A soft rock large deformation tunnel construction method and intelligent support system based on multi-source feedback collaborative regulation

Abstract

This invention relates to the field of tunnel engineering technology and discloses a construction method and intelligent support system for large deformation tunnels in soft rock based on multi-source feedback collaborative control. The method collects multimodal data, uses deep learning algorithms to predict the deformation trend of the soft rock area of ​​the target tunnel based on the multimodal data, performs dynamic risk level classification, and outputs corresponding control commands. The target equipment then performs dynamic advance support, realizing dynamic adjustment during the construction of tunnels in soft rock formations. This solves the technical problems in the prior art of insufficient accuracy of advance geological prediction and lack of real-time dynamic adjustment capability of support methods during the construction of tunnels in soft rock formations.

Description

Technical Field

[0001] This invention relates to the field of tunnel engineering technology, specifically to a construction method and intelligent support system for soft rock tunnels with large deformation based on multi-source feedback collaborative control. Background Technology

[0002] Soft rock formations (such as mudstone, shale, and siltstone) are prone to significant deformation due to their low rock strength and pronounced rheological properties, especially in environments with high ground stress, water abundance, or fault zones. Daily convergence can reach 50-100 mm, and the displacement velocity is rapid. During tunnel construction, traditional support methods struggle to adapt to the rapidly changing mechanical properties of the surrounding rock, easily leading to initial support cracking or even collapse. Traditional drilling and geophysical exploration methods cannot provide real-time feedback on dynamic changes in the surrounding rock, resulting in lagging support parameters and insufficient accuracy in advanced geological prediction. Furthermore, reliance on experience-based support designs means that rigid support structures may fail due to their inability to adapt to deformation, while the lag in flexible support makes it difficult to control the deformation rate. Consequently, there is a lack of real-time dynamic adjustment capabilities during tunnel construction in soft rock formations. Summary of the Invention

[0003] The purpose of this invention is to provide a construction method and intelligent support system for large deformation tunnels in soft rock based on multi-source feedback collaborative control. This method collects multimodal data, uses deep learning algorithms to predict the deformation trend of the soft rock area of ​​the target tunnel based on the multimodal data, performs dynamic risk level classification, and outputs corresponding control commands. The target equipment then performs dynamic advance support, realizing dynamic adjustment during the construction of tunnels in soft rock formations. This solves the technical problems in the prior art where the accuracy of advance geological prediction is insufficient and the support methods lack real-time dynamic adjustment capabilities during the construction of tunnels in soft rock formations.

[0004] This invention is achieved through the following technical solution: Firstly, a construction method for large deformation tunnels in soft rock based on multi-source feedback synergistic control is provided: First, real-time acquisition of acoustic emission signals from the surrounding rock in the target frequency band, as well as acquisition of the three-dimensional displacement field and advanced detection data in the soft rock area of ​​the target tunnel; Then, based on the acoustic emission signal of the surrounding rock, the three-dimensional displacement field and the advanced detection data, the deep learning algorithm is used to predict the deformation trend of the soft rock area of ​​the target tunnel, to classify the dynamic risk level and output the corresponding dynamic risk level control instructions. Finally, according to the control instructions, a hydraulic and electronically coupled support method was used to provide advance support for the soft rock area of ​​the target tunnel.

[0005] To better realize the present invention, the acoustic emission signals of the surrounding rock in the target frequency band are collected in real time by deploying distributed optical fiber acoustic wave sensors (DAS). Specifically, distributed optical fiber acoustic wave sensors (DAS) are deployed along the center line of the arch, waist and invert of the target tunnel at preset intervals and the monitoring frequency band is set to 10-1000Hz. In addition, an acoustic emission amplitude threshold is set to trigger a micro-vibration event alarm, so as to collect the acoustic emission signals of the surrounding rock in the target frequency band in real time.

[0006] To better realize the present invention, furthermore, according to a preset grid specification, a microelectromechanical system (MEMS) inclinometer is gridded on the sidewall of the target tunnel to obtain the three-dimensional displacement field of the soft rock region of the target tunnel.

[0007] To better realize the present invention, further, obtaining the three-dimensional displacement field of the soft rock region of the target tunnel includes: The target tunnel point cloud data is acquired and preprocessed. The tilt data collected by the MEMS inclinometer is converted into displacement in the coordinate system corresponding to the point cloud data and superimposed. The three-dimensional displacement vector of each point is calculated to establish the three-dimensional displacement field of the soft rock area of ​​the target tunnel.

[0008] To better realize the present invention, a digital twin platform is constructed by standardizing the data and aligning the coordinate systems of the BIM building information model and the GIS geographic information system. In the digital twin platform, based on the acoustic emission signal of the surrounding rock, the three-dimensional displacement field and the advanced detection data, the deformation trend of the soft rock area of ​​the target tunnel is predicted by the deep learning algorithm, the dynamic risk level is classified and the corresponding dynamic risk level control instructions are output.

[0009] To better realize the present invention, a deep learning algorithm is further used to predict the deformation rate, microseismic energy accumulation, microseismic event density, and local crack propagation rate of the soft rock area of ​​the target tunnel. Based on preset deformation rate threshold, microseismic energy accumulation threshold, microseismic event density threshold, and local crack propagation rate threshold, dynamic risk level classification is performed and corresponding control instructions for dynamic risk level are output.

[0010] To better realize this invention, further, according to the control command, the advance support of the soft rock area of ​​the target tunnel using a hydraulic and electronically coupled support method includes: According to the voltage control command, adjust the hole diameter and grouting pressure of the grouting anchor, and detect the grouting filling rate; Adjust the arch radius and the expansion speed of the arch radius according to the hydraulic control command.

[0011] Secondly, an intelligent support system is provided, applied to the soft rock large deformation tunnel construction method based on multi-source feedback collaborative control as described in any of the first aspects, including: The heterogeneous sensing module includes a distributed fiber optic acoustic wave sensor (DAS), a microelectromechanical system (MEMS) inclinometer, and a ground-penetrating radar, used to collect acoustic emission signals of the surrounding rock in the target frequency band in real time, as well as to acquire the three-dimensional displacement field of the soft rock area of ​​the target tunnel and acquire advanced detection data. The digital twin decision-making module is used to predict the deformation trend of the soft rock area of ​​the target tunnel based on the acoustic emission signal of the surrounding rock, the three-dimensional displacement field and the advanced detection data, using deep learning algorithms, to perform dynamic risk level classification and output the corresponding dynamic risk level control instructions. An adaptive execution module, comprising a multi-degree-of-freedom adjustable arch frame, an intelligent grouting anchor group, and a concrete rapid spraying robotic arm, is used to provide advanced support for the soft rock area of ​​the target tunnel using a hydraulic and electrical control coupled support method according to control commands.

[0012] To better realize the present invention, the multi-degree-of-freedom adjustable arch frame further includes: The segmented arch frame units and hydraulic actuators are connected by shape memory alloy (SMA) hinge nodes to form the arch frame body, and the hydraulic actuators are symmetrically arranged inside the arch frame body. The multi-degree-of-freedom adjustable arch frame is configured to include three arch frame adjustment modes: radial extension, axial deflection, and local curvature compensation.

[0013] To better realize the present invention, the intelligent grouting anchor rod is further integrated with a piezoelectric ceramic brake to regulate the grouting hole diameter by voltage; The intelligent grouting anchor is equipped with a miniature pressure sensor at its head to provide real-time feedback on the grouting filling rate.

[0014] Compared with the prior art, the present invention has the following advantages and beneficial effects: By fusing multi-source information, the deformation of the surrounding rock is monitored in real time. Deep learning algorithms are used to predict the deformation trend of the soft rock area of ​​the target tunnel, classify risk levels, and adjust the support parameters of the target equipment in real time. This enables real-time dynamic adjustment during the construction of tunnels in soft rock formations, significantly improving the construction safety and efficiency of soft rock tunnels. It is suitable for complex geological conditions such as high ground stress and water abundance. Attached Figure Description

[0015] The present invention will be further described in conjunction with the following drawings and embodiments. All inventive concepts of the present invention should be considered as disclosed content and within the scope of protection of the present invention.

[0016] Figure 1 This is a schematic flowchart illustrating an embodiment of a method for constructing large deformation tunnels in soft rock based on multi-source feedback collaborative control, as described in this application.

[0017] Figure 2 This is a schematic diagram of an intelligent support system according to this application.

[0018] Figure 3 This is a schematic diagram of a multi-degree-of-freedom adjustable arch frame of an intelligent support system according to this application. Detailed Implementation

[0019] Example 1:

[0020] See Figure 1 A construction method for large deformation tunnels in soft rock based on multi-source feedback synergistic control: First, real-time acquisition of acoustic emission signals from the surrounding rock in the target frequency band, as well as acquisition of the three-dimensional displacement field and advanced detection data in the soft rock area of ​​the target tunnel; Then, based on the acoustic emission signal of the surrounding rock, the three-dimensional displacement field and the advanced detection data, the deep learning algorithm is used to predict the deformation trend of the soft rock area of ​​the target tunnel, to classify the dynamic risk level and output the corresponding dynamic risk level control instructions. Finally, according to the control instructions, a hydraulic and electronically coupled support method was used to provide advance support for the soft rock area of ​​the target tunnel.

[0021] In this embodiment, the acoustic emission signal of the surrounding rock is the elastic wave signal released by the micro-fractures inside the rock mass. Through this signal, parameters such as the accumulation of microseismic energy and the density of microseismic events, which reflect the rate of accumulation of rock mass damage, can be obtained. When acquiring the acoustic emission signal of the surrounding rock in real time, algorithms such as wavelet transform can be used to extract effective signal components to cope with the interference of the complex environment inside the tunnel.

[0022] In one optional implementation, firstly, the wavelet basis function and decomposition scale for matching the soft rock area of ​​the target tunnel are selected. Then, the acquired noisy surrounding rock acoustic emission signal is decomposed by wavelet to obtain approximation coefficients and detail coefficients at different scales. Next, the threshold at each decomposition scale is calculated. Based on the threshold, the detail coefficients are threshold quantized, with coefficients less than the threshold set to zero and coefficients greater than or equal to the threshold remaining unchanged or being shrunk. Finally, based on the thresholded approximation coefficients and detail coefficients, inverse wavelet transform is performed to reconstruct the denoised surrounding rock acoustic emission signal. Of course, wavelet transform can also be combined with other algorithms, such as combining wavelet transform with independent component analysis (ICA). First, ICA is used to separate the different independent source signals in the acoustic emission signal of the surrounding rock. Then, wavelet transform and threshold denoising are performed on each independent source signal. Finally, the denoised independent source signals are recombined to obtain the denoised signal, so as to process the acoustic emission signal of the surrounding rock collected by multiple sensors at the same time.

[0023] In this embodiment, the target frequency band is the sensor monitoring frequency band that can accurately collect acoustic emission signals from the surrounding rock. In one optional implementation, the target frequency band is 10-1000Hz.

[0024] In this embodiment, the three-dimensional displacement field is a collection of displacement states of various points in the soft rock area of ​​the target tunnel in three-dimensional space (x, y, z directions). The advanced detection data can be obtained through methods such as ground-penetrating radar and tunnel seismic wave method. Unlike existing technologies, the advanced detection data will be combined with the acoustic emission signal of the surrounding rock and the three-dimensional displacement field for analysis, so as to avoid the problem that existing technologies cannot provide real-time feedback on the dynamic changes of the surrounding rock, resulting in the lag of support parameters and insufficient accuracy of advanced geological prediction.

[0025] In this embodiment, the deep learning algorithm can be an LSTM long short-term memory network, a CNN-LSTM convolutional neural network-long short-term memory network hybrid model, or a Transformer model, and the deformation trend can include the variation trend of one or more parameters among deformation rate, microseismic energy accumulation, microseismic event density, and local crack propagation rate. Risk levels can be simply divided into high risk, medium risk, and low risk. Alternatively, based on the actual conditions of the soft rock area of ​​the target tunnel, different types of risks can be classified according to the changing trends of one or more parameters. Similarly, since the control instructions correspond to the risk levels, they can be three control instructions that match high risk, medium risk, and low risk, or they can be control instructions that are preset separately for each specific risk. Advanced support may include adjustable arch frames and support using smart anchors, and may also include shotcrete.

[0026] Example 2:

[0027] This embodiment is a further optimization based on the above embodiment 1. In this embodiment, distributed fiber optic acoustic wave sensors (DAS) are deployed to collect acoustic emission signals of the surrounding rock in the target frequency band in real time. Specifically, distributed fiber optic acoustic wave sensors (DAS) are deployed along the center line of the arch, waist and invert of the target tunnel at preset intervals and the monitoring frequency band is set to 10-1000Hz. In addition, an acoustic emission amplitude threshold is set to trigger a micro-vibration event alarm so as to collect acoustic emission signals of the surrounding rock in the target frequency band in real time. According to the preset grid specifications, a microelectromechanical system (MEMS) inclinometer is gridded on the sidewall of the target tunnel to obtain the three-dimensional displacement field of the soft rock area of ​​the target tunnel. Obtaining the three-dimensional displacement field of the soft rock region of the target tunnel includes acquiring point cloud data of the target tunnel and preprocessing it, converting the tilt data collected by the MEMS inclinometer into displacement values ​​in the coordinate system corresponding to the point cloud data and superimposing them, calculating the three-dimensional displacement vector of each point, and establishing the three-dimensional displacement field of the soft rock region of the target tunnel.

[0028] In one optional implementation, the tunnel is located in a low-mountainous area with significant topographic relief. The slopes of the tunnel site are mostly covered by colluvial deposits, with some exposed bedrock. The strata consist of Quaternary Holocene colluvial silty clay, breccia (Q4dl), alluvial-diluvial silty clay, and gravelly silty soil (Q4al+pl). The underlying bedrock consists of Middle Devonian Shujiaba Formation sandy slate, argillaceous slate (D2s), and fault breccia. The area traverses seven fault zones with complex fault structures. The fault zones are generally wide, with strong compression and fracturing on both sides of the fault planes. Brecciatica and mylonitization are obvious, and secondary faults, tectonic faults, and folds are well-developed. The interpenetration of faults and tectonic fissures constitutes a relatively complex structural pattern. The tunnel body is excavated with Class IV surrounding rock accounting for 67.7% and Class V surrounding rock accounting for 24.82%, with a maximum deformation limit of 40 cm.

[0029] Therefore, in this embodiment, a set of distributed fiber optic acoustic sensors is arranged every 8m along the center line of the tunnel arch, arch waist and invert arch, with a monitoring frequency band of 10-1000Hz and an acoustic emission amplitude threshold of 50dB. When the threshold is exceeded, a micro-vibration event alarm is triggered to complete the DAS array deployment. MEMS inclinometers were implanted by drilling holes in the tunnel sidewall according to a 2m×2m grid, with a sampling frequency of 100Hz. The inclinometer data was then fused with a point cloud model to generate a three-dimensional displacement vector field. Every 15m of tunneling, a ground-penetrating radar is used to conduct a 30m deep advance detection to identify water-rich zones with fissure widths greater than 3mm. The ground-penetrating radar uses a 500MHz antenna.

[0030] Using this implementation method, comprehensive data can be collected to enhance the ability to perceive risks at all scales; through the DAS array and the 10-1000Hz monitoring frequency band, it is possible to cover the entire process from microcrack initiation to macroscopic fracture of the surrounding rock; the 50dB amplitude threshold can filter construction machinery noise and accurately identify fault activation. Combining these three types of monitoring data, encompassing acoustic information, mechanical displacement information, and geological structure information, provides a comprehensive, multi-dimensional data foundation for constructing a multi-physics coupled model. This helps to more accurately understand the mechanical behavior and geological conditions of the surrounding rock. In subsequent analysis, compared to using only a single data source, the fused data contains more detailed information about the mechanical properties, deformation trends, and geological structure of the surrounding rock. This helps the model more accurately identify key factors affecting the stability of the surrounding rock, thereby improving the accuracy of predicting the deformation trend and potential risks of the surrounding rock.

[0031] The preprocessing of the target tunnel point cloud data includes point cloud denoising and registration. Through denoising, ground point extraction and voxelization, a digital elevation model (DEM) is generated. Software such as Smart3D is used to perform global registration of multi-view point cloud data, and then feature point matching is used to eliminate splicing errors. After the tilt data is superimposed on the point cloud model, time series analysis is used to compare the point cloud data and tilt angle changes at different time periods. Key feature points on the rock surface, such as the displacement changes of cracks and support nodes, are further extracted. Combined with inclinometer data, the three-dimensional displacement vectors of each measuring point are calculated, including horizontal displacement, settlement and tilt direction.

[0032] In one alternative implementation, fusing tilt angle data and point cloud data includes: First, the MEMS inclinometer and the laser scanning device are synchronized in time through an IoT platform or bus communication, such as RS485, to ensure that the data timestamps are consistent. Then, spatial mapping is achieved by combining rotation and translation matrices to convert the tilt angle (X / Y / Z axis) in the local coordinate system output by the inclinometer into the displacement in the global coordinate system of the laser point cloud. Then, an improved Iterative Closest Point (ICP) algorithm is employed. By introducing rotation angle constraints and dynamic iteration coefficients, multi-site point cloud data, including laser and oblique photogrammetry point clouds, are accurately registered to a unified coordinate system. In the registered point cloud model, the triaxial tilt data are converted into surface displacement changes in the rock strata, such as horizontal offset and settlement, based on the inclinometer position, and then superimposed onto the corresponding point cloud feature points. Next, the inclinometer data and point cloud geometric features were fused using extended Kalman filtering (EKF). Zero-velocity correction and attitude self-observation were then used to eliminate gyroscope drift errors and accelerometer dynamic interference. A time-series analysis framework was then established to compare the trends of point cloud deformation and tilt angle changes over continuous time periods. Regression analysis was used to verify the physical correlation between displacement and tilt. Finally, based on the extraction of key points from the fused data, such as the three-dimensional coordinate changes of rock fissures and support nodes, and combined with the dynamic compensation data from the inclinometer, the displacement vector of each measuring point is calculated. A voxel interpolation method is used to extend the discrete displacement vector into a continuous field. Furthermore, the displacement distribution can be displayed and high-risk areas identified through color gradients or arrow density maps. Example 3:

[0033] This embodiment is a further optimization based on the above embodiment 2. In this embodiment, a digital twin platform is constructed by unifying the data standards and aligning the coordinate systems of the BIM building information model and the GIS geographic information system. In the digital twin platform, based on the acoustic emission signal of the surrounding rock, the three-dimensional displacement field and the advanced detection data, the deformation trend of the soft rock area of ​​the target tunnel is predicted by the deep learning algorithm, the dynamic risk level is divided and the corresponding dynamic risk level control instructions are output.

[0034] Deep learning algorithms are used to predict the deformation rate, microseismic energy accumulation, microseismic event density, and local crack propagation rate in the soft rock area of ​​the target tunnel. Based on preset thresholds for deformation rate, microseismic energy accumulation, microseismic event density, and local crack propagation rate, dynamic risk levels are classified and corresponding control instructions for dynamic risk levels are output.

[0035] In one alternative implementation, building a digital twin platform includes: First, the BIM model is converted using the IFC (Industry Foundation Classes) standard. The GIS data is standardized using CityGML or Geodatabase to achieve data standardization. The local coordinate system of BIM is converted into the global coordinate system of GIS using the seven-parameter method or quaternion rotation matrix to achieve coordinate system alignment. Then, 3D Tiles or Cesium Ion are used to optimize the BIM model at multiple levels of detail (LOD) to reduce the number of triangular patches. TIN (Trivial Irregular Triangulation) or DEM (Digital Elevation Model) is used to compress and store the GIS terrain data, thereby achieving model lightweighting. Finally, the data is synchronized to the digital twin platform in real time via MQTT / HTTP protocol.

[0036] In one optional implementation, the deep learning algorithm is a hybrid model of CNN-LSTM convolutional neural network-long short-term memory network. First, the CNN convolutional neural network extracts the spatial distribution features of the three-dimensional displacement field, such as local deformation concentration areas, and extracts local anomaly patterns from the advanced detection data. Then, the LSTM long short-term memory network integrates the spatial features output by the CNN with temporal data to establish a multi-source data collaborative prediction model. This model can simultaneously analyze the temporal correlation of acoustic emission signals and the spatial heterogeneity of the displacement field, improving the accuracy of deformation prediction under complex soft rock geological conditions. In one optional implementation, the risk levels include high-risk areas, medium-risk areas, and low-risk areas, specifically: High-risk area (marked in red): Deformation rate > 10 mm / d, and microseismic energy accumulation > 1 × 10^4 J or microseismic event density > 5 times / h; Medium-risk area (highlighted in yellow): Deformation rate 5-10 mm / d, local crack propagation rate 2-5 mm / h; Low-risk area (marked in green): Deformation rate <5mm / d.

[0037] By using this implementation method, deep learning algorithms are embedded in the digital twin platform, which enables simulation and prediction based on collected multimodal data, thereby improving the accuracy of deformation trend prediction.

[0038] Example 4:

[0039] This embodiment is a further optimization based on the above embodiment 3. In this embodiment, according to the control command, the advance support of the soft rock area of ​​the target tunnel using a hydraulic and electronically coupled support method includes: According to the voltage control command, adjust the hole diameter and grouting pressure of the grouting anchor, and detect the grouting filling rate; Adjust the arch radius and the expansion speed of the arch radius according to the hydraulic control command.

[0040] In one alternative implementation, the hydraulic control command also includes activating the SMA hinge node in the high-risk arch section, deflecting it by 3° to release stress concentration, while simultaneously expanding the arch radius by 12% at a speed of 0.2 m / min; The voltage control instructions include adjusting the grouting pressure as follows: 3-5 MPa in high-risk areas, 1-3 MPa in medium-risk areas, and 0.5-1 MPa in low-risk areas; the grouting filling rate must be ≥95%, otherwise secondary grouting should be initiated; in medium-risk areas, anchor bolts should be arranged in a 1.2m×1.2m quincunx pattern, with the hole diameter dynamically adjusted to 1.5mm.

[0041] Example 5:

[0042] See Figure 2 An intelligent support system, applied to the soft rock large deformation tunnel construction method based on multi-source feedback collaborative control as described in any of Examples 1 to 4, includes: The heterogeneous sensing module includes a distributed fiber optic acoustic wave sensor (DAS), a microelectromechanical system (MEMS) inclinometer, and a ground-penetrating radar, used to collect acoustic emission signals of the surrounding rock in the target frequency band in real time, as well as to acquire the three-dimensional displacement field of the soft rock area of ​​the target tunnel and acquire advanced detection data. The digital twin decision-making module is used to predict the deformation trend of the soft rock area of ​​the target tunnel based on the acoustic emission signal of the surrounding rock, the three-dimensional displacement field and the advanced detection data, using deep learning algorithms, to perform dynamic risk level classification and output the corresponding dynamic risk level control instructions. An adaptive execution module, comprising a multi-degree-of-freedom adjustable arch frame, an intelligent grouting anchor group, and a concrete rapid spraying robotic arm, is used to provide advanced support for the soft rock area of ​​the target tunnel using a hydraulic and electrical control coupled support method according to control commands.

[0043] See Figure 3 Optionally, the multi-degree-of-freedom adjustable arch frame includes: The segmented arch frame units and hydraulic actuators are connected by shape memory alloy (SMA) hinge nodes to form the arch frame body, and the hydraulic actuators are symmetrically arranged inside the arch frame body. The multi-degree-of-freedom adjustable arch frame is configured to include three arch frame adjustment modes: radial extension, axial deflection, and local curvature compensation.

[0044] Further optionally, the intelligent grouting anchor bolt integrates a piezoelectric ceramic brake to regulate the grouting hole diameter by voltage; The intelligent grouting anchor is equipped with a miniature pressure sensor at its head to provide real-time feedback on the grouting filling rate.

[0045] In one alternative implementation, the segmented arch frame units, each 2.5m long, are connected by shape memory alloy SMA hinge nodes. Hydraulic actuators are symmetrically arranged on the inner side of the arch frame, with two sets of actuators configured in each segment, and a stroke accuracy of ±0.1mm; The arch frame adjustment modes include radial extension (stroke range ±15%), axial deflection (angle ±5°) and local curvature compensation.

[0046] In an alternative implementation, the smart grouting anchor will also inject shear-thickening slurry STF into high-risk areas; Preferably, the composition of the shear-thickening slurry (STF) is as follows: Base liquid: 40%-50% nano-silicate, 30%-40% ultrafine cement, 5%-10% carbon nanotubes; Thickener: 10%-15% polyethylene glycol modified silica; The water-cement ratio is 0.6:1-0.8:1, and the viscosity increases non-linearly with the grouting pressure (viscosity ≥500mPa·s at a pressure of 1MPa).

[0047] In one alternative implementation, the concrete spraying robot arm sprays ultra-early strength nano-concrete to seal the core soil while completing the advanced support and excavating the arc-shaped pilot tunnel. The preferred mix proportions for ultra-early strength nano-concrete are: Cement:Sand:Crushed stone:Nano silica:Steel fiber:Accelerator =1:1.8:2.2:0.15:0.1:0.06 (mass ratio); Initial setting time ≤ 3 minutes, compressive strength ≥ 15 MPa at 2 hours, and strength ≥ 40 MPa at 24 hours.

[0048] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Any simple modifications or equivalent changes made to the above embodiments based on the technical essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for constructing large deformation tunnels in soft rock based on multi-source feedback synergistic control, characterized in that: First, real-time acquisition of acoustic emission signals from the surrounding rock in the target frequency band, as well as acquisition of the three-dimensional displacement field and advanced detection data in the soft rock area of ​​the target tunnel; Then, based on the acoustic emission signal of the surrounding rock, the three-dimensional displacement field and the advanced detection data, the deep learning algorithm is used to predict the deformation trend of the soft rock area of ​​the target tunnel, to classify the dynamic risk level and output the corresponding dynamic risk level control instructions. Finally, according to the control instructions, a hydraulic and electronically coupled support method was used to provide advance support for the soft rock area of ​​the target tunnel.

2. The method for constructing large deformation tunnels in soft rock based on multi-source feedback synergistic control according to claim 1, characterized in that: The acoustic emission signals of the surrounding rock in the target frequency band are collected in real time by deploying distributed fiber optic acoustic wave sensors (DAS). Specifically, distributed fiber optic acoustic wave sensors (DAS) are deployed along the center line of the arch, waist and invert of the target tunnel at preset intervals, and the monitoring frequency band is set to 10-1000Hz. In addition, an acoustic emission amplitude threshold is set to trigger a micro-seismic event alarm, so as to collect the acoustic emission signals of the surrounding rock in the target frequency band in real time.

3. The method for constructing large deformation tunnels in soft rock based on multi-source feedback synergistic control according to claim 1, characterized in that: According to the preset grid specifications, a microelectromechanical system (MEMS) inclinometer is gridded on the sidewall of the target tunnel to obtain the three-dimensional displacement field of the soft rock area of ​​the target tunnel.

4. The method for constructing large deformation tunnels in soft rock based on multi-source feedback synergistic control according to claim 3, characterized in that, Obtaining the three-dimensional displacement field of the soft rock region of the target tunnel includes: The target tunnel point cloud data is acquired and preprocessed. The tilt data collected by the MEMS inclinometer is converted into displacement in the coordinate system corresponding to the point cloud data and superimposed. The three-dimensional displacement vector of each point is calculated to establish the three-dimensional displacement field of the soft rock area of ​​the target tunnel.

5. The method for constructing large deformation tunnels in soft rock based on multi-source feedback synergistic control according to claim 1, characterized in that: By standardizing data and aligning coordinate systems between the BIM building information model and the GIS geographic information system, a digital twin platform is constructed. In this digital twin platform, based on the acoustic emission signals of the surrounding rock, the three-dimensional displacement field, and advanced detection data, a deep learning algorithm is used to predict the deformation trend of the soft rock area of ​​the target tunnel, perform dynamic risk level classification, and output corresponding dynamic risk level control instructions.

6. The method for constructing large deformation tunnels in soft rock based on multi-source feedback synergistic control according to claim 5, characterized in that: Deep learning algorithms are used to predict the deformation rate, microseismic energy accumulation, microseismic event density, and local crack propagation rate in the soft rock area of ​​the target tunnel. Based on preset thresholds for deformation rate, microseismic energy accumulation, microseismic event density, and local crack propagation rate, dynamic risk levels are classified and corresponding control instructions for dynamic risk levels are output.

7. The method for constructing large deformation tunnels in soft rock based on multi-source feedback synergistic control according to claim 1, characterized in that, According to the control instructions, the pre-support of the soft rock area of ​​the target tunnel using a hydraulic and electronically coupled support method includes: According to the voltage control command, adjust the hole diameter and grouting pressure of the grouting anchor, and detect the grouting filling rate; Adjust the arch radius and the expansion speed of the arch radius according to the hydraulic control command.

8. An intelligent support system, applied to the construction method of soft rock large deformation tunnel based on multi-source feedback collaborative control as described in any one of claims 1-7, characterized in that, include: The heterogeneous sensing module includes a distributed fiber optic acoustic wave sensor (DAS), a microelectromechanical system (MEMS) inclinometer, and a ground-penetrating radar, used to collect acoustic emission signals of the surrounding rock in the target frequency band in real time, as well as to acquire the three-dimensional displacement field of the soft rock area of ​​the target tunnel and acquire advanced detection data. The digital twin decision-making module is used to predict the deformation trend of the soft rock area of ​​the target tunnel based on the acoustic emission signal of the surrounding rock, the three-dimensional displacement field and the advanced detection data, using deep learning algorithms, to perform dynamic risk level classification and output the corresponding dynamic risk level control instructions. An adaptive execution module, comprising a multi-degree-of-freedom adjustable arch frame, an intelligent grouting anchor group, and a concrete rapid spraying robotic arm, is used to provide advanced support for the soft rock area of ​​the target tunnel using a hydraulic and electrical control coupled support method according to control commands.

9. The intelligent support system according to claim 8, characterized in that, The multi-degree-of-freedom adjustable arch frame includes: The segmented arch frame units and hydraulic actuators are connected by shape memory alloy (SMA) hinge nodes to form the arch frame body, and the hydraulic actuators are symmetrically arranged inside the arch frame body. The multi-degree-of-freedom adjustable arch frame is configured to include three arch frame adjustment modes: radial extension, axial deflection, and local curvature compensation.

10. The intelligent support system according to claim 8, characterized in that: The intelligent grouting anchor bolt integrates a piezoelectric ceramic brake to regulate the grouting hole diameter via voltage. The intelligent grouting anchor is equipped with a miniature pressure sensor at its head to provide real-time feedback on the grouting filling rate.

Images