A method and system for detecting gas leakage from a vacuum heat treatment furnace

Abstract

The present application relates to the technical field of vacuum equipment operation state monitoring and fault diagnosis, and particularly relates to a vacuum heat treatment furnace gas leakage detection method and system. According to the difference between the total pressure data at the current sampling moment and the historical adjacent sampling moment, the total pressure increment is obtained. According to the difference between the temperature data at the current sampling moment and the historical adjacent sampling moment, the temperature increment is obtained. According to the data fluctuation characteristics of the total pressure increment at the current sampling moment and the historical sampling moment, and the data fluctuation characteristics of the temperature increment at the current sampling moment and the historical sampling moment, the trend correlation characteristic index of the current sampling moment is analyzed. According to the total pressure increment, the sound pressure data and the trend correlation characteristic index at the current sampling moment, the gas leakage judgment index of the current sampling moment is determined, and whether the vacuum heat treatment furnace leaks is judged.

Description

Technical Field

[0001] This invention relates to the field of vacuum equipment operation status monitoring and fault diagnosis technology, specifically to a method and system for detecting gas leakage in a vacuum heat treatment furnace. Background Technology

[0002] Vacuum heat treatment processes require equipment to maintain a strict vacuum environment during the heating and pressure holding stages to prevent oxidation of metal workpieces at high temperatures. In industrial settings, the rate of increase in total furnace pressure is typically used to determine if a leak has occurred. However, under actual high-temperature pressure holding conditions, the pressure rise inside the furnace often involves a mixture of two different gas increase sources. On one hand, the heater's heating causes desorption between the workpiece surface and the porous insulation material, resulting in normal heat-induced gas release. On the other hand, failure of seals such as equipment flanges can allow external atmospheric pressure air to leak into the furnace, creating a genuine leak. These two phenomena exhibit highly overlapping rising characteristics on a single pressure transmitter reading, and due to the significant differences in surface area between different batches of workpieces, the normal heat-induced gas release will fluctuate non-linearly with temperature variations.

[0003] Existing technologies typically use fixed pressure alarm thresholds for leak detection. However, fixed pressure alarm thresholds cannot distinguish between genuine leaks and normal material release, easily leading to false alarms and equipment shutdowns. Furthermore, if the pressure alarm threshold is too high, it can mask genuine, minute leaks. To aid in diagnosis, existing technologies attempt to add ultrasonic sensors to the outside of the flange to detect frictional noises caused by leaks. However, industrial workshops are filled with mechanical vibrations from cranes, pumps, and other equipment, making ultrasonic leak detection alone highly susceptible to false positives. Therefore, existing leak detection technologies face a very high risk of false alarms and missed detections under the high-temperature conditions of vacuum heat treatment. Summary of the Invention

[0004] To address the high risk of false alarms and missed alarms in existing technologies for gas leakage detection under high-temperature conditions in vacuum heat treatment, the present invention aims to provide a gas leakage detection method and system for vacuum heat treatment furnaces. The specific technical solution adopted is as follows: This invention proposes a method for detecting gas leakage in a vacuum heat treatment furnace, the method comprising: Acquire the total pressure, temperature, and sound pressure data inside the furnace at the current sampling moment; The total pressure increment is obtained based on the difference between the total pressure data in the furnace at the current sampling time and the historical adjacent sampling time; the temperature increment is obtained based on the difference between the temperature data at the current sampling time and the historical adjacent sampling time; and the trend correlation characteristic indicators at the current sampling time are analyzed based on the data fluctuation characteristics of the total pressure increment at the current sampling time and the data fluctuation characteristics of the temperature increment at the current sampling time and the historical sampling time. Based on the total pressure increment, sound pressure data, and trend correlation characteristic indicators, a leakage judgment index is determined at the current sampling time; based on the leakage judgment index, it is determined whether the vacuum heat treatment furnace is leaking.

[0005] Further, the step of obtaining the total pressure increment based on the difference between the total furnace pressure data at the current sampling time and those at historical adjacent sampling times includes: The difference between the total pressure data inside the furnace at the current sampling time and the total pressure data inside the furnace at the historical adjacent sampling time is used as the total pressure increment.

[0006] Further, the step of obtaining the temperature increment based on the difference between the temperature data at the current sampling time and the temperature data at historical adjacent sampling times includes: The temperature increment is the difference between the temperature data at the current sampling time and the temperature data at historical adjacent sampling times.

[0007] Furthermore, the analysis of trend correlation indicators for the current sampling time based on the data fluctuation characteristics of the total pressure increment between the current sampling time and historical sampling times, and the data fluctuation characteristics of the temperature increment between the current sampling time and historical sampling times, includes: Based on the data fluctuation characteristics of the total pressure increment between the current sampling time and the historical sampling time within a preset historical window, the air pressure fluctuation state sequence is analyzed; the elements in the air pressure fluctuation state sequence are air pressure fluctuation states. Based on the data fluctuation characteristics of the temperature increment between the current sampling time and the historical sampling time within a preset historical window, the temperature fluctuation state sequence is analyzed; the elements in the temperature fluctuation state sequence are the temperature fluctuation states. Based on the pressure fluctuation state sequence and temperature fluctuation state sequence, trend correlation characteristic indicators are determined.

[0008] Furthermore, determining the trend correlation characteristic indicators based on the pressure fluctuation state sequence and the temperature fluctuation state sequence includes: Any pressure fluctuation state in the pressure fluctuation state sequence is taken as the first pressure state, and the previous pressure fluctuation state adjacent to the first pressure state is taken as the second pressure state; the second pressure state and the first pressure state constitute a pressure state combination; based on the probability of the pressure state combination appearing in the pressure fluctuation state sequence and the probability of the second pressure state appearing in the pressure fluctuation state sequence, the conditional information entropy is calculated. The temperature fluctuation state corresponding to the first air pressure state under a preset delay is taken as the first temperature state; the second air pressure state, the first air pressure state, and the first temperature state constitute a ternary combination, and the second air pressure state and the first temperature state constitute a binary combination; based on the probability of the ternary combination appearing in the air pressure fluctuation state sequence and the temperature fluctuation state sequence, and the probability of the binary combination appearing in the air pressure fluctuation state sequence and the temperature fluctuation state sequence, the joint conditional information entropy is calculated. The difference between the conditional information entropy and the joint conditional information entropy is used as a trend correlation feature indicator.

[0009] Furthermore, before analyzing the trend correlation characteristic indicators of the current sampling time based on the data fluctuation characteristics of the total pressure increment between the current sampling time and the historical sampling time, and the data fluctuation characteristics of the temperature increment between the current sampling time and the historical sampling time, the method further includes: If the number of current sampling times and historical sampling times is less than or equal to the length of the preset historical window, the preset historical window is suspended and trend correlation characteristic index analysis is not performed; if the number of current sampling times and historical sampling times is greater than the length of the preset historical window, trend correlation characteristic index analysis is performed.

[0010] Further, determining the leakage judgment index at the current sampling moment based on the total pressure increment, sound pressure data, and trend correlation characteristic indicators includes: The total pressure increment, sound pressure data, and trend correlation feature index are processed by feature fusion multiplication to obtain the leakage judgment index at the current sampling time.

[0011] Furthermore, before determining the leakage judgment index at the current sampling moment based on the total pressure increment, sound pressure data, and trend correlation characteristic indicators, the method further includes: If the total pressure increment is less than or equal to 0, the leakage judgment index is set to 0; if the total pressure increment is greater than 0, the leakage judgment index is calculated.

[0012] Further, determining whether the vacuum heat treatment furnace is leaking based on the leakage judgment index includes: If the leakage judgment index is less than or equal to the preset upper limit threshold, it is determined that the vacuum heat treatment furnace has not leaked; if the leakage judgment index is greater than the preset upper limit threshold, it is determined that the vacuum heat treatment furnace has leaked.

[0013] The present invention also proposes a gas leakage detection system for a vacuum heat treatment furnace, the system comprising: The data acquisition module is used to acquire the total pressure, temperature and sound pressure data inside the furnace at the current sampling time; The feature analysis module is used to obtain the total pressure increment based on the difference between the total pressure data in the furnace at the current sampling time and the historical adjacent sampling time; to obtain the temperature increment based on the difference between the temperature data at the current sampling time and the historical adjacent sampling time; and to analyze the trend correlation feature indicators at the current sampling time based on the data fluctuation characteristics of the total pressure increment at the current sampling time and the data fluctuation characteristics of the temperature increment at the current sampling time and the historical sampling time. The results analysis module is used to determine the leakage judgment index at the current sampling time based on the total pressure increment, sound pressure data, and trend correlation characteristic indicators; and to determine whether the vacuum heat treatment furnace is leaking based on the leakage judgment index.

[0014] The present invention has the following beneficial effects: This invention obtains the total pressure increment based on the difference between the total pressure data in the furnace at the current sampling time and the historical adjacent sampling time. This total pressure increment represents the actual increase in gas in the furnace within a unit sampling interval, reflecting the rate of gas pressure accumulation caused by workpiece venting or leakage. It also obtains the temperature increment based on the difference between the temperature data at the current sampling time and the historical adjacent sampling time. This temperature increment represents the actual temperature rise in the furnace within a unit sampling interval, reflecting the dynamic driving force of heating on workpiece desorption. Furthermore, based on the data fluctuation characteristics of the total pressure increment and the temperature increment at the current and historical sampling times, the invention analyzes the trend correlation indicators at the current sampling time. This quantifies the closeness of the temporal correlation between the gas pressure fluctuation characteristics and the temperature fluctuation characteristics, providing a key indicator for distinguishing the fundamental physical driving source of gas increase in the furnace.

[0015] Based on the total pressure increment, sound pressure data, and trend correlation characteristic indicators, a leakage judgment index is determined at the current sampling time. The leakage judgment index characterizes the comprehensive severity of abnormal characteristics caused purely by external atmospheric pressure gas leakage, after removing normal heat release components and environmental mechanical noise interference. Based on this leakage judgment index, it is determined whether the vacuum heat treatment furnace is leaking. Simultaneously, leakage events are cross-validated and comprehensively judged from the dimensions of gas mass accumulation, direct acoustic perception, and causal consistency, effectively eliminating the problem of misjudgment and missed judgment caused by single indicators under complex operating conditions due to factors such as temperature rise rate fluctuations, workpiece material differences, and environmental acoustic interference. Attached Figure Description

[0016] Figure 1 This is a flowchart of a gas leakage detection method for a vacuum heat treatment furnace provided in one embodiment of the present invention. Detailed Implementation

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0018] The following description, in conjunction with the accompanying drawings, details the specific scheme of the gas leakage detection method and system for a vacuum heat treatment furnace provided by the present invention.

[0019] Please see Figure 1 The diagram illustrates a flowchart of a gas leak detection method for a vacuum heat treatment furnace according to an embodiment of the present invention. The method includes: Step S1: Obtain the total pressure data, temperature data, and sound pressure data inside the furnace at the current sampling time.

[0020] To ensure the validity of the data and the reliability of the judgment, this embodiment requires that a cold-state ultimate vacuum pressure test be performed before starting the heating program in the vacuum heat treatment furnace. The heating program will only be started if the operator or the host computer confirms that there is no air leakage in the flange under cold conditions.

[0021] In this embodiment of the invention, at a fixed sampling interval, the latest refreshed cached data in the communication registers of each sensor is synchronously captured at each sampling moment. This includes the absolute pressure value output by the vacuum transmitter, the temperature readings of all probes in the thermocouple array distributed inside the heating chamber, and the effective sound pressure value of the ultrasonic sensors arranged at the outer nodes of each flange.

[0022] The absolute pressure value obtained at each sampling time is taken as the total pressure data inside the furnace at each sampling time; the maximum value among all temperature readings obtained at each sampling time is taken as the temperature data at each sampling time; and the maximum value among all effective sound pressure values ​​obtained at each sampling time is taken as the sound pressure data at each sampling time.

[0023] Since the desorption rate of gas on the surface of a metal workpiece directly depends on the highest temperature it receives, and heating inside a furnace is often uneven, extracting the maximum value from all temperature readings can most sensitively and proactively reflect the state of the heat source that will lead to violent gas release.

[0024] Since external air leakage often occurs only at a specific flange gap, if the effective sound pressure values ​​of all ultrasonic sensors are averaged, the actual leakage noise will be diluted by the quiet environment of other channels (i.e., the feature is masked). Therefore, the maximum value among all effective sound pressure values ​​can most quickly capture the highest abnormal friction noise occurring locally, thus amplifying the leakage feature.

[0025] In one specific implementation of this invention, the fixed sampling interval is set to 1 second.

[0026] In this embodiment of the invention, the total number of ultrasonic sensors arranged on the vacuum heat treatment furnace is set to L. Each ultrasonic sensor is assigned a continuous positive integer from 1 to L as a number. When data is collected synchronously at each sampling moment, not only is the maximum value among all effective sound pressure values ​​extracted as the sound pressure data at each sampling moment, but the number of the ultrasonic sensor that generated the sound pressure data is also latched simultaneously. In addition, a static mapping table is pre-stored in the memory of the industrial control computer. This mapping table records the correspondence between each ultrasonic sensor number and the three-dimensional physical position of the vacuum heat treatment furnace.

[0027] Step S2: Obtain the total pressure increment based on the difference between the total pressure data in the furnace at the current sampling time and the historical adjacent sampling time; obtain the temperature increment based on the difference between the temperature data at the current sampling time and the historical adjacent sampling time; analyze the trend correlation characteristic indicators of the current sampling time based on the data fluctuation characteristics of the total pressure increment at the current sampling time and the data fluctuation characteristics of the temperature increment at the current sampling time and the historical sampling time.

[0028] Since the initial vacuum degree and absolute pressure baseline of the vacuum heat treatment furnace vary significantly between different batches of furnace loading or under equipment aging conditions, and the heat release of the workpiece and external leakage are both dynamic accumulation processes of pressure, and this embodiment samples according to a fixed sampling interval, the total pressure increment between two adjacent sampling times can directly reflect the net increase rate of gas mass in the furnace, eliminating the interference of the initial static base pressure difference and absolute pressure fluctuation on the subsequent time series correlation analysis.

[0029] The total pressure increment represents the actual increase in gas inside the furnace within a unit sampling interval, reflecting the rate of gas pressure accumulation caused by workpiece venting or leakage.

[0030] Since the desorption process of workpieces in a vacuum heat treatment furnace is essentially driven by dynamic temperature changes rather than determined by a static absolute temperature value at a certain moment, and the differences in material, furnace loading, and thermocouple installation position of different batches of workpieces will lead to inherent deviations in absolute temperature readings, the temperature increment between two adjacent sampling moments can directly reflect the dynamic driving force of heating on the desorption of gas on the workpiece surface, eliminating the interference of individual equipment differences, thermocouple calibration deviations, and uneven initial temperature distribution of workpieces on subsequent time-series correlation analysis.

[0031] The temperature increment represents the actual temperature rise in the furnace within a unit sampling interval, reflecting the dynamic driving force of heating on the desorption phenomenon of the workpiece. In terms of physical causal logic, it achieves a strict data mapping and alignment with the total pressure increment.

[0032] Under normal operating conditions, a definite causal relationship exists between the amount of gas released by the workpiece through pyrolysis and adsorption in a vacuum heat treatment furnace and the actual temperature rise. Temperature increment is the cause, and total pressure increment is the effect; both should exhibit a stable, coordinated fluctuation pattern over time. However, when gas leakage occurs in the furnace, the abnormal influx of external gas disrupts this inherent causal relationship, causing the fluctuation pattern of total pressure increment to deviate from the expected response trajectory of temperature increment. Therefore, it is necessary to analyze the trend correlation characteristics of the current sampling time by comparing the fluctuation characteristics of total pressure increment and temperature increment data between the current sampling time and historical sampling times.

[0033] The trend correlation characteristic index quantifies the degree of temporal correlation between pressure fluctuation characteristics and temperature fluctuation characteristics, and is a key indicator for distinguishing the fundamental physical driving source of gas increase in the furnace. The larger the trend correlation characteristic index, the more the increase in gas pressure in the vacuum heat treatment furnace is completely controlled by the heating operation of the heater, which is a normal phenomenon of workpiece desorption and gas release due to heat. The smaller the trend correlation characteristic index, the more the pressure fluctuation in the vacuum heat treatment furnace is out of the control of the heat source, indicating that external atmospheric pressure gas independent of temperature changes is being drawn into the furnace, indicating a real external physical gas leakage fault.

[0034] Preferably, in some implementations of the present invention, the total pressure increment is obtained based on the difference between the total furnace pressure data at the current sampling time and the historical adjacent sampling times, including: The total pressure increment at the current sampling time is obtained by subtracting the total pressure data at the previous historical sampling time adjacent to the current sampling time from the total pressure data at the current sampling time.

[0035] Preferably, in some implementations of the present invention, obtaining the temperature increment based on the difference between the temperature data at the current sampling time and the temperature data at historical adjacent sampling times includes: The temperature increment at the current sampling time is obtained by subtracting the temperature data from the temperature data of the previous historical sampling time adjacent to the current sampling time from the temperature data at the current sampling time.

[0036] Step S3: Determine the leakage judgment index at the current sampling time based on the total pressure increment, sound pressure data, and trend correlation characteristic index; determine whether the vacuum heat treatment furnace is leaking based on the leakage judgment index.

[0037] Since the total pressure increment only reflects the net increase in gas mass in the furnace at the current sampling time, it cannot distinguish whether the increment comes from normal venting due to thermal desorption of the workpiece or abnormal gas leaking in through sealing defects; sound pressure data can capture the characteristic acoustic signal generated when gas flows through tiny gaps at high speed through acoustic sensors, and is directly sensitive to leakage events, but it is easily affected by environmental interference factors such as vibration of the vacuum pump in the furnace and mechanical noise generated by thermal expansion and contraction of the workpiece, and there is a risk of false alarm when used alone; although the trend correlation characteristic index can quantitatively characterize the degree of causal coupling between the total pressure increment and the temperature increment, and indirectly reflect whether there is an abnormal pressure source independent of temperature driving from the time series dimension, it is itself a dimensionless relative measure and cannot directly characterize the severity of the leakage.

[0038] Therefore, by integrating the total pressure increment, sound pressure data, and trend correlation characteristic indicators to determine the leakage judgment index at the current sampling time, leakage events can be cross-validated and comprehensively judged from the dimensions of gas mass accumulation, direct acoustic perception, and causal consistency. This effectively eliminates the problem of misjudgment and omission caused by single indicators under complex working conditions due to factors such as temperature rise rate fluctuations, workpiece material differences, and environmental acoustic interference.

[0039] The leakage judgment index represents the overall severity of abnormal characteristics caused by the leakage of external atmospheric pressure gas at the current sampling time, after removing normal heated gas components and environmental mechanical noise interference.

[0040] Preferably, in some implementations of the present invention, determining the leakage judgment index at the current sampling time based on the total pressure increment, sound pressure data, and trend correlation characteristic indicators includes: The total pressure increment, sound pressure data, and trend correlation feature index are processed by feature fusion multiplication to obtain the leakage judgment index at the current sampling time.

[0041] Based on the total pressure increment as the fundamental characteristic, an acoustic amplification factor is constructed using sound pressure data, and a dynamic attenuation damping is constructed using trend correlation characteristic indicators. When the gas increase is controlled by the temperature rise, the dynamic attenuation damping approaches 0, suppressing the leakage judgment index to a minimum value; when the gas increase is independent of the temperature rise and accompanied by abnormal noise, the dynamic attenuation damping approaches 1, and the acoustic amplification factor is activated, resulting in a significant jump in the leakage judgment index, thus achieving decoupling of the aliased gas driving source and extraction of real leakage energy.

[0042] The leakage judgment index retains the same pressure dimension (pressure unit Pa) as the total pressure increment, and in physical semantics represents the net equivalent leakage pressure increase after removing the interference of normal heat release.

[0043] Specifically, by using feature fusion multiplication operations to process the total pressure increment, sound pressure data, and trend correlation feature indicators, the formula for obtaining the leakage judgment index at the current sampling time is as follows: in, The leakage judgment index represents the current sampling time; This represents the total pressure increment at the current sampling moment; Represents the preset acoustic gain coefficient; The sound pressure data represents the current sampling moment; Represents the preset reference sound pressure level; e represents the natural constant; A trend correlation characteristic indicator representing the current sampling time.

[0044] The preset acoustic gain coefficient represents the amplification weight of abnormal sounds. Its dimension is the reciprocal of the sound pressure data dimension (sound pressure unit dB), and the empirical range is 0.5dB. -1 Up to 2.0dB -1 For high-noise-floor burst environments, a smaller value is used; for high-sensitivity scenarios, a larger value is used. In this embodiment, it is set to 0.5dB. -1 .

[0045] This represents the acoustic amplification factor constructed based on sound pressure data, and the result is a dimensionless value. Characterizes the inherent mechanical vibration and background noise constant of a vacuum heat treatment furnace under normal, leak-free conditions. Used to strip away normal mechanical environmental noise and extract abnormal net sound pressure signals generated purely by external gas being squeezed into the gap at high speed; Used to constrain the lower limit of net sound pressure level, if S is less than or equal to If the difference is forcibly truncated to 0, the acoustic amplification factor will degenerate to a constant of 1. This mechanism ensures that in real micro-leakage scenarios where no abnormal noise is detected or in normal pressure holding scenarios, the formula calculation will not cause the leakage judgment index to be abnormally reduced due to the accidental decrease in ambient noise. This ensures that the calculation process is always dominated by the total pressure increment and trend correlation characteristic indicators, maintaining the unidirectional incremental leakage prevention characteristic of the algorithm logic.

[0046] Represents dynamic attenuation damping; when the trend correlation characteristic indicator E is large, The value of the leak detection index decreases sharply and approaches 0, thus keeping the leak detection index within the safe baseline; when the trend correlation characteristic index E approaches 0 (the gas pressure fluctuation is completely independent of the heater state, i.e., a pure leak occurs). The pressure naturally and smoothly tends towards a constant of 1. At this point, the damping fails and the calculation is not overloaded. The leakage judgment index is directly dominated by the total pressure increment and the net sound pressure, which perfectly achieves the decoupling of the aliasing state in mathematics.

[0047] Furthermore, before determining the leakage judgment index at the current sampling moment based on the total pressure increment, sound pressure data, and trend correlation characteristic indicators, the process also includes: If the total pressure increment is less than or equal to 0, it indicates that the vacuum heat treatment furnace is in a state of pressure reduction or constant pressure maintenance at the current sampling time. In a vacuum environment, leakage of external atmospheric pressure air will inevitably lead to an increase in total pressure. This situation physically eliminates the possibility of air leakage, so the leakage judgment index is directly set to 0.

[0048] If the total pressure increment is greater than 0, it indicates that the total pressure inside the furnace is rising at the current sampling time, and the gas mass is substantially increasing. This pressure increase may be caused by normal thermal desorption and gas release from the workpiece, or it may be caused by actual external leakage, which is a typical cascading state. In this case, the calculation of the leakage judgment index mentioned above must be triggered.

[0049] In summary, this invention obtains the total pressure increment based on the difference between the total pressure data inside the furnace at the current sampling time and the historical adjacent sampling times; it obtains the temperature increment based on the difference between the temperature data at the current sampling time and the historical adjacent sampling times; and it analyzes the trend correlation characteristic index of the current sampling time based on the data fluctuation characteristics of the total pressure increment and the temperature increment at the current sampling time and the historical sampling times. Based on the total pressure increment, sound pressure data, and trend correlation characteristic index of the current sampling time, a leakage judgment index is determined to determine whether the vacuum heat treatment furnace is leaking.

[0050] Preferably, in some implementations of the present invention, the trend correlation characteristic index of the current sampling time is analyzed based on the data fluctuation characteristics of the total pressure increment between the current sampling time and the historical sampling time, and the data fluctuation characteristics of the temperature increment between the current sampling time and the historical sampling time, including: Based on the data fluctuation characteristics of the total pressure increment between the current sampling time and the historical sampling time within a preset historical window, the air pressure fluctuation state sequence is analyzed; the elements in the air pressure fluctuation state sequence are air pressure fluctuation states. Based on the data fluctuation characteristics of the temperature increment between the current sampling time and the historical sampling time within a preset historical window, the temperature fluctuation state sequence is analyzed; the elements in the temperature fluctuation state sequence are the temperature fluctuation states. Based on the pressure fluctuation state sequence and temperature fluctuation state sequence, trend correlation characteristic indicators are determined.

[0051] Because the pressure inside a vacuum heat treatment furnace does not exhibit an ideal monotonic linear change during actual operation, but rather experiences high-frequency random disturbances and oscillations due to process characteristics such as pulsating adjustments in heating power, periodic start-stop cycles of the vacuum pump, and gas release from the rupture of the workpiece surface film. If leakage is directly determined based on the total pressure increment at multiple sampling times, normal process oscillations are easily misinterpreted as leakage signals, or slowly developing minute leaks are drowned out by background noise. Therefore, this embodiment transforms continuously changing pressure values ​​into discrete gas pressure fluctuation sequences based on the data fluctuation characteristics of the total pressure increment within a preset historical window. This not only filters out interference from high-frequency random noise but also desensitizes the absolute numerical differences in gas pressure changes.

[0052] The pressure fluctuation state sequence represents the pure evolution process of the pressure change trend over a period of time. Each element in the pressure fluctuation state sequence represents a specific pressure fluctuation state, including continuous rise, continuous fall, rise followed by fall, fall followed by rise, remain unchanged, remain unchanged followed by rise, remain unchanged followed by fall, rise followed by remain unchanged, and fall followed by remain unchanged. This provides a standardized data foundation for the subsequent analysis of trend-related characteristic indicators.

[0053] Because the operating modes of the heaters in a vacuum heat treatment furnace differ fundamentally across different process stages—heating, holding, and cooling—the heating power is continuously output during the heating stage, resulting in a consistently positive temperature increment. During the holding stage, the heating power is intermittently compensated, and the temperature increment oscillates slightly around zero. During the cooling stage, heating ceases and forced cooling occurs, resulting in a consistently negative temperature increment. The absolute values ​​of the temperature increments vary significantly across these stages. Directly performing trend correlation analysis based on temperature increment data from multiple sampling points would lead to distorted correlation calculations due to the magnitude differences between stages. Therefore, this embodiment transforms continuously changing temperature increment values ​​into discrete temperature fluctuation state sequences based on the data fluctuation characteristics of temperature increments within a preset historical window. This eliminates the influence of differences in the magnitude of temperature changes across different process stages, providing normalized and standardized input data for subsequent morphological matching and causal consistency determination with the gas pressure fluctuation state sequences.

[0054] The temperature fluctuation state sequence represents the pure evolution process of the dynamic pushing intensity applied by the heater to the workpiece over a period of time. Each element in the temperature fluctuation state sequence represents a specific temperature fluctuation state, including continuous rise, continuous fall, rise then fall, fall then rise, remain unchanged, remain unchanged then rise, remain unchanged then fall, rise then remain unchanged, and fall then remain unchanged. This provides a standardized data foundation for the subsequent analysis of trend correlation characteristic indicators.

[0055] By calculating the morphological matching degree between the gas pressure fluctuation state sequence and the temperature fluctuation state sequence, the causal mapping relationship between the amount of gas released by the workpiece under thermal desorption and adsorption and the actual temperature rise can be measured, and the trend correlation characteristic index at the current sampling time can be obtained.

[0056] Specifically, the method for analyzing the pressure fluctuation state sequence based on the data fluctuation characteristics of the total pressure increment at the current sampling time and at historical sampling times within a preset historical window is as follows: Construct a preset historical window of length W, taking the current sampling time as the last sampling time within the preset historical window, and extract the total voltage increment between the current sampling time and the historical sampling times within the preset historical window. Set a subsequence window of length m (e.g., m=3), starting from the beginning position of the preset historical window, use the subsequence window to traverse in steps of one sampling time, and sequentially extract m consecutive total voltage increments.

[0057] Based on the changing trend of the total pressure increment in the subsequence window, the air pressure fluctuation state of the current subsequence window is determined. For example, if m is set to 3, and the total pressure increments captured by the subsequence window within the preset historical window are (0.5, 0.5, 0.2) in sequence, then the changing trend of the total pressure increment in the subsequence window is first unchanged and then decreasing. The first unchanged and then decreasing trend is an element in the air pressure fluctuation state sequence, representing the air pressure fluctuation state within the subsequence window.

[0058] The total pressure increments captured by the subsequence window within the preset historical window are (0.5, 0.2, 0.8) in sequence. The trend of the total pressure increment within the subsequence window is first decreasing and then increasing. The first decrease and then increase is an element in the pressure fluctuation state sequence, representing the pressure fluctuation state within the subsequence window.

[0059] After traversing the entire preset history window, the following is obtained: The sequence of atmospheric pressure fluctuation states of each element, that is, the existence of Subsequence windows.

[0060] Using the same method, the temperature fluctuation state sequence is analyzed based on the data fluctuation characteristics of the temperature increment between the current sampling time and the historical sampling time within the preset historical window.

[0061] The elements in the pressure fluctuation state sequence and the temperature fluctuation state sequence are strictly equal in number and correspond one-to-one in time.

[0062] In one specific implementation of this invention, the length of the preset history window is set to 300 sampling times.

[0063] Furthermore, based on the pressure fluctuation state sequence and temperature fluctuation state sequence, trend correlation characteristic indicators are determined, including: Any pressure fluctuation state in the pressure fluctuation state sequence is taken as the first pressure state, and the previous pressure fluctuation state adjacent to the first pressure state is taken as the second pressure state; the second pressure state and the first pressure state constitute a pressure state combination; based on the probability of the pressure state combination appearing in the pressure fluctuation state sequence and the probability of the second pressure state appearing in the pressure fluctuation state sequence, the conditional information entropy is calculated. The temperature fluctuation state corresponding to the first air pressure state under a preset delay is taken as the first temperature state; the second air pressure state, the first air pressure state, and the first temperature state constitute a ternary combination, and the second air pressure state and the first temperature state constitute a binary combination; based on the probability of the ternary combination appearing in the air pressure fluctuation state sequence and the temperature fluctuation state sequence, and the probability of the binary combination appearing in the air pressure fluctuation state sequence and the temperature fluctuation state sequence, the joint conditional information entropy is calculated. The difference between the conditional information entropy and the joint conditional information entropy is used as a trend correlation feature indicator.

[0064] Conditional information entropy characterizes the randomness or uncertainty of subsequent pressure fluctuation states given a known trend of pressure fluctuation state changes. Joint conditional information entropy characterizes the randomness or uncertainty of subsequent pressure fluctuation states given both a known trend of pressure fluctuation state changes and a known trend of temperature fluctuation state changes with physical delays.

[0065] The trend correlation characteristic index represents the magnitude of the randomness or uncertainty of subsequent air pressure fluctuations eliminated after introducing the trend of temperature fluctuation changes. It quantifies the degree of temporal correlation between air pressure fluctuation characteristics and temperature fluctuation characteristics. The larger the trend correlation characteristic index, the more fully the increase in air pressure inside the vacuum heat treatment furnace is controlled by the heater's heating operation, which is a normal phenomenon of workpiece desorption and gas release due to heat. The smaller the trend correlation characteristic index, the more the air pressure fluctuation inside the vacuum heat treatment furnace is out of the control of the heat source, indicating that external atmospheric pressure gas independent of temperature changes is being drawn into the furnace, indicating a real external physical gas leakage fault.

[0066] Specifically, the process of calculating trend correlation characteristic indicators is as follows: 1. As shown in the above process, both the air pressure fluctuation state sequence and the temperature fluctuation state sequence contain W-m+1 fluctuation states. A unified internal sliding traversal index t is set. In the air pressure fluctuation state sequence, the starting position of t is set to the (d+2)th position, and the ending position is set to the (W-m+1)th position. The total number of valid samples participating in the calculation within the traversal interval [d+2, W-m+1] is N=Wmd. Since the effect of temperature rise on air pressure only becomes apparent after d sampling intervals, the air pressure fluctuation state at position t in the air pressure fluctuation state sequence corresponds to the tdth temperature fluctuation state in the temperature fluctuation state sequence. The traversal interval of the temperature fluctuation state sequence is [2, W-m+1-d]. The total number of valid samples participating in the calculation is the same as the total number of valid samples participating in the calculation of the air pressure fluctuation state sequence. The preset delay is d.

[0067] 2. Calculate the entropy of its own conditional information. .

[0068] (1) Taking any pressure fluctuation state in the pressure fluctuation state sequence as an example, let's take the t-th pressure fluctuation state as an example. As the first pressure state, the (t-1)th pressure fluctuation state is... As the second pressure state. The first and second pressure states constitute a pressure state combination ( , The elements in the pressure state combination are arranged in chronological order. For example... It represents an initial rise followed by a fall. If it represents a continuous rise, then ( , (First rise, then fall, then continuous rise).

[0069] statistics( , The total number of times a y-state appears in the traversed interval, divided by the total number of all pressure state combinations in the traversed interval, yields ( ). , The probability of occurrence of ) ;statistics The total number of occurrences within the traversal interval, divided by the total number of all pressure fluctuation states within the traversal interval, yields the result. probability of occurrence .

[0070] (2) Calculate the conditional information entropy using the following formula: in, The conditional information entropy is represented by N; the total number of valid samples is represented by t; the internal traversal index is represented by d; the preset delay is represented by W; the preset length of the history window is represented by m; and the preset length of the subsequence window is represented by m. Indicates the second atmospheric pressure state; Indicates the first atmospheric pressure state; Indicates the combination of air pressure states ( , The probability of occurrence of ) This indicates the probability of the second pressure state occurring.

[0071] 3. Calculate the joint conditional information entropy .

[0072] (1) Similarly, taking any pressure fluctuation state in the pressure fluctuation state sequence as an example, let's take the t-th pressure fluctuation state as an example. As the first pressure state, the (t-1)th pressure fluctuation state is... As the second pressure state, the temperature fluctuation state corresponding to the first pressure state after a preset delay d is defined as the td-th temperature fluctuation state in the temperature fluctuation state sequence. The first temperature state is considered the first atmospheric pressure state. The second atmospheric pressure state, the first atmospheric pressure state, and the first temperature state constitute a ternary combination. The second pressure state and the first temperature state constitute a binary combination. The elements in the ternary and binary combinations are arranged in order of pressure fluctuation state first, followed by temperature fluctuation state, with the pressure fluctuation states in the ternary combination arranged in chronological order.

[0073] statistics( The total number of occurrences of a ternary combination within the traversed interval, divided by the total number of all triplet combinations within the traversed interval, yields ( The probability of occurrence of ) ;statistics The total number of occurrences within the traversed interval, divided by the total number of all binary combinations within the traversed interval, yields the result. probability of occurrence .

[0074] (2) Calculate the joint conditional information entropy using the following formula: in, represents the joint conditional information entropy; N represents the total number of valid samples; t represents the internal traversal index; d represents the preset delay; W represents the preset length of the history window; m represents the preset length of the subsequence window. Indicates the second atmospheric pressure state; Indicates the first atmospheric pressure state; Indicates the first temperature state; Represents a ternary combination ( The probability of ) appearing; Represents a binary combination The probability of its occurrence.

[0075] 4. Subtract the joint conditional information entropy from the obtained conditional information entropy to obtain the trend correlation feature index at the current sampling time.

[0076] In one specific implementation of this invention, the method for obtaining the preset delay is as follows: Before the equipment is officially put into production, a temperature rise calibration test is performed on the vacuum heat treatment furnace under a leak-free condition. The starting time T1 is recorded when the furnace temperature data first shows a continuous upward trend after the heating program starts. The total pressure data inside the furnace is continuously monitored, and the response time T2, which corresponds to the first inflection point in the total pressure data caused by the occurrence of heated gas release, is recorded. The time difference between the response time T2 and the starting time T1 is calculated, and this time difference is divided by the fixed sampling interval in step S1. The result is then rounded to obtain the preset delay. In this embodiment, the preset delay is 5 sampling intervals.

[0077] Preferably, in some implementations of the present invention, before analyzing the trend correlation characteristic index of the current sampling time based on the data fluctuation characteristics of the total pressure increment between the current sampling time and the historical sampling time, and the data fluctuation characteristics of the temperature increment between the current sampling time and the historical sampling time, the method further includes: If the number of current sampling times and historical sampling times is less than or equal to the length of the preset historical window, it indicates that the data is in the data accumulation period and the amount of data obtained is insufficient to support subsequent calculations. In this case, the preset historical window is suspended, and only the data collection operation is performed. If the number of current sampling times and historical sampling times is greater than the length of the preset historical window, it indicates that enough data has been collected. In this case, the preset historical window is closed, and the trend correlation feature index analysis is performed within the preset historical window.

[0078] Preferably, in some implementations of the present invention, determining whether the vacuum heat treatment furnace is leaking based on the leakage judgment index includes: If the leakage judgment index is less than or equal to the preset upper limit threshold, it is determined that the vacuum heat treatment furnace has not leaked; if the leakage judgment index is greater than the preset upper limit threshold, it is determined that the vacuum heat treatment furnace has leaked.

[0079] The preset upper limit threshold represents the maximum leakage judgment index that the workpiece in the vacuum heat treatment furnace can produce under the condition that the workpiece is normally subjected to heat desorption and gas release under fully compliant conditions.

[0080] A leakage judgment index less than or equal to the preset upper threshold indicates that the pressure rise at the current sampling time is entirely caused by the workpiece releasing gas due to heating, and the furnace is in normal working condition. In this case, the vacuum heat treatment furnace is not leaking. A leakage judgment index greater than the preset upper threshold indicates that the pressure rise at the current sampling time has deviated from normal working condition, and there is external atmospheric pressure air leaking into the furnace. In this case, the vacuum heat treatment furnace is leaking.

[0081] In this embodiment of the invention, considering the significant differences in surface area, material, and aging degree of insulation materials among different batches of workpieces, the normal heat release volume of the vacuum heat treatment furnace varies considerably each time, making it impossible to set a universally applicable preset upper limit threshold. Therefore, this embodiment uses the initial heating phase of the vacuum heat treatment furnace (when workpiece heat release is most intense) as the baseline capture period. During the baseline capture period, only the leakage judgment index is collected at each sampling moment, without determining whether the vacuum heat treatment furnace is leaking. This embodiment sets the length of the baseline capture period to 300 sampling moments. The maximum value of all leakage judgment indices within the baseline capture period is used as the preset upper limit threshold.

[0082] After obtaining the preset upper limit threshold, the system enters the online monitoring period, during which the vacuum heat treatment furnace is checked for leaks at each sampling moment.

[0083] The preset upper limit threshold obtained in this embodiment is 1.5 Pa.

[0084] As mentioned above, the preset historical window length is 300 sampling times. That is, the baseline capture period begins from the 301st sampling time after the vacuum heat treatment furnace starts, and the leakage judgment index is obtained. The online monitoring period begins from the 601st sampling time.

[0085] In this embodiment of the invention, if a leak is detected in the vacuum heat treatment furnace, an abnormal event identifier is generated. The number of the ultrasonic sensor that generated the sound pressure data at the current sampling time is extracted, and the three-dimensional physical location of the ultrasonic sensor in the vacuum heat treatment furnace is queried in the static mapping table described in step S1. Subsequently, the abnormal event identifier and the three-dimensional physical location are sent to the on-site maintenance personnel, and a leak alarm is issued, prompting the on-site maintenance personnel to go to the leak point for emergency repair.

[0086] During on-site repairs at the leak point, this invention continuously monitors the leakage judgment index at each sampling time. When the latest leakage judgment index shows a decrease, and the leakage judgment index at three consecutive sampling times is less than or equal to a preset upper limit threshold, the leak point is determined to have been repaired, and the vacuum heat treatment furnace has returned to a compliant pure heat release phenomenon. At this point, the abnormal event flag is removed, and a confirmation record of successful seal repair is generated in the system log of this invention. Otherwise, the abnormal event output and alarm status are maintained, prompting on-site operators to continue repairs at the leak point until the leakage judgment index decreases and meets the above verification conditions.

[0087] The present invention also proposes a gas leakage detection system for a vacuum heat treatment furnace, the system comprising: The data acquisition module is used to acquire the total pressure, temperature and sound pressure data inside the furnace at the current sampling time; The feature analysis module is used to obtain the total pressure increment based on the difference between the total pressure data in the furnace at the current sampling time and the historical adjacent sampling time; to obtain the temperature increment based on the difference between the temperature data at the current sampling time and the historical adjacent sampling time; and to analyze the trend correlation feature indicators at the current sampling time based on the data fluctuation characteristics of the total pressure increment at the current sampling time and the data fluctuation characteristics of the temperature increment at the current sampling time and the historical sampling time. The results analysis module is used to determine the leakage judgment index at the current sampling time based on the total pressure increment, sound pressure data, and trend correlation characteristic indicators; and to determine whether the vacuum heat treatment furnace is leaking based on the leakage judgment index.

Claims

1. A method for detecting gas leakage in a vacuum heat treatment furnace, characterized in that, The method includes: Acquire the total pressure, temperature, and sound pressure data inside the furnace at the current sampling moment; The total pressure increment is obtained based on the difference between the total pressure data in the furnace at the current sampling time and the historical adjacent sampling time; the temperature increment is obtained based on the difference between the temperature data at the current sampling time and the historical adjacent sampling time; and the trend correlation characteristic indicators at the current sampling time are analyzed based on the data fluctuation characteristics of the total pressure increment at the current sampling time and the data fluctuation characteristics of the temperature increment at the current sampling time and the historical sampling time. Based on the total pressure increment, sound pressure data, and trend correlation characteristic indicators, a leakage judgment index is determined at the current sampling time; based on the leakage judgment index, it is determined whether the vacuum heat treatment furnace is leaking.

2. The method for detecting gas leakage in a vacuum heat treatment furnace according to claim 1, characterized in that, The step of obtaining the total pressure increment based on the difference between the total pressure data inside the furnace at the current sampling time and the historical adjacent sampling times includes: The difference between the total pressure data inside the furnace at the current sampling time and the total pressure data inside the furnace at the historical adjacent sampling time is used as the total pressure increment.

3. The method for detecting gas leakage in a vacuum heat treatment furnace according to claim 1, characterized in that, The step of obtaining the temperature increment based on the difference between the temperature data at the current sampling time and the temperature data at historical adjacent sampling times includes: The temperature increment is the difference between the temperature data at the current sampling time and the temperature data at historical adjacent sampling times.

4. The method for detecting gas leakage in a vacuum heat treatment furnace according to claim 1, characterized in that, The analysis of trend correlation indicators for the current sampling time, based on the data fluctuation characteristics of the total pressure increment between the current sampling time and the historical sampling time, and the data fluctuation characteristics of the temperature increment between the current sampling time and the historical sampling time, includes: Based on the data fluctuation characteristics of the total pressure increment between the current sampling time and the historical sampling time within a preset historical window, the air pressure fluctuation state sequence is analyzed; the elements in the air pressure fluctuation state sequence are air pressure fluctuation states. Based on the data fluctuation characteristics of the temperature increment between the current sampling time and the historical sampling time within a preset historical window, the temperature fluctuation state sequence is analyzed; the elements in the temperature fluctuation state sequence are the temperature fluctuation states. Based on the pressure fluctuation state sequence and temperature fluctuation state sequence, trend correlation characteristic indicators are determined.

5. The method for detecting gas leakage in a vacuum heat treatment furnace according to claim 4, characterized in that, The determination of trend correlation characteristic indicators based on the pressure fluctuation state sequence and temperature fluctuation state sequence includes: Any pressure fluctuation state in the pressure fluctuation state sequence is taken as the first pressure state, and the previous pressure fluctuation state adjacent to the first pressure state is taken as the second pressure state; the second pressure state and the first pressure state constitute a pressure state combination; based on the probability of the pressure state combination appearing in the pressure fluctuation state sequence and the probability of the second pressure state appearing in the pressure fluctuation state sequence, the conditional information entropy is calculated. The temperature fluctuation state corresponding to the first air pressure state under a preset delay is taken as the first temperature state; the second air pressure state, the first air pressure state, and the first temperature state constitute a ternary combination, and the second air pressure state and the first temperature state constitute a binary combination; based on the probability of the ternary combination appearing in the air pressure fluctuation state sequence and the temperature fluctuation state sequence, and the probability of the binary combination appearing in the air pressure fluctuation state sequence and the temperature fluctuation state sequence, the joint conditional information entropy is calculated. The difference between the conditional information entropy and the joint conditional information entropy is used as a trend correlation feature indicator.

6. The method for detecting gas leakage in a vacuum heat treatment furnace according to claim 4, characterized in that, Before analyzing the trend correlation indicators of the current sampling time based on the data fluctuation characteristics of the total pressure increment between the current sampling time and the historical sampling time, and the data fluctuation characteristics of the temperature increment between the current sampling time and the historical sampling time, the method further includes: If the number of current sampling times and historical sampling times is less than or equal to the length of the preset historical window, the preset historical window is suspended and trend correlation characteristic index analysis is not performed; if the number of current sampling times and historical sampling times is greater than the length of the preset historical window, trend correlation characteristic index analysis is performed.

7. The method for detecting gas leakage in a vacuum heat treatment furnace according to claim 1, characterized in that, The step of determining the leakage judgment index at the current sampling moment based on the total pressure increment, sound pressure data, and trend correlation characteristic indicators includes: The total pressure increment, sound pressure data, and trend correlation feature index are processed by feature fusion multiplication to obtain the leakage judgment index at the current sampling time.

8. The method for detecting gas leakage in a vacuum heat treatment furnace according to claim 1, characterized in that, Before determining the leakage judgment index at the current sampling moment based on the total pressure increment, sound pressure data, and trend correlation characteristic indicators, the method further includes: If the total pressure increment is less than or equal to 0, the leakage judgment index is set to 0; if the total pressure increment is greater than 0, the leakage judgment index is calculated.

9. The method for detecting gas leakage in a vacuum heat treatment furnace according to claim 1, characterized in that, The step of determining whether the vacuum heat treatment furnace is leaking based on the leakage judgment index includes: If the leakage judgment index is less than or equal to the preset upper limit threshold, it is determined that the vacuum heat treatment furnace has not leaked; if the leakage judgment index is greater than the preset upper limit threshold, it is determined that the vacuum heat treatment furnace has leaked.

10. A gas leak detection system for a vacuum heat treatment furnace, characterized in that, The system includes: The data acquisition module is used to acquire the total pressure, temperature and sound pressure data inside the furnace at the current sampling time; The feature analysis module is used to obtain the total pressure increment based on the difference between the total pressure data in the furnace at the current sampling time and the historical adjacent sampling time; to obtain the temperature increment based on the difference between the temperature data at the current sampling time and the historical adjacent sampling time; and to analyze the trend correlation feature indicators at the current sampling time based on the data fluctuation characteristics of the total pressure increment at the current sampling time and the data fluctuation characteristics of the temperature increment at the current sampling time and the historical sampling time. The results analysis module is used to determine the leakage judgment index at the current sampling time based on the total pressure increment, sound pressure data, and trend correlation characteristic indicators; and to determine whether the vacuum heat treatment furnace is leaking based on the leakage judgment index.

Images