Heat exchange channel cross section design method and system based on artificial intelligence algorithm optimization

Abstract

The application discloses a heat exchange channel section design method and system based on an artificial intelligence algorithm optimization, and belongs to the technical field of heat pipes.The application generates regular section shapes and random section shapes meeting the needs of users according to application scene requirements, expands design materials, and improves the accuracy of a deep learning prediction model.The application can assist in realizing the section structure optimization design of a uniform channel heat exchanger under the condition of multi-physical field coupling in different application scenes.Based on image data driving, through constructing a cloud or private database and training a deep learning model, the complex nonlinear relationship between input pictures, data features and output features can be captured, the performance of the channel heat exchanger can be quickly predicted, and meanwhile, based on the mapping relationship, the multi-scale optimization design of the channel under the demand guidance can be realized, the time and economic consumption required in the channel design process are reduced, and the application is more efficient.

Description

Technical Field

[0001] This invention relates to the field of heat pipe technology, and in particular to the field of heat exchange channel optimization design accelerated by artificial intelligence technology. Specifically, it relates to a heat exchange channel cross-section design method and system based on artificial intelligence algorithm optimization. Background Technology

[0002] As the size of equipment in various fields such as electronic devices, automobiles, and aerospace continues to shrink and power density continues to increase, the requirements for heat dissipation performance are constantly rising, making thermal management technology increasingly critical. For example, lithium-ion batteries, due to their advantages such as high specific energy, high power, long cycle life, and low self-discharge rate, are currently one of the most popular choices for new energy vehicles.

[0003] Extensive efforts have been made to find better electrode materials for lithium-ion batteries to improve specific energy, specific power, and lifespan. Despite this, thermal issues caused by high or low temperatures still significantly impact the performance, lifespan, and safety of lithium-ion batteries. Furthermore, due to consumers' increasing pursuit of comfort and convenience in automobiles, pure electric vehicles still suffer from range anxiety. This necessitates not only high-energy-density lithium-ion batteries for new energy vehicles but also more compact battery packs to accommodate more batteries and extend driving range. This trend increases internal heat generation and heat accumulation within the battery, placing higher demands on the compactness and thermal management performance of the battery thermal management system. Similarly, driven by the combined requirements of performance, strength, weight, and cost, rocket engines are currently evolving towards stronger thrust output and higher energy density release within a smaller space, resulting in a significant increase in engine thermal load. To ensure the stability and safety of the engine under extreme high temperatures, thermal management technology has become a crucial aspect of rocket engine design. Efficient heat dissipation must be provided within a more compact space. This trend is driving the development of rocket engine thermal management technology towards high efficiency, compactness, and high performance to ensure the safe and stable operation of the engine under high-temperature and high-load environments.

[0004] Thermal management methods, depending on the working fluid, are mainly divided into air-based, liquid-based, and phase change material-based methods and their combinations. Among these, liquid channel thermal management is an important choice for solving thermal management problems in high-power, high-energy-density, and high-temperature scenarios due to its superior heat transfer efficiency, compactness, and stability. It has shown significant application potential in fields such as microelectronics, aerospace, nuclear energy construction, new energy industry, chemical engineering, and mechanical manufacturing. This means that liquid channel thermal management technology, especially its application in micro-sized channels, can alleviate the contradiction between high compactness and high thermal management performance in thermal management systems to a certain extent.

[0005] These demands have spurred the ongoing pursuit of efficient cooling channel design for liquid channels, efficient heat dissipation materials, and efficient and precise thermal simulation.

[0006] In terms of efficient cooling channel design, liquid channels suffer from problems such as large pressure drop, significant temperature rise of the refrigerant along the flow direction, and large temperature difference at the wall surface. To further improve liquid channel performance, existing research focuses on structural factors such as channel geometry, cross-sectional shape, channel layout, and interface properties. It indicates that the cross-sectional shape and size of the liquid channel are significant factors affecting the hydraulic and thermal performance of channel heat exchangers. For example, for elliptical and rectangular liquid channels, there is a threshold for the hydraulic diameter. When the hydraulic diameter is less than the threshold, the rectangular liquid channel exhibits greater hydraulic resistance and a higher convective heat transfer coefficient; conversely, when the hydraulic diameter is greater than the threshold, the opposite is true. However, there is currently a lack of research comparing the hydraulic and thermal performance of liquid channels of different shapes within a wide range under certain size constraints, to identify which channel cross-sectional shape offers optimal hydraulic and thermal performance for different application scenarios.

[0007] Regarding high-efficiency heat dissipation materials: depending on the operating temperature and working fluid material, the flow within the liquid channel can be divided into single-phase flow and two-phase flow. Single-phase flow provides stable cooling, but requires enhanced heat transfer performance. Boiling flow in two-phase flow achieves high heat transfer efficiency through the boiling of the working fluid on the channel wall, but further solutions are needed to address inherent issues such as instability in gas-liquid two-phase flow. Latent heat functional fluids in two-phase flow, composed of a base fluid and well-dispersed phase change microcapsule materials (MEPCM), can also achieve high heat transfer efficiency by absorbing / releasing latent heat through phase change during flow using uniformly dispersed MEPCMs, but further improvements are needed to extend the service life of latent heat functional fluids.

[0008] In terms of efficient and precise thermal simulation: channel structure design and material optimization are complex problems involving multiple variables and multiple couplings. Computational fluid dynamics numerical simulations or experimental investigations are economically and temporally inefficient for systematic performance analysis of the vast number of channel structures. Furthermore, in practical applications, the diverse choices of position and spacing within liquid cooling plates due to varying application scenarios mean that the channel cross-section and its arrangement within the liquid cooling plate are synergistic. Therefore, in practical applications, a wide variety of features can be extracted manually, further increasing the economic and time costs of dataset preparation. The development of big data technologies such as artificial intelligence (AI) has provided an opportunity to solve these problems. Based on a database constructed from limited liquid channel heat exchanger performance parameters, rapid performance evaluation of liquid channel heat exchangers can be achieved through model testing and training, significantly shortening the design cycle. Simultaneously, for complex optimization problems such as demand-driven channel structure material optimization, metaheuristic stochastic optimization algorithms in the field of artificial intelligence can achieve good global search optimization based on predictive models.

[0009] In summary, liquid channel thermal management is a thermal management solution with high application potential. Currently, further in-depth research is needed on the design of cooling channels and heat dissipation materials to improve the performance ceiling. However, the research methods commonly used in this field are too costly in terms of economy and time, and new technologies such as artificial intelligence need to be introduced to improve development efficiency. Summary of the Invention

[0010] To address the shortcomings of existing technologies, this invention proposes a heat exchange channel cross-section design method and system based on artificial intelligence algorithm optimization. The aim is to accelerate the design of channel cross-sections and reduce the time and economic cost of channel heat exchanger design by proposing a random cross-section shape generation algorithm and combining deep learning and metaheuristic random optimization algorithms in artificial intelligence technology.

[0011] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0012] A heat exchange channel cross-section design method based on artificial intelligence algorithm optimization includes the following steps:

[0013] S1: Initialize the design conditions for the cross-section of the channel heat exchanger;

[0014] S2: Based on the design space and load characteristic data of the channel heat exchanger cross-sectional structure, the optimal design structure and parameters of the channel heat exchanger cross-sectional structure are obtained;

[0015] S3: Based on the optimal design structure and parameters of the channel heat exchanger cross-section structure, derive the structural model and complete the design of the channel heat exchanger cross-section structure;

[0016] Preferably, in S1, the method for initializing the cross-sectional design conditions of the heat exchanger includes: determining the structural design conditions of the heat exchange channel cross-section, including heat exchange channel structural information, heat exchange channel material information, working fluid information, operating condition information, and performance evaluation parameters;

[0017] S101: Determine the established design variables, variable design variables, and performance evaluation parameters to be optimized in the structural design conditions of the heat exchange channel cross section.

[0018] S102: The structural information of the heat exchange channel includes the dimensions of the channel heat exchanger, the symmetry of the channel heat exchanger, the periodicity of the channel heat exchanger, the dimensional relationship between the channels, and the ratio of the volume inside the channel to the total volume of the channel heat exchanger.

[0019] S103: Heat exchange channel material information includes channel material type, corresponding material density, corresponding material thermodynamic parameters, channel inner wall friction coefficient, corresponding material elastic mechanical parameters, corresponding material plasticity and strength parameters, corresponding material fatigue and fracture parameters, and application scenario contact parameters;

[0020] S104: Information on the working fluid in the heat exchange channel includes working fluid density, working fluid viscosity, working fluid thermal conductivity, and working fluid specific heat capacity;

[0021] S105: The heat exchange channel operating condition information includes the thermal boundary conditions of the channel heat exchanger, the working fluid temperature at the channel inlet, the working fluid flow rate at the channel inlet, the pressure conditions at the channel outlet, and the force and pressure loads.

[0022] S106: Performance evaluation parameters for heat exchange channels include surface heat transfer coefficient, pressure drop, thermal resistance, dimensionless temperature, friction coefficient, Nusselt number, flow resistance characteristic coefficient, comprehensive heat transfer characteristic coefficient, stress distribution, strain analysis, displacement analysis, thermal stress, and fatigue analysis.

[0023] Preferably, in S2, the method for obtaining the optimal design structure and parameters of the channel heat exchanger cross-section structure based on the design space and load characteristic data of the channel heat exchanger cross-section structure includes:

[0024] S201: Generate the corresponding channel cross-section structure based on the design conditions of the heat exchange channel structure cross-section;

[0025] S202: Based on the design conditions of the heat exchange channel cross-section and the channel cross-section structure, the corresponding channel structure model is generated and exported by calling external 3D drawing software;

[0026] S203: Based on the channel structure model, establish data transmission with different domestic and foreign software used in flow, heat transfer, mechanical analysis and calculation; based on the heat transfer channel structure cross-sectional structure design conditions and channel structure model, obtain the performance evaluation parameter results of the corresponding channel structure.

[0027] S204: Based on the structural design conditions of the heat exchange channel cross-section, the channel cross-section structure and performance evaluation parameters, the channel cross-section dataset is obtained;

[0028] S205: Based on the channel cross-section dataset, a deep learning channel cross-section performance prediction and evaluation infrastructure template is pre-designed to obtain the channel cross-section performance prediction and evaluation space;

[0029] S206: Based on the channel cross-section performance prediction and evaluation space, initialize the channel cross-section structure, design parameters, and operating parameter weights;

[0030] S207: Based on the channel cross-section performance prediction and evaluation space, the optimal channel cross-section structure, design parameters and operating parameters are obtained;

[0031] Preferably, in S3, the method for deriving the structural model based on the optimal design structure and parameters of the channel heat exchanger cross-section structure, and completing the design of the channel heat exchanger cross-section structure includes:

[0032] S301: Optimized channel cross-section structure based on channel cross-section, exporting channel cross-section structure that can be called by external 3D drawing software;

[0033] S302: Optimal channel design parameters based on channel cross-section and channel cross-section structure that can be called by external 3D drawing software. By calling external 3D drawing software, the corresponding optimal channel structure model can be generated and exported.

[0034] S303: Based on the optimized channel cross-section, the optimized channel operating parameters and optimized channel structure model are used to establish data transmission with different domestic and foreign software used in flow, heat transfer, mechanical analysis and calculation, and obtain the performance evaluation parameter results of the corresponding channel structure.

[0035] Optionally, in S104, the working fluid in the heat exchange channel can be two or more immiscible materials;

[0036] Preferably, in S104, if the working fluid is two or more immiscible materials, for non-solid working fluids, the working fluid information should include the density, viscosity, thermal conductivity, specific heat capacity, and volume fraction of each working fluid.

[0037] Preferably, in S104, if the working fluid consists of two or more immiscible materials and includes solid particles, the working fluid information should also include particle size, particle density, particle thermal conductivity, particle specific heat capacity, and particle volume concentration.

[0038] Preferably, in S201, the method for generating the corresponding channel cross-sectional structure based on the design conditions of the heat exchange channel structure includes:

[0039] S201-1: Based on the structural design conditions of the hot channel cross section and the symmetry and periodicity of the channel heat exchanger, the minimum design unit of the channel is determined. Under the premise of applying the structural design conditions of the hot channel cross section, the minimum design unit can be reassembled into a channel heat exchanger through symmetrical and periodic arrangement.

[0040] S201-2: The cross section of the minimum design unit is the channel cross section, which includes the working fluid domain cross section and the wall domain cross section. The working fluid domain cross section should not exceed the range of the wall domain area.

[0041] S201-3: Determine the length-to-width ratio range and the thinnest wall thickness of the minimum design unit channel cross-section wall region;

[0042] S201-4: Select the planned regular working medium domain cross-sectional shape and random working medium domain cross-sectional shape. The regular working medium domain cross-sectional shapes and random working medium domain cross-sectional shapes that can be generated include: circle, ellipse, semicircle, circular segment, triangle, four- to seventeen-sided convex or concave shapes, and three- to nine-sided Reichelk polygons.

[0043] S201-5: Based on the requirements of S102, S201-2 and S201-3, the size and area parameters of the planned working fluid domain cross-sectional shape are predefined respectively, and the corresponding working fluid domain cross-sectional shape that meets the requirements and the working fluid domain cross-sectional geometry that can be called by external 3D drawing software are generated.

[0044] S201-6: Based on the requirements of S102, S201-1 and S201-4, create the wall domain cross-sectional shape that surrounds the corresponding working fluid domain cross-sectional shape, as well as the wall domain cross-sectional geometry that can be called by external 3D drawing software;

[0045] S201-7: Merge the working fluid domain section geometry and the wall domain section geometry into a composite shape and export it as an image file format, which can be used by external 3D drawing software as a channel section geometry, and export the channel section image;

[0046] S201-8: In the channel cross-sectional image, the channel wall region is black, and the working fluid region is white;

[0047] Preferably, in S204, based on the heat exchange channel structure cross-section design conditions, channel cross-section structure and performance evaluation parameter results, the channel cross-section dataset includes: channel cross-section structure image, corresponding channel cross-section structure variable design variables, and the results of the corresponding channel structure performance evaluation parameters to be optimized based on S203 and the above two points;

[0048] Preferably, in S205, the method for pre-designing a deep learning channel cross-section performance prediction and evaluation infrastructure template based on the channel cross-section dataset to obtain the channel cross-section performance prediction and evaluation space includes:

[0049] S205-1: Adjust the channel cross-sectional image to a square of a specified size while maintaining the aspect ratio information to ensure that the content of the image is not distorted due to scaling. The reference size for image resolution is 128~1600 pixels.

[0050] S205-2: Based on the channel cross-section dataset, the adjusted cross-section image and the variable design variables of the corresponding channel cross-section structure are used as inputs, and the results of the performance evaluation parameters of the corresponding channel structure to be optimized are used as outputs;

[0051] S205-3: Divide the cross-sectional image into several blocks, with the block size ranging from 16 to 64 pixels, and use a convolutional layer to convert these block images into embedding vectors with dimensions ranging from 128 to 512.

[0052] S205-4: Add position encoding to the image blocks to ensure that the transformer can understand the relative positions of the image blocks;

[0053] S205-5: Convert the variable design variables of the corresponding channel cross-section structure into an embedding vector with the same dimension as the image block;

[0054] S205-6: Input the above embedding vectors into the Transformer encoder to obtain the final feature representation; the Transformer encoder is composed of multiple stacked Transformer encoder layers, each encoder containing a self-attention mechanism and a feedforward neural network; the number of Transformer encoder layers is 4 to 12; the number of self-attention heads in each encoder layer is 4 to 12; the scale ratio of hidden layers in the feedforward neural network is 2.0 to 8.0;

[0055] S205-7: The output layer maps the final feature representation of each performance evaluation parameter to be optimized to the specific target prediction value;

[0056] S205-8: Calculate the loss for each performance evaluation parameter to be optimized, and then perform backpropagation to update the model parameters;

[0057] S205-9: Iterate from S205-6 to S205-8, calculate the loss on the validation set, and stop training early if the loss of the main performance evaluation parameter to be optimized in the validation set has not improved within a certain number of rounds. The learning rate decay range is 1e-5~1e-3.

[0058] S205-10: The output includes the model's weights and statistical information related to model training, representing the best performing model.

[0059] Preferably, in S206, the method for initializing the weights of design parameters and operating parameters based on the channel cross-section performance prediction and evaluation space includes:

[0060] S206-1: Set the weights and upper and lower limits of the performance evaluation parameters to be optimized;

[0061] S206-2: Constructing channel cross-section images based on shape models; optionally, shape models include hyperellipses, combinations of hyperellipses and polygons, Bézier curves, and Lagrange interpolation curves.

[0062] S206-3: Define the variable design variables and parameter range of the shape model for the cross-sectional structure;

[0063] S206-4: Construct an image of the channel cross-section based on the shape model and model parameter settings;

[0064] S206-5: Standardize the channel cross-section image and variable design variables of the cross-section structure constructed in S206-4, input them into the channel cross-section performance prediction and evaluation space, and predict the values ​​of the performance evaluation parameters to be optimized;

[0065] Preferably, in S207, the method for obtaining the optimal channel cross-section structure, design parameters, and operating parameters based on the channel cross-section performance prediction and evaluation space includes:

[0066] S207-1: The random initialization method generates a set of channel cross-sectional images and corresponding cross-sectional structure variable design variables that satisfy the settings of S206, as a set of design samples;

[0067] S207-2: Based on the design sample, use the evaluation components to evaluate the design sample and the corresponding performance evaluation parameters to be optimized, and obtain the evaluation component's score;

[0068] S207-3: Calculate the score based on the evaluation component and calculate the weight gradient of each parameter within the design sample;

[0069] S207-4: Based on the weight gradient and optimization rate, correct the weights of the corresponding channel cross-sectional structure, design parameters and operating parameters;

[0070] S207-5: Iteratively optimize S207-1 to S207-4 to obtain the optimal channel cross-section image, design parameters, and operating parameters;

[0071] Optionally, in S206, the method for initializing the weights of design parameters and operating parameters based on the channel cross-section performance prediction and evaluation space includes:

[0072] S206-1: Set the weights and upper and lower limits of the performance evaluation parameters to be optimized; set the range of variable design variables and shape model parameters for the cross-sectional structure;

[0073] S206-2: Based on the generator model, construct the channel cross-section image. The generator generates the corresponding channel cross-section image by inputting a random noise vector.

[0074] S206-3: Set the range of input noise and network structure parameters for the generator to control the diversity and complexity of the generated channel cross-sectional images and adjust the values ​​of variable design variables;

[0075] S206-4: By setting the parameters of the generator, a channel cross-sectional image is generated, and the quality of the generated sample is gradually adjusted to make it approximate the distribution of the real sample;

[0076] S206-5: Input the generated image into the discriminator. The discriminator evaluates the realism of the image and provides feedback to guide the optimization process of the generator. It also combines variable design variables to predict the performance evaluation parameters to be optimized, so that the generated image can meet the interval requirements.

[0077] Optionally, in S207, the method for obtaining the optimal channel cross-section structure, design parameters, and operating parameters based on the channel cross-section performance prediction and evaluation space includes:

[0078] S207-1: A set of noise vectors is randomly initialized and used as input samples. The generator then generates corresponding channel cross-sectional images based on these noise vectors. During the generation process, the generator continuously adjusts the channel cross-sectional images and variable design variables to conform to the set parameters.

[0079] S207-2: The generated images are evaluated by a discriminator, which assigns a realism score to each image;

[0080] S207-3: Based on the discriminator's score, calculate the generator's loss and perform backpropagation. By calculating the loss function, the generator adjusts its weights and design variables to improve the quality of the generated images and generate samples that meet the objective function requirements as much as possible;

[0081] S207-4: Update the generator weights through gradient descent to improve the quality of the generated image, making it more difficult for the discriminator to distinguish it as a fake image;

[0082] S207-5: Iteratively execute the above steps until the channel cross-section image generated by the generator meets the predetermined quality requirements, the discriminator cannot accurately distinguish between real and generated samples, and the generator continuously optimizes image generation based on performance evaluation parameters to achieve the optimal design within the set range, thereby obtaining the optimal channel cross-section image, design parameters, and operating parameters.

[0083] Preferably, in S301, the method for deriving a channel cross-section structure that can be called by external 3D drawing software based on the optimized channel cross-section structure includes:

[0084] S301-1: Based on the optimized channel cross-section image, design parameters, and operating parameters, the image pixels are mapped to the actual physical size;

[0085] S301-2: Based on S301-1, create the optimal wall domain section geometry that can be called by external 3D drawing software;

[0086] S301-3: Based on S301-1, create the optimal working fluid domain cross-sectional geometry that can be called by external 3D drawing software;

[0087] S301-4: Merge the optimal wall domain section geometry and the optimal working medium domain section geometry into a composite shape and export it as an image file format, which can be used as a channel section geometry that can be called by external 3D drawing software;

[0088] This invention also provides a heat exchange channel cross-section design system based on artificial intelligence algorithm optimization, including an initialization module, a design optimization module, and an output module:

[0089] The initialization module is used to initialize the cross-sectional design conditions of the channel heat exchanger;

[0090] The design optimization module is used to obtain the optimal design structure and parameters of the channel heat exchanger cross-section structure based on the design space and load characteristic data of the channel heat exchanger cross-section structure.

[0091] The output module is used to optimize the design structure and parameters based on the cross-sectional structure of the channel heat exchanger, export the structural model, and complete the design of the cross-sectional structure of the channel heat exchanger.

[0092] Preferably, in the initialization module, the process of initializing the design conditions of the channel heat exchanger cross section is the same as step S1;

[0093] Preferably, in the design optimization module, the process of obtaining the optimal design structure and parameters of the channel heat exchanger cross-section structure based on the design space and load characteristic data of the channel heat exchanger cross-section structure is the same as step S2;

[0094] Preferably, in the output module, the process of deriving the structural model based on the optimal design structure and parameters of the channel heat exchanger cross-section structure and completing the design of the channel heat exchanger cross-section structure is the same as step S3.

[0095] The technical solution of the present invention achieves the following beneficial technical effects:

[0096] 1. This invention can generate regular and random cross-sectional shapes that meet the needs of users according to the requirements of application scenarios, expand design materials, and improve the accuracy of deep learning prediction models;

[0097] 2. This invention can assist in the performance analysis of uniform channel heat exchangers under multi-physics coupling conditions in different application scenarios, and help to reveal the influencing factors and laws of channel heat exchanger performance;

[0098] 3. This invention is driven by image data. By constructing a cloud or private database and training a deep learning model, it can capture the complex nonlinear relationship between input images, data features and output features. It can quickly predict the performance parameters of channel heat exchangers based on images of channel cross-section structures and the variable design variables of the corresponding channel cross-section structures, reducing the time and economic consumption required for the parameterization of channel cross-section structures, and making it more efficient.

[0099] 4. This invention can output the best-performing model weights based on deep learning models, adjust the weights and upper and lower limits of the performance evaluation parameters to be optimized according to the user's needs, and combine various optimization algorithms for horizontal comparison to help users achieve demand-oriented multi-scale optimization design.

[0100] 5. This invention has good openness and data conversion capabilities. The system is an integrated software platform that can establish data transmission with most mainstream 3D design software and mainstream domestic and foreign software used in flow, heat transfer, mechanical analysis and calculation, and has strong comprehensive processing capabilities. Attached Figure Description

[0101] Figure 1 This is a structural diagram of the design optimization module of a heat exchange channel cross-section design system based on artificial intelligence algorithm optimization according to the present invention.

[0102] Figure 2 This is a flowchart illustrating the execution of a heat exchange channel cross-section design system based on artificial intelligence algorithm optimization according to the present invention.

[0103] Figure 3 This invention relates to a heat exchange channel cross-section design system based on artificial intelligence algorithms, which generates regular or random cross-sectional shapes.

[0104] Figure 4 This describes the execution process of a predictive model for a heat exchange channel cross-section design system optimized by an artificial intelligence algorithm, according to the present invention.

[0105] Figure 5 This is an iterative process for optimizing the channel cross-section image and geometry in a heat exchange channel cross-section design system based on artificial intelligence algorithms according to the present invention.

[0106] Figure 6 This is a fitting effect diagram of the convective heat transfer coefficient h of the prediction model of the heat exchange channel cross-section design system based on artificial intelligence algorithm optimization according to the present invention.

[0107] Figure 7 This is a fitting effect diagram of the pressure drop pd of the prediction model of the heat exchange channel cross-section design system based on artificial intelligence algorithm optimization according to the present invention. Detailed Implementation

[0108] The present invention will be specifically described below through embodiments. It should be noted that the following embodiments are only used to further illustrate the present invention, but are not limited thereto, unless otherwise stated.

[0109] The specific embodiments of the present invention are described in detail below with reference to the technical solutions:

[0110] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0111] Example 1:

[0112] Please see Figure 1-5 One embodiment of the present invention provides a heat exchange channel cross-section design system based on artificial intelligence algorithm optimization, comprising an initialization module, a design optimization module, and an output module.

[0113] The initialization module is used to initialize the cross-sectional design conditions of the channel heat exchanger;

[0114] The design optimization module is used to obtain the optimal design structure and parameters of the channel heat exchanger cross-section structure based on the design space and load characteristic data of the channel heat exchanger cross-section structure.

[0115] The output module is used to optimize the design of the structure and parameters based on the cross-sectional structure of the channel heat exchanger, export the structural model, and complete the design of the cross-sectional structure of the channel heat exchanger.

[0116] A heat exchange channel cross-section design method based on artificial intelligence algorithm optimization. Figure 2 The flowchart of the method is shown, which specifically includes the following steps:

[0117] S1. Determine the structural design conditions of the heat exchange channel cross section: including heat exchange channel structural information, heat exchange channel material information, working fluid information, operating condition information, and performance evaluation parameters;

[0118] S101: The known design variables in the structural design conditions for the heat exchange channel cross-section are the working fluid, the volume ratio of the liquid domain to the solid domain, the solid domain material, and the thermal boundary conditions; the variable design variables are the Reynolds number, cross-sectional area, cross-sectional shape, and channel length; the performance evaluation parameters to be optimized are the convective heat transfer coefficient h and the pressure drop pd.

[0119] S102: The heat exchange channel structure information indicates that the channel is a straight heat exchanger with a length ranging from 3 to 2000 mm, a width of 2000 mm, and a cross-sectional area of ​​a single channel ranging from 0.02 to 200 mm². 2 The channel heat exchanger has a symmetrical structure and the channels can be considered to satisfy a periodic relationship; the ratio of the channel fluid domain volume to the total volume of the channel heat exchanger is 0.5.

[0120] S103: The heat exchange channel material is pure aluminum;

[0121] S104: The working fluid in the heat exchange channel consists of two immiscible materials: water and a type of heat storage microcapsule. The heat storage microcapsule is a solid particle. It is suspended in water. Due to its small size, the working fluid in the heat exchange channel is considered homogeneous. The working fluid density, viscosity, thermal conductivity, specific heat capacity, and particle volume concentration are measured in the laboratory within the range of 5-85℃. The particle size, density, thermal conductivity, and specific heat capacity of the heat storage microcapsule are measured in the laboratory.

[0122] S105: The heat exchange channel operating condition information includes the heat boundary conditions of the channel heat exchanger as constant heat flow on the heat exchange surface and adiabatic surface, constant working fluid temperature at the channel inlet, Reynolds number of the working fluid at the channel inlet ranging from 50 to 2000, and pressure at the channel outlet of 0.

[0123] S106: Performance evaluation parameters for heat exchange channels include surface heat transfer coefficient h and pressure drop pd;

[0124] S2. Methods for obtaining the optimal design structure and parameters of the channel heat exchanger cross-section structure based on the design space and load characteristic data of the channel heat exchanger cross-section structure include:

[0125] S201: Generate the corresponding channel cross-section structure based on the design conditions of the heat exchange channel structure cross-section;

[0126] S201-1: Based on the structural design conditions of the hot channel cross section and the symmetry and periodicity of the channel heat exchanger, the minimum design unit of the channel is determined. Under the premise of applying the structural design conditions of the hot channel cross section, the minimum design unit can be reassembled into the channel heat exchanger through symmetrical and periodic arrangement.

[0127] S201-2: The cross-section of the minimum design unit is the channel cross-section, which includes the working fluid domain cross-section and the wall domain cross-section, wherein the working fluid domain cross-section should not exceed the range of the wall domain area;

[0128] S201-3: Determine the length-to-width ratio range and the thinnest wall thickness of the channel cross-section wall region of the minimum design unit;

[0129] S201-4: Select the planned regular working medium domain cross-sectional shape and random working medium domain cross-sectional shape as: circular and elliptical.

[0130] S201-5: Based on the requirements of S102, S201-2 and S201-3, the size and area parameters of the planned working fluid domain cross-sectional shape are predefined respectively, and the corresponding working fluid domain cross-sectional shape that meets the requirements and the working fluid domain cross-sectional geometry that can be called by the 3D drawing software SolidWorks are generated.

[0131] S201-6: Based on the requirements of S102, S201-1 and S201-4, create the wall domain cross-sectional shape that surrounds the corresponding working fluid domain cross-sectional shape, as well as the wall domain cross-sectional geometry that can be called by the 3D drawing software SolidWorks;

[0132] S201-7: Merge the working fluid domain section geometry and the wall domain section geometry into a composite shape and export it as a STEP file format, which can be used by the 3D drawing software SolidWorks as a channel section geometry, and export the channel section image.

[0133] S201-8: In the cross-sectional image of the channel, the channel wall region is black, and the working fluid region is white;

[0134] Preferably, in step S204, based on the design conditions of the heat exchange channel cross-section, the channel cross-section structure, and the performance evaluation parameter results, the channel cross-section dataset includes: the channel cross-section structure image, the variable design variables of the corresponding channel cross-section structure, and the results of the corresponding channel structure performance evaluation parameters to be optimized based on step S203 and the above two points.

[0135] S202: Based on the design conditions of the heat exchange channel cross-section and the channel cross-section, the corresponding channel structure model is generated and exported by calling the external 3D drawing software SolidWorks.

[0136] S203: Based on the channel structure model, establish data transmission with the multiphysics coupling simulation software ANSYS, and obtain the performance evaluation parameter results of the corresponding channel structure based on the heat exchange channel structure cross-sectional structure design conditions and channel structure model;

[0137] S204: Based on the structural design conditions of the heat exchange channel cross-section, the channel cross-section structure and performance evaluation parameters, the channel cross-section dataset is obtained;

[0138] S205: Based on the channel cross-section dataset, a deep learning channel cross-section performance prediction and evaluation infrastructure template is pre-designed to obtain the channel cross-section performance prediction and evaluation space;

[0139] S205-1: Adjust the channel cross-sectional image to a square of a specified size while maintaining the aspect ratio information to ensure that the content of the image is not distorted due to scaling. The reference size for the image resolution is 1600 pixels.

[0140] S205-2: Based on the channel cross-section dataset, take the adjusted cross-section image and the variable design variables of the corresponding channel cross-section structure as input, and take the results of the performance evaluation parameters of the corresponding channel structure to be optimized as output;

[0141] S205-3: Divide the cross-sectional image into several blocks, each block being 40 pixels in size, and use a convolutional layer to convert these blocks into embedding vectors, with the embedding vectors having a dimension of 192.

[0142] S205-4: Add position encoding to the image blocks to ensure that the transformer can understand the relative position of the image blocks;

[0143] S205-5: Convert the variable design variables of the corresponding channel cross-section structure into an embedding vector with the same dimension as the image block;

[0144] S205-6: Input the above embedding vector into the Transformer encoder to obtain the final feature representation; the Transformer encoder is composed of multiple stacked Transformer encoder layers, each encoder containing a self-attention mechanism and a feedforward neural network; the Transformer encoder has 8 layers; each encoder layer has 6 self-attention heads; the hidden layer size ratio in the feedforward neural network is 4.0;

[0145] S205-7: The final feature representation of each of the performance evaluation parameters to be optimized is mapped to a specific target predicted value through the output layer;

[0146] S205-8: Calculate the loss for each of the performance evaluation parameters to be optimized, and then perform backpropagation to update the model parameters;

[0147] S205-9: Iterate from S205-6 to S205-8, calculate the loss on the validation set, and stop training early if the loss of the main performance evaluation parameter to be optimized in the validation set has not improved within a certain number of rounds. The learning rate decay range is 1e-4.

[0148] S205-10: The output includes the model's weights and statistical information related to model training, representing the best performing model.

[0149] S206: Based on the channel cross-section performance prediction and evaluation space, initialize the channel cross-section structure, design parameters, and operating parameter weights;

[0150] S206-1: Set the weights of the performance evaluation parameters to be optimized;

[0151] S206-2: Constructing an image of the channel cross-section based on a shape model; the shape model is a hyperellipse model;

[0152] S206-3: Set the variable design variables and parameter range of the shape model of the cross-sectional structure;

[0153] S206-4: Construct an image of the channel cross-section based on the shape model and model parameter settings;

[0154] S206-5: Standardize the channel cross-section image constructed in S206-4 and the variable design variables of the cross-section structure, input the channel cross-section performance prediction and evaluation space, and predict the values ​​of the performance evaluation parameters to be optimized;

[0155] S207: Based on the channel cross-section performance prediction and evaluation space, the optimal channel cross-section structure, design parameters and operating parameters are obtained;

[0156] S207-1: The random initialization method generates a set of variable design variables for the channel cross-section image and corresponding cross-section structure that satisfy the settings of S206, as a set of design samples;

[0157] S207-2: Based on the design sample, use the evaluation component to evaluate the design sample and the corresponding performance evaluation parameters to be optimized, and obtain the evaluation component's calculated score;

[0158] S207-3: Calculate the score based on the evaluation component and calculate the weight gradient of each parameter within the design sample;

[0159] S207-4: Based on the aforementioned weight gradient and optimization rate, correct the weights of the corresponding channel cross-sectional structure, design parameters, and operating parameters;

[0160] S207-5: Iteratively optimize S207-1 to S207-4 to obtain the optimal channel cross-section image, design parameters, and operating parameters.

[0161] S3. Based on the optimal design structure and parameters of the channel heat exchanger cross-sectional structure, the method for deriving the structural model and completing the design of the channel heat exchanger cross-sectional structure includes:

[0162] S301: Optimizes the channel cross-section structure based on the channel cross-section, and exports the channel cross-section structure in STEP format, which can be used by the 3D drawing software SolidWorks.

[0163] S301-1: Based on the optimized channel cross-section image, design parameters, and operating parameters, map the image pixels to the actual physical dimensions;

[0164] S301-2: Based on S301-1, create the optimal wall domain section geometry that can be called by the 3D drawing software SolidWorks;

[0165] S301-3: Based on S301-1, create the optimal working fluid domain cross-sectional geometry that can be called by the 3D drawing software SolidWorks;

[0166] S301-4: Merge the optimal wall domain section geometry and the optimal working fluid domain section geometry into a composite shape and export it as a STEP file format, which can be used as a channel section geometry that can be called by the 3D drawing software SolidWorks.

[0167] S302: Based on the optimal channel design parameters and STEP format channel cross-section structure, the corresponding optimal channel structure model is generated and exported by calling the 3D drawing software SolidWorks.

[0168] S303: Based on the optimized channel operating parameters and optimized channel structure model of the channel cross section, establish data transmission with the multiphysics coupling simulation software ANSYS to obtain the performance evaluation parameter results of the corresponding channel structure.

[0169] Figure 6 This figure shows the fitting effect of the prediction model of the heat exchange channel cross-section design system based on artificial intelligence algorithm optimization of the present invention on the convective heat transfer coefficient h. As can be seen from the figure, all data points are closely distributed on both sides of the reference line, and the overall trend is highly consistent with the reference line, indicating that the predicted value of the variable h by the calibrated model has a very high consistency with the actual value.

[0170] Figure 7 This figure shows the fitting effect of the prediction model of the heat exchange channel cross-section design system based on artificial intelligence algorithm optimization of the present invention on the pressure drop pd. As can be seen from the figure, the data points are generally distributed along the reference line. The points in the low value range are well concentrated. Although there are a few discrete points in the high value range, there is no obvious systematic shift. The overall trend is highly consistent with the reference line, indicating that the model's prediction of the variable pd after calibration has a high degree of consistency with the actual value.

[0171] 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 method for designing the cross-section of a heat exchange channel based on artificial intelligence algorithm optimization, comprising the following steps: S1: Initialize the design conditions for the cross-section of the channel heat exchanger; The structural design conditions for determining the heat exchange channel cross-section include heat exchange channel structural information, heat exchange channel material information, working fluid information, operating condition information, and performance evaluation parameters; And determine the known design variables, variable design variables, and performance evaluation parameters to be optimized in the structural design conditions of the heat exchange channel cross section; S2: Based on the design space and load characteristic data of the channel heat exchanger cross-sectional structure, the optimal design structure and parameters of the channel heat exchanger cross-sectional structure are obtained; Specifically, it includes the following sub-steps: S201: Generate the corresponding channel cross-section structure based on the design conditions of the heat exchange channel structure cross-section; S202: Based on the design conditions of the heat exchange channel cross-section and the channel cross-section structure, the corresponding channel structure model is generated and exported by calling three-dimensional drawing software; S203: Establish data transmission between the channel structure model and the software used in flow, heat transfer, mechanical analysis and calculation; obtain the performance evaluation parameter results of the corresponding channel structure based on the heat transfer channel structure cross-sectional design conditions and the channel structure model. S204: Based on the structural design conditions of the heat exchange channel cross-section, the channel cross-section structure and performance evaluation parameters, the channel cross-section dataset is obtained; The channel cross-section dataset includes: images of the channel cross-section structure, the variable design variables of the corresponding channel cross-section structure, and the results of the performance evaluation parameters of the corresponding channel structure to be optimized. S205: Based on the channel cross-section dataset, a deep learning channel cross-section performance prediction and evaluation infrastructure template is pre-designed to obtain the channel cross-section performance prediction and evaluation space; S206: Based on the channel cross-section performance prediction and evaluation space, initialize the channel cross-section structure, design parameters, and operating parameter weights; S207: Based on the channel cross-section performance prediction and evaluation space, the optimal channel cross-section structure, design parameters and operating parameters are obtained; S3: Based on the optimal design structure and parameters of the channel heat exchanger cross-section structure, derive the structural model and complete the design of the channel heat exchanger cross-section structure.

2. The method according to claim 1, characterized in that: The structural information of the heat exchange channel includes the dimensions of the channel heat exchanger, the symmetry of the channel heat exchanger, the periodicity of the channel heat exchanger, the dimensional relationship between the channels, and the ratio of the volume inside the channel to the total volume of the channel heat exchanger. The heat exchange channel material information is selected from at least one of the following: channel material type, corresponding material density, corresponding material thermodynamic parameters, channel inner wall friction coefficient, corresponding material elastic mechanical parameters, corresponding material plasticity and strength parameters, corresponding material fatigue and fracture parameters, and application scenario contact parameters. The working fluid information for the heat exchange channel is selected from at least one of the following: working fluid density, working fluid viscosity, working fluid thermal conductivity, and working fluid specific heat capacity. The operating information of the heat exchange channel includes the thermal boundary conditions of the channel heat exchanger, the working fluid temperature at the channel inlet, the working fluid flow rate at the channel inlet, the pressure conditions at the channel outlet, and the force and pressure loads. The performance evaluation parameters for heat exchange channels are selected from at least one of the following: surface heat transfer coefficient, pressure drop, thermal resistance, dimensionless temperature, friction coefficient, Nusselt number, flow resistance characteristic coefficient, comprehensive heat transfer characteristic coefficient, stress distribution, strain analysis, displacement analysis, thermal stress, and fatigue analysis.

3. The method according to claim 2, characterized in that: The working fluid in the heat exchange channel is made of two or more immiscible materials; For non-solid working fluids, the working fluid information includes the density, viscosity, thermal conductivity, specific heat capacity, and volume fraction of each working fluid. For working fluids containing solid particles, the working fluid information includes particle size, particle density, particle thermal conductivity, particle specific heat capacity, and particle volume concentration.

4. The method according to claim 3, characterized in that: Step S201 includes the following sub-steps: S201-1: Based on the structural design conditions of the hot channel cross section and the symmetry and periodicity of the channel heat exchanger, determine the minimum design unit of the channel; under the premise of applying the structural design conditions of the hot channel cross section, the minimum design unit is reassembled into a channel heat exchanger through symmetrical and periodic arrangement; S201-2: The cross section of the minimum design unit is the channel cross section, which includes the working fluid domain cross section and the wall domain cross section. The working fluid domain cross section should not exceed the range of the wall domain area. S201-3: Determine the length-to-width ratio range and the thinnest wall thickness of the minimum design unit channel cross-section wall region; S201-4: Select the planned regular working medium domain cross-sectional shape and random working medium domain cross-sectional shape; S201-5: Predefine the size and area parameters of the planned working fluid domain cross-sectional shape, generate the corresponding working fluid domain cross-sectional shape that meets the requirements, and the working fluid domain cross-sectional geometry used in external 3D drawing software; S201-6: Create the wall domain cross-sectional shape that encloses the corresponding working fluid domain cross-sectional shape, as well as the wall domain cross-sectional geometry used in external 3D drawing software; S201-7: Merge the working fluid domain section geometry and the wall domain section geometry into a composite shape, export it as an image file format, use it as a channel section geometry for external 3D drawing software, and export the channel section image.

5. The method according to claim 4, characterized in that: Step S205 includes the following sub-steps: S205-1: Adjust the channel cross-section image to a square of the specified size while maintaining the aspect ratio information; S205-2: Based on the channel cross-section dataset, the adjusted cross-section image and the variable design variables of the corresponding channel cross-section structure are used as input, and the results of the performance evaluation parameters of the corresponding channel structure to be optimized are used as output; S205-3: Divide the cross-sectional image into several blocks and use a convolutional layer to convert these image blocks into embedding vectors with dimensions ranging from 128 to 512. S205-4: Add position encoding to the image blocks so that the transformer can understand the relative positions of the image blocks; S205-5: Convert the variable design variables of the corresponding channel cross-section structure into an embedding vector of the same dimension as the image block; S205-6: Input the above embedding vectors into the Transformer encoder to obtain the final feature representation; the Transformer encoder is composed of multiple stacked Transformer encoder layers, each encoder containing a self-attention mechanism and a feedforward neural network; the number of Transformer encoder layers is 4 to 12; the number of self-attention heads in each encoder layer is 4 to 12; the scale ratio of hidden layers in the feedforward neural network is 2.0 to 8.0; S205-7: The output layer maps the final feature representation of each performance evaluation parameter to be optimized to the specific target prediction value; S205-8: Calculate the loss for each performance evaluation parameter to be optimized, and then perform backpropagation to update the model parameters; S205-9: After iteration, the output is the best-performing model, which includes the model's weights and statistical information related to the model's training.

6. The method according to claim 5, characterized in that: Step S206 includes the following sub-steps: S206-1: Set the weights and upper and lower limits of the performance evaluation parameters to be optimized; S206-2: Constructing images of channel cross-sections based on shape models; S206-3: Define the variable design variables and parameter range of the shape model for the cross-sectional structure; S206-4: Construct an image of the channel cross-section based on the shape model and model parameter settings; S206-5: Standardize the constructed channel cross-section image and the variable design variables of the cross-section structure, input the channel cross-section performance prediction and evaluation space, and predict the values ​​of the performance evaluation parameters to be optimized.

7. The method according to claim 6, characterized in that, Step S207 includes the following sub-steps: S207-1: The random initialization method generates variable design variables for the channel cross-section image and corresponding cross-section structure that meet the set requirements, as a set of design samples; S207-2: Based on the design sample, use the evaluation component to evaluate the design sample and the corresponding performance evaluation parameters to be optimized, and obtain the evaluation component's score; S207-3: Calculate the score based on the evaluation component and calculate the weight gradient of each parameter within the design sample; S207-4: Based on the weight gradient and optimization rate, correct the weights of the corresponding channel cross-sectional structure, design parameters and operating parameters; S207-5: Iterative optimization to obtain the optimal channel cross-section image, design parameters, and operating parameters.

8. The method according to claim 7, characterized in that, Step S3 includes the following sub-steps: S301: Optimized channel cross-section structure based on channel cross-section, deriving channel cross-section structures used in external 3D drawing software; S302: Based on the optimal channel design parameters and channel cross-section structure, generate and export the corresponding optimal channel structure model using 3D drawing software.

9. The method according to claim 8, characterized in that, Step S301 specifically includes the following sub-steps: S301-1: Based on the optimized channel cross-section image, design parameters, and operating parameters, the image pixels are mapped to the actual physical size; S301-2: Create the optimal wall region section geometry for use in 3D drawing software; S301-3: Create the optimal working fluid domain section geometry for use in 3D drawing software; S301-4: Merge the optimal wall domain cross-sectional geometry and the optimal working medium domain cross-sectional geometry into a composite shape and export it as an image file format as a channel cross-sectional structure.

10. A heat exchange channel cross-section design system based on artificial intelligence algorithm optimization, characterized in that: The system includes an initialization module, a design optimization module, and an output module; The initialization module is used to initialize the cross-sectional design conditions of the channel heat exchanger. The design optimization module is used to obtain the optimal design structure and parameters of the channel heat exchanger cross-section structure based on the design space and load characteristic data of the channel heat exchanger cross-section structure. The output module is used to optimize the design structure and parameters based on the cross-sectional structure of the channel heat exchanger, derive the structural model, and complete the design of the cross-sectional structure of the channel heat exchanger. The above modules work together to achieve the method described in claims 1-9.

Images