A municipal pipeline management method and system based on positioning perception function

Abstract

The present invention relates to a municipal pipeline management method and system based on positioning perception function, the method includes collecting underground pipeline geographic coordinates and real-time monitoring images, building a pipeline real-time database; analyzing pipeline aging and damage trends, building a pipeline crack monitoring model, and predicting the development trend of cracks; analyzing and processing pipeline real-time image information, calculating the real-time external stress value and force change trend of the outer wall of the pipeline crack position; evaluating crack detailed information and external stress, quantifying the urgency of pipeline maintenance; combining the emergency maintenance degree, budget and cost of the pipeline, establishing a pipeline maintenance decision model, determining the maintenance priority of the pipeline; implementing corresponding pipeline maintenance strategies according to the maintenance priority. The present invention realizes accurate positioning and real-time monitoring of municipal underground pipelines by collecting underground pipeline geographic coordinates and real-time monitoring images, builds a comprehensive and dynamically updated pipeline database, and greatly improves the accuracy and timeliness of pipeline information management.

Description

Technical Field

[0001] The present invention relates to the technical field of pipeline engineering, and in particular to a municipal pipeline management method and system based on positioning perception function. Background Art

[0002] In urban underground pipeline networks, pipeline aging and damage is a common and serious problem. With the aging of pipeline materials, the impact of the external environment, and the long-term scouring and corrosion of the medium inside the pipeline, pipelines are prone to cracks and breakages, which in turn lead to accidents such as pipeline leakage and breakage. Especially in complex underground pipeline networks, due to the mutual influence and constraints of the spatial positions of pipelines, the damage of a single pipeline is likely to trigger a chain reaction, causing damage to adjacent pipelines and ultimately paralyzing a large area of ​​the pipeline network.

[0003] At present, a technical solution is needed to monitor the status of pipelines in real time, detect pipeline cracks and other problems in time, and quantify and analyze the severity and development trend of cracks, so as to take targeted pipeline maintenance and repair measures. Due to the complexity of underground pipeline networks, pipeline maintenance and repair work faces problems such as high construction difficulty, high cost, and long cycle. How to reasonably formulate the priority of pipeline maintenance on the basis of comprehensive monitoring of pipeline status and adopt economical and efficient pipeline repair technology with little impact on the environment has become a major technical challenge facing urban management departments. It is necessary to make full use of advanced monitoring, sensing, big data analysis, artificial intelligence and other technologies to build an intelligent underground pipeline network supervision and operation and maintenance system to achieve real-time perception, dynamic evaluation, accurate early warning and optimized decision-making of pipeline status, and ultimately achieve the goal of improving pipeline safety, extending pipeline service life, saving maintenance costs, and protecting the ecological environment. Summary of the invention

[0004] In order to solve the problems existing in the above-mentioned prior art, the purpose of the present invention is to provide a municipal pipeline management method and system based on positioning perception function. The municipal pipeline management method based on positioning perception function realizes accurate positioning and real-time monitoring of municipal underground pipelines by collecting the geographic coordinates of underground pipelines and real-time monitoring images, and builds a comprehensive and dynamically updated pipeline database, which greatly improves the accuracy and timeliness of pipeline information management.

[0005] The municipal pipeline management method and system based on positioning perception function described in the present invention include:

[0006] S1. Obtain the geographic coordinates of the underground pipelines under the jurisdiction of the municipality, monitor the outer wall images of the pipelines in real time, and build a real-time pipeline database;

[0007] S2. In combination with the pipeline real-time database, the attribute data of the pipeline is analyzed by using a time series analysis method to obtain the pipeline aging and damage trends, a pipeline crack monitoring model is constructed, and the development trend of the cracks is predicted by combining historical data and real-time monitoring data;

[0008] S3. Analyze and process the real-time image information of the pipeline, determine the depth, width, horizontal angle to the ground and curvature of the pipeline crack, and calculate the real-time external stress value and stress change trend of the outer wall of the pipeline crack position;

[0009] S4. Collecting vibration signals of the pipeline, and performing spectrum analysis on the vibration signals, and quantifying the urgency of pipeline maintenance by combining the external stress value, stress change trend and crack change trend of the outer wall of the pipeline crack;

[0010] S5. Process the pipeline maintenance urgency, maintenance budget and cost, and establish a pipeline maintenance decision model in combination with the grey correlation analysis method to evaluate and adjust the pipeline maintenance priority;

[0011] S6. According to the maintenance priority of the pipeline, different degrees of environmentally friendly maintenance strategies are associated with pipeline sections with different maintenance priorities.

[0012] Preferably, the step S1 specifically includes:

[0013] Using a differential global positioning system to obtain geographic coordinate information of underground pipelines under municipal jurisdiction, the geographic coordinate information including a starting point, an end point and a turning point, and forming a pipeline geographic coordinate data set;

[0014] By comparing and calibrating the differential global positioning system with the actual pipeline path, extracting pipeline depth information from the differential global positioning system and associating it with corresponding geographic coordinates to form a pipeline depth information data set;

[0015] Extract the material information of the pipeline to form a pipeline material information data set;

[0016] Acquire the design working pressure data of the pipeline, and associate the design working pressure data with the geographic coordinate information, the depth information, and the material information to generate a pipeline design working pressure data set;

[0017] Use pressure sensors to monitor the working pressure of the pipeline in real time and transmit the real-time pressure data to the pipeline database;

[0018] An image acquisition device is used to photograph the outer wall of the pipeline to obtain an image of the outer wall of the pipeline, and an image transmission network is established to transmit the acquired image to the pipeline database in real time;

[0019] A data warehouse is used to process the multi-source heterogeneous data of the pipeline, form a consistent data model and format, and build the pipeline real-time database.

[0020] Preferably, the step S2 specifically includes:

[0021] Extracting pipeline attribute data from the pipeline real-time database, the pipeline attribute data including service life, material type and design working pressure, and arranging the attribute data in chronological order to form a time series data set of pipeline attributes;

[0022] Preprocessing the time series data set, identifying outliers using a box plot, and conducting a comprehensive analysis based on the pipeline operation status and environmental factors to determine whether the outliers represent an abnormal state of the pipeline, and marking and alarming the abnormal state of the pipeline;

[0023] The LSTM model is used to model and predict the historical change trend of the pipeline design working pressure. By analyzing the periodicity, trend and randomness of the pressure change, the pipeline aging and damage trend is obtained;

[0024] Based on pipeline aging and damage trends and combined with pipeline attribute data, a pipeline crack monitoring model is constructed using machine learning algorithms;

[0025] Use a pipeline endoscope or a crawling robot to obtain crack images, perform denoising, enhancement and segmentation on the crack images, extract the length, width and direction of the cracks, as well as the texture and color within the preset range of the cracks, and construct a crack feature vector;

[0026] By using crack feature vectors and combining historical crack data of pipelines, a convolutional neural network is used to train a pipeline crack monitoring model to predict the development direction of pipeline cracks.

[0027] Preferably, the step S3 specifically includes:

[0028] Collecting real-time image information of the pipeline through an image acquisition device, and performing image enhancement and noise removal processing on the real-time image information;

[0029] The processed image is segmented by an image segmentation algorithm to extract pixel information of the pipeline crack area and obtain a binary image of the crack area.

[0030] The edge detection algorithm is used to extract the pixel coordinate information of the crack edge and determine the width and depth parameters of the crack;

[0031] Combined with the geometric structure parameters and spatial position information of the pipeline, the angle of the crack area relative to the central axis of the pipeline is calculated to determine the horizontal angle of the crack to the ground;

[0032] The pixel coordinates of the crack edge are fitted using a curve fitting algorithm to obtain the mathematical expression of the crack curve and calculate the curvature of the crack curve.

[0033] Using graph convolutional neural networks, an image processing model is constructed based on the width, depth, angle, and curvature of the crack. The mapping relationship between crack parameters and pipeline stress is established through feature extraction and parameter learning of the image processing model.

[0034] According to the output results of the image processing model, the real-time stress condition of the outer wall at the crack position of the pipeline is obtained to determine the external stress value of the pipeline at the current moment;

[0035] The time series analysis algorithm is used to predict the trend of pipeline stress data and determine the stress change trend of pipeline crack locations.

[0036] Preferably, the step S3 further includes:

[0037] For adjacent pipes with cracks, the ductility score between the two cracks of the adjacent pipes is calculated through the real-time image information of the adjacent pipes.

[0038] Preferably, in step S4, performing spectrum analysis on the vibration signal includes:

[0039] The vibration signal of the outer wall of the pipeline is collected by using an acceleration sensor, and the collected time domain vibration signal is analyzed and preprocessed;

[0040] The fast Fourier transform algorithm is used to perform frequency domain conversion on the preprocessed pipeline vibration signal to obtain the frequency spectrum distribution diagram of pipeline vibration;

[0041] According to the frequency spectrum distribution diagram, the distribution law of the frequency spectrum peak amplitude is obtained through statistical analysis of the pipeline vibration data;

[0042] According to the mean and standard deviation, the threshold of the spectrum peak is set as: mean + 2 times the standard deviation, and the corresponding frequency and amplitude parameters are extracted to obtain the characteristic spectrum peak parameters of pipeline vibration;

[0043] Perform short-time Fourier transform on the pipeline vibration signal to obtain the time-frequency distribution diagram of the vibration signal. Set the window function of the short-time Fourier transform. Through the time-frequency diagram, the transient characteristics and energy distribution changes of the pipeline vibration can be analyzed, and the transient characteristic parameters of the pipeline vibration signal can be extracted.

[0044] The pipeline vibration signal is divided into several time windows by using the order analysis method based on short-time Fourier transform. The signal in each window is subjected to fast Fourier transform to obtain the spectrum diagram at different times.

[0045] For the spectrum graph within each time window, find the frequency point with the largest amplitude in each frequency band as the characteristic frequency of the frequency band;

[0046] Arrange the characteristic frequencies of each frequency band from small to large to obtain a set of frequency sequences, which correspond to the order components of pipeline vibration.

[0047] Preferably, the step S4 further includes:

[0048] For adjacent pipelines with cracks, the resonance phenomenon of the adjacent pipelines is identified through the vibration signals of the adjacent pipelines.

[0049] Preferably, the step S5 specifically includes:

[0050] Extracting data on pipeline maintenance urgency, pipeline maintenance budget and pipeline maintenance cost from the pipeline maintenance decision system;

[0051] Processing data on the pipeline maintenance urgency, the pipeline maintenance budget, and the pipeline maintenance cost;

[0052] A pipeline maintenance decision model is established by using a grey correlation analysis method, the pipeline maintenance urgency, the pipeline maintenance budget and the pipeline maintenance cost are used as input parameters of the model, and the pipeline maintenance priority is used as the output parameter of the model. The importance weight of each parameter is determined by calculating the grey correlation between each input parameter and the output parameter;

[0053] Calculating the grey correlation coefficient of the pipeline maintenance urgency, the pipeline maintenance budget and the pipeline maintenance cost;

[0054] According to the grey correlation coefficient, the grey correlation degree of the pipeline maintenance urgency, the pipeline maintenance budget and the pipeline maintenance cost is calculated, and the grey correlation degree is normalized to obtain a weight vector of each parameter;

[0055] Performing a weighted summation on the pipeline maintenance urgency, the pipeline maintenance budget, and the standardized value of the pipeline maintenance cost and a weight vector to obtain a comprehensive evaluation value of the pipeline maintenance priority;

[0056] Sort the pipelines according to the comprehensive evaluation values ​​to obtain a pipeline maintenance priority list;

[0057] If the maintenance priority of a section of pipeline changes, the maintenance priority of adjacent pipelines will be adjusted in conjunction with the pipeline network topology, combined with the pipeline importance coefficient and risk level coefficient.

[0058] Preferably, the step S6 specifically includes:

[0059] Based on the pipeline maintenance priority list generated by the pipeline maintenance decision system, the pipeline maintenance priority evaluation index system is constructed by using the analytic hierarchy process from multiple dimensions including pipeline importance, failure risk and maintenance cost.

[0060] By comparing two by two, the weight of each indicator is calculated to form a comprehensive score of pipeline maintenance priority. According to the score distribution, the threshold intervals of high and low priorities are divided, and the priority threshold is set. The pipelines with maintenance priorities higher than the threshold are classified as high-priority pipelines, and the pipelines with maintenance priorities lower than or equal to the threshold are classified as low-priority pipelines. Different degrees of the environmentally friendly maintenance strategy are associated with pipeline sections with different maintenance priorities;

[0061] The environmentally friendly maintenance strategy includes using bioremediation agents to repair cracks in high-priority pipelines and taking regular monitoring and tracking and non-invasive detection protection measures for cracks in low-priority pipelines.

[0062] The present invention further proposes a municipal pipeline management system based on positioning perception function, which adopts the municipal pipeline management method based on positioning perception function as described above, including:

[0063] A collection and monitoring module, which is used to collect geographic coordinates, depth and real-time outer wall image information of underground pipelines;

[0064] A pipeline real-time database construction module, wherein the pipeline real-time database construction module is used to store and manage the geographic coordinates, materials, design pressure and image data of the pipeline;

[0065] A time series analysis and crack monitoring module, which is used to analyze pipeline aging trends and predict crack development;

[0066] An image processing and stress analysis module, which is used to determine the specific characteristics of the cracks and analyze the real-time external stress on the outer wall of the pipeline;

[0067] A spectrum analysis and evaluation module, which is used to analyze pipeline status through vibration signals and quantify maintenance requirements;

[0068] The maintenance decision and environmental strategy module is used to determine the maintenance priority of each pipeline section and associate the corresponding maintenance strategy.

[0069] The municipal pipeline management method and system based on positioning perception function described in the present invention have the following advantages:

[0070] A municipal pipeline management method based on positioning perception function of the present invention realizes accurate positioning and real-time monitoring of municipal underground pipelines by collecting geographic coordinates and real-time monitoring images of underground pipelines, constructs a comprehensive and dynamically updated pipeline database, and greatly improves the accuracy and timeliness of pipeline information management; adopts time series analysis and image processing models, which can not only predict the aging trend and damage risk of pipelines, but also monitor the dynamic changes of cracks in real time, discover potential safety hazards in advance, and effectively prevent sudden accidents; by analyzing the spectrum characteristics of cracks and vibration signals of adjacent pipelines, the influence of external stress and resonance phenomenon on pipeline cracks is accurately identified, providing a scientific basis for formulating targeted maintenance strategies; by quantifying the urgency of pipeline maintenance, considering maintenance budget and cost, scientifically formulating maintenance priorities, and realizing efficient allocation and utilization of maintenance resources; implementing differentiated maintenance strategies according to pipeline maintenance priorities, and by dynamically adjusting maintenance priorities and strategies, timely maintenance of key pipelines is ensured, excessive or insufficient maintenance is avoided, maintenance costs are reduced, and stable operation of urban infrastructure is guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 It is a flow chart of a municipal pipeline management method based on positioning perception function described in the present invention. DETAILED DESCRIPTION

[0072] like Figure 1 As shown, the municipal pipeline management method based on positioning perception function described in the present invention includes:

[0073] S1. Obtain the geographic coordinates of the underground pipelines under the jurisdiction of the municipality, monitor the outer wall images of the pipelines in real time, and build a real-time pipeline database;

[0074] S2. Combined with the pipeline real-time database, the time series analysis method is used to analyze the pipeline attribute data, obtain the pipeline aging and damage trends, build a pipeline crack monitoring model, and combine historical data and real-time monitoring data to predict the development trend of cracks;

[0075] S3. Analyze and process the real-time image information of the pipeline, determine the depth, width, horizontal angle to the ground and curvature of the pipeline crack, and calculate the real-time external stress value and stress change trend of the outer wall of the pipeline crack position;

[0076] S4. Collect the vibration signal of the pipeline, perform spectrum analysis on the vibration signal, and quantify the urgency of pipeline maintenance by combining the external stress value, stress change trend and crack change trend of the outer wall of the pipeline crack;

[0077] S5. Process the pipeline maintenance urgency, maintenance budget and cost, and establish a pipeline maintenance decision model in combination with the grey correlation analysis method to evaluate and adjust the pipeline maintenance priority;

[0078] S6. According to the maintenance priority of the pipeline, different degrees of pipeline maintenance strategies are associated with pipeline sections with different maintenance priorities.

[0079] Furthermore, in this embodiment, step S1 specifically includes:

[0080] Use differential global positioning system to obtain the geographic coordinate information of underground pipelines under municipal jurisdiction, including starting point, end point and turning point, and form a pipeline geographic coordinate data set;

[0081] By comparing and calibrating the differential global positioning system with the actual pipeline path, the pipeline depth information is extracted from the differential global positioning system and associated with the corresponding geographic coordinates to form a pipeline depth information data set;

[0082] Extract the material information of the pipeline to form a pipeline material information data set; the material information of the pipeline includes the material, diameter and wall thickness of the pipeline;

[0083] Obtain the design working pressure data of the pipeline, and associate the design working pressure data with geographic coordinate information, depth information, and material information to generate a pipeline design working pressure data set;

[0084] Use pressure sensors to monitor the working pressure of the pipeline in real time and transmit the real-time pressure data to the pipeline database;

[0085] Use image acquisition equipment to shoot the outer wall of the pipeline to obtain the image of the outer wall of the pipeline, and establish an image transmission network to transmit the collected image to the pipeline database in real time;

[0086] A data warehouse is used to process the multi-source heterogeneous data of the pipeline, form a consistent data model and format, and build the pipeline real-time database; the multi-source heterogeneous data includes pipeline coordinates, depth, material, pressure, and image; the multi-source heterogeneous data processing includes cleaning, conversion, and loading;

[0087] Based on the real-time pipeline database, correlation analysis and data mining of pipeline attributes are carried out to clarify the analysis objectives, such as pipeline health status assessment, leak detection, and life prediction;

[0088] Use analytical methods and models, such as pipeline degradation models and machine learning algorithms, combined with potential risks and abnormal conditions of the pipeline, and formulate targeted pipeline maintenance and emergency plans based on the analysis results to ensure the safe operation of the pipeline;

[0089] Here is an example:

[0090] When obtaining the geographic coordinates of the pipeline, the differential global positioning system is used. By setting up a base station and a mobile station, the centimeter-level positioning accuracy is calculated using carrier phase observations to ensure the accuracy of the pipeline coordinates.

[0091] According to the pipeline vector data in the differential global positioning system, such as the coordinates of key nodes such as the starting point, end point, and turning point of the pipeline, the actual coordinates obtained by RTK measurement are compared to calculate the average error and maximum error. When the error exceeds the set threshold (for example, 5 cm), a warning is triggered and re-measurement or correction is performed;

[0092] When extracting pipeline buried depth information, the starting depth and ending depth of each pipeline segment can be queried through the attribute data table in the differential global positioning system, the average buried depth of the pipeline segment can be calculated, and the depth value can be associated with the corresponding pipeline coordinates and stored;

[0093] For the extraction of pipeline material information, the regular expression matching method can be used to extract the pipeline material (such as steel pipe, cast iron pipe), pipe diameter (such as DN500), and wall thickness (such as 10mm) from the text description of the pipeline design drawings or construction records, and construct a structured pipeline material attribute table;

[0094] When obtaining the pipeline design working pressure data, text parsing technology can be used to extract the design pressure value (e.g. 1.6MPa) from the pressure parameter section of the design document and map it with the pipeline ID;

[0095] By installing wireless pressure sensors on the pipeline, the working pressure data of the pipeline is collected in real time. When the difference between the measured pressure value and the design pressure value exceeds 10%, an early warning message is generated and pressure adjustment measures are triggered;

[0096] When collecting images of the outer wall of the pipeline, a high-definition digital camera or industrial endoscope can be used to take 360-degree full-scale images of the outer wall of the pipeline at intervals of 1 meter to obtain a high-definition image sequence of the pipeline surface;

[0097] Using an image stitching algorithm, such as the Stitcher class in OpenCV, stitch the image sequence into a complete panoramic image of the outer wall of the pipeline, and using an image compression algorithm (such as JPEG) to compress the panoramic image to reduce the amount of data for image transmission and storage;

[0098] When merging pipeline data, use ETL tools, such as Kettle or DataStage, to extract, clean, and convert multi-source heterogeneous data such as pipeline coordinates, depth, material, pressure, and images, and unify data formats and units. For example, unify the pressure unit to MPa and convert pipeline materials to enumeration types.

[0099] Load the converted data into the MySQL relational database to build a real-time pipeline data warehouse, and establish the association between pipeline ID, coordinates, depth, material, pressure, image and other attributes to form a consistent data model;

[0100] When analyzing pipeline data, pipeline degradation prediction models, such as Weibull distribution or machine learning algorithms (such as support vector machines), are used to predict the remaining life of the pipeline.

[0101] By extracting characteristic parameters such as pipeline material, burial years, and working pressure, the pipeline degradation model is trained. When a new set of pipeline attribute data is input, the model can output the expected life (e.g., 30 years) and failure probability (e.g., 5%) of the pipeline.

[0102] A deep learning algorithm is used. The deep learning algorithm can select a convolutional neural network to analyze the image of the pipeline outer wall, identify defect features such as cracks and corrosion on the pipeline surface, and calculate quantitative indicators such as the length, width, and area of ​​the defect to evaluate the health status of the pipeline. When the defect indicator exceeds the preset threshold, a work task reminder for pipeline maintenance or replacement is generated.

[0103] Furthermore, in this embodiment, step S2 specifically includes:

[0104] Extracting pipeline attribute data from the pipeline real-time database, including service life, material type, and design working pressure, and arranging the attribute data in chronological order to form a time series data set of pipeline attributes;

[0105] Preprocess the time series data set, use box plots to identify outliers, and conduct a comprehensive analysis based on the pipeline's operating status and environmental factors to determine whether the outliers represent an abnormal state of the pipeline, and mark and alarm the abnormal state of the pipeline; preprocess the time series data set including missing value filling and data normalization;

[0106] The LSTM model is used to model and predict the historical change trend of the pipeline design working pressure. By analyzing the periodicity, trend and randomness of the pressure change, the pipeline aging and damage trend is obtained;

[0107] Based on pipeline aging and damage trends and combined with pipeline attribute data, a pipeline crack monitoring model is constructed using machine learning algorithms;

[0108] Use a pipeline endoscope or a crawling robot to obtain crack images, perform denoising, enhancement and segmentation on the crack images, extract the length, width and direction of the cracks, as well as the texture and color within the preset range of the cracks, and construct a crack feature vector;

[0109] Using crack feature vectors, combined with historical crack data of pipelines, and using convolutional neural networks, pipeline crack monitoring models are trained to predict the development direction of pipeline cracks. Predicting the development direction of pipeline cracks includes predicting the expansion speed and direction of pipeline cracks.

[0110] Furthermore, the prediction results of the pipeline crack monitoring model are correlated with the time series data set, and the pipeline health status is comprehensively evaluated by combining the service life, material type, design working pressure and crack development trend of the pipeline, identifying high-risk pipeline sections and generating maintenance or replacement strategies;

[0111] Here is an example:

[0112] Extract the pipeline attribute data from the pipeline real-time database, including service life, material type and design working pressure, and arrange the attribute data in chronological order according to the timestamp to form a time series data set of pipeline attributes;

[0113] The time series dataset is preprocessed by filling missing values ​​with the nearest neighbor interpolation method and mapping the data to the [0,1] interval using the min-max normalization method;

[0114] A box plot is used to identify outliers, and data points beyond 5 times the upper and lower interquartile range are considered outliers. Combined with the operating status of the pipeline (such as flow rate and pressure changes) and environmental factors (such as temperature and humidity), the isolation forest algorithm is used to determine whether the outlier indicates an abnormal state of the pipeline, and the abnormal state is marked and alarmed;

[0115] The LSTM model is used to model and predict the pipeline design working pressure, and the pressure data of the past 30 days are used to predict the pressure changes in the next 7 days;

[0116] By seasonally decomposing the pressure time series, the periodicity, trend and randomness of pressure changes are obtained; the Hurst index is used to determine the long-term correlation of pipeline pressure. A Hurst index greater than 5 indicates long-term persistence, that is, pipeline aging and damage trends;

[0117] According to the pipeline aging and damage trends, combined with the pipeline material type, service life and other attribute data, the support vector machine algorithm is used to build a pipeline crack monitoring model;

[0118] The crack image was obtained through a pipeline endoscope, the non-local mean filtering algorithm was used to remove image noise, the histogram equalization algorithm was used to enhance image contrast, and the crack area was extracted through Otsu threshold segmentation.

[0119] The length, width and direction of the cracks are obtained by geometrically measuring the crack area, and the texture and color features in the 40 pixel × 40 pixel area around the cracks are extracted to form a 72-dimensional crack feature vector.

[0120] Based on the crack feature vector and the historical crack data of the pipeline, the transfer learning method is used to fine-tune the pre-trained ResNet-50 convolutional neural network to train the pipeline crack monitoring model;

[0121] The model performance was evaluated using 5-fold cross validation, with an accuracy rate of over 90%. The trained model was used to predict the expansion speed and direction of pipeline cracks. Cracks with an expansion speed exceeding 2 mm / day or an angle with the pipeline axis of less than 30 degrees were considered high-risk cracks;

[0122] The prediction results of the crack monitoring model are associated with the pipeline attribute time series data, and the pipeline health status is evaluated using the fuzzy comprehensive evaluation method;

[0123] The evaluation indicators include pipeline service life, material type, design working pressure, pressure change trend and crack expansion risk, etc. The pipeline health index is obtained through weighted average;

[0124] Pipeline segments with health indexes lower than 6 are marked as high-risk segments, and maintenance or replacement strategies are generated;

[0125] Based on the pipeline health status assessment results, the Markov decision process is used to dynamically optimize the pipeline inspection and maintenance strategy;

[0126] For predicted high-risk pipeline sections, the inspection frequency will be increased to once a week, and pressure and vibration sensors will be added within 100 meters at both ends of the pipeline section;

[0127] For the predicted rapidly deteriorating pipe sections, when the health index is lower than 4, the replacement strategy is triggered, a work order is automatically generated and the maintenance personnel are notified.

[0128] Furthermore, in this embodiment, step S3 specifically includes:

[0129] The real-time image information of the pipeline is collected by an image acquisition device, and image enhancement and noise removal are performed on the real-time image information; the real-time image information of the pipeline is transmitted to an image preprocessing module for image enhancement and noise removal;

[0130] The processed image is segmented by an image segmentation algorithm to extract pixel information of the pipeline crack area and obtain a binary image of the crack area.

[0131] The edge detection algorithm is used to extract the pixel coordinate information of the crack edge and determine the width and depth parameters of the crack;

[0132] Combined with the geometric structure parameters and spatial position information of the pipeline, the angle of the crack area relative to the central axis of the pipeline is calculated to determine the horizontal angle of the crack to the ground;

[0133] The pixel coordinates of the crack edge are fitted using a curve fitting algorithm to obtain the mathematical expression of the crack curve and calculate the curvature of the crack curve.

[0134] Using graph convolutional neural networks, an image processing model is constructed based on the width, depth, angle, and curvature of the crack. The mapping relationship between crack parameters and pipeline stress is established through feature extraction and parameter learning of the image processing model.

[0135] According to the output results of the image processing model, the real-time stress condition of the outer wall at the crack position of the pipeline is obtained to determine the external stress value of the pipeline at the current moment;

[0136] Use time series analysis algorithm to predict the trend of pipeline stress data and determine the stress change trend of pipeline crack location;

[0137] If the pipeline stress exceeds the preset safety threshold, the early warning mechanism is triggered, an alarm message is sent to the pipeline operation and maintenance personnel, and the specific coordinates and stress parameters of the crack location are output;

[0138] Here is an example:

[0139] A high-resolution industrial camera is used to collect real-time images of the inner wall of the pipeline, with an image resolution of 1920×1080 pixels and a frame rate of 30 frames per second;

[0140] The image preprocessing module uses an adaptive histogram equalization algorithm to enhance the image and improve the image contrast. It also uses a median filter algorithm to remove image noise. The filter window size is 5×5 pixels.

[0141] The image segmentation algorithm adopts a segmentation method based on region growing, sets the grayscale threshold of the seed point to 100, and the growth criterion is that the grayscale difference of the 8-neighborhood pixels is less than 20, extracts the pixel information of the pipeline crack area, and generates a binary image of the crack area;

[0142] The edge detection algorithm uses the Canny operator, the standard deviation of the Gaussian filter is 4, the low threshold is 50, and the high threshold is 150. The pixel coordinates of the crack edge are extracted, and the Euclidean distance between the edge pixels is calculated to obtain the width of the crack as 5 mm and the depth as 2 mm;

[0143] According to the geometric structural parameters of the pipeline, the inner diameter of the pipeline is 500 mm and the wall thickness is 10 mm. By calculating the angle between the geometric center of the pixel coordinates of the crack area and the central axis of the pipeline, it is determined that the horizontal angle of the crack to the ground is 35 degrees. The curve fitting algorithm uses the cubic spline interpolation method. According to the pixel coordinates of the crack edge, the mathematical expression of the crack curve is fitted as y=02x 3 -15x 2 +3x+1, the calculated curvature of the crack curve is 08;

[0144] The image processing model uses a graph convolutional neural network to input parameters such as crack width, depth, angle and curvature. The convolution layer extracts features, the pooling layer reduces dimensions, and the fully connected layer learns parameters to establish a mapping relationship between crack parameters and pipeline stress. The mean square error of the model is 0.2 and the determination coefficient is 98.

[0145] According to the output results of the image processing model, the real-time stress on the outer wall of the pipeline crack location is 16 MPa, which exceeds the yield strength of the pipeline material by 15 MPa, posing a safety hazard.

[0146] The ARIMA time series model is used to predict the trend of pipeline stress data for 30 consecutive days. It is found that the pipeline stress will continue to increase in the next 7 days, with a growth rate of 8 MPa / day. It is expected that the pipeline stress will reach the ultimate strength of the material of 12 MPa on the 5th day, triggering the early warning mechanism and sending alarm information to the pipeline operation and maintenance personnel. At the same time, the spatial coordinates of the crack location (for example, x=2m, y=5m, z=8m) and stress parameters are output to provide data support for pipeline maintenance and reinforcement.

[0147] Furthermore, in this embodiment, step S3 also includes:

[0148] For adjacent pipelines with cracks, the ductility score between the two cracks of the adjacent pipelines is calculated through the real-time image information of the adjacent pipelines. If the ductility score is greater than the threshold, it is judged that the cause of the cracks in the adjacent pipelines is external stress;

[0149] Specifically, a high-resolution camera is used to collect real-time image information of the outer walls of adjacent pipes, obtain a continuous image sequence of the surface of adjacent pipes, and transmit the image data to an image processing server for real-time analysis;

[0150] Preprocess the collected real-time image information of adjacent pipelines, use the image segmentation algorithm based on deep learning to identify and extract the crack area of ​​the preprocessed adjacent pipeline images, obtain the pixel-level masks of the two cracks on the adjacent pipelines, and determine the precise position and shape of the cracks in the image;

[0151] According to the image segmentation results, the pixel distance and angle relationship between the two cracks are calculated. The Zhang Zhengyou calibration method is used to extract the corner coordinates by shooting calibration plate images at different angles and distances, and to construct a mapping relationship between pixel coordinates and world coordinates. The camera's intrinsic and extrinsic parameter matrices are calculated, and the pixel distance is converted into the actual distance. The actual spatial distance and relative direction between the cracks are calculated, and the positional relationship between the cracks is quantified.

[0152] Extract the morphological features of the two cracks, including geometric parameters such as length, width, and curvature of the cracks, as well as texture features such as texture and color of the crack edges, and construct a crack morphological feature vector;

[0153] Using machine learning algorithms, a pre-ductility assessment model is constructed based on the spatial relationship characteristics and morphological characteristics between cracks;

[0154] The ductility assessment model collects a large amount of pipeline crack case data, including crack location, morphology, stress state, and material property information, and cleans, annotates, and enhances the data;

[0155] Select key factors affecting crack extension, such as crack length, width, spacing, angle, and stress distribution, and construct feature vectors;

[0156] Using the transfer learning method, the CNN model pre-trained on the ImageNet large image dataset is used to extract high-level semantic features of crack images, and then combined with the geometric and physical features of the cracks to train ductility assessment models, such as support vector machines and random forests;

[0157] Through cross-validation and grid search methods, the model's hyperparameters and feature subsets are optimized to improve the model's generalization performance. The model conducts a comprehensive analysis of the input crack characteristics, calculates the ductility score between two cracks, and evaluates the possibility and severity of crack expansion.

[0158] Based on historical data, the danger level of crack extension is divided into low risk, medium risk and high risk, corresponding to different scoring thresholds;

[0159] Determine whether the ductility score exceeds the preset threshold. If the ductility score exceeds the high-risk threshold, it is considered that the risk of crack extension is very high and further analysis of the source and magnitude of external stress is required;

[0160] The stress distribution of the pipeline at the crack location is calculated by finite element analysis, and factors such as the pipeline's burial environment, load conditions, and material properties are considered to determine whether the direction and magnitude of the external stress exceeds the strength limit of the pipeline;

[0161] If the external stress exceeds the strength limit, the position relationship and morphological characteristics of the cracks are combined with the material properties and stress distribution of the adjacent pipelines to determine that the cause of the adjacent pipeline cracks is the external stress, generate a crack cause analysis report, and trigger the corresponding early warning and maintenance decision-making mechanism;

[0162] If the ductility score does not exceed the threshold or the external stress does not exceed the strength limit, the cracks in the adjacent pipelines are considered to be independent individuals, and the mutual influence between the cracks is not considered for the time being. Each crack continues to be monitored and evaluated separately;

[0163] Here is an example:

[0164] When collecting images of adjacent pipelines, an industrial camera with a resolution of 4096×2160 and an industrial lens with a focal length of 50mm are used to scan and image the surface of adjacent pipelines. The camera's acquisition frame rate is set to 20 frames per second, and the images are transmitted to the server for processing via the optical fiber network.

[0165] During image preprocessing, the non-local mean filtering algorithm is used to remove Gaussian noise, the CLAHE algorithm is used to enhance local contrast, and the perspective distortion of the image is corrected by perspective transformation;

[0166] In image segmentation, the U-Net network is used to extract crack areas. The encoder part of the network uses the first four convolution blocks of ResNet34, and the decoder part uses four upsampling convolution layers. The sizes of the feature maps are 256×256, 128×128, 64×64, and 32×32 respectively.

[0167] The network loss function uses binary cross entropy, the optimizer uses Adam, the initial learning rate is 0.001, the batch_size is 16, and the training is 100 epochs;

[0168] When calculating the spatial relationship of cracks, a 7×9 chessboard is used for camera calibration. The coordinates of the chessboard corners are extracted through Harris corner detection and sub-pixel refinement. The camera's intrinsic parameter matrix and distortion coefficient are calculated using the Levenberg-Marquardt optimization algorithm, and the camera's extrinsic parameter matrix is ​​solved using the PnP algorithm. The pixel coordinates of the crack area are projected into the world coordinate system, and the Euclidean distance and angle between the cracks are calculated. The distance accuracy is better than 1mm, and the angle accuracy is better than 0.5°.

[0169] When extracting crack morphological features, morphological filtering is performed on the crack mask image to extract the skeleton line and edge contour of the crack, calculate the geometric parameters of the crack such as length, width, direction, curvature, and the texture parameters such as grayscale mean and variance of the pixels on the skeleton line, and construct a 36-dimensional crack morphological feature vector. When training the ductility assessment model, 500 sets of pipeline crack sample data are collected. Each set of data contains information such as images, spatial relationships, morphological features, and stress states of 2 cracks, and is randomly divided into training sets, validation sets, and test sets in a ratio of 7:2:1;

[0170] ResNet50 is used as the backbone network to extract the 2048-dimensional feature vector of the crack image. After being spliced ​​with the spatial relationship features and morphological features of the crack, it is input into a three-layer fully connected neural network for ductility score prediction. The number of hidden layer nodes of the network is 512, 256 and 128 respectively. The output layer uses the sigmoid activation function to predict the probability of crack extension.

[0171] The loss function of the model uses cross entropy, the optimizer uses SGD, the initial learning rate is 0.01, the momentum factor is 0.9, the weight decay coefficient is 0.0001, and the training is 50 epochs. When setting the ductility score threshold, according to the distribution of crack extension probability, the extension risk is divided into three levels, namely low risk (probability less than 0.3), medium risk (probability between 0.3 and 0.7) and high risk (probability greater than 0.7), and the threshold of the high risk level is set to 0.7;

[0172] In the external stress analysis, ANSYS finite element analysis software was used to establish a three-dimensional solid model of the pipeline and its buried environment. The pipeline material was Q235 steel, with an elastic modulus of 200 GPa, a Poisson's ratio of 0.3, and a yield strength of 235 MPa.

[0173] The working pressure inside the pipeline (the value range is 110MPa) and the external soil pressure (the value range is 50200kPa) are applied to the crack location, and the vonMi ses equivalent stress distribution at the crack is calculated by static analysis. When the equivalent stress exceeds the yield strength of the material, it is judged that the external stress exceeds the strength limit of the pipeline, which may lead to the generation and expansion of cracks.

[0174] Furthermore, in this embodiment, in step S4, performing spectrum analysis on the vibration signal includes:

[0175] An acceleration sensor is used to collect the vibration signal of the outer wall of the pipeline, and the collected time-domain vibration signal is analyzed and preprocessed; the sampling frequency of the sensor is set to 10kHz, and the sampling time is 10 seconds; the collected time-domain vibration signal is transmitted to the computer through a data acquisition card for analysis and preprocessing. The preprocessing includes selecting a 4th-order Butterworth bandpass filter according to the frequency characteristics of pipeline vibration, with a lower cutoff frequency of 10Hz and an upper cutoff frequency of 1000Hz to filter out high-frequency noise and ultra-low-frequency interference components in the vibration signal;

[0176] The pre-processed pipeline vibration signal is converted into frequency domain using the fast Fourier transform algorithm to obtain the spectrum distribution diagram of pipeline vibration. The spectrum resolution is set to 0.1 Hz, and the spectrum range covers 0 to 5000 Hz. The main frequency components and amplitudes of pipeline vibration can be intuitively observed through the spectrum diagram.

[0177] According to the spectrum distribution diagram, the distribution law of the spectrum peak amplitude is obtained through statistical analysis of the pipeline vibration data;

[0178] According to the mean and standard deviation, the threshold of the spectrum peak is set as: mean + 2 times the standard deviation, and the corresponding frequency and amplitude parameters are extracted to obtain the characteristic spectrum peak parameters of pipeline vibration; wherein, the threshold of the spectrum peak is set as: mean + 2 times the standard deviation because the spectrum peak with an amplitude greater than the threshold is considered to be a significant peak;

[0179] Perform short-time Fourier transform on the pipeline vibration signal to obtain the time-frequency distribution diagram of the vibration signal. Set the window function of the short-time Fourier transform. Through the time-frequency diagram, the transient characteristics and energy distribution changes of the pipeline vibration can be analyzed, and the transient characteristic parameters of the pipeline vibration signal can be extracted. The window function of the short-time Fourier transform is the Hanning window, the window length is 1 second, and the overlap rate is 50%;

[0180] The order analysis method based on short-time Fourier transform is used to divide the pipeline vibration signal into several time windows, and the signal in each window is subjected to fast Fourier transform to obtain the spectrum diagram at different times; the speed frequency range is set to 0100Hz, corresponding to the speed range of the pipeline is 0-6000rpm, and the frequency range is divided into several equally spaced frequency bands with a bandwidth of 0.1Hz, corresponding to a speed resolution of 6rpm;

[0181] For the spectrum graph within each time window, find the frequency point with the largest amplitude in each frequency band as the characteristic frequency of the frequency band;

[0182] Arrange the characteristic frequencies of each frequency band from small to large frequency to obtain a set of frequency sequences corresponding to the order components of pipeline vibration;

[0183] Furthermore, the amplitude corresponding to the characteristic frequency in each frequency band is calculated to obtain the amplitude sequence of the order components, which reflects the vibration intensity of the pipeline vibration at different rotation speeds;

[0184] Use non-destructive testing technologies such as ultrasonic testing or eddy current testing to measure the stress distribution at the crack location of the pipeline, obtain the stress cloud map at the crack, extract the characteristic parameters such as the stress gradient and stress concentration factor of the crack, and obtain the morphological characteristics of the crack through regular pipeline endoscope testing, such as the geometric parameters such as the length, width, and depth of the crack, as well as the dynamic parameters such as the expansion speed and change trend of the crack;

[0185] The partial least squares regression (PLSR) method is used to establish the correlation model between pipeline vibration parameters and crack state parameters. The pipeline vibration parameters (characteristic spectrum peaks, order components, etc.) and crack state parameters (stress characteristics, morphology characteristics, etc.) are respectively composed of independent variable matrix X and dependent variable matrix Y. The independent variable matrix X and dependent variable matrix Y are centered and standardized to eliminate the dimension effect.

[0186] Through cross-validation test, such as leave-one-out cross-validation, the optimal number of latent variables for the PLSR model is determined, and the number of latent variables corresponding to the minimum sum of squared errors is taken;

[0187] Calculate the latent variable matrices T and U of the independent variable matrix X and the dependent variable matrix Y, as well as the corresponding loading matrices P and Q, and establish the regression relationship U=TB+E between the latent variables, where B is the regression coefficient matrix and E is the residual matrix. Map the regression coefficient matrix B to the original variable space to obtain the regression coefficient matrix between the independent variables and the dependent variables, and evaluate the contribution and influence direction of each independent variable on the dependent variable.

[0188] According to the goodness of fit and prediction ability of the PLSR model, such as the coefficient of determination R^2 and the root mean square error RMSE index, the validity and reliability of the model are evaluated and used for subsequent pipeline crack trend prediction and risk assessment;

[0189] According to the prediction results of the PLSR model, combined with the change trends of pipeline vibration parameters and crack state parameters, the health status of the pipeline is comprehensively evaluated and a pipeline risk assessment matrix is ​​constructed;

[0190] According to the pipeline crack propagation speed and stress level predicted by the PLSR model, it is divided into three levels: low speed / low stress, medium speed / medium stress, and high speed / high stress;

[0191] According to the change trend of pipeline vibration parameters, such as the frequency shift of characteristic spectrum peak and the amplitude increment of order component, it is divided into three levels: small change, medium change and large change;

[0192] A risk matrix is ​​established, with the horizontal axis representing the crack extension level and the vertical axis representing the vibration change level. Based on the combination of the two indicators, the risk is divided into four levels: low risk, medium risk, high risk, and extremely high risk.

[0193] For pipelines of different risk levels, formulate corresponding maintenance strategies, such as regular inspections and status monitoring for low-risk pipelines, increased inspection frequency and accuracy for medium-risk pipelines, local repair or reinforcement for high-risk pipelines, and overall replacement or shutdown for maintenance for extremely high-risk pipelines;

[0194] According to the number, location, risk level and other factors of pipeline cracks, the analytic hierarchy process (AHP) is used to quantitatively rank the urgency of pipeline maintenance, determine maintenance priorities, and rationally allocate maintenance resources and manpower to improve the timeliness and pertinence of pipeline maintenance;

[0195] Here is an example:

[0196] When collecting pipeline vibration signals, a piezoelectric acceleration sensor with a sensitivity of 100mV / g and a frequency response range of 0.5Hz-5000Hz is selected. The sensor is adsorbed on the outer wall of the pipeline by magnetism or epoxy glue. The sensor output is connected to a 24-bit AD acquisition card, and the sampling frequency is set to 10kHz. The vibration signal is collected for 10 seconds each time, and the data is collected every 1 hour.

[0197] During signal preprocessing, the collected vibration signal is first subjected to DC removal processing, and then the signal is filtered using a 4th-order Butterworth bandpass digital filter with a lower cutoff frequency of 10 Hz and an upper cutoff frequency of 1000 Hz. The filtered signal is amplitude normalized;

[0198] During spectrum analysis, the FFT algorithm is used to perform 4096-point spectrum conversion on the preprocessed vibration signal, with a spectrum resolution of 2.4 Hz. The spectrum amplitudes of 100 sets of pipeline vibration data are statistically analyzed, and the mean of the spectrum peak amplitude is 0.15 and the standard deviation is 0.05. The threshold is set to 0.25, and the spectrum peaks with amplitudes greater than 0.25 are extracted as characteristic peaks.

[0199] In time-frequency analysis, the STFT algorithm is used to transform the vibration signal, with a window length of 1 second, a corresponding frequency resolution of 1 Hz, an overlap rate of 50%, and the average amplitude of the spectrum is calculated in each time window to obtain the time-frequency energy distribution diagram of the vibration signal;

[0200] During order analysis, according to the speed range of the pipeline, the frequency range of the order analysis is set to 0100Hz, and the bandwidth is 0.1Hz, then the corresponding order resolution is 0.1 order, the characteristic frequency and corresponding amplitude in each frequency band are extracted, and the order slice spectrum within 15 orders is generated;

[0201] When detecting pipeline cracks, ultrasonic phased array detection technology is used to obtain a panoramic image of the crack area through linear scanning, and the outline of the crack is extracted using an image segmentation algorithm. The length, width and depth parameters of the crack are calculated, and the stress intensity factor at the crack tip is measured using the ultrasonic diffraction time difference method. At the same time, an eddy current array sensor is used to measure the stress field distribution around the crack.

[0202] In the analysis of the correlation between vibration and cracks, the PLSR algorithm is used to construct a regression model of vibration parameters (spectral peak, order components, etc.) and crack parameters (size, stress, etc.). The optimal number of latent variables is selected as 3 through leave-one-out cross-validation, and the vibration parameter matrix X (100×8) and the crack parameter matrix Y (100×4) are established. The matrices X and Y are automatically mean-centered and variance-normalized, and the latent variable matrices T (100×3) and U (100×3) of X and Y, as well as the load matrices P (8×3) and Q (4×3) are calculated. The multivariate linear regression equation is established through the latent variables T and U, with a determination coefficient R2 of 0.85 and a root mean square error RMSE of 0.12.

[0203] In pipeline risk assessment, according to the crack propagation rate and stress amplitude predicted by the PLSR model, they are divided into three levels: low speed / low amplitude (0-0.1 mm / month, 0-50 MPa), medium speed / medium amplitude (0.1-0.5 mm / month, 50-100 MPa), and high speed / high amplitude (>0.5 mm / month, >100 MPa). According to the frequency shift of the spectrum peak and the change in the order amplitude, they are divided into three levels: small change (0-1 Hz, -10%-10%), medium change (1-5 Hz, -50%--10% or 10%-50%), and large change (>5 Hz, <-50% or >50%). A 3×3 risk matrix is ​​constructed to quantify the risk level of the pipeline.

[0204] According to the AHP hierarchical model, a pipeline crack risk assessment index system was established. The indicators were divided into four levels: crack size (length, width, depth), crack location (longitudinal, circumferential, oblique), number of cracks (single, multiple), and risk level (low, medium, high, and extremely high). A judgment matrix was constructed through pairwise comparison, and the weight vector of the indicators was calculated. The risk ranking and maintenance priority evaluation of the pipeline were carried out to quantify the urgency of pipeline maintenance.

[0205] Furthermore, in this embodiment, step S4 also includes:

[0206] For adjacent pipelines with cracks, the resonance phenomenon of the adjacent pipelines is identified through the vibration signals of the adjacent pipelines; if the resonance phenomenon of the adjacent pipelines exists, the urgency of pipeline maintenance is increased when quantifying the urgency of pipeline maintenance;

[0207] Specifically, an acceleration sensor is used to synchronously collect vibration signals of adjacent pipelines with cracks. The sensor is installed at the same position of the adjacent pipelines, the sampling frequency is set to 10 kHz, and the sampling time is 10 seconds to obtain the time domain vibration signal data of the adjacent pipelines;

[0208] The collected vibration signals of adjacent pipelines were preprocessed, including removing the DC component of the signal, filtering out high-frequency noise and eliminating outliers. A 4th-order Butterworth bandpass filter was used with a lower cutoff frequency of 10 Hz and an upper cutoff frequency of 1000 Hz to filter out interference components in the vibration signal.

[0209] The fast Fourier transform algorithm is used to convert the preprocessed adjacent pipeline vibration signals into frequency domain to obtain the spectrum distribution diagram of the vibration signal. The spectrum resolution is set to 0.1 Hz, and the spectrum range covers 0 to 5000 Hz. The main frequency components and amplitudes of the adjacent pipeline vibration can be observed through the spectrum diagram.

[0210] Compare the frequency spectrum distribution diagrams of adjacent pipeline vibration signals, refer to pipeline vibration standards and specifications, such as ISO10816 series standards, select appropriate vibration velocity or vibration displacement thresholds according to pipeline type, power, speed and other parameters as the basis for resonance judgment, and consider the material, size and support method of the pipeline, correct and optimize the resonance criteria of different pipelines, identify the resonance peak in the frequency spectrum, and if the amplitudes of adjacent pipelines at a certain frequency all exceed the resonance criterion threshold at that frequency, it is determined that resonance exists at that frequency;

[0211] Extract characteristic parameters such as resonance frequency, resonance amplitude and resonance phase from the vibration signals of adjacent pipelines. Use finite element analysis to establish a dynamic model of the pipeline based on the geometric parameters and material properties of the pipeline, calculate the natural frequency and vibration mode of the pipeline, and measure the frequency response function of the pipeline under different excitation conditions through on-site vibration testing. Verify and correct the finite element model to improve the accuracy of the natural frequency calculation. Compare the resonant frequency with the natural frequency of the pipeline, and set a frequency deviation threshold, such as 5%. When the relative error between the two is less than the threshold, the resonant frequency is considered to be close to the natural frequency, indicating that resonance may induce the natural vibration mode of the pipeline and increase the vibration risk of the pipeline.

[0212] The resonance amplitude is compared with the critical vibration amplitude of the pipeline. If it exceeds the critical value, it indicates that the resonance may cause excessive displacement and stress of the pipeline, resulting in fatigue damage to the pipeline. According to the resonance characteristic parameters of the vibration signals of adjacent pipelines, the pipeline resonance risk assessment index system is established by using the hierarchical analysis method. The target layer is the pipeline resonance risk assessment, the criterion layer includes resonance characteristics, pipeline attributes and environmental factors, and the index layer includes resonance frequency, resonance amplitude, resonance phase, pipeline material, pipeline diameter, pipeline wall thickness, pipeline pressure, pipeline temperature, pipeline medium, support stiffness and geological conditions.

[0213] Through pairwise comparison, the judgment matrix of indicators at each level is constructed, the weight vector of the indicators is calculated, and the consistency test is carried out. The membership function of each indicator and the weight vector are weighted and synthesized by fuzzy comprehensive evaluation method to obtain the comprehensive membership of pipeline resonance risk. According to the principle of maximum membership, the resonance risk level of the pipeline is determined and divided into four levels: low risk, medium risk, high risk and extremely high risk.

[0214] When quantifying the urgency of pipeline maintenance, the weighted average method is used to comprehensively consider the resonance risk level and the crack risk level to construct a two-dimensional risk matrix. The horizontal axis is the resonance risk level, the vertical axis is the crack risk level, and the matrix element is the maintenance urgency, which is divided into four levels: general, relatively urgent, urgent and very urgent.

[0215] If there is resonance between adjacent pipelines, and the resonance risk level is high risk or extremely high risk, the urgency of pipeline maintenance will be increased to urgent or very urgent according to the position of the pipeline in the risk matrix. When determining the maintenance priority, factors such as the importance, maintainability, and resource constraints of the pipeline need to be considered. Multi-objective optimization algorithms, such as genetic algorithms and particle swarm algorithms, are used to generate optimal maintenance plans and strategies, giving priority to the inspection and reinforcement of high-risk and extremely high-risk pipelines to avoid pipeline accidents caused by resonance.

[0216] In order to further improve the accuracy of resonance identification and risk assessment, machine learning algorithms such as support vector machines and random forests are introduced. The time domain, frequency domain, and time-frequency domain features of pipeline vibration signals are used as inputs, and the resonance risk level is used as output. The classification model is trained to achieve automatic identification and early warning of pipeline resonance risks.

[0217] Adopt data fusion technology to correlate pipeline vibration data with multi-source heterogeneous data such as pipeline structural parameters, inspection records, and maintenance history, explore the intrinsic relationship between pipeline vibration and pipeline status, establish pipeline health diagnosis and prediction models, and guide pipeline safe operation and maintenance decisions;

[0218] Here is an example:

[0219] When collecting vibration signals from adjacent pipelines, a piezoelectric acceleration sensor with a sensitivity of 100mV / g and a frequency response range of 0.5Hz-5000Hz was selected. The sensor was adsorbed on the outer wall of the adjacent pipeline at four locations 1 meter and 2 meters away from the crack by magnetic force or epoxy glue. The sensor output was connected to a 24-bit AD acquisition card with a sampling frequency of 10kHz. The vibration signal was collected for 10 seconds each time, and the data was collected every hour.

[0220] During signal preprocessing, the 4 collected vibration signals are first subjected to DC removal, and then filtered using a 4th-order Butterworth bandpass digital filter with a lower cutoff frequency of 10 Hz and an upper cutoff frequency of 1000 Hz. The filtered signals are then amplitude normalized.

[0221] During spectrum analysis, the fast Fourier transform algorithm is used to perform 4096-point spectrum conversion on the four pre-processed vibration signals, with a spectrum resolution of 2.4Hz. Referring to the ISO10816-3 standard, for rigidly supported pipes with a rotation speed of 600rpm1800rpm, the vibration velocity threshold is 4.5mm / s. The spectrum peaks below the threshold are considered normal peaks, and the spectrum peaks above the threshold are considered resonance peaks.

[0222] When calculating the natural frequency, ANSYS finite element analysis software was used to establish the beam unit model of the adjacent pipelines. The pipeline material was Q235 steel with an elastic modulus of 210 GPa and a density of 7850 kg / m 3 , Poisson's ratio is 0.3, the pipe diameter is 219 mm, the wall thickness is 6 mm, and the pipe length is 10 m. The first 10 natural frequencies of the pipe are calculated and compared with the field vibration test results. If the relative error is less than 5%, the finite element model is considered to be valid;

[0223] In the resonance risk assessment, the AHP method is used to construct a three-layer indicator system. Three indicators are set at the criterion layer, namely resonance feature A1, pipeline attribute A2 and environmental factor A3, with a weight vector of (0.5, 0.3, 0.2). Three sub-indicators are set under the resonance feature, namely resonance frequency B1, resonance amplitude B2 and resonance phase B3, with a weight vector of (0.4, 0.4, 0.2). Four sub-indicators are set under the pipeline attribute, namely pipeline material C1, pipeline diameter C2, pipeline wall thickness C3 and pipeline pressure C4, with a weight vector of (0.2, 0.3, 0.3, 0.2). Two sub-indicators are set under the environmental factor, namely support stiffness D1 and geological condition D2, with a weight vector of (0.6, 0.4). Through comprehensive weighted calculation, the membership vector of pipeline resonance risk is obtained, and the level corresponding to the maximum membership is taken as the resonance risk level of the pipeline.

[0224] When quantifying the maintenance urgency, a 4×4 risk matrix is ​​used, and the weights of the resonance risk level and the crack risk level are 0.6 and 0.4 respectively. If the pipeline is at both a high risk of resonance and a high risk of crack, its maintenance urgency is very urgent and has the highest priority.

[0225] When machine learning modeling is used, vibration monitoring data of 100 pipelines are selected. 1,000 sets of data are collected from each pipeline. Each set of data contains vibration signals from 4 sensors, and each signal contains 10,000 sampling points. 10 time domain features, 20 frequency domain features and 5 wavelet coefficients are extracted for each signal, and a total of 140 features are extracted. The pipeline resonance risk level is used as a label, and 10-fold cross validation is used. Support vector machine and random forest algorithm are used for classification. The support vector machine uses radial basis kernel function, and the penalty factor C is 10. The random forest algorithm uses 100 decision trees with a maximum tree depth of 5. The hyperparameters are optimized through grid search to make the classification accuracy reach more than 95%;

[0226] Using the same method, a pipeline crack risk prediction model and a pipeline remaining life prediction model are established to achieve comprehensive diagnosis and early warning of pipeline status.

[0227] Furthermore, in this embodiment, step S5 specifically includes:

[0228] Extract data on pipeline maintenance urgency, pipeline maintenance budget, and pipeline maintenance cost from the pipeline maintenance decision system; the pipeline maintenance urgency is calculated by a pipeline status assessment model based on machine learning, with pipeline inspection data (such as pipeline wall thickness, corrosion rate), environmental data (such as soil moisture, temperature), and historical accident data as input features of the model, and pipeline status level or maintenance urgency as output of the model. By training and optimizing support vector machines or random forest models, intelligent assessment and prediction of pipeline status are achieved; the pipeline maintenance budget is determined based on the annual maintenance plan and financial allocation of the area where the pipeline is located, in units of 10,000 yuan; the pipeline maintenance cost is estimated based on factors such as the material, diameter, burial depth, soil quality, etc. of the pipeline, in units of 10,000 yuan

[0229] The data of pipeline maintenance urgency, pipeline maintenance budget and pipeline maintenance cost are processed; the processing adopts the range normalization method within the index, calculates the maximum and minimum values ​​of each index respectively, maps the data to the interval [0, 1], eliminates the influence of dimension and order of magnitude differences, and the normalization formula is: x'=(x-min) / (max-min), where x is the original data, x' is the normalized data, and max and min are the maximum and minimum values ​​of the index respectively;

[0230] A pipeline maintenance decision model is established by using a grey correlation analysis method. The pipeline maintenance urgency, pipeline maintenance budget and pipeline maintenance cost are used as input parameters of the model, and the pipeline maintenance priority is used as the output parameter of the model. The importance weight of each parameter is determined by calculating the grey correlation between each input parameter and the output parameter. The greater the correlation, the greater the influence of the parameter on the output parameter, and the higher the weight.

[0231] The grey correlation coefficients of pipeline maintenance urgency, pipeline maintenance budget and pipeline maintenance cost are calculated; the grey correlation coefficient reflects the correlation between each parameter and pipeline maintenance priority, and the calculation formula is:

[0232] ξ[X 0 (k),X i (k)]=[min i min k |X 0 (k)-X i (k)|+ρmax i max k |X 0 (k)-X i (k)|] /

[0233] [|X 0 (k)-X i (k)|+ρmax i max k |X 0 (k)-X i (k)|]

[0234] Among them, ξ is the grey relational coefficient, X 0 is the reference sequence (pipeline maintenance priority), X i is the comparison series (pipeline maintenance urgency, pipeline maintenance budget, and pipeline maintenance cost), ρ is the resolution coefficient, the value range is [0, 1], usually 0.5, and k represents the position variable;

[0235] According to the grey correlation coefficient, the grey correlation of pipeline maintenance urgency, pipeline maintenance budget and pipeline maintenance cost is calculated, and the grey correlation is normalized to obtain the weight vector of each parameter. The larger the weight, the greater the influence of the parameter on the pipeline maintenance priority. The grey correlation reflects the comprehensive influence of each parameter on the pipeline maintenance priority. The calculation formula is:

[0236]

[0237] where r i is the grey relational degree of the i-th parameter, n is the number of parameters;

[0238] Performing a weighted summation on the standardized values ​​of pipeline maintenance urgency, pipeline maintenance budget, and pipeline maintenance cost and the weight vector to obtain a comprehensive evaluation value of the pipeline maintenance priority;

[0239] The pipelines are sorted according to the comprehensive evaluation values ​​to obtain a pipeline maintenance priority list; the pipeline with the highest priority is scheduled for maintenance first;

[0240] If the maintenance priority of a certain section of pipeline changes, the maintenance priority of adjacent pipelines will be adjusted in conjunction with the pipeline network topology, combined with the pipeline importance coefficient and risk level coefficient;

[0241] Among them, the pipeline importance coefficient is calculated based on the spatial location, flow rate, and supply object of the pipeline, and the network topology indicators (such as betweenness centrality and connectivity reliability) are used to quantify the importance of the pipeline;

[0242] The pipeline risk level coefficient is calculated based on the pipeline status assessment results and accident risk assessment results. The risk matrix method is used to comprehensively consider the possibility and severity of pipeline accidents, and the pipeline risk level is quantified into low risk, medium risk, and high risk.

[0243] For pipelines whose maintenance priority has changed, the adjustment range of their priority is calculated based on their importance coefficient and risk level coefficient. The higher the importance coefficient and the higher the risk level, the greater the adjustment range of the priority.

[0244] According to the pipeline network topology, the network diffusion model is used to transfer the adjustment range of pipeline priority to adjacent pipelines, forming a spatial correlation effect of maintenance priority to ensure the continuity and systematicness of pipeline maintenance.

[0245] In order to further improve the scientificity and explainability of pipeline maintenance decisions, the grey correlation analysis is combined with multi-criteria decision-making methods such as the analytic hierarchy process (AHP) and the fuzzy comprehensive evaluation method to construct a comprehensive evaluation model for pipeline maintenance decisions. From the perspectives of pipeline status, maintenance effect, maintenance cost, and maintenance risk, a quantitative evaluation mechanism for pipeline maintenance priority is formed.

[0246] Introducing visualization technologies such as GIS and BIM, combining pipeline maintenance decision results with pipeline geographic spatial information, 3D models, etc., to intuitively display the spatial distribution and schedule of pipeline maintenance, provide visual auxiliary decision support for pipeline operation and maintenance personnel, and improve pipeline maintenance efficiency and refined management level;

[0247] Here is an example:

[0248] When evaluating the pipeline status, pipeline attribute data, including pipeline material, diameter, burial depth, and service life, are extracted from the pipeline geographic information system (GIS). Pipeline wall thickness data are extracted from the pipeline detection system. Ultrasonic thickness measurement is used to measure every 1 meter to form a pipeline wall thickness distribution curve. Pipeline corrosion rate data are extracted from the pipeline corrosion monitoring system. Resistive corrosion probes are used, one is arranged every 2 meters, and the instantaneous corrosion rate of the pipeline is monitored in real time. At the same time, environmental data such as pH value, moisture content, resistivity, etc. of the soil around the pipeline, as well as historical pipeline leakage, pipe burst and other accident data are collected to construct a pipeline status assessment data set.

[0249] The support vector machine (SVM) algorithm is used, with basic pipeline attributes, test data, and environmental data as feature vectors, and pipeline status levels (such as normal, mild defects, moderate defects, and severe defects) as classification labels. The SVM model is trained using the radial basis kernel function and grid search to optimize model parameters. With 5-fold cross validation, the model accuracy reaches more than 90%;

[0250] The new pipeline data is used to predict the state and obtain the pipeline maintenance urgency score, ranging from 0 to 100 points. The higher the score, the worse the pipeline state and the more urgent the maintenance.

[0251] Based on the number, length, and importance of pipelines in each region, combined with the annual financial budget, determine the amount of pipeline maintenance funds for each region. For example, the 100-kilometer pipeline in Area A accounts for 20% of the total length of pipelines in the city, and is responsible for the water supply task in the main urban area. The maintenance fund accounts for 30%, about 15 million yuan. Based on the previous pipeline maintenance ledger data, the average maintenance unit price of pipelines of different diameters, materials, and burial depths is calculated, and the pipeline maintenance cost is estimated by considering factors such as construction difficulty and price increases. For example, the average maintenance unit price of DN500 ductile iron pipes is 2 meters deep and is 2,000 yuan per meter. If the length of the pipeline is 1 kilometer, the estimated maintenance cost is 2 million yuan;

[0252] The pipeline maintenance urgency, maintenance budget, and maintenance cost indicators are standardized. For each pipeline section, the urgency standardized value = (urgency score - the lowest score of all pipelines) / (the highest score of all pipelines - the lowest score), the budget standardized value = (pipeline section budget amount - pipeline section minimum budget) / (pipe section maximum budget - minimum budget), and the cost standardized value is calculated similarly;

[0253] When calculating the grey correlation between pipeline maintenance level and each indicator, a reference sequence X0 = {x0(k)} = {maintenance level of pipeline 1, maintenance level of pipeline 2, ...} is constructed, and a comparison sequence X1 = {x1(k)} = {standardized value of pipeline 1 urgency, standardized value of pipeline 2 urgency, ...}, X2 = {standardized value of pipeline 1 budget, ...}, X3 = {standardized value of pipeline 1 cost, ...} is constructed. ρ = 0.5 is taken to calculate the grey correlation coefficients ξ(X1, X2, X3) of X0, X1, X2, X3 and X3 respectively. 0 ,X 1 ),ξ(X 0 ,X 2 ),ξ(X 0 ,X 3 ), and then calculate the corresponding grey relational degree r 1 、r 2 、r 3，According to the correlation degree, the weights of urgency, budget and cost on pipeline maintenance level are determined to be 0.5, 0.3 and 0.2 respectively;

[0254] The standardized values ​​of the urgency, budget, and cost of each pipeline section are multiplied by the weights and then added to obtain the comprehensive evaluation value of the pipeline maintenance priority. For example, the standardized value of the urgency, budget, and cost of a pipeline section is 0.8, 0.6, and 0.4, respectively. Then, the comprehensive evaluation value of its maintenance priority = 0.8 × 0.5 + 0.6 × 0.3 + 0.4 × 0.2 = 0.66. The pipeline maintenance priority list is obtained by sorting the comprehensive evaluation values ​​from large to small.

[0255] When the maintenance priority of a pipeline changes, the PageRank algorithm is used to calculate the importance of the pipeline in the network, such as the number of users connected to the pipeline, the proportion of traffic, and the number of downstream key facilities (such as hospitals and schools), which are quantified as an importance coefficient of 0.1. The environmental risk and natural disaster risk of the area where the pipeline is located are calculated by the analytic hierarchy process (AHP), which are quantified as risk levels of 1 to 5. The pipeline maintenance priority is weighted and modified. For a pipeline with an importance coefficient of 0.8 and a risk level of 4, its maintenance priority is increased by 1 level, and the priority of the adjacent upstream and downstream pipelines is also increased by 0.5 level, forming an associated transmission effect of pipeline maintenance.

[0256] Combining pipeline maintenance priorities with GIS pipeline network maps, a pipeline maintenance planning map is generated to visually display the importance of pipeline maintenance in various areas of the city, guide the temporal and spatial arrangements of pipeline inspections, and achieve scientific, refined, and visual pipeline maintenance decision-making.

[0257] Furthermore, in this embodiment, step S6 specifically includes:

[0258] Based on the pipeline maintenance priority list generated by the pipeline maintenance decision system, the pipeline maintenance priority evaluation index system is constructed by using the analytic hierarchy process from multiple dimensions including pipeline importance, failure risk and maintenance cost.

[0259] By comparing two indicators, the weight of each indicator is calculated to form a comprehensive score of pipeline maintenance priority. According to the score distribution, the threshold intervals of high and low priority are divided, and the priority threshold is set. The pipelines with maintenance priorities higher than the threshold are classified as high-priority pipelines, and the pipelines with maintenance priorities lower than or equal to the threshold are classified as low-priority pipelines. In addition, pipeline maintenance strategies of different degrees are associated with pipeline segments with different maintenance priorities.

[0260] The pipeline maintenance strategy includes using bioremediation agents to repair cracks in high-priority pipelines and taking regular monitoring and non-invasive detection protection measures for cracks in low-priority pipelines.

[0261] For high-priority pipelines, environmentally friendly bioremediation technology is used to repair pipeline cracks. A comprehensive evaluation is conducted from the aspects of microbial species, temperature adaptability, pH tolerance, matrix affinity, etc. The formulation of the microbial agent is optimized through orthogonal experiments. While improving the repair efficiency, the negative impact on the pipeline material and the surrounding environment is minimized. Safe and non-toxic microbial agents with little impact on the environment, such as sulfate-reducing bacteria and iron-oxidizing bacteria, are selected. By adding microbial agents to the medium near the cracks, the metabolic activities of microorganisms are used to form dense mineral nodules on the crack surface, gradually filling and sealing the cracks.

[0262] The safety and effectiveness of the microbial agent are verified in small-scale pipelines, and then gradually extended to the entire network of pipelines. The bioremediation agent solution is injected into the cracks of the pipeline using the high-pressure injection method. A high-pressure pump such as a screw pump or a plunger pump is selected. The injection pressure is controlled at 15MPa and the injection speed is controlled at 0.1-0.5m / s. The diameter and number of the injection nozzles are adjusted to achieve uniform distribution of the repair agent in the cracks. At the same time, the injection range and time are estimated according to the crack depth and pipeline pressure, and a scientific injection plan is generated;

[0263] The repair agent solution is composed of cultured microbial agents, pH buffer, nutrient matrix, etc. The required amount of repair agent is calculated according to parameters such as crack depth and length, and the injection pressure and speed are controlled to ensure that the repair agent fully penetrates and diffuses in the crack;

[0264] During the bioremediation process, the pH value, dissolved oxygen, temperature and other environmental parameters of the medium near the cracks are monitored in real time. Data is collected through online sensors, and machine learning algorithms such as support vector machines and random forests are selected to extract key indicators such as microbial growth rate and reproduction cycle from laboratory culture data and field sampling data. Environmental parameters (temperature, pH value, nutrient concentration) are used as independent variables, and growth indicators are used as dependent variables to train regression prediction models. Through cross-validation and hyperparameter optimization, the fitting accuracy and generalization performance of the model are improved, and a prediction model between environmental parameters and microbial growth and reproduction is established. The formula and delivery plan of the repair agent are dynamically optimized to improve microbial activity and repair efficiency.

[0265] Regularly track and monitor the repaired pipeline cracks, use non-destructive testing technologies such as endoscopes and sonar, use computer vision and deep learning technologies to perform semantic segmentation and feature extraction on crack images, use convolutional neural networks to extract local features such as crack texture and edges, and use long short-term memory networks to extract global features such as crack shape and direction. Then use support vector regression algorithms to establish a mapping relationship between crack features and repair quality, evaluate the repair quality and integrity of cracks, achieve quantitative scoring and grading of crack repair quality, generate crack repair quality assessment reports, and promptly identify areas with unsatisfactory repair effects for targeted repairs;

[0266] For low-priority pipelines, we focus on preventive maintenance and condition monitoring, and take preventive maintenance measures such as regular monitoring and non-invasive testing to delay the expansion and deterioration of cracks. We generate differentiated monitoring frequencies and testing plans based on the material, diameter, and burial depth of the pipelines. For example, for metal pipelines, we conduct electromagnetic ultrasonic testing every six months, and for non-metallic pipelines, we conduct geological radar scanning every quarter. Through regular testing and data analysis, we can promptly detect early signs of crack expansion and intervene in time when the cracks have not yet seriously threatened the safety of the pipelines, so as to prevent the cracks from deteriorating to the point where major repairs are required, thereby minimizing maintenance costs and service interruption time.

[0267] Use open source frameworks such as Hadoop and Spark to build a big data analysis platform, perform ETL processing and data integration on pipeline sensor data, business system data, third-party data, etc., and build a pipeline full life cycle data warehouse;

[0268] Based on the data warehouse, we use association rule mining and time series analysis data mining algorithms to discover the key factors and rules that affect pipeline degradation, and use machine learning algorithms such as random forest and GBDT to train pipeline degradation prediction models to provide early warning of the development trend of pipeline cracks;

[0269] Adopt reinforcement learning algorithms, such as DQN and DDPG, to train the optimal pipeline maintenance strategy in a simulation environment and generate dynamic maintenance plans and resource scheduling solutions;

[0270] Correlate and analyze the monitoring data of pipeline cracks with the pipeline GIS system, operating parameters and other data, and shift from passive repair to active prevention, thereby maximizing the service life of the pipeline and reducing the cost of the entire life cycle of the pipeline;

[0271] Here is an example:

[0272] When dividing the pipeline maintenance priority, a three-layer indicator system is constructed through the analytic hierarchy process (AHP). The target layer is the pipeline maintenance priority, the criterion layer includes pipeline importance E1, failure risk E2, and maintenance cost E3, and the sub-criterion layer includes pipe diameter F1, flow F2, and water supply population F3 under pipeline importance, pipe material H1, pressure H2, and sensitivity H3 under failure risk, and burial depth G1, road surface G2, and construction difficulty G3 under maintenance cost.

[0273] The criterion layer indicators are compared pairwise to obtain the judgment matrix, and the maximum eigenvalue λ_max=3.087 is calculated. Through consistency test, the weight vector W=(0.625,0.136,0.239) is calculated, and the weights of pipeline importance, failure risk, and maintenance cost are 0.625, 0.136, and 0.239 respectively. Under the pipeline importance criterion, the weight vectors of pipe diameter, flow rate, and water supply population are (0.127, 0.717, 0.156), that is, the flow rate has the largest weight;

[0274] Similarly, the standardized values ​​of various pipeline attributes are calculated, and the weighted summary is used to obtain the comprehensive score of pipeline maintenance priority, with a value range of 0100 points. A cluster analysis was performed on the comprehensive scores of the 2,189 pipelines in the city. The K-means algorithm was used, and the number of cluster centers k=2 was selected. The number of iterations was 50 times, and the high priority threshold was 72 points. When screening bioremediation agents, from 6 alternative bacterial agents, an orthogonal test of 6 factors and 2 levels was conducted, with bacterial agent ratio, temperature, pH value, substrate concentration, etc. as factors, and microbial density and mineralization deposition rate as indicators, to select a bacterial agent formula suitable for use under the conditions of 1030℃, pH68, and COD50100mg / L;

[0275] A plunger pump is used to inject the repair agent into the pipeline. The plunger diameter is 50 mm, the pressure is 3 MPa, the speed is 0.3 m / s, and the repair agent density is 1.2 g / cm 3 , the crack size is 2 meters long, 2 millimeters wide, and 1 centimeter deep. The dosage of the agent is calculated based on 1.5 times the volume of the crack, and 1080 ml needs to be injected;

[0276] During the repair process, when the online ORP sensor detects that the ORP value at the crack has risen from -150mV to +50mV, it is judged that anaerobic microorganisms have begun to multiply in large numbers.

[0277] The support vector machine algorithm was used to establish a regression model of dissolved oxygen concentration and iron oxidizing bacteria density. The radial basis kernel function was selected, the penalty factor C = 10, and the test set R 2 =0.85, the model predicts that the density of iron oxidizing bacteria reaches the design value of 5×10 6 ce ll / ml, the oxidation precipitation rate of Fe2+ can reach 0.2 mm / day;

[0278] The long short-term memory network is used to build an intelligent crack detection model, with the grayscale histogram and LBP texture features of the crack image block as input, and the crack length, width and depth as output. The training samples are 2000, the learning rate is 0.01, the batch size is 32, the number of iterations is 500, and the relative error of the predicted crack size is less than 5%;

[0279] The FasterR-CNN algorithm was used to extract semantic information of images before and after crack repair, and quantitative indicators such as crack closure and surface roughness were calculated. The repair qualification rate reached more than 95%.

[0280] The 2,189 pipelines were divided into 1,753 low-priority pipelines and 436 high-priority pipelines. Endoscopic inspection was carried out every two months for high-priority pipelines, and electromagnetic inspection was carried out every six months for low-priority pipelines to form pipeline health records. Based on pipeline inspection big data, the April association rule mining algorithm was used, with pipeline attributes and state parameters as data items, crack occurrence as the association target, support degree of 0.05, and confidence level of 0.8. Rules such as "DN300 ductile iron pipe && pressure > 1MPa && corrosion potential <-350mV → longitudinal cracks" were obtained to construct a causal logic diagram for crack development;

[0281] The GBDT algorithm is used to predict the remaining life of pipelines. 38 features such as pipeline attributes, detection data, and crack parameters are used as input variables, and failure time is used as the target variable. 10,000 samples are trained. The top three features in terms of importance are pipe age, corrosion current, and crack length. Pipelines with a failure probability of more than 10% within 5 years account for 12%;

[0282] The DDPG algorithm is used to optimize the pipeline inspection path. The state space includes 12 state variables such as the health of 1753 pipeline nodes, crack risk, and detection cycle. The action space includes a combination of 5 detection methods and 6 repair strategies. Through continuous interaction with the environment, the average detection efficiency is improved by 8% and the average maintenance cost is reduced for every 1000 epi sodes trained.

[0283] The present invention further proposes a municipal pipeline management system based on positioning perception function, which adopts the municipal pipeline management method based on positioning perception function as described above, including:

[0284] A collection and monitoring module, which is used to collect geographic coordinates, depth and real-time outer wall image information of underground pipelines;

[0285] A pipeline real-time database construction module, wherein the pipeline real-time database construction module is used to store and manage the geographic coordinates, materials, design pressure and image data of the pipeline;

[0286] A time series analysis and crack monitoring module, which is used to analyze pipeline aging trends and predict crack development;

[0287] An image processing and stress analysis module, which is used to determine the specific characteristics of the cracks and analyze the real-time external stress on the outer wall of the pipeline;

[0288] A spectrum analysis and evaluation module, which is used to analyze pipeline status through vibration signals and quantify maintenance requirements;

[0289] The maintenance decision and environmental strategy module is used to determine the maintenance priority of each pipeline section and associate the corresponding maintenance strategy.

[0290] In the description of the present invention, it is necessary to understand that the directions or positional relationships indicated by directional words such as "front, back, up, down, left, right", "lateral, vertical, horizontal" and "top, bottom" are usually based on the directions or positional relationships shown in the drawings. They are only for the convenience of describing the present invention and simplifying the description. Unless otherwise specified, these directional words do not indicate or imply that the device or element referred to must have a specific direction or be constructed and operated in a specific direction. Therefore, they cannot be understood as limiting the scope of protection of the present invention.

[0291] For those skilled in the art, various other corresponding changes and deformations can be made according to the technical solutions and concepts described above, and all of these changes and deformations should fall within the protection scope of the claims of the present invention.

Claims

1. A municipal pipeline management method based on positioning perception function, characterized in that: include: S1. Obtain the geographic coordinates of the underground pipelines under the jurisdiction of the municipality, monitor the outer wall images of the pipelines in real time, and build a real-time pipeline database; S2. In combination with the pipeline real-time database, the attribute data of the pipeline is analyzed by using a time series analysis method to obtain the pipeline aging and damage trends, a pipeline crack monitoring model is constructed, and the development trend of the cracks is predicted by combining historical data and real-time monitoring data; S3. Analyze and process the real-time image information of the pipeline, determine the depth, width, horizontal angle to the ground and curvature of the pipeline crack, and calculate the real-time external stress value and stress change trend of the outer wall of the pipeline crack position; S4. Collecting vibration signals of the pipeline, and performing spectrum analysis on the vibration signals, and quantifying the urgency of pipeline maintenance by combining the external stress value, stress change trend and crack change trend of the outer wall of the pipeline crack; S5. Process the pipeline maintenance urgency, maintenance budget and cost, and establish a pipeline maintenance decision model in combination with the grey correlation analysis method to evaluate and adjust the pipeline maintenance priority; S6. according to the maintenance priority of the pipeline, associating pipeline maintenance strategies of different degrees with pipeline sections of different maintenance priorities; Step S2 specifically includes: Extracting pipeline attribute data from the pipeline real-time database, the pipeline attribute data including service life, material type and design working pressure, and arranging the attribute data in chronological order to form a time series data set of pipeline attributes; Preprocessing the time series data set, identifying outliers using a box plot, and conducting a comprehensive analysis based on the pipeline operation status and environmental factors to determine whether the outliers represent an abnormal state of the pipeline, and marking and alarming the abnormal state of the pipeline; The LSTM model is used to model and predict the historical change trend of the pipeline design working pressure. By analyzing the periodicity, trend and randomness of the pressure change, the pipeline aging and damage trend is obtained; Based on pipeline aging and damage trends and combined with pipeline attribute data, a pipeline crack monitoring model is constructed using machine learning algorithms; Use a pipeline endoscope or a crawling robot to obtain crack images, perform denoising, enhancement and segmentation on the crack images, extract the length, width and direction of the cracks, as well as the texture and color within the preset range of the cracks, and construct a crack feature vector; Using crack feature vectors, combined with historical crack data of pipelines, and using convolutional neural networks, pipeline crack monitoring models are trained to predict the development direction of pipeline cracks. Step S3 specifically includes: Collecting real-time image information of the pipeline through an image acquisition device, and performing image enhancement and noise removal processing on the real-time image information; The processed image is segmented by an image segmentation algorithm to extract pixel information of the pipeline crack area and obtain a binary image of the crack area. The edge detection algorithm is used to extract the pixel coordinate information of the crack edge and determine the width and depth parameters of the crack; Combined with the geometric structure parameters and spatial position information of the pipeline, the angle of the crack area relative to the central axis of the pipeline is calculated to determine the horizontal angle of the crack to the ground; The pixel coordinates of the crack edge are fitted using a curve fitting algorithm to obtain the mathematical expression of the crack curve and calculate the curvature of the crack curve. Using graph convolutional neural networks, an image processing model is constructed based on the width, depth, angle, and curvature of the crack. The mapping relationship between crack parameters and pipeline stress is established through feature extraction and parameter learning of the image processing model. According to the output results of the image processing model, the real-time stress condition of the outer wall at the crack position of the pipeline is obtained to determine the external stress value of the pipeline at the current moment; Use time series analysis algorithm to predict the trend of pipeline stress data and determine the stress change trend of pipeline crack location; For adjacent pipelines with cracks, the ductility score between two cracks in the adjacent pipelines is calculated through the real-time image information of the adjacent pipelines; Step S4 also includes: For adjacent pipelines with cracks, the resonance phenomenon of the adjacent pipelines is identified through the vibration signals of the adjacent pipelines.

2. According to claim 1, the municipal pipeline management method based on positioning perception function is characterized in that: Step S1 specifically includes: Using a differential global positioning system to obtain geographic coordinate information of underground pipelines under municipal jurisdiction, the geographic coordinate information including a starting point, an end point and a turning point, and forming a pipeline geographic coordinate data set; By comparing and calibrating the differential global positioning system with the actual pipeline path, extracting pipeline depth information from the differential global positioning system and associating it with corresponding geographic coordinates to form a pipeline depth information data set; Extract the material information of the pipeline to form a pipeline material information data set; Acquire the design working pressure data of the pipeline, and associate the design working pressure data with the geographic coordinate information, the depth information, and the material information to generate a pipeline design working pressure data set; Use pressure sensors to monitor the working pressure of the pipeline in real time and transmit the real-time pressure data to the pipeline database; An image acquisition device is used to photograph the outer wall of the pipeline to obtain an image of the outer wall of the pipeline, and an image transmission network is established to transmit the acquired image to the pipeline database in real time; A data warehouse is used to process the multi-source heterogeneous data of the pipeline, form a consistent data model and format, and build the pipeline real-time database.

3. According to claim 1, the municipal pipeline management method based on positioning perception function is characterized in that: In step S4, performing spectrum analysis on the vibration signal includes: The vibration signal of the outer wall of the pipeline is collected by using an acceleration sensor, and the collected time domain vibration signal is analyzed and preprocessed; The fast Fourier transform algorithm is used to perform frequency domain conversion on the preprocessed pipeline vibration signal to obtain the frequency spectrum distribution diagram of pipeline vibration; According to the frequency spectrum distribution diagram, the distribution law of the frequency spectrum peak amplitude is obtained through statistical analysis of the pipeline vibration data; According to the mean and standard deviation, the threshold of the spectrum peak is set as: mean + 2 times the standard deviation, and the corresponding frequency and amplitude parameters are extracted to obtain the characteristic spectrum peak parameters of pipeline vibration; Perform short-time Fourier transform on the pipeline vibration signal to obtain the time-frequency distribution diagram of the vibration signal. Set the window function of the short-time Fourier transform. Through the time-frequency diagram, the transient characteristics and energy distribution changes of the pipeline vibration can be analyzed, and the transient characteristic parameters of the pipeline vibration signal can be extracted. The pipeline vibration signal is divided into several time windows by using the order analysis method based on short-time Fourier transform. The signal in each window is subjected to fast Fourier transform to obtain the spectrum diagram at different times. For the spectrum graph within each time window, find the frequency point with the largest amplitude in each frequency band as the characteristic frequency of the frequency band; Arrange the characteristic frequencies of each frequency band from small to large to obtain a set of frequency sequences, which correspond to the order components of pipeline vibration.

4. According to claim 1, the municipal pipeline management method based on positioning perception function is characterized in that: Step S5 specifically includes: Extracting data on pipeline maintenance urgency, pipeline maintenance budget and pipeline maintenance cost from the pipeline maintenance decision system; Processing data on the pipeline maintenance urgency, the pipeline maintenance budget, and the pipeline maintenance cost; A pipeline maintenance decision model is established by using a grey correlation analysis method, the pipeline maintenance urgency, the pipeline maintenance budget and the pipeline maintenance cost are used as input parameters of the model, and the pipeline maintenance priority is used as the output parameter of the model. The importance weight of each parameter is determined by calculating the grey correlation between each input parameter and the output parameter; Calculating the grey correlation coefficient of the pipeline maintenance urgency, the pipeline maintenance budget and the pipeline maintenance cost; According to the grey correlation coefficient, the grey correlation degree of the pipeline maintenance urgency, the pipeline maintenance budget and the pipeline maintenance cost is calculated, and the grey correlation degree is normalized to obtain a weight vector of each parameter; Performing a weighted summation on the pipeline maintenance urgency, the pipeline maintenance budget, and the standardized value of the pipeline maintenance cost and a weight vector to obtain a comprehensive evaluation value of the pipeline maintenance priority; Sort the pipelines according to the comprehensive evaluation values ​​to obtain a pipeline maintenance priority list; If the maintenance priority of a section of pipeline changes, the maintenance priority of adjacent pipelines will be adjusted in conjunction with the pipeline network topology, combined with the pipeline importance coefficient and risk level coefficient.

5. According to the municipal pipeline management method based on positioning perception function as claimed in claim 1, it is characterized in that: Step S6 specifically includes: Based on the pipeline maintenance priority list generated by the pipeline maintenance decision system, the pipeline maintenance priority evaluation index system is constructed by using the analytic hierarchy process from multiple dimensions including pipeline importance, failure risk and maintenance cost. By comparing two indicators in pairs, the weight of each indicator is calculated to form a comprehensive score of pipeline maintenance priority. According to the score distribution, the threshold intervals of high and low priorities are divided, and the priority threshold is set. The pipelines with maintenance priorities higher than the threshold are classified as high-priority pipelines, and the pipelines with maintenance priorities lower than or equal to the threshold are classified as low-priority pipelines. In addition, different degrees of pipeline maintenance strategies are associated with pipeline segments with different maintenance priorities. The pipeline maintenance strategy includes using bioremediation agents to repair cracks in high-priority pipelines and taking regular monitoring and tracking and non-invasive detection and protection measures for cracks in low-priority pipelines.

6. A municipal pipeline management system based on positioning perception function, adopting the municipal pipeline management method based on positioning perception function as claimed in any one of claims 1 to 5, characterized in that: include: A collection and monitoring module, which is used to collect geographic coordinates, depth and real-time outer wall image information of underground pipelines; A pipeline real-time database construction module, wherein the pipeline real-time database construction module is used to store and manage the geographic coordinates, materials, design pressure and image data of the pipeline; A time series analysis and crack monitoring module, which is used to analyze pipeline aging trends and predict crack development; An image processing and stress analysis module, which is used to determine the specific characteristics of the cracks and analyze the real-time external stress on the outer wall of the pipeline; A spectrum analysis and evaluation module, which is used to analyze pipeline status through vibration signals and quantify maintenance requirements; The maintenance decision and environmental strategy module is used to determine the maintenance priority of each pipeline section and associate the corresponding maintenance strategy.

Images