Pipeline valve leakage temperature and humidity signal adaptive dynamic threshold monitoring method

Abstract

The present application belongs to the technical field of nuclear equipment pipeline valve leakage temperature and humidity signal analysis, and particularly relates to a pipeline valve leakage temperature and humidity signal adaptive dynamic threshold monitoring method. Data is collected and a data benchmark is calculated according to non-leakage data; firstly, the data trend is determined through linear regression; then, the threshold initial value is set, and the leakage state is divided into non-leakage, attention leakage, suspected leakage and occurrence leakage through the data benchmark and the threshold initial value; finally, the probabilities of different leakage states are calculated; the weights of benchmark methods 1 and 2 are initialized, and the leakage state is comprehensively analyzed; the threshold in benchmark method 1 and benchmark method 2 and the weights of the two benchmark methods are updated through the real label and the diagnostic label, and the final leakage state is output. The present application effectively characterizes and judges the pipeline valve state by adaptively extracting the dynamic threshold from the real-time collected temperature and humidity data, thereby providing support for nuclear equipment pipeline valve leakage monitoring.

Description

Technical Field

[0001] This invention belongs to the technical field of temperature and humidity signal analysis for pipeline valve leakage in nuclear equipment, and specifically relates to an adaptive dynamic threshold monitoring method for temperature and humidity signals of pipeline valve leakage. Background Technology

[0002] In recent years, numerous leaks, including spills and drips, have occurred in nuclear equipment pipelines and valves both domestically and internationally. Existing methods for detecting leaks in nuclear equipment pipelines and valves mainly fall into two categories: one relies on post-incident inspections of pipeline and valve status. This method cannot promptly identify the leak source and facilitate repairs, potentially leading to escalation of the leak and more severe economic losses. The other relies on parameters such as system pressure and water level to determine leaks; however, small leaks are unlikely to cause changes in system parameters. Furthermore, due to the complexity of pipeline and valve types and operating conditions, experience-based judgments or fixed threshold alarm methods based on system pressure and other parameters are not applicable to all pipeline and valve status assessments and are prone to false alarms.

[0003] This invention proposes an adaptive dynamic threshold monitoring method for temperature and humidity signals of pipeline valve leakage by reviewing relevant domestic and international standards and combining field testing experience. Summary of the Invention

[0004] The purpose of this invention is to provide an adaptive dynamic threshold monitoring method for temperature and humidity signals of pipeline valve leakage. By adaptively extracting dynamic thresholds from real-time collected temperature and humidity data, the method effectively characterizes and judges the status of pipeline valves, providing support for pipeline valve leakage monitoring in nuclear equipment.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: An adaptive dynamic threshold monitoring method for temperature and humidity signals in pipeline valve leakage: 1. Data acquisition, and calculation of data baseline based on leak-free data; II. Leakage Judgment Baseline Method 1: First, determine the data trend through linear regression; then, set an initial threshold value, and classify the leakage status into no leakage, leakage of concern, suspected leakage, and leakage that has occurred based on the data baseline and the initial threshold value; finally, calculate the probability of different leakage statuses. III. Leakage Judgment Criteria Method 2: First, design a rectangular window and set the initial values ​​for the window length and movement step size; then calculate the mean, standard deviation, and normalized mean of the data within the rectangular window; then, classify the leakage status into no leakage, leakage of concern, suspected leakage, and leakage by the number of times the mean and standard deviation increase and the relationship between the normalized mean and the initial threshold; finally, calculate the probability of different leakage statuses. IV. Comprehensive Method: Initialize the weights of baseline methods 1 and 2, and comprehensively analyze the leakage status; V. Threshold and Weight Optimization: Update the thresholds and weights of the two benchmark methods in benchmark method 1 and benchmark method 2 by using real labels and diagnostic labels, and output the final leakage state.

[0006] Step 1: Data acquisition and data benchmark selection calculation; (1) Data acquisition: Use fiber optic sensors to acquire pipeline temperature and humidity signals to obtain a one-dimensional time series of temperature and humidity. , Represents a time series of temperature and humidity. Indicates the first in the sequence (2) Benchmark analysis: The average value of the temperature and humidity data at the two sampling frequency lengths before the initial working condition without leakage is used as the data benchmark. .

[0007] Step 2: Pipeline valve leakage status analysis based on linear regression analysis and state classification; (3) Regression analysis: use linear regression to fit temperature and humidity signals. To determine whether the trend is increasing, the slope is calculated through regression. The formula is as follows: In the formula: —Regression coefficients, reflecting temperature and humidity signals Over time The rate of change, if This indicates that the data shows an increasing trend. This indicates that the data shows a downward trend. b 2 Indicates the intercept; Based on dimensional data benchmarks Define the state category using the following expression: In the formula: State 1—No leakage; State 2—Leakage of concern; State 3—Suspected leakage; State 4—Leakage has occurred; —Threshold 1; —Threshold 2, and ; (5) Calculate the probability of different states, as shown in the following expression: In the formula: —Different state probabilities, when This represents the probability of a leak-free state. This indicates a focus on the probability of a leakage state. This indicates the probability of a suspected leak. Indicates the probability of a leakage occurring; —Different state data lengths, Indicates the length of the leak-free state data. This indicates a concern about the length of the leaked status data. Indicates the length of the suspected leak status data. Indicates the length of the data indicating the leakage status; —Total data length.

[0008] The initial value is set to 5. The initial value is set to 10.

[0009] Step 3: Analysis of pipeline valve leakage status based on dynamic threshold of window function; (6) Design a rectangular window to move on the time domain signal, with a window length of With sampling frequency Related, its domain is The domain is reset according to the actual situation; its movement step size As the window slides along the sequence, the starting position of each window moves forward by one step. Therefore, the first Starting index of the window satisfy: In the formula: —No. One window; Calculate the mean and standard deviation of the signal within the rectangular window, and normalize the mean. For each sliding window, start from the initial position. Initially, the data in the window is Mean within the window Calculated using the following formula: In the formula: —The starting index of the window, —Window length; Window mean normalization results The calculation is as follows: In the formula: —The normalized mean within the window is a dimensionless quantity; —Data benchmark, —Maximum range of temperature and humidity sensor; In the time series Standard deviation in each window Calculated using the following formula: In the formula: —In the time series Standard deviation in each window; —The starting index of the window; —Window length; —Normalized mean within the window.

[0010] (8) State analysis: When the mean and standard deviation increase for 5 consecutive times, but the 5th time within the window When the mean and standard deviation increase for 5 consecutive times, and the 5th increase occurs within the window, it indicates that leakage needs to be monitored; When the mean and standard deviation increase for 5 consecutive times and the 5th increase occurs within the window, a suspected leak is detected; An alarm will sound if a leak has occurred; otherwise, it will indicate no leak. Simultaneously, if the mean and standard deviation increase 10 times consecutively, but the 10th increase occurs within the specified window... When the mean and standard deviation increase for 10 consecutive times and the 10th increase occurs within the window, it indicates that leakage needs to be monitored; When the mean and standard deviation increase for 10 consecutive times and the 10th increase occurs within the window, a suspected leak is detected; An alarm will sound if a leak has occurred; otherwise, no leak will be detected. The above means that if the mean and standard deviation within the window increase consecutively in 5 or 10 cycles and meet the following conditions, they will be classified into different types: In the formula: State 1—No leakage; State 2—Leakage of concern; State 3—Suspected leakage; State 4—Leakage has occurred; —Threshold 3; —Threshold 4, —Threshold 5, and .

[0011] The initial value is set to 0.2. The initial value is set to 0.4. The initial value is set to 0.6.

[0012] (10) Window length adjustment and probability calculation: If the final judgment indicates no leakage, the baseline data remains unchanged. The subsequent data are calculated using the data within the 10th window as the starting value. The mean, standard deviation and normalized mean of the data within the window are recalculated. The leakage state classification and the judgment process of leakage state probability statistics are performed. The probability of each type of leakage state is calculated by the following formula: In the formula: —Probability of occurrence of various leakage conditions; —Number of occurrences of various leakage conditions; —Total number of events related to all leaked states ; (11) Comprehensive judgment: The probability of the final leakage state is calculated using the following formula: In the formula: —The probability of occurrence of various leakage states determined by a combination of the two methods; —The weight for Method 1 is set to an initial value of 0.5; —The probability of occurrence of various leakage states determined by Method 1; —The weights for Method 2 are initially set to 0.5; —The probability of occurrence of various leakage states determined by Method 2.

[0013] Step 4: Weight Optimization: Adaptively adjust the weights of the leakage states obtained from different methods to comprehensively analyze the pipeline state, and provide feedback adjustments to the thresholds of different methods; assuming that in actual use... The next alarm will have Each sample has a true label of TrueLabels= The leak status diagnostic labels for each method are as follows: ,in m Indicates the method number, and the accuracy of each method. The calculation formula is as follows: In the formula: -method The accuracy rate; —Total number of samples; —No. The true label of each sample; -method For the first Prediction for a single sample; —Indicator function, when A value of 1 indicates a correct prediction; otherwise, a value of 0 indicates an incorrect prediction. Comparative analysis is performed on the actual and diagnostic labels each time. When the actual and diagnostic labels match, i.e., the leakage status diagnosis is correct, the threshold remains unchanged. When the actual and diagnostic labels do not match, i.e., the leakage status diagnosis is incorrect, two scenarios are considered: 1) If the leakage severity of the actual label is less than that of the diagnostic label (e.g., the actual label indicates no leakage while the diagnostic label indicates leakage), then in Method 1, the threshold values ​​of threshold1 and threshold2 are incremented by 1 until the threshold value at which the diagnostic label matches the actual label is updated. In Method 2, the threshold values ​​of threshold3, threshold4, and threshold5 are incremented by 0.1 until the threshold value at which the diagnostic label matches the actual label is updated. 2) If the leakage severity of the actual label is greater than that of the diagnostic label (e.g., the actual label indicates leakage while the diagnostic label indicates no leakage), then in Method 1, the threshold values ​​of threshold1 and threshold2 are decremented by 1 until the threshold value at which the diagnostic label matches the actual label is updated. In Method 2, the threshold values ​​of threshold3, threshold4, and threshold5 are decremented by 0.1 until the threshold value at which the diagnostic label matches the actual label is updated.

[0014] (13) Calculation of normalized weights: weights It is the ratio of the accuracy of this method to the accuracy of all methods, calculated using the following formula: In the formula: -method Normalized weights; -method The accuracy rate; —The sum of the accuracies of all methods. This method provides two methods for determining leakage status, and the updated weights of the two methods are: .

[0015] The beneficial effects achieved by this invention are as follows: (1) Two leakage status judgment benchmark methods were constructed, which combined the advantages of dimensional and dimensionless indicators. At the same time, the moving rectangular window and adjustable window length were used for comprehensive analysis to effectively reduce the misdiagnosis rate caused by random fluctuations. (2) Compared with existing fixed threshold method models, this invention combines the advantages of two benchmark methods and uses on-site maintenance data to compare and analyze diagnostic labels and real labels to adaptively optimize the threshold and weight parameters of the benchmark method. This method is based on on-site data and maintenance results to adaptively adjust the method model, which is convenient for use and promotion in different equipment. (3) By using probability to classify the leakage status into "no leakage", "leakage of concern", "suspected leakage" and "leakage occurred", effective maintenance guidance can be provided to on-site personnel. Attached Figure Description

[0016] Figure 1 Flowchart of an adaptive dynamic threshold monitoring method for temperature and humidity signals in pipeline valve leakage; Figure 2 This is a schematic diagram of state classification for baseline method 1; Figure 3 A schematic diagram showing the relationship between the rectangular window and the movement position of real-time monitoring data in baseline method 2; Figure 4 This is a schematic diagram of the dynamic mean curve and standard deviation curve of the monitoring data of the baseline method 2 within a rectangular window. Detailed Implementation

[0017] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0018] An adaptive dynamic threshold monitoring method for temperature and humidity signals in pipeline valve leakage is proposed, comprising the following steps: (1) Data acquisition, and calculation of data baseline based on leak-free data; (2) Leakage judgment benchmark method 1: First, the data trend is determined by linear regression; then, the initial threshold value is set, and the leakage status is divided into "no leakage", "leakage of concern", "suspected leakage" and "leakage occurred" by the data benchmark and the initial threshold value; finally, the probability of different leakage status is calculated. (3) Leakage Judgment Criterion Method 2: First, design a rectangular window and set the initial values ​​of the window length and movement step size; then, calculate the mean, standard deviation and normalized mean of the data in the rectangular window; then, divide the leakage status into "no leakage", "leakage of concern", "suspected leakage" and "leakage occurred" by the relationship between the growth number of the mean and standard deviation and the normalized mean and the initial threshold; finally, calculate the probability of different leakage statuses. (4) Comprehensive method: Initialize the weights of baseline methods 1 and 2, and comprehensively analyze the leakage status; (5) Threshold and weight optimization: Update the thresholds in benchmark method 1 in step 2 and benchmark method 2 in step 3 and the weights of the two benchmark methods in step 4 by using real labels and diagnostic labels, and output the final leakage status.

[0019] The key points and areas to be protected in this invention are: (1) This invention proposes an adaptive dynamic threshold analysis and alarm method for one-dimensional time series of temperature and humidity collected by sensors such as optical fibers; (2) This invention proposes a pipeline valve leakage status analysis method based on linear regression analysis and status classification, and a pipeline valve leakage status analysis method based on window function dynamic threshold. The former designs a dimensional index based on benchmark data, while the latter designs a dimensionless index based on normalized mean. The leakage status is divided into "leakage of concern", "suspected leakage" and "leakage occurred" through probability. Because data changes caused by random fluctuations do not easily change the overall probability distribution of the data, the above two methods can effectively avoid false alarms caused by small random fluctuations in numerical values ​​due to on-site interference. On the other hand, the pipeline valve leakage status analysis method based on window function dynamic threshold further reduces the false alarm rate of leakage status by combining the leakage status probability with rectangular windows of different window lengths and the number of increases.

[0020] (3) Based on the pipeline valve leakage status analysis of linear regression analysis and state classification and the pipeline valve leakage status analysis benchmark method of window function dynamic threshold, this invention proposes an adaptive dynamic threshold monitoring method for pipeline valve leakage temperature and humidity signals. This method comprehensively utilizes the pipeline valve leakage status judgment probability of the two benchmark methods and iteratively updates the thresholds of the two benchmark methods through weight optimization, thereby adaptively realizing the monitoring of leakage status of different types of pipeline valves, which is easy to promote and apply the monitoring method and strategy.

[0021] This invention addresses the problem that fixed thresholds in current nuclear equipment pipeline valve monitoring processes can easily lead to false alarms, and proposes an adaptive dynamic threshold monitoring method for temperature and humidity signals of nuclear equipment pipeline valve leakage.

[0022] like Figure 1 As shown, this method constructs two benchmark methods and combines these two methods to effectively characterize and judge the status of pipeline valves by adaptively extracting dynamic thresholds from real-time collected temperature and humidity data, providing support for monitoring pipeline valve leaks in nuclear equipment. The specific implementation steps are as follows: Step 1: Data acquisition and data benchmark selection and calculation; (1) Data acquisition: Use optical fiber and other sensors to acquire humidity / temperature signals in the pipeline to obtain a one-dimensional time series of temperature and humidity. ( Represents a time series of temperature and humidity. Indicates the first in the sequence Temperature and humidity measurements at various time points); (2) Benchmark analysis: The average value of temperature and humidity data at the two sampling frequency lengths before the initial operating condition without leakage is used as the data benchmark. ; Step 2: Pipeline valve leakage status analysis based on linear regression analysis and state classification; (3) Regression analysis: Linear regression was used to fit the temperature and humidity signals. To determine if the trend is increasing, the slope is calculated using regression. The formula is as follows: In the formula: —Regression coefficients (slopes) reflect temperature and humidity signals Over time The rate of change. If This indicates that the data shows an increasing trend. This indicates that the data shows a downward trend. b 2 This represents the intercept.

[0023] (4) such as Figure 2 As shown, based on dimensional data benchmarks Define the state category; the specific expression is as follows: In the formula: State 1 – No leakage; State 2 – Leakage of concern; State 3 – Suspected leakage; State 4 – Leakage has occurred; —Threshold 1; —Threshold 2, and In this method, The initial value is set to 5. The initial value is set to 10, and its update value is determined iteratively based on the comparison between the diagnostic label and the real label in step four.

[0024] (5) Calculate the probability of different states, the specific expression is as follows: In the formula: —Probabilities of different states, when This represents the probability of a leak-free state. This indicates a focus on the probability of a leakage state. This indicates the probability of a suspected leak. Indicates the probability of a leakage occurring; —Data length for different states, Indicates the length of the leak-free state data. This indicates a concern about the length of the leaked status data. Indicates the length of the suspected leak status data. Indicates the length of the data indicating the leakage status; —Total length of data.

[0025] Step 3: Analysis of pipeline valve leakage status based on window function dynamic threshold; (6) For example Figure 3 As shown, a rectangular window is designed to move across the signal in the time domain, and its window length is... With sampling frequency Related, its domain is This domain can be reset according to the actual situation; its movement step size As the window slides along the sequence, the starting position of each window moves forward by one step. Therefore, the first Starting index of the window satisfy: In the formula: ——No. A window.

[0026] (7) Calculate the mean and standard deviation of the signal in the rectangular window, and normalize the mean.

[0027] like Figure 4 As shown, for each sliding window, starting from the initial position Initially, the data in the window is Mean within the window It can be calculated using the following formula: In the formula: —The starting index of the window, — Window length.

[0028] Window mean normalization results The calculation is as follows: In the formula: —The normalized mean within the window is a dimensionless quantity; —Data benchmark, —Maximum range of temperature and humidity sensor.

[0029] In the time series Standard deviation in each window It can be calculated using the following formula: In the formula: —In the time series Standard deviation in each window; —The starting index of the window; —Window length; — Normalized mean within the window.

[0030] (8) State analysis: When the mean and standard deviation increase for 5 consecutive times, but the 5th time within the window When the mean and standard deviation increase for 5 consecutive times, and the 5th increase occurs within the window, a "leakage monitoring" message is displayed; When the mean and standard deviation increase for 5 consecutive times, and the 5th increase occurs within the window, a "suspected leak" message is displayed; If the mean and standard deviation increase 10 times consecutively, but the 10th increase does not occur within the specified timeframe, an alarm will be triggered indicating a leak; otherwise, a message will be displayed indicating no leak. Additionally, if the mean and standard deviation increase 10 times consecutively, but the 10th increase does not occur within the specified timeframe, an alarm will be triggered. When the mean and standard deviation increase for 10 consecutive times and the 10th increase occurs within the window, a "leakage monitoring" message is displayed; When the mean and standard deviation increase for 10 consecutive times and the 10th increase occurs within the window, a "suspected leak" message is displayed; If the system detects a leak, it will alert the user; otherwise, it will display "No leak". The initial value is 0.2. The initial value is 0.4. The initial value is 0.6. The above can be expressed as follows: if the mean and standard deviation within the window satisfy the following conditions in 5 or 10 consecutive increases, then they are classified into different types: In the formula: State 1 – No leakage; State 2 – Leakage of concern; State 3 – Suspected leakage; State 4 – Leakage has occurred; —Threshold 3; —Threshold 4, —Threshold 5, and In this method, The initial value is set to 0.2. The initial value is set to 0.4. The initial value is set to 0.6, and its update value is determined iteratively based on the comparison between the diagnostic label and the real label in step four.

[0031] (10) Window length adjustment and probability calculation: Adjust the window length according to (6), and calculate according to the process of (7)-(9) to calculate the probability of "no leakage", "leakage of concern", "suspected leakage" and "leakage occurred" and give the final judgment. If the final judgment indicates "no leakage", the baseline data remains unchanged, and the subsequent data are calculated with the data in the 10th window as the starting value. The mean, standard deviation and normalized mean of the data in the window are re-executed, and the judgment process of leakage status classification and leakage status probability statistics is performed. The probability of each type of leakage status can be calculated by the following formula: In the formula: —Probability of various leakage conditions occurring; —Number of occurrences of various leakage conditions; —Total number of events in all leak states .

[0032] (11) Comprehensive judgment: Based on the comprehensive analysis and judgment of the above two methods, the probability of the final leakage state can be calculated by the following formula: In the formula: —The probability of occurrence of various leakage states determined by a combination of the two methods; —The weight for Method 1 can be initially set to 0.5; —The probability of occurrence of various leakage states determined by Method 1; —The weight for Method 2 can be initially set to 0.5; — The probability of occurrence of various leakage states determined by Method 2. Step 4: Weight Optimization: Adaptive weight adjustment is used to comprehensively analyze the pipeline status based on the leakage status obtained from different methods, and the thresholds of different methods are adjusted accordingly to make the model more accurate.

[0033] (12) Assuming that during actual use, The next alarm will have Each sample has a true label of TrueLabels= The leak status diagnostic labels for each method are as follows: , where m represents the method number. The accuracy of each method. The calculation formula is as follows: In the formula: --method The accuracy rate; —Total number of samples; ——No. The true label of each sample; --method For the first Prediction for a single sample; —Indicator function, when The value is 1 if the prediction is correct, and 0 otherwise if the prediction is incorrect.

[0034] Comparative analysis is performed on the real label and the diagnostic label each time. When the real label and the diagnostic label are consistent, that is, the leakage status diagnosis is correct, the threshold does not change. When the real label and the diagnostic label are inconsistent, that is, the leakage status diagnosis is incorrect, there are two cases: 1) The leakage severity of the real label is less than the leakage severity of the diagnostic label. For example, if the real label is no leakage and the diagnostic label is leakage, then in Method 1, the threshold values ​​of threshold1 and threshold2 are gradually increased by 1 until the threshold value when the diagnostic label and the real label are consistent is the updated value; in Method 2, the values ​​of threshold3, threshold4, and threshold5 are gradually increased by 0.1 until the threshold value when the diagnostic label and the real label are consistent is the updated value. 2) If the severity of leakage in the real label is greater than that in the diagnostic label, and the real label shows no leakage while the diagnostic label shows no leakage, then in Method 1, the threshold values ​​of threshold1 and threshold2 are gradually reduced by 1 until the threshold value at which the diagnostic label matches the real label is updated. In Method 2, the threshold values ​​of threshold3, threshold4, and threshold5 are gradually reduced by 0.1 until the threshold value at which the diagnostic label matches the real label is updated.

[0035] (13) Calculation of normalized weights: After obtaining the accuracy, a weight can be assigned to each method. It is the ratio of the accuracy of this method to the accuracy of all methods, calculated using the following formula: In the formula: --method Normalized weights; --method The accuracy rate; —The sum of the accuracies of all methods. This design provides two methods for determining leakage status, and the updated weights of the two methods are: .

Claims

1. A method for adaptive dynamic threshold monitoring of temperature and humidity signals in pipeline valve leakage, characterized in that:

1. Data acquisition, and calculation of data baseline based on leak-free data; II. Leakage Judgment Baseline Method 1: First, determine the data trend through linear regression; then, set an initial threshold value, and classify the leakage status into no leakage, leakage of concern, suspected leakage, and leakage that has occurred based on the data baseline and the initial threshold value; finally, calculate the probability of different leakage statuses. III. Leakage Judgment Criteria Method 2: First, design a rectangular window and set the initial values ​​for the window length and movement step size; then calculate the mean, standard deviation, and normalized mean of the data within the rectangular window; then, classify the leakage status into no leakage, leakage of concern, suspected leakage, and leakage by the number of times the mean and standard deviation increase and the relationship between the normalized mean and the initial threshold; finally, calculate the probability of different leakage statuses. IV. Comprehensive Method: Initialize the weights of baseline methods 1 and 2, and comprehensively analyze the leakage status; Weight optimization: Adaptive weight adjustment is performed on the leakage states obtained by different methods to comprehensively analyze the pipeline state, and feedback adjustment is made to the thresholds of different methods; assuming that in actual use... The next alarm will have Each sample has a true label of TrueLabels= The leak status diagnostic labels for each method are as follows: ,in m Indicates the method number, and the accuracy of each method. The calculation formula is as follows: In the formula: -method Accuracy; —Total number of samples; —No. The true label of each sample; -method For the Prediction for a single sample; —Indicator function, when A value of 1 indicates a correct prediction; otherwise, a value of 0 indicates an incorrect prediction. Comparative analysis is performed on the actual and diagnostic labels each time. When the actual and diagnostic labels match, indicating a correct leak diagnosis, the threshold remains unchanged. When the actual and diagnostic labels do not match, indicating an incorrect leak diagnosis, two scenarios are considered: 1) If the actual label's leak severity is less than the diagnostic label's (e.g., the actual label indicates no leak, but the diagnostic label indicates a leak), then in Method 1, thresholds threshold1 and threshold2 are incremented by 1 until the threshold value at which the diagnostic label matches the actual label is updated. In Method 2, thresholds threshold3, threshold4, and threshold5 are incremented by 0.1 until the threshold value at which the diagnostic label matches the actual label is updated. 2) If the actual label's leak severity is greater than the diagnostic label's (e.g., the actual label indicates a leak, but the diagnostic label indicates no leak), then in Method 1, thresholds threshold1 and threshold2 are decremented by 1 until the threshold value at which the diagnostic label matches the actual label is updated. In Method 2, thresholds threshold3, threshold4, and threshold5 are decremented by 0.1 until the threshold value at which the diagnostic label matches the actual label is updated. V. Threshold and Weight Optimization: Update the thresholds and weights of the two benchmark methods in benchmark method 1 and benchmark method 2 by using real labels and diagnostic labels, and output the final leakage state.

2. The adaptive dynamic threshold monitoring method for temperature and humidity signals of pipeline valve leakage according to claim 1, characterized in that: Step 1: Data Acquisition and Data Baseline Selection and Calculation; Data Acquisition: Use fiber optic sensors to acquire temperature and humidity signals from the pipeline, obtaining a one-dimensional time series of temperature and humidity. , Represents a time series of temperature and humidity. Indicates the first in the sequence Temperature and humidity measurements at specific time points; benchmark analysis: the average value of temperature and humidity data from the two sampling frequency lengths prior to the initial leak-free operating conditions was used as the data benchmark. .

3. The adaptive dynamic threshold monitoring method for temperature and humidity signals of pipeline valve leakage according to claim 2, characterized in that: Step 2: Pipeline valve leakage status analysis based on linear regression analysis and state classification; Regression analysis: using linear regression to fit temperature and humidity signals. To determine whether the trend is increasing, the slope is calculated through regression. The formula is as follows: In the formula: —Regression coefficients reflect temperature and humidity signals Over time The rate of change, if This indicates that the data shows an increasing trend. This indicates that the data shows a downward trend. b 2 Indicates the intercept; Based on dimensional data benchmarks Define the state category using the following expression: In the formula: State 1—No leakage; State 2—Leakage of concern; State 3—Suspected leakage; State 4—Leakage has occurred; —Threshold 1; —Threshold 2, and ; The probability of different states is calculated using the following expression: In the formula: —Different state probabilities, when This represents the probability of a leak-free state. This indicates a focus on the probability of a leakage state. This indicates the probability of a suspected leak. Indicates the probability of a leakage occurring; —Different state data lengths, Indicates the length of the leak-free state data. This indicates a concern about the length of the leaked status data. Indicates the length of the suspected leak status data. Indicates the length of the data indicating the leakage status; —Total data length.

4. The adaptive dynamic threshold monitoring method for temperature and humidity signals of pipeline valve leakage according to claim 3, characterized in that: The initial value is set to 5. The initial value is set to 10.

5. The adaptive dynamic threshold monitoring method for temperature and humidity signals of pipeline valve leakage according to claim 3, characterized in that: Step 3: Pipeline valve leakage status analysis based on window function dynamic threshold; design a rectangular window that moves across the time domain signal, with a window length of... With sampling frequency Related, its domain is The domain is reset according to the actual situation; its movement step size As the window slides along the sequence, the starting position of each window moves forward by one step. Therefore, the first Starting index of the window satisfy: In the formula: —No. One window; Calculate the mean and standard deviation of the signal within the rectangular window, and normalize the mean. For each sliding window, start from the initial position. Initially, the data in the window is Mean within the window Calculated using the following formula: In the formula: —The starting index of the window, —Window length; Window mean normalization results The calculation is as follows: In the formula: —The normalized mean within the window is a dimensionless quantity; —Data benchmark, —Maximum range of temperature and humidity sensor; In the time series Standard deviation in each window Calculated using the following formula: In the formula: —In the time series Standard deviation in each window; —The starting index of the window; —Window length; —Normalized mean within the window.

6. The adaptive dynamic threshold monitoring method for temperature and humidity signals of pipeline valve leakage according to claim 5, characterized in that: State analysis: When the mean and standard deviation increase five times consecutively, but within the fifth window... When the mean and standard deviation increase for 5 consecutive times, and the 5th increase occurs within the window, it indicates that leakage needs to be monitored; When the mean and standard deviation increase for 5 consecutive times and the 5th increase occurs within the window, a suspected leak is detected; An alarm will sound if a leak has occurred; otherwise, it will indicate no leak. Simultaneously, if the mean and standard deviation increase 10 times consecutively, but the 10th increase occurs within the specified window... When the mean and standard deviation increase for 10 consecutive times and the 10th increase occurs within the window, it indicates that leakage needs to be monitored; When the mean and standard deviation increase for 10 consecutive times and the 10th increase occurs within the window, a suspected leak is detected; An alarm will sound if a leak has occurred; otherwise, no leak will be detected. The above means that if the mean and standard deviation within the window increase consecutively in 5 or 10 cycles and meet the following conditions, they will be classified into different types: In the formula: State 1—No leakage; State 2—Leakage of concern; State 3—Suspected leakage; State 4—Leakage has occurred; —Threshold 3; —Threshold 4, —Threshold 5, and .

7. The adaptive dynamic threshold monitoring method for temperature and humidity signals of pipeline valve leakage according to claim 6, characterized in that: The initial value is set to 0.

2. The initial value is set to 0.

4. The initial value is set to 0.

6.

8. The adaptive dynamic threshold monitoring method for temperature and humidity signals of pipeline valve leakage according to claim 6, characterized in that: Window length adjustment and probability calculation: If the final determination indicates no leakage, the baseline data remains unchanged. The subsequent data is calculated using the data from the 10th window as the starting point. The mean, standard deviation, and normalized mean of the data within the window are recalculated. The leakage state classification and probability statistics judgment process are then performed. The probability of each leakage state is calculated using the following formula: In the formula: —Probability of occurrence of various leakage conditions; —Number of occurrences of various leakage conditions; —Total number of events related to all leaked states ; Overall assessment: The probability of the final leakage state is calculated using the following formula: In the formula: —The probability of occurrence of various leakage states determined by a combination of the two methods; —The weight for Method 1 is set to an initial value of 0.5; —The probability of occurrence of various leakage states determined by Method 1; —The weights for Method 2 are initially set to 0.5; —The probability of occurrence of various leakage states determined by Method 2.

9. The adaptive dynamic threshold monitoring method for temperature and humidity signals of pipeline valve leakage according to claim 1, characterized in that: Calculation of normalized weights: weights It is the ratio of the accuracy of this method to the accuracy of all methods, calculated using the following formula: In the formula: -method Normalized weights; -method Accuracy; —The sum of the accuracies of all methods. This method provides two methods for determining leakage status, and the updated weights of the two methods are: 。

Images