Fuel gas monitoring fault analysis method and system and storage medium

Abstract

The invention relates to the technical field of data processing, and discloses a fuel gas monitoring fault analysis method and system and a storage medium. The method comprises the following steps: acquiring multi-dimensional data through a sensor group, and transmitting the multi-dimensional data to a processing unit for pre-screening through an Internet of Things protocol; carrying out filtering, feature calculation and dimension reduction processing on the data to obtain a feature vector; establishing a fault database and generating a mapping relation; time sequence and space data are fused to calculate a fault judgment value; an early warning result is obtained by combining sensor state data evaluation; historical data is retrieved, similarity is calculated, and a disposal instruction is generated. According to the method, the accuracy of gas pipe network fault diagnosis is improved through fusion analysis of multi-source data, and early recognition and early warning of faults are achieved.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to a gas monitoring fault analysis method, system and storage medium. Background Art

[0002] At present, a single sensor or a simple combination of multiple sensors is generally used for data collection in gas pipeline monitoring, focusing on the monitoring of basic parameters such as gas concentration and pressure. These monitoring systems usually use fixed threshold judgment methods for fault detection, and trigger alarms when the monitoring parameters exceed the preset threshold. In terms of data processing, most of them use simple filtering and statistical analysis methods to perform basic processing and storage on the collected data. At the same time, the existing fault analysis methods mainly rely on manual experience, and fault diagnosis and early warning are performed by comparing historical data and expert knowledge. The formulation of disposal plans is also mainly based on manual judgment.

[0003] However, the existing gas monitoring fault analysis methods have the following shortcomings: a single data collection method is difficult to fully reflect the operating status of the gas pipeline network; the fixed threshold judgment method is prone to false alarms and missed alarms; simple data processing methods cannot fully mine the fault feature information contained in the data; it is highly dependent on manual experience and it is difficult to ensure the objectivity and accuracy of the analysis results; it lacks the ability to predict the development trend of faults and it is difficult to achieve early warning; the formulation of disposal plans lacks systematicity and pertinence, which affects the efficiency and effectiveness of fault disposal. Summary of the invention

[0004] The present application provides a gas monitoring fault analysis method, system and storage medium, which are used to improve the accuracy of gas pipeline network fault diagnosis through fusion analysis of multi-source data and realize early identification and warning of faults.

[0005] In a first aspect, the present application provides a gas monitoring fault analysis method, the gas monitoring fault analysis method comprising:

[0006] The gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data and pipeline vibration data are collected through the gas monitoring sensor group, and the data is transmitted to the processing unit according to the Internet of Things communication protocol, and the data is pre-screened to obtain the original gas monitoring data;

[0007] The raw gas monitoring data is calculated according to the median filter formula and outliers are removed, characteristic parameters are calculated for the filtered data, and dimension reduction processing is performed on the calculation results to obtain a gas characteristic vector;

[0008] Establishing a fault type database based on the gas characteristic vector, classifying and calculating the characteristic data, comparing the calculation results with preset parameters, and generating a gas fault mapping relationship;

[0009] Extracting time series data according to the gas feature vector and the gas fault mapping relationship, performing weight fusion calculation on the time series data and the spatial data, and outputting a gas fault determination value;

[0010] The gas fault judgment value is compared with the preset threshold value for classification, weighted calculation is performed on the classification result, and credibility evaluation is performed in combination with the sensor status data to obtain a gas fault warning result;

[0011] Based on the gas fault warning result, historical data is retrieved, similarity calculation and priority sorting are performed on the retrieved data, and a gas fault handling instruction is generated.

[0012] In a second aspect, the present application provides a gas monitoring fault analysis system, the gas monitoring fault analysis system comprising:

[0013] The transmission module is used to collect gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data and pipeline vibration data through the gas monitoring sensor group, transmit the data to the processing unit according to the Internet of Things communication protocol, pre-screen the data, and obtain the original gas monitoring data;

[0014] A dimension reduction module is used to calculate the raw gas monitoring data according to the median filter formula and remove abnormal values, calculate characteristic parameters for the filtered data, and perform dimension reduction processing on the calculation results to obtain a gas feature vector;

[0015] A comparison module, used to establish a fault type database according to the gas characteristic vector, classify and calculate the characteristic data, compare the calculation results with preset parameters, and generate a gas fault mapping relationship;

[0016] A fusion module, used to extract time series data according to the gas feature vector and the gas fault mapping relationship, perform weight fusion calculation on the time series data and the spatial data, and output a gas fault judgment value;

[0017] An evaluation module is used to compare and classify the gas fault judgment value with a preset threshold value, perform weighted calculation on the classification result, and perform credibility evaluation in combination with the sensor status data to obtain a gas fault warning result;

[0018] The sorting module is used to retrieve historical data based on the gas fault warning result, perform similarity calculation and priority sorting on the retrieved data, and generate a gas fault handling instruction.

[0019] A third aspect of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned gas monitoring fault analysis method.

[0020] In the technical solution provided by the present application, the gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data and pipeline vibration data are collected simultaneously through the gas monitoring sensor group, thereby realizing the coordinated monitoring of multi-dimensional data and comprehensively reflecting the operation status of the gas pipeline network. In addition, the Internet of Things communication protocol is used for data transmission to ensure the reliability and real-time performance of data transmission. The pre-screening process effectively reduces the influence of data noise. The median filter formula calculation can effectively remove outliers and improve data quality. The characteristic parameter calculation and dimensionality reduction processing realize the effective compression and feature extraction of data. The establishment of a fault type database enables the fault characteristics to be systematically organized and managed, and the classification calculation of characteristic data The accuracy of fault identification is improved by comparison and analysis. The weighted fusion calculation of time series data and spatial data makes full use of the spatiotemporal correlation of fault characteristics, making fault judgment more comprehensive and accurate. The hierarchical comparison and weighted calculation of preset thresholds realize the accurate division of fault levels. The credibility assessment combined with sensor status data ensures the reliability of early warning results. By retrieving historical data and performing similarity calculation, the most similar historical cases are found, which provides a reliable reference for fault handling. Priority sorting ensures the optimal selection of handling solutions, thus forming a complete, automated and intelligent gas monitoring fault analysis process, which significantly improves the accuracy of gas pipeline fault diagnosis and the efficiency of handling. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0022] Figure 1 A schematic diagram of an embodiment of a method for analyzing a gas monitoring fault in an embodiment of the present application;

[0023] Figure 2 This is a schematic diagram of an embodiment of a gas monitoring fault analysis system in an embodiment of the present application. DETAILED DESCRIPTION

[0024] The embodiments of the present application provide a method, system and storage medium for analyzing gas monitoring faults. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0025] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 , an embodiment of the gas monitoring fault analysis method in the embodiment of the present application includes:

[0026] Step S101: Collect gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data and pipeline vibration data through a gas monitoring sensor group, transmit the data to a processing unit according to an Internet of Things communication protocol, pre-screen the data, and obtain gas monitoring raw data;

[0027] Step S102: Calculate the original gas monitoring data according to the median filter formula and remove outliers, calculate characteristic parameters for the filtered data, perform dimensionality reduction processing on the calculation results, and obtain a gas characteristic vector;

[0028] Step S103: Establish a fault type database based on the gas characteristic vector, classify and calculate the characteristic data, compare the calculation results with the preset parameters, and generate a gas fault mapping relationship;

[0029] Step S104: extracting time series data according to the gas feature vector and the gas fault mapping relationship, performing weight fusion calculation on the time series data and the spatial data, and outputting a gas fault determination value;

[0030] Step S105: Compare and grade the gas fault judgment value with the preset threshold, perform weighted calculation on the graded result, perform credibility assessment in combination with the sensor status data, and obtain a gas fault warning result;

[0031] Step S106: Retrieve historical data based on the gas fault warning result, perform similarity calculation and priority sorting on the retrieved data, and generate a gas fault handling instruction.

[0032] It is understandable that the execution subject of the present application may be a gas monitoring fault analysis system, or a terminal or a server, which is not limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0033] Specifically, the gas monitoring sensor group includes a gas concentration sensor, a pressure sensor, a temperature sensor, an acoustic sensor, and an optical fiber vibration sensor, which are respectively used to collect different types of monitoring data. The gas concentration sensor uses a metal oxide semiconductor sensor to measure the gas concentration by detecting the change in conductivity generated when the gas molecules undergo an oxidation-reduction reaction with the sensor surface, with a sampling frequency of 100ms. The pressure sensor uses a piezoresistive sensor to measure the pipeline pressure by using the change in resistance value under pressure, with a sampling frequency of 200ms. The temperature sensor uses a thermocouple to measure the gas temperature through the thermoelectric electromotive force, with a sampling frequency of 500ms. The acoustic sensor uses a piezoelectric sensor to convert the acoustic wave vibration into an electrical signal to measure the acoustic characteristics of the leakage, with a sampling frequency of 50ms. The optical fiber vibration sensor is based on the principle of fiber Bragg grating to measure the vibration characteristics of the pipeline network, with a sampling frequency of 20ms. The collected data is transmitted through the Internet of Things communication protocol MQTT, which adopts a publish / subscribe mode and has the characteristics of low bandwidth occupancy and high reliability. After the data is transmitted to the processing unit, it is first pre-screened. Set a range threshold of 0-100ppm for gas concentration data, 0-1.6MPa for pressure data, -20℃-60℃ for temperature data, 30-120dB for acoustic data, and 0-5mm for vibration data. Abnormal data is eliminated through range check, and the data is time-aligned, data packets are constructed in units of 1s, and check bits are added to ensure data integrity.

[0034] The pre-screened raw gas monitoring data is processed by median filtering. A sliding time window is used for each type of data to calculate the median and interquartile range of the data in the window, and outliers are determined and eliminated based on the 3-times interval of the interquartile range. The filtered data is subjected to characteristic parameter calculation, including: calculation of mean, standard deviation, kurtosis, and skewness for gas concentration data; calculation of maximum fluctuation amplitude and fluctuation period for pressure data; calculation of rate of change and temperature gradient for temperature data; calculation of sound pressure level and frequency distribution for acoustic data; and calculation of vibration amplitude and vibration frequency for vibration data. The calculation results are mapped to a uniform scale through data normalization, and then the dimension is reduced through principal component analysis, retaining the feature dimensions with a cumulative variance contribution rate of 95%, and finally a linear combination is performed based on the feature importance weights to generate a gas feature vector.

[0035] A fault type database is constructed based on the generated gas feature vector. First, the feature vector is divided into 24-hour time windows, and the statistical features in each window are calculated. The feature data is grouped according to the fault type, and the mean center and variance distribution of each group are calculated. The feature data is segmented and the probability distribution characteristics of different fault types in each feature dimension are calculated. The calculation results are compared with the historical fault case data, the feature similarity coefficient is calculated, and the mapping relationship between the feature and the fault type is established. Time series-spatial data fusion analysis is performed based on the gas feature vector and the fault mapping relationship. The feature vector is divided into three time scales: hour, day, and week, and the change trend is calculated respectively to extract the time series features. At the same time, the spatial correlation matrix is ​​constructed using the sensor location information to analyze the propagation characteristics of the fault features in space. After standardizing the time series features and spatial features, the weight coefficients of the two types of features are calculated, and weighted fusion is performed to output the fault judgment value in the range of 0-1.

[0036] Compare and grade the fault judgment value with the preset four threshold nodes of 0.3, 0.5, 0.7, and 0.9. Calculate the frequency and duration of the grading results, and weight them with weights of 0.4 and 0.6. Evaluate the reliability by combining the three indicators of signal strength, response time, and data integrity rate in the sensor status data. Combine the weighted results with the reliability score to calculate the confidence of the fault warning and form a graded warning result. Finally, retrieve historical data based on the warning results. Extract the characteristics of the fault type, degree, location, etc. to establish the retrieval conditions and match them with the historical case library. Calculate the Euclidean distance between the feature parameters of the matched historical cases and evaluate the case similarity. Select cases with a similarity greater than 0.8 and extract their disposal plans. Score and sort the plans according to the three indicators of disposal time, resource consumption, and repair effect, and select the best plan as the disposal instruction output.

[0037] For example: A gas pipeline monitoring point detected that the gas concentration increased from 5ppm to 45ppm within 10 minutes, the pressure decreased from 1.2MPa to 0.9MPa, and the temperature increased by 2°C. At the same time, the acoustic sensor detected a continuous noise of 50-60dB, and the vibration sensor showed an amplitude of 0.8mm. After pre-screening and filtering, the extracted characteristic parameters of this set of data showed typical trace leakage characteristics. After time-space analysis, it was found that the adjacent monitoring points also showed slight abnormalities, and the fault judgment value was 0.65. Combined with the sensor status evaluation, it was determined to be a medium-level warning. The system matched 3 similar historical cases, among which the solution with the shortest disposal time and the lowest resource consumption was selected as the optimal disposal instruction.

[0038] In the embodiment of the present application, the gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data and pipeline vibration data are collected simultaneously through the gas monitoring sensor group, thereby realizing the coordinated monitoring of multi-dimensional data and comprehensively reflecting the operation status of the gas pipeline network. In addition, the Internet of Things communication protocol is used for data transmission to ensure the reliability and real-time performance of data transmission. The pre-screening process effectively reduces the influence of data noise. The median filter formula calculation can effectively remove outliers and improve data quality. The characteristic parameter calculation and dimensionality reduction processing realize effective data compression and feature extraction. The establishment of a fault type database enables the fault characteristics to be systematically organized and managed, and the classification calculation and comparison of characteristic data are carried out. The analysis improves the accuracy of fault identification. The weighted fusion calculation of time series data and spatial data makes full use of the spatiotemporal correlation of fault characteristics, making fault judgment more comprehensive and accurate. The hierarchical comparison and weighted calculation of preset thresholds realize the accurate division of fault levels. The credibility assessment combined with sensor status data ensures the reliability of the early warning results. By retrieving historical data and performing similarity calculations, the most similar historical cases are found, providing a reliable reference for fault handling. Priority sorting ensures the optimal selection of handling plans, thus forming a complete, automated and intelligent gas monitoring fault analysis process, which significantly improves the accuracy of gas pipeline fault diagnosis and the efficiency of handling.

[0039] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0040] (1) The gas concentration data is collected at a sampling interval of 100 ms, the pipeline pressure data is collected at a sampling interval of 200 ms, the gas temperature data is collected at a sampling interval of 500 ms, the leakage acoustic data is collected at a sampling interval of 50 ms, and the pipeline network vibration data is collected at a sampling interval of 20 ms;

[0041] (2) Perform time stamp alignment on gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data, and pipeline network vibration data, construct data packets in units of 1 second, and add check bits to the data packets according to the IoT communication protocol;

[0042] (3) Pre-screening the numerical range of the gas concentration data in the data packet by setting the concentration range threshold of 0-100ppm, the pipeline pressure data by setting the pressure range threshold of 0-1.6MPa, the gas temperature data by setting the temperature range threshold of -20°C-60°C, the leakage acoustic data by setting the sound pressure range threshold of 30-120dB, and the pipeline network vibration data by setting the amplitude range threshold of 0-5mm;

[0043] (4) The pre-screened gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data, and pipeline network vibration data are correspondingly merged into data records, and the collection time and sensor number are marked for each data record;

[0044] (5) Perform data integrity checks on data records, remove records with missing data, mark abnormal data records, and establish data quality scores;

[0045] (6) Organize the data records that have passed the integrity check into a time series data table, add data source identification, and generate gas monitoring raw data.

[0046] Specifically, the gas concentration data is collected using a metal oxide semiconductor sensor, which works on the principle of using the conductivity change generated when gas molecules undergo redox reactions with the surface of semiconductor materials for detection. The sampling interval is set to 100ms, which can capture small changes in gas concentration in a timely manner. Pipeline pressure data is collected using a piezoresistive pressure sensor, which measures pressure by measuring the resistance change of the strain gauge under pressure. The sampling interval is 200ms, which meets the needs of monitoring pipeline pressure changes. Gas temperature data is collected using a K-type thermocouple, which uses the thermoelectric potential generated by the temperature difference to measure temperature. The sampling interval is 500ms, taking into account the relatively slow temperature change. Leakage acoustic data collection uses a piezoelectric acoustic sensor to convert acoustic wave vibrations into electrical signals. The sampling interval is 50ms to ensure the integrity of the collection of high-frequency acoustic signals. Pipeline network vibration data collection uses a fiber Bragg grating sensor, which measures vibration characteristics based on the periodic changes of the fiber grating. The sampling interval is 20ms, which meets the requirements for detecting small vibrations. Time alignment is performed on the different types of data collected. Since the sampling intervals of various sensors are different, the data needs to be unified to the same time base. The timestamp alignment uses the interpolation method. For data points with a sampling interval greater than 1s, linear interpolation is used to supplement the intermediate value. For data points with a sampling interval less than 1s, the average value within 1s is taken. The data packet is constructed using the MQTT (Message Queuing Telemetry Transport) protocol, which is a lightweight message transmission protocol that uses a publish / subscribe model. Each data packet contains all sensor data within 1s, and a CRC (Cyclic Redundancy Check) check bit is added for data transmission error detection.

[0047] In the data pre-screening stage, reasonable range thresholds are set for various types of data. The threshold range of gas concentration data is 0-100ppm, which is determined based on the common combustible gas leakage concentration limit. The threshold range of pipeline pressure data is 0-1.6MPa, which takes into account the typical working pressure of the urban gas pipeline network. The threshold range of gas temperature data is -20℃-60℃, which covers the temperature change range under different climatic conditions. The threshold range of leakage acoustic data is 30-120dB, which includes the sound pressure level changes from small leaks to serious leaks. The threshold range of pipeline network vibration data is 0-5mm, which is set based on the safety limit of the pipeline structure. Data exceeding the threshold range will be marked as abnormal data for special processing. The data record merging process integrates the pre-screened multidimensional data into a unified data structure. Each data record contains a complete set of sensor data, with timestamp and sensor identification information attached. The timestamp is accurate to milliseconds and records the specific time of data collection. The sensor number uses a unique identification code, which contains sensor type and installation location information, to facilitate subsequent data traceability and analysis.

[0048] Data integrity check ensures the quality of data records. First, check whether there are missing values ​​in the data record. When a certain type of sensor data is completely missing, the record will be eliminated. For records with partial data missing, mark the missing items and record the reasons for the missing. The marking of abnormal data records is based on multiple dimensions, including numerical anomalies (exceeding the threshold range), change anomalies (sudden changes or excessive fluctuations), and association anomalies (inconsistent with other sensor data). The data quality score adopts a weighted method, considering the three aspects of data integrity, accuracy and timeliness. The scoring results are used for subsequent data credibility analysis.

[0049] The data records that have been checked for integrity are organized into a standardized time series data table. Each row of the data table represents a complete monitoring record at a time point, and the columns include timestamps, various sensor data, and data quality scores. The data source identifies the acquisition node and transmission path of the recorded data for data traceability. The organized data table is the original gas monitoring data, which serves as the basic data set for subsequent fault analysis.

[0050] For example: During continuous monitoring, the gas concentration sensor records data every 100ms, and records 100 data points in 10 seconds, showing that the concentration gradually increases from 3ppm to 12ppm. During the same period, the pressure sensor records data every 200ms, records 50 data points, and shows that the pressure decreases from 1.2MPa to 1.1MPa. The temperature sensor records data every 500ms, records 20 data points, and the temperature fluctuates around 28℃. The acoustic sensor records data every 50ms, records 200 data points, and detects a continuous sound pressure of 40-45dB. The vibration sensor records data every 20ms, records 500 data points, and shows a small vibration of 0.2-0.3mm. These data are processed by time alignment and integrated into 10 complete data records in units of 1s. Data pre-screening shows that all data are within the set threshold range, and the data integrity check does not find missing and abnormal records. The generated time series data table fully records the change process of all sensor data within 10s, and the data quality score shows that the data reliability is good. This set of data reflects a typical early leak scenario, with a slow increase in gas concentration accompanied by a slight drop in pressure, accompanied by weak acoustic and vibration signals.

[0051] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0052] (1) The raw gas monitoring data were grouped according to 10-minute time windows, the median value and interquartile range of each group of data were calculated, and outliers were determined and eliminated based on the interval of 3 times the interquartile range to obtain filtered processed data;

[0053] (2) Calculate the mean, standard deviation, kurtosis, and skewness of the gas concentration data in the filtered data, calculate the maximum fluctuation amplitude and fluctuation period of the pipeline pressure data, calculate the change rate and temperature gradient of the gas temperature data, calculate the sound pressure level and frequency distribution of the leakage acoustic data, and calculate the vibration amplitude and vibration frequency of the pipeline network vibration data to generate a feature data set;

[0054] (3) normalizing the mean, standard deviation, kurtosis, and skewness data in the feature data group to the range of 0-1, normalizing the maximum fluctuation amplitude and fluctuation period data to the range of -1-1, linearly transforming the change rate and temperature gradient data to the standard normal distribution, logarithmically transforming the sound pressure level and frequency distribution data, and performing minimum-maximum transformation on the vibration amplitude and vibration frequency data to obtain normalized feature data;

[0055] (4) Sort the normalized feature data by variance contribution rate, select the feature dimensions whose cumulative variance contribution rate reaches 95% to form the main feature data set;

[0056] (5) Calculate the feature importance of each feature data in the main feature data set according to the correlation coefficient matrix, perform linear combination of the feature data based on the importance weight, and construct compressed feature data;

[0057] (6) Convert the compressed feature data into a numerical vector form, add the time index and sensor location identifier, and generate a gas feature vector.

[0058] Specifically, in the data filtering and feature extraction stage, the gas monitoring raw data is first grouped into time windows. 10 minutes is used as a fixed time window to process the data in segments. In each time window, the median value and interquartile range of the data are calculated respectively. The median is calculated by sorting the data in the window by size and taking the value in the middle position, and the interquartile range is obtained by calculating the difference between the upper quartile (75th percentile) and the lower quartile (25th percentile). The outlier judgment limit is set based on the 3 times interval of the interquartile range, that is, the data points that exceed the range of (Q1-3IQR) to (Q3+3IQR) are judged as outliers and eliminated, where Q1 is the lower quartile, Q3 is the upper quartile, and IQR is the interquartile range.

[0059] Multi-dimensional feature extraction is performed on the filtered data. For gas concentration data, mathematical statistical features are calculated, including: mean (reflecting concentration level), standard deviation (reflecting degree of fluctuation), kurtosis (reflecting distribution sharpness), skewness (reflecting the degree of deviation of distribution from symmetry). For pipeline pressure data, the focus is on analyzing dynamic change characteristics, calculating the maximum fluctuation amplitude (difference between the highest pressure and the lowest pressure) and fluctuation period (time interval of pressure fluctuation). The feature extraction of gas temperature data focuses on the change trend, calculating the temperature change rate (temperature change per unit time) and temperature gradient (temperature change in spatial position). Leakage acoustic data analysis includes the calculation of sound pressure level (reflecting sound wave intensity) and frequency distribution (reflecting sound wave spectrum characteristics). Pipeline network vibration data extracts vibration amplitude (maximum value of vibration displacement) and vibration frequency (inverse of vibration period) features. In order to make feature data of different dimensions and ranges comparable, feature normalization processing is required. Statistics such as mean, standard deviation, kurtosis, and skewness are normalized by 0-1 interval, and the original value is converted to the range of [0,1] through linear mapping. The maximum fluctuation amplitude and fluctuation period characteristics are normalized in the [-1,1] interval, which is suitable for describing physical quantities with bidirectional changes. The temperature change rate and temperature gradient data are mapped to data that conform to the standard normal distribution through linear transformation, which is convenient for subsequent statistical analysis. Since the sound pressure level and frequency distribution data have a large value span, logarithmic transformation is used to compress the data range. The vibration amplitude and frequency data are normalized using the minimum-maximum method to maintain the relative proportional relationship of the data.

[0060] The normalized feature data is subjected to dimensionality reduction, mainly using the principal component analysis (PCA) method. The features are ranked by calculating their variance contribution rate, and the variance contribution rate calculation formula is:

[0061]

[0062] Among them: VCR i is the variance contribution rate of the i-th feature; i is the eigenvalue of the i-th feature; n is the total number of features; k is the feature number.

[0063] The feature dimensions with a cumulative variance contribution rate of 95% are retained to form the main feature data set. During the selection of the main feature data set, the feature importance is evaluated by calculating the correlation coefficient matrix between the features. Each element of the correlation coefficient matrix represents the degree of correlation between two features, and the importance weight of each feature is calculated based on this matrix. The importance weight calculation takes into account the correlation between the feature and other features and the variance contribution of the feature itself. The obtained feature importance weights are standardized so that the sum of the weights is 1, and then these weights are applied to the linear combination of the feature data to construct the compressed feature data.

[0064] The compressed feature data finally needs to be converted into a standard numerical vector form. Each vector contains the feature value after dimension reduction and weight combination, and also adds the time index (accurate to seconds) and sensor location identifier (including installation point coordinate information). This vector form facilitates subsequent fault mapping and diagnostic analysis.

[0065] For example, at a gas network monitoring point, 10 minutes of monitoring data are processed. The original data contains 600 sampling points (sampling frequency 1Hz), and the outliers are first determined by calculating the median and interquartile range. Taking the gas concentration data as an example, the median is 15ppm, the upper quartile is 20ppm, the lower quartile is 10ppm, and the interquartile range is 10ppm. Based on this, the outlier determination interval is determined to be [-20ppm, 50ppm], and data points outside this range are removed.

[0066] The filtered data was feature extracted, and the gas concentration data was calculated to have a mean of 16ppm, a standard deviation of 5ppm, a kurtosis of 2.1, and a skewness of 0.3. The pipeline pressure data showed a maximum fluctuation amplitude of 0.2MPa, and a fluctuation period of about 120 seconds. The rate of change of the temperature data was 0.1℃ / min, and the temperature gradient was 0.05℃ / m. The sound pressure level of the acoustic data was 45dB, and the frequency was mainly distributed in the range of 200-500Hz. The amplitude of the vibration data was 0.3mm, and the main frequency was 2Hz. Through normalization, these features were converted to a unified numerical scale. The statistical features of gas concentration were mapped to the [0,1] interval, the pressure fluctuation features were mapped to the [-1,1] interval, the temperature features were adjusted to a distribution with a mean of 0 and a standard deviation of 1 through linear transformation, the acoustic features were compressed through logarithmic transformation, and the vibration features were maintained in relative proportion using minimum-maximum transformation.

[0067] After variance contribution analysis, it was found that among all the features, the cumulative variance contribution rate of the four features of gas concentration mean, pressure fluctuation amplitude, sound pressure level and vibration frequency reached 95%, so these features were retained as the main features. Further analysis of the correlation coefficient matrix showed that the gas concentration showed a strong negative correlation with the pressure fluctuation, and the sound pressure level showed a weak positive correlation with the vibration frequency. Based on this, the importance weights of the four features were calculated to be 0.35, 0.3, 0.2 and 0.15 respectively. These features are linearly combined according to the weights and converted into a four-dimensional feature vector. At the same time, the timestamp (such as 2024-01-0110:00:00) and the sensor location identifier (such as X=100m, Y=50m, Z=2m) are added to form a gas feature vector. It not only retains the most important information in the original data, but also greatly reduces the data dimension.

[0068] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0069] (1) Divide the gas feature vector into 24-hour data segments according to the timestamp, calculate the statistical features of the gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data and pipeline network vibration data in each data segment, and generate a feature statistical table;

[0070] (2) Group the data in the feature statistics table according to the fault type label, calculate the mean center and variance distribution of each group of data, and form a fault feature statistics matrix;

[0071] (3) The data in the fault feature statistical matrix are segmented according to the gas concentration range of 0-100 ppm, the pressure fluctuation range of 0-1.6 MPa, the temperature change range of -20°C-60°C, the sound pressure level of 30-120 dB, and the vibration amplitude of 0-5 mm, and segmented statistical data are constructed;

[0072] (4) Calculate the probability density of the segmented statistical data to obtain the characteristic distribution curve of each fault type and generate a fault characteristic probability table;

[0073] (5) Compare the fault feature probability table with the historical fault data, calculate the feature similarity coefficient, and establish a fault type database;

[0074] (6) Based on the fault type database, data mapping is performed on the input gas feature vector, the fault type probability value is calculated, and a gas fault mapping relationship is generated.

[0075] Specifically, the gas feature vector is organized into 24-hour data segments according to the time series. The choice of this time scale is based on the daily cycle variation law of gas usage characteristics. For each 24-hour data segment, the gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data and pipeline network vibration data are processed separately. The average value, maximum value, minimum value and change trend of the gas concentration data within 24 hours are calculated; the pressure fluctuation range and fluctuation frequency of the pipeline pressure data are analyzed; the temperature change amplitude and change rate of the gas temperature data are statistically analyzed; the sound pressure characteristics and spectrum characteristics of the leakage acoustic data are extracted; and the vibration intensity and frequency characteristics of the pipeline network vibration data are recorded. These statistical features constitute the basic content of the feature statistical table. On the basis of the feature statistical table, according to historical experience and expert knowledge, the data are grouped by labels according to different fault types. Fault types include: pipeline leakage (micro leakage, medium leakage, large leakage), pressure abnormality (too high pressure, too low pressure, pressure fluctuation), temperature abnormality (too high temperature, too low temperature), pipeline blockage, pipeline corrosion, etc. The mean center is calculated for each set of data to reflect the typical characteristic value of this type of fault; the variance distribution is calculated to reflect the fluctuation range of the characteristic. These statistics constitute the fault characteristic statistical matrix. The data in the fault characteristic statistical matrix needs to be segmented according to the reasonable range of physical quantities. The gas concentration is divided into 5 intervals within the range of 0-100ppm, corresponding to different degrees of leakage; the pressure fluctuation range of 0-1.6MPa is divided into 4 intervals, corresponding to different pressure abnormalities; the temperature change range of -20℃-60℃ is divided into 6 intervals, reflecting different temperature abnormalities; the sound pressure level of 30-120dB is divided into 5 intervals, corresponding to different leakage acoustic characteristics; the vibration amplitude of 0-5mm is divided into 4 intervals, reflecting different mechanical fault states.

[0076] The probability density function is calculated for the segmented statistical data. The calculation formula of the probability density function is:

[0077]

[0078] Where: P(x) is the probability density of the eigenvalue x; M is the number of mixed distributions; ω i is the weight coefficient of the ith distribution; μi is the mean of the ith distribution; σ i is the standard deviation of the ith distribution; N is the number of constraints; is the jth constraint function.

[0079] Based on the probability density calculation results, a characteristic distribution curve is established for each fault type to form a fault characteristic probability table. The probability table records the distribution characteristics of different fault types in various characteristic dimensions, including the center position, distribution shape, distribution range and other information of the distribution. The probability table is compared with the historical fault data to calculate the similarity coefficient. The similarity calculation takes into account the numerical difference of the characteristic value, the difference in distribution form and the correlation between the characteristics. The similarity coefficient is used as a quantitative indicator to measure the similarity between the new fault and the historical fault. The establishment of the fault type database is based on the fault characteristic probability table and the similarity analysis results. Each record in the database contains information such as fault type identification, characteristic parameter range, probability distribution characteristics, and similar fault cases. The database structure adopts the form of a graph database, which is convenient for establishing the association relationship between features and recording the evolution law of features. When a new gas feature vector is input, it is matched with the records in the database through data mapping, and the probability values ​​of various fault types are calculated to generate a gas fault mapping relationship.

[0080] For example, the data of a gas pipeline monitoring point is analyzed. First, the 24-hour continuous monitoring data is sorted according to the time series to form data fragments containing five dimensions: gas concentration, pressure, temperature, acoustics and vibration. Through statistical analysis, it is found that the gas concentration shows a slow upward trend from 3ppm to 15ppm during the period of 8:00-10:00; the pressure data during the same period shows a slight decrease, from 1.2MPa to 1.1MPa; the temperature is basically stable around 28℃; the acoustic data fluctuates in the range of 40-45dB; the vibration data shows a small amplitude of 0.2-0.3mm. These characteristic data are compared with the records in the fault type database. In the fault feature statistical matrix, the characteristic distribution of this set of data falls into the characteristic interval of the trace leakage fault type. Specifically, the gas concentration change amplitude and change rate are highly consistent with the characteristic distribution of the trace leakage fault; the pressure drop amplitude is small but consistent with the leakage state; the acoustic and vibration characteristics also show typical trace leakage characteristics. Through probability density calculation, the probability of this set of data belonging to the trace leakage fault type is 0.85, which is significantly higher than the probability value of other fault types.

[0081] When establishing the mapping relationship, the feature vector is associated with the trace leakage cases in the historical database, and its feature evolution process is recorded. This association includes both numerical similarity and similarity in temporal change trends. For example: a slow increase in gas concentration, a slight drop in pressure, specific frequency characteristics of acoustic and vibration signals, etc. It not only reflects the fault diagnosis results of the current state, but also includes the dynamic characteristics of fault development. This dynamic mapping relationship is useful for timely detection and disposal of various faults in the gas pipeline network. The process of establishing the mapping relationship fully considers the correlation and temporal evolution characteristics of multi-dimensional feature data, ensuring the accuracy and reliability of fault judgment.

[0082] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0083] (1) Divide the gas characteristic vector into data segments according to the three time scales of hour, day, and week, calculate the change trend of the data at each time scale, and generate time series change data;

[0084] (2) Extracting the time association rules of fault characteristics from the gas fault mapping relationship, performing time association analysis on the time series change data, and obtaining time series feature data;

[0085] (3) Convert the sensor location information in the gas feature vector into coordinate data, calculate the distance matrix between adjacent sensors, and construct a spatial association table;

[0086] (4) Perform spatial correlation analysis on the sensor data in the spatial association table, calculate the propagation characteristics of the fault characteristics in space, and form spatial feature data;

[0087] (5) Numerically normalize the temporal feature data and the spatial feature data, calculate the weight coefficients of the two types of feature data, and generate feature fusion data;

[0088] (6) Numerical calculation is performed on the feature fusion data and the gas fault mapping relationship, the fault degree is quantified according to the 0-1 interval, and the gas fault judgment value is output.

[0089] Specifically, the gas feature vector is decomposed into multiple time scales, and the data is divided into three time scales: hourly (reflecting short-term changes), daily (reflecting daily rules), and weekly (reflecting long-term trends). The change trend is calculated for the data of each time scale, including: calculating the change rate of gas concentration, pressure, and temperature, and the frequency change of acoustic and vibration characteristics on the hourly scale; analyzing the intraday fluctuation law of each parameter on the daily scale, including the characteristics of peak period, stable period, and trough period; and statistically analyzing the long-term change trend on the weekly scale to identify periodic fluctuations and cumulative effects. For the generated time series change data, the time association rules are extracted from the gas fault mapping relationship. The association rules include the time pattern of fault occurrence, the evolution law of fault development, and the time series dependency relationship between faults. By performing time correlation analysis on the time series change data, the following aspects are mainly focused on: the order of parameter changes (such as whether the pressure drop precedes the concentration increase), the duration of the change (the duration of the abnormal state), the periodicity of the change (whether there is a regular fluctuation), and the time series coupling relationship between multiple parameters.

[0090] In the spatial dimension, the sensor location information in the gas feature vector is first converted into three-dimensional coordinate data. The coordinate conversion takes into account the spatial layout of the pipeline network, including factors such as the direction, burial depth, and branch structure of the pipeline. The Euclidean distance between adjacent sensors is calculated based on the coordinate data to form a distance matrix. The distance matrix is ​​used to characterize the spatial relationship between sensor nodes and construct a spatial association table. The sensor data in the spatial association table is subjected to spatial correlation analysis. The analysis includes: the propagation direction of the fault characteristics in space (diffusion path along the pipeline network), the propagation speed (spatial gradient of feature changes), and the impact range (spatial distribution range of abnormal features). Based on the analysis results, spatial feature data is formed to describe the distribution and evolution characteristics of the fault in the spatial dimension.

[0091] Time series feature data and spatial feature data need to be standardized before they can be fused. The standardization process maps data of different dimensions to a unified numerical range to ensure data comparability. The weight coefficient calculation of the two types of feature data takes into account the reliability, representativeness and contribution of the data to fault diagnosis. Feature fusion data is generated through weighted fusion to comprehensively reflect the temporal and spatial characteristics of the fault.

[0092] Finally, the feature fusion data and the gas fault mapping relationship are numerically calculated, and the fault degree quantification formula is:

[0093]

[0094] Where: F(d) is the fault severity value (0-1 interval); T is the time scale number; S is the space scale number; α t is the time feature weight; β s is the spatial feature weight; d is the feature fusion data; γts is the standard value of space-time characteristics; η ts is the characteristic fluctuation range; R is the number of constraints; θ r (d) is the constraint function.

[0095] For example, a gas pipeline network has a leakage fault. First, the gas concentration is observed to increase slowly from 3ppm to 15ppm on an hourly scale, and the adjacent pressure sensor shows that the pressure drops from 1.2MPa to 1.1MPa. On a daily scale, this change trend is more obvious during the day and relatively gentle at night. Weekly data show that anomalies begin to accumulate. Time series analysis shows that the pressure drop precedes the concentration increase by about 30 minutes, which is consistent with the typical time series characteristics of trace leakage. Spatial analysis shows that the anomaly first appears at the sensor at point A, followed by the sensors at point B (10 meters from point A) and point C (15 meters from point A). According to the spatial layout of the sensors and the time difference between the detection of the anomaly, the propagation speed of the fault feature is calculated to be about 0.2 meters per minute, and the propagation direction is consistent with the direction of the pipeline. This spatial propagation feature is consistent with the characteristics of pipeline interface leakage. The time series and spatial features are fused after standardization and weight calculation. In this example, the weight of the time series feature is 0.6 (the dynamic feature reflecting the development of the leakage is more important), and the weight of the spatial feature is 0.4. The calculated fault judgment value is 0.65, indicating that there is a medium-level leakage risk.

[0096] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0097] (1) The gas fault judgment value is divided into five intervals according to the four numerical nodes of 0.3, 0.5, 0.7, and 0.9, and the data in each interval are numerically counted to generate a fault level distribution table;

[0098] (2) Calculate the occurrence frequency and duration of the data in the fault level distribution table, and perform numerical weighting according to the frequency weight of 0.4 and the duration weight of 0.6 to form weighted statistical data;

[0099] (3) The three indicators of signal strength, response time, and data integrity rate in the sensor status data are numerically quantified to construct a sensor reliability index table;

[0100] (4) Performing 0-1 interval normalization on the data in the sensor reliability index table, calculating the comprehensive reliability score of each sensor, and obtaining a sensor score table;

[0101] (5) Numerically combine the weighted statistical data with the sensor score table, calculate the confidence of the fault judgment according to the weighted coefficients of the three indicators, and generate fault credibility data;

[0102] (6) Numerically associate the fault credibility data with the fault level distribution table, correct the fault level according to the confidence level, and output the gas fault warning result.

[0103] Specifically, the gas fault judgment value is divided into five intervals according to the four key nodes of 0.3, 0.5, 0.7, and 0.9, corresponding to normal state (0-0.3), slight abnormality (0.3-0.5), moderate abnormality (0.5-0.7), severe abnormality (0.7-0.9), and critical state (0.9-1.0). The data in each interval are counted, and the number of data points, numerical distribution, occurrence period and other information are recorded to form a fault level distribution table. The data in the fault level distribution table are deeply analyzed, and two key indicators are calculated: frequency of occurrence and duration. The frequency of occurrence reflects the proportion of the number of times the fault state occurs in the monitoring period, which is obtained by counting the cumulative number of data points in each level interval and dividing it by the total number of monitoring times. The duration reflects the continuity of the fault state, and the total duration is obtained by calculating the duration of each fault state and accumulating it. When calculating weighted statistical data, a frequency weight of 0.4 and a duration weight of 0.6 are used. This weight configuration pays more attention to the continuity of the fault state, because persistent abnormalities often indicate more serious fault risks.

[0104] The analysis of sensor status data involves three core indicators: signal strength, response time, and data integrity rate. Signal strength is evaluated by measuring the amplitude of the sensor output signal, reflecting the working status of the sensor; response time indicates the sensor's response speed to environmental changes, which is determined by measuring the lag time of signal changes; data integrity rate is calculated by counting the proportion of valid data points to the total sampling points. These three indicators are quantified to form a sensor reliability index table. The data in the sensor reliability index table is normalized to the interval of 0-1 to make indicators of different dimensions comparable. The normalization process takes into account the physical meaning of each indicator, such as higher signal strength indicates better reliability, shorter response time indicates better performance, and higher data integrity rate indicates better stability. By combining these normalized indicators, a comprehensive reliability score is calculated for each sensor to form a sensor score table.

[0105] The combination of weighted statistical data and sensor scoring table needs to consider the correlation of data. First, analyze the relationship between fault level and sensor reliability. When the sensor reliability is low, the corresponding fault judgment result needs to be compensated for reliability. The confidence of fault judgment is calculated by the weighted coefficients of three indicators (signal strength coefficient, response time coefficient, data integrity coefficient) to generate fault credibility data. Finally, the fault credibility data is correlated with the fault level distribution table. Correct the initial fault level according to the confidence level. For example, when the confidence is low, the fault level judgment result needs to be appropriately lowered. The corrected result is output as the gas fault warning result.

[0106] For example: In the continuous 24-hour monitoring data, the fault judgment value fluctuates in different periods. From 6 to 8 in the morning, the judgment value fluctuates between 0.35-0.45, which belongs to the slight abnormal range; from 10 am to 2 pm, the judgment value rises to 0.55-0.65, entering the moderate abnormal range; the rest of the time is maintained at the normal level of 0.2-0.3. Statistics show that the frequency of slight abnormalities is 0.08 (2 hours / 24 hours) and lasts for 2 hours; the frequency of moderate abnormalities is 0.17 (4 hours / 24 hours) and lasts for 4 hours. According to the set weight calculation, the weighted value of slight abnormalities is 0.08×0.4+2 / 24×0.6=0.082, and the weighted value of moderate abnormalities is 0.17×0.4+4 / 24×0.6=0.168.

[0107] At the same time, the sensor status shows that the signal strength is maintained at more than 85% of the full scale, the response time is stable within 1.2 times the design value, and the data integrity rate reaches 96%. These indicators are normalized to obtain scores of 0.85, 0.83 and 0.96 respectively. Combining the weighted coefficients of the three indicators (0.3, 0.3, 0.4), the comprehensive reliability score of the sensor is calculated to be 0.886. Combining the weighted statistical data with the sensor score, considering the high sensor reliability, the preliminary fault level is confirmed. The generated early warning results show that the monitoring point is in a moderate abnormal state with a confidence level of 0.886, and personnel need to be arranged for on-site inspection.

[0108] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0109] (1) Extract the fault type, fault degree, and fault location information from the gas fault warning results, classify them according to the fault characteristics, and construct fault retrieval conditions;

[0110] (2) Establish a retrieval index for historical data based on three dimensions: fault type, location, and fault severity, match the fault retrieval conditions with the retrieval index, and obtain historical case data;

[0111] (3) Compare the fault characteristic parameters in the historical case data with the gas fault warning results, calculate the Euclidean distance between the characteristic parameters, and generate similarity data;

[0112] (4) Arrange the similarity data in descending order, select historical cases with a similarity greater than 0.8, extract their disposal plans and disposal effect information, and form candidate disposal plans;

[0113] (5) Quantify the candidate disposal plans according to the three indicators of disposal time, resource consumption, and restoration effect, and generate a plan scoring table;

[0114] (6) Prioritize the candidate disposal solutions based on the solution scoring table, select the solution with the highest score, and output the gas fault disposal instructions.

[0115] Specifically, the extraction of fault type information includes the specific category of the fault (such as leakage, pressure anomaly, temperature anomaly, etc.), the development stage of the fault (early, middle, late); the extraction of fault degree information includes the severity level of the fault (minor, medium, severe), the impact range of the fault; the extraction of fault location information includes the specific pipe network coordinates, surrounding environmental conditions, geological conditions, etc. The extracted information is classified and sorted according to the fault characteristics, and the fault retrieval conditions containing multiple key feature dimensions are established. The retrieval index of historical data is established based on three core dimensions: the fault type dimension records the characteristic description, judgment criteria and disposal requirements of different types of faults; the location dimension contains spatial information such as pipe network layout, geographical environment, population density, etc.; the fault degree dimension reflects the degree of harm, development speed and impact range of the fault. The fault retrieval conditions are matched with the retrieval index of these three dimensions. The matching process adopts a multi-level screening strategy. First, the fault type is roughly screened, then the region is screened according to the location information, and finally the fault degree is accurately matched to obtain historical case data with similar characteristics to the current fault.

[0116] The comparison process between historical case data and current gas fault warning results is a multi-dimensional feature similarity calculation process. First, the feature parameters are standardized to make parameters of different dimensions comparable. Then the Euclidean distances on each feature dimension are calculated, including: gas concentration feature distance, pressure feature distance, temperature feature distance, acoustic feature distance, and vibration feature distance. The calculated similarity data not only reflects the overall similarity, but also retains the difference information on each feature dimension. After the similarity data is sorted in descending order, historical cases with a similarity greater than 0.8 are selected as reference cases. The setting of this threshold ensures that the selected cases have a sufficiently high similarity with the current fault. Disposal plan information is extracted from these high-similarity cases, including specific disposal steps, required equipment, personnel configuration, etc., and disposal effect information is extracted, including repair time, resource investment, fault elimination degree, etc., to form a set of candidate disposal plans.

[0117] The scoring calculation of candidate disposal solutions considers three key indicators: the disposal time indicator reflects the time efficiency of the solution execution, including response time, repair time, and recovery time; the resource consumption indicator reflects the economy of the solution, including manpower input, equipment use, and material consumption; the repair effect indicator evaluates the effectiveness of the solution, including the degree of fault elimination, system recovery status, and subsequent stability. Through the quantitative calculation of these indicators, a scoring table for each candidate solution is generated. When prioritizing based on the solution scoring table, a comprehensive evaluation method is adopted. First, the weight coefficient of each indicator is calculated to reflect the importance of different indicators in decision-making. Then the score of each indicator is multiplied by the weight coefficient and summed to obtain the comprehensive score of the solution. The priority order of the solution is determined according to the comprehensive score, and the solution with the highest score is selected as the disposal solution to generate detailed fault disposal instructions.

[0118] For example, a gas pipeline monitoring point detected a fault warning, and the warning result showed a medium-level leakage fault. The following characteristic information was obtained during the data extraction process: the fault type was a trace leakage, characterized by a slow increase in gas concentration (from 3ppm to 15ppm) and a slight decrease in pressure (from 1.2MPa to 1.1MPa); the fault location was located at an interface of the main pipeline, surrounded by non-dense residential areas; the fault level was medium, and it had lasted for 4 hours but did not show a sharp deterioration trend. The historical database was retrieved, and 300 historical records were obtained by first filtering by the type of trace leakage; further filtering by the location of the pipeline interface, 80 records remained; and finally filtering by medium-level faults to obtain 15 similar cases. The characteristic parameters of these cases were compared, and key parameters such as gas concentration change rate, pressure change trend, and fault duration were focused on when calculating the similarity. After similarity calculation and sorting, 5 cases with a similarity greater than 0.8 were selected for in-depth analysis.

[0119] The disposal plans for these five cases have their own characteristics: Plan A uses temporary sealing and then replacing the interface, which takes 8 hours to dispose and requires interruption of gas supply; Plan B uses pressure repair technology, which takes 12 hours to dispose and does not require interruption of gas supply; Plan C uses an interface modification plan, which takes 24 hours to dispose and requires local pipe network modification; Plan D uses glue injection and plugging technology, which takes 6 hours to dispose and is a temporary repair; Plan E adopts an overall replacement plan, which takes 16 hours to dispose and has the most thorough treatment effect. Through the scoring calculation, considering the different requirements for disposal time, resource consumption and repair effect under the current situation, Plan B received the highest score. The generated disposal instructions clearly stipulate the use of pressure repair technology for disposal, and list in detail the required special equipment, technical personnel configuration, operating procedures and safety precautions.

[0120] In a specific embodiment, the process of performing numerical comparison between the fault characteristic parameters in the historical case data and the gas fault warning result may specifically include the following steps:

[0121] (1) Decompose the historical case data and gas fault warning results according to five dimensions: gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data, and pipeline network vibration data to obtain characteristic component data;

[0122] (2) Perform zero-mean normalization on the feature component data, map each dimension of data to a unified numerical interval, and generate standardized feature data;

[0123] (3) Assigning weight coefficients to each dimensional feature in the standardized feature data according to the numerical importance, weighting the data to form weighted feature data;

[0124] (4) constructing a distance calculation matrix for the weighted feature data, calculating the Euclidean distance value between each pair of historical cases and the current fault, and generating a distance matrix;

[0125] (5) Convert the values ​​in the distance matrix into similarity coefficients according to an inverse relationship, normalize the similarity coefficients, and obtain similarity degree data;

[0126] (6) Group and count the similarity data, calculate the similarity contribution rate of each feature dimension, and output the similarity data.

[0127] Specifically, the historical case data and the current gas fault warning results are feature decomposed to extract the characteristics of five dimensions: gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data, and pipeline vibration data. The gas concentration data reflects the change in the degree of leakage, including the concentration value, change rate, and fluctuation range; the pipeline pressure data reflects the operation status of the pipeline network, including the pressure value, pressure drop rate, and fluctuation amplitude; the gas temperature data records the temperature change characteristics, including the temperature value, temperature gradient, and fluctuation law; the leakage acoustic data describes the acoustic characteristics, including the sound pressure level, spectrum distribution, and duration; the pipeline vibration data characterizes the mechanical characteristics, including amplitude, frequency, and vibration mode.

[0128] The extracted feature component data is subjected to zero mean standardization to unify data of different dimensions and ranges to a comparable scale. The standardization process first calculates the mean of each dimension data, then subtracts the mean from the original data and divides it by the standard deviation, so that the processed data has the characteristics of zero mean and unit standard deviation. This processing method eliminates the influence of dimension while maintaining the distribution characteristics of the data. The standardized data is convenient for subsequent similarity calculation. The weight allocation of standardized feature data is based on the importance of each dimensional feature in fault judgment. Gas concentration data usually has a higher weight because it directly reflects the leakage condition; pressure data is second, reflecting the operation status of the pipeline network; the weights of temperature, acoustic and vibration data are dynamically adjusted according to the fault type. The determination of the weight coefficient takes into account the stability, response speed and anti-interference ability of the feature. Through weight weighting, the role of key features in similarity calculation is highlighted.

[0129] The weighted feature data is used to construct a distance calculation matrix. Each element of the matrix represents the Euclidean distance between a pair of historical cases and the current fault on a certain feature dimension. The distance calculation takes into account the absolute difference in the feature values ​​and the difference in the change trend. A smaller distance value indicates a higher similarity. A complete distance matrix is ​​formed by calculating the distances on all feature dimensions. An inverse relationship is used when the distance matrix is ​​converted to a similarity coefficient, that is, the smaller the distance, the higher the similarity. Nonlinear relationships are considered in the conversion process so that the similarity coefficient better reflects the actual degree of similarity. The similarity coefficient is normalized and the value is mapped to the range of 0-1 for intuitive understanding and comparison. The normalized similarity data clearly shows the similarity relationship between different cases.

[0130] The grouping statistics of the similarity data analyze the contribution of each feature dimension to the overall similarity. The statistical process calculates the similarity contribution rate of each dimension feature, reflecting the importance of different features in fault identification. This contribution rate information helps to optimize the feature weight configuration and improve the similarity calculation method. The final output similarity data fully reflects the similarity between the historical case and the current fault.

[0131] For example, in the analysis of a gas pipeline leakage fault, the characteristic data of the current fault shows that the gas concentration rises slowly from 3ppm to 15ppm, the pressure drops from 1.2MPa to 1.1MPa, the temperature remains at around 28℃, the acoustic data shows a continuous sound pressure of 40-45dB, and the vibration data shows an amplitude of 0.2-0.3mm. When retrieving from the historical database, these features are decomposed and standardized. Considering the characteristics of the leakage fault, the gas concentration feature weight is set to 0.4, the pressure feature weight is set to 0.3, the acoustic feature weight is set to 0.2, and the temperature and vibration features are set to 0.05 each. Through distance calculation and similarity conversion, the three most similar historical cases are identified, which all show similar gas concentration rising trends and pressure drop characteristics, providing valuable references for fault diagnosis and disposal. The similarity analysis results show that the similarity contribution rates of gas concentration and pressure features are 45% and 35% respectively, which verifies the rationality of feature weight setting.

[0132] The above describes the gas monitoring fault analysis method in the embodiment of the present application. The following describes the gas monitoring fault analysis system in the embodiment of the present application. Figure 2 In the embodiment of the present application, an embodiment of the gas monitoring fault analysis system includes:

[0133] The transmission module is used to collect gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data and pipeline vibration data through the gas monitoring sensor group, transmit the data to the processing unit according to the Internet of Things communication protocol, pre-screen the data, and obtain the original gas monitoring data;

[0134] A dimension reduction module is used to calculate the raw gas monitoring data according to the median filter formula and remove abnormal values, calculate characteristic parameters for the filtered data, and perform dimension reduction processing on the calculation results to obtain a gas feature vector;

[0135] A comparison module, used to establish a fault type database according to the gas characteristic vector, classify and calculate the characteristic data, compare the calculation results with preset parameters, and generate a gas fault mapping relationship;

[0136] A fusion module, used to extract time series data according to the gas feature vector and the gas fault mapping relationship, perform weight fusion calculation on the time series data and the spatial data, and output a gas fault judgment value;

[0137] An evaluation module is used to compare and classify the gas fault judgment value with a preset threshold value, perform weighted calculation on the classification result, and perform credibility evaluation in combination with the sensor status data to obtain a gas fault warning result;

[0138] The sorting module is used to retrieve historical data based on the gas fault warning result, perform similarity calculation and priority sorting on the retrieved data, and generate a gas fault handling instruction.

[0139] Through the coordinated cooperation of the above components, the gas monitoring sensor group simultaneously collects gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data and pipeline vibration data, realizing the coordinated monitoring of multi-dimensional data, comprehensively reflecting the operation status of the gas pipeline network, and using the Internet of Things communication protocol for data transmission to ensure the reliability and real-time performance of data transmission. The pre-screening process effectively reduces the impact of data noise. The median filter formula calculation can effectively remove outliers and improve data quality. The characteristic parameter calculation and dimensionality reduction processing realize effective data compression and feature extraction. The establishment of a fault type database enables fault characteristics to be systematically organized and managed, and the classification of characteristic data Calculation and comparative analysis improve the accuracy of fault identification. The weighted fusion calculation of time series data and spatial data makes full use of the temporal and spatial correlation of fault characteristics, making fault judgment more comprehensive and accurate. The hierarchical comparison and weighted calculation of preset thresholds realize the accurate division of fault levels. The credibility assessment combined with sensor status data ensures the reliability of the early warning results. By retrieving historical data and performing similarity calculation, the most similar historical cases are found, providing a reliable reference for fault handling. Priority sorting ensures the optimal selection of handling plans, thus forming a complete, automated and intelligent gas monitoring fault analysis process, which significantly improves the accuracy of gas pipeline fault diagnosis and the efficiency of handling.

[0140] The present application also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the gas monitoring fault analysis method.

[0141] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0142] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0143] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for analyzing gas monitoring failures, characterized in that: The gas monitoring fault analysis method comprises: The gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data and pipeline vibration data are collected through the gas monitoring sensor group, and the data is transmitted to the processing unit according to the Internet of Things communication protocol, and the data is pre-screened to obtain the original gas monitoring data; The raw gas monitoring data is calculated according to the median filter formula and outliers are removed, characteristic parameters are calculated for the filtered data, and dimension reduction processing is performed on the calculation results to obtain a gas characteristic vector; Establishing a fault type database based on the gas characteristic vector, classifying and calculating the characteristic data, comparing the calculation results with preset parameters, and generating a gas fault mapping relationship; Extracting time series data according to the gas feature vector and the gas fault mapping relationship, performing weight fusion calculation on the time series data and the spatial data, and outputting a gas fault determination value; The gas fault judgment value is compared with the preset threshold value for classification, weighted calculation is performed on the classification result, and credibility evaluation is performed in combination with the sensor status data to obtain a gas fault warning result; Based on the gas fault warning result, historical data is retrieved, similarity calculation and priority sorting are performed on the retrieved data, and a gas fault handling instruction is generated.

2. The gas monitoring fault analysis method according to claim 1, characterized in that: The gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data and pipeline vibration data are collected by the gas monitoring sensor group, and the data are transmitted to the processing unit according to the Internet of Things communication protocol, and the data are pre-screened to obtain the gas monitoring raw data, including: The gas concentration data is collected at a sampling interval of 100ms, the pipeline pressure data is collected at a sampling interval of 200ms, the gas temperature data is collected at a sampling interval of 500ms, the leakage acoustic data is collected at a sampling interval of 50ms, and the pipe network vibration data is collected at a sampling interval of 20ms; Perform timestamp alignment on the gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data, and pipeline network vibration data, construct a data packet in units of 1s, and add a check bit to the data packet according to the Internet of Things communication protocol; The gas concentration data in the data packet is set to a concentration range threshold of 0-100ppm, the pipeline pressure data is set to a pressure range threshold of 0-1.6MPa, the gas temperature data is set to a temperature range threshold of -20°C-60°C, the leakage acoustic data is set to a sound pressure range threshold of 30-120dB, and the pipeline network vibration data is set to an amplitude range threshold of 0-5mm for numerical range pre-screening; The pre-screened gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data and pipeline network vibration data are correspondingly merged into data records, and the collection time and sensor number are marked for each data record; Performing data integrity checks on the data records, removing records with missing data, marking abnormal data records, and establishing data quality scores; The data records that have passed the integrity check are organized into a time series data table, and a data source identifier is added to generate the gas monitoring raw data.

3. The gas monitoring fault analysis method according to claim 1, characterized in that: The gas monitoring raw data is calculated according to the median filter formula and outliers are removed, characteristic parameters are calculated for the filtered data, and dimension reduction processing is performed on the calculation results to obtain a gas characteristic vector, including: The raw gas monitoring data is grouped according to a 10-minute time window, the median value and the interquartile range are calculated for each group of data, and outliers are determined and eliminated based on an interval of three times the interquartile range to obtain filtered processed data; Calculate the mean, standard deviation, kurtosis and skewness of the gas concentration data in the filtered data, calculate the maximum fluctuation amplitude and fluctuation period of the pipeline pressure data, calculate the rate of change and temperature gradient of the gas temperature data, calculate the sound pressure level and frequency distribution of the leakage acoustic data, and calculate the vibration amplitude and vibration frequency of the pipeline network vibration data to generate a characteristic data group; The mean, standard deviation, kurtosis, and skewness data in the characteristic data group are normalized to the interval of 0-1, the maximum fluctuation amplitude and fluctuation period data are normalized to the interval of -1-1, the rate of change and temperature gradient data are linearly transformed to the standard normal distribution, the sound pressure level and frequency distribution data are logarithmically transformed, and the vibration amplitude and vibration frequency data are subjected to minimum-maximum transformation to obtain normalized characteristic data; The normalized feature data are sorted according to the variance contribution rate, and feature dimensions whose cumulative variance contribution rate reaches 95% are selected to form a main feature data set; Calculate the feature importance of each feature data in the main feature data set according to the correlation coefficient matrix, perform linear combination on the feature data based on the importance weight, and construct compressed feature data; The compressed characteristic data is converted into a numerical vector form, and a time index and a sensor position identifier are added to generate the gas characteristic vector.

4. The gas monitoring fault analysis method according to claim 1, characterized in that: The method of establishing a fault type database based on the gas characteristic vector, classifying and calculating the characteristic data, comparing the calculation results with preset parameters, and generating a gas fault mapping relationship includes: The gas feature vector is divided into 24-hour data segments according to the timestamp, and statistical features are calculated for the gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data and pipeline network vibration data in each data segment to generate a feature statistical table; The data in the characteristic statistics table are grouped according to the fault type labels, and the mean center and variance distribution of each group of data are calculated to form a fault characteristic statistics matrix; The data in the fault feature statistical matrix are segmented according to the gas concentration range of 0-100ppm, the pressure fluctuation range of 0-1.6MPa, the temperature change range of -20℃-60℃, the sound pressure level of 30-120dB, and the vibration amplitude of 0-5mm, to construct segmented statistical data; Performing probability density calculation on the segmented statistical data to obtain characteristic distribution curves of each fault type and generating a fault characteristic probability table; Compare the fault feature probability table with the historical fault data, calculate the feature similarity coefficient, and establish a fault type database; The input gas feature vector is data mapped based on the fault type database, the fault type probability value is calculated, and the gas fault mapping relationship is generated.

5. The method for analyzing gas monitoring failure according to claim 1, characterized in that: The extracting of time series data according to the gas feature vector and the gas fault mapping relationship, performing weight fusion calculation on the time series data and the spatial data, and outputting a gas fault determination value includes: Divide the gas characteristic vector into data segments according to three time scales: hour, day, and week, calculate the change trend of the data at each time scale, and generate time series change data; Extracting the time association rule of the fault feature from the gas fault mapping relationship, performing time association analysis on the time series variation data, and obtaining time series feature data; Converting the sensor position information in the gas feature vector into coordinate data, calculating the distance matrix between adjacent sensors, and constructing a spatial association table; Performing spatial correlation analysis on the sensor data in the spatial association table, calculating the propagation characteristics of the fault features in space, and forming spatial feature data; Numerical standardization is performed on the temporal feature data and the spatial feature data, weight coefficients of the two types of feature data are calculated, and feature fusion data is generated; The feature fusion data and the gas fault mapping relationship are numerically calculated, the fault degree is quantified according to the interval of 0-1, and the gas fault judgment value is output.

6. The method for analyzing gas monitoring failure according to claim 1, characterized in that: The gas fault judgment value is compared with a preset threshold value for grading, weighted calculation is performed on the grading result, and credibility evaluation is performed in combination with the sensor status data to obtain a gas fault warning result, including: The gas fault judgment value is divided into five intervals according to four numerical nodes of 0.3, 0.5, 0.7 and 0.9, and numerical statistics are performed on the data of each interval to generate a fault level distribution table; The occurrence frequency and duration of the data in the fault level distribution table are calculated, and numerical weighting is performed according to a frequency weight of 0.4 and a duration weight of 0.6 to form weighted statistical data; The three indicators of signal strength, response time and data integrity rate in the sensor status data are numerically quantified to construct a sensor reliability index table; The data in the sensor reliability index table are normalized to a range of 0-1, and the comprehensive reliability score of each sensor is calculated to obtain a sensor score table; The weighted statistical data is numerically combined with the sensor scoring table, and the confidence of the fault judgment is calculated according to the weighted coefficients of the three indicators to generate fault credibility data; The fault credibility data is numerically associated with the fault level distribution table, the fault level is corrected according to the confidence level, and the gas fault warning result is output.

7. The method for analyzing gas monitoring failure according to claim 1, characterized in that: The retrieving historical data based on the gas fault warning result, performing similarity calculation and priority sorting on the retrieved data, and generating a gas fault handling instruction includes: Extract the fault type, fault degree, and fault location information from the gas fault warning result, classify them according to the fault characteristics, and construct fault retrieval conditions; Establishing a retrieval index for the historical data according to three dimensions: fault type, occurrence location, and fault severity, matching the fault retrieval condition with the retrieval index, and obtaining historical case data; Numerically compare the fault characteristic parameters in the historical case data with the gas fault warning results, calculate the Euclidean distance between the characteristic parameters, and generate similarity data; Arrange the similarity data in descending order, select historical cases with a similarity greater than 0.8, extract their disposal plans and disposal effect information, and form candidate disposal plans; Quantitatively calculate the candidate disposal solutions according to three indicators: disposal time, resource consumption, and restoration effect, and generate a solution scoring table; The candidate disposal solutions are prioritized based on the solution scoring table, the disposal solution with the highest score is selected, and the gas fault disposal instruction is output.

8. The method for analyzing gas monitoring failure according to claim 7, characterized in that: The method of comparing the fault characteristic parameters in the historical case data with the gas fault warning results, calculating the Euclidean distance between the characteristic parameters, and generating similarity data includes: Decomposing the historical case data and the gas fault warning results according to five dimensions, namely, gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data, and pipeline network vibration data, to obtain characteristic component data; Performing zero-mean normalization processing on the feature component data, mapping each dimension of data to a unified numerical interval, and generating standardized feature data; Assigning weight coefficients to the features of each dimension in the standardized feature data according to the importance of the values, weighting the data, and forming weighted feature data; Constructing a distance calculation matrix for the weighted feature data, calculating the Euclidean distance value between each pair of historical cases and the current fault, and generating a distance matrix; The values ​​in the distance matrix are converted into similarity coefficients according to an inverse relationship, and the similarity coefficients are normalized to obtain similarity degree data; The similarity data are grouped and counted, the similarity contribution rate of each feature dimension is calculated, and the similarity data is output.

9. A gas monitoring fault analysis system, used to implement the gas monitoring fault analysis method according to any one of claims 1 to 8, characterized in that: The gas monitoring fault analysis system comprises: The transmission module is used to collect gas concentration data, pipeline pressure data, gas temperature data, leakage acoustic data and pipeline vibration data through the gas monitoring sensor group, transmit the data to the processing unit according to the Internet of Things communication protocol, pre-screen the data, and obtain the original gas monitoring data; A dimension reduction module is used to calculate the raw gas monitoring data according to the median filter formula and remove abnormal values, calculate characteristic parameters for the filtered data, and perform dimension reduction processing on the calculation results to obtain a gas feature vector; A comparison module, used to establish a fault type database according to the gas characteristic vector, classify and calculate the characteristic data, compare the calculation results with preset parameters, and generate a gas fault mapping relationship; A fusion module, used to extract time series data according to the gas feature vector and the gas fault mapping relationship, perform weight fusion calculation on the time series data and the spatial data, and output a gas fault judgment value; An evaluation module is used to compare and classify the gas fault judgment value with a preset threshold value, perform weighted calculation on the classification result, and perform credibility evaluation in combination with the sensor status data to obtain a gas fault warning result; The sorting module is used to retrieve historical data based on the gas fault warning result, perform similarity calculation and priority sorting on the retrieved data, and generate a gas fault handling instruction.

10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the gas monitoring fault analysis method according to any one of claims 1 to 8 is implemented.

Images