Vacuum brazing temperature field real-time prediction and control method based on digital twinning

Abstract

The invention discloses a vacuum brazing temperature field real-time prediction and control method based on digital twinning, particularly relates to the technical field of program control, and is used for solving the problem of unnecessary shutdown caused by production interruption forcibly triggered by a safety protection unit in an existing vacuum brazing control system under an extreme working condition. According to the method, temperature field data and pressure data in a furnace are obtained in real time, extreme working conditions approaching a safety threshold are dynamically recognized, a weighted graph model is constructed based on a high-temperature area to analyze a heat influence relation, then a minimum intervention area is accurately positioned, and a flexible regulation and control strategy is generated; a strategy execution effect is simulated in a virtual space by using a digital twin model, a safety recovery condition is intelligently evaluated by analyzing the energy change of a graph signal frequency domain, and finally a programmed decision is realized to execute flexible regulation or trigger production interruption, so that the control precision and the operation continuity are improved.

Description

Technical Field

[0001] This invention relates to the field of program control technology, and in particular to a method for real-time prediction and control of vacuum brazing temperature field based on digital twins. Background Technology

[0002] Vacuum brazing is a core process for manufacturing key components of high-end equipment such as aerospace engines and precision instruments. Its process quality directly determines the service performance and reliability of the components. The uniformity and stability of the temperature field are the primary factors affecting the brazing quality. In the field of vacuum brazing, digital twin technology can realize real-time monitoring and prediction of the temperature field inside the furnace by constructing a virtual mapping of the physical brazing process, thus providing a new technical path for precise temperature control. In existing technologies, the integration of digital twin systems with underlying actuators, such as heaters and vacuum systems, to form a complete program control system is a practice. This system executes a preset temperature control program based on the predicted data output by the digital twin model.

[0003] However, when faced with extreme conditions such as sudden increases in furnace pressure and excessive regional temperatures approaching safety thresholds, the internal operating logic of such integrated control systems is contradictory. To ensure absolute safety, the independently set safety protection unit in the system will follow its inherent design and forcibly trigger production interruption. However, this standardized protection mechanism fails to consider the flexible control judgments that the digital twin system may make based on its global perception and prediction capabilities, which could avoid downtime. This decision-making conflict between safety logic and control intelligence may cause the production process to encounter unnecessary interruptions due to recoverable transient disturbances, resulting not only in the scrapping of work-in-process and energy waste, but also restricting the full realization of the continuous and stable operation efficiency of the intelligent manufacturing system. Summary of the Invention

[0004] This invention addresses the technical problems existing in the prior art by providing a method for real-time prediction and control of vacuum brazing temperature field based on digital twins.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: A method for real-time prediction and control of vacuum brazing temperature field based on digital twins includes: S1. Obtain real-time temperature field data and furnace pressure data of the vacuum brazing furnace; S2. Based on real-time temperature field data and furnace pressure data, determine whether extreme operating conditions approaching the safety threshold have occurred; S3. When it is determined that an extreme working condition has occurred, a weighted graph model is constructed based on real-time temperature field data, with high-temperature regions as nodes and the thermal influence relationship between nodes as weights. S4. Based on the weighted graph model, the set of nodes that play a key role in maintaining the structure of the anomalous temperature field is identified by analyzing the topological importance of nodes in the graph, and the physical region corresponding to the set of nodes is determined as the minimum intervention region; at the same time, a flexible control strategy for the minimum intervention region is generated. S5. Using a digital twin model, analyze the frequency domain energy changes of the graph signal caused by the flexible control strategy on the minimum intervention area in a weighted graph model; S6. Determine whether the flexible control strategy meets the safety recovery conditions based on the frequency domain energy change; if it does, execute the flexible control strategy; if it does not, trigger the production interruption of the safety protection unit.

[0006] Furthermore, real-time temperature field data and internal pressure data of the vacuum brazing furnace are acquired, including: Based on the multiple temperature acquisition zones divided inside the vacuum brazing furnace, temperature values ​​are collected in real time from each temperature acquisition zone to form temperature field data. Acquiring furnace pressure data includes real-time acquisition of pressure values ​​from pressure monitoring points in the vacuum brazing furnace.

[0007] Furthermore, based on real-time temperature field data and furnace pressure data, it is determined whether extreme operating conditions approaching the safety threshold have occurred, including: Identify whether there are high-temperature regions exceeding the temperature threshold in real-time temperature field data; Monitor whether the pressure data inside the furnace exceeds the pressure threshold; When a high-temperature zone is detected simultaneously and the pressure data inside the furnace exceeds the pressure threshold, it is determined that an extreme operating condition approaching the safety threshold has occurred.

[0008] Furthermore, when an extreme operating condition is identified, a weighted graph model is constructed based on real-time temperature field data, with high-temperature regions as nodes and the thermal influence relationships between nodes as weights. This model includes: When an extreme working condition is determined to occur, the temperature acquisition area with a temperature value exceeding the high temperature threshold is extracted from the real-time temperature field data as a node. Calculate the weights of thermal influence relationships based on the spatial distance between nodes and the temperature gradient; The spatial distance is determined by the geometric center distance of the temperature acquisition area corresponding to the node, and the temperature gradient is determined by the temperature change rate between nodes. A weighted graph model is constructed using all nodes and the calculated weights of the thermal influence relationships.

[0009] Furthermore, a weighted graph model is constructed using all nodes and the calculated thermal influence relationship weights, including: using the node set as the vertices of the graph and establishing edges between nodes according to the thermal influence relationship weights, thereby forming a weighted graph structure; where nodes correspond to high-temperature regions, and edge weights represent the intensity of the thermal influence relationship between nodes.

[0010] Furthermore, based on a weighted graph model, the set of nodes that play a key role in maintaining the structure of the anomalous temperature field is identified by analyzing the topological importance of nodes in the graph, and the physical region corresponding to the set of nodes is determined as the minimum intervention region; simultaneously, a flexible control strategy for the minimum intervention region is generated, including: Calculate the topological importance index of each node in the weighted graph model. The topological importance index is obtained by weighted combination of node degree centrality and eigenvector centrality. Nodes whose topological importance index exceeds a preset importance threshold are selected to form a set of key nodes; The key node set is mapped back to the physical space of the vacuum brazing furnace, and the corresponding temperature acquisition area is determined as the minimum intervention area. Based on the heat load distribution characteristics and heat influence relationship weights of the minimum intervention area, a flexible control strategy including power adjustment amplitude and action timing is generated.

[0011] Furthermore, the set of key nodes is mapped back to the physical space of the vacuum brazing furnace, and the corresponding temperature acquisition area is determined as the minimum intervention area. This includes: locating the temperature acquisition area in the physical space by means of node identifiers based on the correspondence between key nodes and temperature acquisition areas, and defining the set of temperature acquisition areas as the minimum intervention area.

[0012] Furthermore, using a digital twin model, the frequency domain energy changes of the graph signal caused by the flexible control strategy in the minimum intervention area are analyzed on a weighted graph model, including: The temperature field distribution difference between the minimum intervention area and its adjacent nodes before and after the application of the flexible control strategy is defined as the graph signal; The graph Fourier transform of the graph signal is performed based on the Laplace matrix of the weighted graph model to obtain the corresponding spectral distribution. Extract high-frequency components from the spectral distribution and calculate their energy values; By comparing the energy values ​​of high-frequency components before and after applying the flexible control strategy, the high-frequency energy attenuation rate is obtained as the frequency domain energy change of the graphical signal.

[0013] Furthermore, a graph Fourier transform is performed on the graph signal based on the Laplace matrix of the weighted graph model to obtain the corresponding spectral distribution. This includes: calculating the Laplace matrix of the weighted graph model, and then applying the graph Fourier transform to the graph signal to obtain the spectral distribution; where the graph signal represents the temperature field distribution difference, and the spectral distribution contains frequency domain component information.

[0014] Furthermore, the system determines whether the flexible control strategy meets the safety recovery conditions based on frequency domain energy changes; if it does, the flexible control strategy is executed; otherwise, the production interruption of the safety protection unit is triggered, including: Compare the high-frequency energy decay rate with a preset safe recovery threshold; When the high-frequency energy attenuation rate is greater than or equal to the safe recovery threshold, the flexible control strategy is determined to meet the safe recovery condition and the corresponding strategy is executed. When the high-frequency energy attenuation rate is less than the safety recovery threshold, it is determined that the safety recovery condition is not met and the production interruption of the safety protection unit is triggered.

[0015] The beneficial effects of this invention are: 1. By continuously collecting temperature and pressure data inside the furnace, the system dynamically identifies extreme operating conditions approaching the safety threshold. Based on a weighted graph model, it performs topological analysis on the thermal influence relationship in the high-temperature region, thereby accurately locating the minimum intervention area. The programmable control mechanism based on digital twins enables the system to quickly generate targeted flexible control strategies under extreme conditions. The effectiveness of the strategies is evaluated through frequency domain energy changes, ultimately achieving synergistic optimization of control decisions and safety logic. This significantly improves the control accuracy and response efficiency of the vacuum brazing process, effectively avoids unnecessary production interruptions caused by transient disturbances, and ensures the continuity and stability of the manufacturing process of key components for high-end equipment.

[0016] 2. By analyzing the frequency domain of the graph signal, the temperature field distribution difference is transformed into a quantifiable energy attenuation index, providing a reliable basis for the safety assessment of flexible control strategies. By using a digital twin model to simulate the execution effect of the control strategy in virtual space, and by real-time monitoring of the high-frequency energy attenuation in the minimum intervention area, it is possible to intelligently determine whether the safety recovery conditions are met. This not only ensures the accuracy and timeliness of control actions, but also significantly reduces the risk of production interruption caused by the overly conservative nature of traditional safety protection mechanisms. Thus, while improving the consistency of process quality, it also optimizes energy utilization efficiency and equipment operating performance. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method for real-time prediction and control of vacuum brazing temperature field based on digital twins according to the present invention; Figure 2 This is a flowchart for determining the minimum intervention area in this invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example: Figure 1This invention presents a method for real-time prediction and control of vacuum brazing temperature field based on digital twins, comprising: S1. Obtain real-time temperature field data and furnace pressure data of the vacuum brazing furnace; S2. Based on real-time temperature field data and furnace pressure data, determine whether extreme operating conditions approaching the safety threshold have occurred; S3. When it is determined that an extreme working condition has occurred, a weighted graph model is constructed based on real-time temperature field data, with high-temperature regions as nodes and the thermal influence relationship between nodes as weights. S4. Based on the weighted graph model, the set of nodes that play a key role in maintaining the structure of the anomalous temperature field is identified by analyzing the topological importance of nodes in the graph, and the physical region corresponding to the set of nodes is determined as the minimum intervention region; at the same time, a flexible control strategy for the minimum intervention region is generated. S5. Using a digital twin model, analyze the frequency domain energy changes of the graph signal caused by the flexible control strategy on the minimum intervention area in a weighted graph model; S6. Determine whether the flexible control strategy meets the safety recovery conditions based on the frequency domain energy change; if it does, execute the flexible control strategy; if it does not, trigger the production interruption of the safety protection unit.

[0020] S1. Obtain real-time temperature field data and furnace pressure data of the vacuum brazing furnace. Specifically, this is implemented as follows: Acquiring real-time temperature field data of the vacuum brazing furnace involves dividing the internal space of the furnace into multiple temperature acquisition zones. Each zone is defined based on the furnace's geometry and heat distribution characteristics. For example, the furnace space is uniformly divided into multiple cubic sub-regions along its length, width, and height. The dimensions of each sub-region are determined based on the actual furnace dimensions and the temperature gradient range. The actual furnace dimensions are obtained from design drawings or on-site measurements, for example, a length range of 1 to 5 meters, a width range of 0.5 to 3 meters, and a height range of 0.5 meters. The temperature gradient range, up to 2 meters, is obtained through historical process data or thermal simulation analysis. For example, during brazing, the temperature gradient is typically 10 to 100 degrees Celsius per meter. The length direction is divided into 10 to 50 intervals, the width direction into 10 to 50 intervals, and the height direction into 5 to 20 intervals. The specific number is adjusted according to the furnace volume and the required temperature resolution. For example, when the furnace volume is large, the number of intervals is increased to improve the resolution. When high temperature resolution is required, a smaller interval size is selected, such as 0.1 meters by 0.1 meters by 0.The temperature acquisition area is divided into 1-meter sections. During the division, it is ensured that the physical space covered by each temperature acquisition area has consistent thermal conductivity characteristics, avoiding areas where the boundaries are located at abrupt temperature changes. For example, a thermal imager is used to verify whether the area boundaries avoid hot or cold spots. Real-time temperature values ​​are acquired from each area using temperature sensing elements placed at the center or representative locations of the area. These sensing elements include thermocouples or infrared thermometers. For example, K-type or S-type thermocouples are used to match the temperature range, while infrared thermometers are suitable for non-contact measurement. The installation location is determined through thermal balance calculations to ensure representativeness. The acquisition frequency is set to the range of 1 Hz to 100 Hz, with the specific value selected based on process response speed and data accuracy requirements. For example, a 10 Hz acquisition frequency is used during high-temperature brazing to capture rapid temperature fluctuations. The acquired temperature values ​​are amplified and filtered by a signal conditioning circuit to eliminate noise interference. For example, a low-pass filter with a cutoff frequency of 5 Hz is used to remove high-frequency noise. The data is then converted to digital value by an analog-to-digital converter. The analog-to-digital converter (ADC) resolution is set to 12-bit to 16-bit to ensure accuracy. The converted digital signal is transmitted to the processing unit via a data bus. When generating temperature field data, the temperature values ​​of each temperature acquisition area are encoded into a multi-dimensional array structure according to their spatial location. The array dimensions correspond to the spatial division of the temperature acquisition area. For example, in a three-dimensional array, the row index corresponds to the region number in the length direction, the column index corresponds to the region number in the width direction, and the depth index corresponds to the region number in the height direction. The array element value is the real-time temperature value of the corresponding temperature acquisition area, with the unit uniformly in degrees Celsius. The data storage format uses floating-point numbers to ensure accuracy; for example, 32-bit floating-point numbers are used to represent temperature values. The temperature field data is updated in real time and temporarily stored in a buffer for subsequent processing steps. The buffer size is set according to the data volume and update frequency; for example, storing temperature field data from the most recent 1000 time points to ensure continuous processing. Data verification is achieved through cyclic redundancy checks to detect transmission errors. Abnormal situations, such as data loss or exceeding limits, trigger re-acquisition or an alarm.

[0021] Acquiring furnace pressure data involves real-time acquisition of pressure values ​​from pressure monitoring points within the vacuum brazing furnace. These monitoring points are located at key positions inside the furnace, including areas near heating elements and vacuum pipe interfaces. The selection of these locations is based on pressure sensitivity and process safety. For example, pressure distribution is analyzed through fluid dynamics simulation to identify sensitive points. A main pressure monitoring point is set at the center of the furnace, and an auxiliary pressure monitoring point is set at the vacuum pump inlet. The number of monitoring points is adjusted according to the furnace size and the complexity of the vacuum system; for example, two monitoring points are set for small furnaces, and five monitoring points are set for large furnaces. Pressure value acquisition is achieved through pressure sensing devices, including piezoresistive sensors and capacitive sensors. For example, piezoresistive sensors have a range of 0 Pa to 1000 Pa, while capacitive sensors have a range of 0 Pa to 5000 Pa. Frequency and temperature acquisition can be set synchronously or independently, ranging from 1 Hz to 50 Hz. For example, a 20 Hz acquisition frequency can be used to match the dynamic response of the vacuum system. The acquired pressure values ​​are calibrated by a circuit to compensate for the influence of ambient temperature, for example, by using a temperature compensation coefficient of 0.1% per degree Celsius. The data is then transmitted to the processing unit via a digital interface, which can be either serial or parallel. Pressure data is recorded as scalar values ​​or arrays, with the unit uniformly in Pascals. When multiple pressure monitoring points are used, the pressure data is represented as a collection of pressure values ​​from each monitoring point, with timestamps added to ensure time consistency. The data transmission protocol uses standard communication protocols such as Modbus or custom protocols. Data is compressed during storage to reduce storage space, for example, by using a lossless compression algorithm. Abnormal pressure values, such as those exceeding the measurement range, trigger calibration or sensor replacement.

[0022] S2. Based on real-time temperature field data and furnace pressure data, determine whether extreme operating conditions approaching the safety threshold have occurred. Specifically, this is implemented as follows: The system identifies high-temperature regions exceeding a temperature threshold in the real-time temperature field data. This threshold is a critical temperature value set based on the safety requirements of the vacuum brazing process and the thermal performance parameters of the materials, used to prevent overheating of the workpiece or degradation of the material. It is obtained through methods such as analyzing the thermal stability data of the materials, conducting thermal simulation experiments, or statistically analyzing historical production data. For example, the temperature threshold can be determined by measuring the melting point of the brazing filler metal with a thermal analyzer and then adding a safety margin. The safety margin ranges from 10°C to 50°C, with specific values ​​adjusted according to material type and process experience. For instance, the temperature threshold is set to 950°C for titanium alloys and 1080°C for copper-based materials. The temperature threshold can be dynamically updated with each process stage. For example, it can be set to 800°C during the heating stage, 1000°C during the holding stage, and 700°C during the cooling stage. The update logic is based on a preset process curve and real-time temperature trends. The identification process is achieved by scanning all temperature acquisition areas in the real-time temperature field data, which comes from a multi-dimensional array generated in step S1, where each array element has a value representing a specific temperature. The temperature data is collected in Celsius. The identification algorithm compares the temperature value of each temperature collection area with a temperature threshold, using the greater than operator for numerical comparison. If the temperature value of a certain temperature collection area is greater than the temperature threshold, the area is classified as a high-temperature area. The identification operation takes into account measurement errors and uses data smoothing processing, such as applying a moving average filtering algorithm to preprocess the temperature data, with the filter window size set to 3 to 7 data points to suppress random fluctuations. The identification results are output in the form of a high-temperature area index list, which records the spatial coordinates and temperature values ​​of all temperature collection areas exceeding the limit. The list is updated at the same frequency as the data collection frequency. Threshold management includes periodic verification, such as recalibrating the temperature threshold every 10 process cycles, based on recent temperature distribution statistics and process quality feedback. The identification process also includes boundary condition processing, such as using interpolation or default values ​​from adjacent areas if data for a temperature collection area is missing, to ensure identification continuity. The identification algorithm optimization includes using parallel computing to accelerate large-scale data comparison, such as dividing the temperature field data into multiple sub-blocks for simultaneous processing to improve real-time performance.

[0023] The monitoring process checks whether the furnace pressure exceeds the pressure threshold. The pressure threshold is a critical pressure value set based on the vacuum system's operating limits and process safety boundaries. It is used to maintain the furnace vacuum and prevent excessive pressure from damaging the equipment. It can be obtained by consulting vacuum pump performance charts, conducting pressure tolerance tests, or calculating based on equipment design parameters. For example, the pressure threshold can be determined through a vacuum system leakage rate test. The test method involves measuring the pressure rise rate under standard operating conditions and calculating the safety upper limit. The pressure threshold ranges from 0.01 Pa to 200 Pa, with the specific value selected based on the vacuum level and furnace structure. For example, it can be set to 0.1 Pa under high vacuum conditions and 50 Pa under low vacuum conditions. The pressure threshold can be adaptively adjusted according to the process state. For example, it can be set to 100 Pa during the evacuation phase, 1 Pa during the stabilization phase, and 101325 Pa during the cavitation phase. The adjustment strategy is based on the real-time pressure curve and system response characteristics. The monitoring process is achieved by reading the furnace pressure data collected in step S1. The pressure data comes from real-time readings of the pressure monitoring points, and the unit is Pascals. If multiple... The pressure monitoring system employs a method that compares the weighted average or maximum pressure value at each point with a pressure threshold. Weights are assigned based on the importance of the monitoring point's location; for example, a monitoring point near the heater has a weight of 0.6, while other points have a weight of 0.4. The comparison is implemented using conditional statements; if the pressure value exceeds the threshold, it is considered an over-limit pressure reading. The monitoring process includes outlier handling, such as checking if the pressure value is within the physically possible range (e.g., 0 Pascal to 110,000 Pascal). If it exceeds this range, it is marked as a sensor malfunction, and a backup monitoring point is activated. To reduce transient interference, digital filtering techniques are used, such as a first-order low-pass filter to process the pressure data, with a cutoff frequency set between 0.1 Hz and 5 Hz, selected based on the system's dynamic characteristics. Monitoring results are output as a pressure over-limit flag, a Boolean value, along with the pressure value at the time of the over-limit, a timestamp, and the monitoring point identifier. The monitoring system also includes calibration functions, such as periodically calibrating the sensor using a standard pressure source to ensure measurement accuracy. Environmental factors are considered during the monitoring process, such as correcting the pressure reading based on ambient temperature; the compensation coefficient is obtained through experimental calibration.

[0024] When a high-temperature zone is detected simultaneously and the furnace pressure exceeds the pressure threshold, it is determined that an extreme operating condition approaching the safety threshold has occurred. The presence of a high-temperature zone is detected by checking if the high-temperature zone index list is empty; if the list contains at least one entry, it indicates the presence of a high-temperature zone. Pressure exceeding the limit is detected by querying the pressure exceeding the limit flag status; if the flag is true, it indicates that the pressure exceeds the pressure threshold. The judgment logic uses a logical AND operation; that is, when both the presence of a high-temperature zone and pressure exceeding the limit are true, the extreme operating condition judgment is output as true. The judgment operation is executed in each control cycle, which is synchronized with the data acquisition cycle. For example, the control cycle is set to 0.1 seconds to 1 second, with the specific value determined based on the process response speed. The judgment result is output in the form of an extreme operating condition flag, which is a Boolean variable; true indicates that the extreme condition has occurred, and false indicates that it has not occurred. To improve... To improve the reliability of judgments, a continuous confirmation mechanism is introduced. For example, extreme operating conditions are required to be met continuously for more than 2 to 5 control cycles before final confirmation; otherwise, they are considered transient disturbances and ignored. Judgment results are transmitted to subsequent processing units in real time. For example, when the extreme operating condition flag is true, the weighted graph model construction of step S3 is triggered, and judgment logs including time, temperature distribution, and pressure values ​​are stored. The judgment system includes fault protection, such as periodically checking the integrity of input data. If data is lost, the cached value of the previous cycle is used or the system self-check is triggered. The judgment process also considers the process context. For example, judgment conditions can be temporarily relaxed during startup or shutdown to avoid false alarms, ensuring the robustness and adaptability of the judgments. Judgment algorithm optimization includes using a state machine to manage the judgment process, such as defining initialization, monitoring, judgment, and recovery states to ensure clear logic and maintainability.

[0025] S3. When an extreme operating condition is determined to have occurred, a weighted graph model is constructed based on real-time temperature field data, with high-temperature regions as nodes and the thermal influence relationships between nodes as weights. The specific implementation is as follows: When an extreme operating condition is detected, temperature acquisition areas whose temperature values ​​exceed the high-temperature threshold are extracted from the real-time temperature field data and designated as nodes. The condition for determining an extreme operating condition comes from the extreme operating condition flag output in step S2; this step is triggered when the flag is true. The extraction process is achieved by scanning the real-time temperature field data, which comes from a multi-dimensional array generated in step S1. The array element values ​​are the Celsius temperatures of each temperature acquisition area. The high-temperature threshold is a critical temperature value defined in step S2, used to distinguish between normal and high-temperature areas. Its value is set according to the material type and process requirements. For example, for stainless steel, the high-temperature threshold is set to 1100 degrees Celsius. This value is determined through material thermal analysis experiments, including measuring the deformation point or phase transition point of the material at high temperatures, with an additional safety margin such as 50 degrees Celsius. The extraction algorithm traverses all temperature acquisition areas, comparing the temperature value of each area with the high-temperature threshold using the greater than operator. If the temperature value of a certain temperature acquisition area is greater than the high-temperature threshold, that area is selected as a node. The data is stored in a list format, with each element containing an identifier, spatial coordinates, and temperature value for the temperature acquisition area. The identifier matches the area index in step S1. The extraction operation considers data integrity; for example, if data is missing for a certain area, interpolated values ​​from neighboring areas or default values ​​are used. Interpolation methods include linear interpolation or nearest neighbor interpolation to ensure the reliability of the node set. The node list is updated in real time to reflect changes in the temperature field, with an update frequency synchronized with the data acquisition frequency, such as 10 times per second. The extraction process also includes deduplication to ensure that each temperature acquisition area is recorded only once, avoiding duplicate nodes. Furthermore, the extraction algorithm is optimized by using indexes to accelerate traversal, for example, processing only the most recently updated temperature acquisition areas to improve efficiency. Boundary conditions are also considered during extraction; for example, if the temperature acquisition area is located at the furnace edge, the comparison logic is adjusted to account for the impact of heat loss. The extraction process also includes a data verification step, such as checking whether the temperature value is within a reasonable range, such as 0 degrees Celsius to 2000 degrees Celsius; values ​​outside this range are marked as abnormal and excluded.

[0026] The weights of the thermal influence relationship are calculated based on the spatial distance between nodes and the temperature gradient. The spatial distance is determined by the geometric center distance of the temperature acquisition area corresponding to the node. The geometric center coordinates are obtained from the region division in step S1. For example, the three-dimensional coordinates of the center point of each temperature acquisition area are calculated from the region boundary, and the coordinate unit is meters. The distance calculation uses the Euclidean distance method to calculate the straight-line distance between two points. For example, for two nodes A and B, their geometric center coordinates are (x1, y1, z1) and (x2, y2, z2) respectively. The expression for the distance is: The unit is meters. The temperature gradient is determined by the rate of temperature change between nodes, which is defined as the ratio of the temperature difference between two nodes to their distance. For example, if node A has temperature T1 and node B has temperature T2, the temperature gradient is calculated as: g = |T2 - T1| / d; where g is the temperature gradient, in degrees Celsius per meter. The weight of the thermal influence relationship is calculated based on spatial distance and temperature gradient. The calculation method for the weight of the thermal influence relationship is: w = α / d + β × g, where w is the weight of the thermal influence relationship, and α and β are weight coefficients used to balance the influence of distance and gradient. The weight coefficients range from 0.1 to 10, and the specific values ​​are determined through experiments or simulations, such as through historical data regression analysis or heat conduction simulation optimization, α = 1.0, β = 0.5. The weight calculation is performed on all node pairs, but can be limited to adjacent nodes or node pairs with a distance less than a threshold to reduce the amount of computation. The distance threshold is... The distance threshold is set to 0.1 meters to 1 meter, depending on the furnace size. For example, for a small furnace, the distance threshold is set to 0.2 meters. The calculation results are stored as a weight matrix or an edge list, where the matrix element wij represents the weight of the thermal influence relationship between node i and node j. The weight values ​​are normalized, for example, scaled to the range of 0 to 1 for easier subsequent processing. Optimization of the calculation process includes using parallel computing to accelerate the processing of large-scale node pairs. The calculation also includes anomaly handling, such as using a minimum value like 0.001 meters to avoid division by zero if the distance d is 0, or skipping the node pair if the temperature gradient is invalid. The weight calculation also considers environmental factor compensation, such as correcting the temperature gradient value based on the gas flow velocity inside the furnace, with the compensation coefficient obtained through experimental calibration. The calculation process also includes data smoothing, such as applying a moving average filter to the weight values, with the filter window size set to 3 to 5 data points to reduce fluctuations.

[0027] A weighted graph model is constructed using all nodes and the calculated heat impact weights. The construction process includes using the node set as the vertices of the graph, with nodes derived from the node list generated in sub-step 1; establishing edges between nodes based on their heat impact weights, where each edge represents a heat impact relationship and its weight comes from the weight values ​​calculated in sub-step 2; the weighted graph structure is represented using an adjacency matrix or adjacency list, for example, using a two-dimensional array as the adjacency matrix, where the array row and column indices correspond to node indices, and the element values ​​are weights. If there are no edges between nodes, the weight is set to 0 or a maximum value such as 9999; after the graph model is constructed, an integrity check is performed, ensuring that all nodes are labeled, weight values ​​are within a reasonable range (e.g., 0 to 100), and outliers are corrected or removed; the weighted graph model is output as a data structure or file. This is used in subsequent steps, such as the topological importance analysis in step S4; the construction process optimization includes using sparse matrix storage to reduce memory consumption, and storing only non-zero weight edges when the number of nodes is large; the graph model is updated and synchronized with temperature field data, for example, reconstructing whenever we judge extreme operating conditions to ensure real-time performance; the construction algorithm also includes graph attribute calculation, such as node degree or connectivity, to verify the effectiveness of the model; in addition, the construction process considers scalability, such as supporting dynamic addition or deletion of nodes to adapt to changes in the temperature field; the construction process also includes error recovery mechanisms, such as using the cached last valid graph model or triggering recalculation if the construction fails; the construction process also includes performance optimization, such as compressing and storing the graph structure or indexing it to improve the efficiency of subsequent access.

[0028] Figure 2 A flowchart for determining the minimum intervention region in this invention is provided. In S4, based on a weighted graph model, the set of nodes that play a key role in maintaining the structure of the anomalous temperature field is identified by analyzing the topological importance of nodes in the graph, and the physical region corresponding to the set of nodes is determined as the minimum intervention region. Simultaneously, a flexible control strategy for the minimum intervention region is generated, specifically implemented as follows: The topological importance index of each node in the weighted graph model is calculated. This index is obtained by weighting the node's degree centrality and eigenvector centrality. Node degree centrality is defined as the ratio of the sum of the weights of all connected edges to the maximum possible sum of weights. The maximum possible sum of weights is calculated by subtracting one from the number of nodes and multiplying by the maximum edge weight. The edge weights are derived from the heat-affect relationship weights calculated in step S3. Eigenvector centrality is obtained by iteratively calculating the principal eigenvectors of the adjacency matrix of the weighted graph model. The elements of the adjacency matrix are the edge weights. The initial iteration value is set to 1 for all node centralities. The convergence condition is that the L2 norm of the difference between the centrality vectors of two consecutive iterations is less than 0.001. The maximum number of iterations is set to 100. The topological importance index is calculated as: Index value = Degree centrality weight coefficient × Degree centrality + Eigenvector centrality weight coefficient. The degree centrality and eigenvector centrality weight coefficients are both between 0 and 1, and their sum is 1. Specific values ​​are determined through historical data analysis or expert experience; for example, the degree centrality weight coefficient is set to 0.4, and the eigenvector centrality weight coefficient is set to 0.6. The calculation process includes data preprocessing, such as normalizing the degree centrality and eigenvector centrality to a range of 0 to 1 using minimum-maximum scaling. The output is the topological importance index value for each node, stored as a list or array for subsequent steps. Optimization includes using parallel algorithms to accelerate the iterative calculation of eigenvector centrality. The calculation also includes exception handling; for example, if the node's degree centrality is 0, the topological importance index is directly set to 0, or if the iteration does not converge, an alternative method such as algebraic connectivity calculation is used.

[0029] Nodes whose topological importance index exceeds a preset importance threshold are selected to form a set of critical nodes. The preset importance threshold is a critical value set according to process safety requirements and temperature field control accuracy, used to screen critical nodes. It is obtained by analyzing historical abnormal operating data, performing sensitivity analysis, or back-calculating based on control objectives. For example, the optimal value can be selected by simulating the control effect under different thresholds. The preset importance threshold ranges from 0.1 to 0.9, and the specific value is adjusted according to the complexity of the graph model and process stability. For example, it is set to 0.5 for simple temperature field structures and 0.7 for complex temperature field structures. The selection process is achieved by traversing the topological importance index values ​​of all nodes, and the index values ​​are derived from the calculation results of sub-step 1. The selection algorithm uses the greater than operator for numerical comparison. If the topological importance index value of a node is greater than a preset importance threshold, the node is selected into the set of key nodes. The selection operation takes into account data fluctuations and uses lag processing to avoid frequent switching. For example, a node is only selected if its index value exceeds the threshold for more than two consecutive calculation cycles. The selection results are output in the form of a list of key nodes, which records the identifier and topological importance index value of all selected nodes. Threshold management includes dynamic adjustment, such as updating the preset importance threshold according to the real-time temperature field change rate. When the change rate is large, the threshold is increased to improve selectivity. The selection process also includes deduplication and sorting, such as arranging key nodes in descending order of topological importance index value for priority processing.

[0030] The key node set is mapped back to the physical space of the vacuum brazing furnace, and the corresponding temperature acquisition area is determined as the minimum intervention area. The mapping process is based on the correspondence between key nodes and temperature acquisition areas. Key nodes come from the key node list generated in sub-step 2, and temperature acquisition areas are the physical areas defined in step S1. The correspondence is realized through node identifiers, which are consistent with the temperature acquisition area indexes, which are obtained from the multidimensional array structure in step S1. The mapping operation is completed by querying the coordinates of the temperature acquisition area corresponding to the node identifier. The coordinates include three-dimensional spatial positions, and the unit is meters. The minimum intervention area is defined as the set of temperature acquisition areas corresponding to the key nodes, and the area boundary is determined by the area division in step S1. The mapping result is output in the form of a minimum intervention area list, which records the identifiers and spatial coordinates of all involved temperature acquisition areas. The mapping process includes integrity verification, such as checking that all key nodes have corresponding temperature acquisition areas, and using nearest neighbor mapping when missing. Mapping optimization includes spatial clustering, such as merging adjacent temperature acquisition areas into larger intervention units to reduce control complexity. The mapping also considers physical constraints, such as avoiding fixed components or sensor blind spots inside the furnace to ensure intervention feasibility.

[0031] Based on the heat load distribution characteristics and thermal influence relationship weights of the minimum intervention area, a flexible control strategy including power adjustment amplitude and timing is generated. The heat load distribution characteristics are obtained by calculating the sum or average of the temperature values ​​of the temperature acquisition areas within the minimum intervention area. The temperature values ​​are from the real-time temperature field data in step S1, and the unit is degrees Celsius. The thermal influence relationship weights are from the weight matrix calculated in step S3, reflecting the thermal interaction between nodes. The power adjustment amplitude is calculated based on the heat load distribution characteristics and weight values. The calculation method is amplitude value = basic adjustment coefficient × heat load value × weight influence factor. The basic adjustment coefficient is determined by the equipment power range and heat capacity parameters. For example, for a heater power range of 0 to 10 kW, the basic adjustment coefficient is set to 0.1 kW per kilowatt. The temperature is measured in degrees Celsius, and the weighting factor is taken as the maximum weight value of the relevant nodes. The timing of the action is determined based on the weight of the thermal influence relationship, with nodes of higher weight being adjusted first. The timing arrangement is to execute the nodes sequentially after sorting the weight values ​​from high to low, and the interval between adjacent adjustments is set to 1 to 10 seconds, selected according to the system response speed. The generation strategy also includes safety constraints, such as the power adjustment range not exceeding 80% of the maximum power of the equipment and the total duration of the action sequence not exceeding 50% of the process window. The strategy output is a sequence of control instructions, including the power adjustment value and execution time point for each minimum intervention area. The generation process optimization includes simulation verification, such as testing the strategy effect in a digital twin model. The generation also considers adaptive adjustment, such as correcting the power adjustment range and the timing of the action based on real-time feedback.

[0032] S5. Using a digital twin model, analyze the frequency domain energy change of the graph signal caused by the flexible control strategy in the minimum intervention area on a weighted graph model. Specifically, the implementation is as follows: The temperature field distribution difference between the minimum intervention area and its adjacent nodes before and after the application of the flexible control strategy is defined as a graph signal. The minimum intervention area comes from the set of temperature acquisition areas corresponding to the key nodes determined in step S4. Adjacent nodes are obtained by querying nodes directly connected to the minimum intervention area nodes in the weighted graph model, which is constructed in step S3. The temperature field data before the application of the flexible control strategy is taken from the real-time temperature field data before the strategy is executed, and the temperature field data after the application is taken from the real-time temperature field data after the strategy is executed, which comes from step S1. The temperature field distribution difference is calculated by subtracting the temperature before the application from the temperature value after the application for each node, and the unit is degrees Celsius. The graph signal is represented as a vector, and the vector length is equal to the minimum intervention area. The definition process includes the total number of nodes and their neighboring nodes, with each element of the vector corresponding to the temperature difference value of a node. This includes data alignment, such as ensuring that the timestamps of the temperature field data before and after application correspond, with a time deviation of less than 0.1 seconds. Data integrity is also considered; for example, if a node's data is missing, interpolation or a default value of 0 is used. Interpolation methods include linear interpolation based on neighboring node data. Boundary handling is also included; for example, if neighboring nodes do not exist, only nodes in the minimum intervention area are used. Optimization includes using a caching mechanism to store recent temperature field data to improve access efficiency. Data validation is also performed during the definition process, such as checking if the temperature difference value is within a reasonable range, such as -100 degrees Celsius to +100 degrees Celsius; values ​​outside this range are marked as abnormal and data is re-acquired.

[0033] The Laplace matrix based on the weighted graph model is used to perform a graph Fourier transform on the graph signal to obtain the corresponding spectral distribution. The weighted graph model comes from step S3 and includes the node set and edge weights. The formula for calculating the Laplace matrix is: L = D − A, where D is the degree matrix, a diagonal matrix, and the diagonal elements are the weighted degree of each node, which is equal to the sum of the weights of all connected edges of that node; A is the adjacency matrix, and the elements of the adjacency matrix are the edge weights, with an element value of 0 if there are no edges between nodes; L represents the Laplace matrix; the formula for calculating the normalized Laplace matrix is: Where Lsym represents the normalized Laplacian matrix, and I is the identity matrix. The graph Fourier transform is formed by taking the diagonal elements of the degree matrix D to the power of -1 / 2. The graph Fourier transform is achieved by calculating the eigenvalues ​​and eigenvectors of the Laplace matrix. The eigenvalues ​​represent frequency components, and the eigenvectors represent basis functions. The transform process involves projecting the graph signal vector onto the eigenvector space to obtain spectral coefficients, which are the projection values ​​of the graph signal onto each eigenvector. The spectral distribution is represented as a vector of spectral coefficients, with each coefficient corresponding to a different frequency component. Optimization includes using iterative algorithms such as the Lanczos method to approximate the eigenvalue decomposition to reduce computation. The convergence condition is set to an eigenvalue residual of less than 1e-6. The transform also includes data preprocessing, such as removing the mean from the graph signal to eliminate DC components. Computational efficiency is considered during the transform process; for example, a random algorithm is used to accelerate the eigenvalue decomposition when the number of nodes is large. The transform also includes error handling; for example, if the eigenvalue decomposition does not converge, a power iteration method is used as a backup.

[0034] The process extracts high-frequency components from the spectral distribution and calculates their energy values. High-frequency components are defined as spectral coefficients corresponding to larger eigenvalues. The eigenvalue threshold is determined by analyzing the spectral characteristics of the Laplace matrix, for example, by taking the coefficients from the upper half of the sorted eigenvalues ​​or eigenvalues ​​greater than the average eigenvalue. The extraction process is achieved by selecting the spectral coefficients corresponding to eigenvectors with eigenvalues ​​greater than the threshold. The threshold is set as the maximum eigenvalue multiplied by a proportionality coefficient, ranging from 0.5 to 0.9, for example, 0.7. The energy value is calculated as the sum of squares of the spectral coefficients of the high-frequency components, with dimensionless units. The calculation includes data post-processing, such as logarithmic scaling of the energy values ​​to enhance numerical stability. Energy value calculation optimization includes using parallel summation algorithms to accelerate large-scale data processing. The extraction process also includes verification, such as checking whether the high-frequency component energy accounts for a certain proportion of the total energy, with a proportionality threshold set between 0.1 and 0.5, adjusted according to signal characteristics. Frequency band division is considered during extraction, such as dividing the spectrum into multiple sub-bands and calculating their energy separately for more detailed analysis. The extraction process also includes anomaly detection, such as triggering data re-acquisition if the energy value suddenly changes by more than 50%.

[0035] By comparing the high-frequency component energy values ​​before and after applying the flexible control strategy, the high-frequency energy attenuation rate is obtained as the frequency domain energy change of the graphical signal. The high-frequency energy value before application is taken from the energy value calculated before the strategy was implemented, and the high-frequency energy value after application is taken from the energy value calculated after the strategy was implemented. The comparison operation uses difference calculation: attenuation rate = (difference between energy value before application and energy value after application) / energy value before application, and the result is expressed as a percentage. The calculation includes anomaly handling, such as setting the attenuation rate to 100 or a flag value if the energy value before application is 0. The attenuation rate output is a scalar value for use in subsequent steps. The comparison process considers data fluctuations and uses smoothing processing, such as using moving average to calculate energy values ​​to reduce the impact of noise. The comparison also includes threshold checking, such as considering an energy increase when the attenuation rate is less than 0, which may trigger a re-evaluation. The comparison process optimization includes calibration using historical data, such as adjusting the attenuation rate calculation formula based on multiple experiments. The comparison also involves time window management, such as comparing only the energy values ​​within a specific time period after the strategy is implemented, with the time window length set to 5 to 30 seconds, selected according to the process cycle.

[0036] S6. Determine whether the flexible control strategy meets the safety recovery conditions based on the frequency domain energy change; if it does, execute the flexible control strategy; if it does not, trigger the production interruption of the safety protection unit. The specific implementation is as follows: The high-frequency energy attenuation rate is compared with a preset safety recovery threshold. The high-frequency energy attenuation rate is derived from the frequency domain energy change of the graphical signal calculated in step S5 and is expressed as a percentage. The safety recovery threshold is a critical value set according to the safety specifications of the vacuum brazing process and the equipment operating parameters, used to evaluate the effectiveness of the control strategy. It is obtained by analyzing historical safety operating data, performing thermodynamic simulations, or deriving it based on equipment protection settings. For example, the threshold can be determined by statistically analyzing the distribution characteristics of the high-frequency energy attenuation rate under normal operating conditions. The safety recovery threshold ranges from 30% to 90%, with the specific value adjusted according to material characteristics and process requirements. For example, it is set to 70% for heat-sensitive materials and 50% for ordinary materials. The comparison operation uses a numerical comparison algorithm, specifically comparing the high-frequency energy attenuation rate with the safety recovery threshold. The comparison process includes data verification, for example... The comparison process checks whether the high-frequency energy decay rate is within the valid range of 0% to 100%; data outside this range is considered invalid. The comparison also considers measurement errors and employs tolerance processing; for example, when the absolute difference between the high-frequency energy decay rate and the safety recovery threshold is less than 2%, they are considered equal. The comparison result is output in Boolean form, with true indicating that the condition is met and false indicating that it is not. Optimizations to the comparison process include using a sliding window to average the high-frequency energy decay rate, with the window size set to 3 to 5 data points to smooth out instantaneous fluctuations. The comparison also includes a dynamic threshold adjustment mechanism, such as correcting the safety recovery threshold based on changes in ambient temperature; the correction coefficient is obtained through experimental calibration. The comparison process also involves data synchronization, such as ensuring that the high-frequency energy decay rate and the safety recovery threshold use data from the same time base. The comparison operation also includes multiple verifications, such as performing three consecutive comparisons with consistent results before confirming the final output.

[0037] When the high-frequency energy attenuation rate is greater than or equal to the safety recovery threshold, the flexible control strategy is determined to meet the safety recovery condition and the corresponding strategy is executed. The determination logic is based on the comparison result, and the determination is triggered when the comparison output flag is true. Executing the corresponding strategy includes extracting the power adjustment amplitude and action timing parameters from the flexible control strategy generated in step S4, and sending a power adjustment command to the heater through the control interface. The power adjustment command includes a target power value and an execution time. The target power value is calculated based on the power adjustment amplitude, such as the current power + adjustment amplitude, and the execution time is determined based on the action timing. The execution process includes command verification, such as checking whether the target power is within the equipment's allowable range of 0 to maximum power. The execution also considers safety mutual... Locking mechanisms, such as monitoring key temperature parameters during power adjustment and halting execution if an anomaly is detected; execution optimization includes a step-by-step implementation strategy, such as breaking down large power adjustments into multiple small-step adjustments, with each step set to 10% to 20% of the total adjustment; recording operation logs during execution, including timestamps, adjustment parameters, and execution results; execution also includes a feedback mechanism, such as dynamically adjusting power values ​​based on real-time temperature changes, with the adjustment range not exceeding ±10% of the original adjustment value; execution also includes status monitoring, such as tracking the temperature response after power adjustment in real time, with a response delay time set to 1 to 5 seconds; execution optimization also involves resource scheduling, such as coordinating the power allocation of multiple heaters to avoid local overload.

[0038] When the high-frequency energy attenuation rate is less than the safety recovery threshold, the safety recovery condition is deemed not met, triggering a production interruption in the safety protection unit. The decision logic is based on a comparison result; a decision is triggered when the comparison output flag is false. Triggering the production interruption involves sending an interrupt signal to the safety protection unit, which then performs a shutdown operation according to preset safety procedures. The shutdown operation includes sequentially shutting down the heater, starting the cooling system, and activating the vacuum holding device. The operation sequence is set based on equipment safety requirements; for example, the heater is shut down first, followed by a 2-second delay before starting the cooling system. The triggering process includes an urgency assessment, such as determining the shutdown speed based on the difference between the high-frequency energy attenuation rate and the safety recovery threshold. If the difference is greater than 20... Emergency shutdown is executed at a certain percentage; the trigger also includes state saving, such as recording current process parameters and equipment status for subsequent analysis; trigger optimization includes gradual shutdown, such as reducing power to a safe level before complete shutdown to avoid thermal shock; alarm devices are activated simultaneously during the triggering process, including audible and visual alarms and remote notifications, with alarm levels divided into 1 to 3 according to the severity of the operating condition; the trigger also includes system self-checks, such as automatically checking the status of key equipment components after shutdown and generating a test report; the triggering process also considers data backup, such as saving the operating data of the most recent 10 process cycles for fault analysis; trigger optimization also includes recovery plans, such as automatically generating a system restart plan after safety conditions are met.

[0039] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0040] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0041] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions according to the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission; wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission includes infrared, microwave, etc. Computer-readable storage media can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0042] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0043] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0044] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0045] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0046] If a function is implemented as a software module 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 this application, in essence, or the part that contributes to the prior art, or a portion 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 a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this application. 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.

[0047] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0048] In conclusion, the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for real-time prediction and control of vacuum brazing temperature field based on digital twin, characterized in that, include: S1. Obtain real-time temperature field data and furnace pressure data of the vacuum brazing furnace; S2. Based on real-time temperature field data and furnace pressure data, determine whether extreme operating conditions approaching the safety threshold have occurred; S3. When it is determined that an extreme working condition has occurred, a weighted graph model is constructed based on real-time temperature field data, with high-temperature regions as nodes and the thermal influence relationship between nodes as weights. S4. Based on the weighted graph model, the set of nodes that play a key role in maintaining the structure of the anomalous temperature field is identified by analyzing the topological importance of nodes in the graph, and the physical region corresponding to the set of nodes is determined as the minimum intervention region; at the same time, a flexible control strategy for the minimum intervention region is generated. S5. Using a digital twin model, analyze the frequency domain energy changes of the graph signal caused by the flexible control strategy on the minimum intervention area in a weighted graph model; S6. Determine whether the flexible control strategy meets the safety recovery conditions based on the frequency domain energy change; if it does, execute the flexible control strategy; if it does not, trigger the production interruption of the safety protection unit.

2. The method for real-time prediction and control of vacuum brazing temperature field based on digital twin as described in claim 1, characterized in that, Acquire real-time temperature field data and furnace pressure data of the vacuum brazing furnace, including: Based on the multiple temperature acquisition zones divided inside the vacuum brazing furnace, temperature values ​​are collected in real time from each temperature acquisition zone to form temperature field data. Acquiring furnace pressure data includes real-time acquisition of pressure values ​​from pressure monitoring points in the vacuum brazing furnace.

3. The method for real-time prediction and control of vacuum brazing temperature field based on digital twin as described in claim 1, characterized in that, Based on real-time temperature field data and furnace pressure data, determine whether extreme operating conditions approaching the safety threshold have occurred, including: Identify whether there are high-temperature regions exceeding the temperature threshold in real-time temperature field data; Monitor whether the pressure data inside the furnace exceeds the pressure threshold; When a high-temperature zone is detected simultaneously and the pressure data inside the furnace exceeds the pressure threshold, it is determined that an extreme operating condition approaching the safety threshold has occurred.

4. The method for real-time prediction and control of vacuum brazing temperature field based on digital twin as described in claim 1, characterized in that, When an extreme operating condition is identified, a weighted graph model is constructed based on real-time temperature field data, with high-temperature regions as nodes and the thermal influence relationships between nodes as weights. This model includes: When an extreme working condition is determined to occur, the temperature acquisition area with a temperature value exceeding the high temperature threshold is extracted from the real-time temperature field data as a node. Calculate the weights of thermal influence relationships based on the spatial distance between nodes and the temperature gradient; The spatial distance is determined by the geometric center distance of the temperature acquisition area corresponding to the node, and the temperature gradient is determined by the temperature change rate between nodes. A weighted graph model is constructed using all nodes and the calculated weights of the thermal influence relationships.

5. The method for real-time prediction and control of vacuum brazing temperature field based on digital twin as described in claim 4, characterized in that, A weighted graph model is constructed using all nodes and the calculated thermal influence relationship weights. This includes: using the node set as the vertices of the graph and establishing edges between nodes based on the thermal influence relationship weights, thus forming a weighted graph structure; where nodes correspond to high-temperature regions and edge weights represent the intensity of the thermal influence relationship between nodes.

6. The method for real-time prediction and control of vacuum brazing temperature field based on digital twin as described in claim 1, characterized in that, Based on the weighted graph model, the set of nodes that play a key role in maintaining the structure of the anomalous temperature field is identified by analyzing the topological importance of nodes in the graph, and the physical region corresponding to the set of nodes is determined as the minimum intervention region. Simultaneously, flexible control strategies targeting the minimum intervention area are generated, including: Calculate the topological importance index of each node in the weighted graph model. The topological importance index is obtained by weighted combination of node degree centrality and eigenvector centrality. Nodes whose topological importance index exceeds a preset importance threshold are selected to form a set of key nodes; The key node set is mapped back to the physical space of the vacuum brazing furnace, and the corresponding temperature acquisition area is determined as the minimum intervention area. Based on the heat load distribution characteristics and heat influence relationship weights of the minimum intervention area, a flexible control strategy including power adjustment amplitude and action timing is generated.

7. The method for real-time prediction and control of vacuum brazing temperature field based on digital twin as described in claim 6, characterized in that, Mapping the set of key nodes back to the physical space of the vacuum brazing furnace and determining the corresponding temperature acquisition area as the minimum intervention area includes: locating the temperature acquisition area in the physical space by means of node identifiers based on the correspondence between key nodes and temperature acquisition areas, and defining the set of temperature acquisition areas as the minimum intervention area.

8. The method for real-time prediction and control of vacuum brazing temperature field based on digital twin as described in claim 1, characterized in that, Using a digital twin model, we analyze the frequency domain energy changes of the graph signal caused by a flexible control strategy in the minimum intervention area on a weighted graph model, including: The temperature field distribution difference between the minimum intervention area and its adjacent nodes before and after the application of the flexible control strategy is defined as the graph signal; The graph Fourier transform of the graph signal is performed based on the Laplace matrix of the weighted graph model to obtain the corresponding spectral distribution. Extract high-frequency components from the spectral distribution and calculate their energy values; By comparing the energy values ​​of high-frequency components before and after applying the flexible control strategy, the high-frequency energy attenuation rate is obtained as the frequency domain energy change of the graphical signal.

9. The method for real-time prediction and control of vacuum brazing temperature field based on digital twin as described in claim 8, characterized in that, The graph Fourier transform of the graph signal is performed based on the Laplace matrix of the weighted graph model to obtain the corresponding spectral distribution. This includes: calculating the Laplace matrix of the weighted graph model, and then applying the graph Fourier transform to the graph signal to obtain the spectral distribution; where the graph signal represents the temperature field distribution difference, and the spectral distribution contains frequency domain component information.

10. The method for real-time prediction and control of vacuum brazing temperature field based on digital twin as described in claim 1, characterized in that, Determine whether the flexible control strategy meets the safe recovery conditions based on frequency domain energy changes; If the conditions are met, a flexible control strategy will be implemented; otherwise, a production interruption will be triggered by the safety protection unit, including: Compare the high-frequency energy decay rate with a preset safe recovery threshold; When the high-frequency energy attenuation rate is greater than or equal to the safe recovery threshold, the flexible control strategy is determined to meet the safe recovery condition and the corresponding strategy is executed. When the high-frequency energy attenuation rate is less than the safety recovery threshold, it is determined that the safety recovery condition is not met and the production interruption of the safety protection unit is triggered.

Images