Pipe jacking construction digital twinning optimization system and method based on high-precision physical simulation

Abstract

The invention provides a pipe jacking construction digital twinning optimization system and method based on high-precision physical simulation, and relates to the technical field of underground pipe network pipe jacking construction. Construction data of multiple physical entities during pipe jacking construction are collected in real time through a physical simulation engine module and then transmitted to an established digital twin model, virtual mapping synchronously updated with the construction data of the physical entities is established, and an optimization and decision module optimizes construction parameters by adopting a joint simulation algorithm fusing finite element analysis (FEA) and computational fluid dynamics (CFD), so that the construction parameters are optimized. The construction feedback module realizes automatic parameter adjustment based on PLC control according to optimized construction parameters, the analysis report module analyzes time sequence data and generates a construction efficiency evaluation report, a stress concentration area is accurately simulated, the simulation precision is improved, high synchronization of a virtual model and physics is realized, and the precision and real-time performance of system execution are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of underground pipe jacking construction, and in particular to a digital twin optimization system and method for pipe jacking construction based on high-precision physical simulation. Background Art

[0002] Pipe jacking is a trenchless construction technique widely used in urban underground pipeline construction, particularly in complex environments such as those crossing buildings, roads, and railways. During pipe jacking, the pipeline is advanced underground using mechanical equipment, requiring precise control of construction parameters to ensure quality and safety.

[0003] In existing technologies, traditional pipe jacking construction relies on static geological survey reports and is unable to perceive soil changes during the jacking process in real time, such as sudden changes in moisture content and hard rock interfaces. This leads to a mismatch between jacking parameters and geological conditions, causing pipe displacement or surface subsidence. Existing simulations often use a single mechanical model, such as FEA or DEM alone, ignoring the multi-field coupling effects of mud fluid, soil, and pipes, resulting in excessive errors in predicting jacking force fluctuations caused by tool wear. Monitoring data such as stress, posture, and earth pressure are disconnected from the simulation system, and manual import results in significant data delays, making it impossible to support real-time decision-making. Furthermore, passive early warnings rely on threshold alarms and lack the ability to predict early risks based on time series characteristics, such as the rate of change of the jacking force gradient. Construction parameter adjustments rely on engineer experience, and multi-objective optimization lacks a quantitative model, resulting in long adjustment response times.

[0004] The present invention takes into account the influence of various geological conditions and construction parameters by establishing accurate geological models and construction equipment models. In terms of real-time data acquisition and processing, data during the construction process is collected in real time through sensors and monitoring equipment. In terms of digital twin simulation, real-time data is input into the digital twin model for high-precision simulation to predict changes and risks during the construction process. In terms of intelligent decision support, based on the simulation results, the system provides intelligent decision support, such as adjusting propulsion parameters, warning of potential risks, optimizing construction paths, etc. Through the above technical means, the present invention realizes the comprehensive simulation and optimization of the pipe jacking construction process, ensures the efficiency, safety and high quality of construction, and provides scientific decision support for pipe jacking construction. Summary of the Invention

[0005] The main purpose of the present invention is to provide a digital twin optimization system and method for pipe jacking construction based on high-precision physical simulation to solve at least one of the above technical problems.

[0006] To solve the above technical problems, the technical solution adopted by the present invention is: a digital twin optimization system for pipe jacking construction based on high-precision physical simulation, comprising: The physical simulation engine module is used to collect construction data of various physical entities in pipe jacking construction using an adaptive simulation data network, perform data preprocessing, and dynamically adjust the simulation grid based on real-time data and stress concentration. Construction data includes equipment status data and geological data. Equipment status data includes jacking status and pipe posture, and geological data includes geotechnical parameters. The digital twin model construction module is used to build a digital twin model, integrate pre-processed data in real time according to the Internet of Things protocol, and establish a virtual mapping that is updated synchronously with the physical entity; The optimization and decision-making module is used to output the digital twin model and optimize construction parameters using a joint simulation algorithm that integrates finite element analysis (FEA) and computational fluid dynamics (CFD); The construction feedback module is used to optimize construction parameters, realize automatic parameter adjustment based on PLC control, provide interactive operation guidance for abnormal working conditions, and obtain feedback information; The analysis report module is used to analyze the time series data collected by the physical simulation engine module through the LSTM neural network and generate a construction efficiency evaluation report.

[0007] In the preferred solution, the digital twin model includes: The physical component virtual mapping unit performs 3D parametric modeling of relevant simulation parameters of the pipe jacking machine, pipeline, and geotechnical soil. The simulation parameters of the pipe jacking machine include operating parameters such as tool wear, direction, and speed; the simulation parameters of the pipeline include deformation, stress, and strain properties; and the simulation parameters of the geotechnical soil include rock type, viscosity, water content, and its response to external forces. The digital twin model unit is used to build a digital geometric model of a physical entity using 3D modeling tools, assign material properties, mechanical parameters, and dynamic behavior rules, and integrate kinematics and dynamics algorithms to simulate the device's trajectory and mechanical movements. Dynamic synchronization unit, used to implement two-way data interaction between physical entities and virtual models using the MQTT protocol, with a synchronization period of ≤ 1 second; The abnormal calibration unit is used to trigger the automatic calibration process when the deviation between the virtual model and the physical entity is greater than the preset deviation value.

[0008] In a preferred embodiment, the data preprocessing includes: Data cleaning and noise filtering: A sliding time window is used to detect the amount of sensor data mutation. When a single point data deviates from the window mean by ±3 σ When , it is marked as an outlier and completed by linear interpolation; Multi-dimensional data analysis: A cylindrical coordinate system based on the jacking axis is constructed to map pipeline stress, rock and soil pore water pressure, and jacking velocity to a unified space-time grid. The interaction between soil particle displacement and pipeline stress is calculated using the finite element-discrete element coupling algorithm (FEM-DEM). Data fusion: A hierarchical weighted strategy is adopted to weightedly splice mechanical parameters, geological parameters and equipment status parameters through the attention mechanism.

[0009] In a preferred embodiment, the physical simulation engine module includes: Adaptive mesh division unit, dynamically adjusts simulation mesh density according to real-time stress data, and the mesh size in stress concentration area is ≤5mm; Multi-field coupling calculation unit, synchronously processing the mechanical, fluid and deformation field coupling of the soil-pipeline-equipment system; Real-time data interface unit, which receives sensor data with a delay of ≤100ms and a sampling frequency of ≥10Hz through the 5G network.

[0010] In the preferred solution, the optimization and decision module includes The multi-objective optimization unit is used to optimize the construction parameters based on the objective function. The objective function formula is: ; Where, W 1 ,W 2 ,W 3 is the weight coefficient, which indicates the relative importance of each factor. The facility efficiency is E , construction quality is Q , the construction risk is R ; The risk prediction unit is used to predict excessive pipeline deformation and insufficient rock and soil bearing capacity through the FEA module, with an advance warning time of ≥10 minutes; The visualization decision-making unit is used to provide a three-dimensional interactive interface including real-time jacking force thermal map and rock and soil displacement vector field.

[0011] In the preferred solution, the construction quality Q in the multi-objective optimization unit adopts a quadratic function model, and the formula is: ; Where, u is the measured value of mud viscosity, k 2 is the coefficient, u 0 is the mud viscosity value that makes the quality the best. Q 0 is the corresponding optimal quality value; The construction risk is related to the tool configuration. The construction risk is linearly related to the degree of tool wear. The formula is: R=k 3 d ; Where k3 is the proportional coefficient, d The degree of tool wear.

[0012] In a preferred embodiment, the construction feedback module includes: Closed-loop control unit transmits optimized parameters to the pipe jacking machine hydraulic system via OPC UA protocol, with an adjustment accuracy of ±0.5% FS; The graded warning unit has three levels of sound and light alarm thresholds: a sudden increase of 30% in jacking force and a pipeline offset ≥ 80% of the design value; Operation traceability unit records manual intervention events and updates optimization model constraints.

[0013] In the preferred embodiment, the analysis and reporting module adopts an LSTM neural network, and realizes long-term dependency modeling through three gating mechanisms of forget gate, input gate and output gate and a cell state, specifically: The forget gate is used to filter sensor noise. The formula is: ; Where h t-1 is the hidden state at the previous moment, x t is the current input, is the bias term, which adjusts the noise filtering threshold. is the forget gate weight matrix (solidified after training), which is used to identify noise features (such as sensor drift and vibration interference). Information that needs to be forgotten to determine the cell state at the previous moment (such as sensor noise and short-term fluctuations); The input gate is used to capture the sudden change characteristics of geotechnical parameters. The formula is: ; Where i t To select new information needs to be stored in the cell state; For candidate cell states, generate the potential value of new information through tanh; To adjust the model's sensitivity to sudden changes and decide whether to incorporate such changes into long-term memory, is the coupling relationship between the constrained slurry flow rate and soil permeability; Cell state update, combined with the forget gate and input gate to update the cell state, is used to retain long-term key information while incorporating new features. The formula is: ; Where, is element-wise multiplication (Hadamard product); The output gate is used to output the predicted value of the jacking force. The formula is: ; Where, h t For the current moment prediction, b 0 means the model dynamically adjusts the degree of attention paid to key parameters (such as cutterhead torque and ground settlement).

[0014] A digital twin optimization method for pipe jacking construction based on high-precision physical simulation, applicable to the digital twin optimization system for pipe jacking construction based on high-precision physical simulation, comprises the following steps: Deploy a multi-source sensor network to collect jacking force, pipeline posture and geotechnical parameters in real time, with a data sampling interval of ≤0.1 seconds; Construct a layered geological model and define the mechanical parameters of elastic modulus, permeability coefficient and internal friction angle of each soil layer; The digital twin model is used to dynamically simulate the flow of mud annular space and the Herschel-Bulkley fluid model is used to calculate the grouting pressure distribution; The deviation between the simulation and actual data was checked after each meter of jacking was completed, and the LSTM model was retrained when the mean relative error (MRE) was greater than 8%.

[0015] In the preferred embodiment, the dynamic simulation step includes: The pipeline is discretized into an unstructured finite element mesh, and the elastic modulus of the steel is set to 210 GPa and the Poisson's ratio to 0.3; Nonlinear boundary conditions are applied at the cutterhead-soil interface to simulate the friction changes caused by tool wear; The coupled CFD module calculates the inhibitory effect of mud support pressure on formation deformation.

[0016] The present invention provides a digital twin optimization system and method for pipe jacking construction based on high-precision physical simulation. The system collects construction data of various physical entities in real time through a physical simulation engine module, then builds a digital twin model, and establishes a virtual mapping that is updated synchronously with the physical entity according to the Internet of Things protocol. The optimization and decision-making module optimizes construction parameters using a joint simulation algorithm that integrates finite element analysis (FEA) and computational fluid dynamics (CFD). The construction feedback module automatically adjusts parameters based on PLC control according to the optimized construction parameters. The analysis report module analyzes the time series data collected by the physical simulation engine module and generates a construction efficiency evaluation report, accurately simulating stress concentration areas, improving simulation accuracy, achieving a high degree of synchronization between the virtual model and the physical world, and improving the accuracy and real-time performance of system execution. The LSTM neural network is used to analyze the time series data collected by the physical simulation engine module, improving the accuracy of capturing long-term dependencies and sudden changes in geotechnical parameters during the construction process, providing an accurate basis for construction efficiency evaluation, ensuring that the construction reaches the optimal quality state, and predicting and controlling construction process risks in advance, thereby achieving early warning and effective control of risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The present invention will be further described below with reference to the accompanying drawings and examples: Figure 1 It is a schematic diagram of the system architecture of the present invention; Figure 2 It is a schematic diagram of the simulation system of the present invention; Figure 3 This is a physical schematic diagram of the simulation system of the present invention; Figure 4 It is a schematic diagram of the concept of the simulation system of the present invention; Figure 5 is a schematic diagram of the digital twin model of the present invention; Figure 6 This is a real-time monitoring interface diagram of the present invention; Figure 7 It is an operational effect diagram of the present invention.

[0018] In the figure: 1. Pressure sensor, 2. Displacement sensor, 3. Temperature sensor. DETAILED DESCRIPTION

[0019] Example 1 like Figure 1-7 As shown in the figure, a digital twin optimization system for pipe jacking construction based on high-precision physical simulation includes: The physical simulation engine module is used to utilize an adaptive simulation data network to collect construction data of various physical entities in pipe jacking construction, perform data preprocessing, and then dynamically adjust the simulation grid based on real-time data and stress concentration. Construction data includes equipment status data and geological data. Equipment status data includes jacking status and pipe posture, and geological data includes geotechnical parameters.

[0020] The digital twin model construction module is used to build a digital twin model, integrate pre-processed data in real time according to the Internet of Things protocol, and establish a virtual mapping that is updated synchronously with the physical entity; The optimization and decision-making module is used to output the digital twin model and optimize construction parameters using a joint simulation algorithm that integrates finite element analysis (FEA) and computational fluid dynamics (CFD).

[0021] The construction feedback module is used to optimize construction parameters and realize automatic parameter adjustment based on PLC control, interactive operation guidance for abnormal working conditions, and obtain feedback information.

[0022] The analysis report module is used to analyze the time series data collected by the physical simulation engine module through the LSTM neural network and generate a construction efficiency evaluation report.

[0023] This embodiment uses the physical simulation engine module to collect construction data from various physical entities in real time, then builds a digital twin model. Based on the Internet of Things protocol, a virtual mapping is established that is updated synchronously with the physical entity. The optimization and decision-making module uses a joint simulation algorithm that integrates finite element analysis (FEA) and computational fluid dynamics (CFD) to optimize construction parameters. The construction feedback module automatically adjusts parameters based on the optimized construction parameters using PLC control. The analysis and reporting module analyzes the time series data collected by the physical simulation engine module and generates a construction efficiency evaluation report, accurately simulating stress concentration areas, improving simulation accuracy, achieving a high degree of synchronization between the virtual model and the physical world, and enhancing the accuracy and real-time performance of system execution. The LSTM neural network is used to analyze the time series data collected by the physical simulation engine module, improving the accuracy of capturing long-term dependencies and geotechnical parameter mutation characteristics during the construction process. This provides an accurate basis for construction efficiency evaluation, ensures that construction achieves optimal quality, and predicts and controls construction process risks in advance, achieving early warning and effective control of risks.

[0024] In the preferred solution, the virtual mapping unit of the physical components performs three-dimensional parametric modeling of the relevant simulation parameters of the pipe jacking machine, pipeline, and rock and soil. The simulation parameters of the pipe jacking machine include operating parameters such as tool wear, direction, and speed; the simulation parameters of the pipeline include deformation, stress, and strain properties; the simulation parameters of the rock and soil take into account the rock and soil type, viscosity, water content, and its response to external forces; The digital twin model unit is used to establish a digital geometric model of a physical entity through 3D modeling tools, and to assign material properties, mechanical parameters, and dynamic behavior rules, integrating kinematics and dynamics algorithms to simulate the equipment's operating trajectory and mechanical movements.

[0025] Dynamic synchronization unit, used to implement two-way data interaction between physical entities and virtual models using the MQTT protocol, with a synchronization period of ≤1 second.

[0026] The abnormal calibration unit is used to trigger the automatic calibration process when the deviation between the virtual model and the physical entity is greater than the preset deviation value.

[0027] In this embodiment, Figure 4-5 As shown in the figure, the digital twin model unit performs digital simulation on related modeling, including: geometric shape definition, physical property definition, action simulation, algorithm embedding, animation display, parameter adjustment, historical data and analysis module and network interconnection.

[0028] like Figure 2As shown in the figure, the construction and operation process of the digital twin model is as follows: First, the geometric shape of physical entities (such as pipe jacking machines and pipelines) is defined using 3D modeling tools, accurately recreating their size, structure, and spatial layout, and assigning material properties (such as density and stiffness) and mechanical parameters (such as friction coefficient and stress threshold). Then, kinematic and dynamic algorithms are integrated to simulate the equipment's operating trajectory (such as cutterhead rotation) and the dynamic behavior rules of mechanical movements. PID control algorithms and LSTM prediction models are also embedded to support real-time decision-making. A visualization engine is then used to render the equipment's dynamic operating status, overlaying multi-dimensional parameters such as stress cloud maps and flow velocity vectors to support multi-perspective observation. Based on real-time sensor data (such as jacking speed and cutterhead torque) and prediction model results, the system dynamically adjusts construction parameters (such as reducing jacking speed and increasing grouting pressure). If the risk exceeds the threshold, the system automatically intervenes or issues an alarm. Meanwhile, the historical data module continuously trains the model using an LSTM neural network to optimize prediction accuracy. Finally, sensor data is synchronized to the cloud with millisecond latency via the 5G network, forming a closed-loop link of "physical equipment-virtual model-real-time decision-making-inverse control" to ensure a dynamic balance between construction efficiency and safety.

[0029] In this embodiment, the virtual mapped data of the physical components is imported from the cloud into the simulation model module to generate a specific model for real-time simulation, and seamless communication between the virtual model and the physical entity is ensured through data exchange.

[0030] In the preferred embodiment, data preprocessing includes: Data cleaning and noise filtering: A sliding time window is used to detect the sudden change of sensor data. When a single point of data deviates from the window mean by ±3σ, it is marked as an outlier and supplemented by linear interpolation.

[0031] Multi-dimensional data analysis: A cylindrical coordinate system based on the jacking axis is constructed to map the pipeline stress, geotechnical pore water pressure, and jacking velocity to a unified space-time grid. The interaction between soil particle displacement and pipeline stress is calculated using the finite element-discrete element coupling algorithm (FEM-DEM).

[0032] Data fusion: A hierarchical weighted strategy is adopted to weightedly splice mechanical parameters, geological parameters and equipment status parameters through the attention mechanism.

[0033] In this embodiment, each step is specifically implemented as follows: 1. Data cleaning and noise filtering: A sliding time window (window length ≤ 5 seconds) is used to detect sudden changes in sensor data. When a single point of data deviates from the window mean by ±3σ, it is marked as an outlier and supplemented using linear interpolation. The data of the mechanical sensor, attitude meter, and geotechnical probe are time-stamped and then dynamically filtered for noise. For the jacking force signal, wavelet transform is used to separate high-frequency noise components. For the geotechnical moisture data, Kalman filtering is applied to eliminate the influence of ambient temperature drift. Specifically: Step A1: Sliding window anomaly detection and interpolation, including: Window definition: Assume that the time window length T = 5s, the sampling frequency f = 10Hz, and the number of data points in the window N = T × f = 50.

[0034] Mean and standard deviation calculation: ; ; Abnormal judgment: If the data point x t satisfy , marked as abnormal.

[0035] Linear interpolation: for outlier points x t , use the interpolation of the valid points before and after to complete: .

[0036] Step A2: Multi-sensor timestamp synchronization, including: Alignment reference: The clock of the jacking axis motion control system is used as the reference.

[0037] Interpolation alignment: For asynchronous data (such as the sampling frequency of the geotechnical moisture meter is 5Hz), cubic spline interpolation is used to 10Hz.

[0038] Step A3: Dynamic noise filtering, including: Jacking force signal (wavelet transform): Daubechies 4 wavelet basis is selected, the number of decomposition layers is 5, and the high-frequency coefficients are reconstructed after threshold processing. The formula is as follows: ; In the formula λ : Soft threshold, take 0.5×max(DWT(x)) .

[0039] Geosoil moisture (Kalman filter), the state equation is as follows: ; ( h t : True humidity, w t ~N(0,Q),v t ~N(0, R), set up Q=0.01, R=0.1 ).

[0040] 2. Multi-dimensional Data Analysis: A cylindrical coordinate system based on the jacking axis was constructed to map pipeline stress, geotechnical pore water pressure, and jacking velocity to a unified spatiotemporal grid. The interaction between soil particle displacement and pipeline stress was then calculated using the finite element-discrete element method (FEM-DEM). Finally, the DBSCAN clustering algorithm was used to analyze the jacking force-torque combination characteristics in historical construction data, identify areas of abnormal density, and establish a risk feature library. Specifically, Step A1: Cylindrical coordinate space-time grid mapping, including: Coordinate system definition: Take the jacking axis as Z Axis, radial R , circumferential direction is θ, Grid resolution Δr=0.1m, Δr= 0.1m .

[0041] Parameter Mapping: Pipeline Stress σ(r,θ) , pore water pressure p(r,θ) Interpolate to mesh nodes.

[0042] Step A2: FEM-DEM coupling algorithm, including: FEM solution for pipeline stress: equation: [K]{u}={F} , where the stiffness matrix K is generated by the material properties (elastic modulus E = 200 GPa).

[0043] DEM simulation of soil particles, Newton-Euler equation: ; Coupled interaction: Transfer the pipeline displacement to the soil particle boundary conditions through the penalty function method.

[0044] Step A3: DBSCAN cluster analysis, including: Parameter: Neighborhood radius , the minimum number of points minPts=5 .

[0045] Feature extraction: Extract the combination of jacking force F and torque T from historical data (F i ,T i ) .

[0046] Density calculation: Calculate the number of points in the neighborhood of each point, and mark core points, boundary points, and noise points.

[0047] Clustering output: Identify high-density areas (normal operating conditions) and low-density areas (abnormal operating conditions).

[0048] 3. Data fusion: Figure 6 The real-time monitoring interface is shown in Figure 2. This embodiment employs a layered weighted strategy. First, underlying sensor-level fusion is performed, i.e., multi-sensor data from the same monitoring point is integrated. Cross-dimensional feature fusion is then performed, combining mechanical parameters (jacking force, torque), geological parameters (rock and soil internal friction angle, moisture content), and equipment status (tool wear, oil pressure) in a weighted manner through an attention mechanism. Finally, virtual and real data are calibrated to ensure that the relative error between the simulated prediction value and the actual monitoring value is within a preset error range. Specifically, Step A1: Adopt a hierarchical weighting strategy, including: Sensor-level fusion: weighted average of temperature and vibration sensor data at the same monitoring point: .

[0049] Cross-dimensional feature fusion: weights are calculated through the attention mechanism: ; In the formula ，h i is the eigenvector, W∈R d×d ,v∈R d is a trainable parameter.

[0050] Step A2: Calibrate virtual and real data, including: Formula: Dynamically adjust simulation parameters θ Minimize the relative error: ; Where, st .Error ≤ 5%.

[0051] Furthermore, it also includes real-time data quality assessment, specifically: Step A1: Quantify the credibility and define the data quality index as: DQ = λ1·(1-packet loss rate) + λ2·(signal-to-noise ratio / 30dB) + λ3·(time synchronization accuracy); Among them, λ1+λ2+λ3=1.

[0052] A2: Hierarchical processing mechanism: When DQI < 0.7, the data reconstruction process is triggered: For mechanical parameters: Enable LSTM predicted values instead of raw data.

[0053] For geological parameters: switch to interpolation data of adjacent monitoring points.

[0054] For device status: call the latest simulation results of the digital twin model.

[0055] In the preferred solution, the physical simulation engine module includes: Adaptive mesh division unit dynamically adjusts simulation mesh density according to real-time stress data, and the mesh size in stress concentration area is ≤5mm.

[0056] The multi-field coupling calculation unit simultaneously processes the mechanical, fluid and deformation field coupling of the soil-pipeline-equipment system.

[0057] Real-time data interface unit, which receives sensor data with a delay of ≤100ms and a sampling frequency of ≥10Hz through the 5G network.

[0058] The real-time data interface unit receives and pre-processes sensor data through the 5G network. The adaptive mesh division unit dynamically encrypts the local mesh based on real-time stress. Then the multi-field coupling calculation unit synchronously solves the mechanical-fluid-deformation field interaction, and finally updates the simulation parameters through closed-loop feedback, forming a dynamic collaborative link of "data-driven mesh optimization → multi-field coupling solution → reverse synchronization of results".

[0059] In this embodiment, an intelligent algorithm module based on a digital twin model is used to optimize construction parameters during pipe jacking construction.

[0060] The prediction model is based on FEA (Finite Element Analysis Engine ） The module discretizes complex structures such as pipelines, pipe jacking machines, and rock and soil into a finite number of tiny units. By calculating the stress and strain distribution of each unit, it predicts the mechanical behavior of the entire structure (such as pipeline deformation, insufficient rock and soil bearing capacity, etc.), current trends and historical data, and predicts future conditions. The prediction model is a linear regression model with the formula: ; Where: For time t Predicted values (such as pipe jacking status, rock and soil bearing capacity, pipeline location, etc.); x 1 ,x 2 …x n For time t Input variables (such as current construction efficiency, construction quality, construction risks, etc.); β 0 ，β 1 ，⋯，β n is the regression coefficient of the model; is the error term of the model, which represents the deviation between the predicted value and the actual value.

[0061] As shown in Table 1, based on the output of the digital twin model, the FEA module, current trends, and historical data are used to predict future conditions and anticipate and mitigate issues. Subsequently, the CFD (computational fluid dynamics) module, using fluid dynamics principles, simulates the flow and pressure distribution of fluids such as slurry and groundwater during construction. It analyzes the impact of fluid resistance on pipeline advancement and the pore water pressure in the rock and soil, and recommends adjusting relevant parameters to balance jacking speed, slurry viscosity, and tool wear. These parameters can be adjusted automatically or provided to the operator to maintain efficiency and safety.

[0062] Table 1 FEA module data table output

[0063] CFD module data table output

[0064] In the preferred solution, the optimization and decision module includes The multi-objective optimization unit is used to optimize the construction parameters based on the objective function. The objective function formula is: ; Where, W 1 ,W 2 ,W 3 is the weight coefficient, which indicates the relative importance of each factor. The facility efficiency is E , construction quality is Q , the construction risk is R .

[0065] The risk prediction unit is used to predict excessive pipeline deformation and insufficient rock and soil bearing capacity through the FEA module, with an advance warning time of ≥10 minutes.

[0066] The visualization decision-making unit is used to provide a three-dimensional interactive interface including real-time jacking force thermal map and rock and soil displacement vector field.

[0067] In this embodiment, real-time data visualization displays the construction status through dashboards, charts, and 3D models, and predictive analysis visualization uses charts and animations to display prediction results and trend analysis. The support provided by intelligent algorithms includes parameter adjustment suggestions and decision support systems.

[0068] In the preferred solution, the construction quality Q in the multi-objective optimization unit adopts a quadratic function model, and the formula is: ; in, u is the measured value of mud viscosity, k 2 is the coefficient, u 0 is the mud viscosity value that makes the quality the best. Q 0 is the corresponding optimal quality value.

[0069] The construction risk is related to the tool configuration. The construction risk is linearly related to the degree of tool wear. The formula is: R=k 3 d ; Among them, k3 is the proportional coefficient, d The degree of tool wear.

[0070] In this embodiment, the construction quality is related to multiple factors, including mud parameters (such as mud viscosity u As an example, assuming that the mud viscosity is within a certain range ( u min ,u max ), the construction quality and mud viscosity are in a quadratic function relationship.

[0071] In this embodiment, corresponding constraints are set according to the objective function, including: jacking speed constraint, mud parameter constraint and tool configuration constraint, among which: There are certain restrictions on the jacking speed: v min ≤v≤v max , where v min and v max are the minimum and maximum jacking speeds respectively.

[0072] Mud viscosity must meet the following requirements: u min ≤u≤u max .

[0073] Tool configuration constraints: tool wear d There is a reasonable range: d min ≤d≤d max , and tool configuration may also be limited by other factors, such as the number of tools n Integer constraints n∈N is the set of natural numbers.

[0074] .

[0075] In this embodiment, by providing real-time data synchronization and predictive analysis based on the simulation model, operational efficiency is improved, potential problems are predicted, and safety is ensured. At the same time, by analyzing real-time and historical data, machine learning algorithms are used ( LSTM) trains on historical construction data and combines it with real-time simulation results to continuously optimize prediction accuracy. By addressing challenges such as accuracy, real-time processing, and scalability, digital twins offer significant advantages in cost savings and performance optimization. By integrating collected data into a digital twin model (simulation model), enhanced with high-precision physical simulation, engineers can monitor construction in real time and provide decision support based on the following optimization algorithms.

[0076] In the preferred solution, the construction feedback module includes: Closed-loop control unit transmits optimized parameters to the pipe jacking machine hydraulic system via OPC UA protocol, with an adjustment accuracy of ±0.5% FS; The graded warning unit has three levels of sound and light alarm thresholds: a sudden increase of 30% in jacking force and a pipeline offset ≥ 80% of the design value; Operation traceability unit records manual intervention events and updates optimization model constraints.

[0077] In this embodiment, the construction feedback module, serving as the system's output, is responsible for translating optimization and decision-making results into actual construction operations during pipe jacking. By integrating automated equipment and intelligent control systems, it enables automatic or manual adjustment of construction parameters and visualized operational guidance. This real-time feedback mechanism receives field data. This real-time monitoring and feedback mechanism facilitates timely adjustment and optimization of construction strategies, improving both efficiency and safety.

[0078] Through real-time data feedback, model parameters can be continuously updated to improve prediction and optimization capabilities. The model parameter update formula is: ; in: It is t Model parameters for the iteration; is the learning rate; is the objective function J Parameters β i gradient.

[0079] LSTM (Long Short-Term Memory) networks, through their gating mechanisms and cell states, address the vanishing / exploding gradient issues of traditional RNNs in long sequences. They are particularly well-suited for multi-parameter, time-dependent prediction tasks in pipe jacking construction. The weight matrix and bias terms in the formula are trained using historical construction data, ultimately achieving the goal of dynamically optimizing construction parameters.

[0080] In the preferred solution, the analysis and reporting module uses an LSTM neural network to implement long-term dependency modeling through three gating mechanisms: forget gate, input gate, and output gate, as well as a cell state. Specifically: The forget gate is used to filter sensor noise. The formula is: ; Where h t-1 is the hidden state at the previous moment, x t is the current input, is the bias term, which adjusts the noise filtering threshold. is the forget gate weight matrix (solidified after training), which is used to identify noise features (such as sensor drift and vibration interference). To determine the information that needs to be forgotten in the cell state at the previous moment (such as sensor noise, short-term fluctuations).

[0081] The input gate is used to capture the sudden change characteristics of geotechnical parameters. The formula is: ; Where i t To select new information needs to be stored in the cell state; For candidate cell states, generate the potential value of new information through tanh; To adjust the model's sensitivity to sudden changes and decide whether to incorporate such changes into long-term memory, It is the coupling relationship between the constrained slurry flow rate and soil permeability.

[0082] Cell state update, combined with the forget gate and input gate to update the cell state, is used to retain long-term key information while incorporating new features. The formula is: ; Where, is element-wise multiplication (Hadamard product); The output gate is used to output the predicted value of the jacking force. The formula is: ; Where h t For the current moment prediction, To dynamically adjust the model's focus on key parameters (such as cutterhead torque and ground settlement).

[0083] During pipe jacking construction, LSTM The output layer is usually a fully connected layer (Dense Layer) , will hide the state h t Mapping to target value (such as jacking force): ; in, : Predicted jacking force (or other construction parameter).

[0084] In this embodiment LSTM After model training, the training goal is to minimize the mean square error (MSE) between the predicted value and the true value: ; The corresponding relationship between the formula and engineering application is: Cell state C t : Long-term memory of pipeline posture and cumulative effects of geotechnical parameters (such as changes in friction caused by continuous jacking).

[0085] Forget Gate f t : Ignore short-term interference (such as sensor noise) and retain historical data of key construction stages.

[0086] Input Gate i t : Dynamically learn unexpected events (such as sudden changes in jacking force when encountering hard rock layers).

[0087] Output Gate o t : Generate real-time prediction results and trigger early warnings or parameter adjustments (such as excessive jacking force alarms).

[0088] In this embodiment, the parameter dimension is as follows: Assume that the input feature number is 3 (jacking force, inclination angle, moisture content), LSTM The number of hidden units is 64: The weight matrix dimensions are: ; Where, input dimension: 3 features + 64-dimensional hidden state → 64-dimensional gated output.

[0089] The bias dimension is: .

[0090] The analysis report module optimizes the digital twin model through machine learning algorithms (LSTM), comprehensively analyzes construction technology, equipment and operations, evaluates quality, efficiency and cost, provides data support for intelligent optimization and scientific decision-making, and improves construction safety and efficiency.

[0091] Figure 7 shows the operational effect of this embodiment. After optimizing the relevant parameters according to the optimization and decision-making module, the relevant operating methods are provided. The specific mechanism is as follows: The alarm system prioritizes safety and risk management and notifies operators of potential dangers. Thresholds are set for key parameters. When exceeded, the system will trigger an alarm or adjust the pipe jacking process to prevent accidents. When performing relevant operations, feedback information from the site is continuously received according to the algorithm provided by the feedback mechanism. The real-time monitoring and feedback mechanism helps to timely adjust and optimize construction strategies, improving construction efficiency and safety.

[0092] Example 2 Further described in conjunction with Example 1, a digital twin optimization method for pipe jacking construction based on high-precision physical simulation is applicable to a digital twin optimization system for pipe jacking construction based on high-precision physical simulation in Example 1, comprising the following steps: Deploy a multi-source sensor network to collect jacking force, pipeline posture and geotechnical parameters in real time, with a data sampling interval of ≤0.1 seconds; Construct a layered geological model and define the mechanical parameters of elastic modulus, permeability coefficient and internal friction angle of each soil layer; The digital twin model is used to dynamically simulate the flow of mud annular space and the Herschel-Bulkley fluid model is used to calculate the grouting pressure distribution; The deviation between the simulation and actual data was checked after each meter of jacking was completed, and the LSTM model was retrained when MRE>8%.

[0093] In the preferred embodiment, the dynamic simulation step includes: The pipeline is discretized into an unstructured finite element mesh, and the elastic modulus of the steel is set to 210 GPa and the Poisson's ratio to 0.3; Nonlinear boundary conditions are applied at the cutterhead-soil interface to simulate the friction changes caused by tool wear; The coupled CFD module calculates the inhibitory effect of mud support pressure on formation deformation.

[0094] In this embodiment, Figure 3 As shown in the figure, the sensor network deployment involves installing multiple types of sensors at key locations on the pipe jacking machine (such as the thrust cylinder and cutterhead), along pipe interfaces, and in the surrounding rock and soil. These sensors include: force sensors for real-time monitoring of mechanical parameters such as jacking force and torque; attitude sensors for collecting spatial attitude data such as pipe inclination and azimuth; and geotechnical parameter sensors for measuring rock and soil moisture, density, and pore water pressure. Sensor data is transmitted in real time to a data processing center via a 5G network, ensuring a transmission delay of less than 100ms and a sampling frequency of at least 10Hz.

[0095] This embodiment is used in the pipe jacking construction process, and the specific implementation steps are as follows: Step 1: Sensor selection and installation. Mechanical parameter monitoring: Dynamic force sensors (range 0-5000kN, accuracy ±0.5%) were installed on the pipe jacking machine's propulsion cylinder and cutterhead spindle to monitor jacking force and torque changes. Attitude monitoring: A dual-axis inclinometer (resolution 0.01°) was affixed to the inner wall of each pipe section to collect real-time pipe inclination and azimuth data. Geotechnical parameter collection: Geotechnical moisture sensors (range 0-100%) and pore water pressure gauges (range 0-2MPa) were buried every 5 meters along the jacking axis. Soil density was measured using a micro-probe densitometer.

[0096] Step 2: Data transmission and preprocessing, using 5GIndustrial gateway, which encapsulates sensor data into JSON format, through MQTT The protocol is uploaded to the data processing center in real time and saved in the cloud server. The transmission frequency is 10Hz , latency ≤ 50ms. Remove outliers (e.g., data exceeding three times the range) and generate standardized data packets.

[0097] Step 3: 3D modeling and data mapping. Use the system to build models of pipe jacking machine, pipeline and layered geological related models. Define the following attributes: pipe jacking machine: hydraulic system pressure threshold, cutter head speed safety range; pipeline: single section length (2m), maximum allowable bending angle ( 0.5° Geotechnical data: internal friction angle and permeability coefficient of each soil layer. Import data into the simulation model through the API interface, achieving dynamic binding with physical simulation data.

[0098] Step 4: Multi-objective parameter optimization, call FEA (Finite Element Analysis) module: Import the pipeline 3D model (STEP format), set the steel elastic modulus ( 210GPa ), Poisson's ratio ( 0.3 ) and other properties, divide the unstructured grid (stress concentration area grid ≤5mm ), with the goal of minimizing jacking resistance, pipe deviation and mud consumption, the optimal jacking speed and grouting pressure are iteratively calculated.

[0099] Step 5: Calculation CFD (Fluid Mechanics) Configuration, defining the mud as a non-Newtonian fluid, using Herschel- Bulkley The model sets rheological parameters to simulate the flow of mud in annular space (inlet velocity 0.5-2.0m / s, outlet constant pressure).

[0100] Step 6: Dynamic simulation and prediction: receive data from the data processing center every 10 seconds, update the simulation input conditions (jacking force, rock and soil moisture), and use FEA and CFD The module calculates and outputs relevant data, including mud pressure distribution and predicted jacking trajectory. If the predicted pipeline deviation exceeds the design threshold (±30mm), an early warning signal is triggered and sent to the operator console.

[0101] Step 7: Real-time synchronization and calibration: Use a time series database (InfluxDB) to store sensor data and simulation results, with timestamp alignment accuracy ≤ 1ms. Perform a data consistency check every 30 seconds. If the virtual model deviates from the physical entity by more than 5%, an automatic calibration process is triggered.

[0102] Step 8: Automated Control and Manual Intervention: OPC UA The protocol sends the optimized parameters to the pipe jacking machine PLCThe controller automatically adjusts the hydraulic system proportional valve opening. If manual operator intervention (such as pausing jacking) is detected, the cause of the deviation is recorded and the optimization model constraints are updated.

[0103] Step 9: Real-time data closed-loop verification: After each meter of jacking is completed, compare the actual data with the simulation prediction value and calculate the average relative error (MRE) .like MRE>8% , triggering the simulation model of the digital twin system and recalculating.

[0104] Step 10: Automated report generation: Use Jupyter Notebook scripts to generate PDF reports daily. The content covers: key indicators such as average jacking speed, maximum deviation value, and mud consumption; abnormal events such as the number of warning triggers and the effectiveness of treatment measures.

[0105] The above embodiments are merely preferred technical solutions of the present invention and should not be construed as limiting the present invention. The scope of protection of the present invention shall be the technical solutions set forth in the claims, including equivalent alternatives to the technical features of the technical solutions set forth in the claims. In other words, equivalent alternatives and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A digital twin optimization system for pipe jacking construction based on high-precision physical simulation, characterized by: include: The physical simulation engine module is used to collect construction data of various physical entities in pipe jacking construction using an adaptive simulation data network, perform data preprocessing, and dynamically adjust the simulation grid based on real-time data and stress concentration. Construction data includes equipment status data and geological data. Equipment status data includes jacking status and pipe posture, and geological data includes geotechnical parameters. The digital twin model construction module is used to build a digital twin model, integrate pre-processed data in real time according to the Internet of Things protocol, and establish a virtual mapping that is updated synchronously with the physical entity; The optimization and decision-making module is used to output the digital twin model and optimize construction parameters using a joint simulation algorithm that integrates finite element analysis (FEA) and computational fluid dynamics (CFD); The construction feedback module is used to optimize construction parameters, realize automatic parameter adjustment based on PLC control, provide interactive operation guidance for abnormal working conditions, and obtain feedback information; The analysis report module is used to analyze the time series data collected by the physical simulation engine module through the LSTM neural network and generate a construction efficiency evaluation report.

2. The digital twin optimization system for pipe jacking construction based on high-precision physical simulation according to claim 1 is characterized in that: The digital twin model includes: The physical component virtual mapping unit performs 3D parametric modeling of relevant simulation parameters of the pipe jacking machine, pipeline, and geotechnical soil. The simulation parameters of the pipe jacking machine include operating parameters such as tool wear, direction, and speed; the simulation parameters of the pipeline include deformation, stress, and strain properties; and the simulation parameters of the geotechnical soil include rock type, viscosity, water content, and its response to external forces. The digital twin model unit is used to build a digital geometric model of a physical entity using 3D modeling tools, assign material properties, mechanical parameters, and dynamic behavior rules, and integrate kinematics and dynamics algorithms to simulate the device's trajectory and mechanical movements. Dynamic synchronization unit, used to implement two-way data interaction between physical entities and virtual models using the MQTT protocol, with a synchronization period of ≤ 1 second; The abnormal calibration unit is used to trigger the automatic calibration process when the deviation between the virtual model and the physical entity is greater than the preset deviation value.

3. The digital twin optimization system for pipe jacking construction based on high-precision physical simulation according to claim 1 is characterized in that: The data preprocessing includes: Data cleaning and noise filtering: A sliding time window is used to detect the amount of sensor data mutation. When a single point data deviates from the window mean by ±3 σ When , it is marked as an outlier and completed by linear interpolation; Multi-dimensional data analysis: A cylindrical coordinate system based on the jacking axis is constructed to map pipeline stress, rock and soil pore water pressure, and jacking velocity to a unified space-time grid. The interaction between soil particle displacement and pipeline stress is calculated using the finite element-discrete element coupling algorithm (FEM-DEM). Data fusion: A hierarchical weighted strategy is adopted to weightedly splice mechanical parameters, geological parameters and equipment status parameters through the attention mechanism.

4. The digital twin optimization system for pipe jacking construction based on high-precision physical simulation according to claim 1 is characterized in that: The physical simulation engine module includes: Adaptive mesh division unit, dynamically adjusts simulation mesh density according to real-time stress data, and the mesh size in stress concentration area is ≤5mm; Multi-field coupling calculation unit, synchronously processing the mechanical, fluid and deformation field coupling of the soil-pipeline-equipment system; Real-time data interface unit, which receives sensor data with a delay of ≤100ms and a sampling frequency of ≥10Hz through the 5G network.

5. The digital twin optimization system for pipe jacking construction based on high-precision physical simulation according to claim 1 is characterized in that: The optimization and decision-making module includes The multi-objective optimization unit is used to optimize the construction parameters based on the objective function. The objective function formula is: ； Where, W 1 ,W 2 ,W 3 is the weight coefficient, which indicates the relative importance of each factor. The facility efficiency is E , construction quality is Q , the construction risk is R ; The risk prediction unit is used to predict excessive pipeline deformation and insufficient rock and soil bearing capacity through the FEA module, with an advance warning time of ≥10 minutes; The visualization decision-making unit is used to provide a three-dimensional interactive interface including real-time jacking force thermal map and rock and soil displacement vector field.

6. The digital twin optimization system for pipe jacking construction based on high-precision physical simulation according to claim 5 is characterized in that: The construction quality Q in the multi-objective optimization unit adopts a quadratic function model, and the formula is: ； Where, u is the measured value of mud viscosity, k 2 is the coefficient, u 0 is the mud viscosity value that makes the quality the best. Q 0 is the corresponding optimal quality value; The construction risk is related to the tool configuration. The construction risk is linearly related to the degree of tool wear. The formula is: R=k 3 d ； Where k3 is the proportional coefficient, d The degree of tool wear.

7. The digital twin optimization system for pipe jacking construction based on high-precision physical simulation according to claim 1 is characterized in that: The construction feedback module includes: Closed-loop control unit transmits optimized parameters to the pipe jacking machine hydraulic system via OPC UA protocol, with an adjustment accuracy of ±0.5% FS; The graded warning unit has three levels of sound and light alarm thresholds: a sudden increase of 30% in jacking force and a pipeline offset ≥ 80% of the design value; Operation traceability unit records manual intervention events and updates optimization model constraints.

8. The digital twin optimization system for pipe jacking construction based on high-precision physical simulation according to claim 1 is characterized in that: The analysis report module uses an LSTM neural network to implement long-term dependency modeling through three gating mechanisms: forget gate, input gate, and output gate, as well as a cell state. Specifically: The forget gate is used to filter sensor noise. The formula is: ； Where h t-1 is the hidden state at the previous moment, x t is the current input, is the bias term, which adjusts the noise filtering threshold. is the forget gate weight matrix (solidified after training), which is used to identify noise features (such as sensor drift and vibration interference). Information that needs to be forgotten to determine the cell state at the previous moment (such as sensor noise and short-term fluctuations); The input gate is used to capture the sudden change characteristics of geotechnical parameters. The formula is: ； Where i t To select new information needs to be stored in the cell state; For candidate cell states, generate the potential value of new information through tanh; To adjust the model's sensitivity to sudden changes and decide whether to incorporate such changes into long-term memory, is the coupling relationship between the constrained slurry flow rate and soil permeability; Cell state update, combined with the forget gate and input gate to update the cell state, is used to retain long-term key information while incorporating new features. The formula is: ； Where, is element-wise multiplication (Hadamard product); The output gate is used to output the predicted value of the jacking force. The formula is: ； Where, h t For the current moment prediction, b 0 means the model dynamically adjusts the degree of attention paid to key parameters (such as cutterhead torque and ground settlement).

9. A digital twin optimization method for pipe jacking construction based on high-precision physical simulation, applicable to the digital twin optimization system for pipe jacking construction based on high-precision physical simulation as described in any one of claims 1 to 8, characterized in that: The following steps are involved: Deploy a multi-source sensor network to collect jacking force, pipeline posture and geotechnical parameters in real time, with a data sampling interval of ≤0.1 seconds; Construct a layered geological model and define the mechanical parameters of elastic modulus, permeability coefficient and internal friction angle of each soil layer; The digital twin model is used to dynamically simulate the flow of mud annular space and the Herschel-Bulkley fluid model is used to calculate the grouting pressure distribution; The deviation between the simulation and actual data was checked after each meter of jacking was completed, and the LSTM model was retrained when the mean relative error (MRE) was greater than 8%.

10. The digital twin optimization method for pipe jacking construction based on high-precision physical simulation according to claim 9 is characterized in that: The dynamic simulation steps include: The pipeline is discretized into an unstructured finite element mesh, and the elastic modulus of the steel is set to 210 GPa and the Poisson's ratio to 0.3; Nonlinear boundary conditions are applied at the cutterhead-soil interface to simulate the friction changes caused by tool wear; The coupled CFD module calculates the inhibitory effect of mud support pressure on formation deformation.

Images