Design optimization method for industrial-grade printed circuit board type heat exchanger based on structured grid division and artificial neural network

Abstract

The invention discloses an industrial-grade printed circuit board heat exchanger design optimization method based on structured grid division and an artificial neural network, and the method comprises the steps: S1, carrying out the modeling of a heat exchanger based on a structured grid division method, and segmenting an end socket of the heat exchanger into a plurality of hexahedron structures, a flow channel of the heat exchanger is built into an agent flow channel model in the axial direction; s2, key parameters of a heat exchanger runner are extracted through a proxy runner model, and flowing and pressure drop characteristics of fluid in the runner are calculated; S3, based on the computational fluid mechanics simulation technology, a velocity field, a pressure field and flow distribution of the fluid in a heat exchanger end socket are obtained; s4, dividing control bodies for the heat exchanger based on a finite volume method, establishing a heat exchange amount and efficiency calculation program of the heat exchanger, and obtaining the efficiency of the heat exchanger; and S5, an artificial neural network training model is introduced, a multi-objective optimization algorithm is applied, the thermal hydraulic performance of the heat exchanger is optimized, a Pareto leading edge is generated, and multiple design selections are provided for designers to refer to.

Description

Technical Field

[0001] The present invention relates to the technical field of heat exchanger design, and particularly to an industrial printed circuit board heat exchanger design optimization method based on structured grid division and artificial neural network. Background Art

[0002] As a key component in the energy conversion system, heat exchangers play an important role in industrial production, energy utilization, nuclear energy systems and other fields. Industrial printed circuit board heat exchangers (PCHE for short) have been widely used in intermediate heat exchangers of supercritical carbon dioxide Brayton cycles in nuclear energy systems due to their compact volume, high energy density, and ability to withstand high temperatures and pressures. However, in the actual use process of existing PCHEs, the following technical bottlenecks are faced:

[0003] Flow non-uniformity leads to a decrease in heat transfer efficiency: Inside the heat exchanger, due to the limitations of the flow channel distribution and geometric shape, the phenomenon of uneven flow distribution often occurs, resulting in excessive flow in some channels and insufficient flow in other channels. This flow non-uniformity will reduce the heat transfer efficiency and even lead to a decline in the overall performance of the heat exchanger. The research and prediction of the variation law of flow non-uniformity in industrial printed circuit board heat exchangers are still in their infancy.

[0004] High-precision modeling and simulation consume a large amount of computing resources: The design of heat exchangers usually relies on computational fluid dynamics (CFD) simulation to analyze the internal flow and heat transfer characteristics. However, for industrial PCHEs, due to the large number of flow channels, complex three-dimensional geometry and flow field calculations will consume a large amount of computing resources, seriously affecting the design efficiency.

[0005] Insufficient optimization of thermohydraulic characteristics: The current heat exchanger design is mostly based on empirical formulas or single objectives for optimization, and fails to fully consider the comprehensive balance between heat transfer efficiency and pressure drop. In addition, most existing studies are aimed at simplified models and cannot accurately reflect the complex structural characteristics of industrial PCHEs, resulting in insufficient applicability of the optimization scheme in actual applications.

[0006] The application of artificial intelligence in heat exchanger design is not yet mature: Although artificial intelligence technologies such as artificial neural network (ANN) have been gradually applied to heat exchanger design in recent years, most studies are only limited to data prediction or single-objective optimization, lacking a comprehensive exploration of multi-objective optimization.

[0007] In summary, there are still obvious deficiencies in the existing technologies in solving problems such as flow non-uniformity, heat transfer efficiency optimization, and simulation calculation efficiency of PCHEs. Summary of the Invention

[0008] The purpose of the present invention is to provide an industrial-grade printed circuit board heat exchanger design optimization method based on structured grid division and artificial neural network, covering fluid flow distribution, heat exchange efficiency optimization, and the realization of multi-objective design, aiming at the defects of the existing technology.

[0009] To solve the above technical problems, the present invention provides the following technical solutions:

[0010] An industrial-grade printed circuit board heat exchanger design optimization method based on structured grid division and artificial neural network, characterized by comprising the following steps:

[0011] S1: Model the heat exchanger based on the structured grid division method, including dividing the head of the heat exchanger into several hexahedron structures, and establishing a proxy flow channel model along the axial direction of the flow channels of the heat exchanger. The cross-section of the hexahedron structure perpendicular to the fluid flow direction corresponds one-to-one with the inlets and outlets of the proxy flow channels;

[0012] S2: Use the proxy flow channel model to extract the key parameters of the flow channels of the heat exchanger, calculate the flow and pressure drop characteristics of the fluid in the flow channels. The input quantities of each proxy flow channel are the length, hydraulic diameter, cross-sectional area, fluid velocity, and physical properties of the flow channel, and the output quantities are the friction coefficient, mass flow rate, and pressure drop of each flow channel:

[0013] S3: Based on the computational fluid dynamics simulation technology, obtain the velocity field, pressure field, and flow distribution of the fluid in the head of the heat exchanger;

[0014] S4: Divide the control volume of the heat exchanger based on the finite volume method, establish a heat transfer amount and efficiency calculation program for the heat exchanger, and obtain the efficiency of the heat exchanger;

[0015] S5: Introduce an artificial neural network training model, apply a multi-objective optimization algorithm, optimize the thermohydraulic performance of the heat exchanger, generate a Pareto front, and provide multiple design options for designers to refer to.

[0016] Further, the pressure drop, friction coefficient, and mass flow rate formulas of the proxy flow channel in S2 are:

[0017]

[0018] q m,i = ρ i A c u i

[0019] where Δp i is the pressure drop of the flow channel, f is the friction coefficient, L is the length of the flow channel, ρ is the fluid density, u is the fluid velocity, d h is the hydraulic diameter of the flow channel, Re is the fluid Reynolds number, q m,i is the mass flow rate of the flow channel, Ac is the cross-sectional area of the flow channel.

[0020] Furthermore, in S3, the Realizable k-ε turbulence model is adopted to calculate the velocity field, pressure field and flow rate distribution of the fluid.

[0021] Furthermore, in S4, the heat exchanger efficiency calculation formula is

[0022]

[0023] where ρ is the fluid density, c p is the specific heat capacity of the fluid, l is the flow channel length, V is the control volume, Q is the heat transfer amount by conduction between adjacent control volumes, ε is the heat exchanger efficiency, is the temperature difference between the inlet and outlet of the fluid on the side with a large temperature change, ΔT max is the maximum temperature difference in the heat exchanger.

[0024] Furthermore, in S5, the artificial neural network training model adopts a backpropagation neural network, and its input variables are the flow channel length, head size, and number of flow channels, and the output variables are the heat transfer efficiency and the head inlet pressure.

[0025] Furthermore, in S5, the multi-objective optimization algorithm adopts the Non-dominated Sorting Genetic Algorithm II, and the objective functions are to maximize the heat transfer efficiency and minimize the pressure loss.

[0026] Furthermore, the number of flow channels is 20 to 50, the flow channel length is 0.2 m to 1.2 m, and the head size is 200 mm to 300 mm.

[0027] Furthermore, the initial design parameters of the heat exchanger head are: the fluid inlet flow velocity is 60 m / s, the nozzle diameter is 120 mm, the ratio of the cross-sectional area of the flow channel part to the cross-sectional area of the head is 2:9, the flow channel length is 0.8 m, the number of flow channel rows is 50, and the side length of the head is 300 mm.

[0028] Furthermore, the hyperparameters of the artificial neural network training model include the number of hidden layers and the number of neurons. The number of hidden layers is 2, and the number of neurons is 5 for each layer.

[0029] Furthermore, the optimization method is applied to the design of the intermediate heat exchanger in the supercritical carbon dioxide cooled Brayton cycle system.

[0030] Compared with the prior art, the beneficial effects of the present invention are:

[0031] 1. Improve the flow characteristics and enhance the heat transfer efficiency: By optimizing the head design and flow channel configuration, the head pressure drop is effectively optimized, and the overall heat transfer performance is improved;

[0032] 2. Reduce computational cost: By adopting a surrogate flow channel model and structured grid division technology, the grid division and CFD simulation calculation time are significantly reduced;

[0033] 3. Achieve the balance between heat transfer efficiency and pressure drop during the design process: Through a multi-objective optimization method, while ensuring heat transfer efficiency, the pressure drop is reduced as much as possible to improve the system economy;

[0034] 4. Improve design efficiency: Combining with the ANN prediction model, the heat transfer performance under different design parameters can be quickly evaluated to provide guidance for the design of industrial-grade PCHEs.

[0035] 5. The present invention is applicable to the design and optimization of industrial-grade PCHEs and the design of heat exchangers for supercritical carbon dioxide Brayton cycle systems. In addition, it can also be widely applied to industrial fields such as chemical engineering and waste heat recovery that require high-efficiency heat exchangers.

[0036] Other features and advantages of the present invention will be described in the following specification, or understood by implementing the present invention. Brief Description of the Drawings

[0037] Figure 1 It is a schematic structural diagram of an industrial-grade PCHE.

[0038] Figure 2 It is a simplified flowchart of the design optimization process of an industrial-grade PCHE.

[0039] Figure 3 It is a flowchart of the structured grid division and surrogate flow channel model of the heat exchanger in the embodiment.

[0040] Figure 4 It is a schematic diagram of the control volume division for calculating the temperature field and heat transfer efficiency of an industrial-grade PCHE based on the finite volume method.

[0041] Figure 5 It is a flowchart for generating the differential equation group of the finite volume method.

[0042] Figure 6 It is the variation law of the total pressure drop of an industrial-grade PCHE obtained by using the control variable method.

[0043] Figure 7 It is a relationship diagram of the heat transfer efficiency and inlet pressure of an industrial-grade PCHE varying with design variables.

[0044] Figure 8 It is a Pareto front diagram in the multi-objective optimization process. Detailed Embodiment

[0045] To deepen the understanding of the present invention, the following will further elaborate on the present invention in conjunction with the accompanying drawings. This embodiment is only used to explain the present invention and does not limit the protection scope of the present invention.

[0046] Structured mesh generation:

[0047] As Figure 3 shown, the head region of the industrial PCHE is divided into structured meshes to improve the calculation efficiency and numerical accuracy. During the mesh generation process, the meshes in the boundary layer region are appropriately refined to capture the details of the flow boundary layer. For the flow channel region, the geometry is simplified through a surrogate flow channel model, so that only the flow rate and pressure data need to be input for each flow channel to obtain information such as pressure drop and flow rate, without the need for precise mesh generation of the entire flow channel.

[0048] Establishment of the surrogate flow channel model:

[0049] The heat exchanger flow channels are transformed into a surrogate flow channel model to simplify the input and output variables of each flow channel. In the surrogate flow channel model, the input quantities are the mass flow rate at the inlet of the flow channel and the pressure at the outlet of the flow channel, and the output quantities are the mass flow rate at the outlet of the flow channel and the pressure at the inlet of the flow channel:

[0050]

[0051] q m,i = ρ i A c u i

[0052] where Δp i is the pressure drop of each flow channel, f is the friction coefficient, L is the length of the flow channel, ρ is the fluid density, u is the fluid velocity, d h is the hydraulic diameter of the flow channel, Re is the fluid Reynolds number, q m,i is the mass flow rate of each flow channel, and A c is the cross-sectional area of the flow channel.

[0053] CFD-based hydrodynamic simulation

[0054] A three-dimensional simulation of the heat exchanger head is performed using computational fluid dynamics (CFD) software (such as Ansys Fluent); the Realizable k-ε turbulence model is used to calculate the velocity field, pressure field, and flow distribution of the fluid; the total pressure drop of the head is analyzed through simulation to evaluate the flow distribution between the flow channels.

[0055] Based on the finite volume method, control volumes are divided for the industrial PCHE, and a temperature field distribution and efficiency calculation program are established to obtain the temperature field distribution and the heat exchanger efficiency:

[0056]

[0057] where ρ is the fluid density, c p$c_p$ is the specific heat capacity of the fluid, $l$ is the length of the flow channel, $V$ is the volume of the control volume, $Q$ is the heat transfer rate by conduction between a certain control volume and the adjacent control volume, $\varepsilon$ is the heat exchanger efficiency, which is the ratio of the temperature difference of the fluid on the side with a larger temperature change to the maximum temperature difference in the heat exchanger. $\Delta T_{in - out}$ is the temperature difference between the inlet and outlet of the fluid on the side with a larger temperature change. max $\Delta T_{max}$ is the maximum temperature difference in the heat exchanger.

[0058] Training of the artificial neural network (ANN) model for describing the pressure drop and efficiency of the heat exchanger: Generate data on flow rate, pressure, and heat transfer performance under different working conditions using CFD simulation; Train the ANN model using the backpropagation neural network (BPNN): The input variables are: the number of flow channels, the head size, and the flow channel length; The output variables are: the heat transfer efficiency and the inlet pressure of the head; Adjust the hyperparameters of the ANN model (such as the number of hidden layers and the number of neurons) to ensure a small training error for the model.

[0059] Multi - objective optimization of the flow and heat transfer characteristics:

[0060] Use the non - dominated sorting genetic algorithm II (NSGA - II) for optimization. The optimization parameters include the number of flow channels, the head size, and the flow channel length, with the heat transfer efficiency and the pressure drop as the objective functions; According to the Pareto - front design results, select the best combination between the heat transfer efficiency and the pressure drop: For high - performance scenarios, select the design with high heat transfer efficiency; For low - energy - consumption scenarios, select the design with low pressure drop; And generate specific design parameters.

[0061] The present invention will be further described in detail below with reference to the accompanying drawings.

[0062] Structured grid generation:

[0063] First, perform pre - partitioning. As shown in Figure 3 (a), according to the need to generate a structured grid, the head is divided into several hexahedrons. Hexahedral grids have the advantages of high computational efficiency, high numerical accuracy, and suitability for complex geometries in CFD simulations, and can significantly improve the simulation efficiency and accuracy. Therefore, in the present invention, choosing to divide the head region into hexahedral grids is to reduce the consumption of computing resources while ensuring the calculation accuracy and improving the design optimization efficiency.

[0064] After generating the grid, the cross - section of the head is composed of quadrilaterals, as shown in Figure 3 (b). For each rectangular cross - section of the flow channel, several quadrilaterals that make up the above - mentioned head cross - section are assigned to the parts corresponding to the inlet and outlet of the flow channel, as shown in Figure 3As shown in (c). The inlets and outlets of the rectangular cross-section flow channels correspond to the quadrilaterals of the head cross-section through mesh generation, and the inlet and outlet cross-sections of the flow channels are mapped onto the quadrilateral meshes of the head. The surrogate flow channel is a simplified model of the actual flow channel, and the flow characteristics inside the flow channel are simulated by inputting the inlet and outlet parameters of the flow channel. The pitch is the spacing between actual flow channels, which is the same as the spacing between the quadrilateral meshes of the head cross-section.

[0065] On the other hand, in order to simplify the mesh in the axial length direction of the flow channel, each flow channel is transformed into a surrogate flow channel. The input quantities of each surrogate flow channel are the mass flow rate at the inlet of the upstream flow channel, the average surface pressure at the outlet of the downstream flow channel, and the physical properties of the fluid in the flow channel. The output quantities are the mass flow rate at the outlet of the downstream flow channel, the pressure drop of each flow channel, and the average surface pressure at the inlet of the upstream flow channel, without the need to generate a mesh, as Figure 3 shown in (d). This function is implemented through a user-defined function UDF. The formulas for pressure drop, friction coefficient, and mass flow rate are:

[0066]

[0067]

[0068] q m,i =ρ i A c u i

[0069] where Δp i is the pressure drop of each flow channel, f is the friction coefficient, L is the length of the flow channel, ρ is the fluid density, u is the fluid velocity, d h is the hydraulic diameter of the flow channel, Re is the Reynolds number of the fluid, q m,i is the mass flow rate of each flow channel, A c is the cross-sectional area of the flow channel. Known quantities: flow channel length L, fluid density ρ, fluid velocity u, hydraulic diameter d h of the flow channel, and cross-sectional area A c of the flow channel; unknown quantities: mass flow rate q m,i of each flow channel, pressure drop Δp i of each flow channel, friction coefficient f, and Reynolds number Re of the fluid.

[0070] Function: By calculating the pressure drop, friction coefficient, and Reynolds number, the CFD - surrogate flow channel model can work properly, evaluate the flow performance of the heat exchanger, optimize the flow channel design, reduce the pressure drop, and improve the heat transfer efficiency. These parameters play a key role in subsequent overall pressure drop calculations and multi-objective optimizations.

[0071] The main dimensional parameters of the head are as Figure 3 indicated by the text annotation. The reference dimensions of the head in this embodiment are shown in Table 1 below:

[0072] Table 1: Reference Dimensions of the Head

[0073]

[0074] The reference dimensions are the given initial design parameters and serve as the starting point for simulation and optimization. The reference dimensions are selected through a large number of preliminary experiments based on engineering experience, design requirements, and preliminary geometric constraints. The final design results will have a certain fluctuation based on the reference dimensions.

[0075] Research on Flow Characteristics Based on CFD (Computational Fluid Dynamics) Simulation:

[0076] The research on the flow characteristics of the head driven by CFD simulation is an important part of the design and optimization of heat exchangers. When we carry out the design and optimization of heat exchangers, we need to obtain the pressure drop and velocity distribution of the heat exchanger, which can be obtained through CFD simulation.

[0077] Perform three-dimensional simulation on the heat exchanger head using Ansys Fluent, load the head with structured grids, and calculate using the Realizable k-ε turbulence model. The variation law of the total pressure drop of the head is as Figure 6 shown.

[0078] According to Figure 4 , 5 's method, divide the control volume of the industrial-grade PCHE based on the finite volume method, establish a calculation program for the heat transfer amount and efficiency of the heat exchanger, and obtain the heat exchanger efficiency:

[0079]

[0080] where ρ is the fluid density, c p is the specific heat capacity of the fluid, l is the flow channel length, V is the control volume, Q is the heat transfer amount by conduction between a certain control volume and the adjacent control volume, ε is the heat exchanger efficiency, which is the ratio of the temperature difference of the fluid on the side with a larger temperature change to the maximum temperature difference in the heat exchanger, is the inlet and outlet temperature difference of the fluid on the side with a larger temperature change, and ΔT max is the maximum temperature difference in the heat exchanger.

[0081] The selected parameter range for the heat exchanger design is shown in Table 2 below and is changed based on the reference dimensions of this embodiment.

[0082] Table 2: Selected Parameter Range for Heat Exchanger Design

[0083]

[0084] This parameter range is also further selected based on engineering experience, design requirements, and preliminary geometric constraints through a large number of preliminary experiments on the basis of the reference dimensions of this embodiment. This parameter range can ensure that the geometric dimensions of the heat exchanger within the range are in line with engineering practice.

[0085] The total pressure drop and the velocity fields of each flow channel are obtained using the proxy flow channels, and the velocity fields of each flow channel are input into the temperature field distribution and efficiency calculation program to obtain the heat exchanger efficiency. The results when the side length of the head is 300 mm are as Figure 7 shown.

[0086] By comparing these two figures, it can be concluded that when the side length of the head is 300 mm, the variation of the flow and heat transfer characteristics of this industrial-grade PCHE with the design variables is generally monotonic, and the heat transfer efficiency increases with the increase in the number of flow channel rows and the flow channel length. When the flow channel length and the number of flow channel rows are 1.2 m and 50 respectively, the heat transfer efficiency reaches the maximum value of 0.882. When the flow channel length and the number of flow channel rows are 0.2 m and 20 respectively, the heat transfer efficiency reaches the minimum value of 0.572. The variation trend of the PCHE inlet pressure is the same as that of the heat transfer efficiency, but the growth rate varies greatly. The maximum value of the inlet pressure is 22649.99 Pa, and the minimum value is 9288.94 Pa. The maximum value within the design variable range is 2.44 times that of the minimum value. When the heat transfer efficiency changes to 1.54 times the original value, the pressure loss will increase rapidly. Therefore, the pressure drop needs to be optimized in the thermohydraulic design of industrial-grade PCHE.

[0087] Train the ANN model using the backpropagation neural network (BPNN): The input variables are: the number of flow channels, the head size, and the flow channel length; the output variables are: the heat transfer efficiency and the head inlet pressure. Adjust the hyperparameters of the ANN model (such as the number of hidden layers and the number of neurons) to ensure that the training error of the model is small. In the present invention, the method for selecting hyperparameters is parameter sensitivity analysis. For the dataset with a small amount of data in the present invention, an ANN with 2 hidden layers can meet the requirements. In the prediction of the head inlet pressure and the heat exchanger efficiency, gradually increase the number of neurons in the hidden layer. When it exceeds 5, the improvement of the training effect is very small, and at the same time, the probability of training divergence increases. Based on this, the present invention selects 2 hidden layers, with 5 neurons in each layer, and also an input layer and an output layer

[0088] The hyperparameters are shown in Table 3 below:

[0089] Table 3: Hyperparameters of the ANN model

[0090]

[0091] The non-dominated sorting genetic algorithm II (NSGA-II) is used for optimization. Three head sizes are selected, with the number of flow channels and the flow channel length as the optimization parameters, and the heat transfer efficiency and pressure drop as the objective functions, generating three Pareto fronts. The Pareto fronts are as shown in Figure 6 . For each head size, there is no other combination of the number of flow channels and the flow channel length that can simultaneously have a higher heat transfer efficiency and a lower pressure drop than the design points on the Pareto front. Then, three representative heat transfer efficiency values are selected from the set of Pareto fronts, covering the high, medium, and low ranges. High heat transfer efficiency (η = 0.85): prioritize performance and accept a higher pressure drop; medium heat transfer efficiency (η = 0.75): balance efficiency and energy consumption; low heat transfer efficiency (η = 0.65): prioritize reducing the pressure drop and sacrifice some efficiency. By selecting three heat transfer efficiency values and extracting the individual values of the three Pareto fronts respectively, nine design results are obtained, and each individual value is still a non-dominated solution on the Pareto front. As shown in Table 4 below. It enables designers to comprehensively consider the requirements of the heat exchanger for volume, heat transfer efficiency, and pressure drop, select the most suitable heat exchanger parameters, and provides a reference for the optimization of the design of industrial-grade PCHEs.

[0092] Table 4: Design Results

[0093]

[0094] The above specific implementation manners are only for illustrating the technical concept and structural features of the present invention, aiming to enable those skilled in the art to implement it accordingly. However, the above content does not limit the protection scope of the present invention. Any equivalent changes or modifications made according to the spirit of the present invention shall fall within the protection scope of the present invention.

Claims

1. A design optimization method for industrial printed circuit board heat exchanger based on structured grid division and artificial neural network, characterized in that: The steps include: S1: Modeling the heat exchanger based on the structured meshing method, including dividing the head of the heat exchanger into a number of hexahedral structures, establishing the flow channel of the heat exchanger as a proxy flow channel model along the axial direction, and the cross section of the hexahedral structure perpendicular to the fluid flow direction corresponds one to one with the inlet and outlet of the proxy flow channel; S2: Use the proxy flow channel model to extract the key parameters of the heat exchanger flow channel and calculate the flow and pressure drop characteristics of the fluid in the flow channel. The input of each proxy flow channel is the length, hydraulic diameter, cross-sectional area, flow rate and physical properties of the fluid. The output is the friction coefficient, mass flow rate and pressure drop of each flow channel: S3: Based on computational fluid dynamics simulation technology, the velocity field, pressure field and flow distribution of the fluid in the heat exchanger head are obtained; S4: Divide the control volume of the heat exchanger based on the finite volume method, establish the heat exchanger heat transfer and efficiency calculation program, and obtain the heat exchanger efficiency; S5: An artificial neural network training model is introduced and a multi-objective optimization algorithm is applied to optimize the thermal-hydraulic performance of the heat exchanger, generate the Pareto frontier, and provide a variety of design options for designers’ reference.

2. The design optimization method of industrial printed circuit board heat exchanger based on structured grid division and artificial neural network according to claim 1, characterized in that: The pressure drop, friction coefficient, and mass flow rate formulas for the proxy flow channel described in S2 are: q m,i =ρ i A c you i Where Δp i is the flow channel pressure drop, f is the friction coefficient, L is the flow channel length, ρ is the fluid density, u is the fluid flow rate, d h is the hydraulic diameter of the flow channel, Re is the fluid Reynolds number, q m,i is the mass flow rate of the flow channel, A c is the cross-sectional area of ​​the flow channel.

3. The design optimization method of industrial printed circuit board type heat exchanger based on structured grid division and artificial neural network according to claim 1, characterized in that: In S3, the Realizable k-ε turbulence model is used to calculate the velocity field, pressure field and flow distribution of the fluid.

4. The design optimization method of industrial printed circuit board type heat exchanger based on structured grid division and artificial neural network according to claim 1, characterized in that: In S4, the heat exchanger efficiency calculation formula is: Where ρ is the fluid density, c p is the specific heat capacity of the fluid, l is the length of the flow channel, V is the volume of the control body, Q is the heat transfer amount between adjacent control bodies, ε is the efficiency of the heat exchanger, is the inlet and outlet temperature difference of the fluid on the side with greater temperature change, ΔT max is the maximum temperature difference in the heat exchanger.

5. The design optimization method of industrial printed circuit board type heat exchanger based on structured grid division and artificial neural network according to claim 1, characterized in that: In S5, the artificial neural network training model adopts a back propagation neural network, whose input variables are the flow channel length, the head size, and the number of flow channels, and the output variables are the heat exchange efficiency and the head inlet pressure.

6. The design optimization method of industrial printed circuit board type heat exchanger based on structured grid division and artificial neural network according to claim 1, characterized in that: In S5, the multi-objective optimization algorithm adopts a non-dominated sorting genetic algorithm II, and the objective function is to maximize the heat exchange efficiency and minimize the pressure loss.

7. The design optimization method of industrial printed circuit board type heat exchanger based on structured grid division and artificial neural network according to claim 1, characterized in that: The number of flow channels is 20 to 50, the length of the flow channels is 0.2m to 1.2m, and the size of the head is 200mm to 300mm.

8. The design optimization method of industrial printed circuit board type heat exchanger based on structured grid division and artificial neural network according to claim 1, characterized in that: The initial design parameters of the heat exchanger head are: fluid inlet flow rate 60m / s, pipe diameter 120mm, ratio of the cross-sectional area of ​​the flow channel to the cross-sectional area of ​​the head 2:9, flow channel length 0.8m, number of flow channel rows 50, and head side length 300mm.

9. The design optimization method of industrial printed circuit board type heat exchanger based on structured grid division and artificial neural network according to claim 1, characterized in that: The hyperparameters of the artificial neural network training model include the number of hidden layers and the number of neurons. The number of hidden layers is 2, and the number of neurons is 5 per layer.

10. The method for designing and optimizing an industrial printed circuit board heat exchanger based on structured grid division and artificial neural network according to claim 1, characterized in that: The optimization method is applied to the design of intermediate heat exchangers in a supercritical carbon dioxide cooled Brayton cycle system.

Images