Analysis method and system for wet-mix shotcrete over-consumption in intelligent construction of tunnel primary support
By using 3D data capture and multiphysics analysis, combined with intelligent algorithms to optimize spraying parameters, the problem of excessive consumption of wet sprayed concrete in tunnel construction has been solved. This has enabled dynamic and precise control of material usage and interpretable diagnosis of the causes of excessive consumption, thereby reducing construction costs and material waste.
Patent Information
- Application Number
- PCT/CN2025/111378
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-05-14
- Filing Date
- 2025-07-30
- Publication Date
- 2026-11-19
AI Technical Summary
In tunnel construction, the excessive consumption of wet shotcrete leads to material waste and difficulty in cost control. Existing technologies are unable to accurately guide construction and lack the ability to conduct multi-dimensional correlation analysis of process parameters, material flow state and environmental physical field, making it impossible to effectively optimize measures.
By installing a 3D motion capture system to acquire spray trajectory and material flow fluctuation data, and combining it with a geological BIM model and a dynamic weighing system, a 3D feature tensor is constructed. Convolutional neural network and multilayer perceptron algorithms are used to generate an over-consumption pattern map. Combined with multiphysics coupling analysis and reinforcement learning algorithms, a closed-loop control loop is formed to dynamically adjust the spray parameters.
It enables dynamic adaptation to geological conditions and interpretable diagnosis of multi-factor coupled overconsumption mode in wet spraying construction, dynamically and accurately controls material usage, and reduces overall rebound rate and material waste.
Smart Images

Figure CN2025111378_19112026_PF_FP_ABST
Abstract
Description
A method and system for analyzing excessive wear in wet spraying during intelligent tunnel initial support construction. Technical Field
[0001] This application relates to the field of tunnel construction technology, and in particular to a method and system for analyzing excessive consumption of wet spraying in intelligent tunnel initial support construction. Background Technology
[0002] In underground engineering, tunnel construction, and mining, wet-mix shotcrete has become a crucial material for support structures due to its excellent bonding properties, strength, and durability. Wet-mix shotcrete technology involves spraying grout onto the construction surface under high pressure, which not only rapidly forms a support layer but also effectively prevents geological disasters such as landslides and debris flows. However, in actual construction, the use of wet-mix shotcrete often results in excessive consumption, leading to material waste and posing significant challenges to project management and cost control.
[0003] Excessive material consumption stems primarily from multiple factors, including improper material proportioning, unreasonable equipment parameter settings, changes in environmental conditions, and insufficient operational experience among construction personnel. When using wet-mixed shotcrete, construction workers often rely on visual inspection and experience to determine the amount of concrete needed. This method is easily influenced by subjective factors, leading to improper material usage. Furthermore, during construction, the real-time changes in the over- and under-excavation morphology of the surrounding rock, fluctuations in environmental temperature and humidity, and differences in surrounding rock permeability create a coupling effect, making it difficult to dynamically revise theoretical material consumption benchmarks. Existing fluid dynamics models are mostly based on static geological parameters, failing to consider the spatiotemporal heterogeneity of the surrounding rock's adsorption capacity and the time-varying characteristics of the material's setting process. This results in low confidence levels for theoretical consumption ranges, hindering accurate guidance for on-site construction. Moreover, material rebound and slurry segregation are often caused by a combination of factors, including deviations in mechanical motion trajectories, mismatch in material flow, and environmental interference. Current technologies lack the ability to perform multi-dimensional correlation analysis of process parameters, material flow patterns, and environmental physical fields, making it difficult to extract explainable causes of excessive consumption from composite data, resulting in insufficiently targeted optimization measures. Summary of the Invention
[0004] To address the aforementioned problems, embodiments of the present invention provide a method for analyzing excessive wear during wet spraying in intelligent tunnel initial support construction. The method includes:
[0005] By acquiring spatial distribution data of the jetting trajectory through a three-dimensional motion capture system installed on the jetting unit, and combining it with the material flow fluctuation curve generated by the dynamic weighing system, a three-dimensional feature tensor including material flow characteristics, jetting process characteristics and environmental disturbance characteristics is constructed.
[0006] Based on the geological BIM model, the theoretical volume of the initial support structure is extracted. The over-excavation and under-excavation data of the surrounding rock obtained by point cloud scanning are integrated. A dynamic usage benchmark model considering rebound loss is established through fluid dynamics simulation, and theoretical usage interval values with confidence intervals are generated.
[0007] The material flow fluctuation curve is dynamically compared with the theoretical usage range value to extract the material flow deviation, spray trajectory matching degree and environmental interference intensity. The material flow deviation, spray trajectory matching degree and environmental interference intensity are input into the convolutional neural network to generate a three-dimensional feature map of the over-consumption mode.
[0008] Based on the three-dimensional feature map of the over-consumption mode, the multilayer perceptron algorithm is used to match the preset typical over-consumption mode library to generate a diagnostic report including material loss path, process deviation node and environmental interference intensity, and output a set of parameter optimization suggestions.
[0009] The optimized suggestion set is input into the self-learning model of injection parameters. The injection pressure, nozzle movement speed and material flow matching coefficient are dynamically adjusted through reinforcement learning algorithm to form a closed-loop control loop, and the results are stored in the knowledge graph database.
[0010] Furthermore, the environmental disturbance features include three physical field data: the surrounding rock permeability gradient field, the airflow disturbance intensity field, and the temperature gradient field. An environmental disturbance feature subspace is constructed using a multi-physics coupling analysis method and mapped to the third dimension of the three-dimensional feature tensor.
[0011] Furthermore, a time series analysis module is introduced into the dynamic usage benchmark model construction step to establish a spatiotemporal correlation between the over-excavation and under-excavation data of the surrounding rock and the motion trajectory of the jetting unit, generating a dynamic benchmark surface with timestamps.
[0012] Furthermore, the calculation of material flow deviation adopts a dual-channel convolution kernel structure. The first channel extracts the time-frequency characteristics of the material flow fluctuation curve, and the second channel captures the spatial distribution characteristics of the jet trajectory. The composite deviation index is generated through feature fusion.
[0013] Furthermore, the typical over-consumption mode library adopts a hierarchical topology structure, with top-level nodes classified according to material loss paths, middle-level nodes associated with process deviation types, and bottom-level nodes storing environmental interference intensity parameters.
[0014] Furthermore, a constraint satisfaction algorithm is introduced during the parameter optimization suggestion set generation process to establish a dynamic constraint relationship between injection pressure, moving speed and material flow matching coefficient, and generate an optimized solution set that satisfies multi-objective balance.
[0015] Furthermore, the knowledge graph database adopts a spatiotemporal dual index structure, with the time dimension dividing knowledge units according to construction stages and the spatial dimension dividing knowledge nodes according to tunnel cross-section grids.
[0016] Furthermore, the closed-loop control loop is equipped with a three-level optimization mechanism, which includes a primary optimization layer, an intermediate optimization layer, and a high-level optimization layer. The primary optimization layer adjusts the nozzle moving speed, the intermediate optimization layer corrects the material flow matching coefficient, and the high-level optimization layer reconstructs the injection pressure gradient field, forming a progressive parameter optimization system.
[0017] A method for analyzing excessive consumption in wet spraying during intelligent tunnel initial support construction also includes:
[0018] By real-time monitoring of the adsorption rate of the surrounding rock using an embedded moisture sensor array, and by dynamically correcting the permeability parameters of the hydrodynamic model using an extended Kalman filter, a dynamic benchmark surface update mechanism with geological adaptability is generated. The dielectric constant variation curve of the sprayed concrete is determined by the time-domain reflectometry method, the hydration reaction process of the material is inverted, and a dosage compensation coefficient considering the time-varying characteristics of the material is established to achieve online self-calibration of the theoretical dosage range.
[0019] A smart tunnel initial support construction wet spraying excess consumption analysis system, the system includes:
[0020] The multi-source data fusion module acquires the spatial distribution data of the injection trajectory through the three-dimensional motion capture system installed on the injection unit, and combines it with the material flow fluctuation curve generated by the dynamic weighing system to construct a three-dimensional feature tensor that includes material flow characteristics, injection process characteristics and environmental interference characteristics.
[0021] The benchmark model construction module extracts the theoretical volume of the initial support structure based on the geological BIM model, integrates the over- and under-excavation data of the surrounding rock obtained by point cloud scanning, and establishes a dynamic usage benchmark model that considers rebound loss through fluid dynamics simulation, generating theoretical usage interval values with confidence intervals.
[0022] The graph construction module dynamically compares the material flow fluctuation curve with the theoretical usage range value, extracts the material flow deviation, the spray trajectory matching degree and the environmental interference intensity, and inputs the material flow deviation, the spray trajectory matching degree and the environmental interference intensity into the convolutional neural network to generate a three-dimensional feature map of the over-consumption mode.
[0023] The decision generation module is based on the three-dimensional feature map of the over-consumption mode. It uses a multilayer perceptron algorithm to match a preset library of typical over-consumption modes, generate a diagnostic report that includes material loss paths, process deviation nodes, and environmental interference intensity, and outputs a set of parameter optimization suggestions.
[0024] The closed-loop feedback optimization module inputs the set of optimization suggestions into the self-learning model of injection parameters. Through reinforcement learning algorithms, it dynamically adjusts the injection pressure, nozzle movement speed and material flow matching coefficient to form a closed-loop control loop, which is then stored in the knowledge graph database.
[0025] The technical effects and advantages of the intelligent tunnel initial support construction wet spraying excess consumption analysis method and system provided by this invention are as follows:
[0026] This invention solves the problems of inaccurate dynamic adaptation of geological conditions and unanalyzable multi-factor coupled over-consumption modes in wet spraying construction by constructing a three-dimensional feature map and closed-loop optimization system through the fusion of multi-physics dynamic data. It achieves dynamic and precise control of material usage and interpretable diagnosis of over-consumption causes. The invention integrates geological BIM models, over- and under-excavation point clouds of surrounding rock, and multi-physics environmental data, generating a four-dimensional dynamic reference surface through spatiotemporal interpolation algorithms to achieve real-time calibration of theoretical usage range values. It introduces online identification of permeability gradient fields and monitoring of material dielectric properties, combined with extended Kalman filtering to dynamically correct hydrodynamic boundary conditions, enabling the theoretical usage reference to synchronously respond to geological conditions and material performance evolution. Based on three-dimensional feature tensors... (Material flow state / process trajectory / environmental interference) and a dual-channel convolutional network generate a three-dimensional feature map of space-time-intensity, breaking through the limitations of traditional single-dimensional data analysis; a hierarchical over-consumption pattern library is constructed, and the causes of over-consumption are accurately located through a three-level inference link of material loss path → process deviation node → environmental interference intensity; a parameter self-learning model driven by reinforcement learning is adopted, combined with a spatiotemporal dual-index knowledge graph, to form a closed loop of "perception-diagnosis-optimization-precision", realizing the continuous evolution of injection parameters under complex working conditions; a three-level optimization mechanism (speed leveling → material flow stabilization → pressure plastic field) is progressively advanced, and a non-dominated solution set is generated through multi-objective constraint solving, reducing the overall rebound rate. Attached Figure Description
[0027] Figure 1 is a flowchart of a method for analyzing excessive consumption of wet spraying in intelligent tunnel initial support construction according to Example 1.
[0028] Figure 2 is a flowchart of a method for analyzing excessive consumption of wet spraying in intelligent tunnel initial support construction in Example 2.
[0029] Figure 3 is a schematic diagram of the connection of a tunnel intelligent initial support construction wet spraying over-consumption analysis system in Example 3. Detailed Implementation
[0030] 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.
[0031] Example 1:
[0032] Please refer to Figure 1. An embodiment of the present invention provides a method for analyzing excessive consumption of wet spraying during intelligent initial support construction of tunnels. The method includes:
[0033] By acquiring spatial distribution data of the jetting trajectory through a three-dimensional motion capture system installed on the jetting unit, and combining it with the material flow fluctuation curve generated by the dynamic weighing system, a three-dimensional feature tensor including material flow characteristics, jetting process characteristics and environmental disturbance characteristics is constructed.
[0034] Based on the geological BIM model, the theoretical volume of the initial support structure is extracted. The over-excavation and under-excavation data of the surrounding rock obtained by point cloud scanning are integrated. A dynamic usage benchmark model considering rebound loss is established through fluid dynamics simulation, and theoretical usage interval values with confidence intervals are generated.
[0035] The material flow fluctuation curve is dynamically compared with the theoretical usage range value to extract the material flow deviation, spray trajectory matching degree and environmental interference intensity. The material flow deviation, spray trajectory matching degree and environmental interference intensity are input into the convolutional neural network to generate a three-dimensional feature map of the over-consumption mode.
[0036] Based on the three-dimensional feature map of the over-consumption mode, the multilayer perceptron algorithm is used to match the preset typical over-consumption mode library to generate a diagnostic report including material loss path, process deviation node and environmental interference intensity, and output a set of parameter optimization suggestions.
[0037] The optimized suggestion set is input into the self-learning model of injection parameters. The injection pressure, nozzle movement speed and material flow matching coefficient are dynamically adjusted through reinforcement learning algorithm to form a closed-loop control loop, and the results are stored in the knowledge graph database.
[0038] The environmental disturbance features include three physical field data: the surrounding rock permeability gradient field, the airflow disturbance intensity field, and the temperature gradient field. An environmental disturbance feature subspace is constructed using a multi-physics coupling analysis method and mapped to the third dimension of the three-dimensional feature tensor.
[0039] In the process of implementing multi-source data perception and fusion, the motion trajectory of the nozzle in three-dimensional space is captured in real time by a three-dimensional motion capture device. Combined with the concrete material flow fluctuation data collected by dynamic weighing sensors, a three-dimensional feature tensor reflecting the comprehensive state of the spraying process is constructed. Among them, the quantitative analysis of environmental interference characteristics adopts a multi-physics coupling modeling method. That is, the permeability gradient field data of the surrounding rock is obtained in real time by a geological sensor array deployed on the tunnel working face to characterize the difference in the adsorption capacity of the surrounding rock to sprayed concrete in different sections. The airflow disturbance intensity field of the spraying operation area is monitored by an airflow velocity sensor network to reflect the degree of interference of airflow caused by ventilation system and mechanical movement on the spraying trajectory. At the same time, the temperature gradient field data of the spraying surface is collected by an infrared thermal imaging device to characterize the influence of environmental temperature distribution on the material setting rate.
[0040] These three physical field datasets, after feature dimensionality reduction using principal component analysis, were integrated into an environmental disturbance feature subspace. This subspace was then mapped to the third dimension of a three-dimensional feature tensor in tensor encoding form, forming a three-dimensional data representation structure encompassing material flow characteristics (first dimension), jet trajectory characteristics (second dimension), and environmental disturbance characteristics (third dimension).
[0041] This composite feature tensor provides comprehensive data support for the construction of subsequent dynamic usage benchmark models, covering geological conditions, mechanical motion, and environmental factors.
[0042] The dynamic usage benchmark model construction process introduces a time series analysis module to establish a spatiotemporal correlation between the over-excavation and under-excavation data of the surrounding rock and the motion trajectory of the jetting unit, generating a dynamic benchmark surface with timestamps.
[0043] During the implementation of the dynamic usage benchmark model, the dynamic coupling relationship between geological data and construction process was strengthened through spatiotemporal correlation modeling technology. Specifically, the time series analysis module synchronizes the millimeter-level over- and under-excavation data obtained by the laser scanning device on the surrounding rock surface with the six-degree-of-freedom trajectory data recorded by the jetting unit motion capture system in spatiotemporal time. That is, spatial coordinate labels are added to each frame of point cloud data by a high-precision Beidou positioning device, and millisecond-level timestamps are generated by the motion control signal output by the jetting robotic arm controller to establish a dual mapping relationship of "spatial deformation-time trajectory".
[0044] Based on a spatiotemporal interpolation algorithm, discrete over- and under-excavation point cloud data are matched with continuous spraying trajectories in three dimensions to generate a dynamic reference surface with temporal continuity. This dynamic reference surface uses the theoretical initial support contour of the BIM model as its base, and superimposes real-time changes in the surrounding rock morphology along the time axis of the spraying trajectory, forming a four-dimensional data model (three-dimensional space + time dimension) that can dynamically characterize the interaction between geological conditions and construction status. When simulating dynamic material consumption on this surface using a fluid dynamics simulation engine, environmental disturbance characteristic data (such as permeability gradient field and temperature field) from different time slices are simultaneously integrated. The final output is a theoretical consumption interval value with confidence intervals that matches the construction progress, providing a spatiotemporally correlated quantitative benchmark for the dynamic comparison of over-consumption characteristics.
[0045] The calculation of material flow deviation adopts a dual-channel convolution kernel structure. The first channel extracts the time-frequency characteristics of the material flow fluctuation curve, and the second channel captures the spatial distribution characteristics of the jet trajectory. The composite deviation index is generated by feature fusion.
[0046] The first channel connects to the dynamic weighing system and transmits the material flow fluctuation curve in real time. Through time-frequency transformation and sliding window analysis, the periodic fluctuation characteristics and instantaneous change characteristics of the material flow velocity are extracted. The feature vectors output by the two channels are normalized and then dynamically weighted and fused through a learnable weight matrix to form a composite deviation index that characterizes the degree of matching between the actual flow state of the material and the process parameters.
[0047] The second channel uses the point cloud data of the jet trajectory collected by the three-dimensional motion capture system to construct a spatial distribution feature matrix. It then uses multi-scale convolution kernels to extract spatial morphological features such as the continuity and coverage density of the jet trajectory, capturing uneven thickness or local accumulation of the jet layer.
[0048] The material flow deviation, the jet trajectory matching degree, and the environmental temperature and humidity interference intensity together constitute a three-dimensional feature vector, which is input into a deep convolutional network for feature dimensionality upscaling and pattern mapping, and finally generates an over-consumption pattern feature map with three-dimensional spatial-temporal-intensity attributes.
[0049] The typical over-consumption mode library adopts a hierarchical topology structure. The top-level nodes are classified according to the material loss path, the middle-level nodes are associated with the process deviation type, and the bottom-level nodes store environmental interference intensity parameters.
[0050] The typical over-consumption mode library is organized into three levels: top, middle and bottom. The top-level nodes are divided into main categories such as aggregate rebound, slurry segregation and interface adhesion failure according to the material loss path. Under each main category, there are middle-level nodes that are directly related to the spraying process parameters, including process deviation types such as spraying pressure mismatch, nozzle movement speed deviation and abnormal accelerator dosage. The bottom-level nodes are bound to quantitative parameters of environmental disturbance intensity, such as dynamic monitoring data such as surrounding rock seepage rate, air temperature and humidity gradient and equipment vibration amplitude.
[0051] When the three-dimensional feature tensor is input into the multilayer perceptron model, the multilayer perceptron model first filters the main anomaly direction through the material loss path activation function of the top node, then calculates the deviation weight of the feature vector from the preset process threshold along the middle node, and finally combines the environmental parameters of the bottom node to correct the interference intensity, and outputs multi-dimensional diagnostic labels such as "aggregate rebound - nozzle movement speed overload - water seepage interference > 45%".
[0052] The three-tier architecture supports drill-down from macroscopic loss types to microscopic process parameters, while preserving the corrective effect of environmental factors on decision-making, providing an interpretable reasoning path for the generation of parameter optimization suggestion sets.
[0053] In the process of generating the parameter optimization suggestion set, a constraint satisfaction algorithm is introduced to establish a dynamic constraint relationship between injection pressure, moving speed, and material flow matching coefficient, and generate an optimized solution set that satisfies multi-objective balance.
[0054] Based on the process deviation nodes and environmental interference intensity parameters in the diagnostic report, a dynamic constraint matrix is first established for the injection pressure, nozzle moving speed, and material flow matching coefficient. The establishment method includes:
[0055] Spatial constraints are formed by associating the injection pressure threshold with the over- and under-excavation data of the surrounding rock. Kinematic constraints are established by the consistency between the nozzle movement speed range and the trajectory in the three-dimensional feature tensor. The material flow matching coefficient is used to form a mass conservation constraint by combining the dynamic weighing system data with the fluid dynamics model.
[0056] The constraint satisfaction algorithm seeks a balance between suppressing aggregate rebound (material loss path), ensuring spray density (process objective), and controlling equipment energy consumption (environmental constraints) by iteratively solving the Pareto front of the spray pressure, nozzle moving speed, and material flow matching coefficient.
[0057] Each iteration generates a feasible solution that simulates the injection efficiency through a reinforcement learning agent, calculating indicators such as the rebound rate decrease gradient, the injection efficiency improvement coefficient, and the interface density increment. The final output includes 3-5 sets of parameter optimization schemes for non-dominated solutions. Each scheme is labeled with a "pressure-speed-material flow" collaborative adjustment strategy and expected improvement indicators. For example, the "increase moving speed by 12% + dynamically match material flow coefficient" scheme corresponds to a predicted effect of reducing the rebound rate by 8.2% and improving the injection efficiency by 15%. Operators can choose to implement the scheme according to the on-site working conditions, and the optimization results are fed back to the knowledge graph database to improve the constraint relationship model.
[0058] The knowledge graph database adopts a spatiotemporal dual index structure. The time dimension divides knowledge units according to construction stages, and the spatial dimension divides knowledge nodes according to tunnel cross-section grids.
[0059] The time dimension index uses the tunnel construction stage as the dividing benchmark, and divides the spraying parameters, environmental monitoring data and process optimization schemes into dynamic knowledge units according to the process stages such as initial spraying support, re-spraying reinforcement and invert construction. Each dynamic knowledge unit embeds a construction time sequence chain to record the evolution trajectory of process parameters.
[0060] The spatial dimension index is based on the tunnel's three-dimensional laser scanning data to construct a cross-sectional grid coordinate system. Each 1m×1m grid cell is used as an independent knowledge node to store spatial feature data such as historical jet trajectory point clouds, material loss heat maps, and the execution effect of optimization parameters in the area.
[0061] When performing quality traceability or parameter tuning, multidimensional retrieval is achieved through a spatiotemporal correlation engine; for example:
[0062] When the spatiotemporal coordinates of “re-spraying reinforcement stage + mileage K32+150” are entered, and a local density deficiency is detected in the invert section at mileage K32+150, the operator enters the spatiotemporal coordinate parameters through the human-machine interface: the time dimension is selected as “re-spraying reinforcement stage (2023 / 05 / 12-05 / 15)”, the spatial dimension is located to the 7th circumferential partition, and the vertical grid is G-235. The spatiotemporal correlation engine performs the following multidimensional search.
[0063] The accelerator dosage record (fluctuation range of 3.8%-4.2%) and corresponding interface strength test data (average of 28MPa) of the grid during the initial spraying stage (May 8, 2023) were retrieved. It was found that the dosage was increased to 4.5% during the re-spraying stage, but the strength did not meet expectations (31MPa compared to the design value of 35MPa).
[0064] Retrieving the time-series data (0.15-0.25g) of the vibration amplitude of the spraying robotic arm in adjacent grids G-234 and G-236, it was found that grid G-235 exhibited an abnormal vibration peak (0.38g@2023 / 05 / 13 14:23) during re-spraying.
[0065] By combining equipment operation and maintenance records in the knowledge graph, the abnormal vibration period corresponds to the hydraulic pump pressure fluctuation event of the shotcrete truck. It is also associated with the failure mode "vibration amplitude > 0.3g causing slurry segregation" in the historical case library, and automatically generates optimization suggestions with spatiotemporal tags: "In the re-spraying operation of grid G-235, reduce the spraying pressure by 8%-10% to compensate for the impact of hydraulic fluctuation, and simultaneously increase the accelerator dosage to 5.0% and extend the curing time by 2 hours." This solution is displayed through a 3D visualization interface, where the red highlighted area shows the current problem grid, the blue semi-transparent layer overlays the vibration data trend line of adjacent grids, and the yellow arrow marks the direction of optimization parameter adjustment. The quality re-inspection data after execution (such as 34.7MPa) is automatically transmitted back to the knowledge node, forming a closed-loop learning link.
[0066] The spatiotemporal dual index structure enables construction personnel to quickly locate "spatiotemporal blind spots," such as material segregation patterns caused by equipment vibration at the junction of the inverted arch and the sidewall. It also generates targeted process optimization solutions through cross-stage knowledge transfer. All newly generated anomaly diagnosis results and optimization parameter sets are automatically archived according to spatiotemporal tags, forming a knowledge network with spatiotemporal evolution characteristics, providing spatiotemporally correlated experience support for intelligent decision-making in subsequent construction sections.
[0067] The closed-loop control circuit is equipped with a three-level optimization mechanism, which includes a primary optimization layer, an intermediate optimization layer, and a high-level optimization layer. The primary optimization layer adjusts the nozzle moving speed, the intermediate optimization layer corrects the material flow matching coefficient, and the high-level optimization layer reconstructs the injection pressure gradient field, forming a progressive parameter optimization system.
[0068] The primary speed optimization layer uses the spray trajectory deviation data obtained by the laser scanner to drive the dynamic compensation of the nozzle movement speed. When the 3D point cloud analysis shows an over-dug area, speed adjustment commands are generated according to the principle of "under-spraying deceleration - over-drilling acceleration". For example, the nozzle translation rate is automatically reduced at the arch joint to enhance the material accumulation effect. At the same time, the vibration spectrum of the robotic arm is monitored in real time through the inertial measurement unit (IMU) to prevent sudden speed changes from causing equipment resonance.
[0069] Based on speed adjustment, the intermediate material flow optimization layer uses a fuzzy control algorithm to correct the material flow matching coefficient according to the aggregate consumption rate fed back by the dynamic weighing system and the slurry rheological characteristics data from the pressure sensor. When the slurry viscosity increases after the addition of the accelerator is detected, the coupling coefficient between the conveying screw speed and the air supply of the air compressor is automatically increased to ensure that the jet flow maintains a stable jet shape during the acceleration stage. The interface temperature field distribution is detected by infrared thermal imager to verify the effect of material flow matching on improving the uniformity of hydration reaction.
[0070] The advanced pressure field optimization layer, based on the spatiotemporal parameter set formed by the first two optimization stages, constructs a dynamic pressure gradient field model through computational fluid dynamics simulation. The dynamic pressure gradient field model divides the tunnel cross-section into pressure control domains and applies a gradient pressurization strategy in the weak surrounding rock area (such as increasing the pressure value by 12%-15% at the arch foot to form a compaction and anchoring zone). At the same time, combined with the rock fracture development direction detected by ground-penetrating radar, it intelligently generates a pressure vectorized distribution map to guide the multi-axis coordinated movement of the nozzles under different spatial poses. The optimized pressure field parameters are verified by virtual injection through a digital twin system, and finally, a global pressure configuration scheme with weighted coefficients is generated.
[0071] The three-level optimization mechanism forms a progressive optimization path of "speed leveling - material flow stabilization - pressure molding field" by sharing the process constraints in the spatiotemporal index knowledge nodes. The vibration spectrum, material consumption rate and interface strength data generated by each spraying operation are fed back to the knowledge graph, driving the adaptive update of the constraint matrix and realizing the continuous evolution of process parameters in the spatiotemporal dimension.
[0072] Example 2:
[0073] As shown in Figure 2, this embodiment further improves the design based on Embodiment 1. The difference is that Embodiment 1, in actual operation, found that the dynamic usage benchmark model exhibited confidence drift in the theoretical usage range in complex bedding geological sections. This was mainly due to the anisotropy of the surrounding rock permeability causing distortion of the fluid simulation boundary conditions. Based on this, a method for analyzing excessive consumption of wet spraying in intelligent tunnel initial support construction further includes:
[0074] By real-time monitoring of the adsorption rate of the surrounding rock using an embedded moisture sensor array, and by dynamically correcting the permeability parameters of the hydrodynamic model using an extended Kalman filter, a dynamic benchmark surface update mechanism with geological adaptability is generated. The dielectric constant variation curve of the sprayed concrete is determined by the time-domain reflectometry method, the hydration reaction process of the material is inverted, and a dosage compensation coefficient considering the time-varying characteristics of the material is established to achieve online self-calibration of the theoretical dosage range.
[0075] The dynamic correlation correction mechanism between the adsorption rate of the surrounding rock and the permeability gradient field enhances the adaptability of the fluid dynamics simulation model to complex geological conditions. Meanwhile, the time-varying compensation model of materials based on dielectric property inversion constructs a closed-loop feedback link between material performance evolution and usage benchmark. The synergistic effect of the two enables the dynamic benchmark surface to respond synchronously to changes in the adsorption capacity of the surrounding rock and the performance decay of the sprayed material. Finally, through the spatiotemporal index structure of the knowledge graph database, the corrected benchmark data is compared with historical construction parameters across stages to form a usage prediction system with self-evolution capabilities.
[0076] Example 3:
[0077] As shown in Figure 3, based on the same inventive concept as the intelligent tunnel initial support construction wet spraying excess consumption analysis method in the aforementioned embodiments, this application provides an intelligent tunnel initial support construction wet spraying excess consumption analysis system. The system and method embodiments in this application are based on the same inventive concept. The system includes:
[0078] The multi-source data fusion module acquires the spatial distribution data of the injection trajectory through the three-dimensional motion capture system installed on the injection unit, and combines it with the material flow fluctuation curve generated by the dynamic weighing system to construct a three-dimensional feature tensor that includes material flow characteristics, injection process characteristics and environmental interference characteristics.
[0079] The benchmark model construction module extracts the theoretical volume of the initial support structure based on the geological BIM model, integrates the over- and under-excavation data of the surrounding rock obtained by point cloud scanning, and establishes a dynamic usage benchmark model that considers rebound loss through fluid dynamics simulation, generating theoretical usage interval values with confidence intervals.
[0080] The graph construction module dynamically compares the material flow fluctuation curve with the theoretical usage range value, extracts the material flow deviation, the spray trajectory matching degree and the environmental interference intensity, and inputs the material flow deviation, the spray trajectory matching degree and the environmental interference intensity into the convolutional neural network to generate a three-dimensional feature map of the over-consumption mode.
[0081] The decision generation module is based on the three-dimensional feature map of the over-consumption mode. It uses a multilayer perceptron algorithm to match a preset library of typical over-consumption modes, generate a diagnostic report that includes material loss paths, process deviation nodes, and environmental interference intensity, and outputs a set of parameter optimization suggestions.
[0082] The closed-loop feedback optimization module inputs the set of optimization suggestions into the self-learning model of injection parameters. Through reinforcement learning algorithms, it dynamically adjusts the injection pressure, nozzle movement speed and material flow matching coefficient to form a closed-loop control loop, which is then stored in the knowledge graph database.
[0083] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.
[0084] The above description is merely a preferred embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present application, based on the technical solution and concept of the present application, should be covered within the scope of protection of the present application.
Claims
1. A wet spraying super-consumption analysis method for tunnel intelligent primary support construction, characterized in that, include: By acquiring spatial distribution data of the jetting trajectory through a three-dimensional motion capture system installed on the jetting unit, and combining it with the material flow fluctuation curve generated by the dynamic weighing system, a three-dimensional feature tensor including material flow characteristics, jetting process characteristics and environmental disturbance characteristics is constructed. Based on the geological BIM model, the theoretical volume of the initial support structure is extracted. The over-excavation and under-excavation data of the surrounding rock obtained by point cloud scanning are integrated. A dynamic usage benchmark model considering rebound loss is established through fluid dynamics simulation, and theoretical usage interval values with confidence intervals are generated. The material flow fluctuation curve is dynamically compared with the theoretical usage range value to extract the material flow deviation, spray trajectory matching degree and environmental interference intensity. The material flow deviation, spray trajectory matching degree and environmental interference intensity are input into the convolutional neural network to generate a three-dimensional feature map of the over-consumption mode. Based on the three-dimensional feature map of the over-consumption mode, the multilayer perceptron algorithm is used to match the preset typical over-consumption mode library to generate a diagnostic report including material loss path, process deviation node and environmental interference intensity, and output a set of parameter optimization suggestions. The optimized suggestion set is input into the self-learning model of injection parameters. The injection pressure, nozzle movement speed and material flow matching coefficient are dynamically adjusted through reinforcement learning algorithm to form a closed-loop control loop, and the results are stored in the knowledge graph database. 2.The wet spraying ultra-consumption analysis method for tunnel intelligent primary support construction according to claim 1, characterized in that, The environmental disturbance features include three physical field data: the surrounding rock permeability gradient field, the airflow disturbance intensity field, and the temperature gradient field. An environmental disturbance feature subspace is constructed using a multi-physics coupling analysis method and mapped to the third dimension of the three-dimensional feature tensor.
3. The wet spraying ultra-consumption analysis method for intelligent initial support construction of a tunnel according to claim 1, characterized in that, The dynamic usage benchmark model construction step introduces a time series analysis module to establish a spatiotemporal correlation between the over-excavation and under-excavation data of the surrounding rock and the motion trajectory of the jetting unit, generating a dynamic benchmark surface with timestamps.
4. The wet spraying ultra-consumption analysis method for intelligent initial support construction of a tunnel according to claim 1, characterized in that: The calculation of the material flow deviation adopts a dual-channel convolution kernel structure. The first channel extracts the time-frequency characteristics of the material flow fluctuation curve, and the second channel captures the spatial distribution characteristics of the jet trajectory. A composite deviation index is generated by feature fusion.
5. The wet spraying ultra-consumption analysis method for intelligent initial support construction of a tunnel according to claim 1, characterized in that: The typical over-consumption mode library adopts a hierarchical topology structure. The top-level nodes are classified according to the material loss path, the middle-level nodes are associated with the process deviation type, and the bottom-level nodes store environmental interference intensity parameters.
6. The wet spraying ultra-consumption analysis method for intelligent initial support construction of a tunnel according to claim 1, characterized in that: The parameter optimization suggestion set generation process introduces a constraint satisfaction algorithm to establish a dynamic constraint relationship between injection pressure, moving speed and material flow matching coefficient, and generate an optimized solution set that satisfies multi-objective balance.
7. The wet spraying ultra-consumption analysis method for intelligent initial support construction of a tunnel according to claim 1, characterized in that: The knowledge graph database adopts a spatiotemporal dual index structure. The time dimension divides knowledge units according to construction stages, and the spatial dimension divides knowledge nodes according to tunnel cross-section grids. 8.The wet spraying ultra-consumption analysis method for tunnel intelligent primary support construction according to claim 1, characterized in that: The closed-loop control circuit is equipped with a three-level optimization mechanism, which includes a primary optimization layer, an intermediate optimization layer, and a high-level optimization layer. The primary optimization layer adjusts the nozzle moving speed, the intermediate optimization layer corrects the material flow matching coefficient, and the high-level optimization layer reconstructs the injection pressure gradient field, forming a progressive parameter optimization system.
9. The wet spraying ultra-consumption analysis method for intelligent initial support construction of a tunnel according to claim 1, characterized in that, Also includes: By real-time monitoring of the adsorption rate of the surrounding rock using an embedded moisture sensor array, and by dynamically correcting the permeability parameters of the hydrodynamic model using an extended Kalman filter, a dynamic benchmark surface update mechanism with geological adaptability is generated. The dielectric constant variation curve of the sprayed concrete is determined by the time-domain reflectometry method, the hydration reaction process of the material is inverted, and a dosage compensation coefficient considering the time-varying characteristics of the material is established to achieve online self-calibration of the theoretical dosage range.
10. A tunnel intelligent primary support construction wet spraying super-consumption analysis system, characterized in that, The system includes: The multi-source data fusion module acquires the spatial distribution data of the injection trajectory through the three-dimensional motion capture system installed on the injection unit, and combines it with the material flow fluctuation curve generated by the dynamic weighing system to construct a three-dimensional feature tensor that includes material flow characteristics, injection process characteristics and environmental interference characteristics. The benchmark model construction module extracts the theoretical volume of the initial support structure based on the geological BIM model, integrates the over- and under-excavation data of the surrounding rock obtained by point cloud scanning, and establishes a dynamic usage benchmark model that considers rebound loss through fluid dynamics simulation, generating theoretical usage interval values with confidence intervals. The graph construction module dynamically compares the material flow fluctuation curve with the theoretical usage range value, extracts the material flow deviation, the spray trajectory matching degree and the environmental interference intensity, and inputs the material flow deviation, the spray trajectory matching degree and the environmental interference intensity into the convolutional neural network to generate a three-dimensional feature map of the over-consumption mode. The decision generation module is based on the three-dimensional feature map of the over-consumption mode. It uses a multilayer perceptron algorithm to match a preset library of typical over-consumption modes, generate a diagnostic report that includes material loss paths, process deviation nodes, and environmental interference intensity, and outputs a set of parameter optimization suggestions. The closed-loop feedback optimization module inputs the set of optimization suggestions into the self-learning model of injection parameters. Through reinforcement learning algorithms, it dynamically adjusts the injection pressure, nozzle movement speed and material flow matching coefficient to form a closed-loop control loop, which is then stored in the knowledge graph database.
