Automatic pressure protection method for air pressure furnace, platform and medium

Abstract

The invention discloses an automatic pressure protection method, platform and medium for a gas pressure furnace, and relates to the technical field of furnace body pressure control, and the method comprises the steps: starting a multi-source sensor to carry out the collection of the hearth pressure, pressure change rate, temperature data, flow data and furnace body strain signals of the gas pressure furnace, and building a collection data set; executing multi-scale working condition identification of the collected data set, and establishing an extreme working condition criterion; constructing a virtual pressure gauge family, performing joint trust fusion of extreme working condition criteria by using the virtual pressure gauge family, and establishing a pressure output result; predicted pressure triggering protection management of the air pressure furnace is conducted through the dynamic pressure fence. The technical problems that in the prior art, air pressure furnace pressure monitoring is not accurate or comprehensive enough, extreme working conditions are difficult to recognize in advance, and consequently operation safety and stability are poor are solved, and the technical effects that multi-dimensional accurate monitoring of the air pressure furnace pressure is achieved, and the operation safety and stability of the air pressure furnace are effectively improved are achieved.

Description

Technical Field

[0001] This invention relates to the field of furnace pressure control technology, specifically to an automatic pressure protection method, platform, and medium for pressure furnaces. Background Technology

[0002] In the industrial application of pressure furnaces, the stable control of furnace pressure is directly related to processing quality and equipment operation safety. Currently, commonly used pressure protection technologies for pressure furnaces in the industry mostly rely on a single physical pressure gauge to collect furnace pressure data. This only enables real-time monitoring of a single pressure point and cannot cover multi-dimensional operating conditions related to pressure, such as temperature, flow rate, and furnace strain. This results in one-sided pressure monitoring, failing to fully reflect the true pressure state within the furnace. Furthermore, many use fixed threshold triggering protection mechanisms, which lack the ability to predict pressure change trends, making it difficult to identify potential extreme conditions such as abnormal pressure amplitudes or sudden changes in the slope of pressure changes. They also cannot dynamically adjust the protection threshold based on the real-time operating parameters of the pressure furnace, easily leading to delayed protection response or false triggering. Ultimately, this affects the safety and stability of the pressure furnace operation and fails to meet the stringent requirements of high-precision industrial production for pressure protection systems.

[0003] Existing technologies suffer from insufficient accuracy and comprehensiveness in pressure monitoring of gas pressure furnaces, making it difficult to identify extreme operating conditions in advance, resulting in poor operational safety and stability. Summary of the Invention

[0004] This application provides an automatic pressure protection method, platform, and medium for pressure furnaces, which addresses the technical problem that existing pressure monitoring of pressure furnaces is not accurate and comprehensive enough, making it difficult to identify extreme operating conditions in advance, resulting in poor operational safety and stability.

[0005] In view of the above problems, this application provides an automatic pressure protection method, platform and medium for pressure furnaces.

[0006] A first aspect of this application provides an automatic pressure protection method for a pressure furnace, the method comprising: The process involves activating multi-source sensors to acquire furnace pressure, pressure change rate, temperature data, flow rate data, and furnace strain signals from the pressure furnace, establishing a dataset. The multi-source sensors include heterogeneous physical pressure gauges. Under sliding window driving, multi-scale operating condition identification is performed on the acquired dataset, extracting pressure amplitude, slope, and second-order derivative features to establish extreme operating condition criteria. A virtual pressure gauge family is constructed based on the heterogeneous physical pressure gauges and the acquired dataset. This virtual pressure gauge family includes a time-series evolution virtual table and a cloned virtual table. Joint trust fusion of the extreme operating condition criteria is performed using this virtual pressure gauge family to establish a pressure output result, which is equipped with a credibility identifier. A dynamic pressure fence, including pressure upper limit, pressure change rate upper limit, and energy upper limit, is configured using the pressure output result and the acquired dataset. This dynamic pressure fence is then used for predictive pressure trigger protection management of the pressure furnace.

[0007] A second aspect of this application provides an automatic pressure protection platform for a pressure furnace, the platform comprising: The data acquisition module is used to initiate the acquisition of furnace pressure, pressure change rate, temperature data, flow rate data, and furnace body strain signals from multi-source sensors to establish a data acquisition dataset. The multi-source sensors include heterogeneous physical pressure gauges. The extreme condition criterion establishment module is used to perform multi-scale condition identification of the data acquisition dataset under sliding window driving, extracting pressure amplitude, slope, and second derivative features to establish extreme condition criteria. The pressure output result establishment module is used to construct a virtual pressure gauge family based on the heterogeneous physical pressure gauges and the data acquisition dataset. The virtual pressure gauge family includes a time-series evolution virtual table and a clone virtual table. The virtual pressure gauge family is used for joint trust fusion of extreme condition criteria to establish pressure output results, which are set with a credibility identifier. The protection management module is used to configure a dynamic pressure fence containing pressure upper limit, pressure change rate upper limit, and energy upper limit using the pressure output results and the data acquisition dataset, and then use the dynamic pressure fence to perform predictive pressure trigger protection management for the pressure furnace.

[0008] In a third aspect of this application, a computer-readable storage medium is provided storing a computer program for executing the automatic pressure protection method for a pressure furnace provided in this application.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages: A multi-source sensor system is activated to acquire furnace pressure, pressure change rate, temperature data, flow rate data, and furnace body strain signals of the pressure furnace, establishing a dataset. Under sliding window driving, multi-scale operating condition identification of the acquired dataset is performed to establish extreme operating condition criteria. A virtual pressure gauge family is constructed based on the heterogeneous physical pressure gauges and the acquired dataset. This virtual pressure gauge family is used for joint trust fusion of extreme operating condition criteria to establish pressure output results. Using the pressure output results and the acquired dataset, a dynamic pressure fence is configured, including pressure upper limit, pressure change rate upper limit, and energy upper limit. This dynamic pressure fence is then used for predictive pressure trigger protection management of the pressure furnace. This achieves multi-dimensional and accurate monitoring of pressure in the pressure furnace, effectively improving the safety and stability of its operation. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic flowchart of an automatic pressure protection method for a pressure furnace provided in an embodiment of this application.

[0012] Figure 2 This is a schematic diagram of the structure of an automatic pressure protection platform for a pressure furnace provided in an embodiment of this application.

[0013] Figure labeling: Data set establishment module 10, extreme working condition criterion establishment module 20, pressure output result establishment module 30, protection management module 40. Detailed Implementation

[0014] This application provides an automatic pressure protection method, platform, and medium for pressure furnaces, which addresses the technical problem that existing pressure monitoring of pressure furnaces is not accurate and comprehensive enough, making it difficult to identify extreme operating conditions in advance, resulting in poor operational safety and stability.

[0015] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0016] Example 1, as Figure 1As shown, this application provides an automatic pressure protection method for a pressure furnace, the method comprising: Step S100: Start the multi-source sensor to collect furnace pressure, pressure change rate, temperature data, flow data and furnace body strain signal of the pressure furnace, and establish a collection dataset, wherein the multi-source sensor includes a heterogeneous physical pressure gauge.

[0017] Specifically, a multi-source sensor system is activated to comprehensively collect key parameters during the operation of the pressure furnace: real-time furnace pressure data is acquired using heterogeneous physical pressure gauges containing physical pressure detection devices with different accuracies and measurement principles; the pressure change rate is simultaneously derived using a pressure sensor calculation module; temperature data of different areas within the furnace is collected using a temperature sensor array; gas input / output flow data is recorded using a flow sensor; and strain signals of the furnace structure are captured using a strain sensor. All collected parameters are then systematically integrated according to timestamps to ensure synchronized acquisition times and consistent data formats, ultimately constructing a dataset covering the core operating status of the pressure furnace.

[0018] Step S200: Perform multi-scale working condition identification of the collected dataset under the sliding window drive, extract pressure amplitude, slope, and second derivative features, and establish extreme working condition criteria.

[0019] Specifically, multi-scale operating condition identification was performed on the established dataset. During processing, furnace pressure-related data in the dataset were analyzed in time-segmented units using a sliding window approach, extracting three key features: pressure amplitude, slope, and second derivative features. Pressure amplitude reflects the fluctuation range of furnace pressure within the corresponding window period; slope reflects the trend and rate of pressure change over time; and second derivative features are used to capture the acceleration of pressure changes to identify potential sudden changes. Subsequently, combined with historical safe operating condition data and extreme dangerous operating condition data accumulated from the gas pressure furnace's operation, statistical analysis was performed on the extracted pressure amplitude, slope, and second derivative features. This clarified the safety boundaries and danger thresholds for each of the three features under different operating conditions. Finally, the feature parameters and corresponding thresholds were combined to establish extreme operating condition criteria that can be used to determine whether the gas pressure furnace is in an extreme risk state.

[0020] Step S300: Construct a virtual pressure gauge family based on the heterogeneous physical pressure gauges and the collected dataset. The virtual pressure gauge family includes a time-series evolution virtual table and a clone virtual table. Use the virtual pressure gauge family to perform joint trust fusion of extreme working condition criteria to establish pressure output results. The pressure output results are set with a credibility identifier.

[0021] Specifically, a virtual pressure gauge family is first constructed based on heterogeneous physical pressure gauges and the acquired dataset. The time-series measurement signals of each heterogeneous physical pressure gauge are acquired, and an autoregressive prediction model is built for each heterogeneous physical pressure gauge using the correlation parameters in the acquired dataset, thus establishing a time-series evolution virtual table. Subsequently, the acquired signals of each heterogeneous physical pressure gauge in the acquired dataset are filtered, denoised, and zero-biased corrected. Based on the historical residuals and transient noise levels of each heterogeneous physical pressure gauge, a fusion weighting is performed to construct a physics-dominated virtual table. Simultaneously, temperature and flow data from the acquired dataset are used as the first feature set, and the effective volume and heating power data of the pressure furnace are read as the second feature set. These two feature sets are used for thermodynamic model prediction to establish a model-dominated virtual table. The physics-dominated virtual table and the model-dominated virtual table are combined as a clone virtual table, forming a complete virtual pressure gauge family with the time-series evolution virtual table. Next, the virtual pressure table family is used to carry out joint trust fusion of extreme working condition criteria. First, the extreme working condition criteria of each virtual table in the virtual pressure table family are calculated separately to establish the risk index vector of each table. Then, the risk index vector of each table is adaptively weighted and fused with the extreme working condition criteria. The fusion result is used as the pressure output result. At the same time, a credibility label is set for the pressure output result to intuitively reflect the reliability of the pressure data.

[0022] Step S400: After configuring a dynamic pressure fence including pressure upper limit, pressure change rate upper limit and energy upper limit using the pressure output result and the collected dataset, the dynamic pressure fence is used to perform predictive pressure trigger protection management of the pressure furnace.

[0023] Specifically, a dynamic pressure fence is configured using pressure output results and the acquired dataset: The acquired dataset is analyzed to extract the circumferential / axial strain change rate and local strain energy accumulation from the furnace strain signal; the furnace temperature uniformity (temperature difference between regions) and temperature time-series abrupt change rate are extracted from the temperature data. These parameters are normalized and used as structural safety correction terms. Combined with the reliability indicator of the pressure output results and the rated safe pressure value of the pressure furnace, a compensation correction is made to obtain the dynamically changing pressure upper limit; the pressurization rate benchmark is calculated based on the temperature and flow rate in the acquired dataset, and the pressure change rate in the pressure output results is extracted. By combining time-series prediction characteristics with residual analysis of historical safe operating condition curves based on furnace strain signals, residual coupling corrections are obtained. These parameters are normalized and used as rate safety correction terms. The pressure change rate upper limit is then obtained by compensation correction based on the credibility identifier. Instantaneous gas energy storage is calculated by combining pressure output results with the effective volume of the pressure furnace. Local energy characteristics are obtained by accumulating instantaneous gas energy storage within a sliding window. These characteristics are compared with historical local energy peaks to obtain comparison correction results. The energy upper limit is reconstructed by combining the accumulated local strain energy. Finally, a dynamic pressure fence containing the upper limits of pressure, pressure change rate, and energy is formed. Then, the dynamic pressure fence is used to perform predictive pressure trigger protection management: based on the pressure output results and the time-series evolution virtual table, the furnace pressure in the near future is predicted, i.e., the predicted pressure. The predicted pressure is compared with the three upper limits of the dynamic pressure fence to determine whether protection should be triggered. If the predicted pressure exceeds a certain upper limit, a trigger identification result is generated, including the trigger value level, such as slight over-limit or severe over-limit; and the trigger time level, such as immediate trigger or imminent trigger. According to the trigger identification result, the corresponding protection decision is activated, such as executing an early warning decision for slight over-limit and issuing an audible and visual warning; executing a forced protection decision for severe over-limit, automatically cutting off the gas input and opening the pressure relief valve; executing an emergency switching decision in case of emergency, switching to the backup safety system; and executing an abnormal degradation decision in case of equipment abnormality, reducing the operating power, thereby realizing automatic protection management of the pressure furnace pressure.

[0024] In one possible implementation, step S400 further includes: Step S410: Analyze the collected dataset and obtain the strain change rate and local strain energy accumulation characteristics of the pressure furnace in the circumferential or axial direction based on the furnace body strain signal.

[0025] Step S420: Extract the furnace body temperature uniformity characteristics and temperature time-series abrupt change rate based on the temperature data in the collected dataset.

[0026] Step S430: After normalizing the strain change rate, local strain energy accumulation, furnace temperature uniformity characteristics and temperature time-series abrupt change rate, the normalization result is used as a structural safety correction term. The pressure output result is used as a confidence indicator and a safe pressure value to perform compensation correction based on the structural safety correction term, and the pressure upper limit is established.

[0027] Specifically, the collected dataset is first structured and analyzed to filter and extract the furnace body strain signals acquired in real time by strain sensors. The extracted furnace body strain signals are then processed using a time-domain differential algorithm. By calculating the ratio of the difference in strain values ​​between adjacent sampling times to the time interval, the strain change rate in the circumferential (along the furnace body's circumference) or axial (along the furnace body's height) direction of the pressure furnace body is obtained, quantifying the dynamic change trend of the furnace body's structural strain over time. Simultaneously, based on Hooke's law and a strain energy calculation model, the furnace body strain signals are converted into stress data for the corresponding region. Then, through integral calculations, the stress-strain relationship within a specific time interval is accumulated to obtain the total energy generated by the deformation of the local area of ​​the pressure furnace body under stress, thereby extracting the characteristics of the accumulated local strain energy.

[0028] Temperature data from the pressure furnace collected by the temperature sensor array was selected from the dataset. This data covers real-time temperature values ​​in different areas of the furnace chamber and key points on the furnace wall. For extracting the furnace temperature uniformity characteristics, a regional temperature difference analysis method was used to calculate the real-time average temperature of all monitoring points. Then, the absolute deviation of each monitoring point's temperature from the average was calculated. By statistically analyzing the maximum value, standard deviation, and percentage of deviation, the uniformity of temperature distribution within the furnace chamber was quantified, forming the furnace temperature uniformity characteristic. For extracting the temperature temporal abrupt change rate, a sliding time window combined with a difference algorithm was used. The difference between adjacent temperature values ​​within each time window was calculated at preset time intervals (e.g., 1 second). The ratio of this difference to the time interval was used as the instantaneous temperature change rate. Values ​​with instantaneous temperature change rates exceeding a preset normal fluctuation range were selected, and their frequency and maximum amplitude were statistically analyzed. Finally, a temperature temporal abrupt change rate reflecting drastic temperature fluctuations over a short period was formed.

[0029] The Min-Max normalization algorithm is used to map the four parameters—the strain change rate, the cumulative local strain energy, the extracted furnace body temperature uniformity characteristics, and the temperature time series abrupt change rate—to the numerical range of [0, 1], respectively, to eliminate the influence of differences in the dimensions and numerical ranges of different parameters. The four parameters after normalization are integrated into a structural safety correction term to comprehensively reflect the influence of the current furnace body structural state and temperature state on the pressure bearing capacity. Subsequently, the confidence level indicators attached to the established pressure output results are retrieved. For example, high confidence level corresponds to a value range of 0.8-1.0, and medium confidence level corresponds to a value range of 0.5-0.8. Combined with the preset safety pressure value of the pressure furnace, and based on the equipment's rated parameters and historical safe operation data, the safety pressure value is compensated and corrected through weighted calculation according to the rule that the higher the confidence level indicator value, the greater the correction weight of the structural safety correction item on the safety pressure value. If the structural safety correction item value is too low, it means that the furnace structure is stable and the temperature distribution is balanced, so the correction is slightly increased based on the safety pressure value; if the structural safety correction item value is too high, it means that there are potential risks in the furnace structure and the temperature fluctuation is large, so the correction is appropriately decreased based on the safety pressure value. Finally, a dynamic pressure upper limit adapted to the current operating conditions is established.

[0030] In one possible implementation, step S400 further includes: Step S440: Analyze the collected dataset and obtain the charging rate benchmark based on the temperature data and flow rate data.

[0031] Step S450: Extract the pressure change rate and time-series prediction features from the pressure output results.

[0032] Step S460: Perform residual analysis under the historical safe operating condition curve based on the furnace body strain signal, and establish residual coupling correction amount.

[0033] Step S470: After normalizing the charging rate benchmark, pressure change rate, time-series prediction characteristics, and residual coupling correction, a rate safety correction term is established.

[0034] Step S480: Use the confidence identifier to compensate and correct the rate safety correction item to establish an upper limit for the rate of pressure change.

[0035] Specifically, the collected dataset is structured and filtered to extract temperature data, including real-time temperatures at multiple monitoring points within the furnace and furnace wall temperature, as well as flow rate data, including gas input flow rate, output flow rate, and net flow rate. Subsequently, based on the ideal gas law, PV=nRT, where P is pressure, V is the effective volume of the furnace, n is the amount of gas, R is the gas constant, and T is the absolute temperature, the combined effect of gas flow rate and temperature changes per unit time on furnace pressure is calculated to determine the benchmark pressurization rate for safe operation of the pressure furnace under the current temperature and flow conditions. This benchmark reflects the reasonable range of stable pressure rise rates under the current operating conditions.

[0036] The generated pressure output results are retrieved, and two key parameters are extracted from them: first, the pressure change rate, which is obtained by calculating the ratio of the pressure difference between adjacent time nodes in the pressure output results to the time interval, directly reflecting the real-time rate of change of the current furnace pressure; second, the time-series prediction features. For the extraction of time-series prediction features, an autoregressive time series prediction algorithm is used. Based on the pressure data of multiple consecutive time nodes in the pressure output results, a pressure time-series prediction model is constructed. The model calculates and derives the possible trend and numerical range of furnace pressure changes within a preset time period (such as the next 10 seconds), and integrates this trend and range into time-series prediction features.

[0037] The furnace body strain signal of the pressure furnace is extracted from the collected dataset. Simultaneously, historical safe operating condition curves stored in the historical safe operation database of the pressure furnace are retrieved. These curves contain historical furnace body strain signals, pressure data, temperature data, and other related parameters under operating conditions matching the current detection scenario. The extracted furnace body strain signal is then compared point-by-point with historical furnace body strain signals of the corresponding time period and operating condition type in the historical safe operating condition curve. The numerical deviation between the two on the same time dimension, i.e., the residual, is calculated. Next, the pattern of residual change over time and the correlation between residual and pressure and temperature changes in the historical safe operating condition curve are analyzed to identify the main influencing factors of residual generation, such as minor deformation of the furnace body structure and slight sensor drift. Finally, the residual data is processed using a data fitting algorithm. Combining the coupling correlation characteristics between the residual and operating condition parameters, a quantitative parameter reflecting the degree of deviation between the current furnace body strain state and the historical safe state is constructed. This establishes the residual coupling correction quantity, providing data support for the subsequent construction of the rate safety correction term.

[0038] The Z-score standardization method is used to convert the four parameters—the obtained pressurization rate benchmark, the extracted pressure change rate and time series prediction features, and the established residual coupling correction amount—into standardized values ​​with a mean of 0 and a standard deviation of 1, eliminating interference caused by differences in the dimensions and magnitudes of different parameters. Subsequently, through weighted summation, weights are set according to the degree of influence of each parameter on rate safety. For example, if the weight of the pressure change rate is higher than that of the pressurization rate benchmark, the four standardized parameters are integrated to establish a rate safety correction term that can comprehensively reflect rate safety risks.

[0039] The reliability identifier attached to the pressure output result generated by the joint trust fusion of the virtual pressure gauge family reflects the reliability of the pressure output result. Then, based on the different levels of the reliability identifier (high reliability, medium reliability, low reliability), corresponding correction weights are assigned to the established rate safety correction item. The higher the reliability identifier level, the more reliable the pressure output result, and the greater the influence weight of the rate safety correction item on the final pressure change rate upper limit. The lower the reliability identifier level, the lower the influence weight of the rate safety correction item is appropriately reduced to avoid bias caused by unreliable data. Next, the weighted rate safety correction item is fused with the preset basic pressure change rate safety value of the pressure furnace (determined based on equipment structural strength standards and historical safe operation data). If the weighted rate safety correction item value is low, it indicates that the current rate-related parameters are more in line with safety requirements, and the basic safety value can be appropriately increased to adapt to the operating conditions. If the value is high, it indicates a tendency for the rate to exceed the risk limit, and the basic safety value is correspondingly decreased to ensure safety, ultimately establishing a pressure change rate upper limit adapted to the current operating conditions.

[0040] In one possible implementation, step S400 further includes: Instantaneous gas energy storage is calculated based on the pressure output results and the effective volume of the pressure furnace.

[0041] Instantaneous gas energy storage accumulation is performed within a sliding window to establish local energy characteristics.

[0042] Historical local energy peak values ​​are compared using the local energy characteristics to establish a comparison and correction result.

[0043] The upper limit of energy is reconstructed based on the comparison and correction results and the cumulative amount of local strain energy.

[0044] Specifically, the generated pressure output result is first retrieved, which includes the real-time furnace pressure value P after being fused with the virtual pressure gauge family. Simultaneously, the effective volume V of the pressure furnace is read, representing the space volume of gas that the furnace can hold, as specified by the equipment. An energy storage calculation algorithm based on the thermodynamic characteristics of ideal gases is employed, with the gas energy storage formula E=k×P×V as its core, where k is the gas characteristic coefficient. Based on the preset gas type within the pressure furnace, the real-time furnace pressure value P and the effective volume V are substituted into the formula. Numerical calculations are then performed to obtain the energy value contained in the gas within the furnace at the current operating moment of the pressure furnace. This energy value is the instantaneous gas energy storage. During the calculation process, the reliability indicator of the pressure output result must be referenced simultaneously. If the reliability indicator is in the high reliability range, the current pressure value is directly used for calculation. If the reliability indicator is low, it is corrected using the fine-tuning coefficient of the k value in the formula to ensure that the accuracy of the instantaneous gas energy storage calculation result matches the actual operating state of the equipment.

[0045] A suitable sliding window duration is determined based on the current operating conditions. This duration needs to be preset in conjunction with the fluctuation frequency of parameters such as pressure and temperature in the pressure furnace to ensure complete coverage of the critical energy change cycle. Subsequently, within each sliding window cycle, instantaneous gas energy storage data calculated moment by moment is continuously collected and acquired. By accumulating the instantaneous gas energy storage values ​​within the window, the total energy value corresponding to that sliding window is obtained. This total energy value is defined as the local energy characteristic. This characteristic not only reflects the overall cumulative level of furnace gas energy within the window period, but also captures the trend of energy change by comparing the local energy characteristics between different sliding windows.

[0046] The historical safe operation database of the pressure furnace is retrieved, and historical operation records that are consistent with or highly similar to the current operating conditions, such as temperature range, process stage, and gas type, are selected. Local energy peak data for each historical period's corresponding sliding window are extracted to form a historical local energy peak dataset. Subsequently, the local energy characteristics calculated for the current sliding window are compared one by one with the historical local energy peak dataset. Quantitative indicators such as the difference, ratio, and percentage deviation between the current local energy characteristics and historical peaks are calculated to analyze the degree of deviation of the current local energy characteristics from the historical safe peak. If the current local energy characteristics are significantly lower than the historical peak, it indicates sufficient energy safety redundancy; if they are close to or reach the historical peak, the risk of energy exceeding the safe range needs to be considered. A graded correction rule is set according to the degree of deviation: deviation within 10% is considered slight deviation, 10%~30% is moderate deviation, and more than 30% is severe deviation. Corresponding correction coefficients are assigned to different deviation levels. The current local energy characteristics are combined with the correction coefficients to calculate the final comparison and correction results that reflect the difference between the current energy state and the historical safe state.

[0047] The comparison and correction results are retrieved, including the deviation level of the current local energy characteristics from the historical local energy peak and the corresponding correction coefficient. At the same time, the cumulative amount of local strain energy previously obtained from the furnace body strain signal is extracted. A multi-factor weighted fusion algorithm is used to construct an energy upper limit calculation model. The correction coefficient of the comparison and correction results is used as the weight of the gas energy dimension. For example, the correction coefficient is 0.9 when the deviation level is slight, 0.7 when it is moderate, and 0.5 when it is severe. The ratio of the cumulative amount of local strain energy to the preset strain energy threshold is used as the weight of the structural energy dimension. For example, the ratio is 0.8 when the cumulative amount reaches 80% of the threshold and 0.6 when it reaches 60%. The initial energy benchmark value is multiplied by the two dimension weights respectively and then summed to obtain the preliminary energy upper limit value. The preliminary value is then compared with the historical safe energy upper limit under the same operating conditions. If the preliminary value exceeds the historical limit, it is further reduced by 5% to 10%. Finally, the reconstructed energy upper limit is determined and output to ensure that the upper limit meets the requirements of both gas energy safety and furnace body structural safety.

[0048] In one possible implementation, step S300 further includes: Step S310: Obtain the time-series measurement signal of the heterogeneous physical pressure gauge.

[0049] Step S320: Based on the time-series measurement signal and the acquired dataset, construct an autoregressive prediction model for each heterogeneous physical pressure gauge, and use the construction results to establish a time-series evolution virtual table.

[0050] Step S330: After filtering, denoising, and zero-bias correction of the acquired signals of each heterogeneous physical pressure gauge in the acquired dataset, the data is fused and weighted according to the historical residuals and transient noise levels of each heterogeneous physical pressure gauge to construct a physical-dominated virtual table.

[0051] Step S340: Use the temperature data and flow rate data in the collected dataset as the first feature set, read the effective volume and heating power data of the pressure furnace as the second feature set, and use the first feature set and the second feature set to perform thermodynamic model prediction and establish a model-dominated virtual table.

[0052] Step S350: The physical-dominated virtual table and the model-dominated virtual table are used as clone virtual tables, which together with the time-series evolution virtual table form a virtual pressure table family.

[0053] Specifically, a real-time communication connection is established with a multi-source sensor system to receive pressure data collected by various heterogeneous physical pressure gauges during continuous operation at a preset sampling frequency (e.g., 10 times per second). This data is automatically sorted in chronological order to form a sequence containing timestamps and corresponding pressure measurements; this sequence is the time-series measurement signal of the heterogeneous physical pressure gauges.

[0054] First, for each heterogeneous physical pressure gauge, the corresponding time-series measurement signal is retrieved. Simultaneously, correlation parameters related to the monitoring area of ​​that pressure gauge are extracted from the acquired dataset, such as temperature data, flow data, and furnace strain signals for the corresponding area. Using historical pressure values ​​from the time-series measurement signals as core training data and correlation parameters as auxiliary features, the optimal lag order of the model is determined through autoregressive analysis. The pressure values ​​from the previous N time points and correlation parameters are used to predict the pressure at the current time, constructing a dedicated autoregressive prediction model for each heterogeneous physical pressure gauge. After model training, real-time time-series measurement signals and correlation parameters are input to generate a pressure prediction sequence for a preset future time period. This prediction sequence is then integrated with the real-time acquired time-series measurement signals along the time axis to form a complete pressure time-series dataset that simultaneously includes historical measured data, current data, and future prediction data. This allows the creation of a virtual time-series evolution table that dynamically reflects pressure evolution trends.

[0055] The raw acquisition signals of each heterogeneous physical pressure gauge are extracted from the acquired dataset. These signals are then subjected to filtering, noise reduction, and bias correction operations. Digital filtering technology removes high-frequency noise caused by environmental interference and sensor vibration, while the bias correction algorithm eliminates the pressure gauge's own systematic errors, resulting in a purified signal with improved accuracy. Subsequently, the historical operating database of each heterogeneous physical pressure gauge is retrieved to extract its historical measurement residuals—the deviation between historical actual pressure values ​​and calibration values—while simultaneously monitoring the transient noise level during the current acquisition process. Based on the principle that smaller historical residuals and lower transient noise levels indicate higher measurement reliability, corresponding fusion weights are assigned to the purified signals of each pressure gauge. Finally, the purified signals of all pressure gauges are weighted and summed to obtain a set of pressure data sequences that comprehensively reflect the measurement results of each pressure gauge. This constructs a physical-dominated virtual table supported by physical measurement data.

[0056] Temperature and flow data were selected from the collected dataset. Z-score standardization was used to preprocess the two types of data to eliminate dimensional differences, and they were then combined to form the first feature set. Next, the effective volume and heating power data of the pressure furnace were retrieved. After verifying the accuracy of the parameter units (effective volume in cubic meters, heating power in kilowatts) using a data format validation algorithm, this constituted the second feature set. Subsequently, a gradient boosting decision tree (GBDT)-based machine learning algorithm was used to construct a thermodynamic prediction model. The first and second feature sets were divided into a training set (70%), a validation set (20%), and a test set (10%). The model was trained using the training set, and the learning rate, tree depth, and other hyperparameters were adjusted using 5-fold cross-validation on the validation set to ensure that the pressure prediction error of the model on the test set was controlled within a preset range. After the model training is completed, the first feature set and the second feature set acquired in real time are input into the model. The forward propagation algorithm outputs the furnace pressure prediction value for the future preset time step. The prediction value is organized into a structured data sequence according to the time series to establish a model-dominated virtual table. At the same time, the confidence interval of the prediction value is calculated in real time through the model evaluation algorithm to reflect the reliability of the virtual table data.

[0057] The functional roles of the three types of virtual tables constructed in the early stage were clearly defined. The physical-driven virtual tables and the model-driven virtual tables were classified as clone virtual tables. These two types of virtual tables rely on the fusion of physical measurement data and the prediction of thermodynamic models, respectively, to provide support for pressure monitoring from different dimensions. Subsequently, these two types of clone virtual tables were integrated with the established time-series evolution virtual table. Data compatibility among the three types of virtual tables was ensured through a unified data format, ultimately forming a family of virtual pressure tables covering physical measurement verification, model-assisted prediction, and pressure time-series evolution trends.

[0058] In one possible implementation, step S300 further includes: Step S360: Calculate the extreme working condition criteria for a single gauge based on the virtual pressure gauge family, and establish a single gauge risk index vector.

[0059] Step S370: Adaptively weightedly fuse the single-table risk index vector and the extreme working condition criterion, and output the adaptive weighted fusion result as the stress output result.

[0060] Specifically, the system utilizes data from three types of virtual pressure tables within the virtual pressure table family: time-series evolution virtual tables, physical-dominated virtual tables, and model-dominated virtual tables. For each type of virtual table, extreme condition criteria are calculated: based on the pressure time-series data of the time-series evolution virtual table, it calculates whether the pressure amplitude exceeds a preset safety threshold, whether the pressure change slope is abnormal, and whether the pressure fluctuation acceleration reflected by the second derivative exceeds the standard; for the physical-dominated virtual table, it analyzes whether there are extreme condition characteristics such as instantaneous pressure mutations or sustained high pressure by combining its fused and weighted pressure data; for the model-dominated virtual table, it determines whether there is a risk of predicted values ​​exceeding limits by comparing the predicted pressure with the historical safe pressure range. The extreme condition judgment results of each type of virtual table are converted into quantified risk indicators (e.g., 0-100 points, with higher scores indicating higher risk), arranged sequentially by virtual table type, forming a single-table risk indicator vector containing all single-table risk indicators.

[0061] Based on the characteristics of each virtual table and the reliability of historical data, initial weights are assigned to time-series evolution virtual tables, physical-dominated virtual tables, and model-dominated virtual tables. These weights are dynamically adjusted in conjunction with the credibility rating of the stress output results; if a virtual table has a high credibility rating, its weight is increased, and vice versa. Then, each risk indicator in the single-table risk indicator vector is multiplied by its corresponding weight, and all weighted risk indicators are summed to obtain a weighted and integrated comprehensive risk value. Simultaneously, referencing preset extreme condition criteria (e.g., a comprehensive risk value exceeding 80 is considered high risk, 50-80 is medium risk, and below 50 is low risk), the comprehensive risk value is matched to the condition level. Finally, the result of the comprehensive risk value minus the condition level is output as the stress output result.

[0062] In one possible implementation, step S400 further includes: The pressure output results are used to predict the pressure of the pressure furnace and establish the predicted pressure.

[0063] Based on the predicted pressure and the dynamic pressure fence, pressure trigger identification is performed, and a pressure trigger identification result is established. The pressure trigger identification result includes the trigger value level and the trigger time level.

[0064] The trigger identification result is used to activate the protection decision to perform protection management.

[0065] Specifically, the pressure output obtained through joint trust fusion of virtual pressure gauge families, including real-time pressure values, confidence indicators, and associated time-series data, is retrieved. This data, along with temperature data, flow data, and furnace strain signals from the collected dataset, are used as input features. A long short-term memory network machine learning algorithm is employed to construct a pressure prediction model for the pressure furnace. During model training, historical pressure output results and corresponding operating condition data are used as training samples. The Adam optimizer minimizes the mean square error between the predicted and actual pressures, while a Dropout layer is introduced to prevent overfitting. Samples are weighted according to the confidence indicator of the pressure output results, with a weight coefficient of 1.2 for high-confidence samples and 0.8 for low-confidence samples, improving the model's learning efficiency on reliable data. After model training, the real-time pressure output results and current operating condition features are input, and a prediction step size is set, such as outputting a predicted value every 5 seconds for the next 30 seconds. The long short-term memory network model's gating mechanism captures long-term dependencies in the pressure time-series data, generating a continuous pressure prediction sequence, which is ultimately used as the predicted pressure.

[0066] The dynamic pressure fence is defined by three criteria: pressure upper limit, pressure change rate upper limit, and energy upper limit. The real-time pressure value, pressure change rate, and calculated energy value corresponding to the predicted pressure are then compared one by one with these three upper limits. If the real-time pressure value, pressure change rate, and energy value do not exceed the pressure upper limit, no trigger is detected. If any one of these exceeds the upper limit, a trigger value level classification is initiated: a difference of less than 10% of the upper limit is Level 1, 10%–30% is Level 2, and more than 30% is Level 3. Simultaneously, the trigger time level is determined based on the time point at which the predicted pressure reaches the trigger condition: reaching the trigger condition within the next 5 seconds is Emergency Level, reaching it within the next 5–15 seconds is Warning Level, and reaching it within the next 15–30 seconds is Alert Level. Finally, the determined trigger value level and trigger time level are integrated to form a trigger identification result containing both pieces of information.

[0067] The system retrieves the trigger value level and trigger time level from the trigger identification results and activates them according to the preset protection decision matching rules: If the trigger identification result is trigger value level 1 and trigger time warning level, the system activates the early warning decision, sends a text warning message to the operation terminal through the control system, and starts a low-frequency audible and visual alarm to remind the operator to pay attention to changes in the pressure furnace's operating conditions; if the trigger value is trigger value level 2 and trigger time warning level, the system activates the forced protection decision, automatically opens the pressure relief valve of the pressure furnace for gradient pressure relief, and reduces the power of the heating system to control the pressure within a safe range; if the trigger value is trigger value level 3 and trigger time emergency level, the system activates the emergency switching decision, immediately cuts off the heating power supply, closes the air inlet valve, and starts the emergency cooling system to quickly reduce the furnace temperature and pressure to prevent dangerous accidents; if the trigger identification result shows abnormal pressure fluctuations but does not reach the high-risk level, i.e., trigger value level 1 and trigger time warning level, the system activates the abnormal degradation decision, adjusts the operating parameters of the pressure furnace, such as reducing the air inlet flow rate, to maintain the equipment operating under low-load safe conditions. During the implementation of protection decisions, key parameters such as pressure and temperature of the pressure furnace are monitored in real time, and protection measures are dynamically adjusted according to parameter changes until the equipment condition returns to normal, thereby completing the protection management of the pressure furnace.

[0068] In one possible implementation, step S400 further includes: The protection decisions include early warning decisions, mandatory protection decisions, emergency switching decisions, and abnormal degradation decisions.

[0069] Specifically, the early warning decision is applicable to low-risk scenarios. It is activated when the trigger value level is 1 and the trigger time level is the alert level, reminding operators to pay attention to changes in operating conditions through low-frequency audible and visual alarms and text notifications on the operating terminal. The forced protection decision is for medium-risk situations. It is activated when the trigger value level is 2 and the trigger time level is the early warning level, automatically opening the pressure relief valve for gradient pressure relief and reducing the heating system power to control the pressure within a safe range. The emergency switching decision is used for high-risk situations. It is executed when the trigger value level is 3 and the trigger time level is the emergency level, immediately cutting off the heating power supply, closing the air intake valve, and simultaneously activating the emergency cooling system to quickly contain the danger. The abnormal degradation decision is for situations where the pressure fluctuation is abnormal but does not reach the high-risk level. For example, when the trigger value level is 1 and the trigger time level is the early warning level, it switches the equipment to a low-load safe operating condition by adjusting operating parameters such as the air intake flow rate to maintain stable operation.

[0070] Example 2, based on the same inventive concept as the automatic pressure protection method for the pressure furnace in the foregoing examples, such as... Figure 2 As shown, this application provides an automatic pressure protection platform for a pressure furnace. The platform and method embodiments in this application are based on the same inventive concept. The platform includes: The dataset acquisition module 10 is used to initiate the acquisition of furnace pressure, pressure change rate, temperature data, flow rate data and furnace body strain signal of the gas pressure furnace by multi-source sensors, and to establish the acquisition dataset. The multi-source sensors include heterogeneous physical pressure gauges.

[0071] The extreme working condition criterion establishment module 20 is used to perform multi-scale working condition identification of the collected dataset under the sliding window drive, extract pressure amplitude, slope and second derivative features, and establish extreme working condition criteria.

[0072] The pressure output result establishment module 30 is used to construct a virtual pressure table family based on the heterogeneous physical pressure table and the collected dataset. The virtual pressure table family includes a time-series evolution virtual table and a clone virtual table. The virtual pressure table family is used to perform joint trust fusion of extreme working condition criteria to establish pressure output results. The pressure output results are set with a credibility identifier.

[0073] The protection management module 40 is used to configure a dynamic pressure fence, which includes a pressure upper limit, a pressure change rate upper limit, and an energy upper limit, using the pressure output result and the collected dataset, and then to perform predictive pressure trigger protection management of the pressure furnace using the dynamic pressure fence.

[0074] Furthermore, the platform is also used to implement the following functions: The collected dataset is analyzed to obtain the strain change rate and local strain energy accumulation characteristics of the pressure furnace in the circumferential or axial direction based on the furnace body strain signal. The furnace body temperature uniformity characteristics and temperature time-series abrupt change rate are extracted based on the temperature data in the collected dataset. After normalizing the strain change rate, local strain energy accumulation, furnace body temperature uniformity characteristics, and temperature time-series abrupt change rate, the normalization result is used as a structural safety correction term. The pressure upper limit is established by using the confidence indicator of the pressure output result and the safe pressure value to perform compensation correction based on the structural safety correction term.

[0075] Furthermore, the platform is also used to implement the following functions: The collected dataset is analyzed to obtain a pressurization rate benchmark based on temperature and flow data; the pressure change rate and time-series prediction features are extracted from the pressure output results; residual analysis is performed on the historical safe operating condition curve based on the furnace body strain signal to establish a residual coupling correction quantity; after normalizing the pressurization rate benchmark, pressure change rate, time-series prediction features, and residual coupling correction quantity, a rate safety correction term is established; the rate safety correction term is compensated and corrected using the confidence identifier to establish an upper limit for the pressure change rate.

[0076] Furthermore, the platform is also used to implement the following functions: Instantaneous gas energy storage is calculated based on the pressure output results and the effective volume of the pressure furnace; instantaneous gas energy storage is accumulated within a sliding window to establish local energy characteristics; historical local energy peak values ​​are compared using the local energy characteristics to establish comparison correction results; and the energy upper limit is reconstructed based on the comparison correction results and the accumulated local strain energy.

[0077] Furthermore, the platform is also used to implement the following functions: The process involves acquiring time-series measurement signals from the heterogeneous physical pressure gauges; constructing an autoregressive prediction model for each heterogeneous physical pressure gauge based on the time-series measurement signals and the acquired dataset; establishing a time-series evolution virtual table using the construction results; filtering, denoising, and zero-bias correction on the acquired signals of each heterogeneous physical pressure gauge in the acquired dataset; fusing and weighting the historical residuals and transient noise levels of each heterogeneous physical pressure gauge to construct a physics-dominated virtual table; using the temperature and flow data from the acquired dataset as the first feature set, and reading the effective volume and heating power data of the pressure furnace as the second feature set; using the first and second feature sets to perform thermodynamic model prediction to establish a model-dominated virtual table; and combining the physics-dominated virtual table and the model-dominated virtual table as clone virtual tables to form a virtual pressure gauge family with the time-series evolution virtual table.

[0078] Furthermore, the platform is also used to implement the following functions: Based on the virtual pressure gauge family, the extreme working condition criteria for a single gauge are calculated, and a single gauge risk index vector is established. The single gauge risk index vector and the extreme working condition criteria are then adaptively weighted and fused, and the adaptive weighted fusion result is output as the pressure output result.

[0079] Furthermore, the platform is also used to implement the following functions: The pressure output results are used to predict the pressure of the pressure furnace and establish the predicted pressure; based on the predicted pressure and the dynamic pressure fence, pressure trigger identification is performed and a pressure trigger identification result is established, which includes the trigger value level and the trigger time level; the trigger identification result is used to activate protection decisions to perform protection management.

[0080] Furthermore, the platform is also used to implement the following functions: The protection decisions include early warning decisions, mandatory protection decisions, emergency switching decisions, and abnormal degradation decisions.

[0081] Example 3: Based on the same inventive concept as the automatic pressure protection method for a pressure furnace in the previous examples, this example provides a computer-readable storage medium for storing software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the automatic pressure protection method for a pressure furnace in this application. The processor executes the software programs, instructions, and modules stored in the memory to perform various functional applications and data processing of the computer device, thereby implementing the aforementioned automatic pressure protection method for a pressure furnace.

[0082] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0083] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0084] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. An automatic pressure protection method for a pressure furnace, characterized in that, The method includes: The multi-source sensor is activated to collect furnace pressure, pressure change rate, temperature data, flow data and furnace body strain signal of the pressure furnace, and a collection dataset is established. The multi-source sensor includes a heterogeneous physical pressure gauge. Under the sliding window drive, multi-scale working condition identification of the collected dataset is performed, and pressure amplitude, slope and second derivative features are extracted to establish extreme working condition criteria. A virtual pressure table family is constructed based on the heterogeneous physical pressure table and the collected dataset. The virtual pressure table family includes a time-series evolution virtual table and a clone virtual table. The virtual pressure table family is used to perform joint trust fusion of extreme working condition criteria to establish pressure output results. The pressure output results are set with a credibility identifier. After configuring a dynamic pressure fence that includes an upper limit on pressure, an upper limit on the rate of pressure change, and an upper limit on energy using the pressure output results and the collected dataset, the dynamic pressure fence is used for predictive pressure trigger protection management of the gas pressure furnace.

2. The automatic pressure protection method for a pressure furnace as described in claim 1, characterized in that, The configuration of a dynamic pressure fence, including a pressure upper limit, a pressure change rate upper limit, and an energy upper limit, using the pressure output results and the collected dataset includes: The collected dataset is analyzed to obtain the strain change rate and local strain energy accumulation characteristics of the pressure furnace in the circumferential or axial direction based on the furnace body strain signal. Based on the temperature data in the collected dataset, extract the furnace body temperature uniformity characteristics and temperature time-series abrupt change rate; After normalizing the strain change rate, local strain energy accumulation, furnace temperature uniformity characteristics, and temperature time-series abrupt change rate, the normalization result is used as a structural safety correction term. The pressure upper limit is established by using the reliability indicator of the pressure output result and the safe pressure value to perform compensation correction based on the structural safety correction term.

3. The automatic pressure protection method for a pressure furnace as described in claim 2, characterized in that, The configuration of a dynamic pressure fence, including a pressure upper limit, a pressure change rate upper limit, and an energy upper limit, using the pressure output results and the collected dataset, further includes: Analyze the collected dataset to obtain the charging rate benchmark based on temperature and flow data; The pressure change rate and time-series prediction features are extracted from the pressure output results; Based on the furnace body strain signal, residual analysis was performed under the historical safe operating condition curve to establish residual coupling correction quantity; After normalizing the charging rate benchmark, pressure change rate, time-series prediction characteristics, and residual coupling correction, a rate safety correction term is established. The reliability identifier is used to compensate and correct the rate safety correction term, thereby establishing an upper limit for the rate of pressure change.

4. The automatic pressure protection method for a pressure furnace as described in claim 3, characterized in that, The configuration of a dynamic pressure fence, including a pressure upper limit, a pressure change rate upper limit, and an energy upper limit, using the pressure output results and the collected dataset, further includes: The instantaneous gas energy storage is calculated based on the pressure output results and the effective volume of the pressure furnace. Instantaneous gas energy storage accumulation is performed within a sliding window to establish local energy characteristics; The local energy characteristics are used to compare historical local energy peaks, and a comparison correction result is established. The upper limit of energy is reconstructed based on the comparison and correction results and the cumulative amount of local strain energy.

5. The automatic pressure protection method for a pressure furnace as described in claim 1, characterized in that, The step of constructing a virtual pressure gauge family based on the heterogeneous physical pressure gauges and the collected dataset includes: Acquire the time-series measurement signal of the heterogeneous physical pressure gauge; Based on the time-series measurement signals and the acquired dataset, an autoregressive prediction model is constructed for each heterogeneous physical pressure gauge, and a time-series evolution virtual table is established using the construction results; After filtering, denoising, and zero-bias correction of the acquired signals of each heterogeneous physical pressure gauge in the acquired dataset, a physical-dominated virtual table is constructed by fusing and weighting the historical residuals and transient noise levels of each heterogeneous physical pressure gauge. The temperature data and flow data in the collected dataset are used as the first feature set, and the effective volume and heating power data of the pressure furnace are read as the second feature set. The first feature set and the second feature set are used to perform thermodynamic model prediction and establish a model-dominated virtual table. The physical-dominated virtual table and the model-dominated virtual table are used as clone virtual tables, which together with the time-series evolution virtual table constitute a family of virtual pressure tables.

6. The automatic pressure protection method for a pressure furnace as described in claim 5, characterized in that, The joint trust fusion of extreme condition criteria using the virtual pressure gauge family to establish pressure output results includes: Based on the virtual pressure gauge family, the extreme working condition criteria for a single gauge are calculated, and a single gauge risk index vector is established. The single-table risk indicator vector and the extreme working condition criterion are adaptively weighted and fused, and the adaptive weighted fusion result is output as the stress output result.

7. The automatic pressure protection method for a pressure furnace as described in claim 1, characterized in that, The method of using the dynamic pressure fence for predictive pressure triggering protection management of the pressure furnace includes: The pressure output results are used to predict the pressure of the pressure furnace and establish the predicted pressure. Based on the predicted pressure and the dynamic pressure fence, pressure trigger identification is performed, and a pressure trigger identification result is established, which includes the trigger value level and the trigger time level. The trigger identification result is used to activate the protection decision to perform protection management.

8. The automatic pressure protection method for a pressure furnace as described in claim 7, characterized in that, The protection decisions include early warning decisions, mandatory protection decisions, emergency switching decisions, and abnormal degradation decisions.

9. An automatic pressure protection platform for a pressure furnace, characterized in that, The platform is used to implement the automatic pressure protection method for a pressure furnace according to any one of claims 1-8, and the platform comprises: The dataset acquisition module is used to initiate the acquisition of furnace pressure, pressure change rate, temperature data, flow rate data and furnace body strain signal of the pressure furnace by multi-source sensors, and to establish the acquisition dataset. The multi-source sensors include heterogeneous physical pressure gauges. The extreme working condition criterion establishment module is used to perform multi-scale working condition identification of the collected dataset under the sliding window drive, extract pressure amplitude, slope, and second derivative features, and establish extreme working condition criteria. The pressure output result establishment module is used to construct a virtual pressure table family based on the heterogeneous physical pressure table and the collected dataset. The virtual pressure table family includes a time-series evolution virtual table and a clone virtual table. The virtual pressure table family is used to perform joint trust fusion of extreme working condition criteria to establish pressure output results. The pressure output results are set with a credibility identifier. The protection management module is used to configure a dynamic pressure fence, which includes a pressure upper limit, a pressure change rate upper limit, and an energy upper limit, using the pressure output results and the collected dataset. Then, it uses the dynamic pressure fence to perform predictive pressure trigger protection management for the pressure furnace.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the automatic pressure protection method for a pressure furnace as described in any one of claims 1-8.

Images