Heat-conducting gel performance evaluation method and system based on computer vision
The thermal conductive gel performance evaluation system using computer vision technology solves the problem of inaccurate evaluation under dynamic temperature conditions by existing methods. It realizes full-cycle performance evaluation and micro-defect identification of thermal conductive gels, improving the accuracy and practicality of the evaluation.
Patent Information
- Application Number
- CN202610014501.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-07
- Publication Date
- 2026-02-10
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing methods for evaluating the performance of thermally conductive gels cannot accurately assess them under dynamic temperature conditions, ignore the thermal response during equipment startup and load changes, and have difficulty identifying microscopic defects, resulting in discrepancies between the evaluation results and actual performance.
A computer vision-based thermal conductive gel performance evaluation system is adopted, which includes modules for image acquisition, thermal field analysis, mesh generation, defect identification, and performance evaluation. The system analyzes the temperature distribution through infrared thermal imaging image sequence analysis, generates a multi-scale mesh structure, identifies microscopic defects and calculates defect evolution indices, and generates evaluation results by combining the temperature gradient distribution.
It enables full-cycle performance evaluation of thermally conductive gels under dynamic temperature environments, improving the accuracy and efficiency of the evaluation. It can identify the impact of microscopic defects and provide a more scientific basis for performance evaluation, making it suitable for production line quality inspection and equipment operation and maintenance.
Smart Images
Figure CN121499593A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of thermal conductive material testing technology, specifically to a method and system for evaluating the performance of thermal conductive gels based on computer vision. Background Technology
[0002] As electronic devices evolve towards higher integration and higher power density, the large amount of heat generated during operation, if not promptly conducted and dissipated, can lead to increased temperatures in core components, thereby affecting the stability and lifespan of the device. Thermally conductive gels, as thermal interface materials with excellent filling and thermal conductivity, are widely used in the thermal management systems of electronic devices such as chips, power batteries, and communication base stations. Their thermal conductivity directly determines the efficiency of the entire thermal management system. Currently, the industry's methods for evaluating the performance of thermally conductive gels mainly fall into two categories: one is the indirect measurement method based on physical parameters, which evaluates the performance of the thermally conductive gel by detecting basic physical quantities such as thermal conductivity and thermal diffusivity. Commonly used testing equipment includes hot-wire thermal conductivity meters and laser flash thermal conductivity meters. This type of method requires sampling and processing of the thermally conductive gel, and the testing process must be completed in a specific laboratory environment. This not only fails to achieve performance evaluation of the thermally conductive gel under actual working conditions, but also suffers from long testing cycles and insufficient sample representativeness: some samples may have significantly different physical parameters obtained from the tests due to stress changes during the preparation process compared to the performance in actual applications.
[0003] Another type is the direct observation method based on the temperature field. This method uses an infrared thermal imager to acquire images of the temperature distribution on the surface of the thermally conductive gel, and then combines this with manual analysis or simple image processing algorithms to determine its heat conduction effect. However, existing temperature field-based evaluation methods have significant limitations: most methods only analyze the temperature distribution of the thermally conductive gel under stable temperature conditions, ignoring the performance during dynamic temperature changes such as device startup and load variations. In practical applications, the thermal response speed and heat distribution uniformity of the thermally conductive gel during temperature changes are crucial for the transient thermal protection of electronic devices. Existing image processing algorithms mostly use a single-scale mesh division method, which cannot adjust the analysis accuracy according to the resolution differences in temperature distribution. In areas with large temperature gradients, a single-scale mesh is difficult to accurately capture the details of local heat distribution, while in areas with uniform temperature distribution, it will waste computational resources. Existing methods cannot effectively identify microscopic defects inside the thermally conductive gel (such as microbubbles, local cracks, etc.). Although these defects are difficult to detect visually, they can significantly affect the local heat conduction path, leading to uneven heat distribution and potentially causing local overheating damage to electronic devices. Summary of the Invention
[0004] The purpose of this invention is to provide a computer vision-based method and system for evaluating the performance of thermally conductive gels, in order to solve the problems mentioned in the background art.
[0005] To achieve the above objectives, the present invention provides a computer vision-based system for evaluating the performance of thermally conductive gels, the system comprising: The module includes an image acquisition module, a thermal field analysis module, a mesh generation module, a defect identification module, and a performance evaluation module; among them, The image acquisition module is used to acquire infrared thermal imaging image sequences of the thermally conductive gel under different temperature conditions; The thermal field analysis module is used to extract the spatiotemporal sequence of temperature distribution from the infrared thermal imaging image sequence; and to divide the stable temperature stage and the changing temperature stage according to the temperature change trend of each pixel in the spatiotemporal sequence of temperature distribution. The mesh generation module is used to select the region to be analyzed based on the degree of difference in temperature distribution between the stable temperature stage and the changing temperature stage; within the region to be analyzed, a multi-scale mesh structure is generated based on the resolution difference in temperature distribution. The defect identification module is used to identify the microscopic defect features of the thermally conductive gel in a multi-scale grid structure; and to calculate the defect evolution index based on the morphological differences of the microscopic defect features in the stable temperature stage and the temperature variation stage. The performance evaluation module is used to combine the defect evolution index with the temperature gradient distribution of the multi-scale grid structure to generate the thermal conductivity performance evaluation results of the thermally conductive gel.
[0006] Preferably, the thermal field analysis module extracts the spatiotemporal sequence of temperature distribution from the infrared thermal imaging image sequence, including: The infrared thermal imaging image sequence was standardized to obtain normalized temperature data. Based on a pre-defined convolutional neural network, spatial temperature field features and temporal temperature change features are extracted from normalized temperature data to generate a spatiotemporal sequence of temperature distribution.
[0007] Preferably, the thermal field analysis module divides the temperature field into a stable temperature stage and a changing temperature stage based on the temperature change trend of each pixel in the spatiotemporal sequence of temperature distribution, including: For any pixel, obtain its temperature value as a function of time, and calculate the first difference of the temperature sequence. Select the moments with positive first-order difference values as the heating moments; sort the heating moments in ascending order according to their corresponding first-order difference values, and calculate the second-order difference values of the sorted sequence. The time of temperature rise corresponding to the maximum value of the second-order difference is taken as the critical time. The time period before the critical time is the stable temperature stage, and the time period after the critical time is the changing temperature stage.
[0008] Preferably, the mesh generation module filters out the region to be analyzed based on the degree of difference in temperature distribution between the stable temperature stage and the changing temperature stage, including: Calculate the difference between the average temperature during the stable temperature phase and the average temperature during the changing temperature phase within the region to be analyzed; The temperature difference coefficient is obtained by normalizing the difference; if the temperature difference coefficient is greater than the preset difference threshold, the region is marked as the region to be analyzed.
[0009] Preferably, the mesh generation module generates a multi-scale mesh structure within the region to be analyzed based on the resolution differences in temperature distribution, including: The region to be analyzed is divided into multiple sub-regions; the grid cell size and grid density are set according to the temperature distribution resolution of each sub-region. For high-resolution sub-regions, generate fine meshes; for low-resolution sub-regions, generate sparse meshes. By performing boundary synchronization optimization on the meshes of each sub-region, a multi-scale mesh structure is obtained.
[0010] Preferably, the defect identification module identifies microscopic defect features of the thermally conductive gel in a multi-scale grid structure, including: Within each grid cell of the multi-scale grid structure, extract the temperature gradient anomaly region; Based on the geometric morphology and temperature distribution continuity of the temperature gradient anomaly region, the type and location of micro-defect features are identified.
[0011] Preferably, the defect identification module calculates defect evolution indices based on the morphological differences of microscopic defect characteristics in the stable temperature stage and the changing temperature stage, including: Obtain the area percentage of micro-defect features during the stable temperature stage and the area percentage during the changing temperature stage. Calculate the ratio of the area proportions of the two stages, and combine it with the defect type weighting coefficient to generate a defect evolution index.
[0012] Preferably, the performance evaluation module combines defect evolution indices with the temperature gradient distribution of the multi-scale grid structure to generate thermal conductivity performance evaluation results for the thermally conductive gel, including: In a multi-scale grid structure, calculate the mean temperature gradient of each grid cell; The corrected temperature gradient is obtained by weighting and correcting the mean temperature gradient based on the defect evolution index. Based on the spatial distribution of the corrected temperature gradient, thermal conductivity performance evaluation results are generated.
[0013] Preferably, the performance evaluation module generates heat conduction performance evaluation results based on the spatial distribution law of the corrected temperature gradient, including: The corrected temperature gradient is divided into multiple levels according to its magnitude; the proportion of grid cells corresponding to each level is calculated. Based on the degree of matching between the proportion distribution and the preset thermal conductivity standard range, the thermal conductivity performance evaluation results are output.
[0014] Preferably, the present invention also includes a computer vision-based method for evaluating the performance of thermally conductive gels, the method comprising all the modules and process flow of the computer vision-based thermally conductive gel performance evaluation system described above.
[0015] Compared with the prior art, the beneficial effects of the present invention are: From a temperature phase analysis perspective, the system's thermal field analysis module can extract the spatiotemporal sequence of temperature distribution from infrared thermal imaging images and divide the temperature change into stable and fluctuating temperature phases based on the temperature change trend of each pixel. This enables full-cycle performance evaluation of the thermally conductive gel under both dynamic and static temperature environments. Compared to traditional evaluation methods that only focus on stable temperature states, this system can comprehensively capture the thermal response process of the thermally conductive gel in actual working scenarios such as equipment startup and load fluctuations. Whether it's the rapidly rising temperature phase or the stable temperature phase, targeted analysis can be performed, making the evaluation results more consistent with the actual application scenarios of the thermally conductive gel and avoiding evaluation bias caused by neglecting the dynamic temperature phase. In terms of balancing accuracy and efficiency, the mesh generation module filters the analysis area based on the difference in temperature distribution between stable and changing temperature phases, and generates a multi-scale mesh structure based on the resolution differences in temperature distribution. This multi-scale mesh design can flexibly adjust the analysis accuracy according to the temperature distribution characteristics of different regions: in areas with large temperature gradients and complex thermal distributions, fine-scale meshes achieve high-precision detail capture, accurately reflecting local differences in heat conduction; in areas with uniform temperature distribution and small thermal gradients, coarse-scale meshes are used to reduce computational complexity and minimize unnecessary computational resource consumption. This approach not only resolves the contradiction between accuracy and efficiency inherent in traditional single-scale meshes, but also ensures the accuracy of analysis in key areas, improving the efficiency and reliability of the overall evaluation process. In terms of defect identification and performance correlation, the defect identification module can accurately identify the microscopic defect characteristics of thermally conductive gels in multi-scale grid structures and calculate defect evolution indices based on the morphological differences of defects at stable and varying temperatures. Microscopic defects such as tiny bubbles and localized cracks, which are difficult to detect using traditional evaluation methods, can be effectively captured by this module. Furthermore, by analyzing the morphological changes of defects at different temperature stages (such as changes in the temperature diffusion range of the defect area and dynamic adjustments of the defect boundary), the impact of defects on thermal conductivity can be further determined. This analytical approach, which combines microscopic defects with dynamic temperature changes, allows the evaluation results to not only reflect the overall thermal conductivity of the thermally conductive gel but also reveal the root causes of performance anomalies. This provides more targeted references for optimizing the production process of thermally conductive gels (such as reducing bubble generation during preparation) and for application maintenance (such as early detection of areas prone to failure). From the perspective of comprehensive performance evaluation, the performance evaluation module combines defect evolution indicators with the temperature gradient distribution of multi-scale grid structures to generate evaluation results, achieving a multi-dimensional evaluation of "macroscopic thermal distribution + microscopic defects + dynamic response". Traditional methods rely only on a single physical parameter or static temperature distribution for evaluation, which cannot fully reflect the actual performance of thermally conductive gels. However, this system integrates temperature gradient distribution (reflecting macroscopic heat conduction uniformity), defect evolution indicators (reflecting the influence of microstructure on performance), and analysis results at different temperature stages (reflecting dynamic thermal response capability), which can form more complete and accurate performance evaluation conclusions, providing a more scientific basis for the selection, application optimization, and quality control of thermally conductive gels. The entire system utilizes computer vision technology to automate analysis, eliminating the need for manual sampling or subjective judgment. This not only reduces errors caused by human factors but also significantly shortens the evaluation cycle. Compared to traditional evaluation methods that require specific laboratory environments and rely on professional personnel, this system can be more easily applied to real-world scenarios such as production line quality inspection and equipment maintenance. It enhances the practicality and operability of thermal conductive gel performance evaluation, further promoting the industrial application of thermal management system performance optimization. Attached Figure Description
[0016] Figure 1 This is a timing diagram of the computer vision-based thermal conductive gel performance evaluation system described in this invention. Figure 2 A flowchart for dividing the steady-temperature stage and the changing temperature stage; Figure 3 A flowchart for generating multi-scale mesh structures; Figure 4 A flowchart for calculating defect evolution indices; Figure 5 This is a flowchart for generating thermal conductivity performance evaluation results based on a modified temperature gradient. Detailed Implementation
[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0018] Please see Figure 1 The present invention provides a computer vision-based thermal conductive gel performance evaluation system, the system comprising: an image acquisition module, a thermal field analysis module, a mesh generation module, a defect identification module, and a performance evaluation module.
[0019] The image acquisition module acquires infrared thermal imaging image sequences of the thermally conductive gel under different temperature conditions. The thermal field analysis module extracts the spatiotemporal sequence of temperature distribution from the infrared thermal imaging image sequences and divides the temperature distribution into stable temperature stages and varying temperature stages based on the temperature change trend of each pixel in the spatiotemporal sequence. The mesh generation module selects the region to be analyzed based on the degree of difference in temperature distribution within the stable and varying temperature stages, and generates a multi-scale mesh structure within the region to be analyzed based on the resolution difference of temperature distribution. The defect identification module identifies the microscopic defect features of the thermally conductive gel in the multi-scale mesh structure and calculates the defect evolution index based on the morphological differences of the microscopic defect features in the stable and varying temperature stages. The performance evaluation module combines the defect evolution index with the temperature gradient distribution of the multi-scale mesh structure to generate the thermal conductivity performance evaluation results of the thermally conductive gel.
[0020] Example 1: See Figure 2 The thermal field analysis module first standardizes the input infrared thermal imaging image sequence. This process includes unifying the dimensions of temperature values and normalizing image sizes. The temperature value of each pixel is converted to a uniform temperature unit, and all images are adjusted to the same spatial resolution. Through a linear transformation, the original temperature data is mapped to the range of zero to one, resulting in normalized temperature data. After obtaining the normalized temperature data, the module uses a pre-defined convolutional neural network for deep feature extraction. This network architecture employs a multi-branch design, with one branch focusing on spatial feature extraction and another handling temporal features. The spatial feature extraction branch uses two-dimensional convolutional layers, scanning each frame of thermal imaging images using a sliding window approach. The design of the convolutional kernels enables it to capture local temperature distribution patterns, such as regions of uniform temperature or gradient changes. The stacking of multiple convolutional layers allows the network to progressively construct complex spatial temperature field feature representations from simple features.
[0021] The temporal feature extraction branch employs 3D convolution operations, considering both spatial and temporal dimensions. This branch processes multiple consecutive frames of thermal imaging images, analyzing the evolution of the temperature field over time. The sliding of the convolutional kernel along the temporal dimension enables it to detect features such as the rate, direction, and periodicity of temperature changes. The network learns typical patterns of dynamic temperature field changes through training, reflecting the thermal response characteristics of the thermally conductive gel. The features extracted by the two branches are integrated through a fusion layer to generate a spatiotemporal sequence of temperature distribution. This sequence not only contains the temperature values at different time points for each spatial location but also includes information on the spatial structure and temporal evolution of the temperature field. The data structure of the spatiotemporal sequence is organized in tensor form, where the three dimensions correspond to the image height, width, and time series length, respectively.
[0022] After obtaining the spatiotemporal sequence of temperature distribution, the module begins to divide the temperature into stages. This process is performed independently for each pixel in the sequence. For any selected pixel, the sequence data of its temperature value changing over time is extracted. This sequence contains the temperature measurements of that pixel across all time frames, forming a time-temperature curve. The first-order difference values of this temperature sequence are analyzed, i.e., the change in temperature value between adjacent time points is calculated. These difference values reflect the instantaneous rate of temperature change. All moments with positive first-order difference values are selected; these moments correspond to the temperature rise process. These temperature rise moments are sorted in ascending order according to their corresponding first-order difference values, forming a new sequence.
[0023] Calculate the second-order difference value of this sorted sequence, i.e., analyze the change of the first-order difference value itself. The extreme points of the second-order difference value correspond to the inflection points of the temperature change rate. Find the maximum point of the second-order difference value, which marks the moment when the temperature change pattern changes significantly. Determine this moment as the critical moment, and divide the entire time series into two stages: the period before the critical moment is characterized by relatively gentle temperature changes, which is classified as the stable temperature stage; the period after the critical moment is characterized by more drastic temperature changes, which is classified as the changing temperature stage. This division method is based on the thermal response characteristics of the thermally conductive gel. In the stable temperature stage, the material's heat conduction reaches a relatively balanced state, and the temperature distribution exhibits stable characteristics; in the changing temperature stage, the material is undergoing thermal shock or temperature jumps, and its thermal behavior exhibits dynamic characteristics.
[0024] The entire processing is automated, requiring no manual intervention. The parameters of the convolutional neural network are trained using a large number of samples, enabling it to adapt to the differences in thermal properties of different types of thermally conductive gels. The temperature segmentation algorithm is based on rigorous mathematical derivation, ensuring the objectivity and repeatability of the segmentation results. The final output spatiotemporal sequence of temperature distribution and segmentation results are passed to subsequent modules for further mesh generation and defect identification analysis.
[0025] The key feature of this implementation is the combination of deep learning technology with traditional thermal analysis methods. The application of convolutional neural networks enables the system to automatically learn the complex characteristics of the temperature field without relying on manually designed feature extraction rules. The mathematical method for dividing temperature stages provides objective criteria, avoiding errors caused by subjective judgment. The entire processing flow is designed to consider various factors in practical applications, including temperature measurement noise, environmental interference, and individual differences in material properties.
[0026] Example 2: See Figure 3 This involves the specific operational flow of the mesh generation module in the thermally conductive gel performance evaluation system. This module receives processing results from the thermal field analysis module, including the spatiotemporal sequence of temperature distribution and pre-defined stable and varying temperature phases. The core task of the module is to locate the thermally sensitive regions requiring focused analysis within the entire sample area and construct mesh structures adapted to different temperature distribution characteristics within these regions.
[0027] During the screening of the region to be analyzed, the module first divides the entire surface of the thermally conductive gel sample into several basic analysis units. Each unit corresponds to a rectangular area containing several pixels. For each basic analysis unit, the average temperature value is calculated in both the stable temperature phase and the temperature variation phase. The average temperature in the stable temperature phase is obtained by averaging the temperature values of all pixels within the unit across all time frames during that phase. The average temperature in the temperature variation phase is calculated using the same method. The absolute difference between the two average temperatures is recorded.
[0028] The absolute difference is then converted into a temperature difference coefficient. The conversion process uses a normalization method, dividing the absolute difference by the maximum temperature difference across the entire sample surface. The temperature difference coefficient reflects the thermal response intensity of the analytical unit during the temperature transition process. A higher coefficient value indicates a stronger sensitivity of the region to temperature changes. A preset difference threshold is set according to actual application requirements to distinguish between ordinary and heat-sensitive areas. When the temperature difference coefficient of a basic analytical unit exceeds this threshold, that unit is marked as an area to be analyzed. For example, if the surface of a thermally conductive gel sample is divided into 100 basic analytical units, and 15 units have a temperature difference coefficient exceeding the set threshold of 0.35, these units are identified as areas to be analyzed.
[0029] After determining the region to be analyzed, the module begins constructing a multi-scale mesh structure. This process first involves dividing the region into sub-regions. The division is based on the temperature distribution resolution characteristics of each region. Resolution is evaluated by analyzing the spatial frequency of temperature changes within that region. In regions with drastic temperature gradients, the temperature differences between adjacent pixels are significant, resulting in higher resolution; in regions with uniform temperature distribution, adjacent pixels have similar temperatures, leading to lower resolution. Each region to be analyzed is automatically divided into several sub-regions, each with relatively consistent resolution characteristics.
[0030] For sub-regions with different resolutions, the module employs differentiated mesh generation strategies. For high-resolution sub-regions, a fine mesh structure is generated. The cell size of the fine mesh is set to be small, typically between 0.5 mm and 2 mm, with the specific size dynamically adjusted according to the actual resolution. The mesh density is correspondingly increased, and the number of cells increases. This design enables the mesh to accurately capture microscopic changes in the temperature field. For example, in a sub-region with microscopic defects, the temperature distribution exhibits a complex and irregular pattern. Using a 0.8 mm mesh cell size ensures that each temperature anomaly point is covered by an independent mesh cell.
[0031] For low-resolution sub-regions, a sparse mesh structure is generated, with larger cell sizes, typically ranging from 3 to 5 millimeters. Mesh density is reduced, and the number of cells is decreased. This design adapts to the characteristics of a gradual temperature distribution, avoiding unnecessary computational resource consumption. For example, in a sub-region with uniform material composition and a smooth temperature gradient, a 4-millimeter mesh cell size can characterize the overall temperature trend while maintaining computational efficiency.
[0032] After meshing each sub-region, the module performs boundary synchronization optimization. This step primarily addresses the connection issue between meshes of different scales. First, the mesh boundaries of adjacent sub-regions are detected, identifying differences in node positions at the boundaries. By inserting transition nodes or adjusting node spacing, meshes of different scales are seamlessly connected at the boundaries. Node connectivity is re-examined to ensure all mesh cells form a continuous topology. The optimization process employs an iterative algorithm, gradually adjusting the positions of boundary nodes until the requirements for smooth mesh transition are met. The resulting multi-scale mesh structure retains detail capture capabilities in high-resolution regions while maintaining reasonable computational complexity in low-resolution regions.
[0033] The entire mesh generation process is fully automated. The temperature difference coefficient is calculated using a sliding window algorithm, which can efficiently process large-area sample data. Sub-region division is based on spatial autocorrelation analysis of temperature distribution, automatically identifying continuous regions with similar resolution. The mesh generation algorithm uses an adaptive quadtree method, dynamically adjusting the mesh density according to resolution requirements. The boundary optimization process includes a mesh quality check to avoid malformed mesh cells. The final multi-scale mesh structure provides a spatial analysis framework for subsequent defect identification and performance evaluation.
[0034] This implementation method is characterized by combining temperature response characteristics with spatial resolution characteristics. Through temperature difference coefficient filtering, the system can focus on thermally sensitive regions, avoiding redundant analysis of non-critical areas. The construction of multi-scale grids fully considers the temperature distribution characteristics of different regions, optimizing computational efficiency while ensuring analytical accuracy. Boundary synchronization optimization technology ensures the continuity and integrity of the overall grid structure, providing a reliable spatial reference frame for subsequent processing steps. The entire workflow design balances scientific rigor and practicality, adapting to the analytical needs of different types of thermally conductive gels.
[0035] Example 3: See Figure 4 This paper describes the specific operational process of the defect identification module in a thermally conductive gel performance evaluation system. This module receives multi-scale mesh structure data from the mesh generation module and the spatiotemporal sequence of temperature distribution from the thermal field analysis module. The core task of the module is to identify the microscopic defect characteristics of the thermally conductive gel within the constructed mesh structure and quantify the evolution behavior of these defects during temperature changes. The defect identification process first performs temperature gradient analysis within each mesh cell of the multi-scale mesh structure. For each mesh cell, the temperature gradient vector of all nodes within it is calculated. The magnitude of the temperature gradient vector reflects the intensity of temperature change within that cell. By comparing the differences in gradient magnitudes between adjacent mesh cells, abnormal temperature gradient regions are identified. The determination of abnormal regions is based on a preset gradient threshold. When the gradient magnitude of a mesh cell is significantly higher than that of the surrounding cells, the region is marked as abnormal. For example, in a fine mesh cell, the gradient magnitude reaches 45 K / mm, while the gradient magnitudes of the eight surrounding cells are all in the range of 15-20 K / mm. This cell is identified as a temperature gradient abnormal region.
[0036] After identifying regions with abnormal temperature gradients, the module further analyzes the geometric characteristics of these regions. Geometric analysis includes calculating parameters such as the area, perimeter, and shape factor of the abnormal region. Simultaneously, the module assesses the continuity of temperature distribution within the abnormal region by analyzing the smoothness of temperature changes at nodes within the region. Based on the combined characteristics of geometric shape and temperature distribution continuity, the system identifies the specific type and precise location of microscopic defects. Common defect types include voids, cracks, and inclusions. Each defect type has unique geometric shapes and temperature distribution characteristics. For example, voids typically appear as approximately circular regions with significant temperature discontinuities; cracks, on the other hand, exhibit a long and narrow shape with a significant temperature gradient along the crack direction.
[0037] During defect type identification, the module employs a feature-matching-based classification algorithm. This algorithm calculates the similarity between the morphological and temperature features of each abnormal region and a predefined defect feature template. The template type with the highest similarity is identified as the defect type for that abnormal region. All identified defect features, their types, and location information are recorded in the defect feature database.
[0038] After obtaining defect feature information, the module begins calculating defect evolution indices. This process requires analyzing the morphological changes of each defect feature during the stable temperature and varying temperature phases. For each identified defect, the module calculates its area percentage in both phases. The area percentage is calculated by dividing the area of the mesh cells occupied by the defect by the total mesh area of the sub-region containing the defect. This ratio reflects the relative size of the defect in that local region.
[0039] The defect evolution index is calculated using the following formula:
[0040] in: Indicates the defect evolution index, The area percentage representing defects during the stable temperature phase. The percentage of area representing defects during temperature variations. These are weighting coefficients determined based on the defect type. This formula quantifies the relative evolution of defects during temperature changes. Weighting coefficients The value is set based on the degree of influence of different defect types on thermal conductivity. For example, the weight coefficient of void defects is usually higher than that of surface crack defects.
[0041] In the specific calculation process, the module independently calculates the evolution index for each identified defect feature. First, it extracts the area data of the defect at two temperature stages from the defect feature database, and then calculates... and The numerical value. Then, based on the type of defect, find the corresponding... The evolution index value of the defect is then calculated using a formula. All the evolution index values of the defects are compiled to form a defect evolution index distribution map, which reflects the behavioral characteristics of defects at different locations during temperature changes.
[0042] The entire defect identification and evolution analysis process is fully automated. Temperature gradient analysis employs a grid-based differential calculation method, accurately capturing local temperature change characteristics. The defect type identification algorithm combines morphological processing and pattern matching techniques to ensure the accuracy of classification results. During the calculation of evolution indicators, the area percentage is statistically based on precise accumulation of grid cell faces, and the weighting coefficients are assigned according to pre-defined rules regarding the importance of defect types.
[0043] The key feature of this implementation is the combination of spatial morphology analysis and temperature response analysis. Through a multi-scale grid structure providing a spatial framework, the system can accurately locate and quantify microscopic defect characteristics. The design of the defect evolution index comprehensively considers the influence of defect size changes and types, providing a comprehensive characterization of defect behavior. The entire processing flow is designed to fully consider various defect types and temperature scenarios that may be encountered in practical applications, exhibiting good versatility and adaptability.
[0044] Example 4: This describes the specific operation flow of the performance evaluation module in the thermally conductive gel performance evaluation system. This module receives defect evolution index data from the defect identification module, as well as multi-scale mesh structure and corresponding temperature distribution data from the mesh generation module. The core task of the module is to integrate defect characteristics and temperature gradient information to generate the thermal conductivity performance evaluation results of the thermally conductive gel.
[0045] The module processing first performs temperature gradient analysis within a multi-scale grid structure. For each grid cell, the temperature gradient data of all nodes within it is extracted. The temperature gradient is obtained by calculating the ratio of the temperature difference between adjacent nodes to their spatial distance. The mean temperature gradient of each grid cell is calculated by taking the arithmetic mean of the temperature gradient values of all nodes within the cell. This mean reflects the overall thermal conductivity characteristics within the cell region. For example, in a fine grid cell containing 9 temperature measurement nodes, with temperature gradient values of 42K / mm, 38K / mm, 45K / mm, 40K / mm, 43K / mm, 39K / mm, 44K / mm, 41K / mm, and 37K / mm respectively, the mean temperature gradient of this cell is (42+38+45+40+43+39+44+41+37) / 9=41K / mm.
[0046] After obtaining the average temperature gradient of each grid cell, the module performs a weighted correction process, using the defect evolution index from the defect identification module as the correction factor. Each grid cell is associated with one or more defect features, and each defect feature has a corresponding evolution index value. For a grid cell containing multiple defects, the maximum value of its evolution index is taken as the correction weight for that cell. The correction formula uses a linear weighting method: Corrected temperature gradient = Original average temperature gradient × (1 - Defect evolution index). This correction process reduces the temperature gradient value in areas with severe defects, reflecting the negative impact of defects on thermal conductivity. For example, if the original average temperature gradient of a grid cell is 35 K / mm and the evolution index of the associated defect is 0.2, then the corrected temperature gradient is 35 × (1 - 0.2) = 28 K / mm.
[0047] Table 1: Mesh Cell Temperature Gradient Correction Grid cell number Original temperature gradient mean (K / mm) Defect Evolution Indicators Corrected temperature gradient (K / mm) G-045 41.0 0.00 41.0 F-128 38.5 0.15 32.7 C-307 29.2 0.08 26.9 D-512 22.7 0.00 22.7 E-216 18.3 0.12 16.1 The corrected temperature gradient data was organized into a spatial distribution map, which maintained the basic framework of a multi-scale grid structure, with each grid cell filled with its corresponding corrected temperature gradient value. A smooth temperature gradient transition was generated at the grid boundaries using spatial interpolation techniques. The distribution map employed pseudo-color coding, with different colors representing different temperature gradient ranges, visually demonstrating the spatial variation characteristics of the thermal conductivity of the thermally conductive gel surface.
[0048] Based on the spatial distribution map of the corrected temperature gradient, the module performs an overall performance analysis. This analysis includes calculating the statistical characteristics of the temperature gradient across the entire region: maximum, minimum, average, and standard deviation. Simultaneously, it analyzes the spatial autocorrelation of the gradient distribution to detect the existence of large-scale low-conductivity regions. The spatial clustering of gradient values is also quantitatively evaluated to identify clusters of regions with abnormal heat conduction performance.
[0049] The performance evaluation results are described using a hierarchical approach, dividing the thermal conductivity into multiple levels based on the numerical range of the corrected temperature gradient. Each level corresponds to a gradient value range, with the threshold value set according to material properties and application requirements. The proportion of mesh cells for each performance level is statistically analyzed to form a performance distribution overview. The final evaluation report includes a gradient space distribution map, statistical characteristic data, and performance level distribution data, comprehensively describing the thermal conductivity characteristics of the thermally conductive gel.
[0050] The entire processing flow is implemented using automated algorithms. Temperature gradient calculation is based on differential operations on grid node data, ensuring computational efficiency. The weighted correction process employs matrix operations, processing all grid cell data in one go. Spatial distribution maps are generated using standard grid data visualization techniques. Performance analysis algorithms include standard statistical analysis methods and spatial pattern recognition techniques. The evaluation report automatically generates a standard format output document, including text descriptions and charts. A key feature of this implementation is the introduction of defect impact quantification into heat conduction performance evaluation. Temperature gradient correction using defect evolution indices more accurately reflects actual heat conduction performance. The application of a multi-scale grid structure ensures both accuracy and efficiency in the analysis. Spatial distribution analysis provides intuitive performance visualization, and statistical characteristics quantify the overall performance level. The entire workflow design achieves an effective transformation from microscopic defect characteristics to macroscopic performance evaluation.
[0051] Example 5: See Figure 5 This document describes the final evaluation stage processing flow of the performance evaluation module in a thermally conductive gel performance evaluation system. After correcting the temperature gradient, this module generates thermal conductivity performance evaluation results based on the spatial distribution of the corrected temperature gradient. The core task of this stage is to transform quantified temperature gradient data into intuitive performance evaluation conclusions. The module processing begins with the grading of the corrected temperature gradient. Grading is based on a preset gradient threshold range, which is set according to the material properties of the thermally conductive gel and the requirements of the application scenario. All grid cells in the entire multi-scale grid structure are divided into different performance levels according to the magnitude of their corrected temperature gradient values. For example, cells with gradient values above a certain upper threshold are classified as high conductivity performance level, cells with gradient values below a certain lower threshold are classified as low conductivity performance level, and cells in the middle range belong to medium conductivity performance level. This grading method establishes a direct correspondence between temperature gradient values and thermal conductivity performance levels.
[0052] After grading, the module performs a grid cell count, covering all grid cells in the entire analysis area, calculating the number of cells contained in each performance level. This number is converted into a region area percentage, that is, the proportion of the total area of the grid cells corresponding to each performance level to the total area of the entire analysis area. The percentage calculation takes into account the area differences of grid cells of different sizes in the multi-scale grid structure to ensure the accuracy of the statistical results. For example, when analyzing a thermally conductive gel sample, the high conductivity performance level area accounts for 35% of the total area, the medium conductivity performance level area accounts for 50%, and the low conductivity performance level area accounts for 15%.
[0053] After obtaining the percentage distribution of each performance level, the module performs a matching analysis with preset standards. The thermal conductivity standard ranges are predefined based on industry specifications or specific application requirements, clearly defining the ideal percentage range for different performance levels. The matching process compares the actual statistically obtained percentage distribution with these standard ranges. The comparative analysis includes calculating the deviation of the actual percentage from the standard range boundaries and assessing whether the distribution of each performance level region meets the expected requirements. For example, the standard range might require that the percentage of high conductivity regions be no less than 30%, and if the actual statistical value is 35%, it is considered to meet the standard. The matching analysis further examines the spatial distribution characteristics of each performance level region. The module detects whether there are abnormal clusters of low conductivity regions and assesses the spatial continuity of high conductivity regions. The distribution characteristic analysis uses a spatial autocorrelation algorithm to identify cluster distribution patterns of regions with similar conductivity. The spatial distribution characteristics are compared with the requirements for distribution uniformity in the standard ranges to assess whether the actual distribution meets expectations.
[0054] Based on the above analysis results, the module generates the final thermal conductivity performance evaluation conclusion, which combines qualitative description and quantitative data. The qualitative description summarizes the overall performance, such as "overall thermal conductivity meets requirements" or "there are localized areas with insufficient conductivity." The quantitative data section lists the specific percentage values for each performance level and their matching with the standard range. The conclusion also includes a description of the problem areas, pointing out weak areas of conductivity that require special attention.
[0055] The evaluation results are output in a structured report format. The first part of the report displays a spatial distribution map of the corrected temperature gradient, using pseudo-color rendering of different performance level regions. The second part presents a detailed statistical table, listing the percentage of each performance level and the degree of deviation from the standard. The third part is the textual evaluation conclusion, including a comprehensive performance evaluation and a description of problem areas. The final part of the report generates performance rating labels as needed, such as concise evaluation labels like "Grade A" and "Grade B".
[0056] The entire evaluation process employs a fully automated workflow. Hierarchical processing is completed automatically based on preset thresholds, requiring no manual intervention. Percentage statistics are achieved by traversing the grid database, automatically calculating the area of grids of different sizes. Matching analysis utilizes rule engine technology, automatically performing standard comparisons according to preset logic. Spatial distribution analysis uses Geographic Information System (GIS) algorithms to automatically identify regional cluster characteristics. The report generation module integrates all analysis results and automatically formats and outputs standard-format documents. The key feature of this implementation is that it transforms complex temperature gradient data into intuitive performance evaluations. Hierarchical processing simplifies data analysis dimensions, percentage statistics quantify performance distribution characteristics, and standard matching provides an objective evaluation benchmark. Spatial distribution analysis reveals the clustering patterns of performance issues, and the structured report presents comprehensive evaluation conclusions.
[0057] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.
[0058] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended text and its equivalents.
Claims
1. A computer vision-based system for evaluating the performance of thermally conductive gels, characterized in that, It includes an image acquisition module, a thermal field analysis module, a mesh generation module, a defect identification module, and a performance evaluation module; among which, The image acquisition module is used to acquire infrared thermal imaging image sequences of the thermally conductive gel under different temperature conditions; The thermal field analysis module is used to extract the spatiotemporal sequence of temperature distribution from the infrared thermal imaging image sequence; and to divide the stable temperature stage and the changing temperature stage according to the temperature change trend of each pixel in the spatiotemporal sequence of temperature distribution. The mesh generation module is used to select the region to be analyzed based on the degree of difference in temperature distribution between the stable temperature stage and the changing temperature stage; within the region to be analyzed, a multi-scale mesh structure is generated based on the resolution difference in temperature distribution. The defect identification module is used to identify the microscopic defect features of the thermally conductive gel in a multi-scale grid structure; and to calculate the defect evolution index based on the morphological differences of the microscopic defect features in the stable temperature stage and the temperature variation stage. The performance evaluation module is used to combine the defect evolution index with the temperature gradient distribution of the multi-scale grid structure to generate the thermal conductivity performance evaluation results of the thermally conductive gel.
2. The computer vision-based thermal conductive gel performance evaluation system according to claim 1, characterized in that, The thermal field analysis module extracts the spatiotemporal sequence of temperature distribution from the infrared thermal imaging image sequence, including: The infrared thermal imaging image sequence was standardized to obtain normalized temperature data. Based on a pre-defined convolutional neural network, spatial temperature field features and temporal temperature change features are extracted from normalized temperature data to generate a spatiotemporal sequence of temperature distribution.
3. The computer vision-based thermal conductive gel performance evaluation system according to claim 2, characterized in that, The thermal field analysis module divides the temperature field into a stable temperature stage and a changing temperature stage based on the temperature change trend of each pixel in the spatiotemporal sequence of temperature distribution, including: For any pixel, obtain its temperature value as a function of time, and calculate the first difference of the temperature sequence. Select the moments with positive first-order difference values as the heating moments; sort the heating moments in ascending order according to their corresponding first-order difference values, and calculate the second-order difference values of the sorted sequence. The time of temperature rise corresponding to the maximum value of the second-order difference is taken as the critical time. The time period before the critical time is the stable temperature stage, and the time period after the critical time is the changing temperature stage.
4. The computer vision-based thermal conductive gel performance evaluation system according to claim 3, characterized in that, The mesh generation module filters out the regions to be analyzed based on the degree of difference in temperature distribution between the stable temperature phase and the changing temperature phase, including: Calculate the difference between the average temperature during the stable temperature phase and the average temperature during the changing temperature phase within the region to be analyzed; The temperature difference coefficient is obtained by normalizing the difference; if the temperature difference coefficient is greater than the preset difference threshold, the region is marked as the region to be analyzed.
5. The computer vision-based thermal conductive gel performance evaluation system according to claim 4, characterized in that, The mesh generation module generates a multi-scale mesh structure within the region to be analyzed based on the resolution differences in temperature distribution, including: The region to be analyzed is divided into multiple sub-regions; the grid cell size and grid density are set according to the temperature distribution resolution of each sub-region. For high-resolution sub-regions, generate fine meshes; for low-resolution sub-regions, generate sparse meshes. By performing boundary synchronization optimization on the meshes of each sub-region, a multi-scale mesh structure is obtained.
6. The computer vision-based thermal conductive gel performance evaluation system according to claim 5, characterized in that, The defect identification module identifies microscopic defect features of the thermally conductive gel within a multi-scale grid structure, including: Within each grid cell of the multi-scale grid structure, extract the temperature gradient anomaly region; Based on the geometric morphology and temperature distribution continuity of the temperature gradient anomaly region, the type and location of micro-defect features are identified.
7. The computer vision-based thermal conductive gel performance evaluation system according to claim 6, characterized in that, The defect identification module calculates defect evolution indices based on the morphological differences of microscopic defect characteristics in the stable temperature stage and the changing temperature stage, including: Obtain the area percentage of micro-defect features during the stable temperature stage and the area percentage during the changing temperature stage. Calculate the ratio of the area proportions of the two stages, and combine it with the defect type weighting coefficient to generate a defect evolution index.
8. The computer vision-based thermal conductive gel performance evaluation system according to claim 7, characterized in that, The performance evaluation module combines defect evolution indices with the temperature gradient distribution of the multi-scale mesh structure to generate thermal conductivity performance evaluation results for the thermally conductive gel, including: In a multi-scale grid structure, calculate the mean temperature gradient of each grid cell; The corrected temperature gradient is obtained by weighting and correcting the mean temperature gradient based on the defect evolution index. Based on the spatial distribution of the corrected temperature gradient, thermal conductivity performance evaluation results are generated.
9. The computer vision-based thermal conductive gel performance evaluation system according to claim 8, characterized in that, The performance evaluation module generates heat conduction performance evaluation results based on the spatial distribution of the corrected temperature gradient, including: The corrected temperature gradient is divided into multiple levels according to its magnitude; the proportion of grid cells corresponding to each level is calculated. Based on the degree of matching between the proportion distribution and the preset thermal conductivity standard range, the thermal conductivity performance evaluation results are output.
10. A computer vision-based method for evaluating the performance of thermally conductive gels, characterized in that, It includes all modules and method flows of a computer vision-based thermal conductive gel performance evaluation system as described in any one of claims 1 to 9.