Diamond single crystal directional heat dissipation monitoring method and system based on artificial intelligence

Abstract

The application relates to the technical field of heat dissipation monitoring control, and discloses a diamond single crystal directional heat dissipation monitoring method and system based on artificial intelligence. The method comprises the following steps: collecting diamond single crystal heat dissipation system temperature data through a temperature sensor network, generating a heat flow control model through dynamic modeling, outputting fan and flow control instructions through closed loop calculation, adjusting a cooling system in real time and obtaining feedback, generating a correction amount by analyzing temperature deviation, automatically optimizing a heat dissipation control device, and realizing intelligent and efficient heat dissipation. The application realizes active prediction and dynamic regulation and control of the heat dissipation system through artificial intelligence technology, and improves heat dissipation efficiency and system reliability.

Description

Technical Field

[0001] This application relates to the field of heat dissipation monitoring and control technology, and in particular to a method and system for monitoring the directional heat dissipation of diamond single crystals based on artificial intelligence. Background Technology

[0002] With the rapid development of electronic information technology, electronic components are becoming increasingly integrated, with feature sizes constantly decreasing, leading to ever-increasing demands for heat dissipation. Traditional metal materials and second-generation electronic packaging materials, due to their inherent limitations in thermal conductivity and coefficient of thermal expansion, cannot achieve efficient heat dissipation, resulting in high junction temperatures for power devices and affecting their performance and lifespan. Diamond, with the highest known thermal conductivity and low coefficient of thermal expansion in nature, has broad application potential in electronic devices, aerospace, and semiconductors. Diamond / Cu composite materials, reinforced with diamond, are considered a next-generation electronic packaging material due to their excellent thermal conductivity and low coefficient of thermal expansion.

[0003] Current diamond / Cu composite material preparation processes result in composites exhibiting significant variations in diamond size, low diamond content, and uneven distribution of diamond and copper. These characteristics affect the heat dissipation performance and thermal expansion coefficient of the diamond / Cu composite material. Furthermore, traditional heat dissipation monitoring methods lack precise understanding of the anisotropic heat dissipation characteristics of single-crystal diamond, failing to fully utilize the high thermal conductivity of the 100-plane crystal for directional heat dissipation optimization. Existing monitoring technologies primarily employ temperature sensors at fixed locations, making it difficult to capture key hotspot changes along complex heat flow paths. Moreover, conventional monitoring methods lack material performance prediction capabilities, failing to assess and warn of long-term performance degradation of heat dissipation materials, leading to passive and delayed maintenance, and reducing the reliability and lifespan of high-power electronic devices. Summary of the Invention

[0004] This application provides a method and system for monitoring the directional heat dissipation of diamond single crystals based on artificial intelligence, which is used to achieve active prediction and dynamic control of the heat dissipation system through artificial intelligence technology, thereby improving heat dissipation efficiency and system reliability.

[0005] In a first aspect, this application provides an artificial intelligence-based method for monitoring the directional heat dissipation of diamond single crystals. The method includes: collecting temperature parameters of the diamond single crystal heat dissipation system through a temperature sensor network to obtain heat dissipation temperature distribution data; dynamically modeling the heat dissipation characteristics of the diamond single crystal based on the heat dissipation temperature distribution data to generate a real-time heat flow control model; performing closed-loop calculation of heat dissipation control parameters based on the real-time heat flow control model to output fan speed and flow control commands; adjusting the cooling system actuators in real time based on the fan speed and flow control commands to form heat dissipation control feedback data; analyzing the system temperature deviation in real time based on the heat dissipation control feedback data to generate cooling system parameter correction values; and automatically adjusting the heat dissipation control device based on the cooling system parameter correction values.

[0006] Secondly, this application provides an artificial intelligence-based diamond single crystal directional heat dissipation monitoring system, the artificial intelligence-based diamond single crystal directional heat dissipation monitoring system comprising:

[0007] The acquisition module is used to acquire temperature parameters of the diamond single crystal heat dissipation system through a temperature sensor network and obtain heat dissipation temperature distribution data.

[0008] The modeling module is used to dynamically model the heat dissipation characteristics of diamond single crystal based on the heat dissipation temperature distribution data, and generate a real-time heat flow control model.

[0009] The output module is used to perform closed-loop calculation of heat dissipation control parameters based on the real-time heat flow control model and output fan speed and flow control commands.

[0010] The adjustment module is used to adjust the cooling system actuator in real time based on the fan speed and flow control commands, and to generate heat dissipation control feedback data.

[0011] The analysis module is used to perform real-time analysis of the system temperature deviation based on the heat dissipation control feedback data, generate cooling system parameter correction values, and automatically adjust the heat dissipation control device based on the cooling system parameter correction values.

[0012] Thirdly, an artificial intelligence-based diamond single crystal directional heat dissipation monitoring device is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the artificial intelligence-based diamond single crystal directional heat dissipation monitoring device to execute the aforementioned artificial intelligence-based diamond single crystal directional heat dissipation monitoring method.

[0013] Fourthly, a computer-readable storage medium is provided, wherein instructions are stored therein, which, when executed on a computer, cause the computer to perform the aforementioned artificial intelligence-based diamond single crystal directional heat dissipation monitoring method.

[0014] The technical solution provided in this application utilizes a multi-point arrangement of a temperature sensor network to monitor the diamond single-crystal heat dissipation system, achieving high-resolution dynamic monitoring of heat distribution. Compared to traditional heat dissipation assessment methods that rely solely on a limited number of temperature measurement points, this approach captures more detailed changes in heat diffusion, improving the detection accuracy and response speed of thermal anomalies. Secondly, by employing dynamic modeling technology, a real-time heat flow control model is constructed based on the collected heat dissipation temperature distribution data. This effectively addresses thermal disturbances caused by changes in system load, significantly enhancing the system's adaptability and the real-time performance of thermal control. A multi-layer domain-adaptive neural network structure is introduced into the closed-loop computation stage. Through sub-modules such as feature extraction, domain adaptation, and classification, deep control features are extracted from historical control data and the real-time model, eliminating distribution differences. This algorithm not only enhances the system's ability to learn the relationship between control input and thermal response but also ensures the robustness of the algorithm model to new operating conditions. Especially in intelligent heat dissipation optimization, the AI ​​model does not merely provide a simple regression prediction method but makes causal reasoning-based optimization decisions on heat dissipation control parameters through domain-invariant feature mining and heat flow path modeling, thereby forming the optimal control strategy for fan speed and coolant flow rate. By driving the cooling actuator with PWM control and combining it with a PID feedback adjustment mechanism to achieve precise control, and further enabling intelligent analysis and correction of the system's operating status through an abnormal temperature difference identification mechanism in real-time feedback, the cooling system possesses highly automated and intelligent dynamic regulation capabilities. This invention not only demonstrates the coordinated control capabilities of sensors and actuators, but also achieves intelligent closed-loop control of the entire process from perception, modeling, decision-making to execution in the thermal management system by introducing artificial intelligence algorithms. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of an embodiment of the artificial intelligence-based directional heat dissipation monitoring method for diamond single crystals in this application.

[0017] Figure 2 This is a schematic diagram of one embodiment of the AI-based directional heat dissipation monitoring system for diamond single crystals in this application.

[0018] Figure 3 This is a schematic block diagram of the structure of the diamond single crystal directional heat dissipation monitoring device based on artificial intelligence in an embodiment of the present invention. Detailed Implementation

[0019] This application provides an artificial intelligence-based method and system for monitoring directional heat dissipation of diamond single crystals. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0020] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the artificial intelligence-based diamond single crystal directional heat dissipation monitoring method in this application includes:

[0021] Step S101: Collect temperature parameters of the diamond single crystal heat dissipation system through a temperature sensor network to obtain heat dissipation temperature distribution data;

[0022] Step S102: Dynamically model the heat dissipation characteristics of diamond single crystal based on the heat dissipation temperature distribution data to generate a real-time heat flow control model;

[0023] Step S103: Perform closed-loop calculation of heat dissipation control parameters based on the real-time heat flow control model, and output fan speed and flow control commands;

[0024] Step S104: Adjust the cooling system actuator in real time based on fan speed and flow control commands to generate heat dissipation control feedback data;

[0025] Step S105: Analyze the system temperature deviation in real time based on the heat dissipation control feedback data, generate the cooling system parameter correction amount, and automatically adjust the heat dissipation control device according to the cooling system parameter correction amount.

[0026] It is understood that the executing entity of this application can be an AI-based diamond single-crystal directional heat dissipation monitoring system, or it can be a terminal or a server; the specific implementation is not limited here. This application's embodiments use a server as an example for illustration.

[0027] Specifically, a temperature sensor network is used to collect temperature parameters from the diamond single-crystal heat dissipation system to obtain heat dissipation temperature distribution data. Platinum resistance temperature sensors are arranged at predetermined positions on the 100-facet oriented diamond single-crystal heat sink to form a temperature monitoring point array. The platinum resistance temperature sensors have a measurement range of -50°C to 250°C and an accuracy of ±0.1°C, suitable for accurately monitoring the temperature changes of the diamond heat sink. Simultaneously, a heat flux density sensor is installed at the diamond single-crystal-metal interface to acquire interface heat flux data. The heat flux density sensor uses the thermopile principle and can directly measure the amount of heat flux passing through a unit area, with a range of 0-500 W / cm². The signals collected by the temperature monitoring point array and the heat flux density sensor are synchronously acquired and processed to generate a raw data stream. The raw data stream contains temperature and heat flux data sampled at a frequency of 100Hz, with each data point accompanied by a timestamp and location identifier. The raw data stream undergoes a three-stage cascaded filtering process, sequentially performing Butterworth low-pass filtering, median filtering, and Kalman filtering to obtain noise-reduced temperature data. Butterworth low-pass filtering removes high-frequency interference above 50Hz, median filtering removes burst noise, and Kalman filtering smooths the data through prediction and correction stages. Based on the denoised temperature data, the temperature field spatial distribution matrix, heat flux density vector, and cooling medium parameter vector are calculated to generate multimodal heat dissipation parameters. These parameters are then fused with sensor location information to output heat dissipation temperature distribution data.

[0028] Dynamic modeling of the heat dissipation characteristics of diamond single crystals was performed based on heat dissipation temperature distribution data to generate a real-time heat flow control model. The thermal conductivity values ​​of diamond single crystals at different temperature points were extracted from the heat dissipation temperature distribution data to construct a thermal conductivity temperature response matrix. This matrix contains thermal conductivity values ​​at 36 different temperature points, ranging from 25°C to 200°C, with 5°C intervals. Polynomial regression fitting was performed on the thermal conductivity temperature response matrix to obtain continuous thermal conductivity data. The coefficients of the polynomial regression fitting were determined using the least squares method, expressing the discrete thermal conductivity data points as continuous functions. Based on the anisotropic characteristics of thermal conduction in the 100-plane orientation, the continuous thermal conductivity data was decomposed by crystal orientation to obtain the principal direction thermal conductivity components and the secondary direction thermal conductivity components. The 100-plane of diamond single crystals exhibits significant thermal conductivity anisotropy, with a principal direction thermal conductivity of approximately 2120 W / (m·K), while the secondary direction thermal conductivity decreases by 8-15%. The thermal conductivity characteristics of the diamond / Cu interface region are spatially discretized using a three-dimensional interpolation algorithm to form a set of thermal conductivity boundary conditions. The algorithm divides the interface region into a grid and calculates the local thermal conductivity at each grid point, thus forming the thermal conductivity boundary condition set. The thermal conductivity components in the primary and secondary directions are integrated with the thermal conductivity boundary condition set to establish a three-dimensional thermal conductivity mathematical model for the diamond heat dissipation material. The temperature field distribution under standard heat load is solved using the finite difference method to generate a real-time heat flux control model. The finite difference method discretizes the continuous region into a grid, applies finite difference approximations to the temporal and spatial partial derivatives, and obtains the steady-state temperature distribution through iterative calculation.

[0029] Based on a real-time heat flow control model, closed-loop calculations of heat dissipation control parameters are performed, outputting fan speed and flow control commands. The real-time heat flow control model and historical temperature control data are used as source and target domain data, respectively, and input into a multi-layer domain adaptive neural network structure containing feature extraction, domain adaptation, and classification layers. This multi-layer domain adaptive neural network can handle the distribution differences between source and target domain data, improving heat dissipation control accuracy. The source and target domain data are processed using wavelet packet decomposition and reconstruction algorithms to generate temperature control feature vectors. Wavelet packet decomposition and reconstruction algorithms reduce signal redundancy through multi-scale analysis while preserving key heat dissipation feature information. The temperature control feature vectors are input into the three feature extraction layers of the multi-layer domain adaptive neural network, each containing 64, 128, and 256 neurons respectively. After processing with a nonlinear activation function, deep feature representation data is obtained. The deep feature representation data is then processed in the domain adaptation layer using a multi-kernel maximum mean difference algorithm to calculate the distribution differences between the source and target domains. Backpropagation is used to adjust the network weights to obtain domain-invariant control features. The multi-kernel maximum mean difference algorithm quantifies the degree of distributional difference by calculating the distance between two distributions in the Hilbert space of the regenerating kernel, helping the network extract domain-invariant features. These domain-invariant control features are input into a three-layer classification network containing 128, 64, and 32 neurons respectively, and the maximum probability value is used as a pseudo-label for the target domain data to form a heat dissipation control optimization scheme. Based on this optimization scheme, the optimal fan speed curve and flow control curve are calculated, and fan speed and flow control commands are output.

[0030] The cooling system actuators are adjusted in real time based on fan speed and flow control commands to generate heat dissipation control feedback data. The fan speed and flow control commands are converted into PWM control signals to drive the fan motor and coolant pump, activating the cooling system actuators. The fan motor speed is controlled using a closed-loop PID controller; if the actual speed deviates from the target speed by more than 5%, the PWM duty cycle is adjusted until the speed reaches the set value. The coolant pump flow rate is proportionally adjusted; when a flow rate deviation exceeds a set threshold, it is checked for pipe blockage, and the pump pressure is increased or a warning signal is issued accordingly. Real-time operating parameters of the cooling system actuators are collected, including fan speed, flow rate, temperature, and power consumption, to construct an execution state matrix. The execution state matrix is ​​compared and analyzed with the cooling commands to determine the degree of execution deviation and generate an execution effect evaluation value. The execution effect evaluation value is then correlated with temperature monitoring data over time to generate heat dissipation control feedback data.

[0031] The system temperature deviation is analyzed in real time based on the heat dissipation control feedback data, generating cooling system parameter correction values, and automatically adjusting the heat dissipation control device accordingly. Actual temperature distribution characteristics are extracted from the heat dissipation control feedback data to construct a real-time temperature feature vector. Based on the theoretical heat dissipation characteristics of the 100-plane diamond single crystal, the ideal temperature distribution value under corresponding operating conditions is calculated, generating an ideal temperature reference vector. The difference between the real-time temperature feature vector and the ideal temperature reference vector is calculated to obtain a temperature deviation matrix. If three consecutive points in the temperature deviation matrix show a deviation exceeding 10°C, it is considered an abnormal temperature distribution. Statistical analysis of the temperature deviation matrix is ​​performed to calculate the temperature deviation rate and the proportion of abnormal temperature areas. When the temperature deviation rate exceeds 20%, an emergency cooling procedure is triggered. Based on the temperature deviation rate and the proportion of abnormal temperature areas, the cooling system operating parameters are corrected, generating fan speed correction values, flow rate correction values, and radiator angle correction values. The parameter adjustment control unit converts the fan speed correction values, flow rate correction values, and radiator angle correction values ​​into control signals, automatically adjusting the heat dissipation control device until the temperature deviation rate drops below 5%.

[0032] In this embodiment, a multi-point arrangement of a temperature sensor network is used to deploy the diamond single-crystal heat dissipation system, achieving high-resolution dynamic monitoring of heat distribution. Compared with traditional heat dissipation assessment methods that rely only on a limited number of temperature measurement points, this approach can capture more detailed changes in heat diffusion, improving the detection accuracy and response speed of thermal anomalies. Secondly, dynamic modeling technology is used to construct a real-time heat flow control model based on the collected heat dissipation temperature distribution data. This effectively addresses thermal disturbances caused by changes in system load, significantly enhancing the system's adaptability and the real-time performance of thermal control. A multi-layer domain-adaptive neural network structure is introduced into the closed-loop computation stage. Through sub-modules such as feature extraction, domain adaptation, and classification, deep control features are extracted from historical control data and the real-time model, eliminating distribution differences. This algorithm mechanism not only improves the system's ability to learn the relationship between control input and thermal response but also ensures the robustness of the algorithm model to new operating conditions. Especially in intelligent heat dissipation optimization, the AI ​​model does not merely provide a simple regression prediction method but makes causal reasoning-based optimization decisions on heat dissipation control parameters through domain-invariant feature mining and heat flow path modeling, thereby forming the optimal control strategy for fan speed and coolant flow rate. By driving the cooling actuator with PWM control and combining it with a PID feedback adjustment mechanism to achieve precise control, and further enabling intelligent analysis and correction of the system's operating status through an abnormal temperature difference identification mechanism in real-time feedback, the cooling system possesses highly automated and intelligent dynamic regulation capabilities. This invention not only demonstrates the coordinated control capabilities of sensors and actuators, but also achieves intelligent closed-loop control of the entire process from perception, modeling, decision-making to execution in the thermal management system by introducing artificial intelligence algorithms.

[0033] In one specific embodiment, the process of performing step S101 may specifically include the following steps:

[0034] Platinum resistance temperature sensors are arranged at predetermined positions on a diamond single crystal heat sink with 100 crystal plane orientation growth to form a temperature monitoring point array.

[0035] A heat flux density sensor was installed at the interface between a diamond single crystal grown with 100 crystal planes and a metal to acquire interface heat flux data.

[0036] The signals collected by the temperature monitoring point array and the heat flux density sensor are synchronously acquired and processed to generate the raw data stream;

[0037] The raw data stream is subjected to a three-stage cascaded filtering process, which sequentially executes Butterworth low-pass filtering, median filtering, and Kalman filtering to obtain noise-reduced temperature data.

[0038] Based on the noise-reduced temperature data, the spatial distribution matrix of the temperature field, the heat flux density vector, and the cooling medium parameter vector are calculated to generate multimodal heat dissipation parameters.

[0039] Multimodal heat dissipation parameters are fused with sensor location information to output heat dissipation temperature distribution data.

[0040] Specifically, platinum resistance temperature sensors are arranged at predetermined positions on a diamond single-crystal heat sink with 100-plane orientation, forming a temperature monitoring point array. A 100-plane orientation-grown diamond single crystal refers to a diamond single crystal that preferentially grows along the (100) plane direction during crystal growth. This type of diamond with this orientation exhibits excellent thermal conductivity, with a thermal conductivity as high as 2000-2200 W / (m·K). A non-uniform distribution strategy is adopted when arranging the temperature sensors, deploying them in a density ratio of 4:2:1 in the heat flow convergence area, the heat source contact surface, and the heat dissipation boundary area. Specifically, eight sensors are densely arranged near the heat source, four sensors are arranged along the main heat transfer path, and four sensors are arranged in the edge heat dissipation area, totaling 16 monitoring points forming a temperature monitoring network. The platinum resistance temperature sensor uses the PT100 model, which has a wide measurement range of -50°C to 250°C and a high accuracy of ±0.1°C. It is installed using micro-drilling and high thermal conductivity silver paste to ensure good thermal contact with the diamond single crystal.

[0041] A heat flux density sensor was installed at the interface between a diamond single crystal grown on 100 crystal planes and a metal substrate to acquire interfacial heat flux data. The heat flux density sensor operates based on the thermopile principle, consisting of multiple thermocouples connected in series, and can directly measure the heat flux passing through a unit area. An ultrathin, flexible heat flux density sensor with a thickness of less than 200 μm, an area of ​​2 mm × 2 mm, a measurement range of 0-500 W / cm², and an accuracy of ±3% was selected. Four heat flux density sensors were installed at key locations at the interface between the diamond single crystal and the copper substrate to monitor the interfacial region through which the heat flux channels pass, capturing the interfacial heat conduction status. This heat flux density sensor data is crucial for identifying changes in interfacial thermal resistance and detecting interfacial degradation problems at an early stage.

[0042] Signals collected by the temperature monitoring point array and heat flux density sensor are synchronously acquired and processed to generate a raw data stream. Synchronous acquisition is achieved through an industrial-grade data acquisition unit, with a sampling frequency set to 100Hz to ensure the capture of transient thermal response processes. Each data point is accompanied by a millisecond-level timestamp and precise spatial location identifier, forming a spatiotemporally correlated raw data stream. The raw data stream contains a temperature data matrix T(x,y,z,t) and a heat flux density matrix q(x,y,z,t), where x, y, and z are spatial coordinates, and t is the time coordinate. Due to environmental electromagnetic interference, sensor noise, and measurement random errors, the raw data stream typically contains noise components of varying frequencies and amplitudes, requiring further processing.

[0043] The raw data stream undergoes a three-stage cascaded filtering process: Butterworth low-pass filtering, median filtering, and Kalman filtering, sequentially executing to obtain noise-reduced temperature data. The first stage, Butterworth low-pass filtering, primarily handles high-frequency interference, with a cutoff frequency set to 50Hz, effectively filtering out power frequency interference of 50Hz and above. The Butterworth filter is characterized by its flat amplitude-frequency response within the passband and moderate transition band width, providing excellent suppression of high-frequency noise in the temperature signal. The second stage, median filtering, targets impulse noise and sudden outliers. Using a window width of 5 data points, it slides through the data sequence, replacing the center point value with the median value within the window, effectively removing outliers and spike noise. The third stage, Kalman filtering, employs prediction and correction phases to optimally estimate the true temperature based on historical data and current measurements. The Kalman filter is set with a process noise covariance of 0.01 and a measurement noise covariance of 0.1, effectively smoothing the temperature curve while preserving the true temperature change trend. After three-stage cascaded filtering, the noise level of the temperature data was reduced from the original ±0.8°C to ±0.15°C, significantly improving the signal-to-noise ratio of the temperature data.

[0044] Based on the noise-reduced temperature data, the temperature field spatial distribution matrix, heat flux density vector, and cooling medium parameter vector are calculated to generate multimodal heat dissipation parameters. The temperature field spatial distribution matrix is ​​obtained by three-dimensional interpolation of temperature data from 16 monitoring points, using the radial basis function interpolation method to construct a continuous temperature field throughout the entire volume of the diamond single-crystal heat sink. The heat flux density vector is calculated by combining measured interfacial heat flux data with temperature gradient information, representing the direction and intensity of heat propagation in the diamond single crystal. The cooling medium parameter vector includes key parameters such as cooling fan speed, coolant flow rate, and ambient temperature, describing the operating state of the heat dissipation system. The multimodal heat dissipation parameters integrate various data such as the temperature field distribution matrix T(x,y,z), the temperature gradient matrix ∇T(x,y,z), the heat flux density vector q(x,y,z), and the cooling medium parameter vector C(t) to form a comprehensive description of the heat dissipation state.

[0045] Multimodal heat dissipation parameters are fused with sensor location information to output heat dissipation temperature distribution data. The data fusion process employs weighted averaging and covariance cross-validation, assigning weights based on the accuracy and location importance of each sensor to generate more accurate temperature distribution data. Sensor location information includes three-dimensional spatial coordinates and a region type identifier, used to map data from discrete measurement points to continuous space. The fused heat dissipation temperature distribution data is output in JSON format, containing metadata fields (time, location, sensor type, etc.) and data fields (temperature value, heat flux density value, deterministic assessment, etc.). The rate of temperature change ∂T / ∂t and the spatial temperature gradient ∇T(x,y,z) are also calculated as supplementary fields to the data packet.

[0046] In one specific embodiment, the process of performing step S102 may specifically include the following steps:

[0047] The thermal conductivity values ​​of diamond single crystals at different temperature points were extracted from the heat dissipation temperature distribution data to construct a thermal conductivity temperature response matrix.

[0048] Polynomial regression fitting was performed on the thermal conductivity temperature response matrix to obtain continuous thermal conductivity data;

[0049] Based on the anisotropic thermal conductivity characteristics of the 100-plane orientation, the continuous thermal conductivity data is decomposed according to the crystal orientation to obtain the thermal conductivity components of the main direction and the secondary direction.

[0050] The thermal conductivity characteristics of the diamond / Cu interface region are spatially discretized using a three-dimensional interpolation algorithm to form a set of thermal conductivity boundary conditions.

[0051] By integrating the main directional thermal conductivity components, the secondary directional thermal conductivity components, and the thermal conduction boundary condition set, a three-dimensional thermal conduction mathematical model of diamond heat dissipation material is established.

[0052] The temperature field distribution of a three-dimensional heat conduction mathematical model under standard heat load is solved by the finite difference method, and a real-time heat flow control model is generated.

[0053] Specifically, the thermal conductivity values ​​of diamond single crystals at different temperature points are extracted from heat dissipation temperature distribution data to construct a thermal conductivity temperature response matrix. This is achieved through reverse calculation, deriving the material's thermal conductivity from known temperature field distribution and heat flux data. In practice, Fourier's law of heat conduction is used to calculate the thermal conductivity at each point under known temperature gradient and heat flux density conditions. Fourier's law states that heat flux density equals the product of thermal conductivity and temperature gradient. By performing finite difference calculations on the temperature field around the temperature monitoring point, the temperature gradient value is obtained, and then combined with data measured by a heat flux density sensor, the thermal conductivity at the corresponding location is calculated. This calculation is performed within a temperature range of 25°C to 200°C, with 36 temperature points set at 5°C intervals. The calculation is repeated three times at each temperature point, and the average value is taken to form a 36×2 thermal conductivity temperature response matrix. The first column represents the temperature value, and the second column represents the corresponding thermal conductivity value.

[0054] Polynomial regression fitting is performed on the thermal conductivity temperature response matrix to obtain continuous thermal conductivity data. The polynomial regression fitting uses the least squares method to fit discrete thermal conductivity data points to a continuous function. First, the order of the polynomial is determined, typically choosing 3 to 5 to balance fitting accuracy and computational complexity. Then, a least squares problem is constructed to find the polynomial coefficients that minimize the sum of squared residuals. The sum of squared residuals is calculated as the sum of the squares of the differences between the measured and fitted values. In practice, temperature values ​​are standardized to avoid numerical instability caused by excessively large coefficients of higher-order terms. Then, QR decomposition or singular value decomposition is used to solve the least squares problem to obtain the polynomial coefficients. The resulting polynomial function can calculate the corresponding thermal conductivity at any temperature point, forming continuous thermal conductivity data. Continuous thermal conductivity data, represented as a function, more accurately describes the continuous variation of thermal conductivity with temperature compared to the original discrete data points.

[0055] Based on the anisotropic thermal conductivity characteristics of the 100-plane orientation, continuous thermal conductivity data is decomposed according to crystal orientation to obtain the principal and secondary thermal conductivity components. Diamond single crystals are typical anisotropic materials, meaning their thermal conductivity varies in different directions. Diamond single crystals with the 100-plane orientation exhibit the best thermal conductivity in the direction perpendicular to the crystal plane, referred to as the principal direction; while the thermal conductivity is lower in the direction within the crystal plane, referred to as the secondary direction. By analyzing the thermal conductivity data measured in different directions, a model relating crystal orientation to thermal conductivity is established. In practice, utilizing the principle of crystal symmetry, it is assumed that there is a certain proportional relationship between the thermal conductivity in the principal and secondary directions. By decoupling the equations, the measured thermal conductivity values ​​are decomposed into principal and secondary components. The decomposition process must consider the angle of the measurement point relative to the crystal principal axis. After coordinate transformation and mathematical transformation, the principal and secondary thermal conductivity components at different temperatures are obtained. This decomposition makes the thermal conductivity model more accurate and able to reflect the thermal conductivity characteristics of the material in different directions.

[0056] The thermal conductivity characteristics of the diamond / Cu interface region are spatially discretized using a three-dimensional interpolation algorithm to form a set of thermal conductivity boundary conditions. The diamond / Cu interface is a critical region in the heat dissipation system, and its interfacial thermal resistance significantly affects the overall heat dissipation performance. The three-dimensional interpolation algorithm is used to construct a continuous distribution of interfacial thermal conductivity characteristics from a finite number of measurement points. First, the interface region is divided into a three-dimensional grid, typically 80×80×30 pixels, with a total of 192,000 nodes. Then, for each grid point, the thermal conductivity characteristics at that point are calculated using interpolation based on the values ​​of surrounding measurement points. Interpolation methods typically employ trilinear interpolation or radial basis function interpolation, taking into account the spatial distance between points and the differences in measured values. During the calculation, the physical properties of the interface, such as roughness, wetting angle, and bonding strength, must also be considered, as these factors affect local thermal conductivity performance. The final set of thermal conductivity boundary conditions includes the thermal contact thermal conductivity of each grid point in the interface region, forming a complete description of the boundary conditions.

[0057] A three-dimensional mathematical model of heat conduction for diamond heat dissipation materials is established by integrating the principal and secondary thermal conductivity components with the heat conduction boundary condition set. This integration process requires considering the material's physical properties, geometry, and boundary conditions to construct a complete heat conduction equation. First, the geometric boundaries of the computational domain are defined, including the diamond single crystal, the copper matrix, and their interface regions. Then, corresponding material properties are applied within each region; for example, the principal and secondary thermal conductivity components are used in the diamond region, the thermal conductivity of copper is used in the copper matrix region, and the thermal contact thermal conductivity from the heat conduction boundary condition set is used in the interface region. For unsteady-state heat conduction, a partial differential equation for heat conduction including a time term is constructed to describe the temperature field's variation with time and space. For cases with heat sources, a heat source term is also added, representing the heat generation rate per unit volume. After integrating these conditions, a complete three-dimensional mathematical model of heat conduction is formed, which can accurately describe the heat conduction behavior of diamond heat dissipation materials under actual operating conditions.

[0058] A real-time heat flux control model is generated by solving a three-dimensional heat conduction mathematical model under standard heat load using the finite difference method. The finite difference method is a commonly used method in numerical computation. Its basic idea is to discretize the continuous region into a grid and use finite difference approximations to replace the derivative terms in the differential equation. In specific implementation, a grid is used to divide the computational domain into a large number of discrete grid cells, with the center of each grid cell being a computational node. Forward difference is used for the time derivative in the heat conduction equation, and central difference is used for the spatial derivative, constructing a solution format. For boundary conditions, the appropriate difference form is used according to their type. For example, the first type of boundary condition directly specifies the boundary node temperature, the second type of boundary condition uses symmetric difference to handle the heat flux boundary, and the third type of boundary condition considers the boundary heat exchange coefficient. During the calculation process, an initial temperature field is first set, and then the solution is iterated at time steps until a steady-state condition or a specified termination time is reached. The steady-state criterion is usually that the maximum difference between the temperature fields of two adjacent iterations is less than a preset threshold. The final solution generates a real-time heat flow control model, which can accurately predict the temperature distribution and heat flow path inside the diamond heat dissipation material under a given heat load.

[0059] In one specific embodiment, the process of executing step S103 may specifically include the following steps:

[0060] The real-time heat flow control model and historical temperature control data are used as source domain data and target domain data, respectively, and the input is a multi-layer domain adaptation neural network structure containing a feature extraction layer, a domain adaptation layer and a classification layer.

[0061] The source domain data and target domain data are processed using wavelet packet decomposition and reconstruction algorithms to generate temperature control feature vectors;

[0062] The temperature control feature vector is input into the three feature extraction layers of a multi-layer adaptive neural network, each containing 64, 128, and 256 neurons respectively. After processing by a non-linear activation function, deep feature representation data is obtained.

[0063] For deep feature representation data, the multi-kernel maximum mean difference algorithm is used in the domain adaptation layer to calculate the distribution difference between the source domain and the target domain. The network weights are adjusted through backpropagation to obtain domain-invariant control features.

[0064] The domain-invariant control features are input into a three-layer classification network containing 128, 64, and 32 neurons respectively, and the maximum probability value is used as a pseudo-label for the target domain data to form a heat dissipation control optimization scheme.

[0065] The optimal fan speed curve and flow control curve are calculated based on the heat dissipation control optimization scheme, and the fan speed and flow control commands are output.

[0066] Specifically, the real-time heat flux control model and historical temperature control data are used as the source domain data and target domain data, respectively, and input into a multi-layer domain adaptive neural network structure containing a feature extraction layer, a domain adaptation layer, and a classification layer. The source domain data refers to the ideal heat flux distribution and temperature field data generated by the theoretical model, including the anisotropic thermal conductivity characteristics, interfacial thermal resistance characteristics, and ideal heat dissipation effect of the diamond single crystal. The target domain data is the historical temperature control data collected during the actual operation of the heat dissipation system, reflecting the heat dissipation performance in the actual working environment. The multi-layer domain adaptive neural network is a deep learning architecture specifically designed to handle the problem of inconsistent distributions between the source and target domain data. It can learn domain-invariant features, improving the model's applicability in real-world environments. This network structure consists of three main parts: a feature extraction layer for extracting deep features from the raw data; a domain adaptation layer responsible for reducing the distribution differences between the source and target domains; and a classification layer for making control decisions based on the extracted features.

[0067] The source and target domain data are processed using wavelet packet decomposition and reconstruction algorithms to generate temperature control feature vectors. Wavelet packet decomposition is an extension of traditional wavelet transform, providing finer frequency division. The process first determines the number of decomposition levels, typically 3 to 4, and then decomposes the original signal into multiple layers to obtain coefficients for different frequency sub-bands. For temperature data, this decomposition effectively separates different scale features of temperature changes, such as rapid fluctuations, medium-term variations, and long-term trends. After decomposition, sub-bands containing key feature information are selected for reconstruction based on indicators such as energy distribution and information entropy. The reconstruction process is the inverse of decomposition, recombining the retained sub-band coefficients to generate a temperature signal with significant features but reduced redundancy. The processed data forms a temperature control feature vector that retains the key dynamic characteristics of the original data while reducing computational load and noise interference.

[0068] The temperature control feature vector is input into a three-layer feature extraction layer of a multi-layer adaptive neural network. Each layer contains 64, 128, and 256 neurons respectively. After processing by nonlinear activation functions, deep feature representation data is obtained. The feature extraction layer adopts a deep convolutional neural network structure. The first layer of 64 neurons is responsible for extracting low-level features such as temperature gradient and heat flow direction; the second layer of 128 neurons combines low-level features to form mid-level features such as heat conduction path and temperature distribution pattern; the third layer of 256 neurons extracts high-level abstract features such as heat dissipation system status and abnormal patterns. The neurons in each layer are connected by nonlinear activation functions. Commonly used activation functions include ReLU or Leaky ReLU, which can introduce nonlinear transformations and enhance the network's expressive power. The data processing path of the feature extraction layer is: original feature vector → first layer convolution + activation → second layer convolution + activation → third layer convolution + activation → deep feature representation data. Through layer-by-layer transformation, the original temperature control feature vector is transformed into a representation in a high-dimensional abstract feature space, containing the essential characteristics of the heat dissipation system.

[0069] For deep feature representations, a multi-kernel maximum mean difference algorithm is used in the domain adaptation layer to calculate the distribution difference between the source and target domains. The network weights are then adjusted through backpropagation to obtain domain-invariant control features. The multi-kernel maximum mean difference algorithm is a method for measuring the difference between two distributions; its basic principle is to calculate the distance between the two distributions in the regenerating kernel Hilbert space. The algorithm first maps deep features to a high-dimensional space, then calculates the mean difference between the source and target domain features in this space. By using combinations of multiple kernel functions (such as Gaussian kernels, polynomial kernels, etc.), it can capture distribution differences of different scales and forms. The calculated distribution difference is used as one of the network's loss functions, and the network parameters are adjusted through backpropagation, aiming to minimize the distribution difference between the source and target domains. During backpropagation, the network weights are updated according to the gradient direction, gradually reducing the distribution difference. After multiple rounds of training iterations, the features learned by the network have a similar distribution between the source and target domains, i.e., domain-invariant control features. These features are applicable to both theoretical models and real-world environments. Domain-invariant control features are input into a three-layer classification network, containing 128, 64, and 32 neurons respectively. The maximum probability value of the target domain data is used as a pseudo-label to form a heat dissipation control optimization scheme. The classification network adopts a fully connected neural network structure. Starting from the domain-invariant control features, it outputs a heat dissipation control decision through three layers of nonlinear transformation. The first layer of 128 neurons processes the domain-invariant features, the second layer of 64 neurons performs intermediate representation, and the third layer of 32 neurons generates the control output. Since the target domain data usually lacks control labels, a pseudo-label strategy is used to solve the unsupervised learning problem. Specifically, an initial model is first trained using labeled data from the source domain. Then, the model is applied to the target domain data. For each target domain sample, the classification network outputs multiple possible control schemes and their probability values. The scheme with the highest probability is selected as the pseudo-label for that sample. With pseudo-labels, the target domain data can participate in network training, further improving model performance. The training process uses cross-validation, dividing the dataset into training, validation, and test sets to ensure the model's generalization ability. The finally trained network can directly output a heat dissipation control optimization scheme based on the input heat dissipation state data.

[0070] The optimal fan speed and flow control curves are calculated based on the heat dissipation control optimization scheme, and fan speed and flow control commands are output. The heat dissipation control optimization scheme includes the control parameters of each actuator in the heat dissipation system, which need further processing to be converted into specific control commands. First, based on the control objectives and constraints in the heat dissipation control optimization scheme, an optimization problem for fan speed and flow control is constructed. The optimization objective is usually to minimize temperature deviation and energy consumption, and the constraints include fan speed range, flow limit, noise level, etc. Then, by solving this optimization problem, the time-varying fan speed and flow control curves are obtained. The curve generation process considers the inertial characteristics of the thermal system to avoid frequent jumps in control output. Finally, the continuous control curves are discretized into a sequence of commands that the actual control system can execute, including fan PWM control signals and flow valve opening signals. These control commands are sent to the actuators of the heat dissipation system through the control interface to achieve precise control of the heat dissipation process.

[0071] In one specific embodiment, the process of calculating the optimal fan speed curve and flow control curve based on the heat dissipation control optimization scheme may specifically include the following steps:

[0072] The heat dissipation control optimization scheme is converted into a thermal resistance network topology diagram, and the heat source points and heat dissipation boundary points of the diamond single crystal are marked.

[0073] Based on the anisotropic characteristics of the 100 crystal planes of diamond single crystal, a heat transfer weight value is assigned to each node in the thermal resistance network topology diagram.

[0074] Starting from the heat source point, the thermal path tracing algorithm is used to traverse the thermal resistance network topology and calculate all possible heat transfer paths.

[0075] The thermal resistance of all possible heat transfer paths is cumulatively calculated to obtain a set of path thermal resistance values.

[0076] The heat transfer channel with the lowest thermal resistance is selected from the set of path thermal resistance values ​​as the main heat dissipation channel, and the channel with the second lowest thermal resistance value is selected as the auxiliary heat dissipation channel.

[0077] Based on the distribution ratio of 70% of the heat load to 30% of the heat load borne by the main heat dissipation channel, the required heat dissipation power for each channel is calculated. Combining the temperature gradient value and required heat flow rate on each channel, the fan speed and coolant flow rate are determined, and fan speed and flow rate control commands are output.

[0078] Specifically, the temperature distribution and heat flow data in the optimization scheme are mapped to a thermal resistance network through data structure transformation. A thermal resistance network topology diagram is a graphical representation of the heat dissipation path, consisting of nodes and connections. Each node represents a spatial location within the diamond single crystal, and connections represent the path of heat transfer. The transformation process uses a mesh generation method, dividing the diamond single crystal heat sink into a fine mesh, typically 80×80×30, totaling 192,000 nodes. The center of each mesh cell serves as a node in the thermal resistance network, and connections are established between adjacent nodes. For the diamond single crystal heat dissipation system, two types of special nodes need to be clearly marked: heat source points and heat dissipation boundary points. Heat source points refer to the areas where high-power electronic devices contact the diamond single crystal, representing the location of heat input; heat dissipation boundary points are the outer surface points of the diamond single crystal in contact with the heat sink or the environment, representing the location of heat output. These special nodes are identified through geometric boundary conditions and temperature threshold standards. For example, areas with temperatures exceeding a set threshold are marked as heat source points, and outer surface areas with temperatures below another threshold are marked as heat dissipation boundary points. Based on the anisotropic characteristics of the 100-plane diamond single crystal, heat transfer weights are assigned to each node in the thermal resistance network topology. Diamond single crystal is a typical anisotropic thermally conductive material, exhibiting significant differences in thermal conductivity across different crystal orientations. The 100-plane refers to a specific crystal plane of diamond, characterized by optimal thermal conductivity along the direction perpendicular to the plane, while thermal conductivity within the plane is lower. For each node in the thermal resistance network, the directional angle of its location relative to the principal axis of the diamond crystal needs to be calculated, and then a heat transfer weight is assigned based on this angle. The weight allocation uses a cosine square function, meaning the weight value is proportional to the square of the cosine of the angle between the crystal orientation and the principal direction. In practice, the crystal orientation of the diamond single crystal is first determined using X-ray diffraction or electron backscattering diffraction techniques to obtain the relationship between the crystal principal axis and the heat sink coordinate system. Then, the angle between the heat transfer direction and the crystal principal axis is calculated for each node. The weight value along the principal direction is set to 1.0, and the weight value perpendicular to the principal direction is set according to the measured anisotropy ratio, typically around 0.8. In this way, each node in the thermal resistance network has a directional weight, reflecting the anisotropic thermal conductivity of diamond single crystal.

[0079] Starting from the heat source point, a thermal path tracing algorithm is used to traverse the thermal resistance network topology and calculate all possible heat transfer paths. The thermal path tracing algorithm is a special graph traversal algorithm suitable for finding heat transfer paths in a heat dissipation system. This algorithm is similar to the traditional Dijkstra's shortest path algorithm, but with specific improvements for heat conduction characteristics. The algorithm starts from the marked heat source point and progressively explores paths to each heat dissipation boundary point. For each current node, the algorithm considers all neighboring nodes, calculates the thermal resistance values ​​for heat transfer to each neighboring node, and then selects the node with the lowest thermal resistance as the next exploration point. To avoid the algorithm getting trapped in local optima, a simulated annealing strategy is also used, meaning that there is a certain probability of selecting a non-optimal path during the exploration process. Due to the complexity of the thermal resistance network, a complete traversal of all possible paths is computationally intensive; in practice, a pruning strategy is usually used, retaining only paths with thermal resistance less than a certain threshold. After traversal and calculation, multiple possible heat transfer paths from the heat source point to each heat dissipation boundary point are obtained. The thermal resistance of all possible heat transfer paths is cumulatively calculated to obtain a set of path thermal resistance values. Cumulative thermal resistance calculation involves summing the thermal resistance values ​​between all nodes along each heat transfer path. For any two adjacent nodes on the path, the thermal resistance is determined by the distance between the nodes, the average thermal conductivity, and the effective heat transfer cross-sectional area. The formula is: inter-node thermal resistance equals the distance between the nodes divided by (average thermal conductivity multiplied by effective heat transfer cross-sectional area). Here, the distance between the nodes is the Euclidean distance in space; the average thermal conductivity, considering the temperature and crystal orientation of the two nodes, is calculated through interpolation; and the effective heat transfer cross-sectional area is the cross-sectional area perpendicular to the heat flow direction. For heat transfer paths passing through interfaces, the interface thermal resistance must also be included, determined by the thermal conductivity and contact area of ​​the interface. Through this cumulative calculation, the total thermal resistance value for each heat transfer path is obtained, forming a set of path thermal resistance values.

[0080] The heat transfer path with the lowest thermal resistance is selected from the set of path thermal resistance values ​​as the main heat dissipation path, while the path with the second lowest thermal resistance value is selected as the auxiliary heat dissipation path. The selection process first sorts the path thermal resistance values ​​from smallest to largest, selecting the path with the lowest thermal resistance value as a candidate for the main heat dissipation path. However, simply selecting the path with the lowest thermal resistance may lead to excessive heat concentration and localized hotspots. Therefore, when selecting auxiliary heat dissipation paths, not only thermal resistance value but also spatial distribution factors must be considered. In practice, the spatial distance between the candidate path and the main heat dissipation path is calculated, and the path that has both a low thermal resistance value and maintains appropriate spatial separation from the main heat dissipation path is selected as the auxiliary heat dissipation path. The purpose of spatial separation is to balance the heat dissipation load and avoid localized overheating. Typically, 1-3 main heat dissipation paths and 3-5 auxiliary heat dissipation paths are selected to form a multi-parallel heat dissipation path network.

[0081] Based on the heat load distribution ratio of 70% for the main heat dissipation channel and 30% for the auxiliary heat dissipation channel, the required heat dissipation power for each channel is calculated. Combining the temperature gradient value and required heat flow rate for each channel, the fan speed and coolant flow rate are determined, and fan speed and flow rate control commands are output. Heat load distribution is based on the thermal resistance ratio of the main and auxiliary heat dissipation channels; the main heat dissipation channel has lower thermal resistance and receives more heat load, while the auxiliary heat dissipation channel has higher thermal resistance and receives less heat load. The required heat dissipation power for each channel is calculated through heat flow distribution, using the formula: channel heat dissipation power equals total heat load multiplied by the channel's load distribution ratio. The temperature gradient value for each channel is obtained by dividing the temperature difference between the heat source point and the heat dissipation boundary point by the path length. Combining the required heat flow rate and temperature gradient for each channel, the required heat dissipation capacity is calculated. There is a mapping relationship between heat dissipation capacity and fan speed and coolant flow rate, typically obtained through an experimentally measured lookup table. The lookup table input is the required heat dissipation capacity, and the output is the corresponding fan speed and coolant flow rate value. Fan speed values ​​are typically expressed in RPMs and are converted into PWM control signals; coolant flow rates are expressed in L / min and are converted into flow control valve opening signals. These control commands are sent to the corresponding actuators to achieve precise control of the cooling system.

[0082] In one specific embodiment, the process of executing step S104 may specifically include the following steps:

[0083] The fan speed and flow control commands are converted into PWM control signals to drive the fan motor and coolant pump, and start the cooling system actuators.

[0084] The fan motor speed is controlled by a PID controller in a closed loop. If the actual speed deviates from the target speed by more than 5%, the PWM duty cycle is adjusted until the speed reaches the set value.

[0085] The flow rate of the coolant pump is proportionally adjusted. When the flow deviation exceeds the set threshold, it is determined whether there is a blockage in the pipeline, and the pump pressure is increased or a warning signal is issued.

[0086] Collect real-time operating parameters of the cooling system actuators, including fan speed, flow rate, temperature, and power consumption, and construct an execution status matrix;

[0087] The execution status matrix is ​​compared and analyzed with the cooling command to determine the degree of execution deviation and generate an execution effect evaluation value.

[0088] The performance evaluation values ​​are correlated with temperature monitoring data over time to generate heat dissipation control feedback data.

[0089] Specifically, the process of converting fan speed and flow control commands into PWM control signals involves signal conversion and modulation. The fan speed and flow control commands need to be converted into specific execution signals to drive the cooling system actuators. This conversion process is achieved through a control signal mapping table, which represents the correspondence between fan speed and PWM duty cycle, and flow rate and pump speed. For fan speed, the typical mapping relationship is linear or quadratic, with a speed range generally between 1000-4000 RPM, corresponding to a PWM duty cycle range of 20%-100%. The calculation formula is: PWM duty cycle = (current set speed - minimum speed) / (maximum speed - minimum speed) × 80% + 20%. For the coolant pump, control methods include variable frequency speed control and PWM modulation, with a flow rate range typically between 0.5-3.0 L / min. The corresponding control signals are also provided by the mapping table. The converted PWM signal is output through a drive circuit, which includes power amplification and level conversion sections to ensure that the control signal can correctly drive the fan motor and coolant pump. The frequency selection of the PWM signal needs to take into account the response characteristics of the actuator. The PWM frequency for fan control is usually 25kHz, which avoids the range of human hearing while meeting the motor control requirements. The control frequency of the coolant pump is determined according to the pump type. Variable frequency pumps are generally adjusted in the range of 0-60Hz.

[0090] The fan motor speed is controlled using a closed-loop PID controller. The PID controller consists of proportional, integral, and derivative components, used to adjust the control output based on feedback information. During fan control, the actual fan speed is first obtained from the fan speed sensor and then compared with the target speed value to calculate the speed deviation. When the deviation exceeds a set threshold (5%), the PID controller calculates the required PWM duty cycle increment based on the magnitude and trend of the deviation. The control formula is: PWM duty cycle increment = Kp × deviation + Ki × deviation integral + Kd × deviation change rate. Here, Kp, Ki, and Kd are the proportional, integral, and derivative parameters, respectively, and need to be adjusted according to the fan's dynamic characteristics. For most cooling fans, typical PID parameter values ​​are: Kp = 0.8, Ki = 0.05, Kd = 0.1. For example, when the target speed is 3000 RPM and the actual measured speed is 2700 RPM, the deviation is 300 RPM, accounting for 10%, exceeding the 5% threshold, triggering PID adjustment. Assuming the current PWM duty cycle is 60%, and the calculated duty cycle increment is 5%, the adjusted PWM duty cycle will be 65%. The PID control process is iterative, sampling and adjusting every 100ms until the actual speed reaches the target value or the deviation is less than the threshold.

[0091] The coolant pump flow rate is proportionally adjusted using a feedback control strategy simpler than motor speed control. Flow control first acquires the actual flow rate value via a flow sensor and compares it to the target flow rate. When a flow deviation exceeds a set threshold, the control system analyzes the cause of the deviation. Unlike fans, abnormal coolant flow often indicates potential system malfunctions such as pipe blockage. The method for determining pipe blockage is to simultaneously monitor flow and pressure data and analyze the flow-pressure relationship curve. Under normal circumstances, flow rate is directly proportional to pump speed and inversely proportional to pipe resistance. If increasing pump speed does not significantly increase flow rate while pressure rises significantly, pipe blockage is suspected. Two control strategies are employed: for minor blockage, pump pressure is appropriately increased to maintain the target flow rate; for severe blockage, a warning signal is issued to prevent pump overload damage. The method of increasing pump pressure varies depending on the pump type; variable frequency pumps increase pressure by increasing frequency, while PWM-controlled pumps increase pressure by increasing the duty cycle. The warning signal is sent to the control center via the system bus, triggering a maintenance procedure.

[0092] Real-time operating parameters of the cooling system actuators are collected to comprehensively monitor the system's operational status. These parameters include fan speed, flow rate, temperature, and power consumption, all acquired using dedicated sensors. Fan speed is measured using Hall effect sensors or photoelectric encoders with an accuracy of ±1%; flow rate is obtained using turbine or ultrasonic flow meters with an accuracy of ±2%; temperature is obtained from temperature sensors distributed throughout the system; and power consumption is calculated using a precision current detection circuit. All parameters are sampled according to a unified time base at a sampling frequency of 10Hz to ensure data consistency. These parameters are processed to form an execution status matrix, with an n-row, 4-column structure, where n is the number of sampling points, and the four columns correspond to fan speed, flow rate, temperature, and power consumption data, respectively. This execution status matrix provides a digital mapping of the system's operating status, serving as the data foundation for subsequent analysis.

[0093] Comparing the execution state matrix with cooling commands to determine the degree of execution deviation is a crucial step in evaluating the effectiveness of heat dissipation control. The comparative analysis first calculates the differences between the execution state and the control commands, including fan speed deviation, flow rate deviation, temperature deviation, and power consumption deviation. Each deviation has a corresponding weighting coefficient, reflecting its importance to the heat dissipation effect. The deviation calculation uses a weighted mean square error formula, comprehensively considering the differences of each parameter. The calculated deviation values ​​are normalized to obtain an execution deviation metric between 0 and 1. Based on the magnitude of the deviation metric, the execution effect is divided into four levels: excellent, good, average, and poor, corresponding to deviation values ​​less than 0.1, 0.1-0.2, 0.2-0.3, and greater than 0.3, respectively. The execution effect evaluation considers not only static deviation but also dynamic response characteristics, such as rise time, overshoot, and settling time. These indicators are obtained by analyzing the time series characteristics of the execution state matrix. The final generated execution effect evaluation value is a comprehensive indicator that fully reflects the accuracy and responsiveness of heat dissipation control.

[0094] Performing time-series correlation analysis between the performance evaluation values ​​and temperature monitoring data forms the data foundation for heat dissipation control feedback data. The time-series correlation analysis first determines the analysis window, typically 60 seconds wide with a 10-second sliding step. Within each window, the correlation between the performance evaluation values ​​and temperature changes is calculated, including lag correlation and predictive correlation. Lag correlation analyzes the temperature response delay after the control action, usually calculated using a cross-correlation function to determine the time delay corresponding to the maximum correlation. Predictive correlation analyzes the impact of the current control state on future temperatures, using an autoregressive moving average model to predict future temperature changes. The correlation analysis results are combined with actual temperature data to form complete heat dissipation control feedback data. The feedback data structure includes fields such as current temperature state, temperature change trend, control execution state, performance evaluation, temperature-control correlation, and system anomaly flags. This data is integrated according to a unified timestamp and stored in JSON format.

[0095] In one specific embodiment, the process of executing step S105 may specifically include the following steps:

[0096] Extract the actual temperature distribution features from the heat dissipation control feedback data and construct a real-time temperature feature vector;

[0097] Based on the theoretical heat dissipation characteristics of the 100 crystal plane of diamond single crystal, the ideal temperature distribution value under the corresponding working conditions is calculated, and the ideal temperature reference vector is generated.

[0098] The difference between the real-time temperature feature vector and the ideal temperature reference vector is calculated to obtain the temperature deviation matrix. If three consecutive points in the temperature deviation matrix have a deviation of more than 10°C, it is judged as an abnormal temperature distribution.

[0099] Perform statistical analysis on the temperature deviation matrix to calculate the temperature deviation rate and the proportion of abnormal temperature areas. When the temperature deviation rate exceeds 20%, trigger the emergency cooling procedure.

[0100] Based on the temperature deviation rate and the proportion of abnormal temperature areas, the operating parameters of the cooling system are corrected and calculated to generate fan speed correction, flow rate correction, and radiator angle correction.

[0101] The parameter adjustment control unit converts fan speed correction, airflow correction, and radiator angle correction into control signals to automatically adjust the heat dissipation control device.

[0102] Specifically, feature extraction is performed on the raw temperature data. The heat dissipation control feedback data includes time-series temperature data collected by a temperature sensor network, which comes from temperature sensors distributed at key locations on the diamond single-crystal heat sink. The feature extraction process uses principal component analysis to reduce the dimensionality of the high-dimensional temperature data and extract the main temperature distribution patterns. In practice, the temperature data is first organized into a matrix, with each row representing a time point and each column representing a sensor location. Then, the covariance matrix of this matrix is ​​calculated, and the eigenvalues ​​and eigenvectors of the covariance matrix are solved. The magnitude of the eigenvalues ​​indicates the importance of the corresponding eigenvectors, and the eigenvectors corresponding to the k largest eigenvalues ​​are usually selected as principal components. A typical value of k is 4-6, which can capture more than 90% of the variability in the raw data. By projecting the raw temperature data onto these k principal components, a low-dimensional feature representation is obtained. In addition to principal component analysis, the time derivative and spatial gradient of the temperature field also need to be calculated, representing the rate of temperature change and spatial non-uniformity, respectively. Finally, the real-time temperature feature vector contains multiple features such as principal component coefficients, temperature change rate, and temperature gradient, comprehensively describing the current temperature state of the heat dissipation system.

[0103] Based on the theoretical heat dissipation characteristics of the 100-plane diamond single crystal, the ideal temperature distribution under corresponding operating conditions is calculated, generating an ideal temperature reference vector. Diamond single crystal, as a material with extremely high thermal conductivity, exhibits heat dissipation characteristics closely related to its crystal orientation. The 100-plane is a specific crystal plane of diamond, where the thermal conductivity is highest. The ideal temperature distribution calculation is based on a heat conduction theory model, considering the anisotropic thermal conductivity, interfacial thermal resistance, and boundary conditions of the diamond single crystal. The calculation process first determines the current operating load conditions, including heat source power, ambient temperature, and cooling conditions. Then, by solving the heat conduction equation, the ideal temperature values ​​at each monitoring point under steady-state conditions are calculated. The heat conduction equation considers the difference in thermal conductivity of the diamond single crystal in different directions, setting the thermal conductivity of the principal direction of the 100-plane as the highest value and the secondary direction as a relatively lower value. The equation is solved using the finite element method or finite difference method, with a sufficiently fine mesh size to capture heat flow details. Boundary conditions include the heat flow boundary at the heat source interface and the convection boundary at the heat dissipation boundary. The calculated ideal temperature distribution values ​​form an ideal temperature reference vector, which has the same data structure as the real-time temperature feature vector, facilitating subsequent comparison and analysis.

[0104] The temperature deviation matrix is ​​obtained by calculating the difference between the real-time temperature feature vector and the ideal temperature reference vector. If three consecutive points in the temperature deviation matrix show a deviation exceeding 10°C, it is considered an abnormal temperature distribution. The difference calculation is performed element-wise; for each corresponding element in the vector, the difference between the actual value and the ideal value is calculated. The resulting temperature deviation matrix maintains the dimension and structure of the original vector, with each element representing the deviation value of the corresponding feature. For temperature values, the deviation represents the difference between the actual and ideal temperatures; for temperature change rates, the deviation represents the difference between the actual and ideal change rates; for temperature gradients, the deviation represents the difference between the actual and ideal gradients. The temperature deviation matrix reflects the difference between the actual and ideal operating states of the heat dissipation system. Anomaly detection uses a continuous threshold method to monitor consecutive abnormal points in the deviation matrix. The criterion of three consecutive points with a deviation exceeding 10°C considers occasional noise and random fluctuations, avoiding misjudgments caused by single-point anomalies, and enabling timely detection of systemic heat dissipation anomalies. Such abnormal temperature distributions usually indicate potential problems in the heat dissipation system, such as interface degradation, channel blockage, or cooling failure. Statistical analysis is performed on the temperature deviation matrix to calculate the temperature deviation rate and the percentage of abnormal temperature areas. When the temperature deviation rate exceeds 20%, an emergency cooling procedure is triggered. The statistical analysis first calculates the temperature deviation rate, which is the relative deviation between the actual temperature and the ideal temperature. The formula is: deviation rate = deviation value divided by ideal value multiplied by 100%. The deviation rate is calculated for each monitoring point, and the maximum absolute value is taken as the overall temperature deviation rate index. The percentage of abnormal temperature areas is defined as the percentage of monitoring points whose temperature deviation exceeds a preset threshold out of the total number of monitoring points. The preset threshold is typically set to 5%, meaning that a difference of more than 5% between the actual temperature and the ideal temperature is considered abnormal. The statistical analysis also includes calculating the mean, standard deviation, skewness, and kurtosis of the deviation matrix, which reflect the overall distribution characteristics of the temperature deviation. When the temperature deviation rate exceeds 20%, the system determines that the heat dissipation is severely abnormal and automatically triggers an emergency cooling procedure, including measures such as adjusting the cooling fan to maximum speed and the coolant pump to maximum flow rate to prevent equipment damage due to overheating.

[0105] Based on the temperature deviation rate and the proportion of abnormal temperature areas, the operating parameters of the cooling system are corrected, generating fan speed correction, flow rate correction, and radiator angle correction. The correction calculation is based on fuzzy control rules, determining the correction magnitude for different parameters according to the combination of temperature deviation rate and abnormal area proportion. Fuzzy control rules are expressed in "IF-THEN" form, such as "IF large temperature deviation rate AND large abnormal area proportion THEN large fan speed correction". The rule base typically contains 9-16 rules, covering various possible combinations. During fuzzy inference, the temperature deviation rate and abnormal area proportion are first converted into membership degrees of fuzzy sets. Then, the fuzzy set of control output is derived through fuzzy rules. Finally, the specific correction value is obtained through defuzzification calculation. The fan speed correction represents the percentage of speed adjustment required, typically ranging from -20% to +50%; the flow rate correction represents the percentage of flow rate adjustment required, with a similar range; and the radiator angle correction represents the angle change required, typically ranging from -15° to +15°. These corrections take into account the severity and distribution of temperature anomalies, and corresponding adjustment measures are taken for different situations.

[0106] The parameter adjustment control unit converts fan speed correction, flow rate correction, and radiator angle correction into control signals to automatically adjust the cooling control device. This control unit acts as the interface between the analysis and decision-making mechanisms and the execution mechanism, responsible for converting the results of the correction calculations into actual control signals. The conversion process first considers the current system state, applying correction values ​​to calculate target values ​​based on the current fan speed, flow rate, and radiator angle. For example, if the current fan speed is 3000 RPM and the correction is +20%, the target speed is 3600 RPM. Then, the control unit converts the target values ​​into corresponding control signal formats: fan speed is converted to PWM duty cycle, flow rate to pump speed or valve opening, and radiator angle to stepper motor steps. The control signals are sent to the corresponding drive circuits, which amplify and adjust the signal level, ultimately driving the actuators to achieve parameter adjustment. A gradual adjustment strategy is used, especially for fan speed and radiator angle adjustments, to avoid mechanical shocks and noise caused by sudden changes. The control unit also has a self-test function, monitoring the output status of the control signals and the response of the actuators to ensure that control commands are executed correctly.

[0107] The above describes the AI-based directional heat dissipation monitoring method for diamond single crystals in the embodiments of this application. The following describes the AI-based directional heat dissipation monitoring system for diamond single crystals in the embodiments of this application. Please refer to [link to relevant documentation]. Figure 2 One embodiment of the artificial intelligence-based diamond single crystal directional heat dissipation monitoring system in this application includes:

[0108] The acquisition module 201 is used to acquire temperature parameters of the diamond single crystal heat dissipation system through a temperature sensor network and obtain heat dissipation temperature distribution data.

[0109] Modeling module 202 is used to dynamically model the heat dissipation characteristics of diamond single crystal based on the heat dissipation temperature distribution data, and generate a real-time heat flow control model.

[0110] Output module 203 is used to perform closed-loop calculation of heat dissipation control parameters based on the real-time heat flow control model and output fan speed and flow control commands.

[0111] The adjustment module 204 is used to adjust the cooling system actuator in real time based on the fan speed and flow control command, and to generate heat dissipation control feedback data;

[0112] The analysis module 205 is used to perform real-time analysis of the system temperature deviation based on the heat dissipation control feedback data, generate cooling system parameter correction amounts, and automatically adjust the heat dissipation control device based on the cooling system parameter correction amounts.

[0113] Through the collaborative efforts of the aforementioned components, a multi-point arrangement of temperature sensor networks in the diamond single-crystal heat dissipation system enables high-resolution dynamic monitoring of heat distribution. Compared to traditional heat dissipation assessment methods that rely solely on a limited number of temperature measurement points, this approach captures more detailed changes in heat diffusion, improving the detection accuracy and response speed of thermal anomalies. Secondly, utilizing dynamic modeling technology, a real-time heat flow control model is constructed based on the collected heat dissipation temperature distribution data. This effectively addresses thermal disturbances caused by changes in system load, significantly enhancing the system's adaptability and the real-time performance of thermal control. A multi-layer domain-adaptive neural network structure is introduced into the closed-loop computation stage. Through sub-modules such as feature extraction, domain adaptation, and classification, deep control features are extracted from historical control data and the real-time model, eliminating distribution differences. This algorithm not only enhances the system's ability to learn the relationship between control input and thermal response but also ensures the robustness of the algorithm model to new operating conditions. Especially in intelligent heat dissipation optimization, the AI ​​model does not merely provide a simple regression prediction method but makes causal reasoning-based optimization decisions on heat dissipation control parameters through domain-invariant feature mining and heat flow path modeling, thereby forming the optimal control strategy for fan speed and coolant flow rate. By driving the cooling actuator with PWM control and combining it with a PID feedback adjustment mechanism to achieve precise control, and further enabling intelligent analysis and correction of the system's operating status through an abnormal temperature difference identification mechanism in real-time feedback, the cooling system possesses highly automated and intelligent dynamic regulation capabilities. This invention not only demonstrates the coordinated control capabilities of sensors and actuators, but also achieves intelligent closed-loop control of the entire process from perception, modeling, decision-making to execution in the thermal management system by introducing artificial intelligence algorithms.

[0114] above Figure 2 The AI-based diamond single crystal directional heat dissipation monitoring system in this embodiment of the invention will be described in detail from the perspective of modular functional entities. The AI-based diamond single crystal directional heat dissipation monitoring device in this embodiment of the invention will be described in detail from the perspective of hardware processing.

[0115] Figure 3 This is a schematic diagram of the structure of an AI-based diamond single-crystal directional heat dissipation monitoring device 300 provided in an embodiment of the present invention. The AI-based diamond single-crystal directional heat dissipation monitoring device 300 can vary significantly due to different configurations or performance characteristics. It may include one or more central processing units (CPUs) 310 (e.g., one or more processors) and a memory 320, and one or more storage media 330 (e.g., one or more mass storage devices) storing application programs 333 or data 332. The memory 320 and storage media 330 can be temporary or persistent storage. The program stored in the storage media 330 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the AI-based diamond single-crystal directional heat dissipation monitoring device 300. Furthermore, the processor 310 may be configured to communicate with the storage media 330, executing a series of instruction operations in the storage media 330 on the AI-based diamond single-crystal directional heat dissipation monitoring device 300 to implement the steps of the aforementioned AI-based diamond single-crystal directional heat dissipation monitoring method.

[0116] The AI-based diamond single-crystal directional heat dissipation monitoring device 300 may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input / output interfaces 360, and / or one or more operating systems 331, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 3 The illustrated structure of the AI-based diamond single crystal directional heat dissipation monitoring device does not constitute a limitation on the AI-based diamond single crystal directional heat dissipation monitoring device provided by the present invention. It may include more or fewer components than illustrated, or combine certain components, or have different component arrangements.

[0117] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the artificial intelligence-based diamond single crystal directional heat dissipation monitoring method.

[0118] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0119] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an artificial intelligence-based diamond single-crystal directional heat dissipation monitoring device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0120] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A diamond single crystal directional heat dissipation monitoring method based on artificial intelligence, characterized in that, The method comprises: The diamond single crystal heat dissipation system is temperature parameter acquisition by a temperature sensor network, and heat dissipation temperature distribution data is obtained; According to the heat dissipation temperature distribution data, the dynamic modeling of the diamond single crystal heat dissipation characteristics is carried out, and a real-time heat flow control model is generated, including: the thermal conductivity coefficient value of the diamond single crystal at different temperature points is extracted from the heat dissipation temperature distribution data, and a thermal conductivity temperature response matrix is constructed; the thermal conductivity temperature response matrix is subjected to polynomial regression fitting to obtain continuous thermal conductivity data; according to the thermal anisotropy characteristics of the 100 crystal orientation, the continuous thermal conductivity data is decomposed according to the crystal orientation to obtain the main direction thermal conductivity component and the secondary direction thermal conductivity component; the thermal conduction characteristics of the diamond / Cu interface region are subjected to spatial discretization processing by a three-dimensional interpolation algorithm to form a set of thermal conduction boundary conditions; the main direction thermal conductivity component, the secondary direction thermal conductivity component and the set of thermal conduction boundary conditions are integrated to establish a three-dimensional thermal conduction mathematical model of the diamond heat dissipation material; the temperature field distribution of the three-dimensional thermal conduction mathematical model under the standard heat load is solved by the finite difference method to generate a real-time heat flow control model; According to the real-time heat flow control model, the heat dissipation control parameters are calculated in a closed loop, and fan speed and flow control instructions are outputted; Based on the fan speed and flow control instructions, the cooling system actuator is adjusted in real time to form heat dissipation control feedback data; According to the heat dissipation control feedback data, the system temperature deviation is analyzed in real time to generate a cooling system parameter correction amount, and the heat dissipation control device is automatically adjusted according to the cooling system parameter correction amount.

2. The artificial intelligence-based diamond single crystal orientation heat dissipation monitoring method according to claim 1, characterized in that, The temperature parameter acquisition of the diamond single crystal heat dissipation system by the temperature sensor network, and the heat dissipation temperature distribution data obtained, comprises: Platinum resistance temperature sensors are arranged at the preset positions of the 100 crystal face oriented growth diamond single crystal heat dissipator to form a temperature monitoring point array; A heat flux density sensor is installed at the interface between the 100 crystal face oriented growth diamond single crystal and the metal to obtain interface heat flux data; The signals collected by the temperature monitoring point array and the heat flux density sensor are synchronously collected and processed to generate an original data stream; The original data stream is subjected to three-level cascade filtering processing, and in turn, Butterworth low-pass filtering, median filtering and Kalman filtering are performed to obtain noise reduction temperature data; According to the noise reduction temperature data, a temperature field spatial distribution matrix, a heat flux density vector and a cooling medium parameter vector are calculated to generate multi-modal heat dissipation parameters; The multi-modal heat dissipation parameters are fused with the sensor position information to output heat dissipation temperature distribution data.

3. The AI-based diamond single crystal orientation heat dissipation monitoring method according to claim 1, characterized in that, According to the real-time heat flow control model, the heat dissipation control parameters are calculated in a closed loop, and fan speed and flow control instructions are outputted, comprising: The real-time heat flow control model and the historical temperature control data are respectively taken as source domain data and target domain data, and are inputted into a multi-layer domain adaptive neural network structure comprising a feature extraction layer, a domain adaptation layer and a classification layer; The source domain data and the target domain data are processed by wavelet packet decomposition and reconstruction algorithm to generate temperature control feature vectors; The temperature control feature vector is input into a three-layer feature extraction layer of a multi-layer domain adaptation neural network, each layer containing 64, 128 and 256 neurons in turn, and after being processed by a nonlinear activation function, deep feature representation data is obtained; The distribution difference between the source domain and the target domain is calculated using a multi-kernel maximum mean difference algorithm on the deep feature representation data in the domain adaptation layer, and the network weights are adjusted through back propagation to obtain domain-invariant control features; The domain-invariant control features are input into a three-layer classification network containing 128, 64 and 32 neurons, respectively, and the maximum probability value is used as a pseudo label for the target domain data to form a heat dissipation control optimization scheme; Based on the heat dissipation control optimization scheme, the optimal fan speed curve and flow control curve are calculated, and fan speed and flow control instructions are output.

4. The AI-based diamond single crystal orientation heat dissipation monitoring method according to claim 3, characterized in that, The heat dissipation control optimization scheme is converted into a thermal resistance network topology graph, and the heat source point and the heat dissipation boundary point of the diamond single crystal are marked; According to the anisotropic properties of the crystal surface of the diamond single crystal 100, a thermal transfer weight value is assigned to each node in the thermal resistance network topology graph; Starting from the heat source point, a heat path tracking algorithm is used to traverse the thermal resistance network topology graph, and all possible heat transfer paths are calculated; The thermal resistance of all possible heat transfer paths is calculated to obtain a set of path thermal resistance values; The heat transfer channel with the smallest thermal resistance is selected as the main heat dissipation channel from the set of path thermal resistance values, and the channel with the second lowest thermal resistance value is selected as the auxiliary heat dissipation channel; According to the distribution ratio of 70% heat load for the main heat dissipation channel and 30% heat load for the auxiliary heat dissipation channel, the required heat dissipation power of each channel is calculated, and the fan speed value and the cooling liquid flow value are determined based on the temperature gradient value and the required heat flow value of each channel, and the fan speed and flow control instructions are output. The fan speed and flow control instructions are converted into PWM control signals to drive the fan motor and the cooling liquid pump, and the cooling system actuators are started; 5. The artificial intelligence-based diamond single crystal orientation heat dissipation monitoring method according to claim 1, characterized in that, PID controller closed-loop adjustment is performed on the fan motor speed, and if the actual speed deviates from the target speed by more than 5%, the PWM duty cycle is adjusted until the speed reaches the set value; The cooling liquid pump flow is proportionally adjusted, and when the flow deviation exceeds the set threshold, it is determined whether there is pipe blockage, and the pump pressure is increased or a warning signal is sent; Real-time operating parameters of the cooling system actuators are collected, including fan speed, flow value, temperature value and power consumption value, and an execution state matrix is constructed; The execution state matrix is compared and analyzed with the cooling instructions to determine the degree of execution deviation and generate an execution effect evaluation value; The execution effect evaluation value is time-series correlated with the temperature monitoring point data to form heat dissipation control feedback data. ​ ​ 6. The artificial intelligence-based diamond single crystal orientation heat dissipation monitoring method according to claim 1, characterized in that, The real-time analysis of system temperature deviation according to the heat dissipation control feedback data, the generation of cooling system parameter correction amount, and the automatic adjustment of the heat dissipation control device according to the cooling system parameter correction amount, comprise: extracting actual temperature distribution characteristics from the heat dissipation control feedback data to construct a real-time temperature feature vector; calculating the ideal temperature distribution value under the corresponding working condition according to the theoretical heat dissipation characteristics of the diamond single crystal 100 crystal surface to generate an ideal temperature reference vector; differential calculation of the real-time temperature feature vector and the ideal temperature reference vector to obtain a temperature deviation matrix, if the temperature deviation matrix has three consecutive points with a deviation of more than 10°C, it is determined as abnormal temperature distribution; statistical analysis of the temperature deviation matrix to calculate the temperature deviation rate and the temperature abnormal area proportion, when the temperature deviation rate exceeds 20%, the emergency cooling program is triggered; According to the temperature deviation rate and the temperature abnormal area proportion, the cooling system working parameter is corrected and calculated to generate the fan speed correction amount, the flow correction amount and the radiator angle correction amount; The fan speed correction amount, the flow correction amount and the radiator angle correction amount are converted into control signals by the parameter adjustment control unit to automatically adjust the heat dissipation control device.

7. An artificial intelligence-based diamond single crystal directional heat dissipation monitoring system, characterized in that, The AI-based diamond single crystal directional heat dissipation monitoring system for implementing the AI-based diamond single crystal directional heat dissipation monitoring method according to any one of claims 1 to 6, comprises: The acquisition module is used for acquiring temperature parameters of the diamond single crystal heat dissipation system through a temperature sensor network to obtain heat dissipation temperature distribution data. The modeling module is used for dynamically modeling the heat dissipation characteristics of the diamond single crystal according to the heat dissipation temperature distribution data to generate a real-time heat flow control model, comprising: extracting the thermal conductivity value of the diamond single crystal at different temperature points from the heat dissipation temperature distribution data to construct a thermal conductivity temperature response matrix; polynomial regression fitting is performed on the thermal conductivity temperature response matrix to obtain continuous thermal conductivity data; according to the thermal anisotropy characteristics of the 100 crystal surface orientation, the continuous thermal conductivity data is decomposed according to the crystal orientation to obtain the main direction thermal conductivity component and the secondary direction thermal conductivity component; the thermal conduction characteristics of the diamond / Cu interface region are spatially discretized by a three-dimensional interpolation algorithm to form a set of thermal conduction boundary conditions; the main direction thermal conductivity component, the secondary direction thermal conductivity component and the set of thermal conduction boundary conditions are integrated to establish a three-dimensional thermal conduction mathematical model of the diamond heat dissipation material; the temperature field distribution of the three-dimensional thermal conduction mathematical model under standard heat load is solved by finite difference method to generate a real-time heat flow control model; The output module is used for closed-loop calculation of the heat dissipation control parameters according to the real-time heat flow control model to output fan speed and flow control instructions; The adjustment module is used for real-time adjustment of the cooling system actuator based on the fan speed and flow control instructions to form heat dissipation control feedback data; The analysis module is used for real-time analysis of system temperature deviation according to the heat dissipation control feedback data, generation of cooling system parameter correction amount, and automatic adjustment of the heat dissipation control device according to the cooling system parameter correction amount.

8. An artificial intelligence-based diamond single crystal directional heat dissipation monitoring device, characterized in that, An artificial intelligence-based diamond single crystal directional heat dissipation monitoring method is provided, including a memory and a processor, the memory stores a computer program capable of running on the processor, and the processor implements the artificial intelligence-based diamond single crystal directional heat dissipation monitoring method according to any one of claims 1 to 6 when executing the computer program.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program causes the processor to execute the artificial intelligence-based diamond single crystal directional heat dissipation monitoring method according to any one of claims 1 to 6 when the computer program is run by the processor.

Images