Arrayed micro-channel radiator heat flow coupling simulation method, system, device and medium

Abstract

The application discloses an arrayed micro-channel radiator heat flow coupling simulation method, system, equipment and medium, wherein the method comprises the following steps: performing three-dimensional computational fluid dynamics simulation on multiple working conditions of a single micro-channel radiator, generating input samples through Latin hypercube sampling and performing batch simulation, extracting flow field data to form a high-dimensional simulation dataset; training element flow characteristic parameter data through a convolutional neural network and performing dimension reduction processing, so that a reduced-order model is constructed; exporting the trained reduced-order model into a one-dimensional heat network simulation solver recognizable file to form a callable reduced-order element model; based on an actual pipe network of an arrayed system, a one-dimensional pipe topology is constructed, and the reduced-order element model is connected to each branch as a core heat exchange element; a conversion model is constructed, a simulation step is set, and steady-state heat flow coupling simulation is performed, and the pressure drop, temperature distribution and flow distribution of each branch are output. The application can significantly improve the simulation speed of the arrayed system, and support rapid design iteration and parameter optimization.

Description

Technical Field

[0001] This invention relates to the field of thermal management and system simulation technology for electronic devices, and in particular to a thermal flux coupling simulation method, system, device and medium for arrayed microchannel heat sinks. Background Technology

[0002] With the widespread application of highly integrated phased array systems in radar, communications, and other fields, the number of system channels can reach hundreds to thousands. The temperature consistency of the power amplifier chips in each channel directly affects the amplitude and phase consistency of the system. To ensure uniform chip operating temperature, it is necessary to ensure uniform flow distribution of coolant in the arrayed microchannel heat sink. This places high demands on both high accuracy and high efficiency in the thermal-fluid coupling simulation of the heat dissipation system.

[0003] Currently, while mainstream three-dimensional computational fluid dynamics (CFD) simulation methods can accurately simulate the flow and heat transfer processes within microchannels, their complex modeling, high computational resource consumption, and long simulation cycles make them unsuitable for the rapid iteration requirements of engineering design phases. Especially in arrayed systems, if all heat dissipation units are modeled in three dimensions, the computational cost increases exponentially, severely limiting design efficiency. One-dimensional system-level simulation methods offer advantages such as convenient modeling and fast simulation speed, making them suitable for rapid verification of system-level solutions. However, traditional one-dimensional models struggle to accurately characterize the complex three-dimensional flow and heat transfer characteristics inside microchannel heat sinks, limiting simulation accuracy.

[0004] In the existing technology, some studies have attempted to perform thermal analysis on three-dimensional packaging structures through the Reduced-Order Model (ROM) method to solve the problems of large network size and difficulty in simulation based on the finite element method in cross-scale three-dimensional heterogeneous integrated models. However, ROM has not yet effectively solved the problem of efficient coupling simulation of arrayed microchannel heat sinks with one-dimensional pipeline systems under multiple operating conditions.

[0005] Therefore, there is an urgent need for a method that balances simulation accuracy and computational efficiency to achieve fast and accurate thermal-fluid coupling simulation of arrayed microchannel heat sink systems, providing an effective tool for the thermal design of highly integrated electronic systems. Summary of the Invention

[0006] To address the aforementioned issues, this invention proposes a thermal flux coupling simulation method, system, device, and medium for arrayed microchannel heat sinks. This method significantly improves the simulation speed of arrayed systems while ensuring simulation accuracy, and supports rapid design iteration and parameter optimization.

[0007] The technical solution adopted in this invention is as follows: A simulation method for thermal-fluid coupling of arrayed microchannel heat sinks includes: Three-dimensional computational fluid dynamics simulation of a single microchannel heat sink under multiple operating conditions was performed. Input samples were generated by Latin hypercube sampling and batch simulation was performed. Flow field data was extracted to form a high-dimensional simulation dataset. Based on the high-dimensional simulation dataset, a reduced-order model is constructed by training the component flow characteristic parameter data through a convolutional neural network and then performing dimensionality reduction processing. The trained reduced-order model is exported as a file that can be recognized by a one-dimensional thermal network simulation solver, forming a callable reduced-order component model; A one-dimensional pipeline topology is constructed based on the actual pipeline network of the arrayed system, and the reduced-order element model is used as the core heat exchange element and connected to each branch. Construct a conversion model to match the units and dimensions of the system variables and model parameters, set the simulation step size to perform steady-state thermal-fluid coupling simulation, and output the pressure drop, temperature distribution and flow distribution results of each branch.

[0008] Furthermore, the three-dimensional computational fluid dynamics simulation of a single microchannel heat sink under multiple operating conditions includes: combining various local resistance elements, heat exchange elements, power elements and heat conduction elements in the arrayed microchannel heat sink system to perform three-dimensional computational fluid dynamics simulation of a single microchannel heat sink under different operating conditions.

[0009] Furthermore, the step of training the flow characteristic parameter data of the element through a convolutional neural network and performing dimensionality reduction includes: firstly, extracting convolutional features from the high-dimensional data, then using combined pooling to reduce the data dimensionality, and then weighting and integrating the dimensionality-reduced data to form a fully connected layer, thereby fitting the nonlinear mapping relationship between the geometric features of the element, operating conditions and flow characteristic parameters, and completing the construction of the reduced-order model.

[0010] Furthermore, the step of exporting the trained reduced-order model into a file recognizable by a one-dimensional thermal network simulation solver to form a callable reduced-order component model includes: standardizing and encapsulating the trained reduced-order model to obtain a reduced-order component model. The encapsulated reduced-order component model has an input-output parameter system adapted to the one-dimensional thermal network simulation solver. The input-output parameter system matches the structural features and flow heat transfer characteristics of the microchannel radiator and can be directly called by the one-dimensional thermal network simulation solver.

[0011] Furthermore, the construction of a one-dimensional pipeline topology based on the actual pipeline network of the arrayed system, with the reduced-order element model as the core heat exchange element connected to each branch, includes: constructing a one-dimensional pipeline topology based on the actual connection relationship, flow resistance characteristics, and array layout characteristics of the arrayed microchannel radiator system pipeline network. The topology includes various local resistance elements corresponding to the actual pipeline network, and the reduced-order element model is connected to each branch of the one-dimensional pipeline topology according to the heat exchange branch distribution characteristics of the arrayed system.

[0012] Furthermore, the construction of the conversion model performs unit and dimension matching between system variables and model parameters, including: constructing a conversion model based on the basic physical laws and dimensional conversion rules of thermal-fluid coupling simulation, so as to realize the dimensional conversion and physical quantity matching between system thermal flux variables in the one-dimensional thermal network simulation environment and the characteristic parameters of the reduced-order component model.

[0013] Furthermore, the step size for setting the steady-state thermal-fluid coupling simulation and outputting the pressure drop, temperature distribution and flow distribution results of each branch includes: achieving steady-state thermal-fluid coupling simulation through the collaborative solution of the reduced-order element model and the one-dimensional pipeline topology; completing the overall thermal-fluid field solution of the arrayed microchannel radiator system with a preset time step; and outputting results that can fully characterize the flow distribution state, temperature distribution law and fluid resistance characteristics of each branch of the system.

[0014] A thermal-fluid coupling simulation system for arrayed microchannel heat sinks includes: The high-dimensional simulation dataset construction module is configured to perform three-dimensional computational fluid dynamics simulations of a single microchannel radiator under multiple operating conditions. Input samples are generated by Latin hypercube sampling and batch simulations are performed to extract flow field data and form a high-dimensional simulation dataset. The reduced-order model construction module is configured to construct a reduced-order model based on the high-dimensional simulation dataset by training the component flow characteristic parameter data through a convolutional neural network and performing dimensionality reduction processing. The model encapsulation and integration module is configured to export the trained reduced-order model as a file recognizable by a one-dimensional thermal network simulation solver, forming a callable reduced-order component model. The one-dimensional topology construction module is configured to construct a one-dimensional pipeline topology based on the actual pipeline network of the arrayed system, and to connect the reduced-order element model as the core heat exchange element to each branch. The data conversion and coupling simulation module is configured to build a conversion model, perform unit and dimension matching between system variables and model parameters, set the simulation step size to perform steady-state heat-fluid coupling simulation, and output the pressure drop, temperature distribution and flow distribution results of each branch.

[0015] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the thermal flux coupling simulation method for an arrayed microchannel heat sink.

[0016] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the thermal-fluid coupling simulation method for an arrayed microchannel heat sink.

[0017] The beneficial effects of this invention are as follows: This invention combines three-dimensional computational fluid dynamics simulation, data-driven model reduction techniques, and one-dimensional system-level simulation to construct a cross-scale thermal-fluid coupling simulation system adapted to arrayed microchannel heat sinks. It effectively solves the problems in existing technologies, such as the difficulty in balancing simulation accuracy and efficiency, high simulation cost of arrayed systems, and poor coupling between reduced-order models and one-dimensional piping systems. It achieves significant improvements in simulation efficiency, accuracy, and scalability, while also possessing outstanding engineering practical value and platform adaptability. The specific details are as follows.

[0018] 1. Significantly improves the overall efficiency of thermal-fluid coupling simulation of arrayed systems, adapting to the needs of rapid engineering iteration. This invention first conducts multi-condition three-dimensional computational fluid dynamics simulations using a single microchannel radiator as the object. Input samples are efficiently generated and batch simulations are completed through Latin hypercube sampling, avoiding sample redundancy and data loss caused by random sampling. Then, a convolutional neural network is used to extract features and reduce the dimensionality of the high-dimensional data. The constructed reduced-order model can replace the complex three-dimensional simulation model to quickly calculate the flow and heat transfer characteristics, eliminating the need for three-dimensional modeling of each heat dissipation unit in the arrayed system. Simultaneously, the reduced-order model is standardized and encapsulated into a component model that can be directly called by a one-dimensional thermal network simulation solver. Combined with the actual pipe network, a one-dimensional pipe topology can be quickly constructed, eliminating the cumbersome process of model format conversion and parameter re-adjustment. The entire process, from high-dimensional dataset generation to steady-state thermal-fluid coupling simulation, achieves improved computational efficiency, significantly shortening the simulation cycle of the arrayed microchannel radiator system and meeting the needs of rapid adjustment and iteration of schemes during the engineering design phase.

[0019] 2. Ensure high accuracy of simulation results to accurately reflect the actual heat flow characteristics of arrayed microchannel heat sinks. The high-dimensional simulation dataset constructed in this invention combines various local resistance elements, heat exchange elements, power elements, and heat conduction elements in an arrayed microchannel heat dissipation system, fully covering the core component characteristics of the system and providing comprehensive raw data support for training the reduced-order model. A convolutional neural network accurately mines the correlation between component geometric features, operating conditions, and flow characteristic parameters in the high-dimensional data through convolutional feature extraction. Combinatorial pooling dimensionality reduction avoids overfitting while preserving core data characteristics. A fully connected layer fits the nonlinear mapping relationship between the three, enabling the reduced-order model to accurately characterize the complex three-dimensional flow and heat exchange characteristics inside the microchannel. Simultaneously, a transformation model is constructed based on the fundamental physical laws and dimensional transformation rules of heat-fluid coupling simulation, achieving accurate matching between one-dimensional system variables and reduced-order model parameters, avoiding errors in parameter transfer. Finally, the overall system thermal flow field is solved through the collaborative solution of the reduced-order component model and the one-dimensional pipeline topology. The simulation results accurately reflect the actual flow distribution, temperature distribution, and fluid resistance characteristics of the arrayed microchannel heat sink, and the error between the simulation results and the full-order three-dimensional computational fluid dynamics simulation results meets the accuracy requirements of engineering design.

[0020] 3. It possesses excellent scalability, adapting to the simulation needs of different types of arrayed microchannel heat dissipation systems. The simulation method of this invention adopts a phased process design, from the construction of high-dimensional simulation datasets and the training of reduced-order models, to the encapsulation of reduced-order component models and the construction of one-dimensional pipeline topology. Each step is independent yet seamlessly connected. The reduced-order component model of a single microchannel heat sink can be directly reused in microchannel heat dissipation systems of different array sizes and layouts without retraining the model. The one-dimensional pipeline topology can be flexibly constructed and adjusted according to the actual connection relationship of the arrayed system's pipeline network, the flow resistance characteristics, and the array layout characteristics. Various local resistance components can be added to the topology according to actual needs, and the reduced-order component model can also be flexibly integrated according to the heat exchange branch distribution characteristics, adapting to the simulation needs of arrayed microchannel heat sink systems with different structural forms. Furthermore, the simulation system of this invention adopts a modular design, and each functional module can be independently upgraded and optimized, facilitating the subsequent expansion of the model's predictive capabilities and simulation range to meet new simulation requirements.

[0021] 4. It possesses high engineering practical value, providing precise data support for the design optimization of arrayed microchannel heat sinks. The simulation method of this invention can output the pressure drop, temperature distribution, and flow distribution results of each branch of the arrayed microchannel heat sink system, fully characterizing the system's flow distribution state, temperature distribution law, and fluid resistance characteristics. It can directly establish the mapping relationship between coolant flow distribution and chip temperature consistency. Based on the simulation results, engineering designers can accurately locate problems in the heat sink flow channel design and pipe network layout that lead to uneven flow distribution and excessively high local temperatures. It provides scientific and accurate data basis for the structural optimization of the distribution network, the adjustment of flow channel parameters, and the layout design of heat exchange branches, enabling the thermal design of arrayed microchannel heat sinks to shift from experience-based design to digital and precise design, effectively improving the heat dissipation effect and engineering design quality of the heat sink.

[0022] 5. Eliminating dependence on specific commercial simulation software, adapting to the application needs of domestic platforms, and enhancing the independent controllability of technology. The reduced-order model of this invention is encapsulated in a universal file format recognizable by a one-dimensional thermal network simulation solver. The constructed input-output parameter system does not depend on specific commercial simulation software. The construction of the one-dimensional pipeline topology, the design of the transformation model, and the execution of steady-state thermal-fluid coupling simulation are all based on the fundamental physical laws of thermal-fluid coupling simulation, rather than the exclusive functions of specific software. This allows the simulation method to be quickly migrated to domestic system simulation tools and adapted to the application needs of domestic software and hardware platforms. Simultaneously, by solidifying the simulation method into a computer program and storing it in a computer-readable storage medium, cross-device portability and promotion can be achieved, reducing the engineering field's dependence on foreign commercial simulation software and enhancing the independent controllability of thermal management simulation technology for electronic equipment. Attached Figure Description

[0023] Figure 1 This is a flowchart of a thermal flux coupling simulation method for an arrayed microchannel heat sink according to Embodiment 1 of the present invention.

[0024] Figure 2 This is a schematic diagram of the cubic convolution operation in Embodiment 2 of the present invention.

[0025] Figure 3 This is a schematic diagram of the component parameterization list and parameterization structure of Embodiment 3 of the present invention.

[0026] Figure 4 This is a schematic diagram of the components added to the system according to Embodiment 3 of the present invention.

[0027] Figure 5 This is a three-dimensional schematic diagram of the heterogeneous integrated multiphysics simulation model of Embodiment 3 of the present invention.

[0028] Figure 6 This is a schematic diagram of a one-dimensional fluid model of the heterogeneous integrated multiphysics simulation model of Embodiment 3 of the present invention.

[0029] Figure 7 This is a comparison chart of the one-dimensional and three-dimensional simulation results of the total inlet pressure in Embodiment 3 of the present invention.

[0030] Figure 8 This is a comparison chart of the total outlet pressure results of one-dimensional and three-dimensional simulations in Embodiment 3 of the present invention. Detailed Implementation

[0031] To provide a clearer understanding of the technical features, objectives, and effects of the present invention, specific embodiments are now described. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention; that is, the described embodiments are only a part of the embodiments of the invention, not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0032] Example 1 like Figure 1 As shown, this embodiment provides a thermal-fluid coupling simulation method for arrayed microchannel heat sinks, including: Three-dimensional computational fluid dynamics simulation of a single microchannel heat sink under multiple operating conditions was performed. Input samples were generated by Latin hypercube sampling and batch simulation was performed. Flow field data was extracted to form a high-dimensional simulation dataset. Based on the high-dimensional simulation dataset, a reduced-order model is constructed by training the component flow characteristic parameter data through a convolutional neural network and then performing dimensionality reduction processing. The trained reduced-order model is exported as a file that can be recognized by a one-dimensional thermal network simulation solver, forming a callable reduced-order component model; A one-dimensional pipeline topology is constructed based on the actual pipeline network of the arrayed system, and the reduced-order element model is used as the core heat exchange element and connected to each branch. Construct a conversion model to match the units and dimensions of the system variables and model parameters, set the simulation step size to perform steady-state thermal-fluid coupling simulation, and output the pressure drop, temperature distribution and flow distribution results of each branch.

[0033] It should be noted that this method combines the high precision of 3D computational fluid dynamics simulation with the high computational efficiency of the reduced-order model, enabling cross-scale simulation of the thermal-fluid coupling of arrayed microchannel heat sinks. This ensures the accuracy of the simulation results while significantly improving simulation efficiency. Furthermore, by constructing a one-dimensional pipe topology and building a transformation model, effective coupling between the reduced-order model and the one-dimensional thermal network simulation system is achieved. This allows for the complete output of the thermal-fluid characteristics of the arrayed system, providing comprehensive data support for the design and optimization of arrayed microchannel heat sinks.

[0034] Preferably, three-dimensional computational fluid dynamics simulation of a single microchannel heat sink is performed under multiple operating conditions, including: combining various local resistance elements, heat exchange elements, power elements and heat conduction elements in the arrayed microchannel heat sink system to perform three-dimensional computational fluid dynamics simulation of a single microchannel heat sink under different operating conditions.

[0035] Specifically, before conducting three-dimensional computational fluid dynamics simulations of a single microchannel radiator, the types and structural characteristics of the local resistance elements, heat exchange elements, power elements, and heat conduction elements contained in the arrayed microchannel heat dissipation system are first sorted out. The structure and working characteristics of various elements are then integrated into the simulation model of a single microchannel radiator. Then, different actual working conditions such as heat dissipation, medium flow state, and inlet temperature are simulated. Three-dimensional computational fluid dynamics simulations are then carried out on the microchannel radiator model that incorporates the characteristics of various elements in sequence to fully simulate the flow field and heat exchange state of various elements under different working conditions.

[0036] It should be noted that this step integrates the features of various core components in the arrayed system into the simulation process of a single heat sink, so that the generated high-dimensional simulation dataset can fully cover the component features of the arrayed system. This avoids insufficient training of the subsequent reduced-order model due to missing component features, effectively improves the adaptability of the reduced-order model to the arrayed microchannel cooling system, and lays a complete data foundation for the accurate training of the subsequent model.

[0037] Preferably, the flow characteristic parameter data of the component is trained by a convolutional neural network and then dimensionality reduced, including: firstly, performing convolutional feature extraction on the high-dimensional data, then using combined pooling to reduce the data dimensionality, and then weighting and integrating the dimensionality-reduced data to form a fully connected layer, thereby fitting the nonlinear mapping relationship between the component's geometric features, operating conditions and flow characteristic parameters, and completing the construction of the reduced-order model.

[0038] Specifically, after inputting the component flow characteristic parameter data from the high-dimensional simulation dataset into the convolutional neural network, the network's convolutional layers first extract local features from the high-dimensional data to uncover the correlation features between component geometric features, operating conditions, and flow characteristics. Then, combined pooling is used to reduce the dimensionality of the extracted convolutional feature data, reducing the number of data parameters while retaining core data features. Subsequently, the dimensionality-reduced feature data is weighted and integrated to construct a fully connected layer of the neural network. Through the fitting operation of the fully connected layer, a nonlinear mapping relationship between component geometric features, operating conditions, and flow characteristic parameters is established, completing the training and construction of the entire reduced-dimensional model.

[0039] It should be noted that this step accurately mines the core correlation features of high-dimensional data through convolutional feature extraction, and achieves efficient dimensionality reduction by combining pooling, thus avoiding the overfitting problem. At the same time, the nonlinear mapping relationship is fitted through a fully connected layer, so that the constructed reduced-order model can accurately characterize the complex flow and heat transfer characteristics of microchannel heat sink components, ensuring the prediction accuracy of the reduced-order model and realizing accurate prediction of the flow and heat transfer process of the components.

[0040] Preferably, the trained reduced-order model is exported as a file recognizable by a one-dimensional thermal network simulation solver to form a callable reduced-order component model, including: standardizing and encapsulating the trained reduced-order model to obtain a reduced-order component model. The encapsulated reduced-order component model has an input-output parameter system adapted to the one-dimensional thermal network simulation solver. The input-output parameter system matches the structural features and flow heat transfer characteristics of the microchannel radiator and can be directly called by the one-dimensional thermal network simulation solver.

[0041] Specifically, after the reduced-order model is trained, it is standardized and packaged according to the file format and calling specifications of the one-dimensional thermal network simulation solver. Based on the actual structural characteristics and flow heat transfer characteristics of the microchannel radiator, an input-output parameter system that matches the prediction capability of the reduced-order model is built. This allows the input parameters of the parameter system to correspond to the output variables of the one-dimensional thermal network simulation solver, and the output parameters to be directly recognized and called by the solver. Finally, a reduced-order component model that can be directly accessed in the one-dimensional thermal network simulation environment is formed.

[0042] It should be noted that this step, through standardized encapsulation, adapts the reduced-order model to the calling requirements of the one-dimensional thermal network simulation solver. The parameter system built matches the actual characteristics of the microchannel heat sink, achieving seamless integration between the reduced-order model and the one-dimensional simulation system. This avoids the problem of the model being unable to be called due to format or parameter mismatch, greatly improving the integration efficiency of the reduced-order model in system-level simulation. At the same time, the encapsulated model can be directly called, which also provides convenience for subsequent simulations of different topologies.

[0043] Preferably, a one-dimensional pipeline topology is constructed based on the actual pipeline network of the arrayed system, and a reduced-order element model is used as the core heat exchange element and connected to each branch. This includes: constructing a one-dimensional pipeline topology based on the actual connection relationship, flow resistance characteristics and array layout characteristics of the arrayed microchannel radiator system pipeline network. The topology includes various local resistance elements corresponding to the actual pipeline network. The reduced-order element model is connected to each branch of the one-dimensional pipeline topology according to the heat exchange branch distribution characteristics of the arrayed system.

[0044] Specifically, the actual piping network of the arrayed microchannel radiator system is first analyzed to clarify the actual connection relationships of each component, the resistance characteristics of the flow channels, and the array layout characteristics of the entire system. Based on this, a corresponding one-dimensional piping topology is built in a one-dimensional thermal network simulation environment. Various local resistance elements corresponding one-to-one with the actual piping network are set in the topology to restore the flow resistance characteristics of the actual piping network. Then, according to the distribution location and layout characteristics of each heat exchange branch in the arrayed system, the encapsulated reduced-order element models are respectively connected to each heat exchange branch of the one-dimensional piping topology to make the topology structure consistent with the heat exchange structure of the actual arrayed system.

[0045] It should be noted that this step constructs a one-dimensional pipeline topology based on the actual characteristics of the arrayed system network, restoring the structure and resistance characteristics of the actual system. At the same time, it connects the reduced-order component model according to the actual layout, so that the simulation model is highly consistent with the actual system, ensuring the authenticity and accuracy of the system-level thermal-fluid coupling simulation. Moreover, the construction of the topology structure fits the actual layout, which also facilitates the rapid reconstruction and adjustment of the system topology structure in the future.

[0046] Preferably, the conversion model is constructed to match the units and dimensions of the system variables and model parameters, including: constructing a conversion model based on the basic physical laws and dimensional conversion rules of thermal-fluid coupling simulation, so as to realize the dimensional conversion and physical quantity matching between the system thermal flux variables in the one-dimensional thermal network simulation environment and the characteristic parameters of the reduced-order component model.

[0047] Specifically, we first sort out the basic physical laws of heat-fluid coupling simulation, clarify the dimensional conversion rules of heat-fluid related physical quantities, and build a corresponding conversion model by combining the unit and dimensional characteristics of the system heat flux variables in the one-dimensional thermal network simulation environment, as well as the unit and dimensional characteristics of the input and output characteristic parameters of the reduced-order element model. Through this conversion model, we establish the conversion relationship between different units and dimensions. During the coupling simulation process, we automatically convert the units and dimensions of the heat flux variables of the one-dimensional thermal network simulation system so that the converted parameters match the characteristic parameters of the reduced-order element model. At the same time, we can also convert the output parameters of the reduced-order element model into a parameter form that the one-dimensional simulation system can recognize.

[0048] It should be noted that this step constructs a conversion model based on physical laws and dimensional rules, ensuring the scientific nature and accuracy of parameter conversion. It achieves seamless matching between one-dimensional simulation system variables and reduced-order component model parameters, avoiding simulation parameter transfer errors caused by inconsistent units and dimensions, ensuring the consistency of parameter transfer during thermal-fluid coupling simulation, and thus improving the accuracy of the entire system-level simulation.

[0049] Preferably, a simulation step size is set to perform steady-state thermal-fluid coupling simulation, and the pressure drop, temperature distribution and flow distribution results of each branch are output. This includes: achieving steady-state thermal-fluid coupling simulation by co-solving the reduced-order component model and the one-dimensional pipeline topology, completing the overall thermal-fluid field solution of the arrayed microchannel radiator system with a preset time step, and outputting results that can fully characterize the flow distribution state, temperature distribution law and fluid resistance characteristics of each branch of the system.

[0050] Specifically, after completing the construction of the one-dimensional pipeline topology, the access of the reduced-order component model, and the construction of the conversion model, the collaborative solution process of the reduced-order component model and the one-dimensional pipeline topology is initiated. The simulation calculation is gradually advanced according to the preset time step. During the simulation, the parameters are converted and transferred in real time through the conversion model, so that the heat transfer calculation of the reduced-order component model and the flow calculation of the one-dimensional pipeline topology cooperate with each other to complete the steady-state solution of the overall thermal flow field of the arrayed microchannel radiator system. After the simulation is completed, the solution results are sorted out and relevant data results that can reflect the flow distribution status, temperature distribution law, and fluid resistance characteristics of each branch of the system are output.

[0051] It should be noted that this step achieves steady-state simulation of the overall thermal flow field of the arrayed system through the joint solution of the reduced-order component model and the one-dimensional pipeline topology. The solution method with a preset time step ensures the stability of the simulation process. The output results can completely and comprehensively characterize the thermal flow characteristics of the arrayed system, providing accurate and reliable data for engineering operations such as the optimization of the distribution network and the adjustment of the structural design of the arrayed microchannel radiator system, thereby enhancing the engineering practical value of the simulation results.

[0052] Accordingly, this embodiment also provides a thermal-fluid coupling simulation system for arrayed microchannel heat sinks, including: The high-dimensional simulation dataset construction module is configured to perform three-dimensional computational fluid dynamics simulations of a single microchannel radiator under multiple operating conditions. Input samples are generated by Latin hypercube sampling and batch simulations are performed to extract flow field data and form a high-dimensional simulation dataset. The reduced-order model construction module is configured to construct a reduced-order model based on the high-dimensional simulation dataset by training the component flow characteristic parameter data through a convolutional neural network and performing dimensionality reduction processing. The model encapsulation and integration module is configured to export the trained reduced-order model as a file recognizable by a one-dimensional thermal network simulation solver, forming a callable reduced-order component model. The one-dimensional topology construction module is configured to construct a one-dimensional pipeline topology based on the actual pipeline network of the arrayed system, and to connect the reduced-order element model as the core heat exchange element to each branch. The data conversion and coupling simulation module is configured to build a conversion model, perform unit and dimension matching between system variables and model parameters, set the simulation step size to perform steady-state heat-fluid coupling simulation, and output the pressure drop, temperature distribution and flow distribution results of each branch.

[0053] Specifically, the various modules of the simulation system work collaboratively. First, the high-dimensional simulation dataset construction module completes the multi-condition three-dimensional computational fluid dynamics simulation of a single microchannel heat sink and generates the high-dimensional simulation dataset, and then transmits the dataset to the reduced-order model construction module. The reduced-order model construction module uses this dataset as a basis to complete the training of the convolutional neural network and the construction of the reduced-order model, and then transmits the constructed reduced-order model to the model encapsulation and integration module. The model encapsulation and integration module standardizes and encapsulates the reduced-order model to form a callable reduced-order component model, and then transmits it to the one-dimensional topology construction module. The one-dimensional topology construction module completes the construction of the one-dimensional pipeline topology and connects the reduced-order component model to each branch of the topology, and then transmits the topology model to the data conversion and coupling simulation module. Finally, the data conversion and coupling simulation module builds the conversion model, completes parameter matching, starts the steady-state thermal-fluid coupling simulation, and outputs the final simulation results.

[0054] It should be noted that this simulation system breaks down each stage of thermal-fluid coupling simulation into independent functional modules. The functions of each module are clearly defined and smoothly connected, enabling automated progress of the simulation process and significantly improving the efficiency of thermal-fluid coupling simulation of arrayed microchannel heat sinks. At the same time, the modular design also facilitates functional upgrades and adjustments to individual modules, improving the maintainability and scalability of the system and enabling it to adapt to the simulation needs of different types of arrayed microchannel heat sink systems.

[0055] Example 2 This embodiment provides a thermal-fluid coupling simulation method for arrayed microchannel heat sinks. Based on sample data generated from thermal simulation and testing, it utilizes the nonlinear fitting capability of neural networks to reduce the complex three-dimensional shunt network to an accurate one-dimensional network model, greatly improving the simulation modeling speed of complex microsystems. Specifically, the method of this embodiment includes the following steps.

[0056] Step 1: Multi-condition CFD simulation and data generation For a single microchannel radiator structure, three-dimensional CFD simulations were performed under different heat dissipation, coolant flow rate and inlet temperature conditions. Then, the Latin hypercube sampling (LHS) method was used to generate multiple sets of input samples within the preset parameter range, and the simulation was performed in batches to extract the flow field temperature, velocity and pressure distribution data to form a high-dimensional simulation dataset, which provides the original dataset support for step two.

[0057] Specifically, the model CFD data extraction covers 90-degree T-shaped tee components, non-smooth pipe components, gradual change components, heat source components, sudden change components, elbow components, pressure source components, mass source components, microchannel heat exchange components, heat conduction components, and local loss components.

[0058] Step 2: Building a Reduced-Order Model Using a data-driven approach, Convolutional Neural Networks (CNNs) are employed to train the model on high-dimensional data of component flow characteristic parameters obtained from CFD simulations, enabling the prediction and calculation of results. After model training, for the problem of predicting component flow characteristics, the outlet velocity, Reynolds number, and pressure loss coefficient can be output by inputting geometric features and calculation conditions.

[0059] In CNNs, convolutional layers are used to perform convolution operations, which are integral transformations that transform the local convolutional features of each pixel in an image to obtain the corresponding convolutional feature response output value. This process can be represented as:

[0060] in, I(x, y) Input for a two-dimensional image ( input ), w(s,t) For the kernel function of the corresponding size ( kernel function The output is a feature map (), feature map Taking three convolution operations as an example, as shown in the attached diagram. Figure 2 The output two-dimensional tensor after convolution calculation is shown to be 3×3, which corresponds to the spatial position meaning of its convolution calculation result.

[0061] Overfitting can easily occur between multiple consecutive convolutional layers. Therefore, pooling layers are introduced to reduce the number of parameters while maximizing the invariance of data characteristics, effectively achieving dimensionality reduction. This invention uses a combination of average pooling and max pooling to decompose the original data into multiple sparse subarrays in a 2×2 matrix manner, and then uses pooling methods to reduce the dimensionality of the data.

[0062] 1) Average pooling:

[0063] 2) Max pooling:

[0064] The data after dimensionality reduction in the pooling layer is weighted and integrated to form a fully connected layer. Each node in the fully connected layer is connected to all nodes in the previous layer. Each output is represented as the previous layer's node multiplied by a weight coefficient W and a bias value b, as shown in the following formula:

[0065] Among them, a1, a2, and a3 are the outputs.

[0066] Step 3: Model encapsulation and platform integration The trained neural network model is exported as a model file that can be recognized by a one-dimensional thermal network simulation solver, forming a callable reduced-order component model of a microchannel radiator. Specifically, its input parameters include pipe diameter, roughness, wall temperature, Reynolds number correction coefficient, etc., and its output parameters include compressibility loss coefficient, pressure drop, etc.

[0067] Step 4: System-level one-dimensional topology construction Based on the actual pipeline network structure of the arrayed system, a one-dimensional pipeline topology including local resistance elements such as tees, elbows, and tapered / expanding pipes is constructed in a one-dimensional thermal network simulation environment; the encapsulated microchannel radiator reduced-order model is used as the core heat exchange element and connected to each branch.

[0068] Step 5: Data Conversion and Coupled Simulation Construct a left-conversion model, a right-conversion model, and a velocity / mass flow rate conversion model to achieve unit and dimension matching between one-dimensional system variables (pressure, mass flow rate, specific enthalpy) and the input and output of the reduced-order model (velocity, temperature, U coefficient); set the simulation step size (preferably 0.01s) to perform steady-state thermal-fluid coupling simulation and output the pressure drop, temperature distribution, and flow rate distribution results of each branch.

[0069] Example 3 This embodiment is based on embodiment 2: This embodiment provides a thermal-fluid coupling simulation method for arrayed microchannel heat sinks, used for modeling and simulating the current distribution network of a four-channel microchannel heat sink, including the following steps: Step 1: Component Parametric CFD Simulation Parametric simulation calculates multi-dimensional response surfaces for parameters such as Reynolds number and loss coefficient, Reynolds number and outlet velocity, Reynolds number and outlet flow rate, and Reynolds number and outlet temperature, based on the parametric range of different structural parameters and calculation conditions. The selections are shown in Table 1 and Appendix. Figure 3 A parametric model of the component structure shown is established, and the model parameters are as follows: 1) Height: 1~5 (1, 2, 3, 4, 5); 2) Cylinder height: 0.8~2.8 (0.8, 1.3, 1.8, 2.3, 2.8); 3) Width: 3~7 (3, 4, 5, 6, 7); 4) Cylinder diameter: 1.8~3.8 (1.8, 2.3, 2.8, 3.3, 3.8); Table 1 - Component Parameterization List

[0070] The boundary conditions are set as follows: 1) Operating pressure: 101325.0 Pa; 2) Pressure differential: 20, 50.0, 500.0, 5000.0, 50000.0 Pa; 3) Temperature: 30℃ Step 2: Train the neural network and compare the results with CFD simulations. As shown in Tables 2-4, by comparing the predicted results of outlet flow velocity, Reynolds number, and pressure loss coefficient under different operating conditions with the actual CFD simulation results, it was found that the maximum error value of the data is less than 4.5%, and the average error value is about 2.4%, which can effectively prove the reliability of the prediction data model of the present invention.

[0071] Table 2 - Comparison of Outlet Velocity Prediction Results and CFD Calculation Results under Different Operating Conditions

[0072] Table 3 - Comparison of Reynolds number prediction results and CFD calculation results

[0073] Table 4 - Comparison of Predicted Pressure Loss Coefficient Results and CFD Calculation Results

[0074] Step 3: Integrate components in simulation software The integration process of parameterized components within the system consists of the following four steps: 1) Fill in the custom component information on the interface; 2) Generate and edit custom component code templates; 3) Create custom component test cases; 4) Set the case parameters and perform the calculation.

[0075] After integration, it will be displayed in the component library of the tool software as shown in the attached image. Figure 4 As shown.

[0076] Step 4: Construction of a one-dimensional fluid network After integrating all component types into the simulation software, drag and drop component icons into the view area, and add connections between different components to build a one-dimensional fluid network for the heterogeneous integrated model. A screenshot of the final build is attached. Figure 5 As shown, its one-dimensional fluid model is as follows: Figure 6 As shown.

[0077] Step 5: Simulation and Solution of One-Dimensional Fluid Network Run the simulation project, obtain the simulation results, and compare the one-dimensional fluid simulation results with the three-dimensional simulation results (see attached figure). Figure 7 and attached Figure 8 As shown, the maximum error in this case is within 3.0%.

[0078] Example 4 This embodiment is based on embodiment 1: This embodiment provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the thermal flux coupling simulation method for arrayed microchannel heat sinks described in Embodiment 1. The computer program can be in the form of source code, object code, executable file, or some intermediate form.

[0079] Specifically, the computer program for simulating the thermal-fluid coupling of arrayed microchannel heat sinks is stored in the memory of the computer device. When the simulation operation needs to be performed, the processor retrieves the computer program from the memory and starts execution. According to the instructions in the program, the generation of high-dimensional simulation dataset, the construction of reduced-order model, the encapsulation of reduced-order component model, the construction of one-dimensional pipeline topology, the construction of transformation model, and the execution of steady-state thermal-fluid coupling simulation are completed in sequence. Finally, the simulation results are output through the output terminal of the computer device. Throughout the process, the processor completes the program calculation and control, and the memory provides storage space for the program and simulation data.

[0080] It should be noted that by solidifying the thermal flux coupling simulation method of arrayed microchannel heat sinks into a computer program and implementing it with the help of computer equipment, the simulation method can achieve digital and automated operation with the help of hardware equipment, eliminating the cumbersomeness of manual operation and improving the convenience and operability of simulation operation. At the same time, the computing power of the computer equipment can ensure the efficient progress of the simulation process, and computer equipment with different configurations can be adapted to the simulation needs of arrayed systems of different sizes, thus improving the scenario adaptability of the simulation method.

[0081] Example 5 This embodiment is based on embodiment 1: This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the thermal-fluid coupling simulation method for arrayed microchannel heat sinks described in Embodiment 1. The computer program can be in the form of source code, object code, executable file, or some intermediate form. The storage medium includes any entity or device capable of carrying computer program code, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0082] Specifically, the completed computer program that can realize the thermal-fluid coupling simulation method of arrayed microchannel heat sinks is stored in a computer-readable storage medium in a computer-recognizable coded form. This storage medium can be connected to various computer devices for data transfer. When the simulation method needs to be executed, the processor of the computer device reads and loads the computer program from the storage medium, and then executes the various instructions in the program to complete the entire thermal-fluid coupling simulation process.

[0083] It should be noted that this method realizes the program solidification of the thermal flux coupling simulation method for arrayed microchannel heat sinks. With the help of computer-readable storage media, the simulation program can be ported and stored across devices, which facilitates the promotion and application of this simulation method in different engineering design scenarios. At the same time, the storage characteristics of the storage media also facilitate the saving and backup of the simulation program, reducing the risk of program loss. Furthermore, it eliminates the need to write separate programs for each computer device, which greatly reduces the engineering cost of using this simulation method.

[0084] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

[0085] It should be noted that, for the sake of simplicity, the foregoing method embodiments are described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

Claims

1. A simulation method for thermal-fluid coupling of arrayed microchannel heat sinks, characterized in that, include: Three-dimensional computational fluid dynamics simulation of a single microchannel heat sink under multiple operating conditions was performed. Input samples were generated by Latin hypercube sampling and batch simulation was performed. Flow field data was extracted to form a high-dimensional simulation dataset. Based on the high-dimensional simulation dataset, a reduced-order model is constructed by training the component flow characteristic parameter data through a convolutional neural network and then performing dimensionality reduction processing. The trained reduced-order model is exported as a file that can be recognized by a one-dimensional thermal network simulation solver, forming a callable reduced-order component model; A one-dimensional pipeline topology is constructed based on the actual pipeline network of the arrayed system, and the reduced-order element model is used as the core heat exchange element and connected to each branch. Construct a conversion model to match the units and dimensions of the system variables and model parameters, set the simulation step size to perform steady-state thermal-fluid coupling simulation, and output the pressure drop, temperature distribution and flow distribution results of each branch.

2. The thermal flux coupling simulation method for arrayed microchannel heat sinks according to claim 1, characterized in that, The three-dimensional computational fluid dynamics simulation of a single microchannel heat sink under multiple operating conditions includes: combining various local resistance elements, heat exchange elements, power elements and heat conduction elements in the arrayed microchannel heat sink system to perform three-dimensional computational fluid dynamics simulation of a single microchannel heat sink under different operating conditions.

3. The thermal flux coupling simulation method for arrayed microchannel heat sinks according to claim 1, characterized in that, The step of training the flow characteristic parameter data of the component through a convolutional neural network and performing dimensionality reduction includes: firstly, extracting convolutional features from the high-dimensional data, then using combined pooling to reduce the data dimensionality, and then weighting and integrating the dimensionality-reduced data to form a fully connected layer, thereby fitting the nonlinear mapping relationship between the component's geometric features, operating conditions, and flow characteristic parameters, and completing the construction of the reduced-dimensional model.

4. The thermal flux coupling simulation method for arrayed microchannel heat sinks according to claim 1, characterized in that, The step of exporting the trained reduced-order model into a file recognizable by a one-dimensional thermal network simulation solver to form a callable reduced-order component model includes: standardizing and encapsulating the trained reduced-order model to obtain a reduced-order component model. The encapsulated reduced-order component model has an input-output parameter system adapted to the one-dimensional thermal network simulation solver. The input-output parameter system matches the structural features and flow heat transfer characteristics of the microchannel radiator and can be directly called by the one-dimensional thermal network simulation solver.

5. The thermal flux coupling simulation method for arrayed microchannel heat sinks according to claim 1, characterized in that, The construction of a one-dimensional pipeline topology based on the actual pipeline network of the arrayed system, with the reduced-order element model as the core heat exchange element connected to each branch, includes: constructing a one-dimensional pipeline topology based on the actual connection relationship, flow resistance characteristics and array layout characteristics of the arrayed microchannel radiator system pipeline network. The topology includes various local resistance elements corresponding to the actual pipeline network. The reduced-order element model is connected to each branch of the one-dimensional pipeline topology according to the heat exchange branch distribution characteristics of the arrayed system.

6. The thermal flux coupling simulation method for arrayed microchannel heat sinks according to claim 1, characterized in that, The construction of the conversion model performs unit and dimension matching between system variables and model parameters, including: constructing a conversion model based on the basic physical laws and dimensional conversion rules of thermal-fluid coupling simulation, so as to realize the dimensional conversion and physical quantity matching between system thermal flux variables in a one-dimensional thermal network simulation environment and characteristic parameters of the reduced-order component model.

7. The thermal flux coupling simulation method for arrayed microchannel heat sinks according to claim 1, characterized in that, The simulation step size is set to perform steady-state thermal-fluid coupling simulation and output the pressure drop, temperature distribution and flow distribution results of each branch. This includes: achieving steady-state thermal-fluid coupling simulation by co-solving the reduced-order component model and the one-dimensional pipeline topology, completing the overall thermal flow field solution of the arrayed microchannel radiator system with a preset time step, and outputting results that can fully characterize the flow distribution state, temperature distribution law and fluid resistance characteristics of each branch of the system.

8. A thermal flux coupling simulation system for arrayed microchannel heat sinks, characterized in that, include: The high-dimensional simulation dataset construction module is configured to perform three-dimensional computational fluid dynamics simulations of a single microchannel radiator under multiple operating conditions. Input samples are generated by Latin hypercube sampling and batch simulations are performed to extract flow field data and form a high-dimensional simulation dataset. The reduced-order model construction module is configured to construct a reduced-order model based on the high-dimensional simulation dataset by training the component flow characteristic parameter data through a convolutional neural network and performing dimensionality reduction processing. The model encapsulation and integration module is configured to export the trained reduced-order model as a file recognizable by a one-dimensional thermal network simulation solver, forming a callable reduced-order component model. The one-dimensional topology construction module is configured to construct a one-dimensional pipeline topology based on the actual pipeline network of the arrayed system, and to connect the reduced-order element model as the core heat exchange element to each branch. The data conversion and coupling simulation module is configured to build a conversion model, perform unit and dimension matching between system variables and model parameters, set the simulation step size to perform steady-state heat-fluid coupling simulation, and output the pressure drop, temperature distribution and flow distribution results of each branch.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the thermal flux coupling simulation method for arrayed microchannel heat sinks as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the thermal flux coupling simulation method for arrayed microchannel heat sinks as described in any one of claims 1-7.

Images