Pipeline construction process based on digital-intelligent collaborative scheduling

Abstract

The invention belongs to the technical field of pipeline construction, and discloses a pipeline construction process based on digital-intelligent collaborative scheduling, which breaks through the limitation of single-point monitoring and static data processing in the prior art through a cross-domain fusion technology of federated learning and space-time compressed sensing: the problem of privacy disclosure in multi-subject data sharing is solved through federated learning, and the probability of privacy disclosure is lowered; low-bandwidth efficient transmission of high-dimensional data is achieved through space-time compressed sensing, the real-time performance of digital twin is improved to the millisecond level through cooperation of space-time compressed sensing and space-time compressed sensing, a dynamic data basis is provided for multi-subject collaborative decision making, the problems of'equipment waiting for materials', 'sudden shutdown 'and the like are solved through an equipment intelligent connection dispatching system through calculation power dynamic allocation and a fault early warning model, and the system is high in practicability. The resource utilization rate is obviously improved; distributed optical fiber sensing and AR remote processing are combined, the abnormal response period is greatly shortened, for example, when slope displacement exceeds the limit, the system can generate a repair scheme in an extremely short time and synchronously trigger field equipment linkage.

Description

Technical Field

[0001] This invention belongs to the field of pipeline construction technology, specifically a pipeline construction process based on digital and intelligent collaborative scheduling. Background Technology

[0002] Pipeline construction technology is widely used in the laying of underground pipelines in fields such as power, communications, and oil and gas, and is a core technology for ensuring infrastructure connectivity. With the increasing demand for complex urban environments and long-distance pipelines, the requirements for construction precision, efficiency, and full-cycle coordination have significantly increased. A search revealed that the invention patent with publication number CN118798549A discloses a pipeline engineering construction management method based on digital twins. It proposes the application of digital twins in pipeline construction and achieves progress comparison by constructing a sand table model. Its technical advantages include: reducing manual intervention by unifying data sources and improving the consistency between the digital twin and the actual project; and updating the as-built data in real time based on the model, reducing a large amount of manual processing work.

[0003] The patent has significant technical defects: insufficient real-time performance, with data update cycles at the minute level, which cannot meet the millisecond-level temperature field monitoring requirements during the welding stage; lack of dynamic optimization capabilities, making it difficult to cope with dynamic adjustments to construction paths, such as detour decisions caused by weather changes; weak privacy protection mechanisms, with direct data sharing among multiple participants, posing a risk of data leakage; and limited edge computing applications, only implementing data acquisition functions without addressing dynamic allocation of device computing power or edge-side preprocessing of abnormal events. Summary of the Invention

[0004] The purpose of this invention is to provide a pipeline construction process based on digital and intelligent collaborative scheduling to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a pipeline construction process based on intelligent and digital collaborative scheduling, the specific steps of which are as follows: S1. Construction of the Digital Intelligent Central Model: To address the shortcomings of existing technologies with data update cycles of minutes and inability to adapt to the millisecond-level monitoring needs of the welding stage, an IoT sensor array is used to collect microscopic parameters such as the crystal structure of raw materials (sampling frequency 1ms) and the direction of geological stratification. A federated learning dynamic privacy protection mechanism (different from the direct data sharing of existing technologies) is adopted to retain feature correlation during desensitization. Combined with spatiotemporal compression sensing technology, a digital twin is constructed to achieve holographic mapping and millisecond-level dynamic adaptation of construction elements, providing a data foundation for step S2. S2. Generation of multi-agent collaborative scheme: Based on the central model data, the task is decomposed and a decision chain is constructed using smart contracts; an optimal temporal network containing objectives such as resources and risks is generated using a hierarchical reinforcement learning algorithm; when a certain link is delayed, adjacent tasks are automatically adjusted to provide an execution framework for step S3. S3. Intelligent scheduling of resources and equipment: Based on the collaborative solution, edge computing and digital thread technology are integrated to dynamically allocate equipment computing power, track materials and trigger crane pre-positioning, and provide early warning of faults through vibration and energy consumption data to ensure continuous construction and provide a stable environment for step S4. S4. Holographic perception and anomaly handling: During construction, fiber optic sensing and infrared imaging are used to monitor anomalies, digital twins are used to predict trends, AR is used to remotely guide the handling, and quality parameters are collected simultaneously to provide complete data support for step S5. S5. Blockchain quality traceability: High-frequency parameters such as welding parameters are stored on the blockchain, linked to the credit and quality of the construction entity, and rewarded and punished using smart contracts. At the same time, environmental data is collected to provide a basis for step S6. S6. Dynamic Path Real-Time Optimization: Combining environmental data, the graph neural network is used to adjust the path weights, and the cost model is used to select the optimal solution to ensure continuous construction and provide a foundation for intelligent delivery upon completion. S7. Digital and intelligent integration of completed operation and maintenance: Generate archives containing 3D point clouds, build damage and operation and maintenance early warning models, construct them based on LSTM neural networks, connect them to the operation and maintenance system via blockchain, form a closed loop throughout the entire life cycle, and continuously optimize the initial central model.

[0006] Preferably, the specific steps for constructing the digital intelligence hub model in step S1 are as follows: S11. Multi-dimensional data acquisition and intelligent processing: Breaking through the traditional model, it collects microscopic parameters such as the crystal structure and geological stratification of pipeline raw materials, as well as environmental data such as groundwater level and building settlement through IoT sensors; it adopts federated learning algorithm and innovatively designs dynamic privacy protection mechanism to retain the correlation of key features during desensitization, laying the foundation for data fusion. Federated Learning Dynamic Privacy Protection Formula:

[0007] In the formula: For the first Global model parameters, and central model parameters after integrating data from all parties involved in the construction, such as the weights of the geological stratification analysis model; For the first The amount of data from participating parties (such as design units and construction units) (unit: data entries, covering raw material testing data, geological survey data, etc.); Total data volume of participating parties (unit: records); For the first Participant Local model parameters, such as equipment energy consumption prediction model parameters for construction units; Laplace Laplace noise function ( This is a privacy budget parameter, ranging from 0.1 to 0.5, used to control noise levels and ensure that the correlation between crystal structure and sedimentation coefficient is preserved after data anonymization. Formula source: Integrating federated learning weighted average algorithm with differential privacy mechanism to ensure that feature correlation is preserved during data anonymization; S12, Dynamic Adaptation of Digital Twin: Spatiotemporal compression sensing technology is introduced to construct a digital twin, increasing the acquisition frequency to the millisecond level, and edge node preprocessing optimizes transmission efficiency; the model has self-evolution capability and can adjust the data dimensions according to the construction stage. For example, in the welding stage, the weight of temperature field and stress field data is strengthened to achieve holographic mapping and provide accurate data support for step S2. Spatiotemporal compressed sensing reconstruction formula:

[0008]

[0009] In the formula: The original high-dimensional signal vector, dimension It includes spatiotemporal continuous data such as pipe welding temperature field and slope stress field, with a sampling interval of 1ms; For sparse basis matrices, wavelet transform matrix is ​​used, dimension To achieve sparse representation of the original signal, such as converting geological bedding strike data into sparse vectors; Norm measures the sparsity of a signal; the smaller the value, the sparser the signal, which facilitates efficient transmission. For the observation vector, dimension Low-dimensional data actually collected by the sensor, such as stress monitoring values ​​from fiber optic sensors and liquid level data from water level sensors. For the measurement matrix, dimension It employs a random Gaussian matrix to achieve the mapping from high-dimensional signals to low-dimensional observations, thus adapting to the low-bandwidth transmission requirements of IoT sensors. For the noise vector, dimension Noise includes environmental electromagnetic interference, sensor errors, etc., with a mean of 0 and a variance of [missing value]. ; Formula source: Based on compressed sensing theory, millisecond-level data acquisition is achieved through sparse optimization.

[0010] Preferably, the specific steps for generating the multi-agent collaborative scheme in step S2 are as follows: S21. Task Decomposition and Distributed Decision-Making: Based on the holographic data of the central model, the task modules are decomposed according to the construction objectives. A decision alliance chain is constructed through smart contracts. A hierarchical reinforcement learning algorithm is adopted to decompose sub-objective functions such as resource consumption. The optimal time series network is generated by solving through Nash equilibrium. Hierarchical reinforcement learning state transition formula:

[0011] The current state vector has dimensions [not specified]. It includes eight types of parameters, such as the location of construction equipment (unit: m), raw material inventory (unit: pieces), and slope displacement (unit: mm); For the current action, such as crane dispatch instructions: rotation angle (unit: °), lifting height (unit: m); This is a task-level parameter, with values ​​ranging from 1 to 3. 1 represents foundation construction, 2 represents pipeline welding, and 3 represents anti-corrosion treatment. The state transition probability, at the task level. Next, execute the action. After state Transferred to The probability of; Source: Hierarchical reinforcement learning framework for task-based temporal network generation; Nash equilibrium optimal solution formula:

[0012] For the first The utility function of the main body (Designer: Model accuracy score (0-100); Construction: Cost saving rate (%); Supervision: Quality pass rate (%)); For the first The optimal strategy for the main body, such as the welding sequence table of the construction party; For strategy combinations of entities other than i, such as the BIM model parameters of the designer + the sampling frequency of the supervisor; For the i-th subject, there is the set of all possible strategies, such as the three pipeline connection schemes of the construction party; Source: Nash equilibrium conditions in game theory, used for multi-agent task allocation; S22. Dynamic Adjustment and Collaborative Adaptation: The BIM model of the designer can be directly connected to the system through the model lightweight conversion technology, and the equipment capacity of the construction party and the quality standards of the supervision party are automatically matched at the nodes; when a main task is delayed, the system automatically adjusts the adjacent tasks. For example, when the crane dispatch is delayed, the operation sequence of the welding station is optimized simultaneously, breaking the limitations of traditional scheduling and providing a framework for step S3.

[0013] Preferably, the specific steps of intelligent scheduling of resource devices in step S3 are as follows: S31. Precise scheduling of equipment computing power and materials: Based on the task sequence of the collaborative solution, 5G-MEC edge computing and digital thread technology are integrated; a dynamic algorithm for allocating equipment computing power is developed, prioritizing the allocation of computing power during tunnel boring machine excavation and simultaneously adjusting the computing power occupancy of idle equipment; material delivery adopts the "digital thread + unmanned vehicle fleet" mode, automatically triggering the crane's pre-positioning program to ensure docking accuracy; The expression for the dynamic allocation model of device computing power is as follows:

[0014]

[0015] For the task Data volume (unit: GB, e.g., tunnel boring machine excavation parameter data is 50 GB / hour); For the task In the equipment Processing latency (unit: seconds, inversely proportional to device computing power); For equipment The computing power (unit: TOPS, such as 200 TOPS for edge node servers); For equipment Computing power utilization rate (unit: %, when the tunnel boring machine is working) ; For binary variables (1 represents the task) Assigned to device ,0 indicates unallocated); Formula source: Edge computing scheduling algorithm to optimize device computing power utilization; S32. Fault Early Warning and Continuous Support: Design an equipment health prediction model to provide early warning of faults based on vibration spectrum and energy consumption data, coordinate the scheduling of backup equipment, solve the problems of "equipment waiting for materials" and "sudden shutdown", and provide resource support for continuous construction.

[0016] Equipment health prediction model expression:

[0017] For health indicators, such as vibration spectrum values; These are characteristic variables, such as energy consumption and temperature; These are the regression coefficients; This is the error term; Formula source: Equipment degradation modeling based on stochastic processes to achieve fault early warning.

[0018] Preferably, the specific steps for handling holographic perception anomalies in step S4 are as follows: S41. Construction of Holographic Monitoring Network: In the collaborative operation of resources and equipment, break through single-point monitoring, deploy a distributed optical fiber sensor network, collect massive stress monitoring point data per meter of optical fiber, and combine infrared thermal imaging to capture the temperature gradient of the welding area to achieve all-round monitoring. S42. Anomaly Simulation and Remote Collaboration: The digital twin is equipped with an "anomaly simulation engine" that automatically simulates the deformation trend over a short period of time and generates multi-level solutions when the slope displacement rate exceeds the threshold. The AR remote guidance system uses "virtual-real anchoring" technology to precisely align the virtual repair path marked by experts with the actual scene. Combined with force feedback gloves, it enables remote operation guidance and simultaneously collects quality parameters during the disposal process.

[0019] Preferably, the specific steps of blockchain quality traceability in step S5 are as follows: S51. Parameter Uploading and Storage Mechanism: Based on the quality parameters collected during the anomaly handling stage, parameters such as welding current and voltage curves are automatically uploaded to the blockchain for storage at 100ms intervals to ensure that they cannot be tampered with and to build a complete quality data chain. S52. Credit Linkage and Data Linkage: Develop a skill-quality linkage algorithm to link construction personnel skills, equipment performance and quality. Implement rewards and penalties through smart contracts to achieve accurate traceability of quality responsibility. At the same time, receive dynamic environmental data to provide a basis for dynamic path optimization.

[0020] Preferably, the specific steps of the real-time dynamic path optimization in step S6 are as follows: S61. Dynamic adjustment of environmental data weights: Combining environmental data for quality traceability, the spatiotemporal attention mechanism of graph neural network (GNN) is applied to transform meteorological and traffic data into dynamic weight matrices. Meteorological radar data is updated according to the short-cycle rainfall probability distribution, and the traffic monitoring system collects the traffic flow density of the construction section in real time. The graph neural network adjusts the weights of path nodes in real time and controls the weight update delay. S62. Cost Model and Path Optimization: Design a "Construction Unit Migration Cost Model" to calculate the comprehensive cost when detours are required. Automatically calculate the comprehensive costs such as equipment transportation energy consumption and time loss, and prioritize the optimal solution to ensure construction continuity.

[0021] Preferably, the specific steps for the digital and intelligent completion operation and maintenance connection in step S7 are as follows: S71. Construction of As-Built Digital Archives: Dynamic path optimization ensures that after construction is completed, a digital archive containing 3D point clouds is generated with an accuracy of 0.5mm. Semantic segmentation technology is used to mark weak areas of the pipeline (such as the weld heat-affected zone), forming a complete construction data closed loop. S72, Operation and Maintenance Linkage and Full-Cycle Closed Loop: Innovate the "Damage Evolution-Operation and Maintenance Early Warning" model, link construction stress data with operation and maintenance corrosion rate, and train the remaining life prediction model; digital archives and operation and maintenance systems are connected using blockchain cross-chain technology to ensure the integrity of data transmission. Operation and maintenance personnel can use AR glasses to call up key operation records during the construction phase to achieve rapid matching of fault location and repair solutions, and link with the initial central model.

[0022] The beneficial effects of this invention are as follows: 1. This invention overcomes the limitations of existing technologies in single-point monitoring and static data processing by using a cross-domain fusion technology of federated learning and spatiotemporal compressed sensing: federated learning solves the privacy leakage problem in multi-subject data sharing (not addressed in existing technologies), and spatiotemporal compressed sensing enables low-bandwidth and efficient transmission of high-dimensional data (existing technologies only achieve minute-level acquisition). The synergy of the two technologies improves the real-time performance of digital twins to the millisecond level, providing a dynamic data foundation for multi-subject collaborative decision-making.

[0023] 2. This invention solves problems such as "equipment waiting for materials" and "sudden shutdowns" through a smart equipment scheduling system that dynamically allocates computing power and uses a fault early warning model, significantly improving resource utilization. The combination of distributed fiber optic sensing and AR remote handling greatly shortens the abnormal response cycle. For example, when the slope displacement exceeds the limit, the system can generate a repair plan in a very short time and simultaneously trigger the linkage of on-site equipment. Compared with the static progress management of existing technologies, this solution forms a full-chain control of "monitoring-early warning-handling-traceability", effectively reducing the construction accident rate.

[0024] 3. This invention combines blockchain quality traceability with a damage evolution model, enabling parameters such as welding current curves and stress concentration from the construction phase to be directly applied to corrosion prediction during the operation and maintenance phase, significantly improving the accuracy of remaining life assessment. Cross-chain integration of digital archives and the operation and maintenance system allows operation and maintenance personnel to access key construction records via AR, greatly improving fault location efficiency. Compared to the traditional fragmented "construction-operation and maintenance" data model, this closed-loop mechanism not only reduces the overall cost but also forms a construction knowledge graph through data accumulation, providing intelligent decision support for subsequent projects. Attached Figure Description

[0025] Figure 1 This is a flowchart of the pipeline construction process based on digital and intelligent collaborative scheduling according to the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] like Figure 1 This invention provides a pipeline construction process based on intelligent and digital collaborative scheduling. The specific steps of this process are as follows: S1. Construction of the digital and intelligent central model: Data such as the crystal structure and geological stratification of raw materials are collected through the Internet of Things. Federated learning is used to desensitize and fuse the data while retaining its relevance. Combined with spatiotemporal compression sensing, a digital twin is built. The data dimensions can be dynamically adjusted, such as strengthening temperature and stress data during welding, to achieve holographic mapping of construction elements and provide a data foundation for step S2. S2. Generation of multi-agent collaborative scheme: Based on the central model data, the task is decomposed and a decision chain is constructed using smart contracts; an optimal temporal network containing objectives such as resources and risks is generated using a hierarchical reinforcement learning algorithm; when a certain link is delayed, adjacent tasks are automatically adjusted to provide an execution framework for step S3. S3. Intelligent scheduling of resources and equipment: Based on the collaborative solution, edge computing and digital thread technology are integrated to dynamically allocate equipment computing power, track materials and trigger crane pre-positioning, and provide early warning of faults through vibration and energy consumption data to ensure continuous construction and provide a stable environment for step S4. S4. Holographic perception and anomaly handling: During construction, fiber optic sensing and infrared imaging are used to monitor anomalies, digital twins are used to predict trends, AR is used to remotely guide the handling, and quality parameters are collected simultaneously to provide complete data support for step S5. S5. Blockchain quality traceability: High-frequency parameters such as welding parameters are stored on the blockchain, linked to the credit and quality of the construction entity, and rewarded and punished using smart contracts. At the same time, environmental data is collected to provide a basis for step S6. S6. Dynamic Path Real-Time Optimization: Combining environmental data, the graph neural network is used to adjust the path weights, and the cost model is used to select the optimal solution to ensure continuous construction and provide a foundation for intelligent delivery upon completion. S7. Digital and intelligent integration of completed operation and maintenance: Generate archives containing 3D point clouds, build damage and operation and maintenance early warning models, construct them based on LSTM neural networks, connect them to the operation and maintenance system via blockchain, form a closed loop throughout the entire life cycle, and continuously optimize the initial central model.

[0028] In step S1, the construction of the digital central model refers to collecting crystal structure parameters of pipeline raw materials, geological bedding parameters, groundwater level, and building settlement through IoT sensors (including an X-ray crystal structure analyzer (model: BrukerD8), ground-penetrating radar (frequency: 50-2000MHz), and water level sensor (measurement range: 0-10m), deployed at the surface of the raw material warehouse and the construction area and 5m underground, respectively); and using a federated learning algorithm to design a dynamic privacy protection mechanism to retain the correlation of key features during the de-identification process, laying the foundation for data fusion. The specific implementation of the dynamic privacy protection mechanism is as follows: Based on the data sensitivity level (raw material crystal structure data is highly sensitive, geological bedding data is moderately sensitive), the desensitization algorithm is dynamically switched—highly sensitive data uses a differential privacy algorithm (Laplace noise λ is dynamically adjusted according to data sensitivity; λ=0.1 when sensitivity > 0.8, and λ=0.5 when sensitivity ≤ 0.8); moderately sensitive data uses a k-anonymity algorithm (k value = number of participants + 2), ensuring that the correlation of key features after desensitization is ≥ 90%. Federated Learning Dynamic Privacy Protection Formula:

[0029] In the formula: For the first Global model parameters, and central model parameters after integrating data from all parties involved in the construction, such as the weights of the geological stratification analysis model; For the first The amount of data from participating parties (such as design units and construction units) (unit: data entries, covering raw material testing data, geological survey data, etc.); Total data volume of participating parties (unit: records); For the first Participant Local model parameters, such as equipment energy consumption prediction model parameters for construction units; Laplace Laplace noise function ( This is a privacy budget parameter, ranging from 0.1 to 0.5, used to control noise levels and ensure that the correlation between crystal structure and sedimentation coefficient is preserved after data anonymization. A digital twin is constructed by introducing spatiotemporal compressed sensing technology, which increases the acquisition frequency to the millisecond level. Kalman filtering algorithm is used to preprocess edge nodes to optimize transmission efficiency. The data dimensions are adjusted according to the construction stage (geological stratification and groundwater level data dimensions are retained in the foundation construction stage, temperature field and stress field data dimensions are added in the welding stage, and surface roughness and coating thickness data dimensions are retained in the anti-corrosion treatment stage. The welding temperature dimension in the foundation construction stage and the redundant groundwater level dimension in the welding stage are deleted. Based on the correlation analysis of the core parameters of each stage, the adaptive learning module of the random forest algorithm is used to achieve holographic mapping and provide accurate data support for step S2. Edge node preprocessing method: Kalman filtering algorithm is used to reduce noise in the original data (filter window size = 5 sampling periods), and LZ77 compression algorithm is used to compress the data to 30%-50% of the original size to optimize transmission efficiency; Spatiotemporal compressed sensing reconstruction formula:

[0030]

[0031] In the formula: The original high-dimensional signal vector, dimension It includes spatiotemporal continuous data such as pipe welding temperature field and slope stress field, with a sampling interval of 1ms; For sparse basis matrices, wavelet transform matrix is ​​used, dimension To achieve sparse representation of the original signal, such as converting geological bedding strike data into sparse vectors; Norm measures the sparsity of a signal; the smaller the value, the sparser the signal, which facilitates efficient transmission. For the observation vector, dimension Low-dimensional data actually collected by the sensor, such as stress monitoring values ​​from fiber optic sensors and liquid level data from water level sensors. For the measurement matrix, dimension It employs a random Gaussian matrix to achieve the mapping from high-dimensional signals to low-dimensional observations, thus adapting to the low-bandwidth transmission requirements of IoT sensors. For the noise vector, dimension Noise includes environmental electromagnetic interference, sensor errors, etc., with a mean of 0 and a variance of [missing value]. .

[0032] In step S2, the multi-agent collaborative solution generation refers to decomposing task modules based on data from the central model according to construction objectives, constructing a decision-making alliance chain (containing three nodes: the designer, the construction party, and the supervisor, with consensus weights of 30% for the designer, 50% for the construction party, and 20% for the supervisor) through smart contracts, and employing a hierarchical reinforcement learning algorithm (reward function: The process involves breaking down the resource consumption sub-objective function (expression: f(res) = ∑(equipment energy consumption × working time)) and the risk assessment sub-objective function (expression: f(risk) = ∑(slope displacement × weight coefficient), where the weight coefficient is 0.6 when the displacement is > 5mm and 0.3 when it is ≤ 5mm). The optimal time-series network is then generated through Nash equilibrium solution. The specific steps are as follows: ① Initialize the strategy sets for the design, construction, and supervision parties (design strategy includes BIM model accuracy parameters, construction strategy includes equipment scheduling timing, and supervision strategy includes quality sampling frequency); ② Calculate the benefit value of each strategy combination based on the utility function (design utility = model matching degree, construction utility = cost saving rate, supervision utility = quality pass rate); ③ Adjust each main strategy with an iteration step size of 0.01. Calculate the utility function difference after each iteration. Stop when the difference is < 1% for 3 consecutive iterations, and output the current strategy combination as the optimal solution. Sub-objective function weight allocation rules: resource consumption sub-objective function weight accounts for 60% (equipment energy consumption weight 40%, raw material consumption weight 20%), risk assessment sub-objective function weight accounts for 40% (slope displacement weight 25%, equipment failure probability weight 15%). Hierarchical reinforcement learning state transition formula:

[0033] The current state vector has dimensions [not specified]. It includes eight types of parameters, such as the location of construction equipment (unit: m), raw material inventory (unit: pieces), and slope displacement (unit: mm); For the current action, such as crane dispatch instructions: rotation angle (unit: °), lifting height (unit: m); This is a task-level parameter, with values ​​ranging from 1 to 3. 1 represents foundation construction, 2 represents pipeline welding, and 3 represents anti-corrosion treatment. The state transition probability, at the task level. Next, execute the action. After state Transferred to The probability of; Nash equilibrium optimal solution formula:

[0034] For the first The utility function of the main body (Designer: Model accuracy score (0-100); Construction: Cost saving rate (%); Supervision: Quality pass rate (%)); For the first The optimal strategy for the main body, such as the welding sequence table of the construction party; For strategy combinations of entities other than i, such as the BIM model parameters of the designer + the sampling frequency of the supervisor; For the i-th subject, there is the set of all possible strategies, such as the three pipeline connection schemes of the construction party; The designer's BIM model is integrated into the system using IFC format lightweight conversion technology (compression ratio 1:5). The construction party's equipment capacity (parameter: welding speed (m / h)) and the supervisor's quality standards (parameter: weld pass rate (%)) are automatically matched with nodes through a linear matching algorithm. When a main task is delayed, the system automatically adjusts adjacent tasks. For example, when crane dispatch is delayed, the operation sequence of welding stations is optimized synchronously, breaking the limitations of traditional scheduling and providing a framework for step S3. Lightweight BIM Model Conversion: Convert the Revit format BIM model to IFC4 format using the Autodesk BIM360 platform, and use LOD300 hierarchical compression (remove non-construction necessary decorative component layers). After compression, the model size is controlled within 20% of the original size.

[0035] In step S3, intelligent scheduling of resources and equipment refers to integrating 5G-MEC edge computing and digital thread technology (implemented via the 5G NR protocol) according to the task sequence of the collaborative scheme; developing a dynamic allocation algorithm for equipment computing power (prioritizing the allocation of computing power when the tunnel boring machine's excavation speed is >5m / h; determining equipment idleness: if there are no tasks allocated for 10 consecutive minutes, the computing power of idle equipment is adjusted to 30% of the current computing power, and the occupancy is reduced to 50% of the rated value through a dynamic threshold algorithm); and adopting a "digital thread + unmanned vehicle fleet" mode for material delivery (the unmanned vehicle fleet is based on...). The algorithm-based path planning rules enable automatic triggering of the crane's pre-positioning program (achieving ±5mm precision control via UWB positioning sensors) to ensure docking accuracy. Computing power priority determination threshold: The highest priority is triggered when the tunnel boring machine's tunneling speed is >5m / h (computing power allocation ratio ≥60%); medium priority is triggered when the tunneling speed is 3-5m / h (ratio 30%-60%); and low priority is triggered when the tunneling speed is ≤3m / h (ratio ≤30%). 5G-MEC and Digital Thread Data Interaction Process: ① Digital Thread generates material ID, ② Edge nodes receive the ID and associate it with device status via 5GNR protocol, ③ Computing power allocation results are fed back to the collaborative solution module via Digital Thread, with an interaction delay of <20ms; The expression for the dynamic allocation model of device computing power is as follows:

[0036]

[0037] For the task Data volume (unit: GB, e.g., tunnel boring machine excavation parameter data is 50 GB / hour); For the task In the equipment Processing latency (unit: seconds, inversely proportional to device computing power); For equipment The computing power (unit: TOPS, such as 200 TOPS for edge node servers); For equipment Computing power utilization rate (unit: %, when the tunnel boring machine is working) ; For binary variables (1 represents the task) Assigned to device ,0 indicates unallocated); Design an equipment health prediction model to provide early warning of faults based on data such as vibration spectrum (frequency range: 10-1000Hz) and energy consumption (range: 50-500kW·h) (the model training dataset comes from 100,000 historical operating data of 100 similar equipment), coordinate the scheduling of backup equipment, solve the problems of "equipment waiting for materials" and "sudden shutdown", and provide resource guarantee for continuous construction. The equipment health prediction model features three types of variables: vibration spectrum (peak amplitude in the 10-500Hz frequency band), energy consumption data (average power deviation rate per hour), and equipment temperature (temperature difference between motor casing and ambient temperature). The vibration spectrum has a weight of 50%.

[0038] Equipment health prediction model expression:

[0039] For health indicators, such as vibration spectrum values; These are characteristic variables, such as energy consumption and temperature; These are the regression coefficients; This is the error term.

[0040] Among them, the holographic perception anomaly handling in step S4 refers to deploying a distributed optical fiber sensor network (10 monitoring points per meter, optical fiber type: single-mode optical fiber) in the collaborative operation of resources and equipment, combined with an infrared thermal imager (resolution: 640×512, sampling frequency: 30Hz, focusing range: 500×500mm of welding area) to capture the temperature gradient of the welding area and achieve all-round monitoring. The digital twin is equipped with an "anomaly simulation engine" based on a finite element analysis model, using solid elements C3D8R with a mesh density of 10mm / element. When the slope displacement rate exceeds the threshold of 2mm / h, it automatically simulates the deformation trend at a time step of 10s and generates multi-level solutions (early warning level: displacement <5mm; handling level: 5-10mm; emergency level: >10mm). The AR remote guidance system uses "virtual-real anchoring" technology (achieved through ARmarkers with an alignment accuracy of ±3mm). The virtual repair path marked by experts is precisely aligned with the actual scene, and the system, combined with force feedback gloves, enables remote operation guidance. Quality parameters are collected simultaneously during the handling process.

[0041] In step S5, blockchain quality traceability refers to automatically uploading parameters such as welding current and voltage curves to the blockchain for evidence storage at 100ms intervals based on the quality parameters collected during the anomaly handling stage (data format: JSON, encryption algorithm: SHA-256) to ensure immutability and build a complete quality data chain. Develop a skill-quality correlation algorithm (skill level assessment: based on the welding pass rate and training assessment results of the past 3 months, each accounting for 50% weight, quantified into 1-5 levels), linking the skills of construction personnel, equipment performance, and quality. Implement rewards and penalties through smart contracts (when the pass rate is >95% and three consecutive batches meet the standard, reward the construction team with a bonus (amount = base bonus × pass rate excess percentage, base bonus = unit price per meter of construction × batch construction length × 5%); when the pass rate is <80%, trigger an equipment shutdown and maintenance instruction and deduct 20% of the responsible person's monthly performance), to achieve precise traceability of quality responsibility, while receiving dynamic environmental data to provide a basis for dynamic path optimization.

[0042] In step S6, dynamic path real-time optimization refers to combining environmental data from quality traceability and applying the spatiotemporal attention mechanism of graph neural networks (GNNs) (12 nodes in the input layer, 24 nodes in the hidden layer, 8 nodes in the output layer, a time window of 10 minutes, and weights updated every 5 minutes) to transform meteorological (rainfall weight coefficient: 0.3) and traffic (traffic flow weight coefficient: 0.7) data into a dynamic weight matrix (weight update delay < 5 seconds, update trigger condition: data change > 10%). Meteorological radar data is updated according to the short-cycle rainfall probability distribution, the traffic monitoring system collects the traffic flow density of the construction section in real time, and the graph neural network adjusts the weights of the path nodes in real time to control the weight update delay. Attention weight calculation formula:

[0043] Where: q is the meteorological feature vector. For traffic feature vectors, =12; The design includes a "construction unit relocation cost model" (which includes equipment relocation energy consumption (formula: E = power × distance), labor time loss (formula: T = delay time × labor unit price), and detour distance additional cost (formula: C = detour distance × unit distance cost)). When detours are required, the model calculates the comprehensive cost, automatically calculates the comprehensive cost of equipment relocation energy consumption and labor time loss, and prioritizes the solution with the lowest comprehensive cost and construction interruption time of less than 30 minutes to ensure construction continuity.

[0044] In step S7, the digital and intelligent completion operation and maintenance connection refers to the generation of a digital archive containing three-dimensional point clouds after the completion of dynamic path optimization to ensure that the pipeline is marked with weak areas (such as the weld heat-affected zone) through semantic segmentation technology, thus forming a complete construction data closed loop. An innovative "Damage Evolution-Operation and Maintenance Early Warning Model" (trained using an LSTM neural network algorithm, containing two hidden layers with 64 neurons each; the input layer includes five types of features such as construction stress (100-500MPa) and corrosion rate (0.1-1mm / year); the output layer is the remaining lifespan (years); the training dataset contains 5000 sets of historical matching data) is used to correlate construction stress data with operation and maintenance corrosion rate (0.1-1mm / year) to train a remaining lifespan prediction model. Digital archives and the operation and maintenance system are connected using blockchain cross-chain technology to ensure the integrity of data transmission. Operation and maintenance personnel can use AR glasses to access key operation records during the construction phase, enabling rapid matching of fault location and repair solutions, and linking with the initial central model. LSTM neural network parameters: The input layer contains 5 features (construction stress 100-500MPa, corrosion rate 0.1-1mm / year, pipe wall thickness deviation ±0.5mm, number of welding defects, and environmental humidity 30%-90%); the hidden layer has 2 layers, each with 64 neurons, and the activation function is ReLU; the training iterations are 1000 times, the learning rate is 0.001, and the Adam optimizer is used; The cross-chain technology adopts the Polkadot relay chain protocol and realizes asset mapping between the construction blockchain and the operation and maintenance blockchain through XCM (cross-chain messaging). The specific process is as follows: ① The construction end uploads the hash value of the digital file to the Polkadot relay chain; ② The operation and maintenance end verifies the integrity of the hash value through the relay chain; ③ After the verification is successful, the operation and maintenance system calls the smart contract to synchronize the data, and the data transmission latency is controlled within 100ms.

[0045] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0046] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A pipeline construction technology based on intelligent and digital collaborative scheduling, characterized in that: The specific steps of this process are as follows: S1. Construction of the digital and intelligent central model: Collect raw material crystal structure data and geological stratification data through the Internet of Things, use federated learning to desensitize and fuse data while retaining its relevance, combine it with spatiotemporal compression sensing to build a digital twin, and dynamically adjust the data dimensions. S2. Multi-agent collaborative solution generation: Based on the central model data, the task is decomposed and a decision chain is constructed using smart contracts. The optimal temporal network containing resources and risks is generated using a hierarchical reinforcement learning algorithm. S3. Intelligent scheduling of resources and equipment: Based on the collaborative solution, it integrates edge computing and digital thread technology to dynamically allocate equipment computing power, track materials and trigger crane pre-positioning, and provide early warning of faults through vibration and energy consumption data. S4. Holographic Perception and Anomaly Handling: During construction, fiber optic sensing and infrared imaging are used to monitor anomalies, digital twins are used to predict trends, AR is used to remotely guide the handling, and quality parameters are collected simultaneously to provide data support for blockchain quality traceability. S5. Blockchain quality traceability: Welding parameters are stored on the blockchain, linked to the credit and quality of the construction entity, and rewarded or punished using smart contracts, while environmental data is collected. S6. Dynamic path real-time optimization: Combining environmental data, the path weights are adjusted using a graph neural network, and the optimal solution is selected using a cost model to ensure continuous construction. S7. Digital and intelligent integration of completed project operation and maintenance: Generate archives containing 3D point clouds, establish damage and operation and maintenance early warning models, build based on LSTM neural networks, and connect to the operation and maintenance system via blockchain.

2. The pipeline construction technology based on intelligent collaborative scheduling according to claim 1, characterized in that: The specific steps for constructing the digital intelligence hub model in step S1 are as follows: S11. Multi-dimensional data acquisition and intelligent processing: Collect crystal structure parameters of pipeline raw materials, geological bedding parameters, groundwater level, and building settlement through IoT sensors; adopt federated learning algorithm and design dynamic privacy protection mechanism to retain the correlation of key features during desensitization. S12. Dynamic adaptation of digital twins: Introducing spatiotemporal compression sensing technology to construct digital twins, increasing the acquisition frequency to the millisecond level, optimizing transmission efficiency through edge node preprocessing, and adjusting data dimensions according to the construction stage.

3. The pipeline construction technology based on intelligent collaborative scheduling according to claim 1, characterized in that: The specific steps for generating the multi-agent collaborative scheme in step S2 are as follows: S21. Task Decomposition and Distributed Decision-Making: Based on the data of the central model, the task modules are decomposed according to the construction objectives. A decision alliance chain is constructed through smart contracts. A hierarchical reinforcement learning algorithm is adopted to decompose the resource consumption sub-objective function. The optimal time series network is generated by solving the Nash equilibrium. S22. Dynamic Adjustment and Collaborative Adaptation: The designer's BIM model is converted into a lightweight form and integrated into the system, and the construction party's equipment capacity and the supervision party's quality standards are automatically matched at nodes.

4. The pipeline construction technology based on intelligent collaborative scheduling according to claim 1, characterized in that: The specific steps of intelligent scheduling of resource devices in step S3 are as follows: S31. Precise scheduling of equipment computing power and materials: Based on the task sequence of the collaborative solution, integrate 5G-MEC edge computing and digital thread technology to develop a dynamic allocation algorithm for equipment computing power, and prioritize the allocation of computing power when the tunnel boring machine is tunneling. Material delivery adopts a digital delivery network + driverless vehicle fleet model; S32. Fault Early Warning and Continuous Support: Design an equipment health prediction model to provide early warning of faults based on vibration spectrum and energy consumption data, and coordinate the scheduling of backup equipment.

5. The pipeline construction technology based on intelligent collaborative scheduling according to claim 1, characterized in that: The specific steps for handling holographic perception anomalies in step S4 are as follows: S41. Construction of Holographic Monitoring Network: In the collaborative operation of resources and equipment, a distributed optical fiber sensor network is deployed to collect massive stress monitoring point data per meter, and combined with infrared thermal imaging to capture the temperature gradient of the welding area. S42. Anomaly simulation and remote collaboration: The digital twin is equipped with an "anomaly simulation engine" to simulate deformation trends and generate multi-level solutions when standards are exceeded. The AR remote guidance system uses virtual-real anchoring technology to align the virtual path of experts with the site and collect quality parameters synchronously during the handling process.

6. The pipeline construction technology based on intelligent collaborative scheduling according to claim 1, characterized in that: The specific steps of blockchain quality traceability in step S5 are as follows: S51. Parameter Uploading and Evidence Storage Mechanism: Based on the quality data of abnormal handling, the welding current and voltage curves are automatically uploaded to the blockchain for evidence storage at 100ms intervals to ensure that they cannot be tampered with. S52, Credit Linkage and Data Linkage: Develop a skill-quality linkage algorithm to link construction personnel skills, equipment performance and quality, execute rewards and punishments through smart contracts, and receive dynamic environmental data.

7. The pipeline construction technology based on intelligent collaborative scheduling according to claim 1, characterized in that: The specific steps of dynamic path real-time optimization in step S6 are as follows: S61. Dynamic adjustment of environmental data weights: Combining environmental data for quality traceability, the attention mechanism of graph neural network in spatiotemporal space is applied to transform meteorological and traffic data into a dynamic weight matrix and adjust the weights of path nodes in real time. S62. Cost Model and Path Optimization: Design and construction unit migration cost model, calculate the comprehensive cost when detour is required, prioritize the optimal solution, and ensure construction continuity.

8. The pipeline construction technology based on intelligent collaborative scheduling according to claim 1, characterized in that: The specific steps for the digital and intelligent completion operation and maintenance connection in step S7 are as follows: S71. Construction of As-Built Digital Archives: Dynamic path optimization ensures that after construction is completed, a digital archive containing 3D point clouds is generated, and weak areas of the pipeline are marked through semantic segmentation technology; S72, Operation and Maintenance Linkage and Full-Cycle Closed Loop: Design Damage Evolution-Operation and Maintenance Early Warning Model, built on LSTM neural network, links construction stress data with operation and maintenance corrosion rate, and digital archives and operation and maintenance system are connected using blockchain cross-chain technology.

Images