Automatic processing system and method for graphene cooling fins

Abstract

The invention relates to the technical field of graphene, and more particularly discloses an automatic processing system and method for a graphene cooling fin, and the method comprises the steps: obtaining the specification and shape parameters of the cooling fin, a monitoring video, a slicing pressure value and other information, extracting a slicing process change feature vector, a cooling fin parameter correlation feature vector and a pressure feature vector, and obtaining a graphene cooling fin; the feature vectors are classified through a classifier, and whether the pressure value at the current time point needs to be increased or decreased is judged. In this way, automatic machining of the graphene cooling fins is achieved, and the machining applicability and quality are improved.

Description

Technical Field

[0001] The present application relates to the field of graphene technology, and more specifically, to an automatic processing system and method for graphene heat sinks. Background Art

[0002] Graphene is a two-dimensional carbon nanomaterial composed of carbon atoms in a hexagonal honeycomb lattice with sp2 hybrid orbitals. Due to its excellent performance, it is widely made into heat sinks to dissipate heat from electronic components. However, the structure of traditional heat sink processing equipment is relatively fixed and cannot adapt to heat sinks of different specifications, thereby reducing the applicability of the device. Moreover, if the heat sink is trimmed manually, the heat sink may be irreversibly damaged if not paid attention to, which in turn affects the processing quality of the device.

[0003] Therefore, an automatic processing system and method for graphene heat sinks are desired. Summary of the Invention

[0004] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide an automatic processing system and method for graphene heat sinks, which adjusts the slicing pressure according to the characteristics of the slicing process and pressure changes, thereby improving the processing quality and applicability.

[0005] Accordingly, according to one aspect of the present application, there is provided an automatic processing system for graphene heat sinks, comprising:

[0006] a slicing processing data acquisition module, configured to obtain required heat sink specifications and shape parameters, monitoring video of the heat sink slicing process during a predetermined time period, and slicing pressure values ​​at multiple predetermined time points during the predetermined time period, wherein the required heat sink specifications and shape parameters include size and hole location;

[0007] a slicing process data processing module, configured to extract a slicing process variation feature vector, a multi-scale heat sink parameter association feature vector, and a pressure feature vector from monitoring video of the heat sink slicing process during the predetermined time period, the required heat sink specifications and shape parameters, and the slicing pressure values ​​at the plurality of predetermined time points;

[0008] a slicing processing data fusion module, configured to fuse the slicing process variation feature vector, the multi-scale heat sink parameter association feature vector, and the pressure feature vector to obtain an optimized slicing result analysis feature vector;

[0009] The slicing processing data analysis module is used to pass the optimized slicing result analysis feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the pressure value at the current time point should be increased or decreased.

[0010] According to another aspect of the present application, there is provided a method for automatically processing a graphene heat sink, comprising:

[0011] Obtaining required heat sink specifications and shape parameters, monitoring video of the heat sink slicing process within a predetermined time period, and slicing pressure values ​​at multiple predetermined time points within the predetermined time period, wherein the required heat sink specifications and shape parameters include size and hole location;

[0012] Extracting a slicing process variation feature vector, a multi-scale fin parameter association feature vector, and a pressure feature vector from the monitoring video of the fin slicing process during the predetermined time period, the required fin specifications and shape parameters, and the slicing pressure values ​​at the plurality of predetermined time points;

[0013] fusing the slicing process variation feature vector, the multi-scale heat sink parameter association feature vector, and the pressure feature vector to obtain an optimized slicing result analysis feature vector;

[0014] The feature vector of the optimized slicing result analysis is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the pressure value at the current time point should be increased or decreased.

[0015] Compared to existing technologies, the automated graphene heat sink processing system and method provided in this application obtains information such as heat sink specifications and shape parameters, monitoring video, and slicing pressure values. It then extracts slicing process variation feature vectors, heat sink parameter correlation feature vectors, and pressure feature vectors, and fuses these features to generate an optimized slicing result analysis feature vector. A classifier then classifies these feature vectors to determine whether the pressure value at the current point in time should be increased or decreased. This enables automated processing of graphene heat sinks, improving both the applicability and quality of the process. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0017] Figure 1 Schematic diagram of a block diagram of an automatic processing system for graphene heat sinks according to an embodiment of the present application.

[0018] Figure 2 Schematic diagram of a block diagram of a slicing processing data processing module in an automatic processing system for graphene heat sinks according to an embodiment of the present application.

[0019] Figure 3Schematic diagram of a block diagram of a monitoring video processing unit in an automatic processing system for graphene heat sinks according to an embodiment of the present application.

[0020] Figure 4 Schematic diagram of a block diagram of a heat sink parameter processing unit in an automatic processing system of a graphene heat sink according to an embodiment of the present application.

[0021] Figure 5 Flowchart of an automatic processing method of a graphene heat sink according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] Various exemplary embodiments, features, and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise indicated.

[0023] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0024] In addition, numerous specific details are provided in the following detailed description to better illustrate the present application. Those skilled in the art will appreciate that the present application can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art are not described in detail in order to highlight the main purpose of the present application.

[0025] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. Throughout the description of this application, "plurality" means two or more, unless otherwise specifically defined.

[0026] Figure 1 FIG2 is a block diagram of an automatic processing system for graphene heat sinks according to an embodiment of the present application. Figure 1As shown, the automatic processing system 100 of the graphene heat sink according to the embodiment of the present application includes: a slicing processing data acquisition module 110, which is used to obtain the required heat sink specifications and shape parameters, a monitoring video of the heat sink slicing processing process in a predetermined time period, and the slicing pressure values ​​at multiple predetermined time points in the predetermined time period, wherein the required heat sink specifications and shape parameters include size and hole position; a slicing processing data processing module 120, which is used to extract a slicing process change feature vector, a multi-scale heat sink parameter association feature vector and a pressure feature vector from the monitoring video of the heat sink slicing processing process in the predetermined time period, the required heat sink specifications and shape parameters, and the slicing pressure values ​​at the multiple predetermined time points; a slicing processing data fusion module 130, which is used to fuse the slicing process change feature vector, the multi-scale heat sink parameter association feature vector and the pressure feature vector to obtain an optimized slicing result analysis feature vector; a slicing processing data analysis module 140, which is used to pass the optimized slicing result analysis feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the pressure value at the current time point should increase or decrease.

[0027] In this embodiment of the present application, the slicing processing data acquisition module 110 is used to obtain the required heat sink specifications and shape parameters, surveillance video of the heat sink slicing process over a predetermined time period, and slicing pressure values ​​at multiple predetermined time points within the predetermined time period. The required heat sink specifications and shape parameters include dimensions and hole locations. It should be understood that information such as the heat sink specifications and shape parameters, size, and hole location are crucial for determining processing goals and requirements. Dimensional parameters determine the size and adaptability of the heat sink, while the accuracy of hole location is directly related to the heat dissipation performance and installation requirements of the heat sink. By obtaining these specifications and shape parameters, it is possible to ensure that the processed heat sink meets the expected size and shape requirements to meet customer needs. Recording and analyzing surveillance video and slicing pressure values ​​is crucial for controlling and optimizing the slicing process. Surveillance video provides real-time slicing process information, including cutting speed and tool position. Video analysis can identify changes and anomalies during the slicing process. Slicing pressure values ​​reflect the force and pressure distribution during the cutting process, significantly impacting cutting quality and processing results. By recording and analyzing surveillance video and slicing pressure values, feature vectors can be extracted and classified, enabling automatic adjustment and optimization of the slicing process. For example, pressure changes can be used to determine whether tool wear or material deformation is occurring during the cutting process, allowing for timely adjustment of slicing pressure to ensure processing quality and improve the adaptability of the heat sink.

[0028] In this embodiment of the present application, the slicing process data processing module 120 is configured to extract a slicing process variation feature vector, a multi-scale heat sink parameter correlation feature vector, and a pressure feature vector from surveillance video of the heat sink slicing process during a predetermined time period, the desired heat sink specifications and shape parameters, and the slicing pressure values ​​at multiple predetermined time points. It should be understood that by analyzing the surveillance video, variation features of the slicing process, such as cutting speed, tool position, and tool trajectory, can be extracted. These features can be used to describe dynamic changes during the slicing process, such as cutting speed and tool position accuracy, and thus determine whether the slicing process is stable or anomalies exist. Extracting these variation feature vectors facilitates monitoring and control of the slicing process, enabling timely identification of problems and appropriate adjustments to improve processing quality and efficiency. Heat sink specifications and shape parameters include information such as size and hole location, and these parameters are correlated with the slicing process feature vectors. By correlating these heat sink parameters with the slicing process feature vectors, the extent to which different heat sink parameters affect the slicing process can be understood. Slicing process parameter settings can then be adjusted based on the desired heat sink specifications and shape parameters to meet the processing requirements of heat sinks of different specifications. Cutting pressure is a critical parameter during the slicing process, directly impacting the contact force between the tool and the workpiece and the quality of the cut. By recording slicing pressure values ​​at multiple predetermined time points and extracting the associated pressure eigenvectors, we can understand the pressure distribution and pressure trends during the slicing process. These eigenvectors can be used to determine if abnormal pressure conditions, such as excessively high or low pressure, are present during the slicing process, allowing for timely adjustment of slicing parameters to ensure machining quality and tool life.

[0029] Specifically, in one embodiment of the present application, Figure 2 The figure shows a block diagram of a slice processing data processing module in an automatic processing system for graphene heat sinks according to an embodiment of the present application. Figure 2 As shown, in the above-mentioned automatic processing system 100 of the graphene heat sink, the slicing processing data processing module 120 includes: a monitoring video processing unit 121, which is used to extract key frames from the monitoring video of the heat sink slicing processing process in the predetermined time period and then encode them to obtain the slicing process change feature vector; a heat sink parameter processing unit 122, which is used to segment the required heat sink specifications and shape parameters and then encode them to obtain the multi-scale heat sink parameter association feature vector; a slicing pressure processing unit 123, which is used to arrange the slicing pressure values ​​of the multiple predetermined time points according to the time dimension as a slicing pressure value input vector and then pass it through a third convolutional neural network model as a filter to obtain the pressure feature vector.

[0030] Accordingly, in a specific example of the present application, the monitoring video processing unit 121 is configured to extract key frames from the monitoring video of the heat sink slicing process during the predetermined time period and then encode them to obtain a slicing process change feature vector. It should be understood that during the slicing process, the monitoring video is typically recorded in the form of continuous frame images. However, for analyzing and processing the slicing process, not all frame images need to be used, as some frame images may contain repeated or redundant information. Therefore, by extracting key frames, i.e., frames representing the most significant changes during the slicing process, the amount of data can be reduced while retaining critical information. Key frame extraction can be achieved using various methods, such as based on image difference, motion information, or other specific rules. Once the key frames are extracted, the next step is to encode them to obtain a slicing process change feature vector. The purpose of encoding is to convert image data into a more compact and expressive representation for subsequent analysis and processing. Common encoding methods include image feature extraction and image compression algorithms. Image feature extraction can represent image content by calculating features such as color histogram, gradient, and texture. Image compression algorithms, such as JPEG and PNG, can reduce image data size while preserving important information. By encoding keyframes, we can generate feature vectors that describe key changes and characteristics during the slicing process. These feature vectors can be used for subsequent analysis, classification, model training, and other tasks to automatically adjust and optimize the slicing process.

[0031] further, Figure 3 FIG2 is a block diagram of a monitoring video processing unit in an automatic processing system for graphene heat sinks according to an embodiment of the present application. Figure 3 As shown, in the slicing processing data processing module 120 of the above-mentioned automatic processing system 100 of the graphene heat sink, the monitoring video processing unit 121 includes: a key frame extraction sub-unit 1211, which is used to extract multiple slicing process monitoring key frames from the monitoring video of the heat sink slicing processing process in the predetermined time period; a deep and shallow fusion sub-unit 1212, which is used to pass the multiple slicing process monitoring key frames through a first convolutional neural network model including a deep and shallow fusion module to obtain multiple slicing process monitoring feature matrices; a three-dimensional convolution kernel sub-unit 1213, which is used to aggregate the multiple slicing process monitoring feature matrices along the time dimension into a three-dimensional slicing process feature tensor and then obtain a slicing process change feature vector through a second convolutional neural network model using a three-dimensional convolution kernel.

[0032] Specifically, the keyframe extraction subunit 1211 is configured to extract multiple slicing process monitoring keyframes from the surveillance video of the heat sink slicing process during the predetermined time period. It should be understood that the surveillance video of the slicing process records the dynamic changes throughout the entire process, including key information such as tool position, cutting speed, and cutting pressure. However, the slicing process is dynamic and continuous, and the slicing process at different points in time may vary and change. Therefore, extracting only a single keyframe may not fully capture the entire slicing process. By extracting multiple slicing process monitoring keyframes, the variability and uncertainty of the slicing process can be comprehensively considered. These keyframes can temporally and spatially cover the key stages and important changes of the entire slicing process, providing more comprehensive and accurate information. By analyzing the changing trends and correlations between these keyframes, the characteristics and patterns of the slicing process can be better understood, allowing for further optimization of slicing parameters and processing results. Furthermore, extracting multiple slicing process monitoring keyframes can address anomalies and issues during the slicing process. Extracting only a single keyframe may overlook anomalies during the slicing process, such as sudden changes in cutting speed or abnormal cutting pressure. Extracting multiple key frames can better capture these abnormal situations and take timely measures to adjust and correct them.

[0033] Specifically, the deep-shallow fusion subunit 1212 is configured to pass the multiple slicing process monitoring keyframes through a first convolutional neural network model comprising a deep-shallow fusion module to obtain multiple slicing process monitoring feature matrices. It should be understood that the deep-shallow fusion module is a method for fusing multi-level features, effectively combining shallow and deep features. In slicing process monitoring, different keyframes may contain information at different levels, such as color, texture, and shape. Shallow features are generally easier to extract, while deep features are more abstract and have greater semantic representation capabilities. Therefore, the deep-shallow fusion module can comprehensively utilize multi-level feature information to improve feature representation capabilities. Convolutional neural networks are a deep learning model widely used in image processing tasks. By inputting multiple slicing process monitoring keyframes into the convolutional neural network model, its powerful feature extraction and representation capabilities can be leveraged to learn more discriminative feature representations from the image. After the multiple slicing process monitoring keyframes are input into the convolutional neural network model, the deep-shallow fusion module can fuse the features at different levels. This fusion can be performed in parallel, serial, or other forms, depending on the design of the deep-shallow fusion module. The fused features can be further processed, such as through dimensionality reduction using fully connected layers or using other neural network structures for tasks like classification and clustering. Ultimately, the first convolutional neural network model, including the deep and shallow fusion modules, generates multiple slicing process monitoring feature matrices, each representing a feature extracted from a keyframe. These feature matrices can be used for subsequent slicing process analysis, anomaly detection, and classification, enabling automated processing and optimization of the slicing process.

[0034] Correspondingly, the deep-shallow fusion subunit includes: a shallow feature extraction secondary subunit for extracting a shallow feature map from the i-th layer of the first convolutional neural network model, where j is greater than or equal to 1 and less than or equal to 6; a deep feature extraction secondary subunit for extracting a deep feature map from the j-th layer of the first convolutional neural network model, where the ratio between the j-th layer and the i-th layer is greater than or equal to 5 and less than or equal to 10; a feature map fusion secondary subunit for using the deep-shallow feature fusion module of the first convolutional neural network model to fuse the shallow feature map and the deep feature map to obtain a fused feature map; a feature map dimensionality reduction secondary subunit for performing global pooling on the fused feature map along the channel dimension to obtain the multiple slicing process monitoring feature matrices.

[0035] Specifically, the three-dimensional convolution kernel subunit 1213 is configured to aggregate the multiple slicing process monitoring feature matrices along the time dimension into a three-dimensional slicing process feature tensor, and then use the second convolutional neural network model with the three-dimensional convolution kernel to obtain a slicing process variation feature vector. It should be understood that the slicing process is a time-series process, and the slicing process features at different time points may exhibit temporal variations and correlations. Therefore, aggregating the multiple slicing process monitoring feature matrices along the time dimension into a three-dimensional slicing process feature tensor can incorporate temporal information into the feature representation. This has the advantage of better capturing temporal variations in the slicing process, such as acceleration or deceleration of the cutting speed and changes in cutting pressure. Using the second convolutional neural network model with the three-dimensional convolution kernel, convolution operations can be performed on the slicing process feature tensor in three dimensions (time, height, and width). The three-dimensional convolution operation can simultaneously consider the temporal and spatial feature correlations, thereby better capturing local patterns and temporal variations in the slicing process. This enables the model to have stronger perception capabilities and better understand the dynamic characteristics of the slicing process. Through the three-dimensional convolution operation, the slicing process feature tensor can be mapped into a slicing process variation feature vector. This feature vector can contain key change information during the slicing process, such as the changing trend of cutting speed and the peak position of cutting pressure.

[0036] Correspondingly, the three-dimensional convolution kernel subunit includes: an encoding secondary subunit, which is used to use the second convolutional neural network model to perform three-dimensional convolution encoding on the three-dimensional elevator operation process feature tensor to obtain a slicing process change feature map; a dimensionality reduction secondary subunit, which is used to perform global mean pooling on each feature matrix along the channel dimension of the slicing process change feature map to obtain the slicing process change feature vector.

[0037] Furthermore, the secondary encoding subunit is configured to use the second convolutional neural network model to perform three-dimensional convolution encoding on the three-dimensional elevator operation process feature tensor to obtain a slicing process variation feature map. It should be understood that the slicing process is typically a continuous process with a certain temporal order and continuity. By aggregating multiple slicing process monitoring feature matrices along the time dimension into a three-dimensional slicing process feature tensor, temporal information can be incorporated into the feature representation. This approach can better capture temporal variations in the slicing process, such as changes in cutting speed and fluctuations in cutting pressure. Furthermore, features in the slicing process also exhibit certain spatial correlations. Using the second convolutional neural network model with a three-dimensional convolution kernel, convolution operations can be performed on the slicing process feature tensor in the time, height, and width directions. This simultaneously considers temporal and spatial feature correlations, thereby better capturing local patterns and spatiotemporal relationships in the slicing process. The three-dimensional convolution operation effectively extracts spatial and temporal information from the slicing process feature tensor and maps it into a slicing process variation feature vector. This feature vector can contain key variation information in the slicing process, such as the changing trend of cutting speed and the peak location of cutting pressure. Through such feature representation, the dynamic characteristics of the slicing process can be better understood and analyzed, providing more accurate and meaningful feature representation for subsequent slicing process analysis and application.

[0038] Furthermore, the dimensionality reduction secondary subunit is configured to perform global mean pooling on each feature matrix along the channel dimension of the slicing process variation feature map to obtain the slicing process variation feature vector. It should be understood that in the slicing process variation feature map, each feature matrix corresponds to a different feature channel, and each channel can capture different slicing process variation patterns or characteristics. However, sometimes not every channel's features are useful for the final task, and too many channels may increase computational complexity and the risk of model overfitting. Global mean pooling averages the features of each feature matrix along the channel dimension to obtain a global feature vector. Global mean pooling averages the feature values ​​in each channel, fusing the information from each channel's features to obtain a more comprehensive and global feature representation. The benefit of global mean pooling is that it reduces the feature dimensionality, converting high-dimensional feature matrices into low-dimensional feature vectors. This reduces the number of model parameters and computational complexity, improving model efficiency and generalization. Furthermore, global mean pooling can help reduce the risk of model overfitting by averaging features across each channel, reducing feature redundancy and the impact of noise.

[0039] Accordingly, in a specific example of the present application, the heat sink parameter processing unit 122 is configured to segment the required heat sink specifications and shape parameters and then encode them to obtain the multi-scale heat sink parameter association feature vector. It should be understood that heat sink specifications and shape parameters typically include information such as the heat sink's size, material, and surface shape. These parameters have a significant impact on the heat dissipation effect and performance of the heat sink, but directly using the original parameter values ​​may not fully express the relationships and characteristics between the parameters. By segmenting the parameters, that is, breaking the parameters into smaller semantic units, the information and characteristics within the parameters can be better captured. After segmentation, each segmented word can be represented as an encoding vector. The encoding vector can be a fixed-length numerical vector or a low-dimensional continuous vector obtained through techniques such as word embedding. The encoding vector converts the segmented parameters into a machine-processable numerical representation, which facilitates subsequent feature extraction and analysis. By encoding the segmented heat sink parameters, a multi-scale heat sink parameter association feature vector can be obtained. This feature vector can contain feature information at different scales and levels, such as the overall shape of the heat sink, the proportional relationship between dimensions, and the relative importance of various parameters. This feature representation allows for a better understanding and analysis of the correlations between heat sink parameters, providing more accurate and meaningful feature representations for subsequent heat sink design and optimization. Therefore, after tokenizing the required heat sink specifications and shape parameters, encoding them can convert them into manageable numerical representations. Multi-scale heat sink parameter correlation feature vectors capture the correlations and importance between parameters, extracting a more expressive feature representation.

[0040] further, Figure 4 FIG2 is a block diagram of a heat sink parameter processing unit in an automatic processing system of a graphene heat sink according to an embodiment of the present application. Figure 4 As shown, in the slicing processing data processing module 120 of the above-mentioned automatic processing system 100 of the graphene heat sink, the heat sink parameter processing unit 122 includes: a heat sink parameter segmentation subunit 1221, which is used to segment the required heat sink specifications and shape parameters and then pass them through a first context encoder based on a converter to obtain a plurality of heat sink parameter feature vectors; a heat sink one-dimensional arrangement subunit 1222, which is used to arrange the plurality of heat sink parameter feature vectors into a one-dimensional heat sink parameter input vector and then pass it through a multi-scale neighborhood feature extraction module to obtain the multi-scale heat sink parameter association feature vector.

[0041] Specifically, the heat sink parameter segmentation subunit 1221 is configured to segment the required heat sink specifications and shape parameters and then pass them through a transformer-based first context encoder to obtain multiple heat sink parameter feature vectors. It should be understood that heat sink specifications and shape parameters typically include information such as the size, material, and surface shape of the heat sink, and these parameters often have complex relationships. By breaking the parameters into smaller semantic units through segmentation, the information and features within the parameters can be better captured. The transformer-based context encoder can model global contextual dependencies within the parameter sequence and perform context-aware encoding of the representation of each parameter. Transformer-based first context encoders, such as BERT and GPT models, can learn semantic relationships and contextual information within the parameter sequence. These models, through pre-training and fine-tuning, map the parameter sequence into a high-dimensional feature space, where each parameter is represented as a feature vector. These feature vectors contain the semantic information and contextual associations of the parameters, better representing the relationships and features between the parameters. Using the transformer-based first context encoder, multiple heat sink parameter feature vectors can be obtained. Each feature vector corresponds to a context-encoded parameter, capturing the different semantics and contextual information between the parameters. Such feature vectors can be used for subsequent feature extraction, clustering, classification, and other tasks, thereby improving the accuracy and effectiveness of heat sink design and optimization.

[0042] Specifically, the heat sink one-dimensional arrangement subunit 1222 is configured to arrange the multiple heat sink parameter feature vectors into a one-dimensional heat sink parameter input vector and then pass it through the multi-scale neighborhood feature extraction module to obtain the multi-scale heat sink parameter association feature vector. It should be understood that the specifications and shape parameters of a heat sink typically contain information from multiple dimensions, such as size, material, and surface shape. Each parameter feature vector can represent a feature in one dimension. Arranging these feature vectors into a one-dimensional heat sink parameter input vector combines features from multiple dimensions to form a more comprehensive feature representation. The multi-scale neighborhood feature extraction module enables feature extraction of heat sink parameters at different scales. This means that feature relationships at different scales can be considered, such as the relative size and proportionality between parameters. Multi-scale neighborhood feature extraction captures feature associations at different scales, enriching the diversity and expressiveness of the feature representation. After arranging the heat sink parameter feature vectors into a one-dimensional heat sink parameter input vector, the multi-scale neighborhood feature extraction module performs multi-scale convolution, pooling, attention, and other operations on this vector to extract feature information at different scales. The resulting multi-scale heat sink parameter association feature vector can include feature representations at different scales and levels, more comprehensively reflecting the correlation and importance between parameters. Therefore, by arranging multiple heat sink parameter feature vectors into a one-dimensional heat sink parameter input vector, the multi-scale neighborhood feature extraction module can more comprehensively capture feature information at different scales and obtain a multi-scale heat sink parameter association feature vector, improving the expressiveness of features and the correlation between extracted parameters.

[0043] Furthermore, the heat sink one-dimensional arrangement subunit includes: a first-scale heat sink encoding secondary subunit, used to use the first convolution layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolution encoding on the one-dimensional heat sink parameter input vector with a one-dimensional convolution kernel having a first scale to obtain a first-scale heat sink parameter feature vector, wherein the first convolution layer has a first one-dimensional convolution kernel of a first length; a second-scale heat sink encoding secondary subunit, used to use the second convolution layer of the multi-scale neighborhood feature extraction module to perform one-dimensional convolution encoding on the one-dimensional heat sink parameter input vector with a one-dimensional convolution kernel having a second scale to obtain a second-scale heat sink parameter feature vector, wherein the second convolution layer has a second one-dimensional convolution kernel of a second length, and the first length is different from the second length; a multi-scale heat sink cascade secondary subunit, used to cascade the first-scale heat sink parameter feature vector and the second-scale heat sink parameter feature vector to obtain the multi-scale heat sink parameter association feature vector.

[0044] Accordingly, in a specific example of the present application, the slice pressure processing unit 123 is configured to arrange the slice pressure values ​​at the multiple predetermined time points according to the time dimension into a slice pressure value input vector, and then pass the vector through a third convolutional neural network model as a filter to obtain the pressure feature vector. It should be understood that arranging the slice pressure values ​​according to the time dimension into an input vector preserves the order and continuity of the time series. Time is generally an important factor in pressure data, as changes in pressure values ​​over time may contain useful information. By preserving the time dimension, the model can learn temporal patterns and trends, better capturing the dynamic characteristics of pressure data. The convolutional neural network model can effectively process data with local correlations, such as images and time series. In this case, the slice pressure values ​​can be viewed as a time series data. By using the convolutional neural network model as a filter, local features can be extracted in the time dimension. The third convolutional layer typically has a larger receptive field and higher-level feature extraction capabilities, making it suitable for extracting more abstract and advanced features from pressure data. The convolutional neural network model can perform operations such as convolution and pooling on the slice pressure value input vector to extract important features from the pressure data. These features can be pressure amplitude, rate of change, periodicity, trend, etc. The extracted pressure feature vector can capture the pressure change pattern and the correlation between pressures at different time points, thereby better describing the characteristics of pressure data.

[0045] Specifically, the slice pressure processing unit includes: a slice pressure convolution subunit, which is used to perform convolution processing on the input data based on the convolution kernel to generate a convolution feature map; a slice pressure pooling subunit, which is used to perform mean pooling processing on each feature matrix along the channel dimension of the convolution feature map to obtain a pooled feature map; a slice pressure activation subunit, which is used to perform nonlinear activation on the feature values ​​at each position in the pooled feature map to generate an activation feature map; wherein, the input of the first convolutional neural network model is the slice pressure value input vector, the input from the second layer to the last layer of the first convolutional neural network model is the output of the previous layer, and the output of the last layer of the first convolutional neural network model is the pressure feature vector.

[0046] In this embodiment of the present application, the slicing process data fusion module 130 is configured to fuse the slicing process variation feature vector, the multi-scale heat sink parameter association feature vector, and the pressure feature vector to generate an optimized slicing result analysis feature vector. It should be understood that the slicing process variation feature vector may include dynamic information during the slicing process, such as changes in cutting speed and cutting force. These features can reflect the stability, cutting effect, and cutting quality of the slicing process. Fusion of the slicing process variation feature vector into the optimized slicing result analysis feature vector allows consideration of the impact of the slicing process on the slicing result. The heat sink parameter association feature vector can reflect the relationship and importance between heat sink parameters at different scales. Fusion of the multi-scale heat sink parameter association feature vector comprehensively considers the multi-dimensional characteristics of heat sink parameters and captures feature correlations at different scales. This helps understand the impact of heat sink parameters on slicing results and the interactions between parameters. The pressure feature vector extracts pressure information during the slicing process, including pressure amplitude, rate of change, and trend. Pressure is an important monitoring indicator during the cutting process, reflecting cutting force, tool wear, and cutting process stability. Fusion of the pressure feature vector into the optimized slicing result analysis feature vector allows comprehensive consideration of the impact of pressure on slicing results.

[0047] Accordingly, in one embodiment of the present application, the slicing processing data fusion module includes: a fusion slicing process feature unit, used to fuse the slicing process change feature vector and the multi-scale heat sink parameter association feature vector to obtain a slicing process processing analysis feature vector; a fusion slicing result analysis unit, used to fuse the slicing process processing analysis feature vector and the pressure feature vector to obtain the slicing result analysis feature vector; and a fusion slicing feature adjustment unit, used to perform a feature core domain projection adjustment based on intrinsic regression on the slicing result analysis feature vector to obtain the optimized slicing result analysis feature vector.

[0048] Specifically, the fused slicing result analysis unit is used to: fuse the slicing process processing analysis feature vector and the pressure feature vector to obtain the slicing result analysis feature vector. It should be understood that the slicing result analysis feature vector is obtained by fusing the slicing process processing analysis feature vector and the pressure feature vector, aiming to organically combine the key information related to the slicing result in the feature vectors of two different dimensions, give full play to the advantages of each feature vector, and enable the slicing result analysis feature vector to comprehensively reflect the multi-faceted information in the slicing process, thereby providing a more comprehensive and reliable basis for the subsequent accurate judgment of whether the pressure value at the current time point should be increased or decreased through the classifier, and realize the effective optimization and precise control of the slicing process, so as to adapt to the processing requirements of heat sinks of different specifications and improve the processing quality. In one embodiment of the present application, the slicing process processing analysis feature vector and the pressure feature vector are weightedly fused to obtain the slicing result analysis feature vector.

[0049] In particular, in one embodiment of the present application, the fused slice feature adjustment unit is used to: perform feature core domain projection adjustment based on intrinsic regression on the slice result analysis feature vector to obtain the optimized slice result analysis feature vector. It should be understood that since the feature information of each position in the slice result analysis feature vector has a complex correlation relationship and may contain noise and secondary information, it is difficult to accurately reflect the actual state of the synergistic effect of multiple parameters in the heat sink processing by direct use. Therefore, the present application uses the feature core domain projection adjustment based on intrinsic regression to obtain the optimized slice result analysis feature vector, which can more accurately depict the processing state, provide a comprehensive and reliable basis for judging the pressure value adjustment direction, and improve the system's adaptive processing capability and processing quality for heat sinks of different specifications.

[0050] Specifically, the fusion slice feature adjustment unit is used to construct a pixel-by-pixel information correlation matrix of the slice result analysis feature vector, which is expressed as follows:

[0051]

[0052] Wherein, V represents the slice result analysis feature vector, v i and v j Respectively represent the eigenvalues ​​of the i-th and j-th positions of the slice result analysis feature vector, d(v i ,v j ) represents the calculation of Euclidean distance, D i,j Represents the eigenvalue at the (i, j) position of the pixel-wise information correlation matrix.

[0053] That is, by analyzing the correlation strength or similarity of the feature information at each position in the slicing result feature vector, the slicing result analysis features are converted into an explicit relationship graph representation, thereby revealing the potential patterns, grouping rules and underlying organizational structures within the features, providing more refined structural information support for feature fusion. In this way, the slicing result analysis feature vector not only contains the discrete information of each independent feature, but also contains the correlation logic between features, providing a more comprehensive feature representation basis for subsequent fusion and classifier judgment, helping the system to accurately identify the adjustment direction of the pressure value under the current processing state, thereby improving the adaptability to heat sinks of different specifications and the processing quality control accuracy, and avoiding processing deviations caused by ignoring potential correlations between features.

[0054] Specifically, the fusion slice feature adjustment unit is further used to: perform core feature extraction on the pixel-by-pixel information association matrix based on the convolution layer to obtain a slice result analysis feature core domain nonlinear activation matrix, which is expressed as:

[0055] M=Conv(D)

[0056] Wherein, D represents the pixel-by-pixel information association matrix, Conv represents the convolutional layer, and M represents the nonlinear activation matrix of the core domain of the slice result analysis feature.

[0057] That is, by learning "association patterns" rather than spatial visual patterns through convolution kernels, a kernel domain capable of representing complex nonlinear relationships between features is constructed, providing contextual information containing high-order association logic for subsequent projection modulation, enabling the system to perceive deep association structures between features that go beyond simple second-order statistics. The slicing result analysis features are upgraded from a simple set of association strengths to a deep representation containing complex association logic, which can more accurately reflect the actual state of the synergistic effects of multiple parameters such as size, hole position, and slicing pressure in heat sink processing. This provides richer semantic information support for the system to determine the current pressure value adjustment direction, thereby improving the adaptive ability of heat sinks of different specifications during the processing process, avoiding misjudgment of processing status due to reliance on simple association features, achieving refined control of slicing pressure, and ultimately improving the heat sink processing quality and device applicability.

[0058] Specifically, the fusion slice feature adjustment unit is further used to: perform characteristic spectrum decomposition on the pixel-by-pixel information correlation matrix to obtain a set of slice result analysis feature intrinsic component coding vectors, which can be expressed as follows:

[0059]

[0060] Where T represents the transpose of the vector, Λ represents the diagonal matrix, λ1 and λ mThey represent the first and mth eigenvalues ​​of the diagonal matrix respectively, U represents the set of eigencomponent encoding vectors of the slice result analysis features, x1, x2, x m Respectively represent the first, second and mth slice result analysis feature eigencomponent encoding vectors.

[0061] That is, through spectral decomposition technology, the linear space contained in the pixel-by-pixel information association matrix is ​​mapped into a set of orthogonal basis vectors, namely, the encoding vectors of the intrinsic components of the slice result analysis features. These basis vectors, as the basic atomic patterns that constitute the overall association structure, can capture the main change directions of data association (such as global trend association, local contrast association, etc.), and provide a more concise and effective expression form for the feature space. Specifically, the complex correlation information of the pixel-by-pixel information association matrix is ​​converted into a concise representation with an orthogonal basis as the core, eliminating minor correlation noise while retaining key change directions, allowing the system to focus on the core correlation patterns that affect slice pressure adjustment (such as the main correlation features of different sizes and hole positions that are sensitive to pressure), providing low-dimensional and efficient feature input for subsequent analysis.

[0062] Specifically, the fusion slice feature adjustment unit is further used to: input each slice result analysis feature intrinsic component encoding vector in the set of the slice result analysis feature intrinsic component encoding vector into the feature highlighting adjustment unit based on the self-attention mechanism to obtain a set of slice result analysis feature intrinsic component highlighting adjustment encoding vectors, which is expressed as follows:

[0063] Y=Transformer{[x1,x2,…,x m ]}=[y1,y2,…,y m ]

[0064] Among them, Transformer represents a sequence model based on the self-attention mechanism, Y represents the set of encoding vectors that highlight the adjustment of the eigenvalue components of the slice result analysis feature, y1, y2, y m Respectively represent the first, second and mth slice result analysis feature intrinsic component highlight adjustment encoding vector.

[0065] Specifically, the self-attention mechanism is used to globally model the interdependencies between intrinsic components. Based on the context of the current overall feature structure, the "salience" of each intrinsic component is dynamically evaluated. This allows for targeted enhancement of the representation of key structural patterns (such as those strongly correlated with the current heat sink size and hole location), suppressing interference from minor or irrelevant patterns, and enabling feature representation to adapt to the needs of different processing scenarios. The generated slice result analysis features the intrinsic components, highlighting the set of adjusted encoding vectors, avoiding the problem of overwhelmed critical information caused by averaging all intrinsic components.

[0066] Specifically, the fusion slice feature adjustment unit is further used to: project each slice result analysis feature intrinsic component highlight adjustment coding vector in the set of the slice result analysis feature intrinsic component highlight adjustment coding vectors to the slice result analysis feature core domain nonlinear activation matrix to obtain a set of slice result analysis feature intrinsic component core mask coding vectors, which is expressed as follows:

[0067]

[0068] in, represents matrix multiplication, S represents the characteristic scale of the nonlinear activation matrix of the core domain of the slice result analysis feature, y i represents the i-th slice result analysis feature intrinsic component highlight adjustment coding vector, L represents the length of the slice result analysis feature intrinsic component highlight adjustment coding vector, z i Represents the core mask encoding vector of the characteristic intrinsic component of the i-th slice result analysis.

[0069] That is, by using the generalized projection operation, the slicing result analysis feature intrinsic component highlight adjustment coding vector that has been adjusted for significance can absorb the local complex correlation pattern learned by the convolution layer in the core domain, thereby generating a slicing result analysis feature intrinsic component core mask coding vector that has both global structural framework and local detail dependence. It not only retains the global structural backbone of the heat sink specification parameters, but also embeds the local complex information of the dynamic coupling of multiple parameters in the slicing process, forming a more complete semantic description of the heat sink processing state.

[0070] Specifically, the fused slice feature adjustment unit is further used to: fuse the set of core mask coding vectors of the slice result analysis feature intrinsic components to obtain the optimized slice result analysis feature vector, which is expressed as follows:

[0071] V'=Concat{z1,z2,…,z m}

[0072] Among them, Concat represents the cascade function, z1, z2, z m They represent the first, second and mth slice result analysis feature intrinsic component core mask encoding vectors respectively, and V′ represents the optimized slice result analysis feature vector.

[0073] That is, the multi-source information extracted from the global structural perspective, the local nonlinear correlation perspective, and the dynamic saliency perspective is converged into a unified feature vector through a reasonable fusion method, so that it can not only retain the independent semantics of each component, but also reflect the synergistic effect of multi-dimensional information. The generated optimized slicing result analysis feature vector becomes an organic whole that combines global structural information, local nonlinear dependence, and dynamic saliency. It can accurately characterize the actual state of multi-parameter interaction in graphene heat sink processing, thereby improving the system's adaptive processing capabilities for heat sinks of different specifications, avoiding misjudgment of processing parameters due to fragmentation of feature information, and realizing effective mapping from multi-dimensional features to precise processing control, ultimately ensuring the stability of heat sink processing quality and improving the applicability of the device.

[0074] In this embodiment of the present application, the slicing processing data analysis module 140 is configured to pass the optimized slicing result analysis feature vector through a classifier to obtain a classification result. The classification result indicates whether the pressure value at the current time point should be increased or decreased. It should be understood that the optimized slicing result analysis feature vector contains a comprehensive representation of multiple features, including slicing process variation characteristics, heat sink parameter correlation characteristics, and pressure characteristics. After extraction and fusion, these feature vectors have strong representational capabilities and can better reflect the various factors and influences during the cutting process. By using a classifier, the optimized slicing result analysis feature vector can be mapped to predefined categories, such as "increase" and "decrease." The classifier can learn the relationship between the feature vector and different categories and make corresponding classification decisions. The classifier can be various machine learning algorithms, such as support vector machines (SVMs), decision trees, random forests, etc. The classification result can indicate whether the pressure value at the current time point should be increased or decreased. This conclusion is based on an analysis of the impact of slicing process characteristics and parameters on the pressure value. By adjusting according to the classification results, pressure control and optimization during the cutting process can be achieved.

[0075] Accordingly, in one embodiment of the present application, the slicing processing data analysis module 140 is configured to: use the classifier to process the optimized slicing result analysis feature vector using the following formula to obtain the classification result;

[0076] Wherein, the formula is: softmax{(W n ,B n ):…:(W1,B1)|X}, where W1 to W n is the weight matrix, B1 to B n is a bias vector, X is an optimized slicing result analysis feature vector, softmax represents a softmax function, and O represents the classification result.

[0077] It should be understood that before using the above-mentioned neural network model for inference, it is necessary to train the first convolutional neural network model of the deep-shallow fusion module, the second convolutional neural network model of the three-dimensional convolution kernel, the first context encoder, the multi-scale neighborhood feature extraction module, the third convolutional neural network model, and the classifier. In other words, the automatic processing system 100 of the graphene heat sink according to the present application also includes a training module 200 for training the first convolutional neural network model of the deep-shallow fusion module, the second convolutional neural network model of the three-dimensional convolution kernel, the first context encoder, the multi-scale neighborhood feature extraction module, the third convolutional neural network model, and the classifier.

[0078] In summary, the automated graphene heat sink processing system and method described in the embodiments of this application obtain information such as heat sink specifications and shape parameters, monitoring video, and slicing pressure values. These information then extracts slicing process variation feature vectors, heat sink parameter correlation feature vectors, and pressure feature vectors. These feature vectors are then integrated to generate an optimized slicing result analysis feature vector. A classifier is then used to classify these feature vectors, determining whether the pressure value at the current point in time should increase or decrease. This enables automated processing of graphene heat sinks, improving both the applicability and quality of the process.

[0079] As described above, the automated graphene heat sink processing system 100 according to embodiments of the present application can be implemented in various terminal devices, such as a server of the automated graphene heat sink processing system. In one example, the automated graphene heat sink processing system 100 can be integrated into the terminal device as a software module and / or a hardware module. For example, the automated graphene heat sink processing system 100 can be a software module within the terminal device's operating system, or an application developed specifically for the terminal device. Of course, the automated graphene heat sink processing system 100 can also be one of the terminal device's many hardware modules.

[0080] Alternatively, in another example, the automatic processing system 100 of the graphene heat sink and the terminal device may also be separate devices, and the automatic processing system 100 of the graphene heat sink may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0081] Figure 5 FIG. 1 is a flow chart of an automatic processing method of a graphene heat sink according to an embodiment of the present application. Figure 5As shown, the automatic processing method of the graphene heat sink according to the embodiment of the present application includes the steps of: S110, obtaining the required heat sink specifications and shape parameters, the monitoring video of the heat sink slicing process in a predetermined time period, and the slicing pressure values ​​at multiple predetermined time points in the predetermined time period, wherein the required heat sink specifications and shape parameters include size and hole position; S120, extracting the slicing process change feature vector, the multi-scale heat sink parameter association feature vector and the pressure feature vector from the monitoring video of the heat sink slicing process in the predetermined time period, the required heat sink specifications and shape parameters, and the slicing pressure values ​​at the multiple predetermined time points; S130, fusing the slicing process change feature vector, the multi-scale heat sink parameter association feature vector and the pressure feature vector to obtain an optimized slicing result analysis feature vector; S140, passing the optimized slicing result analysis feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the pressure value at the current time point should increase or decrease.

[0082] Here, those skilled in the art will understand that the specific operations of each step in the above-mentioned automatic processing method of graphene heat sinks have been described in the above reference. Figures 1 to 4 The description of the automatic processing system of the graphene heat sink has been described in detail, and therefore, its repeated description will be omitted.

[0083] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0084] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0085] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0086] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0087] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0088] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0089] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. An automatic processing system for graphene heat sinks, characterized in that: include: a slicing processing data acquisition module, configured to obtain required heat sink specifications and shape parameters, monitoring video of the heat sink slicing process during a predetermined time period, and slicing pressure values ​​at multiple predetermined time points during the predetermined time period, wherein the required heat sink specifications and shape parameters include size and hole location; a slicing process data processing module, configured to extract a slicing process variation feature vector, a multi-scale heat sink parameter association feature vector, and a pressure feature vector from monitoring video of the heat sink slicing process during the predetermined time period, the required heat sink specifications and shape parameters, and the slicing pressure values ​​at the plurality of predetermined time points; a slicing processing data fusion module, configured to fuse the slicing process variation feature vector, the multi-scale heat sink parameter association feature vector, and the pressure feature vector to obtain an optimized slicing result analysis feature vector; The slicing processing data analysis module is used to pass the optimized slicing result analysis feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether the pressure value at the current time point should be increased or decreased.

2. The automatic processing system for graphene heat sinks according to claim 1, characterized in that: The slicing data processing module includes: A monitoring video processing unit, configured to extract key frames from the monitoring video of the heat sink slicing process in the predetermined time period and then encode the key frames to obtain a slicing process change feature vector; a heat sink parameter processing unit, configured to segment the required heat sink specifications and shape parameters and then encode them to obtain the multi-scale heat sink parameter associated feature vector; The slice pressure processing unit is used to arrange the slice pressure values ​​at the multiple predetermined time points into a slice pressure value input vector according to the time dimension and then pass it through a third convolutional neural network model as a filter to obtain the pressure feature vector.

3. The automatic processing system for graphene heat sinks according to claim 2, characterized in that: The monitoring video processing unit includes: A key frame extraction subunit is used to extract a plurality of slicing process monitoring key frames from the monitoring video of the heat sink slicing process in the predetermined time period; a deep-shallow fusion subunit, configured to pass the plurality of slicing process monitoring key frames through a first convolutional neural network model including a deep-shallow fusion module to obtain a plurality of slicing process monitoring feature matrices; The three-dimensional convolution kernel unit is used to aggregate the multiple slicing process monitoring feature matrices into a three-dimensional slicing process feature tensor along the time dimension and then obtain a slicing process change feature vector by using a second convolution neural network model with a three-dimensional convolution kernel.

4. The automatic processing system for graphene heat sinks according to claim 3, characterized in that: The deep and shallow fusion subunit includes: a shallow feature extraction secondary subunit, configured to extract a shallow feature map from the i-th layer of the first convolutional neural network model, where j is greater than or equal to 1 and less than or equal to 6; a deep feature extraction secondary subunit, configured to extract a deep feature map from the j-th layer of the first convolutional neural network model, wherein a ratio between the j-th layer and the i-th layer is greater than or equal to 5 and less than or equal to 10; A feature map fusion secondary subunit, configured to fuse the shallow feature map and the deep feature map using the deep and shallow feature fusion module of the first convolutional neural network model to obtain a fused feature map; The feature map dimensionality reduction secondary subunit is used to perform global pooling along the channel dimension on the fused feature map to obtain the multiple slicing process monitoring feature matrices.

5. The automatic processing system for graphene heat sinks according to claim 4, characterized in that: The heat sink parameter processing unit includes: a heat sink parameter segmentation subunit, configured to perform segmentation processing on the required heat sink specifications and shape parameters and then pass the result through a first context encoder based on a converter to obtain a plurality of heat sink parameter feature vectors; The heat sink one-dimensional arrangement subunit is used to arrange the multiple heat sink parameter feature vectors into a one-dimensional heat sink parameter input vector and then pass it through a multi-scale neighborhood feature extraction module to obtain the multi-scale heat sink parameter associated feature vector.

6. The automatic processing system for graphene heat sinks according to claim 5, characterized in that: The slice pressure processing unit comprises: A slice pressure convolution subunit, configured to perform convolution processing on the input data based on the convolution kernel to generate a convolution feature map; a slice pressure pooling subunit, configured to perform mean pooling processing on each feature matrix along the channel dimension of the convolution feature map to obtain a pooled feature map; a slice pressure activation subunit, configured to perform nonlinear activation on the feature values ​​at each position in the pooled feature map to generate an activation feature map; Among them, the input of the first convolutional neural network model is the slice pressure value input vector, the input from the second layer to the last layer of the first convolutional neural network model is the output of the previous layer, and the output of the last layer of the first convolutional neural network model is the pressure feature vector.

7. The automatic processing system for graphene heat sinks according to claim 6, characterized in that: The slicing processing data fusion module includes: A slicing process feature fusion unit is used to fuse the slicing process change feature vector and the multi-scale heat sink parameter association feature vector to obtain a slicing process processing analysis feature vector; a fusion slicing result analysis unit, configured to fuse the slicing process processing analysis feature vector and the pressure feature vector to obtain the slicing result analysis feature vector; The fusion slice feature adjustment unit is used to perform feature core domain projection adjustment on the slice result analysis feature vector based on intrinsic regression to obtain the optimized slice result analysis feature vector.

8. The automatic processing system for graphene heat sinks according to claim 7, characterized in that: The fused slice feature adjustment unit is configured to: Constructing a pixel-by-pixel information correlation matrix of the slice result analysis feature vector; Performing core feature extraction on the pixel-by-pixel information association matrix based on a convolutional layer to obtain a slice result analysis feature core domain nonlinear activation matrix; Performing eigenspectral decomposition on the pixel-by-pixel information correlation matrix to obtain a set of slicing result analysis feature eigencomponent coding vectors; Inputting each slice result analysis feature intrinsic component encoding vector in the set of the slice result analysis feature intrinsic component encoding vectors into a feature prominence adjustment unit based on a self-attention mechanism to obtain a set of slice result analysis feature intrinsic component prominence adjustment encoding vectors; Projecting each slice result analysis feature intrinsic component highlight adjustment coding vector in the set of the slice result analysis feature intrinsic component highlight adjustment coding vectors onto the slice result analysis feature core domain nonlinear activation matrix to obtain a set of slice result analysis feature intrinsic component core mask coding vectors; The set of core mask encoding vectors of the slice result analysis feature intrinsic components are fused to obtain the optimized slice result analysis feature vector.

9. A method for automatically processing a graphene heat sink, characterized in that: include: Obtaining required heat sink specifications and shape parameters, monitoring video of the heat sink slicing process within a predetermined time period, and slicing pressure values ​​at multiple predetermined time points within the predetermined time period, wherein the required heat sink specifications and shape parameters include size and hole location; Extracting a slicing process variation feature vector, a multi-scale fin parameter association feature vector, and a pressure feature vector from the monitoring video of the fin slicing process during the predetermined time period, the required fin specifications and shape parameters, and the slicing pressure values ​​at the plurality of predetermined time points; fusing the slicing process variation feature vector, the multi-scale heat sink parameter association feature vector, and the pressure feature vector to obtain an optimized slicing result analysis feature vector; The feature vector of the optimized slicing result analysis is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the pressure value at the current time point should be increased or decreased.

Images