Integrated ald valve leak test system and method
By dynamically calculating the judgment threshold and using three-dimensional clustering technology, combined with temperature compensation and false positive identification, the consistency and accuracy issues of ALD valve leakage testing were resolved, achieving high-precision leakage detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-26
- Publication Date
- 2026-06-19
AI Technical Summary
In existing technologies, it is difficult to guarantee the consistency of leakage test results for ALD valves. Static thresholds cannot be dynamically adapted to the processing tolerances and aging degrees of different batches of valve bodies. Helium molecules adsorbed on the surface form false positive signals, affecting the accuracy and reliability of the test.
The integrated ALD valve leakage test method uses dynamic calculation of the judgment threshold, combined with three-dimensional density clustering and nitrogen purging to distinguish between real leaks and false positives. It uses a pre-stored ALD valve three-dimensional CAD semantic feature library to accurately determine the leakage type, and combines temperature drift compensation and dynamic background model to improve detection accuracy.
This technology enables high-sealing performance testing of ALD valves, reduces the false alarm rate, improves leak location accuracy and signal recognition accuracy, and ensures the reliability and consistency of test results.
Smart Images
Figure CN122237848A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of valve leakage testing technology, specifically to an integrated ALD valve leakage testing system and method. Background Technology
[0002] In semiconductor front-end processes, atomic layer deposition (ALD) valves are core components of high-purity, high-vacuum fluid control systems, and their sealing integrity directly determines the cleanliness of the process chamber and the quality of thin film deposition. Currently, the industry commonly uses helium mass spectrometry for leak testing of ALD valves. A typical procedure includes: injecting tracer helium gas externally into the valve body and monitoring the response signal using an internal helium detector; or filling the valve with helium gas and using an external probe to scan the surface to locate the leak point.
[0003] However, existing technologies have multiple bottlenecks, including: static thresholds cannot dynamically adapt to the processing tolerances and aging levels of different batches of valve bodies, making it difficult to guarantee the consistency of test results; in addition, helium molecules adsorbed on the surface can form false positive signals, and existing purging processes mostly use fixed duration and flow rates, which cannot be differentiated according to the location, surface area and material characteristics of the leak point, resulting in interference with real leak signals or the inability to effectively eliminate false signals, seriously affecting the accuracy and reliability of the test. Summary of the Invention
[0004] To address the shortcomings of existing technologies, this invention provides an integrated ALD valve leakage testing system and method, which solves the problem that static thresholds cannot dynamically adapt to the processing tolerances and aging degrees of different batches of valve bodies, making it difficult to guarantee the consistency of test results.
[0005] To achieve the above objectives, the present invention provides the following technical solution: an integrated ALD valve leakage testing method, which specifically includes the following steps:
[0006] Step 1: Close the interface of the valve under test, evacuate the vacuum, and run the helium detector to measure the real-time background leakage rate. If the leakage rate does not exceed the preset safety threshold, trigger subsequent tests, and simultaneously collect the peak response and spatial coordinates of the helium detector to generate the original leakage hotspot map.
[0007] Step 2: Based on the real-time background leakage rate, temperature compensation item and product specifications, dynamically calculate the judgment threshold, screen valid points and then perform three-dimensional density clustering to output several leakage candidate clusters;
[0008] Step 3: Calculate the nitrogen purging time dynamically based on the projected area and surface material of each leak candidate cluster, perform vertical purging to remove adsorbed helium from the surface, and distinguish between real leaks and false positives by comparing the signal attenuation ratio and background level before and after purging.
[0009] Step 4: Use platform + trend dual criteria to verify suspected leaks with high confidence. After deducting dynamic background drift, adaptive duration sampling is performed on the compensation leak rate signal. Only when the signal simultaneously meets the intensity, stability and non-attenuation trend at the end of the observation period is it confirmed as a real leak.
[0010] Step 5: Identify the specific type of the actual leakage situation, match the center coordinates of the leakage cluster with the pre-stored ALD valve 3D CAD semantic feature library, automatically identify the functional area to which it belongs, and accurately determine the leakage type by calculating the shortest distance from the point to the surface / curve and setting the matching threshold.
[0011] As a further aspect of the present invention, the method for dynamically calculating the determination threshold is as follows:
[0012] Obtain the background leakage rate Q0, acquire the temperature-background drift compensation coefficient based on the system's built-in temperature drift calibration table, and obtain the corresponding temperature compensation item based on the current real-time temperature. Then according to the formula The threshold Q is calculated. th Q 下限 This indicates the lower limit of the product specification, and k1 represents the background multiple coefficient. Q represents the temperature drift compensation term. 裕度 Indicates the safety margin.
[0013] As a further aspect of the present invention, the method for outputting several leakage candidate clusters is as follows:
[0014] Based on threshold Q th Filter all valid points with a background leakage rate greater than a set threshold, extract the spatial coordinates of the valid points and their corresponding leakage rate values, then set the neighborhood radius and minimum number of points for spatial clustering. For each valid point Pi, calculate its three-dimensional Euclidean distance from other valid points. If there are at least two other points satisfying d ij ≤d y Then, mark the valid point Pi as the core point. Next, starting from any unassigned core point, add that point to the new cluster and search for all points whose distance to it is ≤ d. y Other valid points are selected, and the screening process is repeated until no new neighboring valid points can be found. All included points are combined into a single leakage candidate cluster, and neighboring points are merged into several leakage candidate clusters.
[0015] As a further aspect of the present invention, the nitrogen purging time is calculated as follows:
[0016] Obtain a candidate cluster for leakage to be verified, and extract the spatial bounding box (X) of the cluster. min X max Y min Ymax Z min Z max ), calculate its projected area A on the valve body surface. 投影 The material database of the ALD valve under test was consulted to obtain the helium adsorption correction factor k. m Set the basic purging flow rate and follow the formula. Calculate the adaptive purging time t blow Where t0 is the base delay, This is an empirical coefficient.
[0017] As a further aspect of the present invention, the specific method for distinguishing between real leaks and false positives is as follows:
[0018] After purging, turn off the nitrogen gas and start the helium detector. Collect the background signal of the area for 5 seconds and calculate the average value Q. p Read the original leakage rate Q before purging. 原始 And according to the formula Calculate the signal attenuation ratio R, if the average value Q p If the signal attenuation ratio is less than 1.5Q0, it is considered a false positive, and the area is marked as non-leaking, skipping further verification. Otherwise, the signal attenuation ratio R is compared. If R < 30%, the purging is considered invalid, and true leak verification begins. If R > 30%, but Q < 1.5Q0, the purging is considered invalid. p If the value is greater than 1.5Q0, then a second purging will be performed.
[0019] As a further aspect of the present invention, the specific method for true / false leakage verification is as follows:
[0020] Record the current system baseline Q z Simultaneously activate the background drift monitoring channel and establish a dynamic background model. ,in Let T(t) be the temperature drift coefficient, and T(t) be the real-time temperature. Then, the helium detector is started to continuously collect the leakage signal Q_leak(t), and the leakage signal is calculated according to the formula... Real-time calculation of leakage rate Q after compensation 补偿 (t), and simultaneously set the minimum observation window t. min , and t min The value is set by the operator; if within the smallest observation window t min Inner Q 补偿 (t) always < 0.5Q th If no leakage is detected, the collection process will terminate early, indicating no leakage; otherwise, data collection will continue until the maximum observation window t is reached. max .
[0021] As a further aspect of the present invention, the method of using a platform + trend dual criterion to perform high-confidence verification of suspected leaks is as follows:
[0022] Based on the platform's judgment criteria, calculate the final t.final The average value Q k and standard deviation Q must be satisfied k Q th ,and ≤12%, based on the trend judgment criteria, for the last t y The second-by-second data is linearly fitted to obtain the slope s, and an allowable decay threshold s is set. min And the slope s must be greater than the allowable attenuation threshold s. min Anomaly detection and analysis are performed based on both platform and trend analysis:
[0023] When both the platform and trend criteria are met, a genuine leak is confirmed. When the platform criteria are met but the trend criteria are not, a suspected interference is identified. When the platform criteria are not met but the slope s corresponding to the trend criteria is greater than 0, it is marked as a potential leak.
[0024] As a further aspect of the present invention, the method for accurately determining the leakage type is as follows:
[0025] Acquire a pre-stored 3D CAD model of the ALD valve under test, extract key functional area features, assign semantic labels to form a structural feature library, obtain a leakage candidate cluster, and calculate its weighted center coordinate Pc(x) with leakage rate as the weight. c y c , z c ), and the weighted center coordinates Pc(x) c y c , z c Transform to the valve body assembly coordinate system, traverse all features in the structural feature library, and calculate the shortest Euclidean distance d from Pc to each feature. k At the same time, set the maximum matching distance threshold d. max Choose the option that satisfies d. k ≤d max , and d k The smallest feature is used as the matching result, and the leakage type is determined based on the obtained matching result.
[0026] The integrated ALD valve leakage testing system includes:
[0027] The system self-test and background calibration module is used to verify the system's sealing performance and instrument status before each test. By closing the interface of the valve under test, drawing a vacuum, and running a helium detector to measure the real-time background leakage rate, if the rate exceeds the preset safety threshold, the test will be automatically terminated and an alarm will be triggered. At the same time, the real-time background leakage rate will be transmitted to the hotspot generation and clustering module.
[0028] The hotspot generation and clustering module is used to control the annular helium nozzle array to automatically scan the valve body according to a preset trajectory, synchronously collect the peak response and spatial coordinates of the helium detector, generate the original leakage hotspot map, dynamically calculate the judgment threshold based on the real-time background leakage rate, temperature compensation item and product specifications, perform three-dimensional density clustering after screening effective points, output several leakage candidate clusters, and transmit the leakage candidate clusters to the purging and false positive identification module.
[0029] The purging and false positive identification module is used to dynamically calculate the nitrogen purging time based on the projected area and surface material of each leak candidate cluster, perform vertical purging to remove surface adsorbed helium, and intelligently distinguish between real leaks and false positives by comparing the signal attenuation ratio and background level before and after purging. For real leaks, the signal is transmitted to the real leak dynamic verification module.
[0030] The real leak dynamic verification module is used to verify suspected leaks with high confidence using platform + trend dual criteria. After deducting dynamic background drift, the compensation leak rate signal is sampled for adaptive duration. Only when the signal meets the intensity, stability and non-decay trend at the end of the observation period is it confirmed as a real leak. At the same time, the real leak situation is transmitted to the leak type identification module.
[0031] The leak type identification module is used to identify the specific type of actual leaks. It spatially matches the center coordinates of the leak cluster with the pre-stored ALD valve 3D CAD semantic feature library, automatically identifies the functional area to which it belongs, and accurately determines the leak type by calculating the shortest distance from the point to the surface / curve and setting the matching threshold.
[0032] This invention provides an integrated ALD valve leakage testing system and method. Compared with the prior art, it has the following advantages:
[0033] This invention introduces a built-in temperature drift calibration table to obtain the background drift compensation coefficient corresponding to the real-time temperature, and calculates a dynamic threshold based on the background leakage rate, thereby achieving temperature-adaptive compensation for the background leakage rate. At the same time, a dynamic background model is established in the true leak verification stage to calculate the compensated leakage rate in real time, effectively offsetting the interference of ambient temperature fluctuations on the detection results, ensuring the accuracy of leakage judgment, and meeting the core requirements of high sealing performance testing of ALD valves.
[0034] This invention uses a PLC to control an array of annular helium nozzles to automatically spray helium along a preset trajectory, simultaneously collecting helium detector response data and nozzle spatial coordinates to generate a leak hotspot map. Then, it uses dynamic thresholding to filter valid leak points and combines this with three-dimensional Euclidean distance for spatial clustering to form leak candidate clusters, accurately extracting features such as the center coordinates and maximum leakage rate of the core leak area. Simultaneously, through a purging-delay-retest process, combined with parameters such as helium adsorption correction coefficient and signal attenuation ratio, it effectively distinguishes between false positive leaks and real leaks, reducing the misjudgment rate and improving leak location accuracy and signal recognition accuracy. Attached Figure Description
[0035] Figure 1 This is a flowchart of the integrated ALD valve leakage test method of the present invention;
[0036] Figure 2 This is a block diagram of the integrated ALD valve leakage testing system of the present invention. Detailed Implementation
[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0038] First Embodiment
[0039] Please see Figure 1 This application provides an integrated ALD valve leakage test method, which specifically includes the following steps:
[0040] Step 1: First, close all the interfaces of the valves under test, start the vacuum pump unit, and evacuate it to a vacuum level of at least 5 × 10⁻⁶. - 6 mbar, then run the helium detector and record the background leakage rate Q0. At the same time, compare the background leakage rate Q0 with the preset leakage rate. If the background leakage rate is greater than the preset leakage rate, the test will be terminated and an alarm will be triggered. Otherwise, if the background leakage rate is less than the preset leakage rate, subsequent testing and processing will be performed.
[0041] Next, an array of annular helium nozzles is arranged around the valve body, and the nozzles are automatically injected according to a preset trajectory under PLC control. Helium injection is performed at a preset injection pressure and duration. Specifically, the injection pressure is controlled at 20±2 kPa and the duration is 2.0±0.2 s. The peak response of the helium detector is recorded simultaneously and correlated with the spatial coordinates. The response data of the helium detector and the spatial coordinate information of the helium nozzles are collected synchronously, and the helium concentration values at different locations are compared to identify abnormally high reading areas and generate a leak hotspot map.
[0042] Step 2: Based on the generated original leakage hotspot map, dynamically set the threshold Q. th The specific dynamic setting method is as follows:
[0043] The background leakage rate Q0 is obtained, and the temperature-background drift compensation coefficient is obtained based on the system's built-in temperature drift calibration table. For example, when the temperature is 25℃, the average background increment is 0, and when the temperature is 80℃, the average background increment is 3.0 × 10⁻⁶. -11 (mbar·L / s), at a temperature of 120℃, the average background increment is 8.0×10 -11 (mbar·L / s), at a temperature of 150℃, the average background increase is 1.5×10 -11 (mbar·L / s), the corresponding temperature compensation term is obtained based on the current real-time temperature. Then according to the formula The threshold Q is calculated. th Q 下限 This indicates the lower limit of the product specification, and k1 represents the background multiple coefficient. Q represents the temperature drift compensation term. 裕度 Indicates safety margin;
[0044] Based on threshold Q th Filter all valid points with a background leakage rate greater than a set threshold, and then perform spatial clustering on the obtained valid points. The specific spatial clustering processing method is as follows:
[0045] Extract the spatial coordinates of the valid points and their corresponding leakage rate values. Then, set the neighborhood radius and minimum number of points for spatial clustering. For each valid point Pi, calculate its three-dimensional Euclidean distance from other valid points, specifically according to the formula. Calculate the 3D Euclidean distance if there exist at least two other points that satisfy d ij ≤d y Then, mark the valid point Pi as the core point. Next, starting from any unassigned core point, add that point to the new cluster and search for all points whose distance to it is ≤ d. y Other valid points are identified, and the screening process is repeated until no new neighboring valid points can be found. All included points are then combined into a single leakage candidate cluster. Neighboring points are merged into several leakage candidate clusters, and the center coordinates and maximum leakage rate Q are output. max And the number of points N;
[0046] Step 3: Next, perform a purge-delay-retest process for each candidate region. From the spatial clustering step, obtain a leak candidate cluster to be verified, and extract the spatial bounding box (X) of the cluster. min X max Y min Ymax Z min Z max ), calculate its projected area A on the valve body surface. 投影 Query the material database of the ALD valve under test to obtain the surface material type corresponding to that area. Based on the material type, look up the table to obtain the helium adsorption correction coefficient k. m The details are shown in the table below:
[0047] Material type Helium adsorption correction factor Metal 1.0 polymer 1.8-2.5 Ceramic coating 1.3
[0048] Next, set the basic purging flow rate and follow the formula. Calculate the adaptive purging time t blow Where t0 is the base delay, As an empirical coefficient, the purging process is based on the baseline purging flow rate and adaptive purging time, with the purging direction perpendicular to the valve body surface. After purging, the nitrogen gas is shut off, and after waiting 2 seconds to allow the surface airflow to stabilize, the helium detector is started to collect the background signal in this area for 5 seconds, and the average value Q is calculated. p Read the original leakage rate Q before purging. 原始 And according to the formula Calculate the signal attenuation ratio R, if the average value Q p If the signal attenuation ratio is less than 1.5Q0, it is considered a false positive, and the area is marked as non-leaking, skipping further verification. Otherwise, the signal attenuation ratio R is compared. If R < 30%, the purging is considered invalid, and true leak verification begins. If R > 30%, but Q < 1.5Q0, the purging is considered invalid. p If the value is greater than 1.5Q0, then a second purging process will be performed;
[0049] For true leak verification, record the current system baseline Q before starting the true leak verification. z Simultaneously activate the background drift monitoring channel and establish a dynamic background model. ,in Let T(t) be the temperature drift coefficient, and T(t) be the real-time temperature. Then, the helium detector is started to continuously collect the leakage signal Q_leak(t), and the leakage signal is calculated according to the formula... Real-time calculation of leakage rate Q after compensation 补偿 (t), and simultaneously set the minimum observation window t. min , and t min The value is set by the operator; if within the smallest observation window t min Inner Q 补偿 (t) always < 0.5Q th If no leakage is detected, the collection process will terminate early, indicating no leakage; otherwise, data collection will continue until the maximum observation window t is reached. max ;
[0050] For the collected Q 补偿(t) sequence, based on both platform and trend judgment criteria, calculate the final t based on the platform judgment criteria. final The average value Q k and standard deviation Q must be satisfied k Q th ,and ≤12%, based on the trend judgment criteria, for the last t y The second-by-second data is linearly fitted to obtain the slope s, and an allowable decay threshold s is set. min And the slope s must be greater than the allowable attenuation threshold s. min Anomaly detection and analysis are performed based on both platform and trend analysis:
[0051] When both the platform and trend criteria are met, a genuine leak is confirmed. When the platform criteria are met but the trend criteria are not, a suspected interference is identified. When the platform criteria are not met but the slope s corresponding to the trend criteria is greater than 0, it is marked as a potential leak.
[0052] Step 4: Based on the confirmed actual leakage, firstly, obtain the pre-stored 3D CAD model of the ALD valve under test, extract key functional area features, and assign semantic labels to form a structural feature library. Each feature includes geometric type, spatial parameters, and recommended verification mode. Then, obtain a leakage candidate cluster and calculate its weighted center coordinate Pc(x) with the leakage rate as the weight. c y c , z c ), and the weighted center coordinates Pc(x) c y c , z c Transform to the valve body assembly coordinate system, consistent with the CAD model, traverse all features in the structural feature library, and calculate the shortest Euclidean distance d from Pc to each feature. k For planes or curved surfaces, the distance from a point to the surface is calculated; for curves, the minimum distance from a point to the curve in space is calculated, while a maximum matching distance threshold d is set. max Choose the option that satisfies d. k ≤d max , and d k The smallest feature is used as the matching result, and the leakage type is determined based on the obtained matching result.
[0053] Second Embodiment
[0054] Please see Figure 2 This application provides an integrated ALD valve leakage testing system, including: a system self-test and background calibration module, a hotspot generation and clustering module, a purging and false positive identification module, a real leakage dynamic verification module, and a leakage type identification module, combined with... Figure 2It can be seen that the information between the above functional modules is transmitted in one direction only;
[0055] The system self-test and background calibration module is used to verify the system's sealing and instrument status before each test. By closing the interface of the valve under test, drawing a vacuum, and running a helium detector to measure the real-time background leakage rate, if it exceeds the preset safety threshold, the test will be automatically terminated and an alarm will be triggered. At the same time, the real-time background leakage rate will be transmitted to the hotspot generation and clustering module, and the specific processing method is the same as the processing process in step one.
[0056] The hotspot generation and clustering module is used to control the annular helium nozzle array to automatically scan the valve body according to a preset trajectory, synchronously collect the peak response and spatial coordinates of the helium detector, generate the original leakage hotspot map, dynamically calculate the judgment threshold based on the real-time background leakage rate, temperature compensation item and product specifications, perform three-dimensional density clustering after screening effective points, output several leakage candidate clusters, and transmit the leakage candidate clusters to the purging and false positive identification module. The specific processing method is the same as the processing process in step two.
[0057] The purging and false positive identification module is used to dynamically calculate the nitrogen purging time based on the projected area and surface material of each leak candidate cluster, perform vertical purging to remove surface adsorbed helium, and intelligently distinguish between real leaks and false positives by comparing the signal attenuation ratio and background level before and after purging. For real leaks, the signal is transmitted to the real leak dynamic verification module, and the specific processing method is the same as the processing process in step three.
[0058] The real leak dynamic verification module uses a platform + trend dual criterion to verify suspected leaks with high confidence. After deducting the dynamic background drift, the compensation leak rate signal is sampled for an adaptive duration. Only when the signal meets the intensity, stability and non-attenuation trend at the end of the observation period is it confirmed as a real leak. At the same time, the real leak situation is transmitted to the leak type identification module, and the specific processing method is the same as the processing process in step three.
[0059] The leak type identification module is used to identify the specific type of actual leaks. It spatially matches the center coordinates of the leak cluster with the pre-stored ALD valve 3D CAD semantic feature library, automatically identifies the functional area to which it belongs, and accurately determines the leak type by calculating the shortest distance from the point to the surface / curve and setting the matching threshold.
[0060] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.
[0061] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.
Claims
1. An integrated ALD valve leakage test method, characterized in that, The method specifically includes the following steps: Step 1: Close the interface of the valve under test, evacuate the vacuum, and run the helium detector to measure the real-time background leakage rate. If the leakage rate does not exceed the preset safety threshold, trigger subsequent tests, and simultaneously collect the peak response and spatial coordinates of the helium detector to generate the original leakage hotspot map. Step 2: Based on the real-time background leakage rate, temperature compensation item and product specifications, dynamically calculate the judgment threshold, screen valid points and then perform three-dimensional density clustering to output several leakage candidate clusters; Step 3: Calculate the nitrogen purging time dynamically based on the projected area and surface material of each leak candidate cluster, perform vertical purging to remove adsorbed helium from the surface, and distinguish between real leaks and false positives by comparing the signal attenuation ratio and background level before and after purging. Step 4: Use platform + trend dual criteria to verify suspected leaks with high confidence. After deducting dynamic background drift, adaptive duration sampling is performed on the compensation leak rate signal. Only when the signal simultaneously meets the intensity, stability and non-attenuation trend at the end of the observation period is it confirmed as a real leak. Step 5: Identify the specific type of the actual leakage situation, match the center coordinates of the leakage cluster with the pre-stored ALD valve 3D CAD semantic feature library, automatically identify the functional area to which it belongs, and accurately determine the leakage type by calculating the shortest distance from the point to the surface / curve and setting the matching threshold.
2. The integrated ALD valve leakage test method according to claim 1, characterized in that, The method for dynamically calculating and determining the threshold is as follows: Obtain the background leakage rate Q0, acquire the temperature-background drift compensation coefficient based on the system's built-in temperature drift calibration table, and obtain the corresponding temperature compensation item based on the current real-time temperature. Then according to the formula The threshold Q is calculated. th Q 下限 This indicates the lower limit of the product specification, and k1 represents the background multiple coefficient. Q represents the temperature drift compensation term. 裕度 Indicates the safety margin.
3. The integrated ALD valve leakage test method according to claim 1, characterized in that, The method for outputting several leak candidate clusters is as follows: Based on threshold Q th Filter all valid points with a background leakage rate greater than a set threshold, extract the spatial coordinates of the valid points and their corresponding leakage rate values, then set the neighborhood radius and minimum number of points for spatial clustering. For each valid point Pi, calculate its three-dimensional Euclidean distance from other valid points. If there are at least two other points satisfying d ij ≤d y Then, mark the valid point Pi as the core point. Next, starting from any unassigned core point, add that point to the new cluster and search for all points whose distance to it is ≤ d. y Other valid points are selected, and the screening process is repeated until no new neighboring valid points can be found. All included points are combined into a single leakage candidate cluster, and neighboring points are merged into several leakage candidate clusters.
4. The integrated ALD valve leakage test method according to claim 1, characterized in that, The nitrogen purging time is calculated as follows: Obtain a candidate cluster for leakage to be verified, and extract the spatial bounding box (X) of the cluster. min X max Y min Y max Z min Z max ), calculate its projected area A on the valve body surface. 投影 The material database of the ALD valve under test was consulted to obtain the helium adsorption correction factor k. m Set the basic purging flow rate and follow the formula. Calculate the adaptive purging time t blow Where t0 is the base delay, This is an empirical coefficient.
5. The integrated ALD valve leakage test method according to claim 1, characterized in that, The specific method for distinguishing between real leaks and false positives is as follows: After purging, turn off the nitrogen gas and start the helium detector. Collect the background signal of the area for 5 seconds and calculate the average value Q. p Read the original leakage rate Q before purging. 原始 And according to the formula Calculate the signal attenuation ratio R, if the average value Q p If the signal attenuation ratio is less than 1.5Q0, it is considered a false positive, and the area is marked as non-leaking, skipping further verification. Otherwise, the signal attenuation ratio R is compared. If R < 30%, the purging is considered invalid, and true leak verification begins. If R > 30%, but Q < 1.5Q0, the purging is considered invalid. p If the value is greater than 1.5Q0, then a second purging will be performed.
6. The integrated ALD valve leakage test method according to claim 5, characterized in that, The specific method for true / false leak detection is as follows: Record the current system baseline Q z Simultaneously activate the background drift monitoring channel and establish a dynamic background model. ,in Let T(t) be the temperature drift coefficient, and T(t) be the real-time temperature. Then, the helium detector is started to continuously collect the leakage signal Q_leak(t), and the leakage signal is calculated according to the formula... Real-time calculation of leakage rate Q after compensation 补偿 (t), and simultaneously set the minimum observation window t. min , and t min The value is set by the operator; if within the smallest observation window t min Inner Q 补偿 (t) always < 0.5Q th If no leakage is detected, the collection process will terminate early, indicating no leakage; otherwise, data collection will continue until the maximum observation window t is reached. max .
7. The integrated ALD valve leakage test method according to claim 1, characterized in that, The method of using a platform + trend dual criterion for high-confidence verification of suspected leaks is as follows: Based on the platform's judgment criteria, calculate the final t. final The average value Q k and standard deviation Q must be satisfied k Q th ,and ≤12%, based on the trend judgment criteria, for the last t y The second-by-second data is linearly fitted to obtain the slope s, and an allowable decay threshold s is set. min And the slope s must be greater than the allowable attenuation threshold s. min Anomaly detection and analysis are performed based on both platform and trend analysis: When both the platform and trend criteria are met, a genuine leak is confirmed. When the platform criteria are met but the trend criteria are not, a suspected interference is identified. When the platform criteria are not met but the slope s corresponding to the trend criteria is greater than 0, it is marked as a potential leak.
8. The integrated ALD valve leakage test method according to claim 1, characterized in that, The method for accurately determining the type of leakage is as follows: Acquire a pre-stored 3D CAD model of the ALD valve under test, extract key functional area features, assign semantic labels to form a structural feature library, obtain a leakage candidate cluster, and calculate its weighted center coordinate Pc(x) with leakage rate as the weight. c y c , z c ), and the weighted center coordinates Pc(x) c y c , z c Transform to the valve body assembly coordinate system, traverse all features in the structural feature library, and calculate the shortest Euclidean distance d from Pc to each feature. k At the same time, set the maximum matching distance threshold d. max Choose the option that satisfies d. k ≤d max , and d k The smallest feature is used as the matching result, and the leakage type is determined based on the obtained matching result.
9. An integrated ALD valve leakage testing system, used to perform the integrated ALD valve leakage testing method according to any one of claims 1-8, characterized in that, include: The system self-test and background calibration module is used to verify the system's sealing performance and instrument status before each test. By closing the interface of the valve under test, drawing a vacuum, and running a helium detector to measure the real-time background leakage rate, if the rate exceeds the preset safety threshold, the test will be automatically terminated and an alarm will be triggered. At the same time, the real-time background leakage rate will be transmitted to the hotspot generation and clustering module. The hotspot generation and clustering module is used to control the annular helium nozzle array to automatically scan the valve body according to a preset trajectory, synchronously collect the peak response and spatial coordinates of the helium detector, generate the original leakage hotspot map, dynamically calculate the judgment threshold based on the real-time background leakage rate, temperature compensation item and product specifications, perform three-dimensional density clustering after screening effective points, output several leakage candidate clusters, and transmit the leakage candidate clusters to the purging and false positive identification module. The purging and false positive identification module is used to dynamically calculate the nitrogen purging time based on the projected area and surface material of each leak candidate cluster, perform vertical purging to remove surface adsorbed helium, and intelligently distinguish between real leaks and false positives by comparing the signal attenuation ratio and background level before and after purging. For real leaks, the signal is transmitted to the real leak dynamic verification module. The real leak dynamic verification module is used to verify suspected leaks with high confidence using platform + trend dual criteria. After deducting dynamic background drift, the compensation leak rate signal is sampled for adaptive duration. Only when the signal meets the intensity, stability and non-decay trend at the end of the observation period is it confirmed as a real leak. At the same time, the real leak situation is transmitted to the leak type identification module. The leak type identification module is used to identify the specific type of actual leaks. It spatially matches the center coordinates of the leak cluster with the pre-stored ALD valve 3D CAD semantic feature library, automatically identifies the functional area to which it belongs, and accurately determines the leak type by calculating the shortest distance from the point to the surface / curve and setting the matching threshold.