A method, system, device and medium for calculating PCB temperature field in radiator

Abstract

The present invention discloses a method, system, device and medium for calculating the temperature field of a PCB in a radiator, wherein the method comprises: obtaining training data, the training data comprising the distribution of fins of the radiator, the layout of the chip and the distribution of the temperature field; preprocessing the training data, and combining the corresponding temperature field distribution as a label to form matching data; training to obtain a strong mapping relationship between the distribution of fins of the radiator and the layout of the chip based on a deep convolution generative adversarial neural network, and constructing an agent model for quickly calculating the steady-state temperature field of the PCB under this heat dissipation condition according to the distribution of fins of the radiator and the layout of the chip; inputting the input data consisting of the distribution of fins of the radiator and the layout of the chip into the trained agent model, and quickly calculating the steady-state PCB temperature cloud map. The present invention obtains an agent model by collecting history and training, and is used to help thermal design engineers improve design efficiency while ensuring prediction accuracy; it can be widely used in the field of numerical heat transfer calculation technology.

Description

Technical Field

[0001] The present invention relates to the technical field of numerical heat transfer calculation, and in particular to a method, system, device and medium for calculating the temperature field of PCB in a radiator. Background Art

[0002] As the degree of autonomous driving in cars increases, the number of electronic control units in cars increases rapidly, which not only challenges the integrated circuit integration capability, but also places higher demands on the heat dissipation capability of electronic devices such as controllers. Generally speaking, the temperature field calculation of electronic devices is a complex process of multi-physical field coupling. When performing finite element simulation, it is necessary to divide a large number of grids for entities such as fins and chips, and to set boundary conditions and iterative conditions in combination with fluid mechanics and heat transfer content for solution.

[0003] However, solving the above-mentioned thermal-fluid coupling problem requires high computing resources, takes a long time, and has high requirements on users. Therefore, how to simplify the numerical simulation process and reduce the design cost of thermal simulation solutions will be an urgent problem to be solved in the new generation of design and research. Summary of the invention

[0004] In order to solve at least some extent one of the technical problems existing in the prior art, the object of the present invention is to provide a method, system, device and medium for calculating the temperature field of the PCB in the radiator.

[0005] The technical solution adopted by the present invention is:

[0006] A method for calculating the PCB temperature field in a radiator, including the following steps:

[0007] Acquire training data, which includes radiator fin distribution, chip layout, and temperature field distribution;

[0008] The training data is preprocessed and the corresponding temperature field distribution is combined as the label to form matching data;

[0009] Based on deep convolutional generative adversarial neural networks, a strong mapping relationship between the fin distribution of the heat sink and the chip layout is trained, and a proxy model is constructed to quickly calculate the steady-state temperature field of the PCB under this heat dissipation condition based on the fin distribution of the heat sink and the chip layout.

[0010] The input data composed of radiator fin distribution and chip layout is input into the training proxy model to quickly calculate the steady-state PCB temperature cloud diagram.

[0011] Furthermore, the proxy model is trained by a generative adversarial network method, and the training process consists of two different models, a generator and a discriminator. The generator is used to convert the input radiator fin distribution and chip distribution into a steady-state temperature field cloud map of the PCB, and the discriminator is used to distinguish the temperature field cloud map generated by the generator from the actual label cloud map. The final training balance point is when the temperature cloud map generated by the generator is consistent with or very similar to the label corresponding to the data set, and the probability that the discriminator can distinguish the cloud map generated by the generator is 50%.

[0012] Further, the input data of the proxy model is obtained by:

[0013] Step 1: Determine the fin distribution and fin height, chip distribution and chip heating power, bump distribution and bump height connecting the chip and radiator;

[0014] Step 2: Create a three-channel image with the same proportion as the heat sink being studied, and linearly map the fin height, chip heating power, and bump height to the image channel values ​​to obtain model training data;

[0015] Step 3: Create a corresponding thermal model in ANSYS software according to the fin height, chip heating power, and bump height and perform numerical simulation to obtain the PCB temperature field distribution as training labels, and pair the training labels with the model training data in step 2 one by one.

[0016] Further, the calculation method further comprises a fine-tuning step:

[0017] Migrate the data training weights of historical application scenarios to the current specific application scenario weights.

[0018] Another technical solution adopted by the present invention is:

[0019] A computing system for PCB temperature field in a radiator, including:

[0020] A data acquisition module is used to acquire training data, which includes radiator fin distribution, chip layout, and temperature field distribution;

[0021] A tag generation module is used to preprocess the training data and combine the corresponding temperature field distribution as the tag to form matching data;

[0022] The data training module is used to generate adversarial neural networks based on deep convolution, train to obtain a strong mapping relationship between the radiator fin distribution and chip layout, and build an agent model that can quickly calculate the steady-state temperature field of the PCB under this heat dissipation condition based on the radiator fin distribution and chip layout;

[0023] The temperature calculation module is used to input the input data composed of the radiator fin distribution and chip layout to train the proxy model to quickly calculate the steady-state PCB temperature cloud diagram.

[0024] Furthermore, the proxy model is trained by a generative adversarial network method, and the training process consists of two different models, a generator and a discriminator. The generator is used to convert the input radiator fin distribution and chip distribution into a steady-state temperature field cloud map of the PCB, and the discriminator is used to distinguish the temperature field cloud map generated by the generator from the actual label cloud map. The final training balance point is when the temperature cloud map generated by the generator is consistent with or very similar to the label corresponding to the data set, and the probability that the discriminator can distinguish the cloud map generated by the generator is 50%.

[0025] Further, the input data of the proxy model is obtained by:

[0026] Determine the fin distribution and fin height, chip distribution and chip heating power, bump distribution and bump height connecting the chip and radiator;

[0027] Create a three-channel image with the same ratio as the studied radiator, and linearly map the fin height, chip heating power, and bump height with the image channel values ​​respectively to obtain the model training data;

[0028] According to the fin height, chip heating power, and bump height, the corresponding thermal model is created in the ANSYS software and numerical simulation is performed to obtain the PCB temperature field distribution as training labels, and the training labels are paired with the model training data one by one.

[0029] Further, the computing system further includes a fine-tuning module for migrating data training weights of historical application scenarios to a current specific application scenario weight.

[0030] Another technical solution adopted by the present invention is:

[0031] A calculating device for PCB temperature field in a radiator, including:

[0032] at least one processor;

[0033] at least one memory for storing at least one program;

[0034] When the at least one program is executed by the at least one processor, the at least one processor implements the above method.

[0035] Another technical solution adopted by the present invention is:

[0036] A computer-readable storage medium stores a program executable by a processor, wherein the program executable by the processor is used to execute the method described above when executed by the processor.

[0037] The beneficial effect of the present invention is as follows: the present invention collects historical relevant thermal design scenario structure data and corresponding PCB temperature field distribution data as training data for generating an adversarial network, and obtains a proxy model that can quickly calculate the PCB temperature field corresponding to the corresponding scenario structure data. The proxy model can help thermal design engineers improve design efficiency while ensuring prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the embodiments of the present invention or the drawings of related technical solutions in the prior art are introduced below. It should be understood that the drawings introduced below are only for the convenience of clearly describing some embodiments of the technical solutions of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without paying any creative work.

[0039] Figure 1 It is a schematic diagram of the PCB temperature field rapid calculation system in the radiator in the embodiment of the present invention;

[0040] Figure 2 It is a schematic diagram of the thermal simulation model used to generate the tag in an embodiment of the present invention;

[0041] Figure 3 It is the initial tag diagram obtained in the step of generating a tag in an embodiment of the present invention;

[0042] Figure 4 It is a schematic diagram of several pairs of matching data obtained in an embodiment of the present invention;

[0043] Figure 5 It is a schematic diagram of the generative adversarial network model structure and training process selected in the embodiment of the present invention;

[0044] Figure 6 It is a comparison diagram of model training results and simulation results in an embodiment of the present invention;

[0045] Figure 7 It is a flowchart of the step of calculating the PCB temperature field in a radiator in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and are not to be construed as limitations of the present invention. For the step numbers in the following embodiments, they are only provided for the convenience of explanation, and the order between the steps is not limited in any way, and the execution order of each step in the embodiment can be adaptively adjusted according to the understanding of those skilled in the art.

[0047] In the description of the present invention, it should be understood that descriptions involving orientations, such as up, down, front, back, left, right, etc., and orientations or positional relationships indicated are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present invention.

[0048] In the description of the present invention, "several" means one or more, "more" means more than two, "greater than", "less than", "exceed" etc. are understood as not including the number itself, and "above", "below", "within" etc. are understood as including the number itself. If there is a description of "first" or "second", it is only used for the purpose of distinguishing the technical features, and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0049] In the description of the present invention, unless otherwise clearly defined, terms such as setting, installing, connecting, etc. should be understood in a broad sense, and technicians in the relevant technical field can reasonably determine the specific meanings of the above terms in the present invention based on the specific content of the technical solution.

[0050] This embodiment provides a fast computing system for PCB temperature field in a radiator, the system includes a model preparation module and a front-end usage module, and the system architecture diagram is as follows: Figure 1 Shown. The model preparation module includes steps such as data sampling, generating labels, data training, and storing weights.

[0051] The data sampling steps include model training data preparation and thermal model structural parameter preparation. First, establish a thermal model for common passive heat dissipation scenarios, such as Figure 2As shown. The upper cover 100 includes fins 110 and bumps 120 that contact the chip. A plurality of chipsets 140 are distributed on the PCB module 130. The lower cover 150 is used to form a closed link in which the PCB module 130 is located. It can be understood that in addition to the above modules, other modules consistent with the actual heat dissipation conditions are also included, such as thermal grease, etc. Then, several parameters within a specified range are generated to perform random sampling within a certain range, including fin type, fin thickness, number of fins, fin height, substrate height, bump position parameters, bump height, chip position parameters, and chip power consumption. The thermal model is parameterized according to the random parameter values, and ICEPAK is used to calculate the steady-state temperature field, and the temperature cloud map distribution on the PCB under steady-state conditions is derived. In order to facilitate the subsequent model training, loading and processing, each parameter value is represented by an image with the same proportion as the PCB, where the fin type, fin thickness and other fin position distribution information are represented by different pixel values ​​R of the red channel of the image, and the specific pixel value is obtained by linear mapping the fin height h1 and the substrate height h2 using the following formula (1); the bump position information is represented by the green channel pixel value G of the image, and the specific pixel value is obtained by linear mapping the bump height h3 using the following formula (2); the chip position information is represented by the blue channel pixel value B of the image, and the specific pixel value is obtained by linear mapping the chip power value P and the size r1, r2 using the following formula (3). Note that h1, h2, h3, P, r1, and r2 are not constants, but depend on the parameters of the image corresponding to the heat sink position.

[0052] Formula (1): in A set of points representing the corresponding positions of fins in the image, It represents a set of points in the image where there are no fins and only the corresponding positions of the boss. h1 and h2 are in mm.

[0053] Formula (2): in A set of points representing the corresponding positions of bumps in the image, h3 unit is mm.

[0054] Formula (3): in The set of points representing the corresponding positions of the chip in the image, P units are W, r1, and r2 units are mm.

[0055] The step of generating labels uses the ICEPAK module in the ANSYS workbench platform to establish a model method for parameterized simulation. This embodiment uses the CFD solver and the Bussineske approximation model inside ANSYS to simulate the temperature field of the radiator model under natural cooling, while taking into account the radiation heat transfer phenomenon in natural convection. Taking into account both calculation speed and accuracy, the Discrete Ordinates (DO) radiation model is adopted. It is assumed that the fluid state in the calculation domain is turbulent. Because it is a simulation of the radiator of electronic products, a zero-equation turbulence model with a high cost performance is adopted. In order to ensure the calculation accuracy, the calculation domain is set to 1.5 times the model characteristic length L except for the Ymax direction at the top of the radiator, and the Ymax direction is set to 2 times the model characteristic length L, and the regional attributes of the six directions are all Opening, and the ambient temperature in the calculation domain is set to 85°C. The chip is simulated using two copper blocks (blocks) with no thickness heat source (source) in the middle. The shape parameters and power parameters of the chip are parameterized. The shell is built using the enclosure cavity module, and the six boundary surfaces are set to 2mm thick thin shell features (the thickness actually participates in the calculation of the temperature field, but does not participate in the meshing of the model). The entire shell is made of aluminum alloy. Because the chip on the PCB is at a certain distance from the upper end of the cavity shell, a corresponding number of bumps (blcok) are established to link the cavity shell with the chip, which is conducive to effective heat dissipation. A plate structure with a thickness of 1mm is set at the contact position between the bump and the chip to simulate the thermal conductive silicone in the actual object. The thermal conductivity of the thermal conductive silicone is set to 5W / m·k. After completing the basic modeling of the electronic device, an aluminum profile extruded heat sink is added to the upper end of the cavity shell. The heat sink is made of aluminum alloy with a thermal conductivity of 240W / m·k. Parameters are set for the overall height of the heat sink, the height of the substrate, the number of fins, and the thickness of the fins. The temperature field of various heat sinks can be simulated by changing the parameter values.

[0056] After the establishment of the overall model, the model needs to be meshed. Because the shapes contained in the model are all regular, the Hexa Cartesian structured grid is used, and all grids are vertically orthogonal. After the mesh is divided, the mesh quality is checked and judged according to the values ​​of the face alignment, the volume of the grid, and the skewness of the grid. The values ​​of the three are close to the expected values, so for this model, the mesh quality is good and can be conformal. In addition to good meshing quality, good meshing accuracy is also required. The number of grids directly affects the speed and accuracy of solving the final result. For the overall model, the maximum mesh size in the X, Y, and Z directions is limited to 10mm, 5mm, and 15mm. This is based on 1 / 20 to 1 / 40 of the characteristic size of the computational domain. However, because some sizes of the chip model and the fin model are much smaller than the characteristic size of the computational domain, independent mesh encryption is required to achieve the required calculation accuracy. In order to test the independence of the grid, the number of grids in the chip and fin models is increased in stages to test the changes in the model temperature field under different numbers of grids. According to the results of six different grid encryption models, when the number of grids is 1161576, the model can ensure good calculation accuracy and calculate the temperature field of the model in a shorter calculation time, so this grid division method is finally selected.

[0057] Because it is a finite element analysis simulation, there are requirements for the precise value of the result, so it is necessary to set the corresponding number of iterations and the residual standard of the corresponding parameters. The standard for coupling residuals is set - the flow residual is 1e-3, and the energy residual value is 1e-7. As long as the three residual standards are met, the calculation result is considered to meet the accuracy requirements. In the early stage, a small number of parameter calculations were done, and it was found that most model calculations converged within 250 steps, so the number of iterations was set to 250.

[0058] The original data obtained using ICEPAK is as follows Figure 3 As shown, the PCB area is cropped by digital image technology and saved for use as a label.

[0059] In the training data, multiple fin types are selected for data acquisition, and several matching data are selected, such as Figure 4 shown. Figure 4 Six pairs of matching data are displayed. The left half is the combined image calculated according to the above formula, and the right half is obtained by the cloud image results obtained by the simulation software.

[0060] The data training part will put paired data into a data set like Figure 5The training process starts in the conditional generative adversarial network shown in . The model used is a Spatially-Adaptive Pixelwise Networks (ASAP-Net). There are two training modes: no pre-training data and scene data fine-tuning. No training data means that there is no relevant historical data for pre-training, and only the current scene data is trained. When training without pre-training data, 16 to 32 paired data are selected to train the model. The weights are obtained and stored after the retention method is used to test the prediction quality of the model. Scene data fine-tuning means selecting part of the historical data for training when there is a similar historical database. The number of historical data is not limited here. Generally, the more data, the more accurate it is. After obtaining the pre-training weights, a small learning rate is used for fine-tuning with a small amount of scene data to obtain accurate weights for storage. The results obtained after training without pre-training data are shown in the following example. Figure 6 It is shown, and the input of 4 randomly extracted data from the test set, model calculation results and corresponding labels are displayed.

[0061] The front-end usage module includes deployment weights and display parts. By deploying the above-mentioned stored weights on cloud resources, the fast calculation function can be provided in the WeChat applet through services such as WeChat cloud hosting. Users can simply design the fin structure and chip position in the WeChat applet interface to achieve fast calculation; similarly, given the lightweight model, the model can be placed on a personal computer for offline use, without consuming a lot of computing power to complete the fast calculation process.

[0062] This embodiment collects historical relevant thermal design scenario structure data and corresponding PCB temperature field distribution data as training data for generating adversarial networks, and obtains a proxy model that can quickly calculate the PCB temperature field corresponding to the corresponding scenario structure data. The proxy model can help thermal design engineers improve design efficiency while ensuring prediction accuracy. In addition, the model weight migration of historical data can be realized based on a small amount of new data to adapt to the thermal design accuracy requirements of the corresponding scenario. And it can help engineers quickly estimate the PCB temperature field when it is necessary to quickly evaluate the peripheral selection of the heat dissipation solution. Compared with the complete CFD numerical simulation process in the traditional design process, it greatly reduces the calculation cost, reduces the calculation time and the user's learning cost, and saves manpower and material resources.

[0063] like Figure 7 As shown, this embodiment also provides a method for calculating the PCB temperature field in the radiator, including the following steps:

[0064] S1. Obtain training data, which includes radiator fin distribution, chip layout, and temperature field distribution;

[0065] S2. Preprocess the training data and combine the corresponding temperature field distribution as the label to form matching data;

[0066] S3. Based on deep convolution generative adversarial neural network, a strong mapping relationship between the fin distribution of the heat sink and the chip layout is obtained through training, and a proxy model is constructed to quickly calculate the steady-state temperature field of the PCB under this heat dissipation condition based on the fin distribution of the heat sink and the chip layout;

[0067] S4. Enter the input data composed of radiator fin distribution and chip layout to quickly calculate the steady-state PCB temperature cloud diagram.

[0068] A method for calculating the temperature field of PCB in the radiator of this embodiment, and Figure 1 The calculation system of the PCB temperature field in a radiator shown has corresponding relationships, and therefore has the functions and beneficial effects of the system.

[0069] This embodiment also provides a computing device for the PCB temperature field in a radiator, including:

[0070] at least one processor;

[0071] at least one memory for storing at least one program;

[0072] When the at least one program is executed by the at least one processor, the at least one processor implements the following Figure 7 The method shown.

[0073] A device for calculating the temperature field of a PCB in a radiator in this embodiment can execute a method for calculating the temperature field of a PCB in a radiator provided by an embodiment of the method of the present invention, can execute any combination of implementation steps of the method embodiment, and has the corresponding functions and beneficial effects of the method.

[0074] The present application also discloses a computer program product or a computer program, which includes a computer instruction stored in a computer-readable storage medium. A processor of a computer device can read the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes Figure 7 The method shown.

[0075] This embodiment also provides a storage medium, which stores instructions or programs that can execute a method for calculating the PCB temperature field in a radiator provided by an embodiment of the method of the present invention. When the instructions or program are run, any combination of implementation steps of the method embodiment can be executed, and the corresponding functions and beneficial effects of the method can be obtained.

[0076] In some selectable embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided by way of example, for the purpose of providing a more comprehensive understanding of technology. The disclosed method is not limited to the operation and logic flow presented herein. Selectable embodiments are expected, wherein the order of various operations is changed and the sub-operation of a part for which is described as a larger operation is performed independently.

[0077] In addition, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise specified, one or more of the functions and / or features described may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the present invention. More specifically, in view of the properties, functions, and internal relationships of the various functional modules in the device disclosed herein, the actual implementation of the module will be understood within the conventional skills of the engineer. Therefore, those skilled in the art can implement the present invention set forth in the claims without excessive experimentation using ordinary techniques. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0078] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0079] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0080] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0081] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0082] In the above description of this specification, the description with reference to the terms "one embodiment / example", "another embodiment / example" or "certain embodiments / examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0083] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

[0084] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the above embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A method for calculating the temperature field of a PCB in a radiator, characterized in that: The following steps are involved: Acquire training data, wherein the training data includes heat sink fin distribution, chip layout, and temperature field distribution; Preprocessing the training data and combining the corresponding temperature field distribution as labels to form matching data; Based on deep convolutional generative adversarial neural networks, a strong mapping relationship between the fin distribution of the heat sink and the chip layout is trained, and a proxy model is constructed to quickly calculate the steady-state temperature field of the PCB under this heat dissipation condition based on the fin distribution of the heat sink and the chip layout. Input data consisting of heat sink fin distribution and chip layout are fed into the trained proxy model to quickly calculate the steady-state PCB temperature cloud map; The proxy model is trained by a generative adversarial network. The training process consists of two different models, a generator and a discriminator. The generator is used to convert the input radiator fin distribution and chip distribution into a steady-state temperature field cloud map of the PCB. The discriminator is used to distinguish the temperature field cloud map generated by the generator from the actual label cloud map. The final training balance point is when the temperature field cloud map generated by the generator is consistent with the label corresponding to the data set, and the probability that the discriminator can distinguish the cloud map generated by the generator is 50%; The input data of the proxy model is obtained in the following way: Determine the fin distribution and fin height, chip distribution and chip heating power, and bump distribution and bump height connecting the chip and the heat sink; Create a three-channel image with the same proportion as the heat sink being studied, and linearly map the fin height, chip heating power, and bump height to the image channel values ​​to obtain model training data; According to the fin height, chip heating power, and bump height, the corresponding thermal model is created in the ANSYS software and numerical simulation is performed to obtain the PCB temperature field distribution as training labels, and the training labels are paired with the model training data one by one.

2. The method for calculating the temperature field of a PCB in a radiator according to claim 1, characterized in that: The calculation method also includes a fine-tuning step: Migrate the data training weights of historical application scenarios to the current specific application scenario weights.

3. A system for calculating the temperature field of a PCB in a radiator, characterized in that: include: A data acquisition module, used to obtain training data, wherein the training data includes heat sink fin distribution, chip layout and temperature field distribution; A label generation module, used for preprocessing the training data and combining the corresponding temperature field distribution as a label to form matching data; The data training module is used to generate adversarial neural networks based on deep convolution, train to obtain a strong mapping relationship between the radiator fin distribution and chip layout, and build an agent model that can quickly calculate the steady-state temperature field of the PCB under this heat dissipation condition based on the radiator fin distribution and chip layout; The temperature calculation module is used to input the input data consisting of the heat sink fin distribution and chip layout into the trained proxy model to quickly calculate the steady-state PCB temperature cloud map; The proxy model is trained by a generative adversarial network. The training process consists of two different models, a generator and a discriminator. The generator is used to convert the input radiator fin distribution and chip distribution into a steady-state temperature field cloud map of the PCB. The discriminator is used to distinguish the temperature field cloud map generated by the generator from the actual label cloud map. The final training balance point is when the temperature field cloud map generated by the generator is consistent with the label corresponding to the data set, and the probability that the discriminator can distinguish the cloud map generated by the generator is 50%; The input data of the proxy model is obtained in the following way: Determine the fin distribution and fin height, chip distribution and chip heating power, and bump distribution and bump height connecting the chip and the heat sink; Create a three-channel image with the same proportion as the heat sink being studied, and linearly map the fin height, chip heating power, and bump height to the image channel values ​​to obtain model training data; According to the fin height, chip heating power, and bump height, the corresponding thermal model is created in the ANSYS software and numerical simulation is performed to obtain the PCB temperature field distribution as training labels, and the training labels are paired with the model training data one by one.

4. A calculation system for PCB temperature field in a radiator according to claim 3, characterized in that: The computing system also includes a fine-tuning module, which is used to migrate data training weights of historical application scenarios to current specific application scenario weights.

5. A device for calculating the temperature field of a PCB in a radiator, characterized in that: include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method according to any one of claims 1 to 2.

6. A computer-readable storage medium storing a program executable by a processor, characterized in that: The processor-executable program is used to perform the method according to any one of claims 1 to 2 when executed by the processor.

Images