Metal pipeline corrosion and anticorrosive coating stripping test system

Abstract

The invention discloses a metal pipeline corrosion and anticorrosive coating stripping test system, and relates to the technical field of metal pipeline detection, the system comprises: an acquisition unit configured to acquire multi-dimensional test data in real time; the first processing unit is used for constructing a grading test evaluation model; the second processing unit is used for mapping the multi-dimensional test data into feature vectors of nodes and edges; the third processing unit is used for carrying out distributed collaborative optimization and updating and optimizing parameters of each edge side test model; and the output unit is configured to obtain a grading test result according to the optimized model. According to the invention, by introducing a collaborative mechanism of multi-dimensional test perception, hierarchical evaluation modeling, graph neural network topological association and distributed collaborative optimization, high-precision, hierarchical and spatial coupling perception and test evaluation of metal pipeline corrosion and anticorrosive coating stripping are realized.

Description

Technical Field

[0001] This invention relates to the technical field of metal pipeline inspection, and in particular to a metal pipeline corrosion and anti-corrosion coating peeling test system. Background Technology

[0002] In recent years, with the rapid development of Industrial Internet of Things (IIoT) and smart pipeline technologies, the operational safety and durability of metal pipelines, as core infrastructure in energy transportation and municipal water supply, have received increasing attention. Especially in scenarios such as long-distance oil and gas transportation and urban underground pipeline networks, corrosion and anti-corrosion layer peeling detection of metal pipelines have become crucial means to prevent pipeline leaks, reduce maintenance costs, and ensure public safety. Currently, some industries have deployed testing terminals at pipeline defects, along pipeline sections, and at pipeline hubs to collect operational data, including corrosion depth, anti-corrosion layer impedance, and soil medium parameters, supplemented by basic algorithms for defect analysis. However, existing testing methods generally suffer from the following problems: On the one hand, traditional methods often employ centralized testing architectures, lacking differentiated designs for testing characteristics at different levels, such as local defect level, pipe segment level, and global pipeline network level, making it difficult to achieve refined hierarchical evaluation. On the other hand, existing models generally do not consider the spatial correlation characteristics of medium conduction, electrochemical correlation, and physical connections between metal pipes, failing to effectively characterize the propagation path and mutual coupling relationship of pipeline corrosion and anti-corrosion coating peeling, resulting in insufficient testing accuracy. Furthermore, current systems often rely on centralized data processing models, suffering from transmission bandwidth bottlenecks, real-time response lags, and the risk of sensitive data leakage. Moreover, the testing model parameter updates lack distributed collaborative mechanisms, making it difficult to achieve collaborative optimization across multiple testing nodes. Therefore, it is necessary to design a testing method and system for metal pipeline corrosion and anti-corrosion coating peeling to address the problems existing in current technologies. Summary of the Invention

[0003] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0004] In view of the problems existing in the prior art, the present invention is proposed.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a metal pipeline corrosion and anti-corrosion coating peeling testing system, the system comprising: The acquisition unit is configured to acquire multi-dimensional test data in real time. The multi-dimensional test data includes the local corrosion depth of the pipeline, the peeling area of ​​the anti-corrosion layer, the humidity of the medium in the pipeline section, the pH value of the soil, the pressure inside the pipeline, the resistance value of the anti-corrosion layer, the electrochemical potential of the pipeline welding point, and the cathodic protection potential of the pipeline network. It also acquires pipeline material parameters and laying environment parameters. The first processing unit is configured to construct a graded test and evaluation model based on the laying distribution of the pipeline network, the connection structure of the pipe segment, and the failure characteristics of the corrosion-anti-corrosion layer. The graded test and evaluation model includes a local defect-level edge test model, a pipe segment-level aggregation evaluation model, and a pipeline network-level risk prediction model. The second processing unit is configured to construct a pipeline topology correlation analysis model based on a graph neural network, taking a single pipeline unit as a graph node, the physical connection relationship of the pipeline and the medium conduction path as graph edges, and mapping the multi-dimensional test data into feature vectors of nodes and edges. The third processing unit is configured to periodically collect parameters from the local defect-level edge test model, the pipe segment-level aggregate evaluation model, and the pipeline topology correlation analysis model, perform distributed collaborative optimization, and update and optimize the parameters of each edge-side test model. The output unit is configured to obtain graded test results based on the optimized model, and output local defect warning, pipe section corrosion rate assessment and pipe network failure risk level in conjunction with the graded test results and the pipe network-level risk prediction model.

[0006] As a preferred embodiment of the metal pipeline corrosion and anti-corrosion coating peeling test system described in this invention, the following are included: on the side of local defects in the pipeline, an ultrasonic flaw detector, an infrared thermal imager, and an electrochemical sensor are used to collect corrosion depth, anti-corrosion coating peeling area, and electrochemical potential of weld points, respectively; on the side of the pipeline laying section, a soil sensor and a humidity sensor are used to collect soil pH value and humidity of the medium in the pipeline section. On the main pipeline side, pressure sensors and impedance testers are used to collect the pressure inside the pipeline and the impedance value of the anti-corrosion layer; on the cathodic protection side of the pipeline network, a potential monitor is used to obtain the cathodic protection potential of the pipeline network. Each test terminal device collects corresponding physical quantity signals according to a preset sampling period, and performs local encoding and format standardization on the collected results. The results are then sent to the edge device via the LoRa communication protocol and industrial bus. The edge device performs time calibration, abnormal data filtering, and feature dimension unification processing on the collected data.

[0007] As a preferred embodiment of the metal pipeline corrosion and anti-corrosion coating peeling testing system of the present invention, a graded testing and evaluation model is constructed based on the pipeline network layout, pipe segment connection structure, and corrosion-anti-corrosion coating failure characteristics. The construction process of the local defect level edge testing model includes: On the edge side equipment corresponding to each target pipeline defect area, an input vector containing historical time series data is constructed based on the corrosion depth, anti-corrosion layer peeling area, weld point electrochemical potential, pipeline material parameters and laying environment parameters of the area. A test model is established based on a neural network structure, which is a gated recurrent network with a temporal feature extraction unit. The test model is trained based on the input vector to predict the local defect expansion trend within a preset time window in the future. The trained model is deployed on edge devices, and inference is performed based on real-time test data to output local defect-level test results.

[0008] As a preferred embodiment of the metal pipeline corrosion and anti-corrosion coating peeling testing system of the present invention, the construction process of the pipe section-level aggregation evaluation model includes: On the edge-side device corresponding to each target pipe segment, the local defect level test results output by several local defect level edge test models under the pipe segment are received, and the aggregated input vector is constructed by combining the laying length of the pipe segment, soil type, historical corrosion data, cathodic protection potential and current stray current intensity. The network structure of the pipe segment-level aggregated evaluation model is based on a multilayer perceptron neural network. The model is trained based on the aggregated input vector to predict the average corrosion rate and the overall failure probability of the anti-corrosion layer within a preset time window of the pipe segment. The trained aggregate evaluation model is deployed in the pipe segment-level edge device, and the pipe segment-level aggregate evaluation result is output based on the real-time test results of each local defect-level edge test model.

[0009] As a preferred embodiment of the metal pipeline corrosion and anti-corrosion coating peeling testing system of the present invention, the construction process of the pipeline network-level risk prediction model includes: On the central analysis server, the pipeline segment-level aggregated evaluation results uploaded by edge-side devices at each pipeline segment level are received, and the overall cathodic protection potential of the pipeline network, the pressure at each pipeline segment connection point, the distribution of stray currents in the surrounding area, and environmental temperature and humidity parameters are collected to construct a risk input vector. A pipeline-level risk prediction model is constructed based on a gradient boosting regression model. The pipeline-level risk prediction model is trained according to the risk input vector and is used to predict the distribution of corrosion hotspots and the overall failure risk level of the pipeline within a preset time window in the future. The pipeline-level risk prediction model is deployed in the central analysis server, and the inference results of the pipeline-level risk prediction model are used as the basis for prioritizing local defect repair, strengthening pipeline corrosion protection strategies, and adjusting pipeline cathodic protection parameters.

[0010] As a preferred embodiment of the metal pipeline corrosion and anti-corrosion coating peeling testing system of the present invention, the system includes: constructing a pipeline topology correlation analysis model based on a graph neural network, treating individual pipeline units as graph nodes, and the physical connection relationships of pipelines and the medium conduction paths as graph edges, and mapping the multi-dimensional test data into feature vectors of nodes and edges, including: Using each pipe unit in the test network as a graph node, and the pipe pairs that are related to each pipe unit through physical connection or medium conduction as graph edges, an undirected graph structure containing a set of nodes and a set of edges is constructed. The corrosion depth, anti-corrosion layer impedance value, pipe material parameters, and laying environment parameters of each pipe unit are used as node feature vectors, and the medium humidity of the pipe section, the pressure inside the pipe, and the cathodic protection potential difference value are used as edge feature vectors. A graph neural network model is constructed based on graph convolutional neural networks. The graph structure is embedded and learned, and a structured representation vector for each pipeline node is generated.

[0011] As a preferred embodiment of the metal pipeline corrosion and anti-corrosion coating peeling testing system of the present invention, the following steps are included: periodically collecting parameters of the local defect-level edge test model, the pipe segment-level aggregate evaluation model, and the pipeline topology correlation analysis model, performing distributed collaborative optimization, and updating and optimizing the parameters of each edge-side test model: A distributed collaboration module is deployed on the central analysis server. The distributed collaboration module sends parameter collaboration requests to each edge device at a preset time period to collect local training parameters of the local defect-level edge test model, the pipe segment-level aggregate evaluation model, and the graph neural network model. The distributed collaboration module performs global collaborative optimization by implementing a dynamic weight allocation strategy on the collected model parameters of each edge device, forming unified global model parameters. The global model parameters are updated and synchronously distributed to the corresponding edge devices to replace the local test model parameters.

[0012] As a preferred embodiment of the metal pipeline corrosion and anti-corrosion coating peeling testing system of the present invention, wherein: when the distributed collaborative module performs collaborative optimization on the local parameters uploaded by each edge-side device, it adopts differentiated optimization strategies for different model types, including: For the local defect-level edge testing model and the pipe segment-level aggregate evaluation model, a collaborative averaging algorithm based on parameter sparsity constraints is used to regularize the parameter updates. For the node embedding parameters of the graph neural network in the pipeline topology correlation analysis model, the feature distribution of highly correlated edges is preserved during the optimization process; After optimization, the degree of adaptation of the optimized model to the edge test results is verified by the error consistency function. If the preset adaptation threshold is met, the model parameter synchronization process is carried out. As a preferred embodiment of the metal pipeline corrosion and anti-corrosion coating peeling testing system of the present invention, wherein: when the distributed collaborative module performs collaborative optimization on the local parameters uploaded by each edge-side device, it adopts differentiated optimization strategies for different model types, including: For the local defect-level edge testing model and the pipe segment-level aggregate evaluation model, a collaborative averaging algorithm based on parameter sparsity constraints is used to regularize the parameter updates. For the node embedding parameters of the graph neural network in the pipeline topology correlation analysis model, the feature distribution of highly correlated edges is preserved during the optimization process; After optimization, the degree of adaptation of the optimized model to the edge test results is verified by the error consistency function. If the preset adaptation threshold is met, the model parameter synchronization process is carried out. The beneficial effects of this invention are as follows: By introducing a collaborative mechanism of "multi-dimensional test perception + hierarchical evaluation modeling + graph neural network topology association + distributed collaborative optimization", high-precision, hierarchical, and spatially coupled perception and test evaluation of metal pipeline corrosion and anti-corrosion layer peeling are achieved.

[0013] This application deploys multiple types of testing terminals on the pipeline local defect side, pipe section laying side, and pipeline cathodic protection side to comprehensively collect key parameters in the pipeline corrosion and anti-corrosion layer failure process. By constructing hierarchical testing and evaluation models at the local defect level, pipe section level, and pipeline network level, and combining graph neural networks to mine the topological relationships of medium conduction, electrochemical correlation, and physical connection between pipelines, the system's ability to characterize corrosion propagation trends and spatial coupling effects is improved. The distributed collaborative module further breaks down data silos, enabling collaborative optimization of various testing models while protecting the privacy of sensitive pipeline network data, ensuring the generalization ability of model updates and the real-time responsiveness of edge-side testing. Combined with the pipeline network-level risk prediction model, the system achieves linked evaluation of local defect repair priorities, pipe section anti-corrosion strengthening strategies, and pipeline network cathodic protection parameters. While ensuring testing accuracy, this improves the operation and maintenance efficiency and control flexibility of the metal pipeline testing system, effectively solving the problems of coarse testing granularity, lack of spatial correlation modeling, and bottlenecks in centralized computing architecture in current technologies. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a schematic diagram of the overall structure of a metal pipeline corrosion and anti-corrosion coating peeling testing system proposed in this invention. Detailed Implementation

[0015] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0016] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0017] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0018] Reference Figure 1 In one embodiment of the present invention, a metal pipeline corrosion and anti-corrosion coating peeling test system is provided. The system includes: a data acquisition unit configured to acquire multi-dimensional test data in real time. The multi-dimensional test data includes local corrosion depth of the pipeline, anti-corrosion coating peeling area, medium humidity of the pipeline section, soil pH value, internal pressure of the pipeline, anti-corrosion coating impedance value, electrochemical potential of pipeline welding point and cathodic protection potential of pipeline network, and also acquires pipeline material parameters and laying environment parameters (including soil type, groundwater level and surrounding stray current intensity).

[0019] Specifically, when collecting multi-dimensional test data in real time, this includes: On the local defect side of the pipeline, ultrasonic flaw detectors, infrared thermal imagers, and electrochemical sensors are used to collect corrosion depth, anti-corrosion coating peeling area, and weld point electrochemical potential, respectively. On the pipeline section laying side, soil sensors and humidity sensors are used to collect soil pH value and pipeline medium humidity. On the pipeline trunk side, pressure sensors and impedance testers are used to collect pipeline internal pressure and anti-corrosion coating impedance value. On the pipeline cathodic protection side, a potential monitor is used to obtain the pipeline cathodic protection potential. Each test terminal device collects corresponding physical quantity signals according to a preset sampling period, and the collected results are locally encoded and format standardized, and sent to the edge device via LoRa communication protocol and industrial bus. The edge device performs time calibration, abnormal data filtering, and feature dimension unification processing on the collected data.

[0020] The first processing unit is configured to construct a graded testing and evaluation model based on the pipeline network's laying distribution, pipe segment connection structure, and corrosion-anti-corrosion layer failure characteristics. The graded testing and evaluation model includes a local defect-level edge testing model, a pipe segment-level aggregate evaluation model, and a pipeline network-level risk prediction model.

[0021] The construction process of the local defect-level edge testing model includes: On the edge side equipment corresponding to each target pipeline defect area, an input vector containing historical time series data is constructed based on the corrosion depth, anti-corrosion layer peeling area, weld point electrochemical potential, pipeline material parameters and laying environment parameters of the area. A test model is established based on a neural network structure, which is a gated recurrent network with a temporal feature extraction unit. The test model is trained based on the input vector to predict the local defect expansion trend within a preset time window in the future. The trained model is deployed on edge devices, and inference is performed based on real-time test data to output local defect-level test results (including corrosion propagation rate and anti-corrosion layer peeling risk coefficient).

[0022] The process of constructing the pipe segment-level aggregation evaluation model includes: On the edge-side device corresponding to each target pipe segment, the local defect level test results output by several local defect level edge test models under the pipe segment are received, and the aggregated input vector is constructed by combining the laying length of the pipe segment, soil type, historical corrosion data, cathodic protection potential and current stray current intensity. The network structure of the pipe segment-level aggregated evaluation model is based on a multilayer perceptron neural network. The model is trained based on the aggregated input vector to predict the average corrosion rate and the overall failure probability of the anti-corrosion layer within a preset time window of the pipe segment. The trained aggregate evaluation model is deployed in the pipe segment-level edge device, and the pipe segment-level aggregate evaluation result is output based on the real-time test results of each local defect-level edge test model.

[0023] The process of constructing a pipeline-level risk prediction model includes: On the central analysis server, the pipeline segment-level aggregated evaluation results uploaded by edge-side devices at each pipeline segment level are received, and the overall cathodic protection potential of the pipeline network, the pressure at each pipeline segment connection point, the distribution of stray currents in the surrounding area, and environmental temperature and humidity parameters are collected to construct a risk input vector. A pipeline-level risk prediction model is constructed based on a gradient boosting regression model. The pipeline-level risk prediction model is trained according to the risk input vector and is used to predict the distribution of corrosion hotspots and the overall failure risk level of the pipeline within a preset time window in the future. The pipeline-level risk prediction model is deployed on the central analysis server, and the inference results of the pipeline-level risk prediction model are used as the basis for prioritizing local defect repair, strengthening pipeline corrosion protection strategies, and adjusting pipeline cathodic protection parameters.

[0024] The second processing unit is configured to construct a pipeline topology correlation analysis model based on a graph neural network, taking a single pipeline unit (including straight pipe sections, welding points, and valve nodes) as a graph node, the physical connection relationship of the pipeline and the medium conduction path as graph edges, and mapping multi-dimensional test data into feature vectors of nodes and edges.

[0025] Specifically, this includes: using each pipe unit in the test network as a graph node, and using pipe pairs that are related to each pipe unit through physical connection or medium conduction as graph edges, constructing an undirected graph structure containing a set of nodes and a set of edges; The corrosion depth, anti-corrosion layer impedance value, pipe material parameters, and laying environment parameters of each pipe unit are used as node feature vectors, and the medium humidity of the pipe section, the pressure inside the pipe, and the cathodic protection potential difference value are used as edge feature vectors. A graph neural network model is constructed based on graph convolutional neural networks. The graph structure is embedded and learned, and a structured representation vector for each pipeline node is generated.

[0026] The third processing unit is configured to periodically collect parameters from the local defect-level edge test model, the pipe segment-level aggregate evaluation model, and the pipeline topology correlation analysis model, perform distributed collaborative optimization, and update and optimize the parameters of each edge-side test model.

[0027] Specifically, a distributed collaboration module is deployed on the central analysis server. The distributed collaboration module sends parameter collaboration requests to each edge device at a preset time period to collect local training parameters of the local defect-level edge test model, the pipe segment-level aggregate evaluation model, and the graph neural network model. The distributed collaboration module performs global collaborative optimization by implementing a dynamic weight allocation strategy on the collected model parameters of each edge device, forming unified global model parameters; The global model parameters are updated and synchronously distributed to the corresponding edge devices to replace the local test model parameters.

[0028] When the distributed collaboration module performs collaborative optimization on the local parameters uploaded by each edge device, it adopts differentiated optimization strategies for different model types, including: For the local defect-level edge testing model and the pipe segment-level aggregate evaluation model, a collaborative averaging algorithm based on parameter sparsity constraints is adopted to regularize the parameter updates. For the node embedding parameters of the graph neural network in the pipeline topology correlation analysis model, the feature distribution of highly correlated edges is preserved during the optimization process (highly correlated edges are defined as pipeline connections where the medium conduction efficiency and physical connection strength are higher than a set threshold). After optimization, the degree of adaptation of the optimized model to the edge test results is verified by the error consistency function. If the preset adaptation threshold is met, the model parameter synchronization process is carried out.

[0029] The output unit is configured to obtain graded test results based on the optimized model, and output local defect warnings, pipe section corrosion rate assessments, and pipe network failure risk levels in conjunction with the graded test results and the pipeline network-level risk prediction model.

[0030] Specifically, based on the optimized model, the graded test results are obtained. When the graded test results are combined with the pipeline-level risk prediction model to output local defect warnings, pipe section corrosion rate assessments, and pipeline failure risk levels, the following are included: Based on the updated local defect level edge test model and pipe segment level aggregate evaluation model, the local defect level edge side equipment and pipe segment level edge side equipment respectively output the local defect level test results and pipe segment level aggregate evaluation results within the prediction time window in real time. The pipeline segment-level aggregation assessment results and the pipeline correlation strength information output by the pipeline topology correlation analysis model are jointly input into the pipeline network-level risk prediction model to obtain the predicted value of the pipeline network failure risk level within the target time window. Based on the predicted value of pipeline failure risk level, the aggregated evaluation results of each pipe segment, and the structured representation vector of pipeline nodes generated by the graph neural network model as input features, a global corrosion risk distribution map is constructed. Based on the global corrosion risk distribution map, a risk optimization objective function is established based on a multi-objective decision-making algorithm. The risk optimization objective function is used to minimize the detection and maintenance costs and failure losses while meeting the requirements for safe pipeline operation. Under the constraints of the risk optimization objective function, the local defect repair priority instructions, pipe section anti-corrosion strengthening parameters, and pipeline cathodic protection adjustment scheme are obtained in real time. Instructions and solutions are encoded in a standardized format and sent to the field testing terminal and pipeline operation and maintenance control unit through the cloud-edge collaboration interface.

[0031] This application provides a detailed description of the invention through a preferred embodiment, using a long-distance oil pipeline system as an application scenario. The pipeline is 500km long and divided into 50 independent pipe sections according to the laying area. Each pipe section contains 8-12 local defect monitoring points (covering weld points, valve nodes, and easily corroded sections). The entire pipeline network adopts a three-level architecture of "local defect monitoring - centralized pipe section assessment - overall network early warning". The pipeline material is X80 steel, and the anti-corrosion layer uses a 3PE composite anti-corrosion layer. The laying environment covers various scenarios such as deserts, farmland, and urban roads. Soil types include sandy soil, clay soil, and saline-alkali soil. The groundwater level is 1-8m deep, and there is stray current interference from railways in the surrounding area (interference intensity 0.1-0.8A / m). 2 This system, by deploying test terminals and edge devices at various levels and combining them with the collaborative optimization capabilities of the central analysis server, achieves precise testing and risk assessment of pipeline corrosion and anti-corrosion coating peeling throughout the entire process. Specific implementation details for each unit are as follows: (I) Data Acquisition Unit: Accurate Acquisition and Preprocessing of Multi-Dimensional Test Data The core objective of the data acquisition unit is to achieve comprehensive and synchronous acquisition of parameters related to pipeline corrosion and anti-corrosion coating failure, providing high-quality data support for subsequent model training and inference. The specific implementation is as follows: Test terminal deployment and parameter acquisition; For localized defects in the pipeline: An Olympus EPO CH650 ultrasonic flaw detector (detection accuracy ±0.01mm) was installed at key locations such as pipeline weld points and historical corrosion points, with a preset sampling period of 1 minute, to collect the localized corrosion depth of the pipeline; a Fluke Ti400+ infrared thermal imager (temperature resolution 0.05℃) was used with a sampling period of 5 minutes to identify the area of ​​anti-corrosion coating peeling and calculate the peeling area through thermal radiation differences; a PARSTAT4000 electrochemical sensor (potential measurement range -2V~2V) was used with a sampling period of 30 seconds to collect the electrochemical potential of the weld points, reflecting the localized electrochemical corrosion activity.

[0032] On the pipeline laying side: A Mettler Toledo InPro4800 soil pH sensor (measurement range 0-14, accuracy ±0.02) is used, buried in the soil at a depth of 50cm around the pipeline, with a sampling cycle of 10 minutes to collect soil pH values; a Honeywell HIH9130 humidity sensor (measurement range 0-100%RH, accuracy ±2%RH) is used, deployed in the same location as the pH sensor, with a sampling cycle of 10 minutes to collect humidity of the medium around the pipeline.

[0033] On the main pipeline side: A Rosemount 3051 pressure sensor (measurement range 0-10MPa, accuracy ±0.075%FS) is used, installed 2m downstream of the valve on the main pipeline, with a sampling period of 5 minutes to collect the pressure inside the pipeline; an Agilent E4990A impedance tester (measurement range 20Hz-1MHz) is used, connected between the surface of the anti-corrosion layer and the metal substrate of the pipeline, with a sampling period of 15 minutes to collect the impedance value of the anti-corrosion layer, characterizing the integrity of the anti-corrosion layer.

[0034] For the cathodic protection side of the pipeline network: A Raytek CP-100 potential monitor (measurement range -2V~0V, accuracy ±0.001V) is deployed at the output end of the cathodic protection station of each pipeline section. The sampling period is 3 minutes to obtain the cathodic protection potential of the pipeline network and evaluate the cathodic protection effect.

[0035] Data transmission and preprocessing: After each test terminal device collects physical quantity signals according to the above-mentioned preset sampling period, the format is standardized by local JSON format encoding (fields include device ID, collection timestamp, parameter name, value, and unit). Then, the data is transmitted to the pipe segment edge device using the LoRa communication protocol (transmission distance 3-5km, bandwidth 125kHz). In densely populated urban pipe segment areas, Modbus RTU industrial bus (transmission rate 9600bps) is used to ensure data stability.

[0036] The edge devices use the Huawei IE4000 industrial-grade edge computing gateway to perform three-step preprocessing on the collected data: Time calibration: Based on the NTP network time protocol, millisecond-level time alignment of data from each terminal is achieved, eliminating timing deviations caused by sampling delays of different devices; Outlier filtering: The 3σ criterion is used to remove outliers that deviate from the data mean by 3 times the standard deviation (such as out-of-range data caused by sensor failure), and missing data is supplemented by linear interpolation. Unified feature dimensions: The Z-score standardization method is adopted to map parameters of different dimensions (such as corrosion depth, impedance value, pH value) to the [0,1] interval, generating a standardized feature matrix for subsequent model input.

[0037] (II) First Processing Unit: Construction and Deployment of the Tiered Testing and Evaluation Model The first processing unit constructs differentiated models based on the testing requirements of different pipeline levels, enabling accurate assessment from local defects to the entire pipeline network. The specific implementation is as follows: A local defect-level edge testing model is constructed using input vectors: Historical time-series data from the past 24 hours is selected on the edge-side device (using a Raspberry Pi 4B embedded development board) corresponding to each target pipeline defect area. This data includes corrosion depth (1440 data points), anti-corrosion coating peeling area (288 data points), weld point electrochemical potential (2880 data points), pipeline material parameters (yield strength of X80 steel 620MPa, tensile strength 700MPa), and laying environment parameters (soil pH value 288 data points, groundwater level depth 10m, stray current intensity 0.3A / m). 2 ), construct an input vector with dimension 128.

[0038] Model Construction and Training: A Gated Recurrent Network (GRU) with temporal feature extraction units was used as the neural network structure. The model architecture is as follows: Input layer (128 neurons) → Hidden layer (2 layers of GRU units, 64 neurons per layer, activation function is tanh) → Output layer (2 neurons, corresponding to corrosion spread rate and anti-corrosion layer peeling risk coefficient). The training process used the Adam optimizer (learning rate 0.001), with mean squared error (MSE) as the loss function. The training consisted of 200 epochs with a batch size of 32. Training was completed based on historical test data of local defects from the past 6 months (a total of 8000 samples). The model fit was R0. 2 ≥0.92.

[0039] Model Deployment and Inference: The trained model is packaged and deployed on edge devices in the form of Docker containers. It receives preprocessed real-time test data every 30 minutes, performs forward inference, and outputs the local defect level test results for the next 24 hours. The corrosion spread rate is in mm / year (accuracy ±0.001mm / year), and the anti-corrosion layer peeling risk coefficient ranges from 0 to 1 (0 means no peeling risk, 1 means immediate failure).

[0040] The pipe segment-level aggregated evaluation model is constructed by constructing aggregated input vectors: On the edge-side device (Huawei FusionServerPro2288HV5 server) corresponding to each target pipe segment, the test results (a total of 20 features) output by 10 local defect-level edge test models under that pipe segment are received, and combined with the pipe segment laying length (10km), soil type (sandy soil, corrosion coefficient 0.3), historical corrosion data (average corrosion rate of 0.05mm / year in the past year), cathodic protection potential (-1.0V), and current stray current intensity (0.4A / m). 2 ), construct an aggregated input vector with a dimension of 16.

[0041] Model Construction and Training: A multilayer perceptron (MLP) neural network was used as the network structure, with the following architecture: Input layer (16 neurons) → Hidden layer 1 (64 neurons, ReLU activation function) → Hidden layer 2 (32 neurons, ReLU activation function) → Hidden layer 3 (16 neurons, ReLU activation function) → Output layer (2 neurons, linear activation function). Training was conducted using historical pipeline data (10,000 samples) from the past three months, employing a stochastic gradient descent (SGD) optimizer (learning rate 0.01, momentum factor 0.9), with MSE as the loss function. The training run consisted of 150 epochs, a batch size of 64, and a test set error ≤3%.

[0042] Model Deployment and Inference: The trained model is deployed on the edge side equipment at the pipe segment level. Every hour, the real-time test results of each local defect level model are summarized. Combined with the real-time collected environmental parameters, the aggregated input vector is updated, and the pipe segment-level aggregated evaluation results for the next 48 hours are output, including the average corrosion rate of the pipe segment (mm / year) and the overall failure probability of the anti-corrosion layer (0-1, accuracy ±0.02).

[0043] Pipeline-level risk prediction model Risk input vector construction: On the central analysis server (Huawei TaiShan200 server cluster, including 4 node servers), aggregated evaluation results (100 features in total) uploaded by 50 pipe segment-level edge-side devices are received, and the overall cathodic protection potential of the pipeline network (-0.9V), the pressure at the connection points of each pipe segment (average 3.5MPa), and the distribution of stray currents in the surrounding area (0.1-0.8A / m) are collected. 2 The risk input vector with 32 dimensions is constructed based on ambient temperature and humidity (average temperature 25℃, humidity 60%RH).

[0044] Model Construction and Training: A pipeline-level risk prediction model was built based on the XGBoost gradient boosting regression model. The model parameters were set as follows: tree depth 6, learning rate 0.01, number of iterations 500, minimum sample weight sum 0.1, and 5-fold cross-validation was used to optimize the hyperparameters. The training data consisted of historical pipeline test data from the past year (20,000 samples), which were divided into training and test sets in a 7:3 ratio. The model's prediction accuracy was ≥90%.

[0045] Model Deployment and Application: The model is deployed on the central analysis server and receives aggregated evaluation results of each pipe segment every 2 hours. It infers and outputs the distribution of pipeline corrosion hotspots (the pipeline area is divided into 10m×10m grids and the hotspot locations are marked) and the overall failure risk level (1-5, where level 1 is extremely low risk and level 5 is extremely high risk) for the next 72 hours. The results are used as the core basis for prioritizing local defect repairs, formulating pipeline anti-corrosion strengthening strategies (such as increasing the thickness of the anti-corrosion coating and adjusting the cathodic protection current), and adjusting the cathodic protection parameters of the pipeline network.

[0046] (III) Second Processing Unit: Construction and Learning of Pipeline Topology Correlation Analysis Model The second processing unit uses graph neural networks to mine the spatial correlation characteristics between pipelines, accurately depicting the propagation path of corrosion and anti-corrosion coating peeling. The specific implementation is as follows: The pipeline topology diagram is constructed using 200 pipeline units within the test network as graph nodes. These include 150 straight pipe segment units (divided into 1km lengths), 30 welding point units, and 20 valve node units. Each node is assigned a unique ID (e.g., “Pipe-001”, “Weld-005”, “Valve-012”).

[0047] Using the physical connections between pipe units (such as the rigid connection between straight pipe sections and welded joints, and the flange connection between valves and straight pipe sections) and the media transmission paths (such as the crude oil transportation path between adjacent pipe sections) as graph edges, an undirected graph structure is constructed, forming a total of 180 physical connection edges and 50 media transmission edges. The initial weight of the edges is set to 1.0.

[0048] Feature vector mapping, node feature vector: The four core parameters of each pipeline unit are mapped to a node feature vector with a dimension of 16, specifically including: corrosion depth (normalized value), anti-corrosion layer impedance value (normalized value), pipeline material parameters (yield strength and tensile strength after normalization), and laying environment parameters (soil pH value, groundwater level depth, and stray current intensity after normalization).

[0049] Edge feature vector: The three types of correlation parameters between pipeline units are mapped to edge feature vectors with a dimension of 8, specifically including: pipe segment medium humidity (average humidity between two adjacent nodes), pipe internal pressure (pressure difference between two adjacent nodes), and cathodic protection potential difference (absolute potential difference between two adjacent nodes).

[0050] Graph Neural Network Model Construction and Embedding Learning: Based on Graph Convolutional Neural Network (GCN), a graph neural network model is constructed with the following architecture: Input layer (node ​​feature vector dimension 16) → First graph convolutional layer (output dimension 64, activation function LeakyReLU) → Second graph convolutional layer (output dimension 32, activation function LeakyReLU) → Output layer (node ​​structured representation vector dimension 32).

[0051] The model was trained using the Adam optimizer (learning rate 0.005), cross-entropy loss function, and 100 training epochs. The training was completed based on nearly 3 months of pipeline topology data and test data (5000 sets of samples). Through neighborhood aggregation operation, the representation vector of each node is integrated with its own attributes and the correlation features of neighboring nodes. For example, the structured representation vector of a welding point node will include the corrosion status and medium parameters of the two straight pipe sections it connects to, thereby accurately depicting the corrosion propagation correlation between pipelines.

[0052] (iv) Third processing unit: Distributed collaborative optimization and updating of model parameters The third processing unit optimizes the parameters of each level of the model through a distributed collaborative mechanism, ensuring the generalization ability of the global model and the real-time responsiveness of the edge-side model. The specific implementation is as follows: Distributed collaboration module deployment and parameter acquisition: A distributed collaboration module (developed based on the TensorFlow Federated framework) is deployed on the central analysis server. The preset collaboration cycle is 12 hours, that is, every 12 hours, parameter collaboration requests are sent to each edge device (50 pipe segment level edge devices and 200 local defect level edge devices).

[0053] After receiving the request, each edge device compresses and encodes the locally trained model parameters (weight matrix of the local defect-level GRU model, weights and biases of the pipe segment-level MLP model, and node embedding parameters of the graph neural network) using the LZ77 compression algorithm, and uploads them to the central analysis server via the HTTPS secure communication protocol. The amount of data uploaded is controlled within 10MB to avoid bandwidth consumption.

[0054] A differentiated collaborative optimization strategy is employed for model parameters at the local defect level and pipe segment level: a collaborative averaging algorithm based on parameter sparsity constraints is used to regularize the uploaded model parameters. Specifically, L1 regularization is used to select the top 15% of key parameters by absolute weight. Weighted averaging is performed only on these key parameters (weight allocation is based on the data volume ratio of each edge device: devices with a data volume ratio ≥30% have a weight coefficient of 0.6, those with 10%-30% have a weight coefficient of 0.3, and those with <10% have a weight coefficient of 0.1). Non-key parameters are set to zero to reduce data transmission redundancy and computational load.

[0055] For the node embedding parameters of the graph neural network: during the optimization process, the feature distribution of highly correlated edges is preserved. Highly correlated edges are defined as pipe connections with a medium conduction efficiency ≥ 0.8 (calculated based on the difference between medium humidity and pressure in the edge feature vector) and a physical connection strength ≥ 0.7 (evaluated based on pipe material and laying method). For the node embedding parameters corresponding to these edges, more than 85% of their original weight update amplitude is maintained during aggregation to ensure that the key correlation features between pipes are not weakened.

[0056] After model parameter verification and synchronization updates, the degree of adaptation is verified using an error consistency function (calculating the mean squared error of the test results of the global model and each edge-side model). The preset adaptation threshold is a mean squared error ≤ 5%. If the threshold requirement is met, the global model parameters are converted to TensorFlow Saved Model format and synchronously distributed to the corresponding edge-side devices through a secure channel. If the threshold is not met, the parameter acquisition and optimization process is re-executed.

[0057] After receiving the global parameters, the edge device first verifies the digital signature of the parameters (using the RSA-2048 encryption algorithm). After confirming the integrity and security of the parameters, it replaces the original parameters of the local model and backs up historical parameters (keeping the three most recent versions) so that it can quickly roll back in case of parameter anomalies and ensure the stable operation of the test system.

[0058] (v) Output Unit: Generation and issuance of graded test results and risk assessment instructions Based on the optimized models at each level, the output unit generates accurate test results and operation and maintenance instructions to realize intelligent operation and maintenance of the pipeline network. The specific implementation is as follows: The graded test results are output. Each local defect level edge device outputs the local defect level test results (corrosion propagation rate, peeling risk coefficient) for the next 24 hours every 30 minutes based on the updated GRU model. Each pipe section level edge device outputs the pipe section level aggregate evaluation results (average corrosion rate, anti-corrosion layer failure probability) for the next 48 hours every hour based on the updated MLP model, and uploads them to the central analysis server in real time.

[0059] Pipeline-level risk prediction and global map construction The central analysis server inputs the pipeline segment-level aggregation evaluation results and the pipeline association strength information (high association edges are marked in red, medium association edges are marked in yellow, and low association edges are marked in green) output by the pipeline topology association analysis model into the pipeline network-level XGBoost model to infer the predicted value of the pipeline network failure risk level (level 1-5) for the next 72 hours.

[0060] Based on the predicted values ​​of pipeline failure risk levels, the aggregated assessment results of each pipe segment, and the structured representation vectors of pipeline nodes generated by the graph neural network, a global corrosion risk distribution map is constructed and visualized in the form of a heat map: based on the geographical coordinates of the pipeline network, the risk level of each area is marked with different colors (level 1 blue, level 2 green, level 3 yellow, level 4 orange, level 5 red), and the specific location (accuracy ±10m) and expected spread range of corrosion hotspots are marked.

[0061] Risk optimization objective function and operation and maintenance instructions are generated. Based on the global corrosion risk distribution map, the risk optimization objective function is established based on the TOPSIS multi-objective decision algorithm. The objective function expression is: min(maintenance cost × 0.3 + failure loss × 0.1). The constraints are: pipeline corrosion rate ≤ 0.1 mm / year, anti-corrosion layer failure probability ≤ 0.2, and cathodic protection potential in the range of -1.2V to -0.8V.

[0062] Under constraints, a quadratic programming algorithm is used to solve the problem in real time, yielding three types of core operation and maintenance instructions: Local defect repair priority instructions: sorted by risk coefficient from high to low (priority levels 1-3), with level 1 priority defects requiring processing within 24 hours; Pipeline section corrosion protection enhancement parameters: For high-risk pipeline sections, increase the output anti-corrosion coating thickness (e.g., from 3mm to 5mm) and adjust the cathodic protection current range (e.g., from 0.5A / m). 2 Increased to 0.8A / m 2 ); Adjustment plan for cathodic protection of pipeline network: Optimize the output potential of each cathodic protection station to ensure uniform distribution of the overall protection potential of the pipeline network.

[0063] Command encoding and issuance: The above maintenance commands are encoded in a standardized JSON format, including fields such as command ID, generation timestamp, target pipe section / defect ID, command content, and execution period. They are sent to the field detection terminal (such as handheld inspection equipment) and the pipeline maintenance control unit (such as cathodic protection station controller and anti-corrosion construction equipment) through the cloud-edge collaborative interface (developed based on RESTful API). The command transmission delay is ≤1 second to ensure the timeliness and accuracy of maintenance actions.

[0064] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A metal pipeline corrosion and anti-corrosion coating peeling testing system, characterized in that, The system includes: The acquisition unit is configured to acquire multi-dimensional test data in real time. The multi-dimensional test data includes the local corrosion depth of the pipeline, the peeling area of ​​the anti-corrosion layer, the humidity of the medium in the pipeline section, the pH value of the soil, the pressure inside the pipeline, the resistance value of the anti-corrosion layer, the electrochemical potential of the pipeline welding point, and the cathodic protection potential of the pipeline network. It also acquires pipeline material parameters and laying environment parameters. The first processing unit is configured to construct a graded test and evaluation model based on the laying distribution of the pipeline network, the connection structure of the pipe segment, and the failure characteristics of the corrosion-anti-corrosion layer. The graded test and evaluation model includes a local defect-level edge test model, a pipe segment-level aggregation evaluation model, and a pipeline network-level risk prediction model. The second processing unit is configured to construct a pipeline topology correlation analysis model based on a graph neural network, taking a single pipeline unit as a graph node, the physical connection relationship of the pipeline and the medium conduction path as graph edges, and mapping the multi-dimensional test data into feature vectors of nodes and edges. The third processing unit is configured to periodically collect parameters from the local defect-level edge test model, the pipe segment-level aggregate evaluation model, and the pipeline topology correlation analysis model, perform distributed collaborative optimization, and update and optimize the parameters of each edge-side test model. The output unit is configured to obtain graded test results based on the optimized model, and output local defect warning, pipe section corrosion rate assessment and pipe network failure risk level in conjunction with the graded test results and the pipe network-level risk prediction model.

2. The metal pipeline corrosion and anti-corrosion coating peeling testing system according to claim 1, characterized in that: When collecting multi-dimensional test data in real time, including: On the side of the pipeline with local defects, ultrasonic flaw detectors, infrared thermal imagers, and electrochemical sensors are used to collect corrosion depth, anti-corrosion coating peeling area, and electrochemical potential of weld points, respectively; on the side of the pipeline laying section, soil sensors and humidity sensors are used to collect soil pH value and pipeline medium humidity. On the main pipeline side, pressure sensors and impedance testers are used to collect the pressure inside the pipeline and the impedance value of the anti-corrosion layer; on the cathodic protection side of the pipeline network, a potential monitor is used to obtain the cathodic protection potential of the pipeline network. Each test terminal device collects corresponding physical quantity signals according to a preset sampling period, and performs local encoding and format standardization on the collected results. The results are then sent to the edge device via the LoRa communication protocol and industrial bus. The edge device performs time calibration, abnormal data filtering, and feature dimension unification processing on the collected data.

3. The metal pipeline corrosion and anti-corrosion coating peeling test system according to claim 1, characterized in that, Based on the pipeline network layout, pipe segment connection structure, and corrosion-anti-corrosion coating failure characteristics, a graded testing and evaluation model is constructed. The construction process of the local defect-level edge testing model includes: On the edge side equipment corresponding to each target pipeline defect area, an input vector containing historical time series data is constructed based on the corrosion depth, anti-corrosion layer peeling area, weld point electrochemical potential, pipeline material parameters and laying environment parameters of the area. A test model is established based on a neural network structure, which is a gated recurrent network with a temporal feature extraction unit. The test model is trained based on the input vector to predict the local defect expansion trend within a preset time window in the future. The trained model is deployed on edge devices, and inference is performed based on real-time test data to output local defect-level test results.

4. The metal pipeline corrosion and anti-corrosion coating peeling test system according to claim 3, characterized in that, The construction process of the pipe segment-level aggregation evaluation model includes: On the edge-side device corresponding to each target pipe segment, the local defect level test results output by several local defect level edge test models under the pipe segment are received, and the aggregated input vector is constructed by combining the laying length of the pipe segment, soil type, historical corrosion data, cathodic protection potential and current stray current intensity. The network structure of the pipe segment-level aggregated evaluation model is based on a multilayer perceptron neural network. The model is trained based on the aggregated input vector to predict the average corrosion rate and the overall failure probability of the anti-corrosion layer within a preset time window of the pipe segment. The trained aggregate evaluation model is deployed in the pipe segment-level edge device, and the pipe segment-level aggregate evaluation result is output based on the real-time test results of each local defect-level edge test model.

5. The metal pipeline corrosion and anti-corrosion coating peeling test system according to claim 4, characterized in that, The construction process of the pipeline-level risk prediction model includes: On the central analysis server, the pipeline segment-level aggregated evaluation results uploaded by edge-side devices at each pipeline segment level are received, and the overall cathodic protection potential of the pipeline network, the pressure at each pipeline segment connection point, the distribution of stray currents in the surrounding area, and environmental temperature and humidity parameters are collected to construct a risk input vector. A pipeline-level risk prediction model is constructed based on a gradient boosting regression model. The pipeline-level risk prediction model is trained according to the risk input vector and is used to predict the distribution of corrosion hotspots and the overall failure risk level of the pipeline within a preset time window in the future. The pipeline-level risk prediction model is deployed in the central analysis server, and the inference results of the pipeline-level risk prediction model are used as the basis for prioritizing local defect repair, strengthening pipeline corrosion protection strategies, and adjusting pipeline cathodic protection parameters.

6. The metal pipeline corrosion and anti-corrosion coating peeling test system according to claim 5, characterized in that, A pipeline topology correlation analysis model is constructed based on a graph neural network. Individual pipeline units are treated as graph nodes, and the physical connections between pipelines and the medium conduction paths are treated as graph edges. When mapping the multi-dimensional test data into feature vectors for nodes and edges, the following steps are taken: Using each pipe unit in the test network as a graph node, and the pipe pairs that are related to each pipe unit through physical connection or medium conduction as graph edges, an undirected graph structure containing a set of nodes and a set of edges is constructed. The corrosion depth, anti-corrosion layer impedance value, pipe material parameters, and laying environment parameters of each pipe unit are used as node feature vectors, and the medium humidity of the pipe section, the pressure inside the pipe, and the cathodic protection potential difference value are used as edge feature vectors. A graph neural network model is constructed based on graph convolutional neural networks. The graph structure is embedded and learned, and a structured representation vector for each pipeline node is generated.

7. The metal pipeline corrosion and anti-corrosion coating peeling test system according to claim 6, characterized in that, When periodically collecting parameters from the local defect-level edge test model, the pipe segment-level aggregate evaluation model, and the pipeline topology correlation analysis model, performing distributed collaborative optimization, and updating and optimizing the parameters of each edge-side test model, the following steps are taken: A distributed collaboration module is deployed on the central analysis server. The distributed collaboration module sends parameter collaboration requests to each edge device at a preset time period to collect local training parameters of the local defect-level edge test model, the pipe segment-level aggregate evaluation model, and the graph neural network model. The distributed collaboration module performs global collaborative optimization by implementing a dynamic weight allocation strategy on the collected model parameters of each edge device, forming unified global model parameters. The global model parameters are updated and synchronously distributed to the corresponding edge devices to replace the local test model parameters.

8. The metal pipeline corrosion and anti-corrosion coating peeling test system according to claim 7, characterized in that, When the distributed collaboration module performs collaborative optimization on the local parameters uploaded by each edge device, it adopts differentiated optimization strategies for different model types, including: For the local defect-level edge testing model and the pipe segment-level aggregate evaluation model, a collaborative averaging algorithm based on parameter sparsity constraints is used to regularize the parameter updates. For the node embedding parameters of the graph neural network in the pipeline topology correlation analysis model, the feature distribution of highly correlated edges is preserved during the optimization process; After optimization, the degree of adaptation of the optimized model to the edge test results is verified by the error consistency function. If the preset adaptation threshold is met, the model parameter synchronization process is carried out.

9. The metal pipeline corrosion and anti-corrosion coating peeling testing system according to claim 1, characterized in that, Based on the optimized model, the graded test results are obtained. When the graded test results are combined with the pipeline network-level risk prediction model to output local defect warnings, pipeline section corrosion rate assessments, and pipeline network failure risk levels, the following are included: Based on the updated local defect level edge test model and pipe segment level aggregate evaluation model, the local defect level edge side equipment and pipe segment level edge side equipment respectively output the local defect level test results and pipe segment level aggregate evaluation results within the prediction time window in real time. The pipeline segment-level aggregation evaluation results and the pipeline association strength information output by the pipeline topology association analysis model are jointly input into the pipeline network-level risk prediction model to obtain the predicted value of the pipeline network failure risk level within the target time window. Based on the predicted value of pipeline failure risk level, the aggregated evaluation results of each pipe segment, and the structured representation vector of pipeline nodes generated by the graph neural network model as input features, a global corrosion risk distribution map is constructed. Based on the global corrosion risk distribution map, a risk optimization objective function is established based on a multi-objective decision-making algorithm. The risk optimization objective function is used to minimize the detection and maintenance costs and failure losses while meeting the pipeline safety operation requirements. Under the constraints of the risk optimization objective function, the local defect repair priority instruction, the pipe section anti-corrosion strengthening parameters, and the pipeline cathodic protection adjustment scheme are obtained in real time. The instructions and schemes are encoded in a standardized format and sent to the field testing terminal and pipeline operation and maintenance control unit through the cloud-edge collaboration interface.

Images