System and method for evaluating junction temperature and thermal resistance of integrated circuit based on digital twinning technology

Abstract

The invention discloses an integrated circuit junction temperature and thermal resistance evaluation system and method based on a digital twinning technology, and the method comprises the steps: extracting a temperature cloud picture from a simulation calculation result, and extracting a thermal flow line diagram from the temperature cloud picture based on a heat dissipation path, carrying out discrete sampling on the heat flow line diagram by using a deep learning data analysis and coding technology, and then carrying out embedded coding on each piece of sampled heat flow point data (a flow point position and a heat flow point temperature); in this way, the rectangular frame position data of the hot spot region is automatically identified according to the graph walk significant aggregation representation between the data embedding coding features of each heat flow point. Through the mode, the influence of a fine structure (such as a connecting wire, a welding spot and the like) on a heat transfer path can be captured, the complex heat dissipation requirement can be better met, and it is ensured that the heat dissipation requirement is effectively met.

Description

Technical Field

[0001] The present application relates to the field of intelligent evaluation, and more specifically, to an integrated circuit junction temperature and thermal resistance evaluation system and method based on digital twin technology. Background Art

[0002] In the field of integrated circuits (ICs), especially for complex three-dimensional integrated circuits (3D-ICs), accurate evaluation of junction temperature and thermal resistance is of vital importance to circuit performance and reliability. With the advancement of semiconductor technology, modern ICs not only achieve extremely high integration on a two-dimensional plane, but also form a three-dimensional structure by stacking multiple layers of chips. Therefore, effective thermal management is essential to ensure the reliability, stability and life of integrated circuits.

[0003] Traditional methods rely on simplified geometric models and empirical formulas in integrated circuit thermal analysis, which is inadequate when faced with complex multi-layer stacked structures. Simplified models cannot capture the impact of fine structures on heat transfer paths, resulting in reduced prediction accuracy. Especially under the trend of high-density integration and heterogeneous integration, traditional methods are difficult to adapt to increasingly complex heat dissipation needs. In addition, traditional methods can often only process static or simple time series data. For heat flow problems involving a large amount of spatiotemporal correlation, such as dynamic heat distribution in data centers, it is difficult for them to capture all potential patterns and trends.

[0004] Therefore, an optimized IC junction temperature and thermal resistance evaluation scheme is desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides an integrated circuit junction temperature and thermal resistance evaluation system and method based on digital twin technology, which extracts a temperature cloud map from the simulation calculation results, and extracts a heat flow line diagram from the temperature cloud map based on the heat dissipation path, and uses deep learning data analysis and encoding technology to discretely sample the heat flow line diagram, and then embeds and encodes the sampled heat flow point data (flow point position and heat flow point temperature), so as to automatically identify the rectangular frame position data of the hot spot area according to the graph walk significant aggregation representation between the embedded encoding features of each heat flow point data. In this way, the influence of fine structures (such as connecting wires, solder joints, etc.) on the heat transfer path can be captured, and the complex heat dissipation requirements can be better adapted to ensure the effective satisfaction of the heat dissipation requirements. In addition, it can comprehensively capture the change pattern and trend of heat distribution, and respond to transient changes in real time, so as to find potential hot spots or abnormal conditions, effectively improve the integrated circuit junction temperature monitoring accuracy and speed, and enhance the ability to evaluate the thermal resistance of the integrated circuit.

[0006] According to one aspect of the present application, a method for evaluating junction temperature and thermal resistance of an integrated circuit based on digital twin technology is provided, which includes: Step 1: establishing an integrated circuit digital twin model; Step 2: constructing a single heat dissipation path in the integrated circuit digital twin model; Step 3: monitoring the temperature data collected by the temperature sensor in the integrated circuit, and transmitting the temperature data collected by the temperature sensor to the integrated circuit digital twin model; Step 4: loading input power on the digital twin model based on different test conditions, and determining boundary conditions and grid parameters to perform simulation calculations to obtain simulation calculation results; Step 5: determining hotspot areas based on the simulation calculation results; Step 6: based on the hotspot areas, improving the integrated circuit digital twin model to obtain an improved integrated circuit digital twin model; Step 7: monitoring the temperature data of the actual integrated circuit in real time, and feeding back the temperature data to the improved integrated circuit digital twin model in real time.

[0007] According to another aspect of the present application, a system for evaluating junction temperature and thermal resistance of an integrated circuit based on digital twin technology is provided, which includes: Digital twin model building module, used to build integrated circuit digital twin models; a heat dissipation path construction module, used to construct a single heat dissipation path in the integrated circuit digital twin model; A data transmission module, used to install a temperature sensor in an actual integrated circuit and transmit the temperature data collected by the temperature sensor to the digital twin model of the integrated circuit; A simulation calculation module, used to load input power to the digital twin model based on different test conditions, and determine boundary conditions and grid parameters to perform simulation calculations to obtain simulation calculation results; A hot spot area determination module, used to determine the hot spot area based on the simulation calculation results; A digital twin model improvement module, used for improving the integrated circuit digital twin model based on the hot spot area to obtain an improved integrated circuit digital twin model; A data feedback module is used to monitor the temperature data of the actual integrated circuit in real time and feed back the temperature data to the improved integrated circuit digital twin model in real time.

[0008] Compared with the prior art, the present application provides an integrated circuit junction temperature and thermal resistance evaluation system and method based on digital twin technology, which extracts a temperature cloud map from the simulation calculation results, and extracts a heat flow line diagram from the temperature cloud map based on the heat dissipation path, and uses deep learning data analysis and encoding technology to discretely sample the heat flow line diagram, and then embeds and encodes the sampled heat flow point data (flow point position and heat flow point temperature), so as to automatically identify the rectangular frame position data of the hot spot area according to the graph walk significant aggregation representation between the embedded encoding features of each heat flow point data. In this way, the influence of fine structures (such as connecting wires, solder joints, etc.) on the heat transfer path can be captured, and the complex heat dissipation requirements can be better adapted to ensure the effective satisfaction of the heat dissipation requirements. In addition, it can comprehensively capture the change pattern and trend of heat distribution, and respond to transient changes in real time, so as to discover potential hot spots or abnormal conditions, effectively improve the accuracy and speed of integrated circuit junction temperature monitoring, and enhance the ability to evaluate the thermal resistance of integrated circuits. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 It is a flow chart of a method for evaluating junction temperature and thermal resistance of an integrated circuit based on digital twin technology according to an embodiment of the present application; Figure 2 A data flow diagram of an integrated circuit junction temperature and thermal resistance evaluation method based on digital twin technology according to an embodiment of the present application; Figure 3 It is a flowchart of sub-step S5 of the integrated circuit junction temperature and thermal resistance evaluation method based on digital twin technology according to an embodiment of the present application; Figure 4 It is a block diagram of an integrated circuit junction temperature and thermal resistance evaluation system based on digital twin technology according to an embodiment of the present application. DETAILED DESCRIPTION

[0011] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0012] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0013] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0014] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0015] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0016] In the technical solution of the present application, a method for evaluating the junction temperature and thermal resistance of an integrated circuit based on digital twin technology is proposed. Figure 1 This is a flow chart of a method for evaluating junction temperature and thermal resistance of an integrated circuit based on digital twin technology according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the integrated circuit junction temperature and thermal resistance evaluation method based on digital twin technology according to an embodiment of the present application. Figure 1 and Figure 2As shown, according to the embodiment of the present application, the integrated circuit junction temperature and thermal resistance evaluation method based on digital twin technology includes the following steps: S1, establishing an integrated circuit digital twin model; S2, constructing a single heat dissipation path in the integrated circuit digital twin model; S3, monitoring the temperature data collected by the temperature sensor in the integrated circuit, and transmitting the temperature data collected by the temperature sensor to the integrated circuit digital twin model; S4, loading input power on the digital twin model based on different test conditions, and determining boundary conditions and grid parameters to perform simulation calculations to obtain simulation calculation results; S5, determining hot spots based on the simulation calculation results; S6, based on the hot spots, improving the integrated circuit digital twin model to obtain an improved integrated circuit digital twin model; S7, real-time monitoring of the temperature data of the actual integrated circuit, and feeding back the temperature data to the improved integrated circuit digital twin model in real time.

[0017] In particular, S1, establishes a digital twin model of an integrated circuit. In a specific example, in the process of establishing a digital twin model of an integrated circuit, it is necessary to comprehensively collect and integrate information from design files, packaging files, and test conditions. This information constitutes the basic parameter system of the model, which covers material properties (such as material density ρ, elastic modulus E, Poisson's ratio ν, thermal conductivity λ) and geometric information (such as the geometry, size, and fit of each part of the structure). For those parameters that are not provided in the original data, they need to be supplemented by other measurement methods, such as using X-rays to measure geometric information or using specific experimental devices to measure thermal conductivity. Next, create a digital representation that covers all the above key parameters and can be used for subsequent simulation calculations. The model should be as faithful to the physical entity as possible, including its internal structure, material distribution, and external environmental conditions.

[0018] In particular, S2, a single heat dissipation path is constructed in the digital twin model of the integrated circuit. It should be understood that the purpose of constructing a single heat dissipation path is to ensure that heat can be transferred from the inside of the integrated circuit to the external environment as intended by the designer, rather than being randomly dissipated or concentrated in undesirable areas. Specifically, a controlled heat conduction path is implemented in the digital model to ensure effective heat discharge, and no reduction in heat dissipation efficiency or hot spot formation due to other unexpected paths. To achieve this, in the initial integrated circuit digital model, adiabatic parameters are set for parts of non-target heat dissipation paths, that is, those areas that should not be used as main heat dissipation channels. This means that these parts will be regarded as having no heat conduction capacity during simulation calculations, thereby preventing the possibility of heat loss through them. This approach effectively guides heat to be transferred only through the predetermined main heat dissipation path. For the terminal part selected as the main heat dissipation path, a surface constant heat transfer parameter is set. This usually refers to the interface between the heat sink or housing and the surrounding air, which is set to have a fixed heat transfer coefficient (also called convection heat transfer coefficient). This is done to simulate the heat dissipation conditions in actual applications, such as natural convection, forced air cooling or other forms of cooling mechanisms. In this way, it can be ensured that heat can be dissipated from the specified location at a stable rate during simulation calculations, similar to the situation in the real world. In particular, real-time data transmission and feedback correction can be achieved by using IoT technology. This means that temperature sensors and other necessary monitoring devices are installed in the actual integrated circuit, and the data collected by these devices will be transmitted to the digital twin model in real time. If the monitoring data shows that an abnormal temperature rise has occurred in some places, the thermal management strategy can be optimized by adjusting the heat dissipation path settings within the model, such as changing the material properties or geometric structure on the heat dissipation path, or even re-evaluating the design of the entire heat dissipation solution.

[0019] In particular, the S3 monitors the temperature data collected by the temperature sensor in the integrated circuit, and transmits the temperature data collected by the temperature sensor to the digital twin model of the integrated circuit. It should be understood that in order to obtain the most accurate data, it is necessary to monitor the sensors in those key areas that are sensitive to temperature changes or prone to hot spots. This may include the surface of the chip, the place near the heat source inside the package, and any other parts that may affect the overall performance. In addition, according to the specific application requirements, other types of external auxiliary monitoring equipment, such as stress sensors or humidity sensors, can also be installed to provide more comprehensive working status information. The selection and arrangement of all these sensors must be considered not to interfere with the normal operation of the circuit, and also to facilitate subsequent maintenance and replacement. Then, it is physically connected to the circuit board. Usually, the temperature sensor is directly integrated at a specified location inside the chip, or connected to it through a specially designed interface. For some advanced applications, miniaturized wireless sensor nodes may be used, which can be embedded in a very small space without affecting the circuit layout. Such sensors can not only measure temperature, but also send data through wireless communication protocols, reducing the complexity of wiring. After installation, the sensor begins to collect temperature information from various key points of the integrated circuit. In order to enable this data to be processed in a timely manner in the digital twin model, the document points out that the Internet of Things (IoT) technology should be used to establish a stable data transmission channel. This means that each sensor needs to be equipped with an appropriate communication module, such as Wi-Fi, Bluetooth, Zigbee or other low-power wide area network (LPWAN) technologies, to facilitate communication with the external network. When the sensor detects a temperature change, it immediately converts the data into a digital signal and sends it out through the selected communication protocol. The receiving end is a platform with integrated data processing capabilities, which can be a local server or a cloud service. On this platform runs a software system developed specifically for the digital twin model, which is responsible for receiving data streams from different sensors and parsing, storing and preliminarily analyzing them. Since the number of sensors may be large, the platform also needs to have efficient data management and synchronization capabilities to ensure that all received data can be correctly integrated into the digital twin model. In particular, if the data of a sensor is found to be abnormal, such as inconsistent with other adjacent sensor readings, or exceeds the preset safety range, the system will trigger an alarm and may automatically take measures to correct the problem. For example, increase the speed of the radiator fan, change the coolant flow, or even re-evaluate the design of the heat dissipation path. This dynamic response mechanism helps maintain stable operation of the system and prevents overall failure due to local overheating.

[0020] In particular, the S4 loads the input power to the digital twin model based on different test conditions, and determines the boundary conditions and grid parameters to perform simulation calculations to obtain simulation calculation results. In a specific example, first, in order to simulate various situations in the actual operating environment, a series of test conditions are defined according to specific usage scenarios. These conditions may include different levels of load conditions, such as light load, rated load and heavy load, etc.; they may also cover temperature changes, voltage fluctuations and other factors that may affect the thermal behavior of the circuit. Each condition corresponds to a specific set of input power values, which will be applied to the digital twin model to reproduce the actual operating conditions. Next, set the corresponding boundary conditions for the digital twin model. Boundary conditions refer to the way the model interacts with the external environment, which has a vital impact on the accuracy of the simulation. The document mentions that according to the actual working conditions of the integrated circuit, fixed constraints are set for structures such as bonding and molding that have no relative displacement, that is, the relative movement between these parts is restricted. For those places where there is a possibility of relative movement, friction constraints are used to allow a certain degree of sliding or rolling contact. In addition, adiabatic conditions are set for parts of non-target heat dissipation paths to prevent heat from being lost through these areas, while constant surface heat exchange parameters are set at the terminals of the main heat dissipation paths to simulate the real heat dissipation effect. Such boundary condition settings can effectively simulate the process of heat transfer in the actual environment and improve the authenticity of the simulation. While determining the boundary conditions, it is also necessary to consider how to divide the grid. The selection of grid parameters directly affects the calculation efficiency and accuracy of the simulation. The document points out that coarse grid division is used in the structure of non-key analysis parts, which can reduce the amount of calculation without affecting the overall results; while finer grids are used in hot spots or other important areas to capture changes in local details. After the initial grid division is completed, multiple iterative optimizations will be performed until the error of the calculation results is controlled within 5%, and the most appropriate grid parameter configuration is finally determined. This differentiated grid strategy not only improves the speed of simulation, but also ensures the accuracy of data in key areas. After completing the setting of boundary conditions and grid parameters, you can submit the calculation and start the simulation process. At this stage, the digital twin model will simulate the working state of the integrated circuit based on the previously loaded input power and the set boundary conditions, and calculate the junction temperature and thermal resistance data. These data reflect the temperature distribution at various locations inside the integrated circuit under given conditions, as well as the ease of heat conduction from the heat source to the heat sink. By comparing the simulation results under different test conditions, engineers can obtain important information about thermal management performance, such as which areas are prone to hot spots and whether there are design defects. Finally, in order to verify the effectiveness of the improvement measures, the document emphasizes that the improved design will be simulated again. This step ensures that any design changes can bring the expected performance improvement without introducing new problems.If the new design does achieve the best thermal management performance, it can be applied to the actual product. Otherwise, it needs to be adjusted and optimized until a satisfactory result is achieved.

[0021] In particular, the S5 determines the hot spot area based on the simulation calculation result. Specifically, in a specific example of the present application, Figure 3 As shown, the S5 includes: S51, extracting a temperature cloud map from the simulation calculation results; S52, extracting a heat flow line diagram from the temperature cloud map based on the heat dissipation path; S53, discretely sampling the heat flow line diagram to obtain a sequence of heat flow point data, the heat flow point data including the heat flow point position and the heat flow point temperature; S54, embedding and encoding each heat flow point data in the sequence of heat flow point data to obtain a sequence of heat flow point data embedded coding features; S55, performing graph walk autocorrelation gated significant aggregation on the sequence of heat flow point data embedded coding features to obtain a heat flow point set significant aggregation feature; S56, obtaining a recognition result based on the heat flow point set significant aggregation feature, and the recognition result is the rectangular frame position data of the hot spot area.

[0022] Specifically, the S51 extracts a temperature cloud map from the simulation calculation results. It should be understood that the temperature of different regions inside the integrated circuit has different forms of expression, and considering that the temperature cloud map is a visualization tool used in thermal analysis, it uses color gradients to represent the temperature distribution of different positions on the surface or inside of an object. Based on this, in the technical solution of the present application, a temperature cloud map is extracted from the simulation calculation results. That is, the temperature cloud map can help quickly locate high-temperature areas in the integrated circuit (IC), which may be caused by excessive heat accumulation, which poses a threat to the long-term reliability of the IC. Identifying these hot spots is essential for taking appropriate heat dissipation measures. In a specific example, after the digital twin model is loaded with input power, boundary conditions and grid parameters are determined, and simulation calculations are performed, a series of data on junction temperature and thermal resistance will be obtained. These data contain temperature information of each node or unit inside the integrated circuit. In order to better understand and analyze these data, especially to identify possible hot spots or other potential problem areas, it is necessary to convert these values ​​into a graphical representation-a temperature cloud map. Before generating a temperature cloud map, it is usually necessary to perform some necessary preprocessing on the original simulation calculation results to ensure that the final generated image can accurately reflect the actual situation and is easy for the observer to understand. First, remove outliers or invalid data points to ensure data quality, which is part of data cleaning. Next, ensure that the data format is suitable for generating temperature cloud maps, such as converting the data to a specific coordinate system so that it can correspond to the location in the physical space. This is the process of data format conversion. In addition, if the data density in some areas is insufficient, the interpolation algorithm can be used to increase the data points to make the temperature distribution smoother and more continuous, avoiding misleading visual effects caused by sparse data. This is the work of data interpolation. After completing data preprocessing, the next step is to use professional software or custom scripts to map the processed temperature data into a two-dimensional or three-dimensional space to form a temperature cloud map. The key to this step is to choose an appropriate color scheme, correctly perform spatial mapping, and apply effective rendering techniques. Choosing a suitable color gradient according to the temperature range (such as cold colors represent low temperature areas and warm colors represent high temperature areas) can enhance the visualization effect. Reasonable color selection can help users quickly identify the trend and degree of temperature changes. At the same time, each temperature data point is accurately mapped to the corresponding spatial position. For complex three-dimensional structures, the relative position relationship between layers also needs to be considered to ensure that the temperature cloud map truly reproduces the actual temperature distribution inside the IC. In order to better display the temperature change trend, various rendering techniques can be used, such as isotherms, transparency adjustment, and surface shading. For example, isotherms can clearly show areas with the same temperature; adjusting transparency allows users to more intuitively see the temperature differences between different layers; and surface shading expresses temperature changes by changing the color depth.

[0023] Specifically, the S52 extracts a heat flow diagram from the temperature cloud map based on the heat dissipation path. Here, in order to provide a more detailed perspective to observe and understand the heat transfer process within the IC, so as to optimize the design of the model, in the technical solution of the present application, a heat flow diagram is extracted from the temperature cloud map based on the heat dissipation path. It should be understood that a heat flow diagram is a visualization tool that depicts the flow path of heat inside or on the surface of an object. Heat flow lines are similar to power lines in an electric field or streamlines in fluid mechanics, which indicate the direction and relative intensity of heat transfer. In an integrated circuit (IC), a heat flow diagram can show which paths are the main conduction channels for heat and how heat is transmitted from a high temperature area (such as a transistor node) to a low temperature area (such as a heat sink or ambient air). That is, through a heat flow diagram, it is possible to more intuitively understand how heat is transferred inside the IC, including whether the heat is concentrated in certain areas or whether there are undesirable heat conduction paths, which helps to deeply analyze the effectiveness of the thermal management design of the IC. And identifying the main heat flow paths is very important for optimizing the heat dissipation design of the IC and its packaging. For example, if it is found that most of the heat is conducted through a specific path, then the heat dissipation measures can be enhanced on that path, such as adding thermal conductive materials or improving the design of the heat sink. In a specific example, the thermal gradient field in the entire IC is calculated. The thermal gradient reflects the rate of change of temperature with spatial position, that is, the speed at which the temperature rises or falls at each point. In an ideal case, the heat flow always flows in the direction of the negative thermal gradient, that is, from the high temperature area to the low temperature area. Therefore, by calculating the temperature difference between each grid cell and applying appropriate mathematical formulas (such as the finite difference method), a vector field describing the heat flow in the entire IC can be obtained. Each vector in this vector field points to the direction of the steepest temperature drop at that point, representing the direction of heat flow here. The calculation of the thermal gradient field is the basis for extracting heat flow lines, which provides the necessary information support for subsequent steps. After obtaining the thermal gradient field, the heat flow lines are extracted. The heat flow lines are curves formed by starting from a certain starting point and gradually moving forward in the direction indicated in the thermal gradient field. In order to ensure the authenticity and reliability of the results, multiple representative positions are usually selected as starting points, and a certain step size control is maintained during the tracking process. In addition, some rules can be introduced to handle special situations, such as bypassing or terminating the current line and opening a new heat flow line when encountering an obstacle. The resulting heat flow line diagram will clearly show the flow trajectory of heat throughout the IC, which can help identify the main heat conduction paths and possible bottlenecks or hot spots.

[0024] Specifically, the S53 performs discrete sampling on the thermal flow diagram to obtain a sequence of thermal flow point data, and the thermal flow point data includes the thermal flow point position and the thermal flow point temperature. Considering that the thermal flow diagram is a continuous graphical representation and contains a large amount of information, directly processing the continuous thermal flow diagram may be very computationally intensive. Therefore, in order to convert the continuous heat flow information into a discrete set of data points to facilitate subsequent analysis, processing and application, in the technical solution of the present application, the thermal flow diagram is discretely sampled to simplify the complex information into a limited number of thermal flow point data to obtain a sequence of thermal flow point data. In this way, the temperature information of each position can be analyzed more meticulously and quantitatively, while significantly reducing the amount of data to be processed, so as to better characterize the temperature characteristics of different regions.

[0025] Specifically, the S54 embeds and encodes each heat flow point data in the sequence of the heat flow point data to obtain a sequence of heat flow point data embedding coding features. Considering that there are mutual relationships and complex connections between each heat flow point data, such as adjacent heat flow points should also be close in the embedding space, which helps to retain the continuity and directionality of the heat flow path. Therefore, in order to convert the original heat flow point data into a more advanced and expressive form, so as to capture and mine the complex relationship between the heat flow point data, such as spatial position, temperature gradient, etc., in the technical solution of the present application, each heat flow point data in the sequence of the heat flow point data is embedded and encoded to obtain a sequence of heat flow point data embedding coding vectors, so that the characteristics of each heat flow point and its role in the overall heat flow path can be more accurately described. In particular, in a specific embodiment of the present application, the heat flow point data embedding matrix can be used to embed and encode each heat flow point data in the sequence of the heat flow point data to effectively extract the implicit relationship between the heat flow points, and obtain a sequence of heat flow point data embedding coding vectors as the sequence of heat flow point data embedding coding features. In particular, the heat flow point data embedding matrix is ​​constructed based on a large amount of historical data.

[0026] Specifically, the S55 performs graph walk autocorrelation gating significant aggregation on the sequence of the heat flow point data embedded coding features to obtain the heat flow point set significant aggregation features. Considering that the heat flow point data usually has strong temporal and spatial dependence, the temperature changes between adjacent heat flow points often affect each other to form a complex spatiotemporal pattern. However, it is difficult for traditional feature extraction methods to fully capture these dynamic associations. Based on this, in the technical solution of the present application, a graph walk autocorrelation gating significant aggregation mechanism is introduced to consider the characteristics of each heat flow point data embedded coding feature itself and its interaction with the surrounding nodes, so as to generate a heat flow point set significant aggregation feature that comprehensively reflects the common behavior pattern of all heat flow points. In particular, the graph walk mechanism allows the model to gradually explore the relationship between nodes along the heat flow path, while the autocorrelation gating helps the model identify and emphasize those features with significant autocorrelation, that is, features that show consistency with changes in time or spatial position, thereby enhancing the understanding of the thermal behavior of integrated circuits.

[0027] Specifically, the specific process of performing graph walk autocorrelation gated significant aggregation on the sequence of heat flow point data embedded coding features includes: first, each heat flow point data embedded coding vector in the sequence of heat flow point data embedded coding vectors is input into the hyperbolic space mapper to obtain a sequence of heat flow point data embedded coding vectors after hyperbolic space mapping. In particular, the hyperbolic space mapper uses hyperbolic geometric characteristics to maintain the hierarchical structure and long-tail distribution in the heat flow point data, thereby enhancing the relative distance representation ability between heat flow point data embedded coding vectors, so that similar heat flow point data embedded coding vectors are more closely clustered together, while different heat flow point data embedded coding vectors are effectively separated. In a specific example, each heat flow point data embedded coding vector in the sequence of heat flow point data embedded coding vectors is input into the hyperbolic space mapper to obtain a sequence of heat flow point data embedded coding vectors after hyperbolic space mapping according to the following formula; wherein the formula is: in, is a sequence of heat flow point data embedded coding vectors, , , , and are the first, second, and third heat flow point data embedded in the sequence of the coding vector. , and The heat flow point data is embedded in the coding vector, and are the first weight matrix in hyperbolic space and the second weight matrix in hyperbolic space respectively, for After hyperbolic space mapping, the heat flow point data is embedded in the coding vector.

[0028] Next, based on the sequence of heat flow point data embedded coding vectors after hyperbolic space mapping, the graph walk topology features are extracted to obtain the heat flow point data graph walk topology feature matrix. Specifically, the Poincare distance between any two heat flow point data embedded coding vectors after hyperbolic space mapping in the sequence of heat flow point data embedded coding vectors after hyperbolic space mapping is first calculated to obtain the heat flow point data graph walk topology matrix. The Poincare distance measures the actual distance between two points in the hyperbolic space. This operation helps to identify the topological relationship between the heat flow point data embedded coding vectors after hyperbolic space mapping, and builds a graph structure based on this. The graph structure reflects the intrinsic connectivity of the sequence of heat flow point data embedded coding vectors after hyperbolic space mapping. The heat flow point data graph walk topology feature matrix records the relative proximity between all heat flow point data embedded coding vectors after hyperbolic space mapping in the sequence of heat flow point data embedded coding vectors after hyperbolic space mapping, which not only reflects the direct connection between the heat flow point data nodes after hyperbolic space mapping, but also implies higher-order structural information. Then, the heat flow point data graph walk topology matrix is ​​subjected to hole convolution encoding to capture the topological features at different scales, so as to obtain the heat flow point data graph walk topology feature matrix. In a specific example, based on the sequence of heat flow point data embedding encoding vectors after hyperbolic space mapping, the graph walk topology features are extracted using the following graph walk topology feature extraction formula to obtain the heat flow point data graph walk topology feature matrix; wherein, the formula is: in, for After hyperbolic space mapping, the heat flow point data is embedded in the coding vector. is the square of the vector's magnitude, is the inverse hyperbolic cosine function, For calculation and The Poincare distance between is the first in the heat flow point data graph walk topology matrix The eigenvalues ​​at the positions, is the heat flow point data graph walk topology matrix, is the dilated convolutional coding, It is the topological characteristic matrix of the heat flow point data graph.

[0029] Then, based on the walk topology feature matrix of the heat flow point data graph, the sequence of heat flow point data embedded coding vectors after the hyperbolic space mapping is contextually semantically enhanced to obtain a sequence of heat flow point data contextual semantically enhanced feature vectors. Specifically, the walk topology feature matrix of the heat flow point data graph and the sequence of heat flow point data embedded coding vectors after the hyperbolic space mapping are input into a global context walk encoder based on a graph convolutional neural network model to obtain a sequence of heat flow point data contextual semantically enhanced feature vectors. The global context walk encoder integrates the information of the local neighborhood of the heat flow point data graph through a graph convolution operation, and performs a walk over the entire range of the heat flow point data graph to extract a sequence of heat flow point data contextual semantically enhanced feature vectors with rich contextual information. That is, a graph convolutional neural network (GCN) is used to integrate information from the entire heat flow point data graph. In a specific example, based on the walk topology feature matrix of the heat flow point data graph, the sequence of heat flow point data embedded coding vectors after hyperbolic space mapping is contextually semantically enhanced using the following contextual semantic enhancement formula to obtain a sequence of heat flow point data contextual semantic enhancement feature vectors; wherein the contextual semantic enhancement formula is: in, is the graph convolutional coding, is the first in the sequence of the semantic enhancement feature vector of the hot flow point data context The contextual semantics of the heat flow point data is enhanced with feature vectors.

[0030] Furthermore, based on the sequence of the hot flow point data context semantic enhancement feature vectors, the sequence of the hot flow point data embedded coding vectors is subjected to significant gated weighted aggregation to obtain a hot flow point set significant aggregation feature vector as the hot flow point set significant aggregation feature. Specifically, in the example of the present application, first, each group of corresponding hot flow point data context semantic enhancement feature vectors and hot flow point data embedded coding vectors in the sequence of the hot flow point data context semantic enhancement feature vectors and the sequence of the hot flow point data embedded coding vectors are input into an autocorrelation gating unit to obtain a sequence of hot flow point data autocorrelation gated significant confidence factors; the autocorrelation gating unit is used to measure the consistency and correlation between the hot flow point data context semantic enhancement feature vector and the hot flow point data embedded coding vector, thereby generating a sequence of hot flow point data autocorrelation gated significant confidence factors. Next, the sequence of the autocorrelation gated significant confidence factors of the hot flow point data is input into the normalization unit based on the Softmax function to obtain the sequence of the autocorrelation gated significant confidence weight factors of the hot flow point data. Finally, based on the sequence of the autocorrelation gated significant confidence weight factors of the hot flow point data, the position-weighted sum of the sequence of the hot flow point data embedded coding vectors is calculated to obtain the significant aggregation feature vector of the hot flow point set. The significant aggregation feature vector of the hot flow point set integrates the information of all the hot flow point data embedded coding vectors and emphasizes those parts that best represent the significant characteristics of the input set. In a specific example, based on the sequence of the hot flow point data contextual semantic enhancement feature vectors, the sequence of the hot flow point data embedded coding vectors is significantly gated and weighted aggregated using the following significant gated weighted aggregation formula to obtain the significant aggregation feature vector of the hot flow point set; wherein, the significant gated weighted aggregation formula is: in, is the normalization function, To subtract by position point, It is calculated and The autocorrelation gated significance confidence factor between for and The heat flow point data autocorrelation gated significant confidence factor between represents the exponential function value with the natural constant e as the base, is the number of heat flow point data autocorrelation gated significant confidence factors in the sequence of heat flow point data autocorrelation gated significant confidence factors, yes The corresponding heat flow point data autocorrelation gated significant confidence weight factor, is the number of vectors in the sequence of the heat flow point data embedded in the encoding vector, For point-by-position multiplication, is the significant aggregation feature vector of the heat flow point set.

[0031] Specifically, the S56 obtains a recognition result based on the significant aggregation features of the heat flow point set, and the recognition result is the rectangular frame position data of the hot spot area. In the technical solution of the present application, the significant aggregation feature vector of the heat flow point set is input into a hot spot area identifier based on a decoder to obtain the recognition result. That is, the significant aggregation features of the heat flow point set obtained by graph walk significant aggregation using the sequence of the heat flow point data embedded in the encoding feature are decoded to automatically identify the rectangular frame position data of the hot spot area.

[0032] In a preferred example, considering that each heat flow point data embedded coding vector in the sequence of the heat flow point data embedded coding vectors respectively represents the low-dimensional embedded coding features of the heat flow point data, when performing feature sequence aggregation analysis based on graph walk autocorrelation gating, the node feature heterogeneity and correlation sparsity of each heat flow point will lead to the lack of decision-making of the aggregated feature instance of the significant aggregated feature vector of the heat flow point set, thereby affecting the accuracy of the recognition result obtained by the hot spot area identifier based on the decoder.

[0033] Based on this, before the heat flow point set significant aggregation feature vector is input into the decoder-based hot spot region identifier, the heat flow point set significant aggregation feature vector is first optimized, including the steps of: Determine the number of zero eigenvalues ​​in the significant aggregation eigenvector of the heat flow point set and zero eigenvalue , and the zero eigenvalue and the number of zero eigenvalues Multiply and divide respectively to obtain the significant aggregation field weight value of the first heat flow point set and the second heat flow point set significant aggregation field weight value ; Calculate the square root of the sum of squares of all eigenvalues ​​of the heat flow point set significant aggregation eigenvector to obtain the heat flow point set significant aggregation full mode weight , Each eigenvalue representing a significantly aggregated eigenvector of the heat flow point set; With the zero characteristic factor value Each eigenvalue of the significantly aggregated eigenvector of the set of heat flow points is calculated as an index The power function , and multiplied by the significant aggregation field weight value of the first heat flow point set To obtain the first heat flow point set significant aggregation intermediate value ; Calculate each eigenvalue of the significant aggregation eigenvector of the heat flow point set The second heat flow point set has a significant aggregate field weight value and the heat flow point set significantly aggregates the full mode weight The product of obtains the significant aggregate intermediate value of the second heat flow point set ; The first set of hot flow points is significantly aggregated into the middle value The intermediate values ​​are significantly aggregated with the second set of hot flow points The weighted difference The optimized set of heat flow points is composed of significantly aggregated feature vectors, where and Represents different weight parameters. Finally, the optimized heat flow point set significant aggregation feature vector is input into the decoder-based hot spot area identifier to obtain the identification result.

[0034] Accordingly, based on the vector field fitting law of high-dimensional manifold, the isolated zeros and their differentiable factors in the significantly aggregated feature vector of the heat flow point set are used as the differentiable dimension of the high-dimensional manifold, so as to fix the normal direction pointing to the center of the feature distribution near the local coordinates represented by the eigenvalues ​​of the significantly aggregated feature vector of the heat flow point set, thereby achieving effective alignment of the feature distribution pattern of the significantly aggregated feature vector of the heat flow point set to the probability density differentiable field to avoid invalid repetition in the probability mapping process caused by the unstructured characteristics of the distribution of the significantly aggregated feature vector of the heat flow point set with rich interactive responses, thereby improving the degree of instance determination of the eigenvalues ​​of the significantly aggregated feature vector of the heat flow point set under similarity constraints, that is, the degree of significance of the eigenvalues ​​as instances for decoding regression decisions, and improving the accuracy of the recognition results obtained by inputting the significantly aggregated feature vector of the heat flow point set into the hot spot area identifier based on the decoder.

[0035] In particular, in S6, based on the hot spot area, the integrated circuit digital twin model is improved to obtain an improved integrated circuit digital twin model. Here, the integrated circuit digital twin model is improved to improve the overall thermal management efficiency. In a specific example, if a certain location often has high temperature, it may be because the material used there has poor thermal conductivity; at this time, it can be selected to replace it with a material with a higher thermal conductivity to speed up the heat transfer. For geometric problems, such as insufficient contact area between the chip and the heat sink, the heat conduction efficiency can be improved by adding additional heat conducting sheets or changing the packaging form. In addition, the design of the heat dissipation path can be re-evaluated to ensure that heat can be transferred from the heat source to the external environment in the expected manner, rather than concentrated in some places where heat is not easy to dissipate. In addition to making physical improvements directly to the hot spot area, the digital twin model itself needs to be adjusted accordingly. This means updating the parameter settings in the model to make it closer to the new design scheme. For example, when a new material is introduced, its physical properties (such as density, elastic modulus, Poisson's ratio, thermal conductivity, etc.) must be accurately entered to ensure the accuracy of the simulation. At the same time, any changes in geometric dimensions also need to be reflected in the model to ensure that the simulation results can truly reproduce the actual situation after the improvement. The setting of boundary conditions also needs to be adjusted according to the latest design requirements. For example, if the shape or installation method of the radiator is changed, the corresponding surface heat transfer coefficient also needs to be modified accordingly to maintain the consistency of the simulation conditions.

[0036] In particular, the S7 monitors the temperature data of the actual integrated circuit in real time, and feeds the temperature data back to the improved integrated circuit digital twin model in real time. That is, as new data continues to flow in, the digital twin model will dynamically update its internal state to reflect the latest temperature distribution. This makes the simulation results always close to the actual situation, enhancing the model's predictive ability and decision support function.

[0037] In summary, the integrated circuit junction temperature and thermal resistance evaluation method based on the digital twin technology according to the embodiment of the present application is explained, which extracts the temperature cloud map from the simulation calculation results, and extracts the heat flow line diagram from the temperature cloud map based on the heat dissipation path, and uses the data analysis and encoding technology of deep learning to discretely sample the heat flow line diagram, and then embeds and encodes the sampled heat flow point data (flow point position and heat flow point temperature), so as to automatically identify the rectangular frame position data of the hot spot area according to the graph walk significant aggregation representation between the embedded encoding features of each heat flow point data. In this way, the influence of fine structures (such as connecting wires, solder joints, etc.) on the heat transfer path can be captured, and the complex heat dissipation requirements can be better adapted to ensure the effective satisfaction of the heat dissipation requirements. In addition, it is possible to comprehensively capture the changing patterns and trends of the heat distribution, and respond to transient changes in real time to discover potential hot spots or abnormal conditions.

[0038] Furthermore, a system for evaluating junction temperature and thermal resistance of an integrated circuit based on digital twin technology is also provided.

[0039] Figure 4 FIG. 1 is a block diagram of an integrated circuit junction temperature and thermal resistance evaluation system based on digital twin technology according to an embodiment of the present application. Figure 4 As shown, according to the embodiment of the present application, the integrated circuit junction temperature and thermal resistance evaluation system 300 based on the digital twin technology includes: a digital twin model establishment module 310, used to establish an integrated circuit digital twin model; a heat dissipation path construction module 320, used to construct a single heat dissipation path in the integrated circuit digital twin model; a data transmission module 330, used to monitor the temperature data collected by the temperature sensor in the integrated circuit, and transmit the temperature data collected by the temperature sensor to the integrated circuit digital twin model; a simulation calculation module 340, used to load input power to the digital twin model based on different test conditions, and determine boundary conditions and grid parameters to perform simulation calculations to obtain simulation calculation results; a hot spot area determination module 350, used to determine the hot spot area based on the simulation calculation results; a digital twin model improvement module 360, used to improve the integrated circuit digital twin model based on the hot spot area to obtain an improved integrated circuit digital twin model; a data feedback module 370, used to monitor the temperature data of the actual integrated circuit in real time, and feed back the temperature data to the improved integrated circuit digital twin model in real time.

[0040] As described above, the integrated circuit junction temperature and thermal resistance evaluation system 300 based on digital twin technology according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with an integrated circuit junction temperature and thermal resistance evaluation algorithm based on digital twin technology. In one possible implementation, the integrated circuit junction temperature and thermal resistance evaluation system 300 based on digital twin technology according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or hardware module. For example, the integrated circuit junction temperature and thermal resistance evaluation system 300 based on digital twin technology can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the integrated circuit junction temperature and thermal resistance evaluation system 300 based on digital twin technology can also be one of the many hardware modules of the wireless terminal.

[0041] Alternatively, in another example, the integrated circuit junction temperature and thermal resistance evaluation system 300 based on digital twin technology and the wireless terminal may also be discrete devices, and the integrated circuit junction temperature and thermal resistance evaluation system 300 based on digital twin technology may be connected to the wireless terminal through a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0042] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for evaluating junction temperature and thermal resistance of integrated circuits based on digital twin technology, characterized in that: The method comprises: step 1: establishing an integrated circuit digital twin model; step 2: constructing a single heat dissipation path in the integrated circuit digital twin model; Step 3: monitor the temperature data collected by the temperature sensor in the integrated circuit, and transmit the temperature data collected by the temperature sensor to the digital twin model of the integrated circuit; Step 4: load the input power to the digital twin model based on different test conditions, and determine the boundary conditions and grid parameters to perform simulation calculations to obtain simulation calculation results; Step 5: determine the hot spot area based on the simulation calculation results; Step 6: based on the hot spot area, improve the digital twin model of the integrated circuit to obtain an improved digital twin model of the integrated circuit; Step 7: monitor the temperature data of the actual integrated circuit in real time, and feed the temperature data back to the improved digital twin model of the integrated circuit in real time; Wherein, the step 5 includes: extracting a temperature cloud map from the simulation calculation results; extracting a heat flow line diagram from the temperature cloud map based on the heat dissipation path; discretely sampling the heat flow line diagram to obtain a sequence of heat flow point data, the heat flow point data including the heat flow point position and the heat flow point temperature; embedding and encoding each heat flow point data in the sequence of heat flow point data to obtain a sequence of heat flow point data embedded coding features; performing graph walk autocorrelation gated significant aggregation on the sequence of heat flow point data embedded coding features to obtain a heat flow point set significant aggregation feature; based on the heat flow point set significant aggregation feature, obtaining a recognition result, the recognition result is the rectangular frame position data of the hot spot area.

2. The integrated circuit junction temperature and thermal resistance evaluation method based on digital twin technology according to claim 1 is characterized in that: The simulation calculation results are data of junction temperature and thermal resistance.

3. The integrated circuit junction temperature and thermal resistance evaluation method based on digital twin technology according to claim 2 is characterized in that: Each hot flow point data in the sequence of hot flow point data is embedded and encoded to obtain a sequence of hot flow point data embedded coding features, including: each hot flow point data in the sequence of hot flow point data is embedded and encoded through a hot flow point data embedding matrix to obtain a sequence of hot flow point data embedded coding vectors as the sequence of hot flow point data embedded coding features.

4. The integrated circuit junction temperature and thermal resistance evaluation method based on digital twin technology according to claim 3 is characterized in that: The sequence of the heat flow point data embedded coding features is subjected to graph walk autocorrelation gating significant aggregation to obtain the heat flow point set significant aggregation features, including: Inputting each heat flow point data embedding coding vector in the sequence of heat flow point data embedding coding vectors into a hyperbolic space mapper to obtain a sequence of heat flow point data embedding coding vectors after hyperbolic space mapping; Based on the sequence of heat flow point data embedded in the coding vector after the hyperbolic space mapping, the graph walk topology feature is extracted to obtain the heat flow point data graph walk topology feature matrix; Based on the walk topology feature matrix of the heat flow point data graph, contextual semantic enhancement is performed on the sequence of heat flow point data embedded coding vectors after the hyperbolic space mapping to obtain a sequence of heat flow point data contextual semantic enhancement feature vectors; Based on the sequence of the hot flow point data context semantics enhanced feature vectors, the sequence of the hot flow point data embedded coding vectors is significantly gated and weighted aggregated to obtain a hot flow point set significant aggregation feature vector as the hot flow point set significant aggregation feature.

5. The integrated circuit junction temperature and thermal resistance evaluation method based on digital twin technology according to claim 4 is characterized in that: Based on the sequence of heat flow point data embedded in the coding vector after the hyperbolic space mapping, the graph walk topology feature is extracted to obtain the heat flow point data graph walk topology feature matrix, including: Calculating the Poincare distance between any two embedding coding vectors of the heat flow point data after hyperbolic space mapping in the sequence of embedding coding vectors of the heat flow point data after hyperbolic space mapping to obtain the walk topology matrix of the heat flow point data graph; The heat flow point data graph walk topology matrix is ​​subjected to dilated convolution encoding to obtain a heat flow point data graph walk topology feature matrix.

6. The integrated circuit junction temperature and thermal resistance evaluation method based on digital twin technology according to claim 5 is characterized in that: Based on the wandering topological feature matrix of the heat flow point data graph, the sequence of heat flow point data embedded coding vectors after the hyperbolic space mapping is subjected to contextual semantic enhancement processing to obtain a sequence of heat flow point data contextual semantic enhancement feature vectors, including: inputting the wandering topological feature matrix of the heat flow point data graph and the sequence of heat flow point data embedded coding vectors after the hyperbolic space mapping into a global context wandering encoder based on a graph convolutional neural network model to obtain a sequence of heat flow point data contextual semantic enhancement feature vectors.

7. The integrated circuit junction temperature and thermal resistance evaluation method based on digital twin technology according to claim 6 is characterized in that: Based on the sequence of the hot flow point data context semantics enhanced feature vectors, performing significant gated weighted aggregation on the sequence of the hot flow point data embedded coding vectors to obtain a hot flow point set significant aggregation feature vector as the hot flow point set significant aggregation feature, including: Inputting each group of corresponding hot flow point data context semantics enhancement feature vectors and hot flow point data embedding coding vectors in the sequence of hot flow point data context semantics enhancement feature vectors and the sequence of hot flow point data embedding coding vectors into an autocorrelation gating unit to obtain a sequence of hot flow point data autocorrelation gating significant confidence factors; Inputting the sequence of the hot flow point data autocorrelation gated significant confidence factors into a normalization unit based on a Softmax function to obtain a sequence of the hot flow point data autocorrelation gated significant confidence weight factors; Based on the sequence of the autocorrelation gated significant confidence weight factors of the hot flow point data, the position-weighted sum of the sequence of the hot flow point data embedding encoding vectors is calculated to obtain the significant aggregation feature vector of the hot flow point set.

8. The integrated circuit junction temperature and thermal resistance evaluation method based on digital twin technology according to claim 7 is characterized in that: Based on the significant aggregation features of the heat flow point set, a recognition result is obtained, including: inputting the significant aggregation feature vector of the heat flow point set into a hot spot area identifier based on a decoder to obtain the recognition result.

9. An integrated circuit junction temperature and thermal resistance evaluation system based on digital twin technology, characterized in that: include: Digital twin model building module, used to build integrated circuit digital twin models; a heat dissipation path construction module, used to construct a single heat dissipation path in the integrated circuit digital twin model; A data transmission module, used to monitor the temperature data collected by the temperature sensor in the integrated circuit, and transmit the temperature data collected by the temperature sensor to the digital twin model of the integrated circuit; A simulation calculation module, used to load input power to the digital twin model based on different test conditions, and determine boundary conditions and grid parameters to perform simulation calculations to obtain simulation calculation results; A hot spot area determination module, used to determine the hot spot area based on the simulation calculation results; A digital twin model improvement module, used for improving the integrated circuit digital twin model based on the hot spot area to obtain an improved integrated circuit digital twin model; The data feedback module is used to monitor the temperature data of the actual integrated circuit in real time and feed back the temperature data to the improved integrated circuit digital twin model in real time.

Images