculation, and elite retention strategies. The asymmetric double elliptic structure with the best overall performance is then selected from the Pareto optimal solution set.

[0030] Optimization objective: Establish a multi-objective optimization model with the objective functions being to minimize ΔP, minimize Tave, and minimize Tσ.

[0031] Optimization process: Multi-objective optimization aims to minimize pressure drop (ΔP), average temperature (Tave), and temperature uniformity coefficient (Tσ). It is carried out based on the CFD-ANN-NSGA-II joint framework. After the dataset is formed, the NSGA-II algorithm (population size 200, 1000 iterations, crossover probability 0.9, mutation probability 0.1) is used to perform a global search in the 16-dimensional design space. Pareto optimal solution set is generated through non-dominated sorting, crowding calculation, and elite retention strategy. Finally, the asymmetric double elliptic knot with the best overall performance is selected from the Pareto optimal solution set. After symmetry improvement, a more manufacturable structure is obtained.

[0032] Optimization Algorithm: The NSGA-II algorithm is adopted. The algorithm is based on the fast prediction of the ANN surrogate model, and performs efficient search in the 16-dimensional design space. It maintains the diversity and convergence of solutions through non-dominated sorting and crowding calculation.

[0033] Output: Through iterative optimization, a series of Pareto optimal solutions (i.e., Pareto fronts, such as...) are obtained. Figure 4 As shown in the figure, these solutions represent the optimal structure under different trade-offs among the three objectives (pressure drop, average temperature, and temperature uniformity).

[0034] Step S40: Introduce a comprehensive performance evaluation criterion. Based on the ratio of enhanced heat transfer to increased flow resistance, evaluate the asymmetric double elliptical structure. The asymmetric double elliptical structure that meets the performance requirements is symmetrically processed to obtain a symmetrical double elliptical turbulence structure.

[0035] Specifically, representative points are selected from the frontier of the Pareto optimal solution set for detailed analysis. A comprehensive performance evaluation criterion is introduced, and the asymmetric double elliptical structure is finally evaluated based on the ratio of enhanced heat transfer to increased flow resistance. The selected optimized structure is verified by CFD simulation. The asymmetric double elliptical structure that meets the performance requirements is symmetricized to obtain a symmetrical double elliptical turbulence structure.

[0036] Solution selection: Select representative points from the Pareto frontier (e.g.) Figure 4 A detailed analysis is conducted on points A, B, C, and D in the diagram. A final evaluation is then performed using the Performance Evaluation Criterion (PEC). PEC comprehensively considers the ratio of increased heat transfer (Nu) to increased flow resistance (f).

[0037] Performance Verification: CFD simulation verification was performed on the selected optimized structure (e.g., point D). Based on the geometric parameters of the optimized structure (D), a 3D model was constructed in a local channel of a liquid-cooled plate with dimensions of 60 mm × 4 mm × 2 mm. The liquid-cooled plate was made of aluminum, and the coolant was water. An unstructured polyhedral mesh was used to divide the computational domain, and 350,000 meshes were determined to be optimal after mesh independence verification. Mass flow rate, inlet temperature, and pressure were set at the outlet boundary. A constant heat flux density was applied to the bottom surface (simulating 5C battery discharge). The sides and top were insulated. A laminar flow model and the finite volume method were used to discretize the governing equations. The pressure-velocity coupling was achieved using the SIMPLE algorithm, and the convergence residual was less than 1.0 × 10⁻⁶. -6After solving the problem, the pressure drop (ΔP), average temperature (Tave), temperature uniformity coefficient (Tσ), and Nusselt number (Nu) were extracted and compared with the ANN predictions for verification. Simultaneously, the flow field (velocity vector, boundary layer) and pressure distribution at point D were compared to quantify the heat transfer and drag reduction advantages of the double-elliptical structure, ultimately confirming the reliability and performance improvement of the optimized structure. The results show that the prediction error of the ANN-NSGA-II framework is less than 4%, demonstrating the reliability and accuracy of the optimized framework.

[0038] Engineering Improvements (Key Step): The initial optimized structure D is an asymmetric shape. To facilitate manufacturing and maintain performance, this invention further proposes its symmetric variant D* (e.g., Figure 5 The dashed section shown is structure D, and the actual structure is D*. Symmetry treatment only resulted in a negligible decrease in PEC (from 1.0393 to 1.0372), but further reduced pressure drop and significantly improved manufacturability. Compared to the baseline teardrop structure (E), the final double-elliptical symmetrical structure (D*) achieved a 16.75% reduction in pressure drop, a significant decrease in system pump power requirement, and a 3.72% increase in PEC, while maintaining essentially the same heat transfer capacity. The double-elliptical structure reduces incoming flow impact through its leading-edge ellipse, guides flow in the middle neck region, and weakens the wake separation zone through its trailing-edge ellipse, achieving excellent drag reduction through a synergistic effect, while the multi-vortex structure ensures sufficient flow disturbance.

[0039] In this invention, the specific geometric configuration of the "double ellipse" is an optimized, naturally emerging contour composed of two tangent or adjacent elliptical segments. This structure is not pre-defined but rather a "natural optimal solution" discovered by the optimization algorithm under the objective of balancing heat transfer and drag. This is the core innovation at the product level. Its technical effects are reflected in: the front ellipse achieves streamlined guidance and drag reduction, the rear ellipse regulates wake stabilization, and beneficial vortices are generated in the neck region. The engineering implementation path of "asymmetric optimization-symmetric improvement" involves first using the optimization algorithm to search for the asymmetric optimal solution without limitations, and then performing symmetry processing based on this to obtain a final product that is easy to manufacture. This path ensures both superior performance and engineering practicality.

[0040] This invention uses a teardrop-shaped turbulence column as the initial structure, defines its geometric profile through parametric modeling, and employs an artificial neural network (ANN) to construct a surrogate model to quickly predict heat transfer performance (such as average temperature and temperature uniformity) and flow characteristics (such as pressure drop) under different structural parameters. Based on this, the NSGA-II multi-objective optimization algorithm is combined to efficiently search the Pareto optimal solution set within the global design space, aiming to minimize pressure drop, minimize average temperature, and optimize temperature uniformity, thus achieving high-performance design of the turbulence structure. The results show that the optimized structure can effectively reduce pressure drop while maintaining good heat transfer performance, achieving an overall performance improvement: compared with the initial scheme, the pressure drop is reduced by 20.37 Pa (16.75%), and the overall performance factor PEC is improved by 3.93%. This demonstrates the effectiveness of the optimization strategy under multi-objective equilibrium, improves optimization efficiency, and provides a theoretical basis for the structural design of liquid cooling systems.

[0041] Furthermore, such as Figure 6 As shown, based on the above-mentioned disturbance structure design method based on artificial intelligence and multi-objective optimization, the present invention also provides a disturbance structure design system based on artificial intelligence and multi-objective optimization, wherein the disturbance structure design system based on artificial intelligence and multi-objective optimization includes: The parametric modeling module 51 is used to select the teardrop-shaped turbulence column as the optimization starting point, define a first preset number of control points to describe the outline of the turbulence column, form a second preset number of design variables based on the first preset number of control points, and set upper and lower limits for each design variable. The surrogate model prediction module 52 is used to construct three artificial neural network surrogate models. After inputting the second preset number of design variables into the three artificial neural network surrogate models for prediction, it outputs the pressure drop, average temperature and temperature uniformity coefficient of the double elliptical turbulence structure. The multi-objective optimization module 53 is used to establish a multi-objective optimization model, determine the objective function with the goal of minimizing pressure drop, average temperature and temperature uniformity coefficient, generate Pareto optimal solution set based on NSGA-II algorithm after forming the dataset, and select the asymmetric double elliptic structure with the best comprehensive performance from the Pareto optimal solution set. The structural verification and improvement module 54 is used to introduce comprehensive performance evaluation criteria, evaluate the asymmetric double elliptical structure based on the ratio of enhanced heat transfer to increased flow resistance, and symmetricize the asymmetric double elliptical structure that meets the performance requirements to obtain a symmetrical double elliptical turbulence structure.

[0042] Furthermore, such as Figure 7As shown, based on the above-mentioned disturbance structure design method and system based on artificial intelligence and multi-objective optimization, the present invention also provides a terminal, which includes a processor 10, a memory 20 and a display 30. Figure 7 Only some of the terminal components are shown; however, it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0043] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard drive or memory. In other embodiments, the memory 20 may be an external storage device of the terminal, such as a plug-in hard drive, smart media card (SMC), secure digital card (SD), flash card, etc. Further, the memory 20 may include both internal and external storage devices. The memory 20 is used to store application software and various types of data installed on the terminal, such as the program code installed on the terminal. The memory 20 can also be used to temporarily store data that has been output or will be output. In one embodiment, the memory 20 stores a turbulence structure design program 40 based on artificial intelligence and multi-objective optimization. This turbulence structure design program 40 can be executed by the processor 10 to implement the turbulence structure design method based on artificial intelligence and multi-objective optimization in this application.

[0044] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor, or other data processing chip, used to run program code stored in the memory 20 or process data, such as executing the disturbance structure design method based on artificial intelligence and multi-objective optimization.

[0045] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display 30 is used to display information on the terminal and to display a visual user interface. The terminal's processor 10, memory 20, and display 30 communicate with each other via a system bus.

[0046] In one embodiment, when the processor 10 executes the artificial intelligence and multi-objective optimization-based perturbation structure design program 40 in the memory 20, the following steps are performed: The teardrop-shaped baffle column is selected as the starting point for optimization. A first preset number of control points are defined to describe the outline of the baffle column. A second preset number of design variables are formed based on the first preset number of control points, and upper and lower limits are set for each design variable. Three artificial neural network proxy models are constructed. The second preset number of design variables are input into the three artificial neural network proxy models respectively for prediction. The pressure drop, average temperature and temperature uniformity coefficient of the double elliptical turbulence structure are output. A multi-objective optimization model was established, and the objective function was determined to minimize the pressure drop, average temperature, and temperature uniformity coefficient. After forming the dataset, the Pareto optimal solution set was generated based on the NSGA-II algorithm, and the asymmetric double elliptic structure with the best comprehensive performance was selected from the Pareto optimal solution set. A comprehensive performance evaluation criterion is introduced, and the asymmetric double elliptical structure is evaluated based on the ratio of enhanced heat transfer to increased flow resistance. The asymmetric double elliptical structure that meets the performance requirements is then symmetricalized to obtain a symmetrical double elliptical turbulence structure.

[0047] The first preset number of control points is 8 control points, and the second preset number of design variables is 16 design variables.

[0048] Specifically, the selection of teardrop-shaped baffle columns as the optimization starting point, the definition of a first preset number of control points to describe the outline of the baffle columns, the formation of a second preset number of design variables based on the first preset number of control points, and the setting of upper and lower limits for each design variable, specifically include: A teardrop-shaped turbulence column whose comprehensive performance meets the preset requirements was selected as the starting point for optimization. The outline of the spoiler column is described by defining the x and y coordinates of 8 control points, and a total of 16 independent design variables are formed based on the 8 control points. Based on geometric feasibility and flow rationality, upper and lower limits are set for each design variable.

[0049] The artificial neural network proxy model includes an input layer, a hidden layer, and an output layer. The training of the artificial neural network proxy model uses the Adam optimizer and an early stopping mechanism.

[0050] Specifically, the construction of three artificial neural network surrogate models involves inputting a second preset number of design variables into each of the three models for prediction, and then outputting the pressure drop, average temperature, and temperature uniformity coefficient of the double-elliptical turbulence structure. The second preset number of design variables are normalized and preprocessed before being passed into each three-layer fully connected artificial neural network proxy model; The design variables are received through the input layer, the two hidden layers complete the nonlinear feature mapping through the activation function, and the output layer outputs the predicted values ​​of three performance indicators through the linear activation function, respectively obtaining the pressure drop, average temperature and temperature uniformity coefficient of the double elliptical turbulence structure.

[0051] Specifically, the establishment of a multi-objective optimization model involves determining the objective function to minimize pressure drop, average temperature, and temperature uniformity coefficient. After forming the dataset, a Pareto optimal solution set is generated based on the NSGA-II algorithm. The asymmetric double-elliptic structure with the best overall performance is then selected from this Pareto optimal solution set. A multi-objective optimization model was established, with the objectives being to minimize pressure drop, average temperature, and temperature uniformity coefficient, and the objective function was determined. After the dataset is generated, a global search is performed in the 16-dimensional design space based on the NSGA-II algorithm. After iterative optimization, a Pareto optimal solution set is generated through non-dominated sorting, crowding calculation and elite retention strategy. The asymmetric double elliptic structure with the best overall performance is then selected from the Pareto optimal solution set.

[0052] The introduction of a comprehensive performance evaluation criterion, based on the ratio of enhanced heat transfer to increased flow resistance, evaluates the asymmetric double-elliptical structure. The asymmetric double-elliptical structure that meets the performance requirements is then symmetrically processed to obtain a symmetrical double-elliptical turbulence structure, specifically including: Representative points are selected from the frontier of the Pareto optimal solution set for detailed analysis. A comprehensive performance evaluation criterion is introduced, and the asymmetric double elliptical structure is finally evaluated based on the ratio of enhanced heat transfer to increased flow resistance. The selected optimized structure was verified by CFD simulation. The asymmetric double elliptical structure that met the performance requirements was symmetricized to obtain a symmetrical double elliptical turbulence structure.

[0053] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a turbulence structure design program based on artificial intelligence and multi-objective optimization, and when the turbulence structure design program based on artificial intelligence and multi-objective optimization is executed by a processor, it implements the steps of the turbulence structure design method based on artificial intelligence and multi-objective optimization as described above.

[0054] In summary, this invention provides a method, system, terminal, and computer-readable storage medium for designing a turbulence structure based on artificial intelligence and multi-objective optimization. The method includes: selecting a teardrop-shaped turbulence column as the optimization starting point; defining a first preset number of control points to describe the outline of the turbulence column; forming a second preset number of design variables based on the first preset number of control points; and setting upper and lower limits for each design variable; constructing three artificial neural network surrogate models; inputting the second preset number of design variables into the three artificial neural network surrogate models for prediction; and outputting the pressure drop, average temperature, and temperature uniformity coefficient of the double-elliptical turbulence structure; establishing a multi-objective optimization model; determining the objective function with the goal of minimizing the pressure drop, average temperature, and temperature uniformity coefficient; generating a Pareto optimal solution set based on the NSGA-II algorithm after forming a dataset; and selecting the asymmetric double-elliptical structure with the best overall performance from the Pareto optimal solution set; introducing a comprehensive performance evaluation criterion; evaluating the asymmetric double-elliptical structure based on the ratio of enhanced heat transfer to increased flow resistance; and symmetricizing the asymmetric double-elliptical structure that meets the performance requirements to obtain a symmetrical double-elliptical turbulence structure. This invention uses a teardrop-shaped turbulence column as the initial structure, defines its geometric profile through parametric modeling, and uses an artificial neural network to construct a surrogate model to quickly predict the heat transfer performance (such as average temperature and temperature uniformity) and flow characteristics (such as pressure drop) under different structural parameters. Based on this, the NSGA-II multi-objective optimization algorithm is combined to efficiently search the Pareto optimal solution set in the global design space with the objectives of minimizing pressure drop, minimizing average temperature, and optimizing temperature uniformity, thereby achieving high-performance design of the turbulence structure.

[0055] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal that includes that element.

[0056] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.). The program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The computer-readable storage medium can be a memory, magnetic disk, optical disk, etc.

[0057] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A method for designing a spoiler structure based on artificial intelligence and multi-objective optimization, characterized in that, The disturbance structure design method based on artificial intelligence and multi-objective optimization includes: The teardrop-shaped baffle column is selected as the starting point for optimization. A first preset number of control points are defined to describe the outline of the baffle column. A second preset number of design variables are formed based on the first preset number of control points, and upper and lower limits are set for each design variable. Three artificial neural network proxy models are constructed. The second preset number of design variables are input into the three artificial neural network proxy models respectively for prediction. The pressure drop, average temperature and temperature uniformity coefficient of the double elliptical turbulence structure are output. A multi-objective optimization model was established, and the objective function was determined to minimize the pressure drop, average temperature, and temperature uniformity coefficient. After forming the dataset, the Pareto optimal solution set was generated based on the NSGA-II algorithm, and the asymmetric double elliptic structure with the best comprehensive performance was selected from the Pareto optimal solution set. A comprehensive performance evaluation criterion is introduced, and the asymmetric double elliptical structure is evaluated based on the ratio of enhanced heat transfer to increased flow resistance. The asymmetric double elliptical structure that meets the performance requirements is then symmetrically processed to obtain a symmetrical double elliptical turbulence structure.

2. The method of claim 1, wherein, The first preset number of control points is 8 control points, and the second preset number of design variables is 16 design variables.

3. The disturbance structure design method based on artificial intelligence and multi-objective optimization according to claim 2, characterized in that, The selection of teardrop-shaped baffle columns as the optimization starting point involves defining a first preset number of control points to describe the outline of the baffle columns. Based on these first preset number of control points, a second preset number of design variables are formed, and upper and lower limits are set for each design variable. Specifically, this includes: A teardrop-shaped turbulence column whose comprehensive performance meets the preset requirements was selected as the starting point for optimization. The outline of the spoiler column is described by defining the x and y coordinates of 8 control points, and a total of 16 independent design variables are formed based on the 8 control points. Based on geometric feasibility and flow rationality, upper and lower limits are set for each design variable.

4. The disturbance structure design method based on artificial intelligence and multi-objective optimization according to claim 1, characterized in that, The artificial neural network proxy model includes an input layer, a hidden layer, and an output layer. The training of the artificial neural network proxy model uses the Adam optimizer and an early stopping mechanism.

5. The disturbance structure design method based on artificial intelligence and multi-objective optimization according to claim 4, characterized in that, The construction of three artificial neural network surrogate models involves inputting a second preset number of design variables into each of the three models for prediction, and then outputting the pressure drop, average temperature, and temperature uniformity coefficient of the double-elliptical turbulence structure. Specifically, this includes: The second preset number of design variables are normalized and preprocessed before being passed into each three-layer fully connected artificial neural network proxy model; The design variables are received through the input layer, the two hidden layers complete the nonlinear feature mapping through the activation function, and the output layer outputs the predicted values ​​of three performance indicators through the linear activation function, respectively obtaining the pressure drop, average temperature and temperature uniformity coefficient of the double elliptical turbulence structure.

6. The disturbance structure design method based on artificial intelligence and multi-objective optimization according to claim 1, characterized in that, The process involves establishing a multi-objective optimization model, determining the objective function to minimize pressure drop, average temperature, and temperature uniformity coefficient, generating a Pareto optimal solution set based on the NSGA-II algorithm after forming the dataset, and selecting the asymmetric double-elliptic structure with the best overall performance from the Pareto optimal solution set. Specifically, this includes: A multi-objective optimization model was established, with the objectives being to minimize pressure drop, average temperature, and temperature uniformity coefficient, and the objective function was determined. After the dataset is generated, a global search is performed in the 16-dimensional design space based on the NSGA-II algorithm. After iterative optimization, a Pareto optimal solution set is generated through non-dominated sorting, crowding calculation and elite retention strategy. The asymmetric double elliptic structure with the best overall performance is then selected from the Pareto optimal solution set.

7. The disturbance structure design method based on artificial intelligence and multi-objective optimization according to claim 1, characterized in that, The introduced comprehensive performance evaluation criteria evaluate the asymmetric double-elliptical structure based on the ratio of enhanced heat transfer to increased flow resistance. The asymmetric double-elliptical structure that meets the performance requirements is then symmetrically processed to obtain a symmetrical double-elliptical turbulence structure, specifically including: Representative points are selected from the frontier of the Pareto optimal solution set for detailed analysis. A comprehensive performance evaluation criterion is introduced, and the asymmetric double elliptical structure is finally evaluated based on the ratio of enhanced heat transfer to increased flow resistance. The selected optimized structure was verified by CFD simulation. The asymmetric double elliptical structure that met the performance requirements was symmetricized to obtain a symmetrical double elliptical turbulence structure.

8. A disturbance structure design system based on artificial intelligence and multi-objective optimization, characterized in that, The disturbance structure design system based on artificial intelligence and multi-objective optimization includes: The parametric modeling module is used to select the teardrop-shaped baffle column as the optimization starting point, define a first preset number of control points to describe the outline of the baffle column, form a second preset number of design variables based on the first preset number of control points, and set upper and lower limits for each design variable. The surrogate model prediction module is used to construct three artificial neural network surrogate models. After inputting the second preset number of design variables into the three artificial neural network surrogate models for prediction, the module outputs the pressure drop, average temperature, and temperature uniformity coefficient of the double elliptical turbulence structure. The multi-objective optimization module is used to establish a multi-objective optimization model, determine the objective function with the goal of minimizing pressure drop, average temperature and temperature uniformity coefficient, generate Pareto optimal solution set based on NSGA-II algorithm after forming the dataset, and select the asymmetric double elliptic structure with the best comprehensive performance from the Pareto optimal solution set. The structural verification and improvement module is used to introduce comprehensive performance evaluation criteria. Based on the ratio of enhanced heat transfer to increased flow resistance, the asymmetric double elliptical structure is evaluated. The asymmetric double elliptical structure that meets the performance requirements is symmetrically processed to obtain a symmetrical double elliptical turbulence structure.

9. A terminal, characterized in that, The terminal includes: a memory, a processor, and a turbulence structure design program based on artificial intelligence and multi-objective optimization stored in the memory and executable on the processor. When the turbulence structure design program based on artificial intelligence and multi-objective optimization is executed by the processor, it implements the steps of the turbulence structure design method based on artificial intelligence and multi-objective optimization as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a turbulence structure design program based on artificial intelligence and multi-objective optimization, which, when executed by a processor, implements the steps of the turbulence structure design method based on artificial intelligence and multi-objective optimization as described in any one of claims 1-7.

Images