Thermal module temperature field construction and evaluation system and method based on multi-modal coupling

Abstract

This invention relates to a system and method for constructing and evaluating the temperature field of a heat dissipation module based on multimodal coupling. The method includes: extracting the module region of the heat dissipation module and performing angle correction on the module region; inputting the corrected infrared image into an RGB-temperature mapping model to obtain an initial infrared temperature distribution; using the measured temperature of the thermocouple as a benchmark, performing deviation calibration on the infrared temperature measurement results in each temperature measurement sub-region and allocating them to each temperature measurement point to reconstruct and generate the global temperature field of the heat dissipation module; based on the physical coordinates and heating power of each temperature measurement point, calling a temperature deviation compensation model to calculate the deviation compensation value of each temperature measurement point to correct the global temperature field point by point; performing reverse calculation on the corrected temperature field to obtain the chip-side temperature field distribution; constructing a temperature field quality evaluation model to quantitatively score the corrected temperature field and output the temperature field quality evaluation result. The system of this invention can continuously output a stable and high-precision temperature field under varying operating conditions.

Description

Technical Field

[0001] This invention relates to the field of thermal management and infrared temperature measurement technology for electronic devices, specifically to a system and method for constructing and evaluating the temperature field of a heat dissipation module based on multimodal coupling. Background Technology

[0002] As electronic devices rapidly evolve towards higher integration and higher power density, the heat transfer performance of heat dissipation modules directly determines the operational stability and lifespan of these devices. Temperature field distribution is a crucial reference indicator characterizing the heat transfer performance of heat dissipation modules, and accurate and comprehensive temperature field detection and evaluation are key aspects of optimizing the thermal design and verifying the performance of electronic devices.

[0003] Currently, the mainstream method for temperature measurement of heat dissipation modules is thermocouple contact temperature measurement. This method measures the temperature difference between the evaporation and condensation sections of the module by using temperature data at fixed measurement points. The corresponding thermal resistance is calculated based on the temperature difference under different heating power conditions, and the performance of the entire heat dissipation module is judged based on the thermal resistance value and its trend. Another temperature measurement method is infrared temperature measurement, which uses an infrared camera to achieve non-contact global temperature measurement of the test object.

[0004] For temperature field analysis of heat dissipation modules, the mainstream approach is currently finite element method (FEM) simulation. This method simulates the temperature field distribution of the heat dissipation module under specific conditions by artificially setting heat sources and boundary conditions and dividing the local mesh. The temperature data generated by this method usually needs to be compared with actual experimental data to ensure the accuracy of the simulation parameters. In addition, existing technologies have also developed meshless numerical simulation methods. Meshless methods are a type of numerical calculation that does not require generating a mesh but instead constructs interpolation functions and discrete control equations based on arbitrarily distributed coordinate points.

[0005] The existing technology has the following technical problems: (1) Thermocouples can only achieve single-point temperature measurement, and the spatial coverage is insufficient. They are difficult to capture the global temperature distribution under complex working conditions with multiple hot spots, and their reconfigurability under dynamic working conditions with multiple hot spots is poor.

[0006] (2) Infrared thermometry has problems such as limited spatial resolution, great influence of the surface emissivity of the object being measured, and limited temperature measurement accuracy, making it difficult to apply to occasions that require precise temperature measurement alone.

[0007] (3) The temperature field of the heat dissipation module is analyzed by finite element simulation, which has a long test cycle. At the same time, the reliability of the simulation parameters also depends on the comparison and calibration of the actual experiment, so the reference value is limited.

[0008] (4) The existing technology lacks a unified and standardized quantitative evaluation system for temperature field reconstruction results, which makes it impossible to objectively judge the accuracy and quality of the reconstructed temperature field, resulting in insufficient credibility and comparability of the test results, making it difficult to meet the high-precision, automated, and batch performance testing requirements of electronic device heat dissipation modules. Summary of the Invention

[0009] The purpose of this invention is to provide a system and method for constructing and evaluating the temperature field of a heat dissipation module based on multimodal coupling, so as to solve the problems in the existing heat dissipation module temperature field detection technology, such as insufficient spatial coverage and limited accuracy of a single temperature measurement mode, and poor adaptability of temperature field reconstruction under complex working conditions with multiple hot spots.

[0010] To achieve the above objectives, the present invention adopts the following technical solution: A method for constructing and evaluating the temperature field of a heat dissipation module based on multimodal coupling includes: S1: Acquire the infrared image of the heat dissipation module under test. Based on the difference in infrared emissivity between the heat dissipation module and the background environment in the infrared image, identify and extract the module area of ​​the heat dissipation module, perform angle correction on the module area, and output the corrected infrared image. S2: Input the RGB three-channel values ​​of the calibrated infrared image into the pre-trained RGB-temperature mapping model to obtain the initial infrared temperature distribution; arrange discrete temperature measurement points in the module area, and associate each temperature measurement point with the material thermal resistance model of the heat dissipation module; divide the module area into multiple temperature measurement sub-regions, obtain the measured temperature of the thermocouples in each temperature measurement sub-region, use the measured temperature of the thermocouples as the benchmark to perform deviation calibration on the infrared temperature measurement results in each temperature measurement sub-region, and distribute the calibrated deviation to each temperature measurement point; based on the calibrated temperature data of each temperature measurement point and the thermal resistance model, reconstruct and generate the global temperature field of the heat dissipation module. S3: Obtain the current heating power of the heat dissipation module. Based on the physical coordinates and heating power of each temperature measurement point, call the pre-stored temperature deviation compensation model, calculate the deviation compensation value of each temperature measurement point, use the deviation compensation value to correct the global temperature field point by point, and output the corrected temperature field. S4: Based on the thermal diffusion parameters of the interface thermally conductive material, the corrected temperature field is calculated in reverse to obtain the temperature field distribution on the chip side; a temperature field quality evaluation model is constructed to quantify and score the corrected temperature field and output the temperature field quality evaluation results.

[0011] Further, step S1 involves identifying and extracting the module region of the heat dissipation module, specifically including: The infrared image is converted into a grayscale image. A grayscale judgment threshold is set based on the difference in infrared emissivity between the metal surface of the heat dissipation module and the ambient background. The heat dissipation module is then separated from the background through threshold segmentation to obtain the module area.

[0012] Furthermore, step S1 involves angle correction of the module area, including: The quadrilateral contour of the copper plate in the heating area of ​​the module is identified, and the detected quadrilateral contour is compared with the standard rectangular contour. The tilt angle of the heat dissipation module relative to the infrared thermal imager is calculated. The infrared image is angle-corrected based on the perspective transformation algorithm, and the corrected infrared image with the facing view is output.

[0013] Furthermore, in step S2, the RGB-temperature mapping model is a backpropagation neural network model. The backpropagation neural network model has 3 neurons in the input layer, corresponding to the RGB three-channel values ​​of the infrared image, and 1 neuron in the output layer, corresponding to the temperature value.

[0014] Furthermore, in step S2, the infrared temperature measurement results in each temperature measurement sub-region are calibrated for deviation based on the measured temperature of the thermocouple, including: Calculate the average infrared temperature measurement value of all temperature measurement points in each temperature measurement sub-region, compare the average infrared temperature measurement value with the actual measured temperature of the thermocouples in the temperature measurement sub-region to obtain the regional average deviation of the temperature measurement sub-region, and distribute the regional average deviation to each temperature measurement point in the temperature measurement sub-region through an interpolation algorithm.

[0015] Furthermore, the method for constructing the temperature deviation compensation model in step S3 includes: Temperature deviation data of multiple heat dissipation modules of different specifications were collected for the entire area under multiple heating powers. The temperature deviation data is the difference between the reconstructed temperature and the measured temperature. A three-dimensional deviation compensation model was constructed by surface fitting with physical coordinates and heating power as independent variables and temperature deviation as dependent variable.

[0016] Furthermore, in step S4, based on the thermal diffusion parameters of the interface thermally conductive material, the corrected temperature field is calculated in reverse to obtain the chip-side temperature field distribution, specifically including: The heat source size, thermal diffusion angle, material thickness and thermal conductivity of the interface thermal conductive material are obtained. Based on the principle of thermal diffusion, the temperature distribution of the chip contact surface before heat transfer in the interface thermal conductive material is calculated by back-calculating the corrected temperature field, and the temperature field distribution on the chip side is obtained.

[0017] Furthermore, in step S4, the temperature field quality evaluation model includes five evaluation indicators: temperature accuracy, temperature gradient fidelity, spatial precision, noise suppression capability, and effective pixel ratio. The comprehensive quality score is obtained by weighted summation of each evaluation indicator.

[0018] Furthermore, the calculation method for temperature accuracy is as follows: ; in This represents the average deviation between the temperature values ​​at various points in the temperature field and the actual values. For the scene temperature range; The temperature gradient fidelity is calculated as follows: ,in The average temperature gradient of the temperature field. Standard temperature gradient; The method for calculating spatial precision is as follows: ,in Let be the pixel density of the temperature field. This represents the system's highest pixel density. To effectively measure the lowest pixel density; The noise suppression capability is calculated as follows: ,in The temperature standard deviation of the temperature field. Maximum allowable noise for the scene; The effective pixel percentage is calculated as follows: ,in Effective pixel count This represents the total number of pixels.

[0019] A system for constructing and evaluating the temperature field of a heat dissipation module based on multimodal coupling includes: The object recognition module is used to acquire infrared images of the heat dissipation module under test. Based on the difference in infrared emissivity between the heat dissipation module and the background environment in the infrared image, the module area of ​​the heat dissipation module is identified and extracted, the module area is angle-corrected, and the corrected infrared image is output. The temperature field construction module is used to input the RGB three-channel values ​​of the calibrated infrared image into a pre-trained RGB-temperature mapping model to obtain the initial infrared temperature distribution; discrete temperature measurement points are arranged within the module area, and each temperature measurement point is associated based on the thermal resistance model of the heat dissipation module; the module area is divided into multiple temperature measurement sub-regions, and the measured temperature of the thermocouples in each temperature measurement sub-region is obtained; the infrared temperature measurement results in each temperature measurement sub-region are calibrated based on the measured temperature of the thermocouples, and the calibrated deviation is distributed to each temperature measurement point; based on the calibrated temperature data of each temperature measurement point and the thermal resistance model, the global temperature field of the heat dissipation module is reconstructed and generated. The result correction module is used to obtain the current heating power of the heat dissipation module. Based on the physical coordinates and heating power of each temperature measurement point, it calls the pre-stored temperature deviation compensation model to calculate the deviation compensation value of each temperature measurement point. The deviation compensation value is used to correct the global temperature field point by point and output the corrected temperature field. The results evaluation module is used to perform reverse calculations on the corrected temperature field based on the thermal diffusion parameters of the interface thermally conductive material to obtain the temperature field distribution on the chip side; it constructs a temperature field quality evaluation model, quantifies and scores the corrected temperature field, and outputs the temperature field quality evaluation results.

[0020] In summary, the present invention has the following advantages: (1) Achieved deep fusion of multimodal infrared temperature measurement and thermocouple temperature measurement. Specifically, this invention breaks through the limitations of traditional single temperature measurement mode: on the one hand, it uses the RGB-temperature mapping model to replace the linear correspondence between gray value and temperature in traditional infrared temperature measurement, which significantly improves the accuracy and surface adaptability of infrared temperature measurement; on the other hand, it arranges discrete temperature measurement points in the module area and associates each temperature measurement point based on the material thermal resistance model, and introduces the measured values ​​of partitioned thermocouples to calibrate the deviation of infrared temperature measurement results, organically combining the global coverage advantage of infrared temperature measurement with the single-point high precision advantage of thermocouples, to achieve the effect of global coverage and local precision temperature field reconstruction.

[0021] (2) The entire process of temperature field construction is self-optimized. After reconstructing the temperature field, dynamic deviation compensation, temperature field inversion and quantitative evaluation are performed in sequence, and the algorithm parameters of the temperature field construction module can be optimized in reverse according to the evaluation results. This closed-loop mechanism enables the system to continuously output a stable and high-precision temperature field under varying operating conditions.

[0022] (3) It realizes non-contact inversion calculation from module surface temperature to chip side temperature. Based on the thermal diffusion mechanism of interface thermal conductive materials, combined with physical parameters such as heat source size, TIM thickness, thermal diffusion angle and thermal conductivity, the chip side temperature distribution is calculated in reverse from the reconstructed module surface temperature field. This expands the application scenario of the temperature measurement system from module-level performance testing to chip-level heat source analysis, providing a more direct reference for the thermal design optimization of electronic devices. Attached Figure Description

[0023] Figure 1 This is a flowchart of the method steps of the present invention; Figure 2 This is a schematic diagram of the module identification and angle correction of the present invention; Figure 3 This is a schematic block diagram of the temperature mapping network and multimodal temperature field reconstruction of the present invention; Figure 4 This is a schematic diagram of the three-dimensional model of the test deviation value of the present invention (Z-axis: deviation, Y-axis: effective power, X-axis: position). Figure 5 This is a schematic diagram of temperature field inversion in this invention (module surface temperature field → chip surface temperature field). Figure 6 This is a schematic diagram of the temperature field comparison results of the present invention. Detailed Implementation

[0024] The present invention will now be described in further detail.

[0025] The temperature field construction and evaluation system for heat dissipation modules based on multimodal coupling is an independent operating architecture. It does not need to adapt to the computing power and interface limitations of multiple hardware modules. The system is set up in sequence according to the data processing flow, including object recognition module, temperature field construction module, result correction module and result evaluation module. Among them, the object recognition module realizes automatic identification and angle adaptive correction of heat dissipation module based on infrared emissivity difference; the temperature field construction module integrates RGB-temperature mapping model based on neural network and real-time temperature field reconstruction unit coupled with infrared and thermocouple to realize global temperature field construction by multi-source data fusion; the result correction module corrects system error through temperature calibration and deviation compensation model; and the result evaluation module realizes temperature field inversion approximation judgment and quantitative evaluation of reconstructed temperature field quality.

[0026] This invention significantly improves the accuracy and scenario generalization ability of temperature field reconstruction of heat dissipation modules, realizes full-process automation of temperature field construction and standardized evaluation of results, and provides a precise and efficient systematic solution for heat transfer performance testing and thermal design optimization of heat dissipation modules for electronic devices.

[0027] like Figure 1 As shown below, each step will be described in detail.

[0028] S1: By exploiting the difference in infrared emissivity between the heat dissipation module and the test environment, the infrared image is converted into a grayscale image. A grayscale threshold is then set to isolate the heat dissipation module from the test environment, enabling its identification. Simultaneously, using the core feature of the heat dissipation module (the copper plate in the heating area) as a reference, a bounding box is selected. The selected quadrilateral is compared to a standard rectangle to calculate the module's tilt angle. Based on this calculated tilt angle, a perspective transformation algorithm is used to construct a spatial transformation matrix. This maps the quadrilateral module region detected in the original infrared image to the standard rectangular viewpoint, eliminating geometric distortion caused by the tilted shooting angle. Finally, a corrected infrared image facing the correct viewpoint is output, such as... Figure 2 As shown.

[0029] S2: The basic principle of infrared thermometry is to first convert the infrared image into a grayscale image, and then measure the temperature based on the correspondence between the grayscale values ​​of each pixel in the grayscale image and the temperature value. The correspondence model between grayscale values ​​and temperature values ​​is crucial to the accuracy of infrared camera temperature measurement. Typically, the accuracy of infrared camera temperature measurement is affected by many factors, involving the adjustment of various parameters. The RGB-temperature mapping model based on neural networks, however, is based on mathematical calibration and can be directly trained using existing experimental data (infrared images, corresponding temperature data) to quickly construct the correspondence between the RGB values ​​of the infrared image and the temperature value. Its core feature is that it directly establishes predictive relationships between variables without relying on mechanistic understanding, and the resulting predictive model achieves excellent fitting results in fixed test scenarios. The trained model obtained above will be directly applied to the temperature measurement module of the infrared camera.

[0030] like Figure 3 As shown, the real-time reconstruction of the temperature field based on infrared and thermocouple coupling first requires the arrangement of temperature measurement points from an infrared perspective, similar to the distribution of coordinate points in a meshless method. Since the material properties of the test object are known, a thermal resistance model of the material can be directly constructed to correlate each temperature measurement point, similar to constructing interpolation functions and discrete control equations between coordinate points in a meshless method. After arranging the infrared temperature measurement points, each temperature measurement point is divided into zones, with several points grouped into a single temperature measurement zone. Multiple hot thermocouples are arranged within each temperature measurement zone; the number of hot thermocouples is typically much smaller than the number of temperature measurement points. The average temperature of the temperature measurement zone is compared with the thermocouples in real time, achieving the goal of calibrating the infrared temperature measurement accuracy through thermocouples. The interpolated temperature comparison is fed back to each temperature measurement point to optimize the temperature measurement data. Finally, based on the data from each temperature measurement point and the associated thermal resistance model, the global temperature field of the test object is constructed.

[0031] S3: Test multiple models of heat dissipation modules, compare actual test data with reconstructed temperature field data, collect deviation data under various physical coordinates of the heat dissipation module, and finally fit the data based on the deviation data, test power, and physical coordinates. The resulting fitted model is the temperature calibration and deviation compensation model. When constructing the deviation model of the physical heat dissipation module, it is simplified to a one-dimensional model, that is, the temperature deviation values ​​on the module are uniformly mapped to the x-axis for analysis and processing. A schematic diagram of the three-dimensional model of the test deviation values ​​is shown below. Figure 4 As shown.

[0032] Table 1. Deviation Models for Heat Sources of Each Chip S4: The laptop cooling module is a core component for heat dissipation of laptop chips. The computer chip is attached to the cooling module to transfer heat. The complete heat transfer process includes: chip core - chip casing - thermal conductive medium (TIM) - cooling module. Therefore, the reconstructed temperature field of the cooling module cannot strictly represent the temperature of the chip surface. Here, by calculating the diffusion of heat during the transfer process through the TIM material, a reverse calculation is performed based on the temperature field of the cooling module to obtain an image that more closely approximates the temperature field of the chip surface. For example... Figure 5 As shown, Indicates the size of the heat source at the chip contact surface. This represents the size of the heat diffusion surface on the heat dissipation module after the chip's heat is diffused through the interface thermal conductive material, and θ represents the heat diffusion angle under this interface thermal conductive material. It refers to the thickness of the interfacial thermally conductive material. In the experiment, ,θ, Given this, the corresponding chip surface temperature field can be derived from the module surface temperature field. Furthermore, the chip surface temperature can be inferred from the thermal conductivity and thermal resistance of the interface thermally conductive material.

[0033] By comparing the inverted temperature field with the chip temperature field cloud map, the accuracy of the tested heat source can be determined. The approximate comparison of temperature field images is mainly based on a weighted evaluation of two factors: differences in the spatial distribution of hot spots in the temperature field and differences in the numerical values ​​of various points in the temperature field. Figure 6 As shown.

[0034] Temperature field image quality refers to the comprehensive characterization of the physical fidelity (accuracy of temperature values, reasonableness of gradient distribution) and spatial precision (ability to reproduce details) of an image to the real / theoretical temperature field under a specific thermal scenario. The core is whether it can accurately convey temperature distribution information while minimizing the impact of noise and distortion on thermal analysis, ultimately supporting practical applications such as simulation verification and experimental evaluation. The definition formula for reconstructed temperature field quality is as follows. Typically, the weight parameters of the influencing factors in the following formula are adjusted according to different focuses on the generated temperature field. ): ; Table 2. Parameter List of Temperature Field Mass Calculation Model In the above table, Temperature range for the scene; temperature gradient = (Temperature difference between adjacent pixels / physical distance); Standard gradient; physical density =Number of pixels / Actual size; The standard deviation of temperature; Maximum allowable noise for the scene; Valid pixels. When ≥0.8 indicates that the temperature field is of "ultra-mass quality", 0.6≤ <0.8 indicates "high quality" A value less than 0.4 indicates "low quality".

[0035] This invention utilizes only a single frame of infrared image captured by an infrared thermal imager. Based on the difference in infrared emissivity between the metal surface of the heat dissipation module and the environmental background, the infrared image is converted into a grayscale image, and a grayscale judgment threshold is set to automatically segment and extract the heat dissipation module area. Then, using the rectangular geometric features of the copper plate in the module's heating area as a benchmark, the module's tilt angle is calculated by the degree of distortion between the quadrilateral contour and the standard rectangle. Based on the principle of perspective transformation, the image's angle adaptive correction is completed. No additional visible light hardware is required; automatic target identification and angle correction can be completed solely based on infrared images. The entire process is without manual intervention, eliminating emissivity measurement errors caused by angle deviations, and improving the automation level of testing and the consistency of temperature measurement under different placement scenarios.

[0036] This invention first constructs a backpropagation (BP) neural network, using the RGB three-channel values ​​of an infrared image as input and the corresponding true temperature value as output. An RGB-temperature nonlinear mapping model is obtained through training with measured samples, replacing traditional linear grayscale calibration. Secondly, drawing on the idea of ​​a meshless method, discrete temperature measurement points are arranged within the infrared field of view, and each temperature measurement point is associated based on the thermal resistance model of the material of the heat dissipation module under test. Then, the temperature measurement area is divided into multiple sub-regions, and a small number of thermocouples are arranged in each sub-region. The infrared temperature measurement results within the region are calibrated in real time using the measured values ​​of the thermocouples as a benchmark, and the calibration value is interpolated and diffused to each temperature measurement point within the region. Finally, a global high-precision temperature field is generated by combining a thermal resistance interpolation model. The nonlinear mapping of the neural network improves the adaptability of infrared temperature measurement to different surfaces. Combining partitioned thermocouple calibration and thermal resistance model interpolation, it simultaneously takes into account the global coverage advantage of infrared temperature measurement and the single-point high-precision advantage of thermocouples. The spatial resolution and numerical accuracy of the reconstructed temperature field are significantly better than existing solutions.

[0037] This invention targets various heat dissipation modules of different specifications, collecting data on the deviation between the reconstructed temperature and the measured temperature across the entire area under multiple heating powers. Using physical coordinates and heating power as independent variables and temperature deviation as the dependent variable, a three-dimensional deviation compensation model is constructed through surface fitting. In actual testing, the compensation model is dynamically invoked to correct the reconstructed temperature deviation point-by-point based on the current test power and measurement point location. This overcomes the limitations of calibration under single operating conditions and at a single point, adapting to dynamic deviation compensation under different power and location conditions, significantly improving the model's generalization ability under varying operating conditions and the accuracy of temperature measurement across the entire area.

[0038] On one hand, this invention, based on the thermal diffusion mechanism of thermally conductive interfacial materials (TIMs), and combining the heat source size, TIM thickness, thermal diffusion angle, and thermal conductivity, calculates the temperature distribution on the chip side from the surface temperature field of the heat dissipation module, achieving non-contact inversion of the heat source temperature. On the other hand, it constructs a temperature field quality evaluation model that includes five indicators: temperature accuracy, temperature gradient rationality, spatial resolution, noise level, and effective pixel ratio. A quantitative quality score is obtained through weighted summation, and quality levels are assigned, achieving standardized judgment of the reconstruction results. Simultaneously, the temperature measurement and reconstruction algorithm parameters can be optimized based on the evaluation results, increasing the temperature inversion capability from the module surface to the chip heat source and expanding the application scenarios of the temperature measurement system. Furthermore, it establishes a quantifiable and reproducible temperature field quality evaluation standard, solving the problems of subjective evaluation and lack of unified standards in existing technologies, and improving the credibility and comparability of the detection results.

[0039] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A method for constructing and evaluating the temperature field of a heat dissipation module based on multimodal coupling, characterized in that, include: S1: Acquire the infrared image of the heat dissipation module under test. Based on the difference in infrared emissivity between the heat dissipation module and the background environment in the infrared image, identify and extract the module area of ​​the heat dissipation module, perform angle correction on the module area, and output the corrected infrared image. S2: Input the RGB three-channel values ​​of the corrected infrared image into the pre-trained RGB-temperature mapping model to obtain the initial infrared temperature distribution; arrange discrete temperature measurement points in the module area, and associate each temperature measurement point based on the material thermal resistance model of the heat dissipation module; The module area is divided into multiple temperature measurement sub-regions. The measured temperature of the thermocouples in each temperature measurement sub-region is obtained. The infrared temperature measurement results in each temperature measurement sub-region are calibrated based on the measured temperature of the thermocouples. The calibrated deviation is then distributed to each temperature measurement point. Based on the calibrated temperature data of each temperature measurement point and the thermal resistance model, the global temperature field of the heat dissipation module is reconstructed. S3: Obtain the current heating power of the heat dissipation module. Based on the physical coordinates and heating power of each temperature measurement point, call the pre-stored temperature deviation compensation model, calculate the deviation compensation value of each temperature measurement point, use the deviation compensation value to correct the global temperature field point by point, and output the corrected temperature field. S4: Based on the thermal diffusion parameters of the interface thermally conductive material, the corrected temperature field is calculated in reverse to obtain the temperature field distribution on the chip side; a temperature field quality evaluation model is constructed to quantify and score the corrected temperature field and output the temperature field quality evaluation results.

2. The method for constructing and evaluating the temperature field of a heat dissipation module based on multimodal coupling according to claim 1, characterized in that, Step S1 involves identifying and extracting the module area of ​​the heat dissipation module, specifically including: The infrared image is converted into a grayscale image. A grayscale judgment threshold is set based on the difference in infrared emissivity between the metal surface of the heat dissipation module and the ambient background. The heat dissipation module is then separated from the background through threshold segmentation to obtain the module area.

3. The method for constructing and evaluating the temperature field of a heat dissipation module based on multimodal coupling according to claim 1, characterized in that, Step S1 involves angle correction of the module area, including: The quadrilateral contour of the copper plate in the heating area of ​​the module is identified, and the detected quadrilateral contour is compared with the standard rectangular contour. The tilt angle of the heat dissipation module relative to the infrared thermal imager is calculated. The infrared image is angle-corrected based on the perspective transformation algorithm, and the corrected infrared image with the facing view is output.

4. The method for constructing and evaluating the temperature field of a heat dissipation module based on multimodal coupling according to claim 1, characterized in that, In step S2, the RGB-temperature mapping model is a backpropagation neural network model. The input layer of the backpropagation neural network model has 3 neurons, corresponding to the RGB three-channel values ​​of the infrared image, and the output layer has 1 neuron, corresponding to the temperature value.

5. The method for constructing and evaluating the temperature field of a heat dissipation module based on multimodal coupling according to claim 1, characterized in that, Step S2 involves calibrating the infrared thermometry results in each thermometric sub-region based on the measured temperature of the thermocouple, including: Calculate the average infrared temperature measurement value of all temperature measurement points in each temperature measurement sub-region, compare the average infrared temperature measurement value with the actual measured temperature of the thermocouples in the temperature measurement sub-region to obtain the regional average deviation of the temperature measurement sub-region, and distribute the regional average deviation to each temperature measurement point in the temperature measurement sub-region through an interpolation algorithm.

6. The method for constructing and evaluating the temperature field of a heat dissipation module based on multimodal coupling according to claim 1, characterized in that, The method for constructing the temperature deviation compensation model in step S3 includes: Temperature deviation data of multiple heat dissipation modules of different specifications were collected for the entire area under multiple heating powers. The temperature deviation data is the difference between the reconstructed temperature and the measured temperature. A three-dimensional deviation compensation model was constructed by surface fitting with physical coordinates and heating power as independent variables and temperature deviation as dependent variable.

7. The method for constructing and evaluating the temperature field of a heat dissipation module based on multimodal coupling according to claim 1, characterized in that, In step S4, based on the thermal diffusion parameters of the interface thermally conductive material, the corrected temperature field is calculated in reverse to obtain the chip-side temperature field distribution, specifically including: The heat source size, thermal diffusion angle, material thickness and thermal conductivity of the interface thermal conductive material are obtained. Based on the principle of thermal diffusion, the temperature distribution of the chip contact surface before heat transfer in the interface thermal conductive material is calculated by back-calculating the corrected temperature field, and the temperature field distribution on the chip side is obtained.

8. The method for constructing and evaluating the temperature field of a heat dissipation module based on multimodal coupling according to claim 1, characterized in that, In step S4, the temperature field quality evaluation model includes five evaluation indicators: temperature accuracy, temperature gradient fidelity, spatial precision, noise suppression capability, and effective pixel ratio. The comprehensive quality score is obtained by weighted summation of each evaluation indicator.

9. The method for constructing and evaluating the temperature field of a heat dissipation module based on multimodal coupling according to claim 8, characterized in that: The method for calculating temperature accuracy is as follows: ; in This represents the average deviation between the temperature values ​​at various points in the temperature field and the actual values. For the scene temperature range; The temperature gradient fidelity is calculated as follows: ,in The average temperature gradient of the temperature field. Standard temperature gradient; The method for calculating spatial precision is as follows: ,in Let be the pixel density of the temperature field. This represents the system's highest pixel density. To effectively measure the lowest pixel density; The noise suppression capability is calculated as follows: ,in The temperature standard deviation of the temperature field. Maximum allowable noise for the scene; The effective pixel percentage is calculated as follows: ,in Effective pixel count This represents the total number of pixels.

10. A system for constructing and evaluating the temperature field of a heat dissipation module based on multimodal coupling, characterized in that, The method for constructing and evaluating the temperature field of a heat dissipation module based on multimodal coupling as described in any one of claims 1 to 9 includes: The object recognition module is used to acquire infrared images of the heat dissipation module under test. Based on the difference in infrared emissivity between the heat dissipation module and the background environment in the infrared image, the module area of ​​the heat dissipation module is identified and extracted, the module area is angle-corrected, and the corrected infrared image is output. The temperature field construction module is used to input the RGB three-channel values ​​of the calibrated infrared image into a pre-trained RGB-temperature mapping model to obtain the initial infrared temperature distribution; discrete temperature measurement points are arranged within the module area, and each temperature measurement point is associated based on the thermal resistance model of the heat dissipation module; the module area is divided into multiple temperature measurement sub-regions, and the measured temperature of the thermocouples in each temperature measurement sub-region is obtained; the infrared temperature measurement results in each temperature measurement sub-region are calibrated based on the measured temperature of the thermocouples, and the calibrated deviation is distributed to each temperature measurement point; based on the calibrated temperature data of each temperature measurement point and the thermal resistance model, the global temperature field of the heat dissipation module is reconstructed and generated. The result correction module is used to obtain the current heating power of the heat dissipation module. Based on the physical coordinates and heating power of each temperature measurement point, it calls the pre-stored temperature deviation compensation model to calculate the deviation compensation value of each temperature measurement point. The deviation compensation value is used to correct the global temperature field point by point and output the corrected temperature field. The results evaluation module is used to perform reverse calculations on the corrected temperature field based on the thermal diffusion parameters of the interface thermally conductive material to obtain the temperature field distribution on the chip side; it constructs a temperature field quality evaluation model, quantifies and scores the corrected temperature field, and outputs the temperature field quality evaluation results.

Images