Sealing detection method for ultrahigh vacuum exhaust device

Abstract

The invention provides a sealing detection method for an ultrahigh vacuum exhaust device, and relates to the technical field of sealing detection, and the method comprises the steps: carrying out the leakage risk feature mining of the ultrahigh vacuum exhaust device, and constructing the detection point distribution; performing vacuumizing stable detection on the ultrahigh vacuum exhaust device, and generating a sealing detection starting signal according to a stable detection result; performing sealing detection on the ultrahigh vacuum exhaust device according to the distribution of the detection points and the helium mass spectrometer leak detector to obtain a first result of sealing detection of the device; and performing detection background interference correction, helium spraying cross influence compensation and material helium adsorption correction on the first result of the device sealing detection to generate a fourth result of the device sealing detection. According to the invention, the technical problem of insufficient sealing detection precision caused by misjudgment and missed judgment due to existence of various interference factors is solved, detection is carried out after vacuum stabilization, and the detection result is corrected, so that the sealing detection effect of the ultrahigh vacuum exhaust device is improved.

Description

Technical Field

[0001] This application relates to the field of sealing detection technology, and in particular to a sealing detection method for ultra-high vacuum exhaust devices. Background Technology

[0002] Sealing testing of ultra-high vacuum exhaust systems is a crucial technology for ensuring the integrity and performance of these systems. Because ultra-high vacuum systems require extremely low leakage rates, any infiltration of external gases can affect the system's efficiency and stability. To test the sealing performance of the equipment, a helium mass spectrometer leak detector is typically used. Helium gas is injected into the device, and the leak detector measures the amount of leaked helium to determine the sealing performance. However, when using a helium mass spectrometer for sealing testing, background gases in the external environment may interfere with the helium, affecting the accuracy of the leak detector. Furthermore, interactions between gases or the vacuum environment can cause interference with the helium detection signal, leading to false alarms or missed alarms, thus affecting the accuracy of the sealing test results.

[0003] In summary, existing technologies suffer from various interference factors, which can easily lead to misjudgments and omissions, resulting in insufficient accuracy in seal detection. Summary of the Invention

[0004] The purpose of this application is to provide a sealing detection method for ultra-high vacuum exhaust devices, in order to solve the technical problem in the prior art that the presence of various interference factors easily leads to misjudgment and omission, resulting in insufficient sealing detection accuracy.

[0005] In view of the above problems, this application provides a sealing detection method for an ultra-high vacuum exhaust device, wherein the sealing detection method for the ultra-high vacuum exhaust device includes: mining leakage risk characteristics of the ultra-high vacuum exhaust device and constructing a detection point distribution; performing a vacuum stabilization test on the ultra-high vacuum exhaust device and generating a sealing detection start signal based on the stabilization test result; performing a sealing detection on the ultra-high vacuum exhaust device based on the sealing detection start signal, according to the detection point distribution and a helium mass spectrometer leak detector, to obtain a first sealing detection result; correcting the detection background interference of the first sealing detection result to obtain a second sealing detection result; compensating for helium cross-influence of the second sealing detection result to obtain a third sealing detection result; and correcting the material helium adsorption of the third sealing detection result to generate a fourth sealing detection result.

[0006] Optionally, a device structure feature dataset of the ultra-high vacuum exhaust device is obtained; based on the device structure feature dataset, a multi-point leakage risk prediction is performed on the ultra-high vacuum exhaust device to obtain a first feature point distribution with a leakage risk greater than or equal to a predetermined risk; based on the ultra-high vacuum exhaust device, frequent leakage feature mining is performed to generate a second feature point distribution; the union of the first feature point distribution and the second feature point distribution is calculated to generate the detection point distribution.

[0007] Optionally, a global search for leakage events is performed on the ultra-high vacuum exhaust device to obtain a leakage event retrieval set; the leakage frequency of multiple points of the ultra-high vacuum exhaust device is evaluated based on the leakage event retrieval set to obtain a leakage frequency coefficient for each point; based on the leakage frequency coefficient of each point, the multiple points are selected according to a leakage frequency threshold to obtain the second feature point distribution.

[0008] Optionally, the ultra-high vacuum exhaust device is evacuated according to a reference vacuum level, and the real-time pressure data set of the ultra-high vacuum exhaust device is collected simultaneously; the stability is evaluated based on the real-time pressure data set to obtain the stability detection result; when the real-time vacuum level of the ultra-high vacuum exhaust device meets the reference vacuum level and the stability detection result meets the predetermined stability, the sealing detection start signal is generated.

[0009] Optionally, when the real-time vacuum degree does not meet the reference vacuum degree and / or the stable detection result does not meet the predetermined stability, a sealing detection waiting signal is generated.

[0010] Optionally, based on the sealing detection start signal, helium gas is injected according to the distribution of detection points, and the helium mass spectrometer leak detector is started simultaneously to collect helium concentration parameters to obtain helium detection data at each point; the helium detection data at each point is input into M sealing performance evaluation models to obtain sealing performance evaluation sequences at each point, where M is a positive integer greater than 1; lumped value calculation is performed on the sealing performance evaluation sequences at each point to obtain the sealing performance coefficient at each point, and the sealing performance coefficient at each point is added to the first result of the device sealing detection.

[0011] Optionally, a detection background factor is set, which includes helium characteristics of the detection environment, airflow characteristics of the detection environment, other characteristics of the detection environment, and status characteristics of the detection equipment; detection background parameters are collected at each point according to the detection background factor to obtain detection background data at each point; anomaly identification is performed based on the detection background data at each point to obtain anomaly characteristics of each detection background; sealing detection interference is evaluated based on the anomaly characteristics of each detection background to obtain the background detection interference evaluation results, and the first sealing detection result of the device is corrected based on the background detection interference evaluation results to generate the second sealing detection result of the device.

[0012] Optionally, the pre-helium injection information of each detection point is collected to obtain the pre-helium injection characteristics of each point; the cross-influence analysis of helium injection at each detection point is performed based on the pre-helium injection characteristics of each point to obtain the cross-influence characteristics of helium injection at each point; the helium detection at each point is corrected based on the cross-influence characteristics of helium injection at each point to obtain the corrected detection data of each point; the second result of the device sealing test is corrected based on the corrected detection data of each point to generate the third result of the device sealing test.

[0013] Optionally, a set of pre-injection helium feature samples and a set of helium injection cross-influence feature samples are obtained; a residual neural network is trained under supervision based on the pre-injection helium feature sample set and the helium injection cross-influence feature sample set, and the analytical accuracy of helium injection cross-influence is obtained after each predetermined number of training iterations; if the analytical accuracy of helium injection cross-influence meets the predetermined analytical accuracy, an analytical model of helium injection cross-influence is generated; the pre-injection helium features of each point are input into the analytical model of helium injection cross-influence to obtain the helium injection cross-influence features of each point.

[0014] Optionally, material characteristic data for each location is obtained; helium adsorption is predicted based on the material characteristic data for each location to obtain helium adsorption characteristics for each location; sealing detection impact is identified based on the helium adsorption characteristics for each location to obtain helium adsorption correction factors for each location; the third sealing detection result of the device is corrected based on the helium adsorption correction factors for each location to generate the fourth sealing detection result of the device.

[0015] The technical solution provided in this application has at least the following beneficial effects: By mining the leakage risk characteristics of the ultra-high vacuum exhaust device, a distribution of detection points is constructed. The ultra-high vacuum exhaust device undergoes stable vacuum testing, and a sealing test initiation signal is generated based on the stable testing results. Based on the sealing test initiation signal, and according to the detection point distribution and a helium mass spectrometer leak detector, a sealing test is performed on the ultra-high vacuum exhaust device to obtain a first sealing test result. Background interference correction is applied to the first sealing test result to obtain a second sealing test result. Helium injection cross-influence compensation is applied to the second sealing test result to obtain a third sealing test result. Material helium adsorption correction is applied to the third sealing test result to generate a fourth sealing test result. In other words, by performing stable vacuum testing to ensure the vacuum environment is stable before sealing test, background gas interference during the testing process is corrected to reduce the influence of external gases, and helium injection cross-influence compensation corrects the helium adsorption effect of the material. This approach exhibits high sensitivity, high robustness, and adaptive multi-factor compensation capabilities, thus improving the sealing test effect of the ultra-high vacuum exhaust device.

[0016] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the sealing test method for the ultra-high vacuum exhaust device of this application.

[0019] Figure 2 This is a schematic diagram of the process for obtaining the third result of the sealing test of the ultra-high vacuum exhaust device in the sealing test method of this application. Detailed Implementation

[0020] This application provides a sealing detection method for ultra-high vacuum exhaust devices, solving the technical problem in existing technologies where misjudgments and omissions are prone to occur due to various interfering factors, leading to insufficient sealing detection accuracy. By performing a stable vacuum test, the method ensures the vacuum environment is stable before sealing detection, corrects for background gas interference during the detection process, reduces the influence of external gases, and corrects for the helium adsorption effect of the material through helium injection cross-influence compensation. This method exhibits high sensitivity, high robustness, and adaptive multi-factor compensation capability, thus improving the sealing detection effect of ultra-high vacuum exhaust devices.

[0021] The technical solutions of this application will now be clearly and completely described 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. It should be understood that this application is not limited to the exemplary embodiments described herein. 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. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0022] For examples, please refer to the appendix. Figure 1This application provides a sealing test method for an ultra-high vacuum exhaust device, wherein the sealing test method for the ultra-high vacuum exhaust device specifically includes the following steps: S100: Conduct leakage risk feature mining for ultra-high vacuum exhaust devices and construct the distribution of detection points.

[0023] Furthermore, this application S100 includes: S110: Obtain the device structure feature dataset of the ultra-high vacuum exhaust device; S120: Perform multi-point leakage risk prediction on the ultra-high vacuum exhaust device based on the device structure feature dataset to obtain a first feature point distribution with a leakage risk greater than or equal to a predetermined risk; S130: Perform frequent leakage feature mining on the ultra-high vacuum exhaust device to generate a second feature point distribution; S140: Calculate the union of the first feature point distribution and the second feature point distribution to generate the detection point distribution.

[0024] S131: Perform a global search for leakage events based on the ultra-high vacuum exhaust device to obtain a leakage event retrieval set; S132: Evaluate the leakage frequency of multiple points of the ultra-high vacuum exhaust device based on the leakage event retrieval set to obtain a leakage frequency coefficient for each point; S133: Based on the leakage frequency coefficients of each point, select the multiple points according to the leakage frequency threshold to obtain the second feature point distribution.

[0025] Specifically, by using design drawings and models of the ultra-high vacuum exhaust device, the dimensions, materials, and connection methods of each physical component of the device are obtained, resulting in a structural feature dataset. This dataset helps in analyzing the performance of each part of the device and identifying potential leakage points. Based on this structural feature dataset, multi-point leakage risk prediction is performed on the ultra-high vacuum exhaust device. By analyzing the structural characteristics of each part of the device, such as interfaces, welded areas, and sealing rings, a risk prediction model is used to assess the risk at each point. The leakage probability at each point is predicted based on historical leakage data and material properties.

[0026] Choose a model trained in a domain closely related to ultra-high vacuum exhaust devices, particularly for leak risk prediction or failure prediction of similar equipment, such as leak detection models for other vacuum systems and pressure vessels. Freeze the first few layers (the part used for feature extraction), and then retrain the later layers of the model using target domain data (i.e., the device structural feature dataset of ultra-high vacuum exhaust devices). Fine-tune the model by using the features extracted from the source domain model on the target task, while optimizing specific parts of the target task. First, standardize, clean, and fill the target domain data (i.e., the device structural feature dataset of ultra-high vacuum exhaust devices), and divide it into training and validation sets, typically 80% training set and 20% validation set. According to the requirements of transfer learning, the training set is fed into the transfer learning model, and the validation set is used to evaluate the model trained after transfer learning, including accuracy, recall, precision, and F1 score. Set the convergence criteria for the model, such as the validation set loss changing by less than 0.01 for 5 consecutive rounds or the training set accuracy reaching 95%. Stop training when the model reaches the convergence criteria; the resulting model is the risk prediction model.

[0027] The structural feature dataset of the ultra-high vacuum exhaust device is input into the risk prediction model to predict the leakage probability at each point, i.e., to perform risk analysis on multiple parts of the device. The predetermined leakage risk is a pre-set threshold used to determine whether the leakage risk at a certain point is high enough. The predicted leakage risks of multiple points are compared with the predetermined leakage risk, and those exceeding the predetermined leakage risk are grouped together to form the first feature point distribution.

[0028] A global leak event retrieval is performed on the ultra-high vacuum exhaust system, involving a comprehensive search and analysis of the entire system's leak history to identify all relevant leak events. For example, maintenance logs might show three leak events at connector 10 and two at connector 15 within the past year. The leak event retrieval results are then compiled into a dataset through simple statistical analysis, recording the location (i.e., point) and frequency of each leak event, resulting in a leak event retrieval set. Using the data in the leak event retrieval set, the leak frequency coefficient for each point is evaluated, and the number of leaks at each point is calculated to obtain the leak frequency coefficient for each point.

[0029] A leakage frequency threshold is set to filter out locations with a high leakage frequency, distinguishing which locations are frequently leaking and which can be ignored. By filtering locations with a leakage frequency greater than or equal to the leakage frequency threshold, the final second characteristic point distribution is obtained, which represents the areas requiring focused attention. The union of the first characteristic point distribution (locations with high leakage risk) and the second characteristic point distribution (locations with frequent leaks) is calculated to obtain the detection point distribution, including all areas with high leakage risk or a high number of leaks, serving as key locations for sealing inspection. By accurately predicting and frequently identifying high-risk leakage points, it is possible to ensure that inspection work is focused on key locations, avoiding the omission of important leakage points, accurately locating locations with high leakage risk and frequent leaks, significantly improving detection efficiency and accuracy, and reducing the risk of misjudgment and missed detection.

[0030] S200: Perform a vacuum stabilization test on the ultra-high vacuum exhaust device, and generate a sealing test start signal based on the stabilization test result.

[0031] Furthermore, this application S200 includes: S210: Perform vacuuming on the ultra-high vacuum exhaust device according to the reference vacuum level, and simultaneously collect the real-time pressure data set of the ultra-high vacuum exhaust device; S220: Evaluate the stability based on the real-time pressure data set to obtain the stability detection result; S230: When the real-time vacuum level of the ultra-high vacuum exhaust device meets the reference vacuum level, and the stability detection result meets the predetermined stability level, generate the sealing detection start signal; S240: When the real-time vacuum level does not meet the reference vacuum level and / or the stability detection result does not meet the predetermined stability level, generate the sealing detection waiting signal.

[0032] Specifically, the reference vacuum level refers to the standard vacuum level that an ultra-high vacuum exhaust system should achieve under normal operating conditions. It is the minimum expected vacuum level set during the design of the ultra-high vacuum exhaust system, typically determined by the equipment manufacturer or engineer based on the equipment's performance specifications. The ultra-high vacuum exhaust system performs a vacuuming process to expel internal gases until the set reference vacuum level is reached. During the vacuuming process, vacuum pumps and vacuum sensors continuously monitor pressure changes in the ultra-high vacuum exhaust system, collecting real-time pressure datasets. These datasets record the pressure value at every moment during the vacuuming process. For example, pressure data is recorded every second during the vacuuming process, continuously until the reference vacuum level is reached.

[0033] Analyzing real-time pressure datasets and calculating the standard deviation of pressure fluctuations assesses the degree of pressure data volatility, thus obtaining stable test results. A smaller standard deviation indicates less pressure fluctuation and a more stable ultra-high vacuum exhaust device. Stability evaluation involves analyzing real-time pressure datasets to assess the stability of pressure fluctuations. Evaluation criteria typically include the amplitude, frequency, and trend of pressure fluctuations, aiming to determine whether the ultra-high vacuum exhaust device has reached a stable state suitable for subsequent sealing tests. For example, suppose the real-time pressure dataset contains the following pressure value: P1 = 10. -5 Pa, P2=10 -5.1 Pa, P3 = 10 -5.2 Pa. Its standard deviation is calculated to be 1.51 × 10⁻⁶. -6 Pa indicates a small change in pressure.

[0034] The real-time vacuum level must be less than or equal to the reference vacuum level. If the real-time vacuum level does not reach the reference vacuum level, it indicates that the equipment is still in the process of evacuation and a seal test cannot be performed yet. Simultaneously, it is necessary to assess whether the pressure fluctuation is within the predetermined range. If not, it indicates that the ultra-high vacuum exhaust device is unstable and a seal test cannot be performed.

[0035] When the real-time vacuum level of the ultra-high vacuum exhaust device meets the reference vacuum level, and the stability test result meets the predetermined stability level, it is considered to have entered a stable state and is suitable for sealing testing, generating a sealing testing start signal. The predetermined stability level is a predefined stability level used to determine whether the ultra-high vacuum exhaust device has entered a stable state. If the real-time vacuum level does not reach the reference vacuum level, or the stability evaluation does not meet the predetermined standard, a waiting signal is generated, indicating that further waiting is required for the equipment to reach an appropriate vacuum state before performing sealing testing. By ensuring that the real-time vacuum level and pressure stability meet the requirements, sealing testing errors caused by pressure instability or failure to meet the reference vacuum level are avoided, ensuring accurate test results.

[0036] S300: Based on the sealing detection start signal, the ultra-high vacuum exhaust device is sealed according to the detection point distribution and the helium mass spectrometer leak detector to obtain the first result of the device sealing detection.

[0037] Furthermore, this application S300 includes: S310: Based on the sealing detection start signal, helium gas is injected according to the distribution of detection points, and the helium mass spectrometer leak detector is started simultaneously to collect helium concentration parameters to obtain helium detection data at each point; S320: The helium detection data at each point is input into M sealing performance evaluation models to obtain sealing performance evaluation sequences at each point, where M is a positive integer greater than 1; S330: The lumped value is calculated for each sealing performance evaluation sequence at each point to obtain the sealing performance coefficient at each point, and the sealing performance coefficient at each point is added to the first result of the device sealing detection.

[0038] Specifically, the sealing test begins upon receiving the sealing test activation signal. Based on the distribution of detection points, helium gas is injected into designated points on the ultra-high vacuum exhaust system to simulate potential leak paths and detect any leaks at these points. Helium injection refers to introducing helium gas into the ultra-high vacuum exhaust system and supplying it to various points through an injection device. Helium, due to its small molecular weight and low reactivity with other gases, is commonly used as a tracer gas in leak detection. A helium mass spectrometer leak detector is an instrument used to detect helium concentration, accurately measuring the helium concentration inside the device. The helium mass spectrometer leak detector operates synchronously at each detection point, detecting the presence of leaks and recording the leakage amount, including the helium concentration value at each point. For example, the helium concentration detected at point 1 is 0.08 ppm, at point 2 it is 0.15 ppm, at point 3 it is 0.06 ppm, at point 4 it is 0.07 ppm, and at point 5 it is 0.22 ppm.

[0039] The helium concentration data at each location is input into M sealing performance evaluation models. Each model, based on its specific algorithm and training data, outputs a sealing performance evaluation coefficient for each location, resulting in a sealing performance evaluation sequence for each location. Each sequence includes M sealing performance evaluation coefficients. The M sealing performance evaluation models refer to multiple independent models inputting the helium concentration data for each location, with each model evaluating the sealing performance of that location using different algorithms and feature analysis methods. For example, Model 1 is trained using a support vector machine, Model 2 using a decision tree algorithm, and Model 3 using a neural network. The model training process is similar, collecting historical helium concentration data and corresponding sealing status labels (leakage or sealing), covering various operating conditions (such as different pressures, different temperatures, etc.) and including helium leakage situations at different locations. Missing values, outliers, and duplicate data are removed to ensure data quality. Helium concentrations at different locations are normalized to ensure the data is trained on the same scale. Multiple evaluation models, such as support vector machines, decision trees, neural networks, and random forests, are selected to construct corresponding architectures. For each model, historical datasets are used for training. For example, given helium concentration data for a given location and corresponding leak status labels, each model is trained to predict the probability of a leak based on helium concentration. Cross-validation is typically used to prevent overfitting. The model's generalization ability is evaluated by dividing the dataset into several parts and training and validating the models alternately. Standard evaluation metrics (such as accuracy, precision, recall, and F1 score) are used to evaluate the performance of each model. Based on the evaluation results, the model's hyperparameters are optimized using grid search or random search, such as the kernel parameters of an SVM, the depth of a decision tree, and the learning rate of a neural network. Convergence conditions are set for the models, such as a validation set loss change of less than 0.01 for five consecutive rounds or a training set accuracy reaching 95%. When the model reaches the convergence condition, training stops, resulting in M ​​sealing evaluation models.

[0040] For example, the helium concentration data at points 1 to 5 are input into M sealing performance evaluation models. Assuming there are three models, the resulting sealing performance evaluation coefficients are shown in Table 1. Table 1 Sealing performance evaluation coefficients Testing points Helium concentration data (ppm) Sealing performance evaluation coefficient of Model 1 Sealing performance evaluation coefficient of Model 2 Sealing performance evaluation coefficient of Model 3 Leakage status (1 for yes, 0 for no) 1 0.08 0.89 0.90 0.88 0 2 0.15 0.31 0.32 0.30 1 3 0.06 0.91 0.93 0.92 0 4 0.07 0.88 0.92 0.91 0 5 0.22 0.23 0.25 0.24 1 For each location's sealing performance evaluation sequence, a lumped value method is performed. Specifically, for each location, the average of the M sealing performance evaluation coefficients for that location is calculated to obtain the sealing performance coefficient for that location. The lumped value method processes multiple evaluation results to obtain a representative value. Typically, the average value method is used for this calculation, but the median, weighted average, or other methods can also be used, depending on the model's accuracy.

[0041] The sealing coefficient of each point is added to the first result of the device sealing test, forming the first result of the device sealing test. By using multi-model evaluation and lumped value calculation, the accuracy of the sealing test of the ultra-high vacuum exhaust device is further improved. By combining the data from helium injection and helium mass spectrometry leak detector with the output of multi-model, not only is the diversity and comprehensiveness of the test enhanced, but the final evaluation result is further optimized through lumped value calculation, thereby improving the reliability and accuracy of the test.

[0042] S400: Correct the background interference of the first result of the device sealing test to obtain the second result of the device sealing test.

[0043] Furthermore, this application S400 includes: S410: Set detection background factors, including detection environment helium characteristics, detection environment airflow characteristics, other detection environment characteristics, and detection equipment status characteristics; S420: Collect detection background parameters at each point according to the detection background factors to obtain detection background data at each point; S430: Perform anomaly identification based on the detection background data at each point to obtain anomaly characteristics of each detection background; S440: Perform sealing detection interference evaluation based on the anomaly characteristics of each detection background to obtain the background detection interference evaluation results, and correct the first sealing detection result of the device based on the background detection interference evaluation results to generate the second sealing detection result of the device.

[0044] Specifically, setting background factors—that is, external environmental and equipment factors that affect the sealing test results—can interfere with the detection performance of the mass spectrometer leak detector. These factors include the helium characteristics of the detection environment, the airflow characteristics of the detection environment, other characteristics of the detection environment, and the status characteristics of the detection equipment. The helium characteristics of the detection environment include helium concentration, gas distribution, and airflow conditions. The airflow characteristics of the detection environment are the airflow patterns, such as airflow speed, direction, and temperature, which affect helium diffusion and detection results. Other environmental characteristics include temperature, humidity, pressure, and other environmental features that affect gas distribution and detection results. The status characteristics of the detection equipment include the operating status of the helium mass spectrometer leak detector, such as equipment calibration, sensor accuracy, and equipment uptime.

[0045] Based on the set background factors, background parameters are collected for each detection point, including helium concentration, airflow velocity, ambient temperature and humidity, and equipment status. Real-time detection data is simultaneously recorded using a helium mass spectrometer leak detector and stored along with the background data. Anomaly identification is performed based on the abnormal characteristics of each background detection point, identifying potentially abnormal portions that may interfere with the sealing test and require correction. In other words, statistical analysis is conducted on the collected background data from each point to identify data points significantly different from normal conditions, thus determining the abnormal characteristics of each background detection point. For example, a sudden and significant fluctuation in helium concentration or a drastic change in airflow direction can be considered an anomaly.

[0046] An evaluation of sealing detection interference is conducted based on the identified background anomaly features. The impact of this interference on the sealing detection results is assessed, and corrections are made accordingly. A regression model is built using historical data to predict the impact of background features on the sealing detection results. Historical background anomaly features and their corresponding sealing detection results are collected, and the collected dataset is divided into a training set and a validation set. The training set is used for model training, and the validation set is used to evaluate the model's performance. During training, the model adjusts weights and parameters by minimizing the error (e.g., using least squares) to fit the optimal relationship between background features and detection results. The trained regression model is used to predict each background anomaly feature, evaluating the interference of background anomalies on the sealing detection results. The model is used to predict the sealing detection results at each location under different background conditions, and the difference between the actual detection results and the predicted results is calculated. This difference represents the deviation caused by background interference.

[0047] Based on the interference assessment results of the sealing test, the first result of the device sealing test is corrected to eliminate the influence of background anomalies on the test results, resulting in a more accurate second result of the device sealing test. According to the background interference assessment results for each test point, a correction factor is applied to adjust the first result of the sealing test, ensuring that the sealing performance assessment at each test point is closer to the actual situation. The correction method can be adding or subtracting correction amounts, or more complex weighted averaging or dynamic adjustment, usually depending on the accuracy requirements of the sealing test. For example, for a test point where the helium concentration is abnormally high, the interference assessment yields a correction factor of 0.8, indicating that the sealing performance score for that point needs to be reduced (or its test result adjusted), resulting in a more accurate final sealing performance assessment value.

[0048] By setting background factors and identifying anomalies, interference caused by the external environment and equipment status is eliminated, thereby improving the accuracy and stability of the sealing performance test of ultra-high vacuum exhaust devices and overcoming the influence of background interference on the test results.

[0049] S500: Perform helium injection cross-influence compensation on the second result of the device sealing test to obtain the third result of the device sealing test.

[0050] Further details are attached. Figure 2 As shown, S500 of this application includes: S510: Collect the pre-helium injection information of each detection point to obtain the pre-helium injection characteristics of each point; S520: Analyze the cross-influence of helium injection at each detection point based on the pre-helium injection characteristics of each point to obtain the cross-influence characteristics of helium injection at each point; S530: Correct the helium detection at each point based on the cross-influence characteristics of helium injection at each point to obtain the corrected detection data at each point; S540: Correct the second result of the device sealing test based on the corrected detection data at each point to generate the third result of the device sealing test.

[0051] Specifically, during the sealing test, pre-emptive helium injection information is collected from each point of the ultra-high vacuum exhaust device. This includes, but is not limited to, the start and end times of helium injection at adjacent test points, the helium injection flow rate, the helium injection duration, the relative distance between the nozzle and the current point, and the airflow direction and velocity parameters. In other words, during the test, helium is injected at each point, and the start and end times of the injection, the helium injection flow rate (in L / min), the helium injection duration, the spatial distance between the nozzle and the current point, and the airflow direction and velocity at the nozzle are recorded. The helium injection start and end times refer to the start and end times of helium injection collected at each detection point. This is typically achieved using a timing device or sensor to synchronously record the start and end times of helium injection at each point. The helium injection flow rate is recorded in real-time by a flow meter, usually in helium flow rate units (L / min). The helium injection duration is calculated based on the start and end times. The relative distance between the nozzle and the current point is obtained by measuring the distance between the nozzle and the detection point using a spatial coordinate system (such as a three-dimensional coordinate system). Airflow direction and velocity parameters are measured using airflow velocity sensors or CFD (Computational Fluid Dynamics) simulations during the helium injection process. The collected data is processed to ensure the completeness of the preceding helium injection information for each detection point, including the start and end times, flow rate, duration, relative position, airflow direction, and velocity.

[0052] The pre-application helium injection characteristics of each detection point are stored to obtain the pre-application helium injection characteristics of each point. For example, the helium injection start time of point A is 10:00, the helium injection end time is 10:05, the helium injection flow rate is 3L / min, the helium injection duration is 5 minutes, the spatial distance between the nozzle and the current point is 2m, the airflow direction is from A to B, and the flow velocity parameter is 0.5m / s; the helium injection start time of point B is 10:03, the helium injection end time is 10:08, the helium injection flow rate is 4L / min, the helium injection duration is 5 minutes, the spatial distance between the nozzle and the current point is 5m, the airflow direction is from A to B, and the flow velocity parameter is 0.7m / s.

[0053] Based on the pre-injection helium characteristics at each detection point, the cross-influence of helium injection at each detection point is analyzed to determine the degree of influence of the helium injection flow on each point, thus obtaining the cross-influence characteristics of helium injection at each point. Based on the pre-injection helium characteristics, the impact of helium injection on adjacent detection points is calculated, considering factors such as helium injection flow rate, airflow velocity, direction, and distance, to estimate the spatial range and intensity of the helium injection's influence. Cross-influence characteristics refer to the impact data generated by these cross-effects at each detection point, typically involving factors such as helium diffusion paths, airflow interference, and concentration changes. By analyzing the helium injection characteristics at each detection point, how helium diffuses from one point to another is predicted, considering factors such as airflow direction, spatial location, and helium flow rate. Based on the analysis results, the cross-influence characteristics of helium injection at each point are generated, reflecting the mutual influence between points caused by helium injection, such as the helium interference concentration, the duration of cross-interference, and airflow changes.

[0054] Based on the cross-influence characteristics of helium injection, helium detection data at each point is corrected to eliminate or reduce cross-interference and improve the accuracy of the detection results. In other words, the original helium detection data at each point is corrected according to the cross-influence characteristics of helium injection. If a point is affected by helium injection from adjacent points, its detection data will be adjusted to eliminate this influence. For example, if point A has a strong cross-influence on point B, the correction module can reduce the helium concentration in the detection results of point B to compensate for the influence of helium injection from point A on point B.

[0055] Based on the corrected helium detection data at each point, the second result of the device sealing test is corrected, eliminating helium concentration fluctuations caused by cross-influence of helium injections, resulting in more accurate detection data and the third result of the device sealing test. By analyzing the cross-influence of pre-injected helium and correcting the detection data, the impact of helium injections at adjacent points on the detection results is reduced or eliminated, ensuring the accuracy of the measurement results. The corrected detection data better reflects the actual leakage situation, improving the accuracy of the sealing test and thus enhancing the reliability and safety of the ultra-high vacuum exhaust device.

[0056] Furthermore, this application also includes the following steps: S521: Obtain the pre-helium injection feature sample set and the helium injection cross-influence feature sample set; S522: Perform supervised training on the residual neural network based on the pre-helium injection feature sample set and the helium injection cross-influence feature sample set, and obtain the helium injection cross-influence analysis accuracy after each predetermined number of training iterations; S523: If the helium injection cross-influence analysis accuracy meets the predetermined analysis accuracy, generate the helium injection cross-influence analysis model; S524: Input the pre-helium injection features of each point into the helium injection cross-influence analysis model to obtain the helium injection cross-influence features of each point.

[0057] Specifically, sufficient sample data is needed to train the residual neural network, ensuring the model can capture the cross-influence characteristics between different points during helium injection. Therefore, a pre-injection helium feature sample set and a helium injection cross-influence feature sample set are obtained. The pre-injection helium feature sample set is collected from actual tests, including helium injection start and end times, flow rate, duration, airflow direction, and distance. The helium injection cross-influence feature sample set analyzes the cross-influence between different detection points through simulation of helium injection flow. These influences are due to the diffusion and flow of helium in space, especially when helium injection at one point intersects with helium injection at another point, the cross-influence alters the distribution and diffusion process of helium in space. The helium injection cross-influence feature sample set includes parameters such as the amplitude, duration, and spatial propagation of the cross-influence.

[0058] Using a pre-helium injection feature sample set and a helium injection cross-influence feature sample set as input data, a residual neural network is trained to predict the corresponding helium injection cross-influence features based on the input feature data. The residual neural network is a deep neural network architecture that utilizes residual connections to avoid the vanishing gradient problem during training, enabling the network to be more efficient in handling complex tasks. Supervised training is a machine learning training method that adjusts the model's parameters by comparing the model's predictions with the ground truth labeled targets, allowing the model to gradually learn the relationship between input and output.

[0059] During each training iteration, the network calculates the loss based on the difference between the input features and the actual helium-spray cross-influence results, and updates the model weights using backpropagation to gradually minimize the model's prediction error. After a predetermined number of training iterations (e.g., a certain number of epochs), the model is evaluated. The evaluation metric can be mean squared error (MSE), which calculates the difference between the model's true and predicted values. Training stops when the model's analytical precision for helium-spray cross-influence meets the required precision, resulting in a final analytical model for helium-spray cross-influence. The analytical precision is determined by setting a target precision (e.g., MSE less than a certain threshold) to determine whether the expected results have been achieved.

[0060] The pre-injection helium features at each detection point are input into a trained helium injection cross-influence analysis model. This model predicts the degree of cross-influence based on the input features, such as helium propagation and concentration distribution between points. Through training with the pre-injection and cross-influence features, the model can identify and analyze the cross-influence of helium injections between points, thereby correcting the second result of the device sealing test and making the sealing test results more accurate.

[0061] S600: Perform material helium adsorption correction on the third result of the device sealing test to generate the fourth result of the device sealing test.

[0062] Furthermore, this application S600 includes: S610: Obtain material characteristic data at each location; S620: Perform helium adsorption prediction based on the material characteristic data at each location to obtain helium adsorption characteristics at each location; S630: Perform sealing detection influence identification based on the helium adsorption characteristics at each location to obtain helium adsorption correction factors at each location; S640: Correct the third result of device sealing detection based on the helium adsorption correction factors at each location to generate the fourth result of device sealing detection.

[0063] Specifically, this involves acquiring material characteristic data at each testing point, specifically information about the materials used at each detection point within the ultra-high vacuum exhaust device, including material type, density, surface structure, porosity, and adsorption capacity. In other words, it involves collecting and recording material characteristic data for each testing point, including chemical composition, surface roughness, specific surface area, porosity, and heat of adsorption. Different materials can have different side effects on helium. For example, if aluminum is used at a certain point, higher surface roughness increases the probability of helium molecules physically adsorbing in microscopic depressions or pores, leading to a higher helium concentration reading during testing (false positive leakage signal).

[0064] Based on the material characteristic data at each site, a physical model (such as the Langmuir adsorption model or the BET model) is used to predict the amount of helium adsorbed at each site. By inputting the material characteristic data, the amount of helium that may be adsorbed on the material surface is calculated. The Langmuir adsorption model is a classic model describing the adsorption behavior of gas molecules on a solid surface. It assumes that each adsorption site on the surface is independent, and each adsorption site can only hold one gas molecule (monolayer adsorption). The key assumption of the Langmuir model is that the adsorption sites are homogeneous, have no interactions, and each adsorption site can only hold one molecule. The BET model is an extension of the Langmuir model used to describe multilayer adsorption phenomena, and is particularly widely used in contrast surface area (BET surface area) measurements. It interprets adsorption isotherms by considering the adsorption of multiple layers of gas molecules on the surface. The BET model is suitable for the analysis of gas molecules in multilayer adsorption states, especially in porous materials, and can more accurately reflect the adsorption behavior of gases such as helium. For example, suppose a porous material with high porosity is used at a certain location. The material may have a high helium adsorption capacity. Therefore, according to the adsorption model, it is predicted that the location may adsorb a lot of helium during detection, leading to errors in the detection data.

[0065] Based on helium adsorption prediction results, the impact of helium adsorption at each site on the detection results of the helium mass spectrometer was analyzed. Helium adsorption affects the accuracy of the seal detection. By comparing experimental data and predicted data, the adsorption effects at which sites significantly affect the detection results were identified. Corrections must be made for these sites. Based on the characteristics of helium adsorption, a correction factor, usually a proportionality coefficient, is derived to adjust the actual detection data. The correction factor is typically calculated based on the relationship between adsorption characteristics (such as adsorption amount, adsorption rate, etc.) and the actual detection error. For each site, a correction factor is calculated and generated based on its specific adsorption characteristics.

[0066] Based on the helium adsorption correction factor at each point, the detection data at each point are corrected. Typically, the third result of the device sealing test is multiplied by the correction factor to obtain the fourth result, thereby improving the accuracy of the test results. Through helium adsorption prediction and correction, the influence of materials on helium adsorption is eliminated, avoiding misjudgments or omissions and improving the accuracy of sealing tests. Personalized corrections are made to the detection results at each point according to the adsorption characteristics of different materials, improving the sealing accuracy of ultra-high vacuum exhaust devices, eliminating errors caused by material adsorption effects, and ensuring more reliable test results.

[0067] In summary, the sealing detection method for the ultra-high vacuum exhaust device provided in this application has the following beneficial effects: By mining the leakage risk characteristics of the ultra-high vacuum exhaust device, a distribution of detection points is constructed. The ultra-high vacuum exhaust device undergoes stable vacuum testing, and a sealing test initiation signal is generated based on the stable testing results. Based on the sealing test initiation signal, and according to the detection point distribution and a helium mass spectrometer leak detector, a sealing test is performed on the ultra-high vacuum exhaust device to obtain a first sealing test result. Background interference correction is applied to the first sealing test result to obtain a second sealing test result. Helium injection cross-influence compensation is applied to the second sealing test result to obtain a third sealing test result. Material helium adsorption correction is applied to the third sealing test result to generate a fourth sealing test result. In other words, by performing stable vacuum testing to ensure the vacuum environment is stable before sealing test, background gas interference during the testing process is corrected to reduce the influence of external gases, and helium injection cross-influence compensation corrects the helium adsorption effect of the material. This approach exhibits high sensitivity, high robustness, and adaptive multi-factor compensation capabilities, thus improving the sealing test effect of the ultra-high vacuum exhaust device.

[0068] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0069] Obviously, those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A sealing test method for an ultra-high vacuum exhaust device, characterized in that, include: Leakage risk characteristics of ultra-high vacuum exhaust devices were identified, and the distribution of detection points was constructed. The ultra-high vacuum exhaust device is subjected to a vacuum stabilization test, and a sealing test start signal is generated based on the stabilization test results. Based on the sealing detection start signal, the ultra-high vacuum exhaust device is sealed according to the detection point distribution and the helium mass spectrometer leak detector to obtain the first result of the device sealing detection. The first result of the device sealing test is corrected for background interference to obtain the second result of the device sealing test. The second result of the device sealing test is compensated for by helium injection cross-influence to obtain the third result of the device sealing test; The third result of the device sealing test is corrected by material helium adsorption to generate the fourth result of the device sealing test.

2. The sealing detection method for the ultra-high vacuum exhaust device as described in claim 1, characterized in that, Based on the sealing detection start signal, the ultra-high vacuum exhaust device is sealed according to the detection point distribution and the helium mass spectrometer leak detector to obtain the first result of the device sealing detection, including: Based on the seal detection start signal, helium gas is injected according to the distribution of the detection points, and the helium mass spectrometer leak detector is started simultaneously to collect helium concentration parameters and obtain helium detection data at each point. Input the helium detection data from each location into M sealing evaluation models to obtain sealing evaluation sequences for each location, where M is a positive integer greater than 1; The lumped value calculation is performed on the sealing performance evaluation sequence of each point to obtain the sealing performance coefficient of each point, and the sealing performance coefficient of each point is added to the first result of the sealing test of the device.

3. The sealing detection method for the ultra-high vacuum exhaust device as described in claim 1, characterized in that, The first result of the device sealing test is corrected for background interference to obtain the second result of the device sealing test, including: The detection background factors include the helium characteristics of the detection environment, the airflow characteristics of the detection environment, other characteristics of the detection environment, and the status characteristics of the detection equipment. Based on the aforementioned background detection factor, background parameters for each location are collected to obtain background data for each location. Anomaly identification is performed based on the background data detected at each location to obtain the anomaly features of each detected background. Based on the abnormal characteristics of each detection background, an interference assessment for sealing detection is performed to obtain the interference assessment results for each background. Based on the interference assessment results for each background, the first result of sealing detection of the device is corrected to generate the second result of sealing detection of the device.

4. The sealing detection method for the ultra-high vacuum exhaust device as described in claim 1, characterized in that, The second result of the device sealing test is compensated for by helium injection cross-influence to obtain the third result of the device sealing test, including: Collect helium injection information at each detection point to obtain the helium injection characteristics at each point; Based on the pre-helium injection characteristics of each point, the cross-influence analysis of helium injection at each detection point is performed to obtain the cross-influence characteristics of helium injection at each point. Based on the cross-influence characteristics of helium injection at each point, the helium detection at each point is corrected to obtain the corrected detection data for each point. The second result of the device sealing test is corrected based on the test data at each point, and the third result of the device sealing test is generated.

5. The sealing detection method for the ultra-high vacuum exhaust device as described in claim 4, characterized in that, Based on the pre-helium injection characteristics of each detection point, the cross-influence of helium injection at each detection point is analyzed to obtain the cross-influence characteristics of helium injection at each point, including: Obtain the pre-helium injection feature sample set and the helium injection cross-influence feature sample set; The residual neural network is trained under supervision based on the pre-helium injection feature sample set and the helium injection cross-influence feature sample set. After each predetermined number of training iterations, the accuracy of the helium injection cross-influence analysis is obtained. If the analytical accuracy of the helium-injection cross-influence meets the predetermined analytical accuracy, generate the analytical model of the helium-injection cross-influence. The pre-helium injection characteristics of each point are input into the helium injection cross-influence analysis model to obtain the helium injection cross-influence characteristics of each point.

6. The sealing detection method for the ultra-high vacuum exhaust device as described in claim 1, characterized in that, The third result of the device sealing test is corrected for material helium adsorption to generate the fourth result of the device sealing test, including: Obtain material characteristic data at each location; Based on the material characteristic data of each site, helium adsorption is predicted to obtain the helium adsorption characteristics of each site. Based on the helium adsorption characteristics of each location, the sealing detection impact is identified, and the helium adsorption correction factor for each location is obtained. The third result of the device sealing test is corrected based on the helium adsorption correction factor at each point, and the fourth result of the device sealing test is generated.

7. The sealing detection method for the ultra-high vacuum exhaust device as described in claim 1, characterized in that, Leakage risk characteristics of ultra-high vacuum exhaust devices were analyzed, and a distribution of detection points was constructed, including: Obtain the device structure feature dataset of the ultra-high vacuum exhaust device; Based on the device structural feature dataset, the ultra-high vacuum exhaust device is used to predict the leakage risk at multiple points, and the distribution of first feature points with a leakage risk greater than or equal to the predetermined leakage risk is obtained. Based on the frequent leakage characteristics of the ultra-high vacuum exhaust device, a second feature point distribution is generated. The union of the first feature point distribution and the second feature point distribution is obtained to generate the detection point distribution.

8. The sealing detection method for the ultra-high vacuum exhaust device as described in claim 7, characterized in that, Based on the frequent leakage characteristics of the ultra-high vacuum exhaust device, a second feature point distribution is generated, including: A global leak event retrieval set is obtained by performing a global search of the ultra-high vacuum exhaust device. Based on the leakage event retrieval set, the leakage frequency of multiple points of the ultra-high vacuum exhaust device is evaluated to obtain the leakage frequency coefficient of each point. Based on the leakage frequency coefficient of each location, the multiple locations are selected according to the leakage frequency threshold to obtain the second characteristic location distribution.

9. The sealing detection method for the ultra-high vacuum exhaust device as described in claim 1, characterized in that, The ultra-high vacuum exhaust device is subjected to a vacuum stabilization test, and a sealing test start signal is generated based on the stabilization test results, including: The ultra-high vacuum exhaust device is evacuated according to the reference vacuum level, and the real-time pressure data of the ultra-high vacuum exhaust device is collected simultaneously. The stability is evaluated based on the real-time pressure dataset to obtain the stability detection result; When the real-time vacuum degree of the ultra-high vacuum exhaust device meets the reference vacuum degree, and the stable detection result meets the predetermined stability degree, the sealing detection start signal is generated.

10. The sealing detection method for the ultra-high vacuum exhaust device as described in claim 9, characterized in that, When the real-time vacuum degree does not meet the reference vacuum degree and / or the stable detection result does not meet the predetermined stable degree, a sealing detection waiting signal is generated.

Images