Underground pipeline real-time monitoring device based on BIM technology

Abstract

The invention relates to the technical field of monitoring and early warning, and discloses an underground pipeline real-time monitoring device based on a BIM technology, and the device comprises a data collection module which is used for collecting the sensing feature data of a sub-region, and representing the sensing feature data of each time point through a first feature sequence; the data preprocessing module is used for preprocessing the first feature sequence to obtain a second feature sequence; the region graph structure construction module is used for constructing a region graph structure according to the second feature sequence; the risk score prediction module is used for optimizing the learning rate and the convolution layer number of the graph convolution network according to the optimized genetic algorithm, and constructing a risk score prediction model according to the regional graph structure; and the early warning measure generation module is used for generating early warning measures according to the risk scores of the M sub-regions in the second preset time period G. According to the invention, the accuracy of pipeline risk prediction is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of monitoring and early warning, and in particular relates to a real-time monitoring device for underground pipelines based on BIM technology. Background Art

[0002] Underground pipelines are an important part of modern urban infrastructure, covering water, gas, electricity, communications and other fields. With the acceleration of urbanization, the number and complexity of underground pipelines are increasing. How to effectively monitor and manage these pipelines has become an important issue in urban management. The monitoring of underground pipelines is not only related to public safety, but also to the sustainable development of cities. Once a pipeline leaks or fails, it may cause serious environmental pollution, property loss, and even endanger people's lives. Therefore, establishing an efficient and reliable underground pipeline monitoring system is an urgent problem to be solved in today's urban management.

[0003] Current underground pipeline monitoring methods mainly include traditional manual inspections, ground detection, and remote monitoring. Although manual inspections can provide certain information, they rely on manual time and energy, making it difficult to be comprehensive and real-time, and easily leading to monitoring omissions. Although ground detection technology can detect some pipelines, it usually requires excavation and exposure of pipelines, which increases construction costs and time, and causes interference to the surrounding environment. Although remote monitoring technology improves the convenience of monitoring, its monitoring accuracy is often limited by equipment and environment, and it is difficult to provide real-time feedback on the true status of the pipeline.

[0004] Therefore, there is an urgent need for a real-time monitoring device for underground pipelines to solve the above problems. Summary of the invention

[0005] The present invention provides a real-time monitoring device for underground pipelines based on BIM technology, which solves the technical problems in related technologies of poor timeliness, high cost, low precision, and difficulty in comprehensively, real-timely and accurately reflecting the status of pipelines.

[0006] The present invention provides a real-time monitoring device for underground pipelines based on BIM technology, comprising:

[0007] A data acquisition module, which is used to collect sensor feature data of M sub-areas within a first preset time period T according to a preset time interval t, and represent the sensor feature data at each time point by a first feature sequence;

[0008] The mth sequence unit of the first feature sequence at the nth time point represents the sensing feature data of the mth sub-region at the nth time point, where 1≤m≤M, 1≤n≤N, and N=T / t;

[0009] Sensor characteristic data include: pressure, temperature, flow and vibration;

[0010] A data preprocessing module, which is used to preprocess the first feature sequence to obtain a second feature sequence;

[0011] A regional graph structure construction module, which is used to construct a regional graph structure according to the second feature sequence, the regional graph structure including: nodes, initial features of nodes and edges;

[0012] Establish a mapping relationship between sub-areas and nodes;

[0013] The sensor feature data of the sub-region is used as the initial feature of the node with which the mapping relationship is established;

[0014] The risk score prediction module is used to optimize the learning rate and number of convolutional layers of the graph convolutional network according to the optimized genetic algorithm, and to build a risk score prediction model according to the regional graph structure;

[0015] The input of the risk score prediction model is a regional graph structure, and the output is the risk scores of M sub-regions within a second preset time period G;

[0016] The risk score value ranges from 0 to 100.

[0017] The early warning measure generation module is used to generate early warning measures according to the risk scores of the M sub-areas within the second preset time period G.

[0018] Furthermore, T, t, M, m, N, n, and G are custom parameters.

[0019] Furthermore, the step of preprocessing the first feature data includes:

[0020] Step S201, using Kalman filtering to remove noise signals in the first feature sequence;

[0021] Step S202, detecting outliers in the first feature sequence by using the 3σ criterion, and replacing the outliers with the median;

[0022] Step S203, convert the time domain waveform of the vibration into a frequency domain waveform through fast Fourier transform. In the frequency domain waveform, the horizontal axis represents the frequency and the vertical axis represents the amplitude. The frequency component with the highest amplitude is extracted from the frequency domain waveform as the main frequency. The main frequency energy is obtained according to the main frequency, and the main frequency energy feature is created. It is spliced ​​after the sensor data. The formula of the main frequency energy is: E mf =|A(f mf )| 2 , where A(f mf ) represents the main frequency f mf The corresponding amplitude;

[0023] Step S204, normalizing the first feature sequence processed in the above steps using a z-score normalization method to obtain a second feature sequence.

[0024] Furthermore, the ways to construct edges include:

[0025] Build edges between adjacent sub-regions;

[0026] Build edges based on the consistency of pipeline types;

[0027] The cosine similarity value of the preprocessed sensor feature data of the two sub-regions is calculated, and when the cosine similarity value is greater than a preset threshold, an edge is constructed between the two sub-regions.

[0028] Furthermore, the steps of optimizing the learning rate and the number of convolutional layers of the graph convolutional network according to the improved genetic algorithm include:

[0029] Step S301, constructing an improved genetic algorithm model by introducing a Cauchy distribution mutation strategy and a convergence factor;

[0030] Step S302, finding the optimal solution for the learning rate and the number of convolutional layers, and building a risk score prediction model;

[0031] Step S303: Calculate the risk scores of the M sub-regions within the second preset time period G according to the risk score prediction model.

[0032] Furthermore, the specific steps of step S301 include:

[0033] Step S401, generating an initialization population with Q individuals meeting the constraint conditions, where Q is a custom parameter;

[0034] Step S402, using the mean square error of the risk score prediction model as the objective function;

[0035] Step S403, obtaining the fitness value of each individual in the initialized population through the objective function calculation;

[0036] Step S404, selecting half of the individuals with the largest fitness values ​​in the initialization population as parents;

[0037] Step S405, generating a random number ranging from 0 to 1, and when the random number is less than the convergence factor, performing a crossover operation that satisfies the constraint condition;

[0038] The calculation formula of the convergence factor is: Where a represents the nonlinear convergence factor based on cosine decay, k represents the current number of iterations, K represents the maximum number of iterations, and α represents the parameter that controls the convergence speed;

[0039] Step S406, mutating the individuals in the population based on the Cauchy mutation strategy;

[0040] Step S407, if it is determined that the current number of iterations is greater than or equal to the maximum number of iterations, the individual with the largest fitness value in the initialized population is taken as the optimal solution for the learning rate and the number of convolutional layers, otherwise, return to step S403.

[0041] Furthermore, the specific steps of step S406 include:

[0042] Step S4061, determining the location parameter and scale parameter of the Cauchy distribution;

[0043] Step S4062, determining the variation ratio;

[0044] Step S4063, generating mutant individuals, for the selected individuals, randomly extracting mutation values ​​from the Cauchy distribution and adding them to each dimension of the individual, each dimension of the individual is represented by each element value of the vector.

[0045] Further, the risk score prediction model includes a first hidden layer, a combination layer, a second hidden layer, a segmentation layer and a classifier;

[0046] The first hidden layer includes N first units, the nth first unit inputs the regional graph structure and outputs the first updated features of M nodes;

[0047] The combination layer includes N combination units, and the nth combination unit is used to combine the first updated features output by the nth first unit into a combination vector;

[0048] The second hidden layer includes N second units, the nth second unit inputs the combination vector output by the nth combination unit, and outputs the second update feature;

[0049] The segmentation layer is used to segment the second update feature output by the Nth second unit into M third update features, where each third update feature corresponds to a node update feature of a node;

[0050] The node update features of the M nodes are input into the classifier, and the classification space of the classifier represents the risk scores of the M sub-regions within the second preset time period G.

[0051] Furthermore, the first unit of the first hidden layer is constructed based on a graph convolutional network.

[0052] Furthermore, the second unit of the second hidden layer is constructed based on a gated recurrent unit.

[0053] The beneficial effects of the present invention are as follows: the present invention optimizes the learning rate and the number of convolution layers of the risk score prediction model through an improved genetic algorithm to achieve accurate risk score prediction, wherein the genetic algorithm introduces the Cauchy mutation strategy and the convergence factor to make the algorithm more exploratory and convergent;

[0054] The risk score prediction model of the present invention first extracts the spatial features of each time point through a graph convolutional network, focusing on the relationship between nodes in different regions and the local information of the pipeline structure, and then uses a gated recurrent unit to process the feature sequence of the time dimension, which helps to capture the dynamic changes of data in time, so that the model can generate more explanatory feature representations at different spatiotemporal scales, thereby improving the accuracy of pipeline risk prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a module schematic diagram of a real-time monitoring device for underground pipelines based on BIM technology of the present invention;

[0056] Figure 2 It is a flow chart of preprocessing the first feature sequence of the present invention.

[0057] In the figure: a data collection module 101, a data preprocessing module 102, a regional map structure building module 103, a risk score prediction module 104 and an early warning measure generation module 105. DETAILED DESCRIPTION

[0058] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0059] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0060] like Figure 1-Figure 2 As shown, a real-time monitoring device for underground pipelines based on BIM technology includes:

[0061] A data collection module 101, which is used to collect sensor feature data of M sub-areas within a first preset time period T according to a preset time interval t, and represent the sensor feature data at each time point by a first feature sequence;

[0062] The mth sequence unit of the first feature sequence at the nth time point represents the sensing feature data of the mth sub-region at the nth time point, where 1≤m≤M, 1≤n≤N, and N=T / t;

[0063] Sensor characteristic data include: pressure, temperature, flow and vibration;

[0064] A data preprocessing module 102, which is used to preprocess the first feature sequence to obtain a second feature sequence;

[0065] A region graph structure construction module 103, which is used to construct a region graph structure according to the second feature sequence, the region graph structure including: nodes, initial features of nodes and edges;

[0066] Establish a mapping relationship between sub-areas and nodes;

[0067] The sensor feature data of the sub-region is used as the initial feature of the node with which the mapping relationship is established;

[0068] The risk score prediction module 104 is used to optimize the learning rate and the number of convolutional layers of the graph convolutional network according to the optimized genetic algorithm, and to construct a risk score prediction model according to the regional graph structure;

[0069] The input of the risk score prediction model is a regional graph structure, and the output is the risk scores of M sub-regions within a second preset time period G;

[0070] The risk score value ranges from 0 to 100.

[0071] The early warning measure generation module 105 is used to generate early warning measures according to the risk scores of the M sub-areas within the second preset time period G.

[0072] In one embodiment of the present invention, T, t, M, m, N, n and G are custom parameters. Preferably, T is set to 5 minutes and t is set to 10 seconds.

[0073] In one embodiment of the present invention, pressure sensors are installed at the pipeline inlet, outlet and key turning points to monitor pressure fluctuations inside the pipeline, temperature sensors are installed inside the pipeline to monitor temperature changes, flow sensors are installed at the pipeline inlet and outlet to monitor changes in fluid flow, and vibration sensors are installed at pipeline support positions to monitor vibration frequency and amplitude; and wireless sensor network technology is used to transmit the data collected by the above sensors to a data center in real time to achieve real-time data management.

[0074] In one embodiment of the present invention, the step of preprocessing the first feature data includes:

[0075] Step S201, using Kalman filtering to remove noise signals in the first feature sequence;

[0076] Step S202, detecting outliers in the first feature sequence by the 3σ criterion, and replacing the outliers with the median. Specifically, calculating the mean and standard deviation of the first feature sequence, determining the normal data range as: μ±3σ, detecting outliers outside the range, and replacing the detected outliers with the median to ensure the stability and accuracy of the data, where μ is the mean and σ is the standard deviation;

[0077] Step S203, convert the time domain waveform of the vibration into a frequency domain waveform through fast Fourier transform. In the frequency domain waveform, the horizontal axis represents the frequency and the vertical axis represents the amplitude. The frequency component with the highest amplitude is extracted from the frequency domain waveform as the main frequency. The main frequency energy is obtained according to the main frequency, and the main frequency energy feature is created. It is spliced ​​after the sensor data. The formula of the main frequency energy is: E mf =|A(f mf )| 2 , where A(f mf ) represents the main frequency f mf The corresponding amplitude;

[0078] Step S204, normalizing the first feature sequence processed in the above steps using a z-score normalization method to obtain a second feature sequence.

[0079] In one embodiment of the present invention, the method of constructing an edge includes:

[0080] Build edges between adjacent sub-regions;

[0081] Edges are built based on the consistency of pipeline types; if pipelines in two sub-areas are of the same type (such as water or gas pipelines), edges can be added to reflect their functional relationship.

[0082] The cosine similarity value of the preprocessed sensor feature data of the two sub-regions is calculated. When the cosine similarity value is greater than a preset threshold, an edge is constructed between the two sub-regions, where the formula for cosine similarity is: a and b represent the preprocessed sensing feature data of the two sub-regions, respectively.

[0083] In one embodiment of the present invention, the step of optimizing the learning rate and the number of convolutional layers of the graph convolutional network according to the improved genetic algorithm includes:

[0084] Step S301, constructing an improved genetic algorithm model by introducing a Cauchy distribution mutation strategy and a convergence factor;

[0085] Step S302, finding the optimal solution for the learning rate and the number of convolutional layers, and building a risk score prediction model;

[0086] Step S303: Calculate the risk scores of the M sub-regions within the second preset time period G according to the risk score prediction model.

[0087] In one embodiment of the present invention, the specific steps of step S301 include:

[0088] Step S401, generating an initialization population with Q individuals meeting the constraint conditions, where Q is a custom parameter, preferably, Q is set to 20;

[0089] The constraints include: the learning rate is within a first preset range; the number of convolutional layers is within a second preset range;

[0090] The individuals of the initialized population are represented by vector encoding, which is represented by: X = r, n}, where r and n represent the learning rate and the number of convolutional layers of the risk score prediction model respectively;

[0091] Step S402, using the mean square error of the risk score prediction model as an objective function to evaluate the prediction effect of the risk score prediction model;

[0092] Step S403, obtaining the fitness value of each individual in the initialized population through the objective function calculation;

[0093] Step S404, selecting half of the individuals with the largest fitness values ​​in the initialization population as parents;

[0094] Step S405, generating a random number ranging from 0 to 1, and when the random number is less than the convergence factor, performing a crossover operation that satisfies the constraint condition;

[0095] The calculation formula of the convergence factor is: Wherein a represents a nonlinear convergence factor based on cosine decay, k represents the current number of iterations, K represents the maximum number of iterations, preferably, K is set to 200, and α represents a parameter for controlling the convergence speed; the convergence factor can be gradually reduced as the number of iterations increases, so that the crossover rate is reduced, and the crossover operation is concentrated on the better individuals, thereby enhancing the convergence of the algorithm;

[0096] The crossover operation means randomly selecting the element value of one position of the vector encoding of two individuals in the parent generation to exchange;

[0097] Step S406, mutating the individuals in the population based on the Cauchy mutation strategy;

[0098] The probability density function of the Cauchy distribution is: where x 0 represents the location parameter, γ represents the scale parameter, and x represents the custom variable;

[0099] Step S407, if it is determined that the current number of iterations is greater than or equal to the maximum number of iterations, the individual with the largest fitness value in the initialized population is used as the optimal solution for the learning rate and the convolutional layer, otherwise, return to step S403.

[0100] In one embodiment of the present invention, the specific steps of step S406 include:

[0101] Step S4061, determine the location parameter and scale parameter of the Cauchy distribution. Preferably, the location parameter x 0 Set to 0, the scale parameter γ is adjusted according to the current number of iterations, γ k+1 =γ k *β, where γ k+1 represents the scale parameter when the number of population iterations is k+1, γ k represents the scale parameter when the number of population iterations is k, β represents a decreasing factor between 0 and 1, preferably, the decreasing factor is set to 0.99 and the scale parameter is set to 0.7;

[0102] Step S4062, determining the variation ratio, preferably, the variation ratio is set to 10%;

[0103] Step S4063, generate mutant individuals. For the selected individuals, randomly extract mutation values ​​from the Cauchy distribution and add them to each dimension of the individual. Each dimension of the individual is represented by each element value of the vector. The specific application method is: when the mutation value is greater than the first preset threshold, add the mutation value to each element value of the individual, thereby updating each element value of the vector. Preferably, the first preset threshold is set to 0.5.

[0104] In one embodiment of the present invention, the risk score prediction model includes a first hidden layer, a combination layer, a second hidden layer, a segmentation layer and a classifier;

[0105] The first hidden layer includes N first units, the nth first unit inputs the regional graph structure and outputs the first updated features of M nodes;

[0106] The combination layer includes N combination units, and the nth combination unit is used to combine the first updated features output by the nth first unit into a combination vector;

[0107] The second hidden layer includes N second units, the nth second unit inputs the combination vector output by the nth combination unit, and outputs the second update feature;

[0108] The segmentation layer is used to segment the second update feature output by the Nth second unit into M third update features, where each third update feature corresponds to a node update feature of a node;

[0109] The node update features of the M nodes are input into the classifier, and the classification space of the classifier represents the risk scores of the M sub-regions within the second preset time period G.

[0110] In one embodiment of the present invention, the first unit of the first hidden layer is constructed based on a graph convolutional network, and the calculation formula of the first unit includes:

[0111]

[0112] in represents the first updated feature of node i, h i and h j They all represent the initial features of the nodes. represents the normalized attention weight between the node i of the sth attention head and its adjacent node j, N i represents the set of nodes adjacent to node i, ReLU represents the ReLU activation function, represents the attention coefficient between node i of the sth attention head and its adjacent node j, represents the attention coefficient between node i of the sth attention head and its adjacent node k, LeakyReLU represents the LeakyReLU activation function, exp represents the exponential function, T represents the transposition operation, and concat represents the concatenation operation. and They respectively represent the first weight parameter, the second weight parameter, the third weight parameter and the fourth weight parameter of the sth attention head.

[0113] In one embodiment of the present invention, the second unit of the second hidden layer is constructed based on a gated recurrent unit.

[0114] In one embodiment of the present invention, the risk scores of underground pipeline sub-areas are divided into three levels to correspond to different health status and early warning measures. The risk score between 70 and 100 is a low risk level, indicating that the pipeline is in a normal or slightly aged state and is in good operating condition, and routine monitoring is sufficient; the risk score between 30 and 70 is a medium risk level, indicating that the pipeline is aged, worn or slightly abnormal, and requires enhanced monitoring, and technicians are arranged to conduct on-site inspections; the risk score between 0 and 30 is a high risk level, indicating that the pipeline is at risk of serious failure and must be immediately shut down, and technicians are arranged to conduct emergency inspections and comprehensive overhauls.

[0115] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. A real-time monitoring device for underground pipelines based on BIM technology, characterized in that: include: A data acquisition module, which is used to collect sensor feature data of M sub-areas within a first preset time period T according to a preset time interval t, and represent the sensor feature data at each time point by a first feature sequence; The mth sequence unit of the first feature sequence at the nth time point represents the sensing feature data of the mth sub-region at the nth time point, where 1≤m≤M, 1≤n≤N, and N=T / t; Sensor characteristic data include: pressure, temperature, flow and vibration; A data preprocessing module, which is used to preprocess the first feature sequence to obtain a second feature sequence; A regional graph structure construction module, which is used to construct a regional graph structure according to the second feature sequence, the regional graph structure including: nodes, initial features of nodes and edges; Establish a mapping relationship between sub-areas and nodes; The sensor feature data of the sub-region is used as the initial feature of the node with which the mapping relationship is established; The risk score prediction module is used to optimize the learning rate and number of convolutional layers of the graph convolutional network according to the optimized genetic algorithm, and to build a risk score prediction model according to the regional graph structure; The input of the risk score prediction model is a regional graph structure, and the output is the risk scores of M sub-regions within a second preset time period G; The risk score value ranges from 0 to 100. The early warning measure generation module is used to generate early warning measures according to the risk scores of the M sub-areas within the second preset time period G.

2. According to claim 1, a real-time monitoring device for underground pipelines based on BIM technology is characterized in that: T, t, M, m, N, n, and G are custom parameters.

3. The underground pipeline real-time monitoring device based on BIM technology according to claim 1 is characterized in that: The step of preprocessing the first feature data includes: Step S201, using Kalman filtering to remove noise signals in the first feature sequence; Step S202, detecting outliers in the first feature sequence by using the 3σ criterion, and replacing the outliers with the median; Step S203, convert the time domain waveform of the vibration into a frequency domain waveform through fast Fourier transform. In the frequency domain waveform, the horizontal axis represents the frequency and the vertical axis represents the amplitude. The frequency component with the highest amplitude is extracted from the frequency domain waveform as the main frequency. The main frequency energy is obtained according to the main frequency, and the main frequency energy feature is created. It is spliced ​​after the sensor feature data. The formula of the main frequency energy is: E mf =|A(f mf )| 2 , where A(f mf ) represents the main frequency f mf The corresponding amplitude; Step S204, normalizing the first feature sequence processed in the above steps using a z-score normalization method to obtain a second feature sequence.

4. The underground pipeline real-time monitoring device based on BIM technology according to claim 1 is characterized in that: Ways to construct edges include: Build edges between adjacent sub-regions; Build edges based on the consistency of pipeline types; The cosine similarity value of the preprocessed sensor feature data of the two sub-regions is calculated, and when the cosine similarity value is greater than a preset threshold, an edge is constructed between the two sub-regions.

5. The underground pipeline real-time monitoring device based on BIM technology according to claim 1 is characterized in that: The steps of optimizing the learning rate and number of convolutional layers of the graph convolutional network according to the improved genetic algorithm include: Step S301, constructing an improved genetic algorithm model by introducing a Cauchy distribution mutation strategy and a convergence factor; Step S302, finding the optimal solution for the learning rate and the number of convolutional layers, and building a risk score prediction model; Step S303: Calculate the risk scores of the M sub-regions within the second preset time period G according to the risk score prediction model.

6. The underground pipeline real-time monitoring device based on BIM technology according to claim 1 is characterized in that: The specific steps of step S301 include: Step S401, generating an initialization population with Q individuals meeting the constraint conditions, where Q is a custom parameter; Step S402, using the mean square error of the risk score prediction model as the objective function; Step S403, obtaining the fitness value of each individual in the initialized population through the objective function calculation; Step S404, selecting half of the individuals with the largest fitness values ​​in the initialization population as parents; Step S405, generating a random number ranging from 0 to 1, and when the random number is less than the convergence factor, performing a crossover operation that satisfies the constraint condition; The calculation formula of the convergence factor is: Where a represents the nonlinear convergence factor based on cosine decay, k represents the current number of iterations, K represents the maximum number of iterations, and α represents the parameter that controls the convergence speed; Step S406, mutating the individuals in the population based on the Cauchy mutation strategy; Step S407, if it is determined that the current number of iterations is greater than or equal to the maximum number of iterations, the individual with the largest fitness value in the initialized population is taken as the optimal solution for the learning rate and the number of convolutional layers, otherwise, return to step S403.

7. The underground pipeline real-time monitoring device based on BIM technology according to claim 1 is characterized in that: The specific steps of step S406 include: Step S4061, determining the location parameter and scale parameter of the Cauchy distribution; Step S4062, determining the variation ratio; Step S4063, generating mutant individuals, for the selected individuals, randomly extracting mutation values ​​from the Cauchy distribution and adding them to each dimension of the individual, each dimension of the individual is represented by each element value of the vector.

8. The underground pipeline real-time monitoring device based on BIM technology according to claim 1 is characterized in that: The risk score prediction model includes a first hidden layer, a combination layer, a second hidden layer, a segmentation layer and a classifier; The first hidden layer includes N first units, the nth first unit inputs the regional graph structure and outputs the first updated features of M nodes; The combination layer includes N combination units, and the nth combination unit is used to combine the first updated features output by the nth first unit into a combination vector; The second hidden layer includes N second units, the nth second unit inputs the combination vector output by the nth combination unit, and outputs a second updated feature; The segmentation layer is used to segment the second update feature output by the Nth second unit into M third update features, where each third update feature corresponds to a node update feature of a node; The node update features of the M nodes are input into the classifier, and the classification space of the classifier represents the risk scores of the M sub-areas within the second preset time period G.

9. The underground pipeline real-time monitoring device based on BIM technology according to claim 1 is characterized in that: The first unit of the first hidden layer is built based on a graph convolutional network.

10. The underground pipeline real-time monitoring device based on BIM technology according to claim 1 is characterized in that: The second unit of the second hidden layer is built based on the gated recurrent unit.

Images