Exploration system for underground pipeline faults

Abstract

According to the exploring system for the underground pipeline faults provided by the invention, multi-modal state data acquisition and multi-modal feature fusion can be carried out on each node on an underground pipeline based on the exploring system for the underground pipeline faults provided by the invention; and determining a fault type T and a fault degree S of a corresponding node according to the multi-modal state characteristics based on a platform fault identification model, matching a corresponding node fault response measure by using a membership function according to the fault type T and the fault degree S of the node, and recommending a measure matched with the fault state of the current node for the current node. Therefore, the AI algorithm can be utilized to perform fusion state analysis on each node on the underground pipeline and intelligently recommend the most matched node fault response measure, so that invalid fault recognition caused by a single state parameter can be avoided, the optimal solution measure can be recommended for the fault point, and the fault diagnosis accuracy is improved. Therefore, the fault operation and maintenance feasibility and the fault solving reliability of the underground pipeline are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of pipeline operation and maintenance, and in particular to a system for detecting underground pipeline faults, a method for detecting underground pipeline faults, an electronic device, and a computer-readable storage medium. Background Art

[0002] Underground pipeline fault detection is crucial for ensuring the safe operation of urban infrastructure, preventing resource waste and environmental pollution, and avoiding secondary disasters caused by leaks or blockages. Traditional detection relies primarily on manual inspections, acoustic listening, tracer positioning, and ground-penetrating radar technology. These methods suffer from low efficiency, sensitivity to environmental interference, poor positioning accuracy, high costs, and an inability to detect non-metallic pipelines.

[0003] Although intelligent detection technology has gradually been incorporated into underground pipeline fault detection operations, some shortcomings still exist in practical applications. For example, patent number CN111367972B discloses an artificial intelligence-based community drainage pipeline detection method and system. For each detection point in the drainage pipeline, the site state parameters obtained through closed-circuit detection, periscope detection, and sonar detection are normalized. Then, based on the topological structure of the drainage pipeline and the characteristics of the state parameters of each site, the topological correlation of the detection sites is determined to form a site local network. The site state parameters of each detection site are used to simulate and analyze the changing patterns of the site local network through neural network training. Finally, for the drainage pipeline fault point discovered during the detection, the impact of the fault point on the site local network is predicted, and a visual display under the GIS information interface is used to achieve an intuitive display of the community drainage pipeline fault point and its impact. Although this technology uses neural networks to simulate and analyze the changing patterns of the local area network at a site, and to predict and visualize the impact of the fault point on the local area network, its fault detection technology still relies on "site state parameters obtained from closed-circuit detection, periscope detection, and sonar detection." Therefore, its method of obtaining pipeline parameters is relatively traditional and single-minded, and it is unable to fuse multiple state parameters for comprehensive prediction. The lack of state fusion assessment may cause errors in subsequent troubleshooting and analysis due to the invalidity of a single state parameter (for example, if the closed-circuit detection is fault-free but the periscope detection is faulty, this may lead to troubleshooting errors). In addition, this technology cannot use AI algorithms to intelligently recommend fault resolution measures based on fused state characteristics, making it less practical. Summary of the Invention

[0004] In order to solve the technical problems existing in the prior art, the present invention provides the following technical solutions: In one aspect, a system for detecting underground pipeline faults is provided, the system comprising: The data acquisition layer is used to explore the multimodal sensor data of each node on the underground pipeline; The data transmission layer is used to transmit the multimodal sensing data of each node to the data processing layer; The data processing layer is used to provide distributed storage and computing services, including: distributed storage of the multimodal sensor data of each node, and using a multimodal fusion algorithm to extract the multimodal state characteristics of each node, and using a preset fault identification model to determine the fault type (T) and fault severity (S) of the corresponding node based on the multimodal state characteristics; The application layer is used to use the membership function to match the corresponding node fault response action according to the node fault type: T and fault severity: S and output it for display: , is the membership function, indicating the measure The degree of adaptation to the current node failure state, arg max(·) represents the optimal measure according to the fault type: T and fault severity: S; The data acquisition layer is communicatively connected to the data processing layer via the data transmission layer; The data processing layer is communicatively connected to the application layer.

[0005] Furthermore, the data collection layer includes: Acoustic sensors, which detect pipe leaks with sound waves; Electromagnetic sensors, used to detect metal corrosion currents in pipelines; Pressure sensor, used to monitor pipeline pressure; Thermal imaging sensors to identify pipe temperatures‌; Laser radar, used to scan and obtain internal structural parameters of the pipeline; A mobile detection robot, configured to carry the acoustic sensor, electromagnetic sensor, pressure sensor, thermal imaging sensor, and lidar, and to detect the multimodal sensing data at various nodes on the underground pipeline; An industrial computer, configured to communicate with a host computer, control the mobile detection robot to move along the underground pipeline, and report the multimodal sensing data of each detected node to the host computer; The acoustic sensor, electromagnetic sensor, pressure sensor, thermal imaging sensor, laser radar and mobile detection robot are respectively electrically connected to the industrial control machine; The industrial computer is connected to the host computer for communication.

[0006] Furthermore, the industrial computer is also used for: The leakage sound waves, metal corrosion current, pipeline pressure, pipeline temperature and pipeline internal structure parameters of each node are separated and recombined into two-mode signals to obtain several sets of two-mode sensing data for each node, including: (leakage sound wave, metal corrosion current), (leakage sound wave, pipeline pressure), (leakage sound wave, pipeline temperature), (leakage sound wave, pipeline internal structure parameters); (metal corrosion current, pipeline pressure), (metal corrosion current, pipeline temperature), (metal corrosion current, pipeline internal structure parameters); (Pipeline pressure, pipeline temperature), (Pipeline pressure, pipeline internal structure parameters); (Pipeline temperature, pipeline internal structure parameters); Several groups of dual-modal sensing data of each node are transmitted to the data processing layer through the data transmission network.

[0007] Furthermore, the data transmission layer includes: Communication optical fiber is used to realize data transmission and communication control between the industrial computer and the host computer in the underground pipeline.

[0008] Furthermore, the data processing layer includes an edge server and a cloud AI platform that are communicatively connected, wherein: The edge server is equipped with an edge matrix conversion system, a distributed file storage system, and a feature fusion module, including: The edge matrix conversion system is used to perform matrix conversion processing on the dual-modal sensing data to obtain the corresponding dual-modal raw data matrix; The distributed file storage system is used for distributed storage of multiple groups of bimodal sensor data of each node; The feature fusion module is used to interact with the distributed file storage system, request to traverse several groups of the bimodal original data matrices of each node, and perform bimodal feature fusion to obtain several groups of bimodal state features of each node: Fi, where: the bimodal feature fusion formula is as follows: (n=2), Di is the bimodal original data matrix of group i; PCA(·) is the principal component analysis algorithm, which projects the data into a low-dimensional space; wi is the modal weight, which is set by mutual information or experience; The cloud AI platform is used to interact with the edge server, use a preset fault identification model to traverse and identify several groups of bimodal state features: Fi with fault characteristics at each node, and output the corresponding fault type: T and fault degree: S to the application layer.

[0009] Furthermore, the application layer, before using the membership function to match the corresponding node fault response action according to the fault type: T and fault severity: S of the node and outputting it for display, is further configured to: For each node with fault characteristics, the fault type: T and fault severity: S, respectively match the corresponding measures .

[0010] Furthermore, the method for generating the fault identification model includes: Collect historical fault operation and maintenance big data of each node, including several groups of bimodal raw data matrices with fault characteristics; Using the bimodal feature fusion formula, the bimodal state features of each node are extracted from the bimodal original data matrix. ; For each set of bimodal state features with fault characteristics at each node Perform fault severity evaluation and feature annotation, label the corresponding fault type: T, and label the measures corresponding to the fault type: T and fault severity: S ; Among them, the evaluation method of the fault degree: S is: , Score( ) is the feature scoring function (linear or nonlinear mapping), For feature weights, set through mutual information or experience; Count the bimodal state features of each group of nodes , and construct a feature set;‌ Dividing the feature set into a training set and a validation set according to a preset ratio; Input the training set into the preset RF random forest model to perform feature training and learning to generate the initial fault recognition model; The recognition performance of the fault recognition model is verified using the validation set: If the verification is successful, the fault identification model is deployed and applied to the cloud AI platform; Otherwise, repeat the above steps to regenerate the fault identification model.

[0011] In another aspect, a method for detecting underground pipeline faults is provided. The method for detecting underground pipeline faults is implemented based on the above-mentioned system for detecting underground pipeline faults. The method includes: Through the data acquisition layer, the multimodal sensor data of each node on the underground pipeline is detected and transmitted to the data processing layer; After the data processing layer receives the multimodal sensor data from each node on the underground pipeline being explored: The edge matrix conversion system performs matrix conversion processing on the dual-modal sensor data, obtains the corresponding dual-modal raw data matrix, and stores it in the distributed file storage system; The feature fusion module interacts with the distributed file storage system, requests to traverse several groups of the bimodal original data matrices of each node, and performs bimodal feature fusion to obtain several groups of bimodal state features of each node and send them to the cloud AI platform: Fi, where the bimodal feature fusion formula is as follows: (n=2), Di is the bimodal original data matrix of group i; PCA(·) is the principal component analysis algorithm, which projects the data into a low-dimensional space; wi is the modal weight, which is set by mutual information or experience; The cloud AI platform uses a preset fault identification model to traverse and identify several sets of bimodal state features (Fi) with fault characteristics at each node, and outputs the corresponding fault type (T) and fault severity (S) to the application layer. The application layer matches the corresponding measures to the fault type (T) and fault severity (S) of each node. ; And, using the membership function, according to the node fault type: T and fault severity: S, match the corresponding node fault response measure Action and output it for display: , is the membership function, indicating the measure The degree of adaptation to the current node failure state, arg max(·) represents the optimal action according to the fault type: T and fault severity: S.

[0012] On the other hand, an electronic device is provided, comprising: a processor; and a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the above method is implemented.

[0013] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, and the at least one instruction is loaded and executed by a processor to implement the above method.

[0014] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least: The underground pipeline fault detection system based on the present invention can obtain multimodal state data of each node on the underground pipeline and perform multimodal feature fusion. After determining the fault type: T and fault degree: S of the corresponding node based on the multimodal state features based on the platform fault identification model, the corresponding node fault response measures are matched according to the node fault type: T and fault degree: S using the membership function, thereby recommending measures adapted to the fault state of the current node. Therefore, the present invention can use the AI ​​algorithm to perform fusion state analysis on each node on the underground pipeline and intelligently recommend the most matching node fault response measures. It can not only avoid invalid fault identification caused by a single state parameter, but also recommend the best solution for the fault point, thereby greatly improving the feasibility of underground pipeline fault operation and maintenance and the reliability of fault resolution. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0016] Figure 1 This is a schematic diagram of the architecture of a system for detecting underground pipeline faults provided by an embodiment of the present invention; Figure 2 This is a schematic diagram of a process mechanism for matching optimal measures based on membership provided by an embodiment of the present invention; Figure 3 This is a training flow chart of a fault identification model provided by an embodiment of the present invention; Figure 4 It is a structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0018] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as an "exemplary" in the present invention should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of the word "exemplary" is intended to present concepts in a concrete manner. Furthermore, in the embodiments of the present invention, "and / or" can mean both or either of the two.

[0019] In the embodiments of the present invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, when the distinction is not emphasized, the meanings they convey are the same. The terms "of," "corresponding," and "corresponding" may sometimes be used interchangeably. It should be noted that, when the distinction is not emphasized, the meanings they convey are the same.

[0020] In the embodiments of the present invention, sometimes a subscript such as W1 may be mistakenly written as a non-subscript form such as W1. When the difference is not emphasized, the meanings to be expressed are the same.

[0021] In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0022] The embodiment of the present invention provides a system for detecting underground pipeline faults. Figure 1 The architecture diagram of the underground pipeline fault detection system shown in FIG. 1 , on the one hand, provides a system for detecting underground pipeline faults, the system comprising: The data acquisition layer is used to explore the multimodal sensor data of each node on the underground pipeline; The data transmission layer is used to transmit the multimodal sensing data of each node to the data processing layer; The data processing layer is used to provide distributed storage and computing services, including: distributed storage of the multimodal sensor data of each node, and using a multimodal fusion algorithm to extract the multimodal state characteristics of each node, and using a preset fault identification model to determine the fault type (T) and fault severity (S) of the corresponding node based on the multimodal state characteristics; The application layer is used to use the membership function to match the corresponding node fault response action according to the node fault type: T and fault severity: S and output it for display: , is the membership function, indicating the measure The degree of adaptation to the current node failure state, arg max(·) represents the optimal measure according to the fault type: T and fault severity: S; The data acquisition layer is communicatively connected to the data processing layer via the data transmission layer; The data processing layer is communicatively connected to the application layer.

[0023] The system architecture is as attached Figure 1As shown, the present invention uses a pipe-climbing robot (i.e., a mobile detection robot, a detection robot that can move along an underground pipeline, which can be implemented in combination with a pipeline robot, such as a "wheeled robot" or a "crawler robot", which can walk along the underground pipeline and advance according to the pre-set coordinate positions of each pipeline node under the path planning system of the host computer, reach the corresponding coordinates, explore each node, and use the sensor module or system it carries to collect multimodal sensor data and transmit it back to the edge server).

[0024] In this invention, the mobile exploration robot communicates and controls with the host computer via an industrial computer, and can also report collected exploration data to the host computer. The data transmission layer uses optical fiber communication for stable data transmission over long-distance trunk pipelines. The data processing layer provides distributed storage and AI computing and analysis services, deploying: Edge computing nodes: pre-process sensor data (denoising, normalization) to reduce cloud load; ‌Cloud AI platform‌: Deploys multimodal fusion algorithms and fault identification models, and supports distributed computing.‌ The application layer uses PC, APP, cockpit large screen, etc. to provide a fault diagnosis interface: display the fault type, degree and recommended measures, and support GIS map visualization. The pipeline detection method of the present invention will be described in detail below.

[0025] During the exploration, the upper computer path planning system (specifically planned in conjunction with the background application of the mobile detection robot configured by the user) can control the mobile detection robot to go to the coordinates of each pipeline node on the underground pipeline to explore the data. In order to improve the exploration efficiency and the accuracy of the detection of node faults, this office proposes the following dynamic priority-driven two-stage path planning method, which combines multi-modal fault feature recognition and path optimization algorithm to achieve rapid response and accurate exploration of emergency fault nodes by mobile robots. The core process of the solution is divided into three stages: initial full-node detection → fault urgency assessment → secondary emergency node directional exploration, to ensure that resources are allocated first to high-urgency fault nodes, while taking into account path efficiency: ‌1. Initial full node detection phase‌ Objective: Collect multimodal status data (vibration, temperature, gas concentration, images, etc.) from all pipeline nodes.

[0026] ‌Path Planning Algorithm‌: Use the improved TSP (Traveling Salesman Problem) algorithm to generate the shortest traversal path based on the pipeline node coordinates and minimize the robot's movement distance.

[0027] Introducing a dynamic obstacle avoidance mechanism that uses lidar to detect obstacles in the path in real time and trigger local path replanning (such as the RRT* algorithm).

[0028] Data Collection: The robot is equipped with multimodal sensors, arrives at each node coordinate in turn, collects original status data and uploads it to the host computer in real time.

[0029] ‌2. Fault urgency assessment phase‌ Multimodal feature fusion: The host computer preprocesses the raw data (denoising and normalization) and extracts key features (such as vibration spectrum peaks and temperature anomaly gradients).

[0030] Use multimodal fusion models (such as neural networks based on attention mechanisms) to jointly analyze multi-source data such as vibration, temperature, and gas.

[0031] Fault classification: The system matches the fault characteristics to the preset rule base (e.g., "temperature > 80°C + gas concentration exceeds the limit = high urgency") and outputs the fault level (low / medium / high).

[0032] Generate an urgency distribution heat map and mark the coordinates of high-urgency nodes and fault types.

[0033] ‌3. Secondary emergency node directional exploration phase‌ Dynamic Priority Path Planning: Input: List of high-urgency nodes, current robot position, pipeline topology map.

[0034] Optimization goal: Minimize the total travel distance while giving priority to covering high-urgency nodes.

[0035] ‌Algorithm Design‌: Urgency-weighted A* algorithm: Converts node urgency into path cost weights (high urgency node weight = 10, medium = 5, low = 1) to optimize the global path.

[0036] Real-time replanning mechanism: If a new emergency node (such as a sudden leak) is found during the secondary exploration, the local path insertion logic is triggered.

[0037] ‌Directed Data Exploration‌: The robot arrives at the high-urgency nodes in sequence according to the planned path and starts the depth detection mode (hereinafter referred to as the multimodal exploration mode of the present invention).

[0038] The secondary collected data is transmitted back to the host computer for fault re-examination and maintenance decision support (i.e., the following implementation plan of the present invention).

[0039] This approach reduces the initial full-node detection time to 70% of traditional solutions (TSP optimization + dynamic obstacle avoidance). The secondary detection path length is reduced by 30% (using the urgency-weighted A* algorithm). The detection latency for high-urgency nodes is reduced to under 5 minutes, preventing fault escalation. Dynamic path adjustments are supported in the event of multimodal data anomalies (e.g., switching to redundant paths in the event of sensor failure).

[0040] The robot is integrated with the following sensors: an acoustic sensor (to detect acoustic waves from pipeline leaks), an electromagnetic sensor (to detect metal corrosion currents), a pressure sensor (to monitor pressure anomalies), and a thermal imaging sensor (to identify temperature anomalies). The integrated installation structure, quantity, model, and location of these sensors are not specified here; the user can configure and install them based on the operating parameters of the selected sensor module. The industrial computer can report the multimodal sensor data from each detected node to a host computer. The industrial computer can also communicate with the host computer to control the movement of the mobile detection robot along the underground pipeline and report the multimodal sensor data from each detected node to the host computer.

[0041] The following describes a solution of the present invention in which the industrial computer samples and processes data and then reports it to the host computer.

[0042] Furthermore, the industrial computer is also used for: The leakage sound waves, metal corrosion current, pipeline pressure, pipeline temperature and pipeline internal structure parameters of each node are separated and recombined into two-mode signals to obtain several sets of two-mode sensing data for each node, including: (leakage sound wave, metal corrosion current), (leakage sound wave, pipeline pressure), (leakage sound wave, pipeline temperature), (leakage sound wave, pipeline internal structure parameters); (metal corrosion current, pipeline pressure), (metal corrosion current, pipeline temperature), (metal corrosion current, pipeline internal structure parameters); (Pipeline pressure, pipeline temperature), (Pipeline pressure, pipeline internal structure parameters); (Pipeline temperature, pipeline internal structure parameters); Several groups of dual-modal sensing data of each node are transmitted to the data processing layer through the data transmission network.

[0043] This embodiment uses a dual-modal signal separation and reassembly (at the industrial control computer level) implementation plan: The industrial control computer performs dual-modal signal separation and reassembly on each node's multi-source sensor data (leakage acoustic wave, metal corrosion current, pipeline pressure, temperature, and internal structural parameters), generating multiple pairs of dual-modal sensor data (e.g., "leakage acoustic wave + metal corrosion current") to facilitate state feature fusion and avoid errors caused by invalid state detection due to a single feature. Specifically, the following steps are included: 1) Signal alignment and preprocessing: Synchronize data from different sensors through timestamps, eliminate timing deviations, and perform noise filtering and normalization.

[0044] 2) Bimodal pairing: Based on physical correlation or fault scenario requirements, sensor data is combined in pairs (such as sound waves and current, pressure and temperature, etc.) to form 10 sets of bimodal data pairs (see the combined dataset above).

[0045] 3) Data compression and encapsulation: A lightweight compression algorithm (such as LZW or Delta encoding) is used on the reorganized bimodal data to reduce transmission bandwidth usage.

[0046] Technical Effects: Reduce Single Feature Error: By complementing multimodal data (e.g., acoustic detection for leaks, current detection for corrosion), we avoid misjudgments caused by noise or failure of a single sensor. Reduce Redundant Transmission: Transmitting only dual-modal combined data (rather than all raw data) reduces network bandwidth pressure by approximately 50%. Support Flexible Analysis: Multiple dual-modal combinations cover different failure modes (e.g., leakage, corrosion, structural deformation), enhancing system robustness.

[0047] After the industrial computer pre-groups the data, the pressure of data transmission can be reduced, thereby reducing the AI ​​algorithm computing power of the host computer.

[0048] Furthermore, the data processing layer includes an edge server and a cloud AI platform that are communicatively connected, wherein: The edge server is equipped with an edge matrix conversion system, a distributed file storage system, and a feature fusion module to implement localized data processing and feature fusion. The edge matrix conversion system is used to perform matrix conversion processing on the dual-modal sensor data to obtain the corresponding dual-modal raw data matrix; the edge matrix conversion system performs matrix processing on the dual-modal sensor data, for example: Sound wave signal → time-frequency matrix (STFT transform); Pressure signal → time series matrix (sliding window segmentation); Output format: unified into a standardized N×M matrix to facilitate subsequent feature extraction.

[0049] The distributed file storage system is used to distribute the storage of multiple sets of bimodal sensor data of each node; it adopts a distributed storage architecture (such as HDFS or MinIO) to store bimodal data in shards according to node ID and timestamp, supporting high concurrent reading and writing and disaster recovery backup; The feature fusion module is used to interact with the distributed file storage system, request to traverse several groups of the bimodal original data matrices of each node, and perform bimodal feature fusion to obtain several groups of bimodal state features of each node: Fi, where: the bimodal feature fusion formula is as follows: (n=2), Di is the bimodal original data matrix of group i; PCA(·) is the principal component analysis algorithm, which projects the data into a low-dimensional space; wi is the modal weight, which is set by mutual information or experience; The cloud AI platform is used to interact with the edge server, use a preset fault identification model to traverse and identify several groups of bimodal state features: Fi with fault characteristics at each node, and output the corresponding fault type: T and fault degree: S to the application layer.

[0050] Feature fusion module: uses a multimodal feature fusion algorithm (such as attention mechanism + weighted splicing). The specific bimodal feature fusion formula is shown above: Extract time domain and frequency domain features (such as mean, variance, and spectral entropy) for each set of bimodal matrices; Dynamically assigning modality importance through attention weights (e.g., higher weight for sound waves in leakage scenarios); Output the fused bimodal state feature Fi (feature vector with unified dimension).

[0051] The modal weight can be configured by the administrator through mutual information or experience setting.

[0052] The cloud-based AI platform performs fault identification and assessment based on the dual-modal state features Fi after edge fusion, including: 1) Fault identification model: Model architecture: A dual-channel convolutional neural network (CNN) is used to process features of different modalities separately, and finally a fully connected layer is used for joint classification.

[0053] Input: Dual-modal state features Fi (such as acoustic wave-current fusion features, pressure-temperature fusion features).

[0054] Output: Fault type T (such as leakage, corrosion, blockage) and fault severity S (such as minor, moderate, severe).

[0055] 2) Dynamic model update: Based on incremental data fed back by the edge, the model is trained online regularly (e.g., updated weekly) to adapt to device aging or environmental changes.

[0056] 3) Technical Effects: High-Precision Identification: Dual-modal feature input increases model accuracy to over 95% (compared to 80% for a single modality). Reduced False Alarm Rate: Multimodal cross-validation reduces interference from a single feature, lowering the false alarm rate to below 5%. Adaptability: Dynamically updates model parameters to adapt to long-term factors such as pipeline aging and changes in ambient temperature and humidity.

[0057] Furthermore, the application layer, before using the membership function to match the corresponding node fault response action according to the fault type: T and fault severity: S of the node and outputting it for display, is further configured to: For each node with fault characteristics, the fault type: T and fault severity: S, respectively match the corresponding measures .

[0058] like Figure 2 As shown in the figure, because there are multiple sets of bimodal state features: Fi at each node, and some sets of bimodal state features: Fi may be invalid features (for example, blank features after fusion), it is necessary to use the preset fault identification model to traverse and identify several sets of bimodal state features: Fi with fault features at each node, and then output the fault type: T and fault degree: S of the nodes with fault features to the application layer for analysis. Here, each node may have multiple sets of fault types: T and fault degrees: S; therefore, the subsequent application layer can take corresponding measures of type and degree based on the multiple sets of fault types: T and fault degrees: S of each node. to match.

[0059] After matching, the algorithm of the node failure response action of this solution is finally combined to output the best node failure solution: , is the membership function, indicating the measure The degree of adaptation to the current node fault state, the membership function is selected by the system default or user-defined; arg max(·) represents the optimal action according to the fault type: T and fault severity: S.

[0060] Through the membership function, the optimal measures (such as "high-pressure flushing", "local repair welding", "overall replacement") are matched according to the fault type T and degree S. The system can match the corresponding measures for each group of nodes with several sets of bimodal state characteristics: Fi However, in order to recommend the most effective measures for the failed nodes, we have measured the , find the most suitable measure for the current node failure status And recommend (the system uses AI model to recommend corresponding measures based on fault type: T and fault severity: S ), for example, after calculating the output through arg max(·), it is found that the measure for the current pipeline node is "local repair welding", indicating that the local welding failure is the most obvious, which reminds the inspection and maintenance personnel to perform "local repair welding" on this node. The coordinate positions of the robot walking to each pipeline node are pre-recorded by the system, so the coordinate position of the current pipeline node and the final output measure can be used to calculate the total cost of the pipeline node. Binding is performed and a maintenance table for the current pipeline node is generated (other features can be included in the maintenance table as supplements), which is then sent to the application end (such as the inspection terminal) of the inspection and maintenance personnel to locate the maintenance and improve efficiency.

[0061] Furthermore, the method for generating the fault identification model includes: Collect historical fault operation and maintenance big data of each node, including several groups of bimodal raw data matrices with fault characteristics; Using the bimodal feature fusion formula, the bimodal state features of each node are extracted from the bimodal original data matrix. ; For each set of bimodal state features with fault characteristics at each node Perform fault severity evaluation and feature annotation, label the corresponding fault type: T, and label the measures corresponding to the fault type: T and fault severity: S ; Among them, the evaluation method of the fault degree: S is: , Score( ) is the feature scoring function (linear or nonlinear mapping), For feature weights, set through mutual information or experience; Count the bimodal state features of each group of nodes , and construct a feature set;‌ Dividing the feature set into a training set and a validation set according to a preset ratio; Input the training set into the preset RF random forest model to perform feature training and learning to generate the initial fault recognition model; The recognition performance of the fault recognition model is verified using the validation set: If the verification is successful, the fault identification model is deployed and applied to the cloud AI platform; Otherwise, repeat the above steps to regenerate the fault identification model.

[0062] like Figure 3As shown in the figure, we have deployed an AI model (fault identification model) on the system's cloud AI platform (cloud server), which can implement different levels of strategy recommendations based on AI technology. Specifically: 1) Data collection and preprocessing Multimodal historical fault data collection‌ Collect historical operation and maintenance data from the cloud AI platform and node devices, including metrics (such as CPU load and disk I / O) and logs (such as error logs and operation records), to form a bimodal raw data matrix (primarily multimodal sensor data from nodes identified as faulty).

[0063] The data must cover normal conditions and various failure scenarios (such as network delays and hardware failures) to ensure data diversity.

[0064] Bimodal data alignment and cleaning Align the timestamps of the raw data to eliminate noise (such as invalid log entries and outliers) and ensure data consistency.

[0065] 2) Dual-modal feature fusion and feature extraction Feature fusion formula design‌ Use weighted fusion or tensor decomposition to map indicator data (numeric) and log data (text) into the same feature space.

[0066] For example, perform word embedding (such as BERT) on the log text to generate a vector, which is then concatenated with the indicator value to form a joint feature vector.

[0067] The fusion formula example is shown in the above-mentioned present invention.

[0068] State feature extraction‌ Extract key state features from the fused features, such as time series volatility (indicator variance) and text keyword frequency (log error type statistics).

[0069] 3) Fault annotation and feature set construction measure Annotated with type (T)‌ Fault severity: refer to the evaluation method of S in the present invention; Fault type: This type is labeled based on the operation and maintenance knowledge base (e.g., T1-network fault, T2-hardware fault).

[0070] Action Note: Remedial action associated with fault type and severity (For example, T1 → restart the network device, T2 → replace the hard disk).

[0071] Feature set construction and partitioning‌ Bimodal features are counted by node to construct a structured feature set (feature dimension ≥ 50).

[0072] The training set and validation set are divided into a ratio of 7:3 to ensure balanced distribution of categories.

[0073] 4) Random forest model training and verification Model training Input the training set to the RF model and set the hyperparameters: The number of trees (n_estimators=200), the maximum depth (max_depth=10), and the feature sampling ratio (max_features=√d).

[0074] The splitting nodes are selected by the Gini coefficient to optimize the classification purity.

[0075] Model Validation‌ Verification indicators: Precision>90%, Recall>85%, F1-Score>88%.

[0076] If the standard is not met, you need to: check the feature fusion weights (α, β); Increasing the amount of data or introducing adversarial examples can enhance generalization.

[0077] V) Model deployment and application Cloud AI platform integration‌ The verified models are encapsulated as API services and integrated into the cloud-based intelligent operation and maintenance platform to support real-time fault detection.

[0078] Dynamic optimization mechanism‌ Regularly upload new fault data to the training set to trigger incremental model training and adapt to changes in the system environment.

[0079] This solution can improve fault identification accuracy by 30%-50%, while reducing manual operation and maintenance costs by more than 40%.

[0080] Feature scoring functions can be categorized into two types: linear and nonlinear. Their core goal is to assign a quantitative score to a feature that reflects its predictive power for the target variable. For example, a linear scoring function directly calculates the strength of the association between a feature and the target variable through a linear relationship. A common example is the linear regression coefficient method, which will not be discussed in detail here. Feature importance is measured based on nonlinear assumptions or statistical dependencies. Mutual information (MI) is a common method for measuring the statistical independence of a feature and the target variable.

[0081] Random forest feature importance: based on the split gain or frequency of occurrence of the feature in the decision tree.

[0082] Empirical setting method: ‌‌Domain knowledge guidance‌‌, combining expert experience to manually adjust weights (such as the significant impact of a certain indicator on disease prediction in clinical medicine), or generating a dynamic penalty factor through LLM (such as using a large language model to inject domain knowledge to correct the Lasso λ parameter in the LLM-Lasso framework)‌.

[0083] In another aspect, a method for detecting underground pipeline faults is provided. The method for detecting underground pipeline faults is implemented based on the above-mentioned system for detecting underground pipeline faults. The method includes: Through the data acquisition layer, the multimodal sensor data of each node on the underground pipeline is detected and transmitted to the data processing layer; After the data processing layer receives the multimodal sensor data from each node on the underground pipeline being explored: The edge matrix conversion system performs matrix conversion processing on the dual-modal sensor data, obtains the corresponding dual-modal raw data matrix, and stores it in the distributed file storage system; The feature fusion module interacts with the distributed file storage system, requests to traverse several groups of the bimodal original data matrices of each node, and performs bimodal feature fusion to obtain several groups of bimodal state features of each node and send them to the cloud AI platform: Fi, where the bimodal feature fusion formula is as follows: (n=2), Di is the bimodal original data matrix of group i; PCA(·) is the principal component analysis algorithm, which projects the data into a low-dimensional space; wi is the modal weight, which is set by mutual information or experience; The cloud AI platform uses a preset fault identification model to traverse and identify several sets of bimodal state features (Fi) with fault characteristics at each node, and outputs the corresponding fault type (T) and fault severity (S) to the application layer. The application layer matches the corresponding measures to the fault type (T) and fault severity (S) of each node. ; And, using the membership function, according to the node fault type: T and fault severity: S, match the corresponding node fault response measure Action and output it for display: , is the membership function, indicating the measure The degree of adaptation to the current node failure state, arg max(·) represents the optimal action according to the fault type: T and fault severity: S.

[0084] Please understand the above steps in conjunction with the corresponding applications of the previous system, and will not be repeated here.

[0085] Figure 4 is a schematic structural diagram of an electronic device provided by an embodiment of the present invention, such as Figure 4 As shown, the electronic device can implement the program of the above-mentioned method for detecting underground pipeline faults. Optionally, the electronic device 410 can include a first processor 2001.

[0086] Optionally, the electronic device 410 may further include a memory 2002 and a transceiver 2003 .

[0087] The first processor 2001, the memory 2002 and the transceiver 2003 may be connected via a communication bus.

[0088] The following combination Figure 4 The components of the electronic device 410 are described in detail. The first processor 2001 is the control center of the electronic device 410 and can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 can be one or more central processing units (CPUs), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs) or one or more field programmable gate arrays (FPGAs).

[0089] Optionally, the first processor 2001 can execute various functions of the electronic device 410 by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0090] In a specific implementation, as an embodiment, the first processor 2001 may include one or more CPUs, such as Figure 4 CPU0 and CPU1 are shown in FIG.

[0091] In a specific implementation, as an embodiment, the electronic device 410 may also include multiple processors, such as Figure 41 and 2. The first processor 2001 and the second processor 2004 are shown in FIG. Each of these processors can be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). A processor herein can refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0092] The memory 2002 is used to store the software program for executing the solution of the present invention, and is controlled by the first processor 2001 for execution. The specific implementation method can refer to the above method embodiment and will not be repeated here.

[0093] Alternatively, the memory 2002 may be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM) or other type of dynamic storage device that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, an optical disc storage (including a compact disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 2002 may be integrated with the first processor 2001 or exist independently and accessed through the interface circuit ( Figure 4 (not shown) is coupled to the first processor 2001, which is not specifically limited in this embodiment of the present invention.

[0094] The transceiver 2003 is used to communicate with a network device or a terminal device.

[0095] Optionally, the transceiver 2003 may include a receiver and a transmitter ( Figure 4 The receiver is used to implement a receiving function, and the transmitter is used to implement a sending function.

[0096] Optionally, the transceiver 2003 may be integrated with the first processor 2001, or may exist independently and communicate with the first processor 2001 through the interface circuit ( Figure 4 (not shown) is coupled to the first processor 2001, which is not specifically limited in this embodiment of the present invention.

[0097] It should be noted that Figure 4 The structure of the electronic device 410 shown in the figure does not constitute a limitation on the router. The actual knowledge structure recognition device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0098] In addition, the technical effects of the electronic device 410 can refer to the technical effects of the underground pipeline fault detection system described in the above method embodiment, and will not be repeated here.

[0099] It should be understood that the first processor 2001 in the embodiment of the present invention may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor, or the processor may be any conventional processor, etc.

[0100] It should also be understood that the memory in the embodiments of the present invention may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory may be random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0101] The above embodiments can be implemented in whole or in part via software, hardware (e.g., circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions or computer programs. When loaded or executed on a computer, the processes or functions described in accordance with the embodiments of the present invention are fully or partially performed. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable method. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired means (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium accessible by a computer, or a data storage device such as a server or data center that contains a collection of one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0102] It should be understood that the term "and / or" as used herein simply describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A alone, A and B together, or B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the associated objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0103] In this disclosure, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, "at least one of a, b, or c" can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or plural.

[0104] It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0105] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0106] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices, methods and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0107] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, methods, and methods can be implemented in other ways. For example, the method embodiments described above are merely illustrative. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, method or unit, which can be electrical, mechanical or other forms.

[0108] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0109] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0110] If the functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical disks.

[0111] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A system for detecting underground pipeline faults, characterized in that: The system comprises: The data acquisition layer is used to explore the multimodal sensor data of each node on the underground pipeline; The data transmission layer is used to transmit the multimodal sensing data of each node to the data processing layer; The data processing layer is used to provide distributed storage and computing services, including: distributed storage of the multimodal sensor data of each node, and using a multimodal fusion algorithm to extract the multimodal state characteristics of each node, and using a preset fault identification model to determine the fault type (T) and fault severity (S) of the corresponding node based on the multimodal state characteristics; The application layer is used to use the membership function to match the corresponding node fault response action according to the node fault type: T and fault severity: S and output it for display: ， is the membership function, indicating the measure The degree of adaptation to the current node failure state, arg max(·) represents the optimal measure according to the fault type: T and fault severity: S; The data acquisition layer is communicatively connected to the data processing layer via the data transmission layer; The data processing layer is communicatively connected to the application layer.

2. The underground pipeline fault detection system according to claim 1, characterized in that: The data collection layer includes: Acoustic sensors, which detect pipe leaks with sound waves; Electromagnetic sensors, used to detect metal corrosion currents in pipelines; Pressure sensor, used to monitor pipeline pressure; Thermal imaging sensors to identify pipe temperatures‌; Laser radar, used to scan and obtain internal structural parameters of the pipeline; A mobile detection robot, configured to carry the acoustic sensor, electromagnetic sensor, pressure sensor, thermal imaging sensor, and lidar, and to detect the multimodal sensing data at various nodes on the underground pipeline; An industrial computer, configured to communicate with a host computer, control the mobile detection robot to move along the underground pipeline, and report the multimodal sensing data of each detected node to the host computer; The acoustic sensor, electromagnetic sensor, pressure sensor, thermal imaging sensor, laser radar and mobile detection robot are respectively electrically connected to the industrial control machine; The industrial computer is connected to the host computer for communication.

3. The underground pipeline fault detection system according to claim 2, characterized in that: The industrial computer is also used for: The leakage sound waves, metal corrosion current, pipeline pressure, pipeline temperature and pipeline internal structure parameters of each node are separated and recombined into two-mode signals to obtain several sets of two-mode sensing data for each node, including: (leakage sound wave, metal corrosion current), (leakage sound wave, pipeline pressure), (leakage sound wave, pipeline temperature), (leakage sound wave, pipeline internal structure parameters); (metal corrosion current, pipeline pressure), (metal corrosion current, pipeline temperature), (metal corrosion current, pipeline internal structure parameters); (Pipeline pressure, pipeline temperature), (Pipeline pressure, pipeline internal structure parameters); (Pipeline temperature, pipeline internal structure parameters); Several groups of dual-modal sensing data of each node are transmitted to the data processing layer through the data transmission network.

4. The underground pipeline fault detection system according to claim 1, characterized in that: The data transmission layer includes: Communication optical fiber is used to realize data transmission and communication control between the industrial computer and the host computer in the underground pipeline.

5. The underground pipeline fault detection system according to claim 3, characterized in that: The data processing layer includes an edge server and a cloud AI platform that are communicatively connected, wherein: The edge server is equipped with an edge matrix conversion system, a distributed file storage system, and a feature fusion module, including: The edge matrix conversion system is used to perform matrix conversion processing on the dual-modal sensing data to obtain the corresponding dual-modal raw data matrix; The distributed file storage system is used for distributed storage of multiple groups of bimodal sensor data of each node; The feature fusion module is used to interact with the distributed file storage system, request to traverse several groups of the bimodal original data matrices of each node, and perform bimodal feature fusion to obtain several groups of bimodal state features of each node: Fi, where: the bimodal feature fusion formula is as follows: （n=2）， Di is the bimodal original data matrix of group i; PCA(·) is the principal component analysis algorithm, which projects the data into a low-dimensional space; wi is the modal weight, which is set by mutual information or experience; The cloud AI platform is used to interact with the edge server, use a preset fault identification model to traverse and identify several groups of bimodal state features: Fi with fault characteristics at each node, and output the corresponding fault type: T and fault degree: S to the application layer.

6. The underground pipeline fault detection system according to claim 5, characterized in that: The application layer, before using the membership function to match the corresponding node fault response action according to the node fault type: T and fault severity: S and outputting it for display, is further used to: For each node with fault characteristics, the fault type: T and fault severity: S, respectively match the corresponding measures .

7. The underground pipeline fault detection system according to claim 5, characterized in that: The method for generating the fault identification model includes: Collect historical fault operation and maintenance big data of each node, including several groups of bimodal raw data matrices with fault characteristics; Using the bimodal feature fusion formula, the bimodal state features of each node are extracted from the bimodal original data matrix. ; For each set of bimodal state features with fault characteristics at each node Perform fault severity evaluation and feature annotation, label the corresponding fault type: T, and label the measures corresponding to the fault type: T and fault severity: S ; Among them, the evaluation method of the fault degree: S is: ， Score( ) is the feature scoring function (linear or nonlinear mapping), For feature weights, set through mutual information or experience; Count the bimodal state features of each group of nodes , and construct a feature set;‌ Dividing the feature set into a training set and a validation set according to a preset ratio; Input the training set into the preset RF random forest model to perform feature training and learning to generate the initial fault recognition model; The recognition performance of the fault recognition model is verified using the validation set: If the verification is successful, the fault identification model is deployed and applied to the cloud AI platform; Otherwise, repeat the above steps to regenerate the fault identification model.

8. A method for detecting underground pipeline faults, wherein the method is implemented based on the underground pipeline fault detection system according to any one of claims 1 to 7, and is characterized in that: The method comprises: Through the data acquisition layer, the multimodal sensor data of each node on the underground pipeline is detected and transmitted to the data processing layer; After the data processing layer receives the multimodal sensor data from each node on the underground pipeline being explored: The edge matrix conversion system performs matrix conversion processing on the dual-modal sensor data, obtains the corresponding dual-modal raw data matrix, and stores it in the distributed file storage system; The feature fusion module interacts with the distributed file storage system, requests to traverse several groups of the bimodal original data matrices of each node, and performs bimodal feature fusion to obtain several groups of bimodal state features of each node and send them to the cloud AI platform: Fi, where the bimodal feature fusion formula is as follows: （n=2）， Di is the bimodal original data matrix of group i; PCA(·) is the principal component analysis algorithm, which projects the data into a low-dimensional space; wi is the modal weight, which is set by mutual information or experience; The cloud AI platform uses a preset fault identification model to traverse and identify several sets of bimodal state features (Fi) with fault characteristics at each node, and outputs the corresponding fault type (T) and fault severity (S) to the application layer. The application layer matches the corresponding measures to the fault type (T) and fault severity (S) of each node. ; And, using the membership function, according to the node fault type: T and fault severity: S, match the corresponding node fault response measure Action and output it for display: ， is the membership function, indicating the measure The degree of adaptation to the current node failure state, arg max(·) represents the optimal action according to the fault type: T and fault severity: S.

9. An electronic device, characterized in that: The electronic device comprises: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the method according to claim 8 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program code, which can be called by a processor to execute the method according to claim 8.

Images