Weighing-method-based intelligent detection system for static evaporation rate of welded heat-insulation gas cylinder

Abstract

The invention discloses an intelligent detection system for the static evaporation rate of a welded heat insulation gas cylinder based on a weighing method, and relates to the technical field of gas cylinder detection. The core defects that in traditional static evaporation rate detection, manual dependence is high, data are prone to being interfered, and result reliability is low are effectively overcome. Aiming at the problems of low efficiency and operation errors caused by manual copying, synchronous automatic acquisition of quality, environment temperature and atmospheric pressure data is realized, static evaporation rate calculation and inspection report generation are automatically completed based on detected gas cylinder parameters and screened high-quality data, human intervention is remarkably reduced, and the detection efficiency is improved. And the detection efficiency and the flow traceability are improved.

Description

Technical Field

[0001] This invention relates to the field of gas cylinder testing technology, and in particular to an intelligent detection system for the static evaporation rate of welded insulated gas cylinders based on the weighing method. Background Technology

[0002] The static evaporation rate of welded insulated gas cylinders is a key indicator for evaluating their insulation performance, directly affecting product safety and efficiency. Currently, the weighing method is commonly used to test the static evaporation rate, and this method is listed as an arbitration method in relevant standards. In a typical testing process, inspectors need to go to the site every hour during the test (usually lasting 48 hours) to manually read and record the ambient temperature, atmospheric pressure, and cylinder mass data displayed on the weighing instrument. After the test, the static evaporation rate must be manually calculated according to the physical formulas specified in the standard (involving parameters such as liquid nitrogen saturation temperature and latent heat of vaporization), and the original records and results must be manually entered into the inspection information system to finally generate an inspection report.

[0003] This traditional method has significant limitations: firstly, it relies heavily on manual operation, which is not only labor-intensive and inefficient, but also prone to introducing time deviations and numerical errors during manual reading and recording, affecting the accuracy of intermediate parameters such as average ambient temperature and average atmospheric pressure; secondly, even with automated data acquisition, the obtained mass, ambient temperature, and atmospheric pressure data are still highly susceptible to various interference factors. For example, mass data may experience fluctuations unrelated to actual evaporation due to micro-vibrations of the weighing device, frost or condensation on the cylinder surface, air disturbances, or sensor zero-point drift; ambient temperature is easily affected by air conditioning operation, door and window opening and closing, or personnel movement; and atmospheric pressure may fluctuate due to weather changes or localized ventilation. These interferences are random, coupled, and time-varying, resulting in a large amount of noise in the raw data that is unrelated to actual evaporation behavior.

[0004] The above problems collectively lead to two core defects: First, manual processes are prone to calculation or data entry errors, lack automated record-keeping mechanisms, and the authenticity of the original data and the traceability of the entire process are difficult to guarantee; second, even if automatic data collection is adopted, if the interference data is used directly for calculation without effective screening, it will still introduce significant systematic errors, causing distortion of the static evaporation rate results and affecting the accurate assessment of the gas cylinder's thermal insulation performance.

[0005] Therefore, there is an urgent need for a technical solution for an intelligent detection system for the static evaporation rate of welded insulated gas cylinders based on the weighing method. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides an intelligent detection system for the static evaporation rate of welded insulated gas cylinders based on a weighing method, specifically comprising the following modules:

[0007] Data acquisition module: Used to set up a high-precision weighing device, ambient temperature sensor and atmospheric pressure sensor on site for static evaporation rate detection of welded insulation gas cylinders. The gas cylinder under test is placed on the high-precision weighing device and the data of the gas cylinder under test is collected synchronously at preset time intervals, and the parameters of the gas cylinder under test are obtained at the same time.

[0008] Quality audit module: Connected to the data acquisition module, it is used to process the collected data of the gas cylinders under inspection based on multi-dimensional quality feature extraction, and to perform compound logical judgment based on the data processing results to filter out the data of the gas cylinders under inspection that meet the preset quality conditions.

[0009] Time window sequence division unit: used to divide the quality data, ambient temperature data and atmospheric pressure data into continuous time window sequences according to time order;

[0010] Quality characteristic calculation unit: used to calculate the first quality characteristic, the second quality characteristic, and the third quality characteristic for each time window;

[0011] Quality residual sequence acquisition interface: used to detrend the quality data within each time window to obtain the detrended quality residual sequence;

[0012] First quality feature acquisition interface: used to calculate the moving average of the absolute values ​​of the quality residual sequence and use the moving average as the first quality feature;

[0013] Ambient temperature sequence acquisition interface: used to perform low-pass filtering on ambient temperature data within each time window to suppress high-frequency noise and obtain a smoothed ambient temperature sequence;

[0014] Second quality feature acquisition interface: used to calculate the peak-to-peak value of the smoothed ambient temperature sequence and use the peak-to-peak value as the second quality feature;

[0015] Outlier Removal Interface: Used to remove outliers from atmospheric pressure data within each time window. The outlier removal process includes identifying and removing pressure sampling points that deviate from the local median by more than a preset tolerance.

[0016] Variance calculation interface: used to calculate the variance of the first-order difference sequence of atmospheric pressure data after outlier removal;

[0017] The third quality feature acquisition interface is used to normalize the variance of the first-order difference sequence to obtain a dimensionless stability index, and then use the dimensionless stability index as the third quality feature.

[0018] Anomaly type discrimination unit: used to dynamically determine a first judgment threshold based on the first quality feature and the position of the current time window in the time window sequence, and to identify whether the current time window belongs to the quality mutation anomaly type according to the comparison result of the first quality feature and the first judgment threshold. The quality mutation anomaly type includes no anomaly, instantaneous jump type or continuous drift type.

[0019] Sequence Index Retrieval Interface: Used to retrieve the sequence index of the current time window within the time window sequence;

[0020] First benchmark threshold acquisition interface: used to query the first benchmark threshold associated with the current index from the preset quality anomaly threshold mapping table based on the index number;

[0021] Local stability factor acquisition interface: used to calculate the local stability factor based on the first quality features of N consecutive time windows before the current time window, where N is a positive integer and N≥3;

[0022] First decision threshold acquisition interface: used to multiply the first benchmark threshold by the local stability factor to obtain the dynamically adjusted first decision threshold;

[0023] First anomaly detection interface: used to compare the first quality feature with the first judgment threshold: if the first quality feature is greater than the first judgment threshold, then proceed to the anomaly type detection process; otherwise, determine the quality mutation anomaly type of the current time window as no anomaly.

[0024] The second anomaly detection interface is used in the anomaly type detection process to calculate the absolute value of the difference between the first quality feature of the current time window and the previous time window, and compare it with the preset jump sensitivity threshold.

[0025] The third anomaly detection interface is used to identify the current time window as an instantaneous jump type if the absolute value is greater than the preset jump sensitivity threshold; otherwise, it is identified as a continuous drift type.

[0026] Environmental state discrimination unit: used to dynamically determine the second judgment threshold and the third judgment threshold based on the second quality feature and the third quality feature, respectively, and determine the environmental state category of the current time window based on the joint comparison results of the second quality feature and the second judgment threshold, and the third quality feature and the third judgment threshold. The environmental state category includes stable state, slightly disturbed state, or strongly disturbed state.

[0027] The baseline threshold acquisition interface is used to query the second and third baseline thresholds associated with the current index from a preset environment threshold mapping table based on the index number.

[0028] The benchmark factor acquisition interface is used to calculate the temperature fluctuation benchmark factor and the pressure disturbance benchmark factor based on the second and third quality characteristics of the M consecutive time windows preceding the current time window, respectively, where M is a positive integer and M≥3;

[0029] The threshold acquisition interface is used to multiply the second benchmark threshold by the temperature fluctuation benchmark factor to obtain the dynamically adjusted second threshold; and to multiply the third benchmark threshold by the pressure disturbance benchmark factor to obtain the dynamically adjusted third threshold.

[0030] Judgment result confirmation interface: used to determine whether the second quality feature is greater than the second judgment threshold to obtain the first judgment result; and to determine whether the third quality feature is greater than the third judgment threshold to obtain the second judgment result.

[0031] First environment state discrimination interface: used to determine that the current time window is in a stable state when the first judgment result is less than or equal to and the second judgment result is less than or equal to.

[0032] The second environmental state discrimination interface is used to further determine the current time window when the first judgment result is greater than or the second judgment result is greater than: if only one of the judgment results is greater than, the current time window is determined to be in a slightly disturbed state; if both judgment results are greater than, the current time window is determined to be in a strongly disturbed state.

[0033] Initial validity label assignment unit: used to execute state mapping logic based on the quality mutation anomaly type and environmental state category, and assign an initial validity label to the current time window, wherein the initial validity label includes valid, pending, or invalid;

[0034] Anomaly type and status category acquisition interface: used to obtain the quality mutation anomaly type and environmental status category for the current time window;

[0035] The first validity label discrimination interface is used to set the initial validity label of the current time window to valid when the quality mutation anomaly type is no anomaly and the environmental state category is stable.

[0036] The second validity label discrimination interface is used to set the initial validity label of the current time window to pending when the quality mutation anomaly type is instantaneous jump type and the environmental state category is stable state;

[0037] The third validity label discrimination interface is used to set the initial validity label of the current time window to pending when the quality mutation anomaly type is no anomaly and the environmental state category is mild disturbance.

[0038] The fourth validity label discrimination interface is used to set the initial validity label of the current time window to invalid when the quality mutation anomaly type is continuous drift.

[0039] The fifth validity label discrimination interface is used to set the initial validity label of the current time window to invalid when the environmental state category is strong disturbance dynamic, or the quality mutation anomaly type is instantaneous jump and the environmental state category is not steady state;

[0040] Initial validity label update unit: If the initial validity label of the current time window is pending, and the initial validity labels of the previous and next time windows are both valid, update the initial validity label of the current time window to be valid; if the initial validity label of the current time window is pending, and the initial validity label of the previous or next time window is invalid, update the initial validity label of the current time window to be invalid.

[0041] Data reconstruction unit: used to collect quality data, ambient temperature data, and atmospheric pressure data within the time window when all initial validity labels are updated to be valid and when the initial validity labels are valid, and to construct data for the tested gas cylinder that meets the preset quality conditions.

[0042] Static evaporation rate calculation module: connected to the quality audit module, used to calculate the static evaporation rate based on the data and parameters of the gas cylinder under inspection that meet the preset quality conditions, according to the physical formula specified in the standard, and record the intermediate data generated during the static evaporation rate calculation process;

[0043] Inspection report generation module: Connected to the static evaporation rate calculation module, it is used to associate and store the data of the tested gas cylinders that meet the preset quality conditions, the parameters of the tested gas cylinders, intermediate data, and static evaporation rate according to the number of the tested gas cylinders, and push the static evaporation rate and the data of the tested gas cylinders that meet the preset quality conditions to the external inspection information system, which will automatically generate an inspection report.

[0044] The embodiments of the present invention have the following technical effects:

[0045] This invention effectively overcomes the core shortcomings of traditional static evaporation rate detection, such as strong reliance on manual labor, susceptibility to data interference, and low reliability of results, by constructing an intelligent detection mechanism that integrates multi-dimensional feature extraction, dynamic threshold determination, and state evolution analysis. Addressing the inefficiency and operational errors caused by manual data recording, this invention achieves simultaneous automatic acquisition of quality, ambient temperature, and atmospheric pressure data. Based on the parameters of the tested gas cylinder and the selected high-quality data, it automatically calculates the static evaporation rate and generates the inspection report, significantly reducing human intervention and improving detection efficiency and process traceability. To address the problem of computational distortion caused by various random disturbances affecting raw data, this invention introduces first, second, and third quality features to characterize the local fluctuation intensity of the quality signal, the peak-to-peak variation of ambient temperature, and the stability of atmospheric pressure, respectively, thereby quantifying data quality from multiple dimensions. Based on this, and considering the current time window's position in the detection sequence, a judgment threshold adapted to the detection stage is dynamically generated, avoiding the limitation of fixed thresholds failing to respond to process evolution. Furthermore, by jointly discriminating between the quality anomaly type (including no anomaly, instantaneous jump type, or continuous drift type) and the environmental state category (stable state, mildly disturbed state, or strongly disturbed state), a refined state mapping logic is established. Each time window is assigned an initial validity label, and the contextual relationship between adjacent window labels is used to intelligently repair the "pending" state, thus accurately eliminating unreliable data while retaining valid information. Finally, only high-quality data validated by composite logic is used in the static evaporation rate calculation, significantly improving the accuracy, repeatability, and anti-interference capability of the detection results, achieving intelligent, automated, and highly reliable static evaporation rate detection of welded insulation cylinders. Attached Figure Description

[0046] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0047] Figure 1 This is a framework diagram of the intelligent detection system for static evaporation rate of welded insulated gas cylinders based on the weighing method provided in this embodiment of the invention.

[0048] Figure 2 This is a data reconstruction framework diagram of the intelligent detection system for static evaporation rate of welded insulated gas cylinders based on the weighing method provided in this embodiment of the invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0050] Example 1: As Figure 1 As shown, the present invention provides an intelligent detection system for the static evaporation rate of welded insulating gas cylinders based on the weighing method, comprising the following modules:

[0051] Data acquisition module: Used to set up a high-precision weighing device, ambient temperature sensor and atmospheric pressure sensor on site for static evaporation rate detection of welded insulation gas cylinders. The gas cylinder under test is placed on the high-precision weighing device and the data of the gas cylinder under test is collected synchronously at preset time intervals, and the parameters of the gas cylinder under test are obtained at the same time.

[0052] It is worth noting that, at the site of static evaporation rate testing of welded insulated gas cylinders, the high-precision weighing device is first installed horizontally and stably on a vibration-free testing platform, ensuring sufficient space around it to avoid airflow disturbances caused by personnel movement or equipment operation. Subsequently, the ambient temperature sensor and atmospheric pressure sensor are placed in fixed positions near the gas cylinder under test, within the same testing environment. The ambient temperature sensor should be kept away from heat sources and direct light sources, and the atmospheric pressure sensor should not be placed near ventilation openings or enclosed corners to ensure the representativeness of the measured environmental parameters. The gas cylinder under test is carefully placed at the center of the weighing pan of the high-precision weighing device, ensuring it remains stationary and does not come into contact with any external objects throughout the testing process. After the system is powered on, the data acquisition controller synchronously initializes the high-precision weighing device, ambient temperature sensor, and atmospheric pressure sensor, and sets a unified sampling clock reference. After the testing begins, the system automatically triggers a synchronous data acquisition command at preset time intervals (such as every minute or every 5 minutes). The three types of sensors simultaneously output corresponding mass data, ambient temperature data, and atmospheric pressure data, which are transmitted in real time to the central processing unit via wired or wireless communication links. Simultaneously, the operator inputs or retrieves parameter information for the tested gas cylinder, including nominal volume, nominal working pressure, and rated filling rate, through a human-machine interface. The system automatically associates and stores this information with the acquired data according to timestamps, thus achieving fully unattended automatic data acquisition, time synchronization, and parameter binding, ensuring the integrity, consistency, and timing accuracy of the original data.

[0053] Quality audit module: Connected to the data acquisition module, it is used to process the collected data of the gas cylinders under inspection based on multi-dimensional quality feature extraction, and to perform compound logical judgment based on the data processing results to filter out the data of the gas cylinders under inspection that meet the preset quality conditions.

[0054] Time window sequence division unit: used to divide the quality data, ambient temperature data and atmospheric pressure data into continuous time window sequences according to time order;

[0055] It is worth noting that after data acquisition is completed, this embodiment divides the quality data, ambient temperature data, and atmospheric pressure data arranged in chronological order throughout the entire detection cycle into several continuous and non-overlapping time windows based on a preset time window length (e.g., every 2 hours as a window). Each time window contains all synchronously acquired samples of the three types of data within that time period, and is numbered sequentially according to the acquisition time, forming a structured continuous time window sequence to ensure that subsequent feature calculations are processed based on a unified time granularity.

[0056] Quality characteristic calculation unit: used to calculate the first quality characteristic, the second quality characteristic, and the third quality characteristic for each time window;

[0057] Quality residual sequence acquisition interface: used to detrend the quality data within each time window to obtain the detrended quality residual sequence;

[0058] It is worth noting that, for the mass data within each time window, this embodiment uses the linear least squares method to fit its overall trend and obtain a baseline trend line characterizing slow evaporation or drift. Subsequently, the fitted values ​​at the corresponding times on the trend line are subtracted from the original mass data point by point to eliminate long-term monotonic change components and obtain a detrended mass residual sequence fluctuating around the zero mean, which is used for subsequent local stability analysis.

[0059] First quality feature acquisition interface: used to calculate the moving average of the absolute values ​​of the quality residual sequence and use the moving average as the first quality feature;

[0060] It is worth noting that, for the detrended quality residual sequence obtained within each time window, this embodiment takes the absolute value of each data point and then uses a moving average algorithm with a fixed window length (such as 10 sampling points) to smooth it, thereby obtaining the local mean of the quality fluctuation amplitude within the window; this moving average is used as the first quality feature characterizing the short-term instability of the quality signal and is used for subsequent anomaly detection.

[0061] Ambient temperature sequence acquisition interface: used to perform low-pass filtering on ambient temperature data within each time window to suppress high-frequency noise and obtain a smoothed ambient temperature sequence;

[0062] It is worth noting that, for the ambient temperature data within each time window, this embodiment uses a first-order Butterworth low-pass filter for digital filtering, with a cutoff frequency set to 0.001 Hz (corresponding to a period of approximately 17 minutes) to effectively suppress short-term high-frequency disturbances caused by personnel movement, equipment start-up and shutdown, etc. The filtered output signal retains the slow trend of ambient temperature change, forming a smoothed ambient temperature sequence, which is used for subsequent peak-to-peak value calculation, thereby more accurately reflecting the real ambient temperature change characteristics.

[0063] Second quality feature acquisition interface: used to calculate the peak-to-peak value of the smoothed ambient temperature sequence and use the peak-to-peak value as the second quality feature;

[0064] It is worth noting that, for the smoothed ambient temperature sequence obtained by low-pass filtering within each time window, this embodiment traverses all data points in the sequence, finds the maximum and minimum values, and calculates the difference between them, which is the peak-to-peak value of the window. The peak-to-peak value directly reflects the total fluctuation range of the ambient temperature during the detection period, and serves as the second quality feature characterizing the intensity of the ambient thermal disturbance, which is used for the dynamic determination of the subsequent environmental state category.

[0065] Outlier Removal Interface: Used to remove outliers from atmospheric pressure data within each time window. The outlier removal process includes identifying and removing pressure sampling points that deviate from the local median by more than a preset tolerance.

[0066] It is worth noting that, for the atmospheric pressure data within each time window, this embodiment adopts the sliding window median filtering method: taking the current sampling point as the center, five data points before and after it are taken to form a local window, and the median pressure value within the window is calculated; if the absolute deviation between the current sampling point and the median exceeds the preset tolerance (e.g., 10 Pa), it is determined to be an outlier and is removed, while the remaining data is retained; after processing, clean atmospheric pressure data after removing sudden interference (such as pressure jumps caused by opening and closing doors or starting and stopping equipment) is obtained and used for subsequent stability analysis.

[0067] Variance calculation interface: used to calculate the variance of the first-order difference sequence of atmospheric pressure data after outlier removal;

[0068] It is worth noting that after outlier removal, this embodiment arranges the remaining atmospheric pressure data in each time window in chronological order and calculates the difference between two adjacent sampling points in turn, i.e., the pressure value at the later time minus the pressure value at the previous time, forming a first-order difference sequence. Then, the arithmetic mean of all values ​​in the difference sequence is calculated, and the square of the difference between each difference and the average value is calculated. All squared values ​​are summed and divided by the length of the difference sequence. The result is the variance of the first-order difference sequence, which is used to quantify the instantaneous fluctuation intensity of atmospheric pressure.

[0069] The third quality feature acquisition interface is used to normalize the variance of the first-order difference sequence to obtain a dimensionless stability index, and then use the dimensionless stability index as the third quality feature.

[0070] It is worth noting that, for the variance of the first-order difference sequence calculated for each time window, this embodiment divides it by the square of the average atmospheric pressure data within that window, thereby eliminating the influence of dimensions. Specifically, the arithmetic mean of the atmospheric pressure data after removing outliers is first calculated, then the average value is multiplied by itself to obtain the square value, and finally the first-order difference variance is divided by this square value. The resulting ratio is the dimensionless stability index. This index eliminates the influence of differences in different detection environments or atmospheric pressure benchmarks, can more fairly reflect the relative stability of pressure fluctuations, and is used as a third quality characteristic for subsequent environmental state discrimination.

[0071] It is worth further elaborating that the interface for obtaining the quality residual sequence and the interface for obtaining the third quality feature, by defining the specific calculation methods for the first, second, and third quality features, provide a quantifiable and implementable technical foundation for multidimensional data quality assessment. Specifically, the first quality feature accurately captures local non-evaporation-related fluctuations by taking the moving average of the absolute values ​​of the residuals after detrending the quality data; the second quality feature uses the peak-to-peak value of the ambient temperature after low-pass filtering to effectively reflect the true amplitude of thermal disturbances and avoid high-frequency noise interference; the third quality feature calculates the first-order difference variance and performs dimensional normalization after removing atmospheric pressure anomalies to obtain a dimensionless stability index, eliminating assessment bias under different atmospheric pressure benchmarks. These three types of features characterize data quality from three orthogonal dimensions: the quality signal itself, ambient temperature disturbances, and atmospheric pressure stability, complementing each other without redundancy. Their design fully considers the diversity and coupling of interference sources in actual detection, providing solid feature support for subsequent anomaly identification and environmental discrimination. As a result, the system can more accurately distinguish between real evaporation behavior and spurious fluctuations, providing high-quality input for high-precision static evaporation rate calculation and significantly improving the anti-interference capability and engineering practicality of the entire detection system.

[0072] Anomaly type discrimination unit: used to dynamically determine a first judgment threshold based on the first quality feature and the position of the current time window in the time window sequence, and to identify whether the current time window belongs to the quality mutation anomaly type according to the comparison result of the first quality feature and the first judgment threshold. The quality mutation anomaly type includes no anomaly, instantaneous jump type or continuous drift type.

[0073] Sequence Index Retrieval Interface: Used to retrieve the sequence index of the current time window within the time window sequence;

[0074] It is worth noting that after the time window sequence is divided, this embodiment assigns an integer number starting from 1 to each time window in chronological order. When processing any time window, its corresponding number value is read directly, which is the index of the window in the time window sequence. This index is used to characterize the current detection stage (such as the initial, middle or steady state) and serves as the key basis for subsequent dynamic threshold queries.

[0075] First benchmark threshold acquisition interface: used to query the first benchmark threshold associated with the current index from the preset quality anomaly threshold mapping table based on the index number;

[0076] It is worth noting that the quality anomaly threshold mapping table was constructed before system deployment through statistical analysis of historical detection data: a large number of completed and valid detection processes were aligned according to a time window sequence. For each sequence number (e.g., window 1, window 2, etc.), the mean and standard deviation of its first quality characteristic were calculated, and a first benchmark threshold was set as the mean of that position plus three times the standard deviation. Each sequence number index was paired with its corresponding first benchmark threshold to form a preset quality anomaly threshold mapping table. In actual detection, the system directly looks up the first benchmark threshold associated with the current time window's sequence number index in the table for subsequent dynamic judgment.

[0077] Local stability factor acquisition interface: used to calculate the local stability factor based on the first quality features of N consecutive time windows before the current time window, where N is a positive integer and N≥3;

[0078] It is worth noting that in this embodiment, the first quality feature values ​​corresponding to the N consecutive time windows (N is a preset positive integer and not less than 3) before the current time window are obtained to form a local feature sequence. Then, the standard deviation of the sequence is calculated and divided by the average value of the sequence to obtain the coefficient of variation. The coefficient of variation is used as a local stability factor to characterize the relative stability of recent data fluctuations. The smaller the factor, the more stable the quality feature is and the more stable the current system operation is.

[0079] First decision threshold acquisition interface: used to multiply the first benchmark threshold by the local stability factor to obtain the dynamically adjusted first decision threshold;

[0080] It is worth noting that in this embodiment, the first benchmark threshold obtained from the quality anomaly threshold mapping table is multiplied by the currently calculated local stability factor, and the resulting product is the dynamically adjusted first judgment threshold. Specifically, when the local stability factor is less than 1, it indicates that the recent data fluctuations are small, and the system automatically lowers the judgment threshold to improve sensitivity; when the local stability factor is greater than 1, it indicates that the recent fluctuations are large, and the system raises the judgment threshold accordingly to avoid false alarms. This mechanism realizes the adaptive adjustment of the judgment threshold to the local stability of the detection process.

[0081] First anomaly detection interface: used to compare the first quality feature with the first judgment threshold: if the first quality feature is greater than the first judgment threshold, then proceed to the anomaly type detection process; otherwise, determine the quality mutation anomaly type of the current time window as no anomaly.

[0082] It is worth noting that this embodiment compares the first quality feature of the current time window with the dynamically adjusted first judgment threshold: if the first quality feature is greater than the first judgment threshold, it indicates that the quality fluctuation within the window significantly exceeds the recent normal level, and there may be a sudden anomaly, thus triggering the anomaly type discrimination process; otherwise, the fluctuation is considered to be within a reasonable range and is judged as not abnormal. This judgment logic is based on the statistical principle that "abnormality is significant deviation"—the first quality feature reflects the intensity of local quality fluctuations, while the first judgment threshold has integrated prior knowledge from the detection stage and recent stability, and has adaptability. Using "greater than" as the criterion can effectively eliminate random noise interference and avoid misjudging normal small fluctuations as abnormal; at the same time, fine-grained classification is only activated when the limit is exceeded, balancing computational efficiency and discrimination accuracy, ensuring that there is sufficient basis for subsequent identification of instantaneous jump types or continuous drift types.

[0083] The second anomaly detection interface is used in the anomaly type detection process to calculate the absolute value of the difference between the first quality feature of the current time window and the previous time window, and compare it with the preset jump sensitivity threshold.

[0084] It is worth noting that in the anomaly type discrimination process, this embodiment obtains the first quality feature values ​​of the current time window and the previous time window, calculates the absolute value of the difference between the two, and uses it as a quantitative indicator of the degree of abrupt change in quality fluctuation between adjacent windows. This value is then compared with a preset jump sensitivity threshold. The jump sensitivity threshold is determined by the statistical distribution of the difference in the first quality feature values ​​of adjacent windows in historical normal detection data, and is usually set to its mean plus twice the standard deviation to cover the vast majority of normal fluctuations. This threshold is used to distinguish between "slow drift" and "sudden jump," because instantaneous interference (such as touch or vibration) usually causes drastic changes in the feature values ​​of adjacent windows, while continuous drift is a gradual change. Therefore, this comparison can provide a key criterion for subsequent anomaly type identification.

[0085] The third anomaly detection interface is used to identify the current time window as an instantaneous jump type if the absolute value is greater than the preset jump sensitivity threshold; otherwise, it is identified as a continuous drift type.

[0086] It is worth noting that this embodiment determines whether the absolute value obtained from the second anomaly discrimination interface is greater than a preset jump sensitivity threshold: if so, the current time window is identified as an instantaneous jump type; otherwise, it is identified as a continuous drift type. This criterion is based on the essential difference between the two types of anomalies—instantaneous jumps are caused by sudden external interference (such as touch or vibration), resulting in abrupt changes in the quality characteristics of adjacent windows and significant differences; while continuous drift originates from slow sensor failure or gradual environmental changes, with gradual changes in characteristic values ​​and small differences between adjacent windows. Since all windows entering this process have been confirmed to have excessive fluctuations, only these two types remain, so a single threshold can be used to achieve mutual exclusion determination. This method is logically simple, computationally efficient, avoids introducing complex models, and accurately distinguishes between repairable instantaneous anomalies and systematic drifts that need to be eliminated, providing a reliable basis for subsequent data validity determination.

[0087] It is worth further explaining that the sequence index retrieval interface to the third anomaly discrimination interface effectively solves the problem that traditional methods cannot distinguish between instantaneous jumps and continuous drifts, leading to excessive removal of valid data or misuse of contaminated data by constructing a dynamic judgment mechanism oriented towards quality mutation anomalies. The system first queries the first baseline threshold adapted during the query stage based on the current window's position in the sequence, and then calculates the local stability factor by combining the first quality features of the recent N windows. The two are multiplied to generate an adaptive first judgment threshold, so that the threshold can respond to changes in the detection stage and adapt to local data stability. On this basis, anomaly type subdivision is initiated only when the first quality feature exceeds the limit: by comparing the difference between the current and previous window features with the jump sensitivity threshold, instantaneous jump type (isolated mutation) or continuous drift type (systematic failure) is accurately identified. This criterion is logically rigorous, computationally efficient, and the two types of anomalies are mutually exclusive, covering all cases exceeding the limit. Therefore, the system can retain the "pending" state for instantaneous jumps for subsequent repair, while marking continuous drifts as invalid. This maximizes the retention of usable data while resolutely eliminating unreliable information, significantly improving the intelligence and rationality of data filtering and laying the foundation for subsequent high-precision calculations.

[0088] Environmental state discrimination unit: used to dynamically determine the second judgment threshold and the third judgment threshold based on the second quality feature and the third quality feature, respectively, and determine the environmental state category of the current time window based on the joint comparison results of the second quality feature and the second judgment threshold, and the third quality feature and the third judgment threshold. The environmental state category includes stable state, slightly disturbed state, or strongly disturbed state.

[0089] The baseline threshold acquisition interface is used to query the second and third baseline thresholds associated with the current index from a preset environment threshold mapping table based on the index number.

[0090] It is worth noting that the environmental threshold mapping table was constructed using a large amount of historical valid detection data before system deployment: each detection process was aligned by time window, and for each sequence position (e.g., window 1, window 2, etc.), the normal distribution range of its second quality characteristic (temperature peak-to-peak value) and third quality characteristic (pressure stability index) was statistically analyzed. The second benchmark threshold was set as the mean of the temperature peak-to-peak value plus twice the standard deviation, and the third benchmark threshold was set as the mean of the pressure stability index plus twice the standard deviation. These two thresholds were then bound to their corresponding sequence indexes to form a pre-defined environmental threshold mapping table. In actual detection, the system queries this table based on the sequence index of the current time window in the time window sequence—this index is completely consistent with the sequence index used for quality anomaly detection in the sequence index acquisition interface, i.e., the position number of the same window in a unified sequence—to obtain the corresponding second and third benchmark thresholds for subsequent dynamic determination of environmental status.

[0091] The benchmark factor acquisition interface is used to calculate the temperature fluctuation benchmark factor and the pressure disturbance benchmark factor based on the second and third quality characteristics of the M consecutive time windows preceding the current time window, respectively, where M is a positive integer and M≥3;

[0092] It is worth noting that this embodiment obtains the second and third quality features corresponding to M consecutive time windows (M is a preset positive integer and not less than 3) preceding the current time window, respectively constructing a temperature feature sequence and a pressure feature sequence; the arithmetic mean of the temperature feature sequence is calculated, and this average is used as a temperature fluctuation benchmark factor; similarly, the arithmetic mean of the pressure feature sequence is calculated, serving as a pressure disturbance benchmark factor. These two benchmark factors reflect the average level of recent environmental temperature fluctuations and atmospheric pressure disturbances, and are used to dynamically adjust the judgment threshold to adapt it to the actual environmental conditions of the current detection stage.

[0093] The threshold acquisition interface is used to multiply the second benchmark threshold by the temperature fluctuation benchmark factor to obtain the dynamically adjusted second threshold; and to multiply the third benchmark threshold by the pressure disturbance benchmark factor to obtain the dynamically adjusted third threshold.

[0094] It is worth noting that in this embodiment, the second benchmark threshold obtained from the environmental threshold mapping table is multiplied by the currently calculated temperature fluctuation benchmark factor, and the product is used as the dynamically adjusted second judgment threshold. Similarly, the third benchmark threshold is multiplied by the pressure disturbance benchmark factor, and the product is used as the dynamically adjusted third judgment threshold. This multiplication operation achieves adaptive scaling of the benchmark thresholds: when the recent temperature fluctuation is large, the temperature fluctuation benchmark factor is greater than 1, and the second judgment threshold is increased accordingly to avoid misjudging normal environmental changes as disturbances; conversely, the threshold is reduced to improve sensitivity. The pressure disturbance judgment is handled similarly, thereby ensuring that the environmental state discrimination is both robust and sensitive.

[0095] Judgment result confirmation interface: used to determine whether the second quality feature is greater than the second judgment threshold to obtain the first judgment result; and to determine whether the third quality feature is greater than the third judgment threshold to obtain the second judgment result.

[0096] First environment state discrimination interface: used to determine that the current time window is in a stable state when the first judgment result is less than or equal to and the second judgment result is less than or equal to.

[0097] The second environmental state discrimination interface is used to further determine the current time window when the first judgment result is greater than or the second judgment result is greater than: if only one of the judgment results is greater than, the current time window is determined to be in a slightly disturbed state; if both judgment results are greater than, the current time window is determined to be in a strongly disturbed state.

[0098] It is worth noting that this embodiment first determines whether the second quality characteristic is greater than the second judgment threshold to obtain the first judgment result; simultaneously, it determines whether the third quality characteristic is greater than the third judgment threshold to obtain the second judgment result. If both judgment results are "less than or equal to", that is, neither temperature fluctuation nor pressure disturbance exceeds its respective dynamic threshold, then the current time window is determined to be in a stable state. If either judgment result is "greater than", then the disturbance classification is entered: if only one exceeds the limit, it is judged as a mild disturbance; if both exceed the limit, it is judged as a strong disturbance. This logic is based on the independence and superposition effect of environmental disturbances—temperature and pressure disturbances are often caused by different sources (such as air conditioning affecting temperature, ventilation affecting air pressure), a single exceedance represents local disturbance, and double exceedance indicates overall environmental instability. The "OR" relationship ensures that no disturbance is missed, and the "AND" relationship identifies compound disturbances. The three states are mutually exclusive and cover all possibilities. This criterion structure is clear and does not require a complex model. It can accurately reflect the real state of the environment and provide reliable input for subsequent data validity mapping, significantly improving the system's anti-interference ability and discrimination robustness.

[0099] It is worth further explaining that the dynamic hierarchical discrimination mechanism for environmental states established between the baseline threshold acquisition interface and the second environmental state discrimination interface effectively overcomes the shortcomings of existing technologies in handling environmental interference with a "one-size-fits-all" approach. The system queries the preset second and third baseline thresholds based on the current window number, and calculates the temperature fluctuation baseline factor and pressure disturbance baseline factor respectively by combining the second and third quality characteristics of the recent M windows. This multiplication achieves a dual dynamic adjustment of the thresholds, allowing the judgment criteria to adapt to changes in the detection process and local environmental stability. Subsequently, by jointly judging whether the peak-to-peak temperature and pressure stability exceed limits, the environmental state is finely divided into three categories: stable state, mild disturbance state, or strong disturbance state. Both not exceeding limits indicate stability, a single exceeding limit indicates mild disturbance, and both exceeding limits indicate strong disturbance. This three-category classification logic covers all possible combinations, is mutually exclusive, and conforms to physical reality—a single disturbance source (such as air conditioning start-up / shutdown) causes mild disturbance, while compound disturbances (such as drastic weather changes combined with ventilation) cause strong disturbance. Therefore, the system can implement differentiated data processing strategies for different environmental conditions, avoid data being mistakenly removed due to normal environmental fluctuations, and resolutely block unreliable periods under strong interference, significantly improving environmental adaptability and robustness of judgment.

[0100] Initial validity label assignment unit: used to execute state mapping logic based on the quality mutation anomaly type and environmental state category, and assign an initial validity label to the current time window, wherein the initial validity label includes valid, pending, or invalid;

[0101] Anomaly type and status category acquisition interface: used to obtain the quality mutation anomaly type and environmental status category for the current time window;

[0102] The first validity label discrimination interface is used to set the initial validity label of the current time window to valid when the quality mutation anomaly type is no anomaly and the environmental state category is stable.

[0103] The second validity label discrimination interface is used to set the initial validity label of the current time window to pending when the quality mutation anomaly type is instantaneous jump type and the environmental state category is stable state;

[0104] The third validity label discrimination interface is used to set the initial validity label of the current time window to pending when the quality mutation anomaly type is no anomaly and the environmental state category is mild disturbance.

[0105] The fourth validity label discrimination interface is used to set the initial validity label of the current time window to invalid when the quality mutation anomaly type is continuous drift.

[0106] The fifth validity label discrimination interface is used to set the initial validity label of the current time window to invalid when the environmental state category is strong disturbance dynamic, or the quality mutation anomaly type is instantaneous jump and the environmental state category is not steady state;

[0107] It is worth noting that, after determining the quality state category (including no anomaly, instantaneous jump type, and continuous drift type) and the environmental state category (steady state, mild disturbance, and strong disturbance) in this embodiment, an initial validity label is assigned to the current time window according to the preset state mapping rules. The specific implementation process is as follows: If the quality status is "no anomaly" and the environmental status is "stable state", it indicates that the data is not affected by quality mutations and the environmental conditions are ideal, and the system marks the window as "valid"; if the quality status is "instantaneous jump type" but the environmental status is "stable state", it indicates that the anomaly may be caused by a brief external disturbance (such as human touch), while the environment itself is stable and the data may be repaired, so it is marked as "pending"; if the quality status is "no anomaly" but the environment is in a "mildly disturbed state", it indicates that the quality signal is reliable but the environment has slight fluctuations. Although it does not affect the trend judgment, it should be used with caution, and it is also marked as "pending"; if the quality status is "continuous drift type", regardless of the environment, it is considered that the sensor or system has suffered irreversible failure, and the data is unreliable, and it is directly marked as "invalid"; in addition, if the environmental status is "strongly disturbed state" (regardless of the quality status), or the quality status is "instantaneous jump type" and the environment is not stable (i.e., in a mild or strongly disturbed state), it indicates that the interference source is complex and the data is severely distorted, and it cannot be repaired by subsequent context, and it is also marked as "invalid";

[0108] The mapping logic is designed based on an in-depth analysis of the interference mechanism: "Valid" retains only high-quality data under ideal conditions; "Pending" focuses on single, interpretable, and local anomalies, providing candidate windows for the state evolution and repair of subsequent initial validity label update units; and "Invalid" covers all unreliable scenarios, including systemic failures (continuous drift) or compound interference (strong environmental disturbances superimposed on quality anomalies). These nine combinations are comprehensive, mutually exclusive, and intuition-compliant, avoiding excessive rejection of valid data and preventing contaminated data from entering the calculation, significantly improving the accuracy and robustness of the final static evaporation rate results.

[0109] It is worth further explaining that the interface for obtaining anomaly types and state categories, and the fifth validity label discrimination interface, construct a nine-grid mapping rule of "quality anomaly type - environmental state category" to achieve refined and scenario-based assignment of validity labels for time windows. This solves the problems of traditional methods lacking context awareness and having coarse label judgment. The system comprehensively considers all combinations of whether the quality is abnormal (no anomaly, instantaneous jump, continuous drift) and whether the environment is stable (stable, slight disturbance, strong disturbance). It is only marked as "valid" when both are ideal (no anomaly + stable state). When there is a single controllable anomaly (such as instantaneous jump but stable environment, or no anomaly but slight disturbance), it is marked as "pending" and left for subsequent context verification. Once continuous drift or strong environmental disturbance (regardless of quality) occurs, or instantaneous jump is superimposed on an unstable environment, it is directly marked as "invalid". This mapping logic profoundly reflects the interference coupling mechanism: isolated, explainable anomalies are repairable, while systemic or compound interference is unreliable. Therefore, the initial labels are neither too conservative (to avoid the loss of effective data) nor too lenient (to prevent noise pollution), providing a high-quality candidate set for the state evolution repair of the initial validity label update unit. Ultimately, this mechanism significantly improves the scientific rigor of data screening and the reliability of results, serving as a key guarantee for achieving high-precision static evaporation rate detection.

[0110] Initial validity label update unit: If the initial validity label of the current time window is pending, and the initial validity labels of the previous and next time windows are both valid, update the initial validity label of the current time window to be valid; if the initial validity label of the current time window is pending, and the initial validity label of the previous or next time window is invalid, update the initial validity label of the current time window to be invalid.

[0111] It is worth noting that after assigning initial validity labels to all time windows in this embodiment, the entire time window sequence is traversed, and context verification is performed on each window labeled "pending". If the labels of its preceding and following windows are both "valid", it is considered that the "pending" window is in a stable data segment, and its anomaly is likely caused by isolated transient interference. The data as a whole is reliable, so its label is updated to "valid". If either its preceding or following window is "invalid", it indicates that the "pending" window is adjacent to an unreliable area, may be affected by continuous interference or is on the edge of data breakage, and its authenticity cannot be guaranteed. Therefore, its label is updated to "invalid". This judgment is based on the assumption of temporal continuity—the real evaporation process is smooth and stable, and effective data should be continuously distributed. Isolated "pending" points can be "repaired" if they are surrounded by effective data, and therefore their labels are updated to "effective." Ultimately, only window data labeled "effective" is used to calculate the static evaporation rate. Those windows that are "repaired" (changed from "pending" to "effective") have their original collected data unchanged; they are simply re-accepted due to the credibility of their context, i.e., "repaired." Conversely, if they are adjacent to invalid regions, they are more likely to be a continuation of interference. This rule does not require a complex model; it relies solely on local label relationships to intelligently repair occasional anomalies and remove suspicious edge data. It maximizes the preservation of effective information while avoiding the introduction of noise, significantly improving the quality and completeness of the final subset of data used for calculation.

[0112] Data reconstruction unit: used to collect quality data, ambient temperature data and atmospheric pressure data within the time window when all initial validity labels are updated to be valid and when the initial validity labels are valid, and to construct data of the gas cylinder under test that meets the preset quality conditions;

[0113] It is worth noting that the transition from time window sequence division to data reconstruction effectively solves the core problem of distortion in raw detection data due to various interferences by introducing a multi-dimensional quality feature extraction and composite logic judgment mechanism based on time window sequences. The system divides continuously collected data into ordered time windows and calculates three types of features representing quality fluctuations, temperature changes, and pressure stability for each window, thereby dynamically identifying the type of quality anomaly and the category of environmental state. Based on this, by establishing a joint "quality-environment" mapping rule, each window is assigned an initial validity label (valid, pending, or invalid), and the contextual relationship between adjacent window labels is used to intelligently repair "pending" windows—only those windows that are valid before and after them are adopted; otherwise, they are discarded. This mechanism avoids the insensitivity of traditional fixed threshold methods to process evolution and overcomes the limitations of single-dimensional discrimination, which is prone to misjudgment. Ultimately, only high-quality, high-reliability data is used for evaporation rate calculation, significantly improving the accuracy and repeatability of the results. This technical solution achieves refined, structured, and adaptive judgment of data validity, representing a key innovation in ensuring the scientific rigor and reliability of static evaporation rate detection.

[0114] Static evaporation rate calculation module: connected to the quality audit module, used to calculate the static evaporation rate based on the data and parameters of the gas cylinder under inspection that meet the preset quality conditions, according to the physical formula specified in the standard, and record the intermediate data generated during the static evaporation rate calculation process;

[0115] It is worth noting that, after data screening, this embodiment gathers mass data, ambient temperature data, and atmospheric pressure data within the time window where the initial validity label has been updated to valid and the initial validity label is valid, forming a dataset of the tested gas cylinders that meet the preset mass conditions. Based on this dataset, the average mass change rate over the entire valid period is first calculated, i.e., the mass difference between the start and end of the valid period is divided by the corresponding time interval. Simultaneously, the average ambient temperature over the valid period is calculated, and the saturated vapor pressure and corresponding latent heat of vaporization of liquid nitrogen at that temperature are determined by looking up a table or by fitting a formula based on the average temperature. Then, combined with the nominal volume, nominal working pressure, and rated filling rate of the tested gas cylinder, the above parameters are substituted into the calculation according to the static evaporation rate physical formula specified in the standard to obtain the static evaporation rate result expressed as mass loss rate or equivalent gas volume. Throughout the calculation process, the system automatically records all intermediate data, including the start and end times of the valid period, average ambient temperature, average atmospheric pressure, mass change, latent heat of vaporization value, and calculation results of each step, ensuring that the calculation process is traceable and verifiable.

[0116] Inspection report generation module: connected to the static evaporation rate calculation module, it is used to associate and store the data of the tested gas cylinders that meet the preset quality conditions, the parameters of the tested gas cylinders, intermediate data and static evaporation rate according to the number of the tested gas cylinders, and push the static evaporation rate and the data of the tested gas cylinders that meet the preset quality conditions to the external inspection information system, which will automatically generate the inspection report.

[0117] It is worth noting that, after completing the static evaporation rate calculation, this embodiment unifies the data of the tested gas cylinders that meet the preset quality conditions (including mass, ambient temperature, and atmospheric pressure data within the effective time window), the parameters of the tested gas cylinders (nominal volume, nominal working pressure, and rated filling rate), intermediate data generated during the calculation process (such as the start and end times of the effective period, average environmental parameters, mass change, latent heat of vaporization, etc.), and the final static evaporation rate result, and stores them in a structured association according to the unique number of the tested gas cylinder, ensuring data integrity and traceability. Subsequently, the system automatically pushes the static evaporation rate value and the data of the tested gas cylinders that meet the preset quality conditions to an external inspection information system through a standard communication interface (such as API or message queue). After receiving the data, the external inspection information system automatically fills in the test results, original data summary, and key parameters according to the preset report template, generating a standardized and complete electronic inspection report, and supports review, issuance, and archiving, realizing paperless and automated testing throughout the entire process.

[0118] It is worth further explaining that by constructing an end-to-end intelligent testing system from the data acquisition module to the test report generation module, the system fundamentally solves the problems of low efficiency, operational errors, and poor traceability caused by the heavy reliance on manual recording, calculation, and data entry in traditional static evaporation rate testing. The system automatically deploys high-precision weighing devices, ambient temperature sensors, and atmospheric pressure sensors at the testing site, achieving simultaneous acquisition of these three key data types. Combined with the structural parameters of the tested gas cylinders, the entire process, from raw data acquisition, quality screening, evaporation rate calculation to report generation, can be completed without manual intervention. This automation mechanism not only significantly reduces the workload of testing personnel but also completely eliminates the risks of time deviations in manual readings, recording errors, and manual calculation mistakes, ensuring the accuracy of intermediate parameters (such as average ambient temperature and average atmospheric pressure). Simultaneously, all raw data, intermediate results, and final conclusions are stored and linked by gas cylinder number and automatically pushed to an external testing information system, forming a complete and tamper-proof data chain, greatly improving the transparency of the testing process and the legal validity of the results. Thus, this invention realizes a paradigm shift in static evaporation rate detection from "human experience-driven" to "data intelligence-driven," providing solid support for high-reliability and high-efficiency gas cylinder performance evaluation.

[0119] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. The intelligent detection system for static evaporation rate of welded adiabatic gas cylinder based on weighing method, characterized in that, The application relates to a welding adiabatic gas cylinder static evaporation rate detection system. The system comprises the following modules: a data acquisition module: used for arranging high-precision weighing devices, environment temperature sensors and atmospheric pressure sensors on the site of welding adiabatic gas cylinder static evaporation rate detection, placing the detected gas cylinder on the high-precision weighing device, synchronously collecting the data of the detected gas cylinder at preset time intervals, and simultaneously acquiring the parameters of the detected gas cylinder; a quality auditing module: connected with the data acquisition module, used for performing data processing on the collected data of the detected gas cylinder based on multi-dimensional quality feature extraction, performing compound logic judgment based on the data processing result, and screening the data of the detected gas cylinder meeting preset quality conditions; a static evaporation rate calculation module: connected with the quality auditing module, used for calculating the static evaporation rate according to the data of the detected gas cylinder meeting the preset quality conditions and the parameters of the detected gas cylinder, according to a physical formula stipulated in a standard, and recording intermediate data generated in the static evaporation rate calculation process; 2. The intelligent detection system for static evaporation rate of the welded adiabatic cylinder based on the weighing method according to claim 1, characterized in that, an inspection report generation module: connected with the static evaporation rate calculation module, used for associating and storing the data of the detected gas cylinder meeting the preset quality conditions, the parameters of the detected gas cylinder, the intermediate data and the static evaporation rate according to the number of the detected gas cylinder, and pushing the static evaporation rate and the data of the detected gas cylinder meeting the preset quality conditions to an external inspection information system to automatically generate an inspection report by the external inspection information system.

3. The intelligent detection system for static evaporation rate of the welded adiabatic cylinder based on the weighing method according to claim 2, characterized in that, The data comprises quality data, environment temperature data and atmospheric pressure data; and the parameters comprise nominal volume, nominal working pressure and rated filling rate. The data processing on the collected data of the detected gas cylinder based on multi-dimensional quality feature extraction and the compound logic judgment based on the data processing result to screen the data of the detected gas cylinder meeting the preset quality conditions comprise: a time window sequence division unit: used for dividing the quality data, the environment temperature data and the atmospheric pressure data into continuous time window sequences in time sequence; a quality feature calculation unit: used for calculating a first quality feature, a second quality feature and a third quality feature for each time window; an abnormal type discrimination unit: used for dynamically determining a first determination threshold based on the first quality feature and the position of the current time window in the time window sequence, and identifying whether the current time window belongs to a quality mutation abnormal type according to the comparison result of the first quality feature and the first determination threshold, wherein the quality mutation abnormal type comprises no abnormality, instantaneous jump type or continuous drift type; an environment state discrimination unit: used for dynamically determining a second determination threshold and a third determination threshold based on the second quality feature and the third quality feature respectively, and determining the environment state category of the current time window according to the joint comparison result of the second quality feature and the second determination threshold and the third quality feature and the third determination threshold, wherein the environment state category comprises stable state, light disturbance state or strong disturbance state; an initial validity label assignment unit: used for executing state mapping logic according to the quality mutation abnormal type and the environment state category, and assigning an initial validity label to the current time window, wherein the initial validity label comprises valid, pending or invalid. The initial validity label updating unit is configured to update the initial validity label of the current time window to valid if the initial validity label of the current time window is pending and the initial validity labels of the previous time window and the next time window of the current time window are both valid; and update the initial validity label of the current time window to invalid if the initial validity label of the current time window is pending and the initial validity label of the previous time window or the next time window of the current time window is invalid. The data reconstruction unit is configured to collect the quality data, the environmental temperature data and the atmospheric pressure data in the time window whose initial validity label is updated to valid and the time window whose initial validity label is valid, and form the data of the inspected gas cylinder meeting the preset quality condition.

4. The intelligent detection system for static evaporation rate of the welded adiabatic cylinder based on the weighing method according to claim 3, characterized in that, The first quality feature, the second quality feature and the third quality feature are calculated for each time window, including: The quality residual sequence acquisition interface is configured to perform detrending processing on the quality data in each time window to obtain a detrended quality residual sequence. The first quality feature acquisition interface is configured to calculate a sliding average value of the absolute value of the quality residual sequence and take the sliding average value as the first quality feature. The environmental temperature sequence acquisition interface is configured to perform low-pass filtering on the environmental temperature data in each time window to suppress high-frequency noise and obtain a smoothed environmental temperature sequence. The second quality feature acquisition interface is configured to calculate a peak-to-peak value of the smoothed environmental temperature sequence and take the peak-to-peak value as the second quality feature. The outlier rejection interface is configured to perform outlier rejection processing on the atmospheric pressure data in each time window, and the outlier rejection processing includes identifying and removing pressure sampling points deviating from the local median by more than a preset tolerance. The variance calculation interface is configured to calculate the variance of the first-order difference sequence of the atmospheric pressure data after the outliers are removed. The third quality feature acquisition interface is configured to perform dimensionless normalization processing on the variance of the first-order difference sequence to obtain a dimensionless stability index, and take the dimensionless stability index as the third quality feature.

5. The intelligent detection system for static evaporation rate of welded adiabatic cylinder based on weighing method according to claim 4, characterized in that, The first quality feature and the position of the current time window in the time window sequence are used to dynamically determine the first determination threshold, and the comparison result of the first quality feature and the first determination threshold is used to identify whether the current time window belongs to the quality mutation abnormal type, and the quality mutation abnormal type includes no abnormality, instantaneous jump type or continuous drift type, including: The serial number index acquisition interface is configured to obtain the serial number index of the current time window in the time window sequence. The first reference threshold acquisition interface is configured to query the first reference threshold associated with the current serial number index from the preset quality abnormal threshold mapping table according to the serial number index. The local stability factor acquisition interface is configured to calculate a local stability factor based on the first quality features of the previous N time windows of the current time window, where N is a positive integer and N≥3. The first determination threshold acquisition interface is configured to multiply the first reference threshold and the local stability factor to obtain the dynamically adjusted first determination threshold. The first abnormality discrimination interface is used for comparing the first quality feature with a first discrimination threshold value: if the first quality feature is greater than the first discrimination threshold value, the abnormality type discrimination process is entered; otherwise, it is determined that the quality mutation abnormality type of the current time window is no abnormality; The second abnormality discrimination interface is used for, in the abnormality type discrimination process, calculating an absolute value of a difference between the first quality feature of the current time window and a first quality feature of a previous time window of the current time window, and comparing the absolute value with a preset jump sensitivity threshold value; The third abnormality discrimination interface is used for, if the absolute value is greater than the preset jump sensitivity threshold value, identifying that the current time window belongs to the instantaneous jump type; otherwise, identifying that the current time window belongs to the continuous drift type.

6. The intelligent detection system for static evaporation rate of the welded adiabatic cylinder based on the weighing method according to claim 5, characterized in that, The second discrimination threshold value and the third discrimination threshold value are dynamically determined based on the second quality feature and the third quality feature respectively, and a joint comparison result of the second quality feature and the second discrimination threshold value and the third quality feature and the third discrimination threshold value is used to determine an environment state category in which the current time window is located, and the environment state category includes a stable state, a light disturbance state or a strong disturbance state, and includes the following: The reference threshold value acquisition interface is used for querying, according to a serial number index, a second reference threshold value and a third reference threshold value associated with the current serial number index from a preset environment threshold value mapping table; The reference factor acquisition interface is used for calculating a temperature fluctuation reference factor and a pressure disturbance reference factor based on the second quality feature and the third quality feature of the M continuous time windows before the current time window respectively, where M is a positive integer and M≥3; The discrimination threshold value acquisition interface is used for multiplying the second reference threshold value and the temperature fluctuation reference factor to obtain a dynamically adjusted second discrimination threshold value, and multiplying the third reference threshold value and the pressure disturbance reference factor to obtain a dynamically adjusted third discrimination threshold value; The judgment result confirmation interface is used for judging whether the second quality feature is greater than the second discrimination threshold value to obtain a first judgment result, and judging whether the third quality feature is greater than the third discrimination threshold value to obtain a second judgment result; The first environment state discrimination interface is used for, when the first judgment result is less than or equal and the second judgment result is less than or equal, determining that the current time window is in the stable state; The second environment state discrimination interface is used for, when the first judgment result is greater than or the second judgment result is greater than, further judging: if only one of the judgment results is greater than, it is determined that the current time window is in the light disturbance state; if both of the judgment results are greater than, it is determined that the current time window is in the strong disturbance state.

7. The intelligent detection system for static evaporation rate of the welded adiabatic cylinder based on the weighing method according to claim 6, characterized in that, The state mapping logic is executed according to the quality mutation abnormality type and the environment state category to give an initial validity label to the current time window, and the initial validity label includes valid, pending or invalid, and includes the following: The abnormality type and state category acquisition interface is used for acquiring the quality mutation abnormality type and the environment state category of the current time window; The first validity label discrimination interface is used for, when the quality mutation abnormality type is no abnormality and the environment state category is the stable state, setting the initial validity label of the current time window as valid. The second validity label discrimination interface is configured to set the initial validity label of the current time window as pending when the quality mutation anomaly type is the instantaneous jump type and the environment state category is the stable state; The third validity label discrimination interface is configured to set the initial validity label of the current time window as pending when the quality mutation anomaly type is the no anomaly and the environment state category is the mild disturbance state; The fourth validity label discrimination interface is configured to set the initial validity label of the current time window as invalid when the quality mutation anomaly type is the persistent drift type; The fifth validity label discrimination interface is configured to set the initial validity label of the current time window as invalid when the environment state category is the strong disturbance state, or the quality mutation anomaly type is the instantaneous jump type and the environment state category is not the stable state.

Images