A three-dimensional visualization management and control platform for urban underground pipeline corridors based on digital twins

Abstract

The present invention discloses a three-dimensional visualization management and control platform for an urban underground pipe gallery based on digital twins, and relates to the technical field of visualization management and control of urban underground pipe gallery. A structural data collection module is set to collect three-dimensional modeling data, a twin model construction module is set to convert the three-dimensional modeling data of the urban underground pipe gallery into a digital twin model of the pipe gallery, an abnormal training data collection module is set to collect training sample graph data, a node fault prediction model training module is set to train a node fault prediction model for judging whether each node is faulty and a fault source identification model for judging whether a node is a fault source, a fault point visualization module is set to collect real-time graph samples, the fault source probability of each node is judged, and based on the fault source probability of each node, a visualization display is performed in the digital twin model of the pipe gallery; the accuracy of judging whether each maintenance point and pipeline end point in the urban underground pipe gallery is a fault source is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of visualized management and control of urban underground pipe corridors, and in particular to a three-dimensional visualized management and control platform for urban underground pipe corridors based on digital twins. Background Art

[0002] Digital twin technology is a new type of information technology that has developed rapidly in recent years. It creates virtual copies of physical entities to simulate, analyze and control their corresponding physical entities in a virtual environment. Urban underground pipeline corridors are an important part of urban infrastructure, including various pipelines such as electricity, communications, water supply and drainage, and gas. With the accelerated development of urbanization, the scale and complexity of underground pipeline corridors are increasing, which brings great challenges to the fault monitoring of various pipelines in underground pipeline corridors;

[0003] Traditional pipeline fault monitoring is mainly carried out through manual regular maintenance or data analysis technology. However, manual regular maintenance is obviously difficult to detect pipeline faults in time, which is easy to cause risks such as pipeline leakage. General data analysis is often carried out on a single node in the underground pipeline corridor, but fails to take into account the fluidity of liquids or gases in the pipeline. It may be that one fault source causes failures at nodes in other locations, making it difficult to accurately locate the fault source.

[0004] To this end, the present invention proposes a three-dimensional visualization management and control platform for urban underground pipeline corridors based on digital twins. Summary of the invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention proposes a three-dimensional visualization management and control platform for urban underground pipe corridors based on digital twins, which improves the accuracy of judging whether each maintenance point and pipeline end point in the urban underground pipe corridor is a fault source.

[0006] To achieve the above purpose, a three-dimensional visualization management and control platform for urban underground pipe corridors based on digital twins is proposed, which includes a structural data collection module, a twin model construction module, an abnormal training data collection module, a fault prediction model training module and a fault point visualization module; wherein each module is electrically connected;

[0007] Structural data collection module, used to collect 3D modeling data of urban underground pipe corridors in advance;

[0008] The collecting of three-dimensional modeling data of the urban underground pipe gallery comprises the following steps:

[0009] Step 11: Collect 3D point cloud model data of the urban underground pipeline gallery by using 3D laser scanning;

[0010] Step 12: Collect the construction drawings and design documents of the urban underground pipeline corridor, and use BIM software to perform three-dimensional parametric modeling of the pipeline corridor and its ancillary components to obtain a three-dimensional BIM model;

[0011] Step 13: Collect spatial coordinate data of urban underground pipeline corridors;

[0012] Step 14: Arrange parameter sensors of preset types at preset observation positions in the urban underground pipe gallery, and receive corresponding physical parameters of the preset observation positions in real time through the parameter sensors, and the real-time data of each parameter sensor constitutes real-time observation parameter data; the three-dimensional point cloud model data, the three-dimensional BIM model, the spatial coordinate data and the real-time observation parameter data constitute three-dimensional modeling data;

[0013] The structural data collection module sends the three-dimensional modeling data of the urban underground pipeline gallery to the twin model construction module;

[0014] Twin model building module, used to convert the 3D modeling data of the urban underground pipeline corridor into a digital twin model of the pipeline corridor;

[0015] The method of converting the three-dimensional modeling data of the urban underground pipe gallery into a pipe gallery digital twin model comprises the following steps:

[0016] Step 21: Based on the control field coordinate information, the 3D point cloud model data is preliminarily registered with the 3D BIM model through rigid body transformation to obtain a preliminarily registered 3D model;

[0017] Step 22: Using a 3D modeling tool, finely fuse the initially registered 3D model to obtain a finely registered 3D model;

[0018] Step 23: Integrate the real-time observation parameter data into the precise registration three-dimensional model according to the installation position of the corresponding sensor to obtain a three-dimensional framework model;

[0019] Step 24: Use a 3D visualization engine to perform visualization rendering on the 3D framework model to obtain a digital twin model of the pipeline gallery;

[0020] The twin model construction module sends the digital twin model of the pipeline gallery to the fault point visualization module;

[0021] The abnormal training data collection module is used to collect training sample graph data in advance in the city's underground pipeline corridor;

[0022] The step of collecting training sample graph data in advance in the urban underground pipe gallery comprises the following steps:

[0023] Step 31: Select several urban underground pipe corridors as test pipe corridors, and construct the pipe corridor topology relationship inside each test pipe corridor based on the construction drawings of the test pipe corridors;

[0024] The method of constructing the pipe gallery topology relationship inside each test pipe gallery is as follows:

[0025] Retrieve and check all pipelines and inspection ports from construction drawings;

[0026] Create a node set V, where the node set V contains all pipeline endpoint locations and inspection port locations;

[0027] Create an undirected edge set E, where the undirected edge set E contains all pipelines, and each pipeline corresponds to a node in the node set V;

[0028] Create an undirected graph G, G = (V, E), where the undirected graph G is the topological relationship of the pipe gallery;

[0029] Step 32: traverse each parameter sensor in the test corridor to obtain the position, and map it to the nearest node of the undirected graph G based on the principle of shortest distance;

[0030] Step 33: collect data from various parameter sensors in each test corridor in real time, and obtain the real-time value of each parameter type for each node in the corresponding undirected graph G;

[0031] Step 34: For each undirected graph G, a collection cycle duration is pre-set, and every collection cycle duration, the parameter time series of each parameter of each node within the collection cycle, the fault label of each node, and the fault source label of each node are collected; then the undirected graph G and the parameter time series of each parameter of each node within the collection cycle constitute a group of graph training samples, and the fault label and fault source label of each node in the undirected graph within the collection cycle constitute fault label data; all graph training samples and fault label data constitute training sample graph data; the fault label is one of 0 or 1, when the fault label is 0, it indicates that the node has not failed, and when the fault label is 1, it indicates that the node has failed; the fault source label is one of 0 or 1, when the fault source label is 0, it indicates that the node is not the source of the fault, and when the fault source label is 1, it indicates that the node is the source of the fault;

[0032] The abnormal training data collection module sends the training sample graph data to the fault prediction model training module;

[0033] A fault prediction model training module is used to train a node fault prediction model for determining whether each node is faulty and a fault source identification model for determining whether a node is a fault source based on training sample graph data;

[0034] The training of a node fault prediction model for determining whether each node is faulty and a fault source identification model for determining whether a node is a fault source comprises the following steps:

[0035] Step 41: for each node in each graph training sample, extract the statistical features and frequency domain features of the parameter time series sequence of each physical parameter to form a time series feature vector of the node; the statistical features include mean, standard deviation, and kurtosis; the frequency domain features include the spectrum after FFT transformation;

[0036] Step 42: for each node in each graph training sample, extract the structural features of the node to form a structural feature vector of the node;

[0037] The method of extracting the structural features of the node is:

[0038] According to the undirected graph where the node is located, the adjacent nodes and adjacent edges of the node are sampled to obtain a multi-order neighbor node set; the attributes of each adjacent node and adjacent edge in the multi-order neighbor node set are combined into a structural feature vector; the attributes of the adjacent nodes and adjacent edges include but are not limited to physical attributes such as pipe diameter and material;

[0039] Step 43: for each node in each undirected graph G, concatenate the temporal feature vector and the structural feature vector of the node to form an input feature vector of the node;

[0040] Step 44: using the undirected graph G with input feature vectors on each node as the input of the GNN model to obtain the N-dimensional spatiotemporal feature vector output by the GNN model; N is the dimension of the preset spatiotemporal feature vector;

[0041] Step 45: Based on the time series feature vector and fault label of each node, train a node fault prediction model;

[0042] The method of training the node failure prediction model is:

[0043] The time series feature vector of each node is used as the input of the node fault prediction model, and the node fault prediction model uses the predicted value of the fault probability corresponding to the time series feature vector as the output, and the value range of the fault probability is [0,1]; the fault label of the node corresponding to the time series feature vector is used as the prediction target, the difference between the predicted value of the fault probability and the fault label is used as the first prediction error, and minimizing the sum of squares of the first prediction error is used as the training target; the node fault prediction model is trained until the sum of squares of the first prediction error reaches convergence, and the training is stopped; the node fault prediction model is any one of the classification models;

[0044] Step 46: Combine the spatiotemporal feature vector and the fault label of each node into a fault source input feature vector of the node;

[0045] Step 47: training a fault source identification model based on the fault source input feature vector and fault source label of each node;

[0046] The method of training the fault source identification model is:

[0047] The fault source input feature vector of each node is used as the input of the fault source identification model, and the fault source identification model uses the predicted value of the fault source probability corresponding to the fault source input feature vector as the output, and the value range of the fault source probability is [0,1]; the fault source label of the node corresponding to the fault source input feature vector is used as the prediction target, the difference between the predicted value of the fault source probability and the fault source label is used as the second prediction error, and minimizing the sum of squares of the second prediction errors is used as the training target; the fault source identification model is trained until the sum of squares of the second prediction errors reaches convergence, and the training is stopped; the fault source identification model is any one of the classification models;

[0048] The fault prediction model training module sends the node fault prediction model and the fault source identification model to the fault point visualization module;

[0049] The fault point visualization module is used to collect real-time graph samples of each node in the underground pipe corridor under control in real time, and judge the fault source probability of each node based on the real-time graph samples, node fault prediction model and fault source identification model. Based on the fault source probability of each node, it is visualized in the digital twin model of the pipe corridor;

[0050] The method of determining the probability of the fault source of each node is as follows:

[0051] Based on the real-time graph samples, the actual time series feature vector and the actual spatiotemporal feature vector of each node are obtained;

[0052] The actual time series feature vector is input into the node fault prediction model to obtain the failure probability of each node output by the node fault prediction model;

[0053] The actual spatiotemporal feature vector of each node and the fault probability of the node are combined into an actual input feature vector, and the actual input feature vector is input into the fault source identification model to obtain the fault source probability of each node output by the fault source identification model;

[0054] The method of visualizing the digital twin model of the pipe gallery is as follows:

[0055] A fault source probability threshold is preset, and nodes with a fault source probability greater than the fault source probability threshold are regarded as suspicious nodes. In the digital twin model of the pipeline gallery, the spatial coordinates of each suspicious node are traversed and searched for, and the position of each suspicious node in the digital twin model of the pipeline gallery is highlighted to different degrees according to the size of the fault source probability.

[0056] A three-dimensional visualization control method for urban underground pipe gallery based on digital twin is proposed, which includes the following steps:

[0057] Step 1: Collect 3D modeling data of urban underground pipeline corridors in advance;

[0058] Step 2: Convert the 3D modeling data of the urban underground pipeline corridor into a digital twin model of the pipeline corridor;

[0059] Step 3: Collect training sample graph data in advance in the city's underground pipeline corridor;

[0060] Step 4: Based on the training sample graph data, a node fault prediction model for determining whether each node is faulty and a fault source identification model for determining whether the node is a fault source are trained;

[0061] Step 5: For the urban underground pipeline corridor under control, real-time graph samples of each node are collected in real time, and the fault source probability of each node is determined based on the real-time graph samples, node fault prediction model and fault source identification model. Based on the fault source probability of each node, a visual display is performed in the digital twin model of the pipeline corridor.

[0062] An electronic device is proposed, comprising: a processor and a memory, wherein the memory stores a computer program that can be called by the processor;

[0063] The processor executes the above-mentioned three-dimensional visualization management and control method of urban underground pipeline corridor based on digital twin by calling the computer program stored in the memory.

[0064] A computer-readable storage medium is provided, on which a rewritable computer program is stored;

[0065] When the computer program runs on a computer device, the computer device executes the above-mentioned three-dimensional visualization management and control method of an urban underground pipeline gallery based on digital twins.

[0066] Compared with the prior art, the present invention has the following beneficial effects:

[0067] The present invention sets a structural data collection module to pre-collect the three-dimensional modeling data of the urban underground pipeline corridor, sets a twin model construction module to convert the three-dimensional modeling data of the urban underground pipeline corridor into a digital twin model of the pipeline corridor, sets an abnormal training data collection module to pre-collect training sample graph data in the urban underground pipeline corridor, sets a fault prediction model training module to train a node fault prediction model for judging whether each node is faulty and a fault source identification model for judging whether the node is a fault source based on the training sample graph data, sets a fault point visualization module to collect real-time graph samples of each node in the urban underground pipeline corridor to be controlled in real time, and judges the fault source probability of each node based on the real-time graph samples, the node fault prediction model and the fault source identification model, and performs visualization in the digital twin model of the pipeline corridor based on the fault source probability of each node; comprehensively considers the individual characteristics of each node, and also considers the problem that the internal pipeline structure of the pipeline corridor causes the nodes to influence each other, thereby improving the accuracy of judging whether the node is a fault source. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 This is a module connection relationship diagram of a three-dimensional visualization management and control platform for an urban underground utility gallery based on digital twins in Example 1 of the present invention;

[0069] Figure 2 is an example graph of an undirected graph in Embodiment 1 of the present invention;

[0070] Figure 3 This is a flow chart of a three-dimensional visualization control method for an urban underground utility gallery based on digital twins in Example 2 of the present invention;

[0071] Figure 4 This is a schematic diagram of the structure of an electronic device in Embodiment 3 of the present invention;

[0072] Figure 5 This is a schematic diagram of the computer-readable storage medium structure in Example 4 of the present invention. DETAILED DESCRIPTION

[0073] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0074] Example 1

[0075] like Figure 1As shown, a three-dimensional visualization management and control platform for urban underground pipe corridors based on digital twins includes a structural data collection module, a twin model construction module, an abnormal training data collection module, a fault prediction model training module, and a fault point visualization module; wherein each module is electrically connected;

[0076] Structural data collection module, used to collect 3D modeling data of urban underground pipe corridors in advance;

[0077] The collecting of three-dimensional modeling data of the urban underground pipe gallery comprises the following steps:

[0078] Step 11: Collect 3D point cloud model data of the urban underground pipeline gallery by using 3D laser scanning;

[0079] Specifically, the method of collecting the three-dimensional point cloud model data of the urban underground pipeline gallery is:

[0080] A high-precision 3D laser scanner is used at each measurement and scanning station to conduct full coverage scanning inside and outside the pipe gallery to obtain original point cloud data; the high-precision standard is within ±2mm;

[0081] The elevation control between each measuring and scanning station in the pipeline corridor is measured by a multi-station precision mercury injection level, and the measured elevation is introduced into the original point cloud data of each measuring and scanning station;

[0082] Perform point cloud stitching, filtering, registration and other data processing on the original point cloud data of each measurement scanning station to obtain three-dimensional point cloud model data;

[0083] Step 12: Collect the construction drawings and design documents of the urban underground pipe gallery, and use BIM software to perform three-dimensional parametric modeling on the pipe gallery and its auxiliary components to obtain a three-dimensional BIM model; the BIM software may be Revit, etc.;

[0084] Furthermore, you can edit non-geometric information such as material properties and structural parameters of the 3D BIM model to build an information-rich BIM model, and export the BIM model through the IFC data exchange standard to prepare for subsequent fusion with point cloud data.

[0085] Step 13: Collect spatial coordinate data of urban underground pipeline corridors;

[0086] Specifically, the method of collecting the spatial coordinate data of the urban underground pipe gallery is:

[0087] Use high-precision GPS / Beidou receivers to measure the spatial coordinates of key nodes such as the entrances and exits of the pipeline corridor and inspection wells to obtain spatial measurement data; align the spatial measurement data with the 3D point cloud model data to determine the geographic spatial position of the 3D point cloud model;

[0088] Furthermore, VRS / FKP real-time differential technology can be used to achieve dynamic real-time positioning with centimeter-level accuracy;

[0089] Step 14: Arrange parameter sensors of preset types at preset observation positions in the urban underground pipe gallery, and receive corresponding physical parameters of the preset observation positions in real time through the parameter sensors, and the real-time data of each parameter sensor constitutes real-time observation parameter data; the three-dimensional point cloud model data, the three-dimensional BIM model, the spatial coordinate data and the real-time observation parameter data constitute three-dimensional modeling data;

[0090] The preset parameter sensors include but are not limited to temperature and humidity sensors, pressure sensors, vibration sensors, flow rate sensors, displacement sensors, and gas concentration sensors;

[0091] The structural data collection module sends the three-dimensional modeling data of the urban underground pipeline gallery to the twin model construction module;

[0092] Among them, the twin model construction module is mainly used to convert the three-dimensional modeling data of the urban underground pipeline corridor into a digital twin model of the pipeline corridor;

[0093] Specifically, the three-dimensional modeling data of the urban underground pipe gallery is converted into a pipe gallery digital twin model, comprising the following steps:

[0094] Step 21: Based on the control field coordinate information, the 3D point cloud model data is preliminarily registered with the 3D BIM model through rigid body transformation to obtain a preliminarily registered 3D model;

[0095] Furthermore, some optimization algorithms such as ICP and CPD can be used to further improve the accuracy of the initial registration of the 3D model;

[0096] Step 22: Using a 3D modeling tool, finely fuse the initially registered 3D model to obtain a finely registered 3D model;

[0097] Specifically, the method of finely fusing the initially registered 3D model is as follows:

[0098] Extract structural feature lines from point clouds and BIM models using constraint-based modeling methods;

[0099] Through feature line matching and topological constraints, the fine alignment of the two is achieved; it should be noted that the constraint-based modeling method and feature line matching and topological constraints are conventional functions in three-dimensional modeling tools (such as CAD, etc.), and the present invention will not be repeated here;

[0100] Step 23: Integrate the real-time observation parameter data into the precise registration three-dimensional model according to the installation position of the corresponding sensor to obtain a three-dimensional framework model;

[0101] Step 24: Use a three-dimensional visualization engine to perform visualization rendering on the three-dimensional framework model to obtain a digital twin model of the pipeline gallery; the three-dimensional visualization engine includes but is not limited to WebGL, Unity3D, Unreal Engine, etc.;

[0102] Specifically, the method of using the three-dimensional visualization engine for visualization rendering includes but is not limited to:

[0103] Different types of facilities such as equipment, inspection ports, connection points, pipelines, sensors, etc. in the 3D framework model are marked with different colors;

[0104] Display the corresponding values ​​of the physical parameters collected in real time on each sensor in the three-dimensional framework model;

[0105] The twin model construction module sends the digital twin model of the pipeline gallery to the fault point visualization module;

[0106] The abnormal training data collection module is mainly used to collect training sample graph data in advance in the urban underground pipeline gallery;

[0107] The step of collecting training sample graph data in advance in the urban underground pipe gallery includes the following steps:

[0108] Step 31: Select several urban underground pipe corridors as test pipe corridors, and construct the pipe corridor topology relationship inside each test pipe corridor based on the construction drawings of the test pipe corridors;

[0109] Specifically, the method of constructing the pipe gallery topology relationship inside each test pipe gallery is as follows:

[0110] Retrieve and check all pipelines and inspection ports from construction drawings;

[0111] Create a node set V, where the node set V contains all pipeline endpoint locations and inspection port locations;

[0112] Create an undirected edge set E, where the undirected edge set E contains all pipelines, and each pipeline corresponds to a node in the node set V;

[0113] Create an undirected graph G, G = (V, E), the undirected graph G is the topological relationship of the pipe gallery; Figure 2 An example of an undirected graph is shown, in which each node represents a pipeline or an inspection port, and the two nodes connected by each edge represent the existence of a pipeline connection between the corresponding two nodes;

[0114] Step 32: traverse each parameter sensor in the test corridor to obtain the position, and map it to the nearest node of the undirected graph G based on the principle of shortest distance;

[0115] Step 33: collect data from various parameter sensors in each test corridor in real time, and obtain the real-time value of each parameter type for each node in the corresponding undirected graph G;

[0116] Step 34: For each undirected graph G, the acquisition cycle duration is pre-set, and every acquisition cycle duration, the parameter time series of each parameter of each node in the acquisition cycle, the fault label of each node, and the fault source label of each node are collected; then the undirected graph G and the parameter time series of each parameter of each node in the acquisition cycle constitute a group of graph training samples, and the fault label and fault source label of each node in the undirected graph in the acquisition cycle constitute fault label data; all graph training samples and fault label data constitute training sample graph data; the fault label is one of 0 or 1, and when the fault label is 0, it means that the fault label is 0. The node has not failed. When the fault label is 1, it indicates that the node has failed. The fault source label is one of 0 and 1. When the fault source label is 0, it indicates that the node is not the source of the failure. When the fault source label is 1, it indicates that the node is the source of the failure. It can be understood that the fault label can be added by professionals based on actual experience to nodes with abnormal physical parameters. Since there are pipelines inside the pipe gallery, and the pipelines have fluidity, it is possible that a fault source may cause nodes in other locations to fail. Therefore, professionals also need to find the fault source from the faulty node based on actual experience.

[0117] An example of a specific graph training sample is as follows:

[0118] When there are 3 nodes in the undirected graph G, the acquisition cycle length is 3, and the physical parameters collected by each node in real time are pressure and vibration, a possible graph training sample is:

[0119] 'Node 1':{'Pressure':[98.2,97.8,98.5,],'Vibration':[0.3,0.2,0.5]},

[0120] 'Node 2':{'Pressure':[102.1,101.5,103.3],'Vibration':[1.2,1.5,0.8]},

[0121] 'Node 3':{'Pressure':[95.6,96.1,96.3],'Vibration':[0.1,0.2,0.3]};

[0122] The abnormal training data collection module sends the training sample graph data to the fault prediction model training module;

[0123] The fault prediction model training module is mainly used to train a node fault prediction model for judging whether each node is faulty and a fault source identification model for judging whether a node is a fault source based on training sample graph data;

[0124] Specifically, the training of a node fault prediction model for determining whether each node is faulty and a fault source identification model for determining whether a node is a fault source based on the training sample graph data includes the following steps:

[0125] Step 41: for each node in each graph training sample, extract the statistical features and frequency domain features of the parameter time series sequence of each physical parameter to form a time series feature vector of the node; the statistical features include mean, standard deviation, and kurtosis; the frequency domain features include the spectrum after FFT transformation;

[0126] Step 42: for each node in each graph training sample, extract the structural features of the node to form a structural feature vector of the node;

[0127] Specifically, the method of extracting the structural features of the node is:

[0128] According to the undirected graph where the node is located, the adjacent nodes and adjacent edges of the node are sampled to obtain a multi-order neighbor node set; the attributes of each adjacent node and adjacent edge in the multi-order neighbor node set are combined into a structural feature vector; the attributes of the adjacent nodes and adjacent edges include but are not limited to physical attributes such as pipe diameter and material;

[0129] Preferably, the sampling is to pre-set the number of samples (for example, 2), and then select the first-order adjacent nodes and the second-order adjacent nodes of the node as the selected adjacent nodes, and the edges corresponding to the first-order adjacent nodes and the second-order adjacent nodes are used as adjacent points;

[0130] Step 43: for each node in each undirected graph G, concatenate the temporal feature vector and the structural feature vector of the node to form an input feature vector of the node;

[0131] Step 44: using the undirected graph G with input feature vectors on each node as the input of the GNN model to obtain the N-dimensional spatiotemporal feature vector output by the GNN model; N is the dimension of the preset spatiotemporal feature vector;

[0132] Preferably, the method of obtaining the N-dimensional spatiotemporal feature vector output by the GNN model may be:

[0133] The GNN model selects a GraphSAGE model, and the GraphSAGE model includes two layers of GraphSAGEConv;

[0134] Among them, the first layer: the input dimension is the input feature vector, and the output dimension is set to 256;

[0135] Among them, the second layer: the input dimension is 256, and the output dimension is set to 128;

[0136] Through forward propagation, GraphSAGEConv learns the representation of the spatiotemporal feature vector of each node;

[0137] An example of a typical spatiotemporal feature vector is as follows:

[0138] node1:[0.21,0.58,0.12,...]#128-dimensional spatiotemporal feature vector

[0139] node2:[-0.35,0.27,-0.04,...]

[0140] node3:[0.15,-0.41,0.38,...]

[0141] Step 45: Based on the time series feature vector and fault label of each node, train a node fault prediction model;

[0142] Specifically, the method of training the node fault prediction model based on the time series feature vector and fault label of each node is as follows:

[0143] The time series feature vector of each node is used as the input of the node fault prediction model, and the node fault prediction model uses the predicted value of the fault probability corresponding to the time series feature vector as the output, and the value range of the fault probability is [0,1]; the fault label of the node corresponding to the time series feature vector is used as the prediction target, the difference between the predicted value of the fault probability and the fault label is used as the first prediction error, and minimizing the sum of squares of the first prediction error is used as the training target; the node fault prediction model is trained until the sum of squares of the first prediction error reaches convergence and the training is stopped; the node fault prediction model is any one of the classification models, and the classification model includes but is not limited to the DNN model, the SVM model, etc.; the sum of squares of the first prediction error is the mean square error or the cross entropy error; it can be understood that the node fault prediction model is a model that analyzes whether a node is faulty by learning the time series characteristics of each physical parameter of each node;

[0144] Step 46: Combine the spatiotemporal feature vector and the fault label of each node into a fault source input feature vector of the node;

[0145] Step 47: training a fault source identification model based on the fault source input feature vector and fault source label of each node;

[0146] Specifically, the method of training the fault source identification model based on the fault source input feature vector and fault source label of each node is:

[0147] The fault source input feature vector of each node is used as the input of the fault source identification model. The fault source identification model uses the predicted value of the fault source probability corresponding to the fault source input feature vector as the output, and the value range of the fault source probability is [0,1]; the fault source label of the node corresponding to the fault source input feature vector is used as the prediction target, the difference between the predicted value of the fault source probability and the fault source label is used as the second prediction error, and minimizing the sum of squares of the second prediction error is used as the training target; the fault source identification model is trained until the sum of squares of the second prediction error reaches convergence and stops training; the fault source identification model is any one of the classification models; the sum of squares of the second prediction error is the mean square error or the cross entropy error; it can be understood that the input data of the fault source identification model includes two parts of data of each node, wherein the spatiotemporal feature vector includes the spatial features of the node inside the pipe gallery, that is, it includes the structural features inside the pipe gallery, and the fault label expresses the individual features of each node, so the individual features of each node are comprehensively considered, and the problem that the internal pipeline structure of the pipe gallery causes the nodes to affect each other is also considered, thereby improving the accuracy of judging whether the node is a fault source;

[0148] The fault prediction model training module sends the node fault prediction model and the fault source identification model to the fault point visualization module;

[0149] Among them, the fault point visualization module is mainly used for the underground pipe gallery under control in the city, to collect real-time graph samples of each node in real time, and to judge the fault source probability of each node based on the real-time graph samples, node fault prediction model and fault source identification model, and to visualize the fault source probability of each node in the pipe gallery digital twin model;

[0150] It can be understood that the method of collecting graph training samples of each node in real time is consistent with the above-mentioned method of collecting graph training samples of each node in each test pipe gallery, and will not be repeated here;

[0151] The method of judging the fault source probability of each node based on the real-time graph sample, the node fault prediction model and the fault source identification model is as follows:

[0152] Based on the real-time graph sample, the actual time series feature vector and the actual time-space feature vector of each node are obtained; it can be understood that the actual time series feature vector and the actual time-space feature vector are obtained in the same manner as the time series feature vector and the time-space feature vector of the node in the test corridor;

[0153] The actual time series feature vector is input into the node fault prediction model to obtain the failure probability of each node output by the node fault prediction model;

[0154] The actual spatiotemporal feature vector of each node and the fault probability of the node are combined into an actual input feature vector, and the actual input feature vector is input into the fault source identification model to obtain the fault source probability of each node output by the fault source identification model;

[0155] The method of visualizing the fault source probability of each node in the digital twin model of the pipe gallery is as follows:

[0156] A fault source probability threshold is preset, and nodes whose fault source probability is greater than the fault source probability threshold are regarded as suspicious nodes. In the digital twin model of the pipeline corridor, the spatial coordinates of each suspicious node are traversed and searched out, and the position of each suspicious node in the digital twin model of the pipeline corridor is highlighted to different degrees according to the size of the fault source probability. Specifically, the method of highlighting to different degrees may be that the greater the fault source probability, the greater the degree of highlighting.

[0157] Example 2

[0158] like Figure 3 As shown, a three-dimensional visualization control method for urban underground pipe gallery based on digital twins includes the following steps:

[0159] Step 1: Collect 3D modeling data of urban underground pipeline corridors in advance;

[0160] Step 2: Convert the 3D modeling data of the urban underground pipeline corridor into a digital twin model of the pipeline corridor;

[0161] Step 3: Collect training sample graph data in advance in the city's underground pipeline corridor;

[0162] Step 4: Based on the training sample graph data, a node fault prediction model for determining whether each node is faulty and a fault source identification model for determining whether the node is a fault source are trained;

[0163] Step 5: For the urban underground pipeline corridor under control, real-time graph samples of each node are collected in real time, and the fault source probability of each node is determined based on the real-time graph samples, node fault prediction model and fault source identification model. Based on the fault source probability of each node, a visual display is performed in the digital twin model of the pipeline corridor.

[0164] Example 3

[0165] Figure 4 1 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Figure 4As shown, according to another aspect of the present application, an electronic device 100 is also provided. The electronic device 100 may include one or more processors and one or more memories. The memory stores a computer readable code, and when the computer readable code is executed by one or more processors, the method for three-dimensional visualization management and control of urban underground pipe corridors based on digital twins as described above can be executed.

[0166] The method or device according to the embodiment of the present application can also be used by Figure 4 The electronic device architecture shown in FIG. Figure 4 As shown, the electronic device 100 may include a bus 101, one or more CPUs 102, ROM 103, RAM 104, a communication port 105 connected to a network, an input / output component 106, a hard disk 107, etc. The storage device in the electronic device 100, such as ROM 103 or hard disk 107, may store a three-dimensional visualization management and control method of an urban underground pipe gallery based on digital twins provided in the present application. A three-dimensional visualization management and control method of an urban underground pipe gallery based on digital twins may, for example, include the following steps:

[0167] Furthermore, the electronic device 100 may further include a user interface 108. Figure 4 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 4 One or more components of an electronic device are shown.

[0168] Example 4

[0169] Figure 5 Schematic diagram of a computer-readable storage medium structure provided by an embodiment of the present application. Figure 5 As shown, it is a computer-readable storage medium 200 according to an embodiment of the present application. Computer-readable instructions are stored on the computer-readable storage medium 200. When the computer-readable instructions are executed by the processor, a three-dimensional visualization control method of an urban underground pipe gallery based on digital twins according to an embodiment of the present application described with reference to the above figures can be executed. The computer-readable storage medium 200 includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0170] In addition, according to the implementation of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, the non-transitory machine-readable storage medium stores machine-readable instructions, the machine-readable instructions can be run by a processor to execute instructions corresponding to the method steps provided by the present application, and when the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0171] The methods, apparatuses, and devices of the present application may be implemented in many ways. For example, the methods, apparatuses, and devices of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the order specifically described above unless otherwise specified. In addition, in some embodiments, the present application may also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the method according to the present application. Therefore, the present application also covers recording media storing programs for executing the method according to the present application.

[0172] In addition, the parts of the above-mentioned technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0173] The specific implementation modes as described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0174] The above preset parameters or preset thresholds are all set by technicians in this field according to actual conditions or obtained through large-scale data simulation.

[0175] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A three-dimensional visualization management and control platform for urban underground pipe galleries based on digital twins, characterized in that: It includes a structural data collection module, a twin model construction module, an abnormal training data collection module, a fault prediction model training module and a fault point visualization module; wherein each module is electrically connected; A structural data collection module is used to collect the three-dimensional modeling data of the urban underground pipeline gallery in advance, and send the three-dimensional modeling data of the urban underground pipeline gallery to the twin model construction module; The twin model construction module is used to convert the three-dimensional modeling data of the urban underground pipeline corridor into a digital twin model of the pipeline corridor, and send the digital twin model of the pipeline corridor to the fault point visualization module; The abnormal training data collection module is used to collect training sample graph data in the urban underground pipeline corridor in advance, and send the training sample graph data to the fault prediction model training module; The fault prediction model training module is used to train a node fault prediction model for judging whether each node is faulty and a fault source identification model for judging whether a node is a fault source based on the training sample graph data, and send the node fault prediction model and the fault source identification model to the fault point visualization module; The fault point visualization module is used to collect real-time graph samples of each node in the controlled urban underground pipeline corridor, and judge the fault source probability of each node based on the real-time graph samples, node fault prediction model and fault source identification model. Based on the fault source probability of each node, it is visualized in the pipeline corridor digital twin model.

2. According to claim 1, a three-dimensional visualization management and control platform for urban underground pipe galleries based on digital twins is characterized in that: The collecting of three-dimensional modeling data of the urban underground pipe gallery comprises the following steps: Step 11: Collect 3D point cloud model data of the urban underground pipeline gallery by using 3D laser scanning; Step 12: Collect the construction drawings and design documents of the urban underground pipeline corridor, and use BIM software to perform three-dimensional parametric modeling of the pipeline corridor and its ancillary components to obtain a three-dimensional BIM model; Step 13: Collect spatial coordinate data of urban underground pipeline corridors; Step 14: Parameter sensors of preset types are deployed at preset observation positions in the urban underground pipe gallery, and the corresponding physical parameters of the preset observation positions are received in real time through the parameter sensors. The real-time data of each parameter sensor constitutes the real-time observation parameter data; the three-dimensional point cloud model data, the three-dimensional BIM model, the spatial coordinate data and the real-time observation parameter data constitute the three-dimensional modeling data.

3. According to claim 2, a three-dimensional visualization management and control platform for urban underground utility corridors based on digital twins is characterized in that: The method of converting the three-dimensional modeling data of the urban underground pipe gallery into a pipe gallery digital twin model comprises the following steps: Step 21: Based on the control field coordinate information, the 3D point cloud model data is preliminarily registered with the 3D BIM model through rigid body transformation to obtain a preliminarily registered 3D model; Step 22: Using a 3D modeling tool, finely fuse the initially registered 3D model to obtain a finely registered 3D model; Step 23: Integrate the real-time observation parameter data into the precise registration three-dimensional model according to the installation position of the corresponding sensor to obtain a three-dimensional framework model; Step 24: Use a 3D visualization engine to perform visualization rendering on the 3D framework model to obtain a digital twin model of the pipeline corridor.

4. According to claim 3, a three-dimensional visualization management and control platform for urban underground pipe gallery based on digital twins is characterized in that: The step of collecting training sample graph data in advance in the urban underground pipe gallery comprises the following steps: Step 31: Select several urban underground pipe corridors as test pipe corridors, and construct the pipe corridor topology relationship inside each test pipe corridor based on the construction drawings of the test pipe corridors; The method of constructing the pipe gallery topology relationship inside each test pipe gallery is as follows: Retrieve and check all pipelines and inspection ports from construction drawings; Create a node set V, where the node set V contains all pipeline endpoint locations and inspection port locations; Create an undirected edge set E, where the undirected edge set E contains all pipelines, and each pipeline corresponds to a node in the node set V; Create an undirected graph G, G = (V, E), where the undirected graph G is the topological relationship of the pipe gallery; Step 32: traverse each parameter sensor in the test corridor to obtain the position, and map it to the nearest node of the undirected graph G based on the principle of shortest distance; Step 33: collect data from various parameter sensors in each test corridor in real time, and obtain the real-time value of each parameter type for each node in the corresponding undirected graph G; Step 34: For each undirected graph G, a collection cycle duration is pre-set, and every collection cycle duration, the parameter time series of various parameters of each node within the collection cycle, the fault label of each node, and the fault source label of each node are collected; then the undirected graph G and the parameter time series of various parameters of each node within the collection cycle constitute a group of graph training samples, and the fault label and fault source label of each node in the undirected graph within the collection cycle constitute fault label data; all graph training samples and fault label data constitute training sample graph data; the fault label is one of 0 or 1, when the fault label is 0, it indicates that the node has not failed, and when the fault label is 1, it indicates that the node has failed; the fault source label is one of 0 or 1, when the fault source label is 0, it indicates that the node is not the source of the fault, and when the fault source label is 1, it indicates that the node is the source of the fault.

5. According to claim 4, a three-dimensional visualization management and control platform for urban underground utility corridors based on digital twins is characterized in that: The training of a node fault prediction model for determining whether each node is faulty and a fault source identification model for determining whether a node is a fault source comprises the following steps: Step 41: for each node in each graph training sample, extract the statistical features and frequency domain features of the parameter time series sequence of each physical parameter to form a time series feature vector of the node; the statistical features include mean, standard deviation, and kurtosis; the frequency domain features include the spectrum after FFT transformation; Step 42: for each node in each graph training sample, extract the structural features of the node to form a structural feature vector of the node; The method of extracting the structural features of the node is: According to the undirected graph where the node is located, the adjacent nodes and adjacent edges of the node are sampled to obtain a multi-order neighbor node set; the attributes of each adjacent node and adjacent edge in the multi-order neighbor node set are combined into a structural feature vector; the attributes of the adjacent nodes and adjacent edges include pipe diameter and material; Step 43: for each node in each undirected graph G, concatenate the temporal feature vector and the structural feature vector of the node to form an input feature vector of the node; Step 44: using the undirected graph G with input feature vectors on each node as the input of the GNN model to obtain the N-dimensional spatiotemporal feature vector output by the GNN model; N is the dimension of the preset spatiotemporal feature vector; Step 45: Based on the time series feature vector and fault label of each node, train a node fault prediction model; Step 46: combining the spatiotemporal feature vector and the fault label of each node into a fault source input feature vector of the node; Step 47: Based on the fault source input feature vector and fault source label of each node, a fault source identification model is trained.

6. According to claim 5, a three-dimensional visualization management and control platform for urban underground utility corridors based on digital twins is characterized in that: The method of training the node failure prediction model is: The time series feature vector of each node is used as the input of the node fault prediction model, and the node fault prediction model uses the predicted value of the fault probability corresponding to the time series feature vector as the output, and the value range of the fault probability is [0,1]; The fault label of the node corresponding to the time series feature vector is used as the prediction target, the difference between the predicted value of the fault probability and the fault label is used as the first prediction error, and minimizing the sum of squares of the first prediction error is used as the training target; The node fault prediction model is trained until the sum of squares of the first prediction error reaches convergence and then the training is stopped; the node fault prediction model is any one of the classification models; The method of training the fault source identification model is: The fault source input feature vector of each node is used as the input of the fault source identification model. The fault source identification model uses the predicted value of the fault source probability corresponding to the fault source input feature vector as the output. The value range of the fault source probability is [0, 1]; The fault source label of the node corresponding to the fault source input feature vector is used as the prediction target, the difference between the predicted value of the fault source probability and the fault source label is used as the second prediction error, and minimizing the sum of squares of the second prediction error is used as the training target; The fault source identification model is trained until the sum of squares of the second prediction error reaches convergence, and the training is stopped; the fault source identification model is any one of the classification models.

7. A three-dimensional visualization management and control platform for urban underground utility corridors based on digital twins according to claim 6, characterized in that: The method of determining the probability of the fault source of each node is as follows: Based on the real-time graph samples, the actual time series feature vector and the actual spatiotemporal feature vector of each node are obtained; The actual time series feature vector is input into the node fault prediction model to obtain the failure probability of each node output by the node fault prediction model; The actual spatiotemporal feature vector of each node and the fault probability of the node are combined into an actual input feature vector, and the actual input feature vector is input into the fault source identification model to obtain the fault source probability of each node output by the fault source identification model; The method of visualizing the digital twin model of the pipe gallery is as follows: A fault source probability threshold is preset, and nodes with a fault source probability greater than the fault source probability threshold are regarded as suspicious nodes. In the digital twin model of the pipeline gallery, the spatial coordinates of each suspicious node are traversed and searched for, and the position of each suspicious node in the digital twin model of the pipeline gallery is highlighted to different degrees according to the size of the fault source probability.

8. A three-dimensional visualization management and control method for an urban underground utility gallery based on digital twins, which is implemented based on the three-dimensional visualization management and control platform for an urban underground utility gallery based on digital twins as described in any one of claims 1 to 7, characterized in that: The following steps are involved: Step 1: Collect 3D modeling data of urban underground pipeline corridors in advance; Step 2: Convert the 3D modeling data of the urban underground pipeline corridor into a digital twin model of the pipeline corridor; Step 3: Collect training sample graph data in advance in the city's underground pipeline corridor; Step 4: Based on the training sample graph data, a node fault prediction model for determining whether each node is faulty and a fault source identification model for determining whether the node is a fault source are trained; Step 5: For the urban underground pipeline corridor under control, real-time graph samples of each node are collected in real time, and the fault source probability of each node is determined based on the real-time graph samples, node fault prediction model and fault source identification model. Based on the fault source probability of each node, a visual display is performed in the digital twin model of the pipeline corridor.

9. An electronic device, characterized in that: include: processor and memory, wherein The memory stores a computer program that can be called by the processor; The processor executes in the background a three-dimensional visualization management and control method for urban underground utility corridors based on digital twins as described in any one of claims 1 to 7 by calling the computer program stored in the memory.

10. A computer-readable storage medium, characterized in that: A rewritable computer program is stored thereon; When the computer program runs on a computer device, the computer device executes in the background a three-dimensional visualization management and control method for an urban underground utility corridor based on digital twins as described in any one of claims 1 to 7.

Images