An adaptive pipeline relocation intelligent optimization system

Abstract

The present invention relates to the field of pipeline relocation technology, and in particular to an adaptive pipeline relocation intelligent optimization system, which acquires and integrates underground space data and preprocesses these data, extracts underground occupied positioning information, depth information and volume information, and evaluates geological availability. The preprocessed data is mapped to an underground three-dimensional space model, and the safety logic of pipeline layout is learned through a BP neural network, and the available relocation area is calibrated in the model. Based on the conflicting spatial data and the available relocation area, the relocation space is extracted from the model. The relocation space is cut with a preset step size to generate a two-dimensional pixel matrix, and the optimal relocation route is quickly calculated through a convolutional neural network. If there is no available route, the system increases the relocation space based on a preset gradient and recalculates. This automated optimization significantly improves the efficiency and accuracy of the relocation, ensures the safety and stability of the route, reduces downtime and repeated construction, and improves resource utilization efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of pipeline relocation, and in particular to an adaptive pipeline relocation intelligent optimization system. Background Art

[0002] In the process of urbanization, with the continuous expansion of urban scale and the advancement of infrastructure construction, the layout of underground pipelines has become increasingly complex. Underground pipelines include water supply pipelines, drainage pipelines, gas pipelines, power cables, and communication optical cables. These pipelines are the lifeblood of urban operations, and ensuring their normal operation is crucial. However, with the advancement of new buildings, roads, and other infrastructure projects, existing underground pipelines often need to be relocated to avoid conflicts and ensure the smooth progress of the project. Underground space is limited, and a variety of pipelines and facilities have been laid. Relocating new pipelines requires finding a suitable path within the limited space. During the relocation process, unknown underground cavities and geologically unstable areas may be encountered. When encountering temporary obstacles or other factors at the construction site, it is impossible to integrate the data and geological factors occupied by multiple parties underground in real time, which results in the need to stop work and spend a lot of time to conduct the entire information and re-evaluate the geology and depth to formulate a relocation plan, which consumes a lot of time and resources. Summary of the invention

[0003] In order to solve the above problems, the present invention provides an adaptive pipeline relocation intelligent optimization system.

[0004] To achieve the above object, the technical solution adopted by the present invention is:

[0005] An adaptive pipeline relocation intelligent optimization system includes: a data receiving module, a data preprocessing module, an underground space mapping module, a pipeline relocation learning module and an adaptive strategy generation module, wherein the data receiving module is connected to the data preprocessing module, and the underground space mapping module is connected to the data preprocessing module, the pipeline relocation learning module and the adaptive strategy generation module respectively;

[0006] A data receiving module is used to obtain underground space data, including underground use data, geological data and geological event data;

[0007] A data preprocessing module, used for preprocessing underground space data, wherein the preprocessing includes extracting location information, depth information and volume information of underground occupancy, and evaluating geological availability based on geological data and geological event data;

[0008] The underground space mapping module is used to build an underground 3D space model based on the positioning API and map the underground 3D space model based on the preprocessed underground space data;

[0009] The pipeline relocation learning module is used to learn the pipeline layout safety logic through underground space data based on the BP neural network, and to mark the relocation available area in the underground three-dimensional space model based on the pipeline layout safety logic;

[0010] The adaptive strategy generation module is used to perform the following steps:

[0011] S1. Obtain conflicting spatial data and synchronously update the underground three-dimensional spatial model;

[0012] S2, extracting the relocation space from the underground three-dimensional space model based on the conflicting spatial data and the relocation available area;

[0013] S3, based on the migration space, several two-dimensional pixel matrices are generated by cutting with a preset step size, and the available migration routes are calculated from the continuous two-dimensional pixel matrices through a convolutional neural network; if there is no available migration route in the migration space, the migration space is increased based on the preset gradient, and S3 is re-executed; if there is an available migration route in the migration space, the available migration route is output and synchronized in the underground three-dimensional space model.

[0014] Furthermore, the underground usage data includes:

[0015] Underground pipeline data, including the location, direction, material and usage information of water pipelines, drainage pipelines, gas pipelines, power cables and communication optical cables;

[0016] Underground facilities data, including the location, structure and usage of underground garages, underground shopping malls, subway stations and tunnels;

[0017] Underground cavity data, including the location, size and volume of underground mining areas, karst caves and old underground building ruins;

[0018] The geological data includes underground soil quality at regional depth and corresponding stability data; the geological event data includes the location and depth of collapse and subsidence events.

[0019] Furthermore, the extracting of the location information, depth information and volume information of underground occupancy and evaluating the geological availability based on the geological data and geological event data comprises the following steps:

[0020] Based on the transformer model, the location information, depth information and volume information of the occupied space are extracted from the underground usage data, and the geographic coordinates of each data point are converted into a standard coordinate system through the positioning API to record the latitude and longitude, depth and volume of the occupied space;

[0021] Weights are selected according to the type of geological events, and geological data and the number of occurrences of geological events are weighted to evaluate the geological availability of underground space.

[0022] Furthermore, mapping the underground three-dimensional space model based on the preprocessed underground space data includes the following steps:

[0023] Obtain pre-processed underground space data, including latitude and longitude, depth, volume of occupied space and geological availability of underground space;

[0024] The positioning API is used to locate the latitude, longitude, depth and volume information of the occupied space in three dimensions and map it into the underground three-dimensional space model framework, and the color value is changed according to different data categories.

[0025] Furthermore, learning pipeline layout safety logic through underground space data includes the following steps:

[0026] Obtain the layout distances between different underground pipelines, underground facilities and underground cavities;

[0027] The layout type and the stability data of the corresponding underground soil are used as input, and the layout distance is used as output to train the BP neural network and obtain the layout safety logic model.

[0028] Furthermore, the marking of the migration available area in the underground three-dimensional space model based on the pipeline layout safety logic includes the following steps:

[0029] Import the deployment safety logic model obtained through BP neural network learning;

[0030] Loading the underground three-dimensional space model;

[0031] Based on the layout safety logic model, the data in the underground three-dimensional space model is analyzed, the available areas for relocation are identified, and the available areas for relocation are calibrated using preset available color values.

[0032] Furthermore, the S1 comprises the following steps:

[0033] Acquire real-time conflict space data, including new obstacles, geological changes and emergency data found during the construction process;

[0034] The acquired conflicting spatial data is synchronously updated with the current underground three-dimensional spatial model.

[0035] Furthermore, the S2 includes extracting a preset size area containing conflicting spatial data from the underground three-dimensional space model based on the synchronously updated model, and defining the area as the relocation space.

[0036] Furthermore, the convolutional neural network is trained by the following steps:

[0037] By marking the digital migration pipeline in the migration space, the digital migration pipeline is divided into turning nodes, each divided migration pipeline is traversed in parallel in the migration available area, and the traversal results of each divided migration pipeline are combined to generate a migration route dataset;

[0038] Based on the migration route data set, the migration space is cut with a preset step size to generate several continuous two-dimensional pixel matrices;

[0039] The pixel matrix of the migration space of the uncalibrated digital migration pipeline is used as input and several continuous two-dimensional pixel matrices are used as output to train the convolutional neural network.

[0040] Furthermore, the formula of the convolutional neural network is as follows:

[0041]

[0042] Where R represents the calculated optimal available relocation route; X i，j W represents the input of the jth convolutional layer in the i-th continuous two-dimensional pixel matrix; j represents the jth convolution kernel; * represents the convolution operation; b j represents the bias of the jth convolution kernel; f represents the activation function; arg max i represents the path when i is selected to maximize the sum; n represents the number of layers of the convolutional neural network.

[0043] The beneficial effects of the present invention are as follows: the present invention obtains and integrates underground space data, including underground use data, geological data and geological event data, through a data receiving module. The underground space data is preprocessed, including extracting the location information, depth information and volume information of underground occupancy, and evaluating the geological availability based on the geological data and geological event data. This module ensures the standardization and uniformity of the data, making it suitable for subsequent analysis and processing. An underground three-dimensional space model is constructed based on the positioning API, and the preprocessed underground space data is mapped to the model. The model can intuitively display the layout of the underground space and the location of the existing pipelines, providing visual support for the formulation of the relocation plan. The BP neural network is used to learn the safety logic of pipeline layout through underground space data. The relocation available area is marked in the underground three-dimensional space model to ensure that the layout of the new pipeline meets the requirements of safety and stability. Conflicting space data is obtained and the underground three-dimensional space model is updated synchronously. Multi-party data is integrated in real time to ensure the accuracy and real-time performance of the model. Based on the conflicting space data and the relocation available area, the relocation space is extracted from the underground three-dimensional space model. The relocation area is dynamically adjusted to adapt to on-site changes. By acquiring real-time conflicting spatial data and synchronously updating the underground three-dimensional spatial model, the system can quickly respond and adjust the relocation plan when encountering temporary obstacles or geological changes at the construction site, avoiding shutdowns and resource waste caused by information lag. The system can quickly calculate the optimal relocation route by cutting the relocation space with a preset step size to generate several two-dimensional pixel matrices, and analyzing these matrices through convolutional neural networks. The feature extraction and recognition capabilities of CNN enable the system to efficiently process complex underground space data and find suitable relocation paths. If there is no available relocation route in the relocation space, the system can add relocation space based on the preset gradient and recalculate until an available relocation route is found. This automated optimization process significantly improves the efficiency and accuracy of relocation work and reduces the need for manual intervention. By analyzing continuous two-dimensional pixel matrices, the system can accurately identify the available areas and risk areas of the underground space, ensure the safety and stability of the relocation route, and reduce safety hazards during construction. Since the system can dynamically adjust in real time and quickly find the optimal relocation route, it reduces downtime and the possibility of repeated construction, thereby significantly reducing project costs and improving resource utilization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a structural schematic diagram of an adaptive pipeline relocation intelligent optimization system in the present invention.

[0045] Figure 2 It is a flow chart of the execution steps of the adaptive strategy generation module in the present invention. DETAILED DESCRIPTION

[0046] See also Figure 1-2As shown, the present invention relates to an adaptive pipeline relocation intelligent optimization system, comprising: a data receiving module, a data preprocessing module, an underground space mapping module, a pipeline relocation learning module and an adaptive strategy generation module, wherein the data receiving module is connected to the data preprocessing module, and the underground space mapping module is connected to the data preprocessing module, the pipeline relocation learning module and the adaptive strategy generation module respectively;

[0047] A data receiving module is used to obtain underground space data, including underground use data, geological data and geological event data;

[0048] A data preprocessing module, used for preprocessing underground space data, wherein the preprocessing includes extracting location information, depth information and volume information of underground occupancy, and evaluating geological availability based on geological data and geological event data;

[0049] The underground space mapping module is used to build an underground 3D space model based on the positioning API and map the underground 3D space model based on the preprocessed underground space data;

[0050] The pipeline relocation learning module is used to learn the pipeline layout safety logic through underground space data based on the BP neural network, and to mark the relocation available area in the underground three-dimensional space model based on the pipeline layout safety logic;

[0051] The adaptive strategy generation module is used to perform the following steps:

[0052] S1. Obtain conflicting spatial data and synchronously update the underground three-dimensional spatial model;

[0053] S2, extracting the relocation space from the underground three-dimensional space model based on the conflicting spatial data and the relocation available area;

[0054] S3, based on the migration space, several two-dimensional pixel matrices are generated by cutting with a preset step size, and the available migration routes are calculated from the continuous two-dimensional pixel matrices through a convolutional neural network; if there is no available migration route in the migration space, the migration space is increased based on the preset gradient, and S3 is re-executed; if there is an available migration route in the migration space, the available migration route is output and synchronized in the underground three-dimensional space model.

[0055] In some embodiments, the data receiving module is used to obtain underground space data, specifically including underground usage data, geological data and geological event data. The data is received through the integration of information sent by multiple relevant departments, including underground pipeline layout data provided by the urban planning department, geological data provided by the geological exploration department, and geological event data provided by the emergency management department. After these data are integrated, the comprehensiveness and accuracy of the information are ensured. The data preprocessing module is used to preprocess the received underground space data. The specific steps include: first, extracting the positioning information, depth information and volume information of underground occupancy. Then, the geographic coordinates of each data point are converted into a standard coordinate system through the positioning API, and its longitude and latitude, depth and volume information are recorded. Finally, the geological availability is evaluated based on the geological data and geological event data, and the geological data and the number of occurrences of geological events are weightedly evaluated using the transformer model to determine the geological availability of the underground space. The underground space mapping module constructs an underground three-dimensional space model based on the preprocessed underground space data. The specific steps include: first, the longitude and latitude, depth and volume information of the occupied space are three-dimensionally positioned based on the positioning API, and mapped into the underground three-dimensional space model framework. Next, the color values ​​of different data categories are changed to generate an intuitive three-dimensional visualization model for subsequent analysis and processing. The pipeline relocation learning module is based on the BP neural network and learns the safety logic of pipeline layout through underground space data. The specific steps include: obtaining the layout distances between different underground pipelines, underground facilities and underground cavities; using the layout type and the stability data of the corresponding underground soil as input, and using the layout distance as output to train the BP neural network to obtain a layout safety logic model. The model is imported into the underground three-dimensional space model, and the available area for relocation is analyzed and calibrated, and the calibration is performed using the preset available color values.

[0056] Furthermore, the underground usage data includes:

[0057] Underground pipeline data, including the location, direction, material and usage information of water pipelines, drainage pipelines, gas pipelines, power cables and communication optical cables;

[0058] Underground facilities data, including the location, structure and usage of underground garages, underground shopping malls, subway stations and tunnels;

[0059] Underground cavity data, including the location, size and volume of underground mining areas, karst caves and old underground building ruins;

[0060] The geological data includes underground soil quality at regional depth and corresponding stability data; the geological event data includes the location and depth of collapse and subsidence events.

[0061] It should be noted that underground pipeline data includes the location, direction, material and usage information of water supply pipelines, drainage pipelines, gas pipelines, power cables and communication optical cables. For example, at a large construction site, the urban planning department provides data on all existing underground pipelines. The location data of water supply pipelines includes the coordinates of the starting and ending points of each pipeline, the direction information is represented by a series of connection points, the material information indicates whether the pipeline is made of PVC, steel or cast iron, and the usage information covers the service life, pressure level and maintenance record of the pipeline. Underground facility data includes the location, structure and usage of underground garages, underground shopping malls, subway stations and tunnels. For example, in a subway station project, the data of underground garages is provided by the traffic management department, including the floor plan of the garage, the location of the entrance and exit, the height of each floor and the number of parking spaces. Underground cavity data includes the location, size and volume of underground voids, karst caves and old underground building ruins. For example, the data of an underground void provided by the geological exploration department includes the three-dimensional coordinates, volume information and formation reasons of the void. The data of old underground building ruins is provided by the cultural relics protection department, including the location information of the ruins. Underground soil and corresponding stability data: including soil types and stability assessments of different depth areas. For example, at a construction site, the geological exploration report describes in detail that the soil at a depth of 0-10 meters is sandy and has a medium stability grade; the soil at a depth of 10-20 meters is clay and has a high stability grade. Stability data also includes soil shear strength, compression coefficient, and moisture content. Location and depth of collapse and subsidence events: including the specific location, depth, and impact range of collapse and subsidence events that have occurred in history. For example, the collapse event data for a certain area provided by the emergency management department records the time of the event, the specific location, the depth of the collapse (such as 30 meters), and the radius of the affected area (such as 50 meters). The data on subsidence events include the amount of settlement, duration, and the degree of impact on surrounding buildings.

[0062] Furthermore, the extracting of the location information, depth information and volume information of underground occupancy and evaluating the geological availability based on the geological data and geological event data comprises the following steps:

[0063] Based on the transformer model, the location information, depth information and volume information of the occupied space are extracted from the underground usage data, and the geographic coordinates of each data point are converted into a standard coordinate system through the positioning API to record the latitude and longitude, depth and volume of the occupied space;

[0064] Weights are selected according to the type of geological events, and geological data and the number of occurrences of geological events are weighted to evaluate the geological availability of underground space.

[0065] In some embodiments, first, the system uses the transformer model to extract the location information, depth information and volume information of the occupied space from the underground use data. Taking the city's underground pipelines as an example, the urban planning department provides detailed data on various pipelines, including the starting and ending coordinates (latitude and longitude) of each pipeline, as well as information such as the buried depth and diameter. The transformer model extracts the specific location and geometric parameters of each pipeline by processing these data. For example, the data of a water supply pipeline may include the starting point coordinates (30.2672°N, 97.7431°W), the end point coordinates (30.2675°N, 97.7434°W), the depth of 5 meters and the diameter of 0.5 meters. The model extracts this information to form standardized structured data. Then, these extracted geographic coordinates are converted into a standard coordinate system through the positioning API. The positioning API can handle the conversion between various coordinate systems to ensure that data from different sources can be represented in the same coordinate system. The converted data includes the latitude and longitude, depth and volume of the occupied space. For example, the starting and ending coordinates of a water supply pipeline are converted to points in a standard coordinate system, the depth information is recorded as 5 meters, and the volume is calculated based on the pipe diameter and pipe length. First, the geological event data includes records of collapse, subsidence and other events that have occurred in the history of the region. The record of each event includes the specific location, depth and impact range. For example, the data provided by the emergency management department may record a collapse event that occurred in a certain area at (30.2673°N, 97.7432°W), a depth of 10 meters, and an impact range radius of 20 meters. The system imports this data into the assessment model. Then, weights are selected based on the type of geological event. Different types of geological events have different effects on the stability of underground space, and the system reflects these differences by setting weights. For example, the weight of a collapse event may be set to 1.5, while the weight of a subsidence event may be set to 1.2. The weight is set based on the potential impact of the event on the underground space. The higher the weight, the greater the impact of the event on the underground space. The system performs weighted processing on the extracted geological data and geological event data. Geological data includes soil types at different depths in the region and their stability indicators, such as shear strength, compression coefficient and moisture content. The stability indicators of each soil type are weighted to obtain a weighted stability score. For example, the sand at a depth of 0-10 meters in a certain area has a shear strength of 30kPa and a weight of 1.2. After weighted calculation, the weighted stability score is 36kPa. Geological event data are also weighted to evaluate their impact on the stability of underground space. For example, three collapse events have occurred in an area in the past five years. The impact range and depth of each event are recorded as follows: the first event was at a depth of 5 meters and the impact range radius was 10 meters; the second event was at a depth of 10 meters and the impact range radius was 15 meters; the third event was at a depth of 15 meters and the impact range radius was 20 meters.The system weights and calculates the comprehensive impact of these events on the stability of the underground space in the area based on the type and number of events. Finally, the system combines the weighted geological data and geological event data to evaluate the geological availability of the underground space. The availability score reflects the stability and applicability of the underground space at different depths and locations, and provides a basis for the formulation of pipeline relocation plans. For example, the system evaluates the geological availability score of an area at a depth of 0-10 meters as 85 points (out of 100 points), and the score at a depth of 10-20 meters is 75 points. The higher the score, the more suitable the geological conditions in the area are for pipeline relocation.

[0066] Furthermore, mapping the underground three-dimensional space model based on the preprocessed underground space data includes the following steps:

[0067] Obtain pre-processed underground space data, including latitude and longitude, depth, volume of occupied space and geological availability of underground space;

[0068] The positioning API is used to locate the latitude, longitude, depth and volume information of the occupied space in three dimensions and map it into the underground three-dimensional space model framework, and the color value is changed according to different data categories.

[0069] In some embodiments, first, the processed underground space data is obtained from the data preprocessing module. These data include the latitude and longitude, depth, volume information of the occupied space and the geological availability of the underground space. For example, the data of a water supply pipeline includes the starting point coordinates (30.2672°N, 97.7431°W), the end point coordinates (30.2675°N, 97.7434°W), the depth of 5 meters, the pipe diameter of 0.5 meters, and the assessed geological availability score of 85 points. Similarly, the data of the underground garage may include the location coordinates, the height of each floor, the total building area and its structural stability score. Next, the latitude and longitude, depth and volume information of the occupied space are three-dimensionally positioned using the positioning API. The interface of the positioning software API can handle the conversion between various coordinate systems and unify data from different sources into a standard coordinate system. In the three-dimensional space model, the starting and end point coordinates of the water supply pipeline are converted to points in the standard coordinate system, and three-dimensional line segments are generated through these points to represent the path of the pipeline. The depth and diameter of the pipeline are used to define the vertical position and thickness of the line segment. Similarly, the coordinates of the underground garage are converted to a standard coordinate system, and a three-dimensional cube is generated based on the height and area information to represent the structure of the garage. The converted three-dimensional positioning data is mapped into the underground three-dimensional space model framework. This process includes classifying different types of underground space data (such as pipelines, facilities, cavities, etc.) by category and displaying them separately. In order to facilitate identification and analysis, the system changes the color value of data of different categories. In order to more intuitively display the stability of the underground space, the system also maps the geological availability score to the three-dimensional model through color value coding. Specifically, areas with higher scores are displayed in green, indicating good stability; areas with medium scores are displayed in yellow, indicating general stability; and areas with lower scores are displayed in red, indicating poor stability. For example, if the geological availability score of an area at a depth of 0-10 meters is 85 points, the area is displayed in green in the three-dimensional model. If the score at a depth of 10-20 meters is 65 points, it is displayed in yellow. In this way, the three-dimensional space model not only shows the spatial layout of underground pipelines and facilities, but also intuitively reflects the geological stability of different areas. Engineering personnel can intuitively understand the layout and stability of the underground space through the model, which is convenient for formulating relocation plans and construction plans.

[0070] Furthermore, learning pipeline layout safety logic through underground space data includes the following steps:

[0071] Obtain the layout distances between different underground pipelines, underground facilities and underground cavities;

[0072] The layout type and the stability data of the corresponding underground soil are used as input, and the layout distance is used as output to train the BP neural network and obtain the layout safety logic model.

[0073] In some embodiments, first, the system obtains detailed data of different underground pipelines, underground facilities and underground cavities from the data receiving module. These data include the location, direction, depth and structural information of each pipeline and facility. For example, the underground pipeline data in a certain area includes the location and direction of water supply pipelines, drainage pipelines, gas pipelines, power cables and communication optical cables; the underground facility data includes the structural information of underground garages, underground shopping malls, subway stations and tunnels; the underground cavity data includes the location and volume information of goafs, karst caves and old underground building ruins. The system uses these data to calculate the layout distances between different pipelines, facilities and cavities. Specifically, through the geometric relationship between points and line segments in the three-dimensional space model, the system can accurately calculate the minimum distance between each pipeline and other pipelines, facilities and cavities. For example, the minimum distance between a water supply pipeline and a drainage pipeline is 2 meters, the minimum distance between a water supply pipeline and an underground garage is 5 meters, and the minimum distance between a water supply pipeline and a karst cave is 10 meters. These layout distance data will serve as the basic data for subsequent training of the neural network. The system then uses the layout types of different pipelines and facilities and the stability data of their corresponding underground soil as input features of the BP neural network. The layout types include pipeline types (such as water supply pipelines, gas pipelines, etc.) and facility types (such as underground garages, subway stations, etc.), and the stability data of underground soil includes soil types (such as sand, clay, rock, etc.) and their stability indicators (such as shear strength, compression coefficient and moisture content). For example, the layout type of a water supply pipeline is "water supply pipeline", and its corresponding soil is sand, with a shear strength of 30kPa, a compression coefficient of 0.2, and a moisture content of 10%. After these data are standardized, they form a set of input feature vectors and are input into the BP neural network. Similarly, the data of other pipelines and facilities are also processed to form input features. The system uses the layout distances between different underground pipelines, facilities and cavities as output labels to train the BP neural network. The BP neural network learns the layout safety logic by continuously adjusting weights and biases to minimize the error between the predicted output and the actual layout distance. The specific training process is as follows: Data standardization: standardize the input features and output labels to ensure that the data is trained on a uniform scale. Model initialization: initialize the weights and biases of the BP neural network. The network structure includes an input layer, several hidden layers and an output layer. The number of hidden layer nodes and the number of layers are determined according to the complexity of the data. Forward propagation: forward propagate the input feature vector through each layer of the neural network, calculate the weight and bias of each layer, and obtain the predicted output. Calculate the loss: use the loss function (such as mean square error) to calculate the error between the predicted output and the actual deployment distance. Back propagation: through the back propagation algorithm, adjust the weights and biases of the network according to the loss gradient to minimize the loss function. Iterative training: multiple iterative training until the loss function converges to obtain the optimal deployment safety logic model.

[0074] Furthermore, the marking of the migration available area in the underground three-dimensional space model based on the pipeline layout safety logic includes the following steps:

[0075] Import the deployment safety logic model obtained through BP neural network learning;

[0076] Loading the underground three-dimensional space model;

[0077] Based on the layout safety logic model, the data in the underground three-dimensional space model is analyzed, the available areas for relocation are identified, and the available areas for relocation are calibrated using preset available color values.

[0078] In some embodiments, first, the layout safety logic model obtained by BP neural network training is imported into the system. The model has mastered the safety layout logic of different types of pipelines and facilities under different geological conditions by learning a large amount of underground pipelines, facilities and geological data. The model file usually adopts a standard machine learning model format (such as ONNX, TensorFlow SavedModel, etc.) to ensure that it can be correctly read and applied by the system. Next, the system loads the underground three-dimensional space model. The model is built based on the preprocessed underground space data and contains three-dimensional space information of underground pipelines, facilities, cavities and their geological conditions. During the loading process, the integrity and consistency of the model are checked to ensure that all necessary data (such as pipeline location, depth, volume, geological availability score, etc.) have been correctly mapped into the three-dimensional space. For example, an underground three-dimensional space model may show the spatial distribution of all underground pipelines and facilities in an area, including the specific location and direction of various pipelines (such as water supply pipelines, gas pipelines, power cables, etc.), the structure of underground facilities (such as underground garages, subway stations), and the specific location and size of underground cavities (such as caves and goafs). The system reads all pipeline, facility and cavity data in the three-dimensional spatial model, extracts their location information, depth information, volume information and geological availability score. By matching with the rules in the layout safety logic model, the system analyzes the safe layout distance between each data point and other surrounding pipelines and facilities. Based on the analysis results of the layout safety logic model, the system identifies areas that meet the safety layout requirements. Specifically, the system calculates the minimum safe distance between each pipeline, facility and cavity to ensure that no safety hazards occur within this distance range. For example, if the system identifies that there are no other pipelines or facilities within 10 meters around a water supply pipeline, the area is marked as a relocation available area. Finally, the system marks the identified relocation available areas so that users can intuitively understand which areas are suitable for pipeline relocation. The system presets different color values ​​to represent different availability levels. For example, green indicates a high availability area, yellow indicates a medium availability area, and red indicates a low availability area.

[0079] Furthermore, the S1 comprises the following steps:

[0080] Acquire real-time conflict space data, including new obstacles, geological changes and emergency data found during the construction process;

[0081] The acquired conflicting spatial data is synchronously updated with the current underground three-dimensional spatial model.

[0082] In some embodiments, during the construction process, the construction team may encounter unrecorded underground obstacles, such as unrecorded old pipelines, construction waste, etc. The construction team detects the location, size and depth of these obstacles in real time through on-site survey equipment (such as underground radar, ground penetrating radar, etc.), and uploads the data to the system. Geological changes may be encountered during the construction process, such as changes in soil layers, new cracks or changes in groundwater levels discovered by geological exploration. Geological engineers obtain these geological change data in real time through on-site detection equipment (such as drilling equipment, geological sensors, etc.) and upload them to the system. Emergencies may occur during the construction process, such as collapses and landslides. These events have a significant impact on construction safety and pipeline relocation. The on-site safety monitoring system (such as video surveillance, emergency sensors, etc.) monitors these emergencies in real time and uploads the data to the system. The system needs to synchronize the acquired real-time conflict space data with the current underground three-dimensional space model to ensure the real-time and accuracy of the model. First, receive and integrate real-time conflict space data from different sources. For new obstacle data, the system records its specific location, size and depth; for geological change data, the system records the geological properties and change characteristics of the change area; for emergency data, the system records the specific location, time and impact range of the event. The integrated conflict space data is applied to the current underground three-dimensional space model. Specifically, the system updates the space occupancy information of the corresponding area in the model according to the location and size of the new obstacle; updates the geological properties of the affected area in the model according to the geological change data; and updates the structure and stability information of the event area in the model according to the emergency data. The system ensures the real-time update of the model, that is, when the new conflict space data is uploaded, the system can complete the synchronous update of the model in a short time. In this way, the underground three-dimensional space model is always kept up to date and reflects the actual situation of the construction site. For example, after receiving the newly discovered old pipeline data, the system immediately adds the location and size of the pipeline to the model; after receiving the geological change data, the soil layer information of the corresponding area in the model is updated; after receiving the collapse event data, the terrain and structure information of the collapse area in the model is updated.

[0083] Furthermore, the S2 includes extracting a preset size area containing conflicting spatial data from the underground three-dimensional space model based on the synchronously updated model, and defining the area as the relocation space.

[0084] During the construction process, the system has acquired and synchronously updated the real-time conflict space data, including new obstacles, geological changes, and emergencies, through the S1 step. These data have been integrated into the current underground three-dimensional space model to ensure that the model reflects the latest underground space conditions. According to the construction requirements and actual conditions, the system presets a standard-sized space area, such as a cube area with a side length of 20 meters. This preset-sized area is designed to cover the range of the conflict space data and provide sufficient analysis space. For each conflict space data, the system first determines the coordinates of its center point. For example, for a newly discovered old pipeline, the system records the location of its center point; for a geological change area, the system records the center point of its change area; for an emergency, the system records the center point of the event impact range. Based on the center point of the conflict space data, the system extracts a preset-sized cube area in the three-dimensional space model. For example, assuming that the coordinates of the center point of the conflict spatial data are (30.2672°N, 97.7431°W, depth 5 meters), the system will extract a cube area with a side length of 20 meters with this point as the center, that is, the range from (30.2672°N, 97.7431°W, depth -5 meters) to (30.2672°N, 97.7431°W, depth 15 meters).

[0085] Furthermore, the convolutional neural network is trained by the following steps:

[0086] By marking the digital migration pipeline in the migration space, the digital migration pipeline is divided into turning nodes, each divided migration pipeline is traversed in parallel in the migration available area, and the traversal results of each divided migration pipeline are combined to generate a migration route dataset;

[0087] Based on the migration route data set, the migration space is cut with a preset step size to generate several continuous two-dimensional pixel matrices;

[0088] The pixel matrix of the migration space of the uncalibrated digital migration pipeline is used as input and several continuous two-dimensional pixel matrices are used as output to train the convolutional neural network.

[0089] In some embodiments, first, the system calibrates the digital relocation pipeline in the relocation space. These digital relocation pipelines simulate the real pipeline relocation path and take into account various factors in actual construction, such as the starting and ending positions of the pipeline, the turning nodes on the path, etc. The system divides the calibrated digital relocation pipeline into turning nodes. Specifically, the system identifies each turning node on the pipeline path and divides the pipeline at these nodes. For example, a water supply pipeline runs from point A to point B, passing through two turning nodes C and D in the middle. The system divides the pipeline into three sections: AC, CD and DB. Next, the system traverses each segment of the relocation pipeline in parallel in the relocation available area. The system generates multiple paths parallel to the segmented relocation pipeline to ensure that as many possible paths as possible are covered in the relocation available area. For example, for the AC segment, the system generates multiple parallel paths from point A to point C, and a certain distance is maintained between each path. The system records detailed data for each parallel path, including the starting point, end point, length, geological conditions along the path, etc. Through these records, the system can fully understand the feasibility and potential risks of each path. The system combines the traversal results of each segmented relocation pipeline to generate a complete relocation route data set. The system combines each parallel path to form a complete relocation route. For example, multiple parallel paths of segment AC are combined with multiple parallel paths of segment CD and segment DB to generate multiple complete relocation routes from point A to point B. The system records all the combined relocation route data to form a complete relocation route data set. The data of each route includes the starting point, end point, length, geological conditions along the way, risk assessment, etc. Based on the generated relocation route data set, the system cuts the relocation space with a preset step length to generate several continuous two-dimensional pixel matrices. The system cuts the relocation space according to the preset step length. For example, the system cuts the relocation space vertically with a step length of 1 meter to generate a series of plane images. The system converts the data on each cutting plane into a two-dimensional pixel matrix. Each pixel represents a small area on the plane, and the pixel value reflects the geological conditions, pipeline density and other information in the area. Finally, the system uses the pixel matrix of the relocation space of the uncalibrated digital relocation pipeline as input and several continuous two-dimensional pixel matrices as output to train the convolutional neural network. The system converts the relocation space data of the uncalibrated digital relocation pipeline into a two-dimensional pixel matrix as the input data of the convolutional neural network. Each input matrix reflects the actual situation of the spatial plane. The system inputs the generated two-dimensional pixel matrix into the convolutional neural network model for training. During the training process, the model extracts the features of the input data through the convolution layer and the pooling layer, and performs path optimization calculations through the fully connected layer. The model learns how to select the optimal relocation path under different geological conditions and spatial layouts. Through multiple iterative training, the system adjusts the parameters of the convolutional neural network to minimize the error between the predicted path and the actual optimal path.After sufficient training, the model can accurately predict and optimize pipeline relocation paths.

[0090] Furthermore, the formula of the convolutional neural network is as follows:

[0091]

[0092] Where R represents the calculated optimal available relocation route; X i，j W represents the input of the jth convolutional layer in the i-th continuous two-dimensional pixel matrix; j represents the jth convolution kernel; * represents the convolution operation; b j represents the bias of the jth convolution kernel; f represents the activation function; arg max i represents the path when i is selected to maximize the sum; n represents the number of layers of the convolutional neural network.

[0093] It should be noted that the convolution operation extracts the features of the input data by sliding the convolution kernel on the input matrix and calculating the weighted sum of the local area, thereby obtaining a new feature matrix that contains the feature information of the input data in the local area. The result of the convolution operation will be added with a bias b j Then, a nonlinear transformation is performed through the activation function f. Common types include ReLU (rectified linear unit), Sigmoid, and Tanh. The role of the activation function is to introduce nonlinear factors so that the neural network can better fit complex data distributions. For example, the ReLU function can set the negative part of the input to zero, thereby retaining only the positive part of the input. This process can enhance the expressive power of the model and enable it to handle more complex situations. Convolutional neural networks usually contain multiple convolutional layers, each of which extracts features at different levels of the input data. In the formula Indicates the accumulation of features from all convolutional layers, combining the features extracted from each layer. For example, the first convolutional layer may extract edge features of the image, the second convolutional layer may extract texture features, and the third convolutional layer may extract higher-level shape features. Through multiple layers of convolution, the neural network can extract deep information from the input data layer by layer, thereby more comprehensively understanding the structure of the input data. In the final path selection stage, the arg max in the formula iIndicates selecting the path that maximizes the sum. Specifically, the neural network calculates the total eigenvalues ​​of all possible relocation paths, and then selects the path with the largest sum of eigenvalues ​​as the optimal relocation route. This process ensures that the selected path has the highest feasibility and safety among all possible paths. For example, assuming there are three candidate paths, the neural network will calculate the total eigenvalues ​​of each path respectively, and then select the path with the highest eigenvalue as the final relocation route. Through the above detailed steps and formula explanations, the system realizes the function of optimizing pipeline relocation paths through convolutional neural networks, provides intelligent support for the planning and decision-making of pipeline relocation, and significantly improves the efficiency and safety of relocation work. Further, by mapping the generated relocation route into a three-dimensional model, it is also possible to perform digital positioning export to convert the specific relocation plan into longitude and latitude data to help construction personnel make auxiliary judgments and relocations.

[0094] Furthermore, if there is no available migration route in the migration space, the migration space is increased based on the preset gradient, and S3 is re-executed; if there is an available migration route in the migration space, the available migration route is output and synchronized in the underground three-dimensional space model. Specifically, when the system fails to find an available migration route in the initial migration space, the system will increase the migration space based on the preset gradient. For example, if the range of the initial migration space is 10 meters × 10 meters, the system can expand it to 20 meters × 20 meters, or adjust it according to actual needs, and then execute S3 to achieve adaptive adjustment. Based on dynamic adaptive adjustment and rapid finding of the optimal migration route, the intelligent generation of the migration plan is realized, which significantly reduces the engineering cost and improves the efficiency of resource utilization.

[0095] The above implementation modes are merely descriptions of the preferred implementation modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary engineering and technical personnel in the field shall fall within the protection scope determined by the claims of the present invention.

Claims

1. An adaptive pipeline relocation intelligent optimization system, characterized in that: include: A data receiving module, a data preprocessing module, an underground space mapping module, a pipeline relocation learning module and an adaptive strategy generation module, wherein the data receiving module is connected to the data preprocessing module, and the underground space mapping module is respectively connected to the data preprocessing module, the pipeline relocation learning module and the adaptive strategy generation module; A data receiving module is used to obtain underground space data, including underground use data, geological data and geological event data; A data preprocessing module, used for preprocessing underground space data, wherein the preprocessing includes extracting location information, depth information and volume information of underground occupancy, and evaluating geological availability based on geological data and geological event data; The underground space mapping module is used to build an underground 3D space model based on the positioning API and map the underground 3D space model based on the preprocessed underground space data; The pipeline relocation learning module is used to learn the pipeline layout safety logic through underground space data based on the BP neural network, and to mark the relocation available area in the underground three-dimensional space model based on the pipeline layout safety logic; The adaptive strategy generation module is used to perform the following steps: S1. Obtain conflicting spatial data and synchronously update the underground three-dimensional spatial model; S2, extracting the relocation space from the underground three-dimensional space model based on the conflicting spatial data and the relocation available area; S3, based on the migration space, a plurality of two-dimensional pixel matrices are generated by cutting with a preset step size, and available migration routes are calculated from continuous two-dimensional pixel matrices through a convolutional neural network; if there is no available migration route in the migration space, the migration space is increased based on a preset gradient, and S3 is re-executed; if there is an available migration route in the migration space, the available migration route is output and synchronized in the underground three-dimensional space model; The convolutional neural network is trained by the following steps: By marking the digital migration pipeline in the migration space, the digital migration pipeline is divided into turning nodes, each divided migration pipeline is traversed in parallel in the migration available area, and the traversal results of each divided migration pipeline are combined to generate a migration route dataset; Based on the migration route data set, the migration space is cut with a preset step size to generate several continuous two-dimensional pixel matrices; The pixel matrix of the migration space of the uncalibrated digital migration pipeline is used as input and several continuous two-dimensional pixel matrices are used as output to train the convolutional neural network.

2. The adaptive pipeline relocation intelligent optimization system according to claim 1, characterized in that: The underground usage data includes: Underground pipeline data, including the location, direction, material and usage information of water pipelines, drainage pipelines, gas pipelines, power cables and communication optical cables; Underground facilities data, including the location, structure and usage of underground garages, underground shopping malls, subway stations and tunnels; Underground cavity data, including the location, size and volume of underground mining areas, karst caves and old underground building ruins; The geological data includes underground soil quality at regional depth and corresponding stability data; the geological event data includes the location and depth of collapse and subsidence events.

3. The adaptive pipeline relocation intelligent optimization system according to claim 1, characterized in that: The extraction of location information, depth information and volume information of underground occupancy and the assessment of geological availability based on geological data and geological event data include the following steps: Based on the transformer model, the location information, depth information and volume information of the occupied space are extracted from the underground usage data, and the geographic coordinates of each data point are converted into a standard coordinate system through the positioning API to record the latitude and longitude, depth and volume of the occupied space; Weights are selected according to the type of geological events, and geological data and the number of occurrences of geological events are weighted to evaluate the geological availability of underground space.

4. The adaptive pipeline relocation intelligent optimization system according to claim 3, characterized in that: The mapping of the underground three-dimensional space model based on the pre-processed underground space data comprises the following steps: Obtain pre-processed underground space data, including latitude and longitude, depth, volume of occupied space and geological availability of underground space; The positioning API is used to locate the latitude, longitude, depth and volume information of the occupied space in three dimensions and map it into the underground three-dimensional space model framework, and the color value is changed according to different data categories.

5. The adaptive pipeline relocation intelligent optimization system according to claim 1, characterized in that: The method of learning pipeline layout safety logic through underground space data includes the following steps: Obtain the layout distances between different underground pipelines, underground facilities and underground cavities; The layout type and the stability data of the corresponding underground soil are used as input, and the layout distance is used as output to train the BP neural network and obtain the layout safety logic model.

6. The adaptive pipeline relocation intelligent optimization system according to claim 5, characterized in that: The method of marking the available area for migration in the underground three-dimensional space model based on the pipeline layout safety logic includes the following steps: Import the deployment safety logic model obtained through BP neural network learning; Loading the underground three-dimensional space model; Based on the layout safety logic model, the data in the underground three-dimensional space model is analyzed, the available areas for relocation are identified, and the available areas for relocation are calibrated using preset available color values.

7. The adaptive pipeline relocation intelligent optimization system according to claim 1, characterized in that: The S1 comprises the following steps: Acquire real-time conflict space data, including new obstacles, geological changes and emergency data found during the construction process; The acquired conflicting spatial data is synchronously updated with the current underground three-dimensional spatial model.

8. The adaptive pipeline relocation intelligent optimization system according to claim 1, characterized in that: The S2 includes extracting a preset size area containing conflicting spatial data from the underground three-dimensional space model based on the synchronously updated model, and defining the area as a relocation space.

9. The adaptive pipeline relocation intelligent optimization system according to claim 8, characterized in that: The formula of the convolutional neural network is as follows: ， in, represents the calculated optimal available relocation route; Indicates The first The input of the convolutional layer; Indicates convolution kernels; Represents the convolution operation; Indicates The bias of the convolution kernel; represents the activation function; Indicates that the sum is maximized. Path when Represents the number of layers of the convolutional neural network.

Images