Multi-source data fusion method and system for tunnel surrounding rock grade judgment

Abstract

The application discloses a multi-source data fusion method for tunnel surrounding rock grade judgment, comprising the following steps: S1, selecting multiple positions around the tunnel advancing line to drill holes and install seismic wave systems to obtain construction site geological data; S2, selecting key positions on the basis of the construction site geological data to perform advanced drilling to obtain key position geological data; S3, performing image shooting on the working face and obtaining working face key structure data through an image recognition algorithm; S4, installing fiber optic sensors around the working face to monitor the mechanical property changes and safety and stability performance of the rock mass near the working face during blasting, and obtaining surrounding rock stress state data; S5, constructing a BP neural network model by using the construction site geological data, the key position geological data, the working face key structure data and the surrounding rock stress state data, obtaining surrounding rock grade geological structure indexes, and judging the surrounding rock grade based on the surrounding rock grade geological structure indexes.

Description

Technical Field

[0001] This invention relates to the field of intelligent tunnel construction technology, specifically to a multi-source data fusion method and system for determining the grade of surrounding rock in tunnels. Background Technology

[0002] In tunnel construction, the drill-and-blast method is the most commonly used construction method in rock engineering excavation. This technique involves drilling, charging explosives, and blasting to excavate rock and create tunnel chambers. It is highly adaptable to various geological conditions and tunnel cross-sectional shapes, and can be used for excavating various types of hard and soft rock, as well as tunnels of different cross-sectional shapes and sizes. This method has evolved from early manual drilling with hand-held chisels and hammers, using detonators to detonate individual explosive charges, to drilling with rock drilling rigs or multi-arm drilling rigs, applying millisecond blasting, pre-splitting blasting, and smooth blasting techniques. Before construction, the excavation method must be selected based on geological conditions, cross-sectional size, support methods, schedule requirements, and available equipment and technology. Blasting design, a key technical aspect of the drill-and-blast method, aims to avoid over- and under-excavation, achieve the expected cycle advance, and minimize material and labor consumption. After each blast, the design parameters are compared with the actual blast, and blasting parameters are adjusted promptly to improve blasting effectiveness and technical and economic indicators.

[0003] However, the existing drill-and-blast construction technology still has the following problems:

[0004] 1. Determining the surrounding rock grade by visual observation relies heavily on human experience and judgment, resulting in insufficient accuracy;

[0005] 2. Joints, fissures, karst caves or weak interlayers in the unexcavated tunnel ahead are difficult to detect by visual observation. During blasting, damage may occur along the structural surface or fissure surface, which can easily lead to over-excavation or under-excavation that is difficult to control subjectively.

[0006] 3. Most current blasting design specifications are based on experience, and their accuracy is difficult to guarantee;

[0007] 4. In the initial stage after tunnel excavation, the surrounding rock mass after blasting is not yet stable. Making manual observation prematurely poses a serious safety hazard.

[0008] Therefore, accurately predicting the surrounding rock and tunnel face structure data in front of the tunnel face is a crucial issue in tunnel construction. Summary of the Invention

[0009] The purpose of this invention is to provide a multi-source data fusion method for determining the grade of surrounding rock in tunnels, so as to solve the problems mentioned in the background art.

[0010] To achieve the above objectives, the present invention provides a multi-source data fusion method for determining the surrounding rock grade of tunnels, comprising the following steps:

[0011] S1. Select multiple locations around the tunnel's forward route to drill holes and install seismic wave systems. Utilize the reflection characteristics of seismic waves to obtain geological data of the construction site. The geological data of the construction site includes interfaces of geological and lithological changes, structural fracture zones, karst, and karst development zones.

[0012] S2. Based on the geological data of the construction site, select key locations and conduct advance drilling to obtain geological data of the key locations. The key locations include joints, fissures, karst caves, and weak interlayer geological structures.

[0013] S3. Take images of the working face and obtain key structural data of the working face through image recognition algorithms;

[0014] S4. Install fiber optic sensors around the working face to monitor the changes in the mechanical properties and safety and stability of the rock mass near the working face during blasting, and obtain data on the stress state of the surrounding rock.

[0015] S5. Construct a BP neural network model using geological data of the construction site, geological data of key locations, key structural data of the tunnel face, and stress state data of the surrounding rock to obtain geological structural indicators of the surrounding rock grade. Based on the geological structural indicators of the surrounding rock grade, determine the surrounding rock grade and adjust and optimize the tunnel blasting construction plan according to the surrounding rock grade.

[0016] In a preferred embodiment, step S1 involves selecting multiple locations along the tunnel's forward path for drilling and installing a seismic wave system. The reflection characteristics of seismic waves are used to obtain geological data of the construction site. This includes the following steps:

[0017] S11. Install seismic wave receiving equipment on the ground, select a suitable seismic source location at the top of the tunnel to set up seismic wave transmitting equipment, and set the excitation parameters of the seismic waves. At the same time, the layout of the seismic survey lines should take into account the location of the detection target, geological conditions, and the depth and resolution requirements of the detection. The survey lines are laid along the direction of the pipeline, and the spacing between the survey lines is set evenly.

[0018] S12. Induce seismic waves, which then propagate through the Earth's crust and are reflected at different geological interfaces;

[0019] S13. Use seismic wave receiving equipment to record the reflected signals of seismic waves, process and analyze the signals to determine the location and shape of geological interfaces, and then infer the location, size and shape of geological structures.

[0020] In a preferred embodiment, step S2 involves selecting key locations based on the geological data of the construction site and conducting advance drilling to obtain geological data for these key locations. The key locations include geological structures such as joints, fissures, karst caves, and weak interlayers. The steps include the following:

[0021] S21. During the drilling process, pay attention to the vibration of the drill rod, the color and flow rate of the flushing fluid, the sound of the impactor, the drilling speed and its changes, rock powder, and stuck drill, in order to make a preliminary judgment on whether there are adverse geological structures at the detection location.

[0022] S22. If it is initially determined that there are unfavorable geological structures, further measures shall be taken for detection and verification.

[0023] S23. After the exploration is completed, computer graphics processing technology is used to draw and analyze the exploration results in order to obtain geological data of key locations;

[0024] In step S22, a further measure is to use a rotary coring drill to drill and obtain rock core samples, thereby determining the type and nature of the adverse geological structure. Specifically, for the detection of joints and fissures, an automatic joint and fissure recorder is used for measurement, and for the detection of karst caves and weak interlayers, an advanced horizontal drilling method is used for detection.

[0025] In a preferred embodiment, step S3 involves capturing images of the tunnel face and obtaining key structural data of the tunnel face using an image recognition algorithm, including the following steps:

[0026] S31. Use an image acquisition device to capture images of the working face to obtain images of the working face area;

[0027] S32. Based on the image-based 3D reconstruction method, a high-resolution global working face surface image is obtained by reconstructing the 3D structure of the working face scene, performing texture mapping and post-processing: Motion recovery structure is performed on the input continuous working face region map to obtain the pose of each image at the time of capture, and a sparse feature point cloud of the entire working face is constructed; based on the recovered camera pose, the obtained depth map is depth-projected to restore the dense point cloud information of the scene, obtaining a highly visualized 3D point cloud; surface reconstruction and texture mapping are performed on the dense point cloud to obtain a 3D surface model with the texture features of the original image; orthogonal mapping is performed on the 3D surface model to obtain a 2D global working face image, and post-processing is performed to determine the image scale information, ultimately obtaining a high-resolution global working face surface image.

[0028] S33. Perform image preprocessing, including edge detection and image segmentation, and construct a denoising algorithm for tunnel face joint feature data. Train the algorithm on tunnel face joint samples to identify general features of the tunnel face. Image transmission is achieved by connecting the local area network to the cloud. Image storage is either local or cloud object storage. In cloud processing, use the Caffe visualization toolbox to extract layer-by-layer features and attach feature histograms for each layer to accurately describe the morphological features of the surrounding rock and obtain key structural data of the tunnel face.

[0029] In a preferred embodiment, step S4 involves installing fiber optic sensors around the tunnel face to monitor changes in the mechanical properties and safety stability of the rock mass near the tunnel face during blasting, and to acquire data on the stress state of the surrounding rock. This includes the following steps:

[0030] S41. Drill holes in the rock mass surrounding the working face; S42. Install fiber optic sensors and embed them into the surrounding rock to be monitored; S43. Connect the signal receiving device of the fiber optic sensors to ensure that the signal receiving device can receive and record the reflection spectrum signal of the fiber optic sensors in real time; S44. Set the threshold of the early warning device. When the detected deformation or stress of the surrounding rock exceeds the preset threshold, the early warning device will issue an early warning signal; S45. Process and analyze the monitored data, including demodulating, analyzing and processing the reflection spectrum signal to obtain real-time monitoring data of the deformation and stress state of the surrounding rock; S46. Take corresponding support measures based on the real-time monitoring data to ensure the stability of the surrounding rock.

[0031] In a preferred embodiment, step S5 involves constructing a BP neural network model using geological data of the construction site, geological data of key locations, key structural data of the tunnel face, and stress state data of the surrounding rock to obtain geological structural indicators of the surrounding rock grade. Based on these indicators, the surrounding rock grade is determined, and the tunnel blasting construction plan is adjusted and optimized according to the surrounding rock grade. This includes the following steps:

[0032] S51. Data Processing: Preprocess the collected geological data of the construction site, geological data of key locations, key structural data of the tunnel face, and stress state data of the surrounding rock. The preprocessing includes data cleaning, missing value handling, and outlier handling, and the data is normalized.

[0033] S52. Construct a BP neural network model: Construct a BP neural network model using the processed data. Take the geological data of the construction site, the geological data of key locations, the key structural data of the working face, and the stress state data of the surrounding rock as input variables, and take the integrity of the surrounding rock and the hardness of the rock as output variables. Select an appropriate neural network structure according to the scale and complexity of the data.

[0034] S53. Training the model: Use the training dataset to train the BP neural network model. During the training process, it is necessary to adjust the parameters of the neural network, including the learning rate, number of iterations, and weight matrix, in order to obtain the optimal model performance.

[0035] S54. Model Validation and Optimization: Use the validation dataset to validate and optimize the trained model. If there is a large error between the model's prediction results and the actual results, the model needs to be adjusted and optimized.

[0036] S55. Application Model: Based on the rock integrity and rock hardness obtained from the trained BP neural network model, the BQ rock mass basic quality index formula is used to determine the surrounding rock grade, and the tunnel blasting construction plan is adjusted and optimized according to the surrounding rock grade.

[0037] BQ = 100 + 3Rc + 250Kv

[0038] Where Rc is the uniaxial saturated compressive strength of the rock, and Kv is the integrity coefficient of the rock mass.

[0039] This invention also provides a multi-source data fusion system for determining the surrounding rock grade of tunnels, comprising:

[0040] The seismic wave system involves drilling at multiple locations around the tunnel's forward route and installing the seismic wave system. By utilizing the reflection characteristics of seismic waves, geological data of the construction site can be obtained. Based on the geological data of the construction site, key locations are selected for advanced drilling to obtain geological data of those key locations.

[0041] The face structure data acquisition system includes an image acquisition device and an image recognition and processing system. The image acquisition device is used to capture images of the face to obtain images of the face area. The image recognition and processing system is used to reconstruct the three-dimensional structure of the face scene based on the image three-dimensional reconstruction method, and perform texture mapping and post-processing to finally obtain a high-resolution global face surface image. The global face surface is then processed to accurately describe the morphological characteristics of the surrounding rock and obtain key structural data of the face.

[0042] The fiber optic monitoring system includes multiple fiber optic sensors symmetrically installed around the working face to monitor changes in the mechanical properties and safety and stability of the rock mass near the working face during blasting, in order to obtain data on the stress state of the surrounding rock.

[0043] The data fusion system is used to process data acquired from seismic wave systems, face structure data acquisition systems, and fiber optic monitoring systems, combined with key location geological data supplemented by advanced drilling. It uses a BP neural network algorithm to obtain geological structural indicators of the surrounding rock grade, determines the surrounding rock grade based on these indicators, and adjusts and optimizes the tunnel blasting construction plan according to the surrounding rock grade.

[0044] In a preferred embodiment, multiple locations are selected along the tunnel's forward path for drilling and installation of a seismic wave system. The reflection characteristics of seismic waves are utilized to obtain geological data of the construction site, including:

[0045] Seismic wave receiving equipment is installed on the ground, and seismic wave transmitting equipment is set at a suitable source location at the top of the tunnel. The excitation parameters of the seismic waves are also set. Meanwhile, the layout of the seismic survey lines should take into account the location of the target and the geological conditions, as well as the depth and resolution requirements of the detection. The survey lines are laid along the direction of the pipeline, and the spacing between the survey lines is set evenly.

[0046] This generates seismic waves, which propagate through the Earth's crust and are reflected at different geological interfaces.

[0047] By using seismic wave receiving equipment to record the reflected signals of seismic waves, and processing and analyzing the signals, the location and shape of geological interfaces can be determined, thereby inferring the location, size and shape of geological structures.

[0048] In a preferred embodiment, a high-resolution global face image is obtained by reconstructing the three-dimensional structure of the face scene, performing texture mapping and post-processing. The method includes: performing motion reconstruction on the input continuous face region map to obtain the pose of each image at the time of capture, and constructing a sparse feature point cloud of the entire face; performing depth projection on the obtained depth map according to the recovered camera pose to restore the dense point cloud information of the scene and obtain a highly visualized three-dimensional point cloud; performing surface reconstruction and texture mapping on the dense point cloud to obtain a three-dimensional surface model with the texture features of the original image; performing orthogonal mapping on the three-dimensional surface model to obtain a two-dimensional global face image, and performing post-processing to determine the image scale information, ultimately obtaining a high-resolution global face image.

[0049] The entire tunnel face surface is then processed to accurately describe the morphological characteristics of the surrounding rock and obtain key structural data of the tunnel face. This includes: image preprocessing, edge detection and image segmentation of the preprocessed images, constructing a denoising algorithm for tunnel face joint feature data, training tunnel face joint samples to identify general features of the tunnel face, image transmission via local area network to the cloud, and image storage as local or cloud object storage. In cloud processing, the Caffe visualization toolbox is used to extract layer-by-layer features and attach feature histograms for each layer to accurately describe the morphological characteristics of the surrounding rock and obtain key structural data of the tunnel face.

[0050] In a preferred embodiment, based on data acquired by the seismic wave system, the tunnel face structure data acquisition system, and the fiber optic monitoring system, combined with key location geological data supplemented by advanced drilling, a BP neural network algorithm is used for data processing to obtain geological structural indicators of the surrounding rock grade. The surrounding rock grade is then determined based on these indicators, and the tunnel blasting construction plan is adjusted and optimized according to the surrounding rock grade, including:

[0051] Data processing: The collected geological data of the construction site, geological data of key locations, key structural data of the tunnel face, and stress state data of the surrounding rock are preprocessed. The preprocessing includes data cleaning, missing value handling, and outlier handling, and the data is normalized.

[0052] Constructing a BP neural network model: Using the processed data, a BP neural network model is constructed. The geological data of the construction site, the geological data of key locations, the key structural data of the working face, and the stress state data of the surrounding rock are used as input variables, and the integrity of the surrounding rock and the hardness of the rock are used as output variables. The appropriate neural network structure is selected according to the scale and complexity of the data.

[0053] Training the model: The BP neural network model is trained using the training dataset. During the training process, the parameters of the neural network need to be adjusted, including the learning rate, number of iterations, and weight matrix, in order to obtain the optimal model performance.

[0054] Model validation and optimization: The trained model is validated and optimized using a validation dataset. If there is a large error between the model's prediction results and the actual results, the model needs to be adjusted and optimized.

[0055] Application Model: Based on the rock integrity and rock hardness obtained from the trained BP neural network model, the basic quality index formula of BQ rock mass is used to determine the surrounding rock grade, and the tunnel blasting construction plan is adjusted and optimized according to the surrounding rock grade.

[0056] BQ = 100 + 3Rc + 250Kv

[0057] Where Rc is the uniaxial saturated compressive strength of the rock, and Kv is the integrity coefficient of the rock mass.

[0058] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention utilizes a seismic wave system to acquire geological data of the construction site and employs advanced drilling technology as an auxiliary means to detect geological information at key locations, effectively detecting joints, fissures, karst caves, or weak interlayers in the unexcavated tunnel ahead, thus avoiding damage along structural or fissure surfaces during blasting; it rapidly acquires key structural data of the tunnel face through image recognition algorithms, avoiding close-range observation of the tunnel face by on-site construction personnel, thereby improving the safety of tunnel construction; it monitors changes in the mechanical properties and safety stability of the rock mass near the tunnel face during blasting; and through multi-source data fusion, this invention can highly informatize, automate, and intelligently determine the grade of the surrounding rock of the tunnel, providing a solid foundation for refined construction using the tunnel drilling and blasting method, thereby creating a safer, more reliable, economical, and efficient new generation of tunnel blasting technology system. Attached Figure Description

[0059] Figure 1 This is a flowchart of a method according to one embodiment of the present invention.

[0060] Figure 2 This is a schematic diagram of the location of a seismic wave system borehole / advance drilling borehole in an unexcavated tunnel according to an embodiment of the present invention.

[0061] Figure 3 This is a schematic diagram of the arrangement of fiber optic sensors around the working face according to one embodiment of the present invention.

[0062] Explanation of reference numerals in the attached figures:

[0063] 201-Seismic wave borehole, 202-Unexcavated tunnel, 203-Excavated tunnel, 301-Advanced drilling borehole, 302-Working face, 303-Image acquisition equipment, 304-Fiber optic sensor, 305-Surrounding rock mass. Detailed Implementation

[0064] The technical solutions in the embodiments of the present invention will be clearly and completely described below. All other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of the present invention.

[0065] Example 1

[0066] like Figure 1-3 As shown, the multi-source data fusion method for determining the surrounding rock grade of tunnels according to the present invention includes the following steps:

[0067] Step S1: Select representative locations around the tunnel's forward route to drill holes and install a seismic wave system. Utilize the reflection characteristics of seismic waves to obtain geological data of the construction site, such as interfaces of geological and lithological changes, structural fracture zones, karst, and karst development zones.

[0068] Specifically, step S1 includes the following steps:

[0069] S11. Install seismic wave receiving equipment on the ground, select a suitable seismic source location at the top of the tunnel to set up seismic wave transmitting equipment, and set the excitation parameters of the seismic waves. At the same time, the layout of the seismic survey lines should take into account the location of the detection target, geological conditions, and the depth and resolution requirements of the detection. The survey lines are laid along the direction of the pipeline, and the spacing between the survey lines is set evenly.

[0070] S12. Induce seismic waves, which then propagate through the Earth's crust and are reflected at different geological interfaces;

[0071] S13. Use seismic wave receiving equipment to record the reflected signals of seismic waves, process and analyze the signals to determine the location and shape of geological interfaces, and then infer the location, size and shape of geological structures.

[0072] Step S2: Based on the geological data of the construction site, select key locations and conduct advanced drilling to obtain geological data of the key locations. The key locations include the locations of joints, fissures, karst caves, and weak interlayer geological structures.

[0073] Specifically, step S2 includes the following steps:

[0074] S21. During the drilling process, pay attention to the vibration of the drill rod, the color and flow rate of the flushing fluid, the sound of the impactor, the drilling speed and its changes, rock powder, and stuck drill, in order to make a preliminary judgment on whether there are adverse geological structures at the detection location.

[0075] S22. If it is initially determined that there are unfavorable geological structures, further measures shall be taken for detection and verification.

[0076] S23. After the exploration is completed, computer graphics processing technology is used to draw and analyze the exploration results in order to obtain geological data of key locations;

[0077] In step S22, a further measure is to use a rotary coring drill to drill and obtain rock core samples, thereby determining the type and nature of the adverse geological structure. Specifically, for the detection of joints and fissures, an automatic joint and fissure recorder is used for measurement, and for the detection of karst caves and weak interlayers, an advanced horizontal drilling method is used for detection.

[0078] Step S3: Take images of the working face and obtain key structural data of the working face through image recognition algorithms.

[0079] Specifically, step S3 includes the following steps:

[0080] S31. Use an image acquisition device to capture images of the working face to obtain images of the working face area.

[0081] S32. Based on the image-based 3D reconstruction method, a high-resolution global working face surface image is obtained by reconstructing the 3D structure of the working face scene, performing texture mapping and post-processing: Motion recovery structure is performed on the input continuous working face region map to obtain the pose of each image at the time of capture, and a sparse feature point cloud of the entire working face is constructed; based on the recovered camera pose, the obtained depth map is depth-projected to restore the dense point cloud information of the scene, obtaining a highly visualized 3D point cloud; surface reconstruction and texture mapping are performed on the dense point cloud to obtain a 3D surface model with the texture features of the original image; by orthogonally mapping the 3D surface model, a 2D global working face image is obtained, and post-processing is performed to determine the image scale information, ultimately obtaining a high-resolution global working face surface image.

[0082] S33. Image preprocessing is performed, including edge detection and image segmentation. A denoising algorithm for tunnel face joint feature data is constructed. For high-quality images, the Candy algorithm is used; for poor-quality images or those with poor recognition performance, a robust deep learning algorithm is used. The algorithm is trained on tunnel face joint samples to identify general features of the tunnel face. Image transmission is achieved via a local area network connected to the cloud, and images are stored either locally or in cloud object storage. In cloud processing, the Caffe visualization toolbox is used to extract layer-by-layer features and attach feature histograms for each layer. These feature histograms are insensitive to image scaling, translation, and rotation, exhibiting visual invariance and high robustness, accurately describing the morphological characteristics of the surrounding rock. Simultaneously, a skeletonization algorithm is used to analyze the tunnel face joints to determine joint parameters (number of joint groups and spacing, etc.), thereby obtaining key structural data of the tunnel face.

[0083] Based on the actual working environment and application needs of the hardware equipment, the image acquisition device of this invention mainly consists of four parts. First, it features a well-designed hardware structure and platform fixing method, utilizing materials such as carbon fiber to achieve overall lightweighting while maintaining high mechanical strength. Second, it selects an edge computing platform suitable for the software platform, enabling the practical deployment of the tunnel face digital image reconstruction algorithm with low power consumption. Third, it incorporates certain expansion capabilities, making it more suitable for operation in tunnel environments, such as adapting to modules like supplementary lighting, carrying frames, and embedded displays. Fourth, it prioritizes external sensors suitable for the working scenario, such as depth cameras, industrial cameras, and inertial measurement units, which can be securely fixed to the overall platform.

[0084] Furthermore, this invention also achieves visualization of the surrounding rock at the tunnel face based on 3D reconstruction technology. By acquiring real-time images of the tunnel face, 3D modeling is performed using colmap technology or by scanning with other devices. The model is then exported and displayed in a browser using WebGL (three.js technology), simultaneously annotating the integrity of the surrounding rock and potential collapse areas at the tunnel face. The annotated results are correlated with the 3D model of the tunnel face, using color rendering, texture mapping, and other methods to intuitively display the integrity of the surrounding rock. The 3D spatial visualization model of the tunnel face and a large dataset are integrated to form a multifunctional tunnel face platform. This platform can achieve 3D visualization of information such as joint parameters, the integrity of the surrounding rock, and the location and morphology of geological bodies at the tunnel face, providing a solid foundation for intelligent design of blasting parameters.

[0085] Step S4: Install fiber optic sensors around the tunnel face to monitor the changes in the mechanical properties and safety and stability of the rock mass near the tunnel face during blasting, and obtain data on the stress state of the surrounding rock.

[0086] Specifically, step S4 includes the following steps:

[0087] S41. Drill holes in the rock mass surrounding the working face; S42. Install fiber optic sensors and embed them into the surrounding rock to be monitored; S43. Connect the signal receiving device of the fiber optic sensors to ensure that the signal receiving device can receive and record the reflection spectrum signal of the fiber optic sensors in real time; S44. Set the threshold of the early warning device. When the detected deformation or stress of the surrounding rock exceeds the preset threshold, the early warning device will issue an early warning signal; S45. Process and analyze the monitored data, including demodulating, analyzing and processing the reflection spectrum signal to obtain real-time monitoring data of the deformation and stress state of the surrounding rock; S46. Take corresponding support measures based on the real-time monitoring data to ensure the stability of the surrounding rock.

[0088] Step S5: Construct a BP neural network model using geological data of the construction site, geological data of key locations, key structural data of the tunnel face, and stress state data of the surrounding rock to obtain geological structural indicators of the surrounding rock grade. Based on the geological structural indicators of the surrounding rock grade, determine the surrounding rock grade and adjust and optimize the tunnel blasting construction plan according to the surrounding rock grade.

[0089] Specifically, step S5 includes the following steps:

[0090] S51. Data Processing: Preprocess the collected geological data of the construction site, geological data of key locations, key structural data of the tunnel face, and stress state data of the surrounding rock. The preprocessing includes data cleaning, missing value handling, and outlier handling, and the data is normalized.

[0091] S52. Construct a BP neural network model: Construct a BP neural network model using the processed data. Take the geological data of the construction site, the geological data of key locations, the key structural data of the working face, and the stress state data of the surrounding rock as input variables, and take the integrity of the surrounding rock and the hardness of the rock as output variables. Select an appropriate neural network structure according to the scale and complexity of the data.

[0092] S53. Training the model: Use the training dataset to train the BP neural network model. During the training process, it is necessary to adjust the parameters of the neural network, including the learning rate, number of iterations, and weight matrix, in order to obtain the optimal model performance.

[0093] S54. Model Validation and Optimization: Use the validation dataset to validate and optimize the trained model. If there is a large error between the model's prediction results and the actual results, the model needs to be adjusted and optimized.

[0094] S55. Application Model: Based on the rock integrity and rock hardness obtained from the trained BP neural network model, the BQ rock mass basic quality index formula is used to determine the surrounding rock grade, and the tunnel blasting construction plan is adjusted and optimized according to the surrounding rock grade.

[0095] BQ = 100 + 3Rc + 250Kv

[0096] Where Rc is the uniaxial saturated compressive strength of the rock, and Kv is the integrity coefficient of the rock mass.

[0097] Example 2

[0098] The present invention also provides a multi-source data fusion system for determining the grade of surrounding rock in tunnels, including: a seismic wave system, a tunnel face structure data acquisition system, an optical fiber monitoring system, and a data fusion system.

[0099] Multiple locations were selected along the tunnel's forward path for drilling (e.g.) Figure 2The seismic wave borehole 201 shown is installed with a seismic wave system. Utilizing the reflection characteristics of seismic waves, geological data of the construction site is obtained, including interfaces of geological lithological changes, tectonic fracture zones, karst, and karst development zones. Based on the geological data of the construction site, key locations are selected for advance drilling borehole 301 to obtain geological data for these key locations.

[0100] Specifically, multiple locations along the tunnel's forward path are selected for drilling and the installation of seismic wave systems. The reflection characteristics of seismic waves are used to obtain geological data of the construction site. This includes: installing seismic wave receiving equipment on the ground; selecting suitable seismic source locations at the tunnel top and setting seismic wave transmitting equipment, along with setting the excitation parameters; considering the location of the target, geological conditions, and the required depth and resolution of the seismic survey lines; laying the survey lines along the pipeline's direction with uniform spacing; exciting seismic waves to propagate through the Earth's crust and reflect at different geological interfaces; recording the reflected signals using the seismic wave receiving equipment; processing and analyzing the signals to determine the location and morphology of the geological interfaces, and thus inferring the location, size, and shape of geological structures.

[0101] Furthermore, based on the geological data of the construction site, key locations were selected for advance drilling borehole 301 to obtain geological data for these key locations. This included monitoring drill rod vibration, changes in the color and flow rate of the flushing fluid, the sound of the impactor, drilling speed and its changes, rock powder, and stuck drill bit conditions during drilling to preliminarily determine whether adverse geological structures existed at the detection location. If adverse geological structures were preliminarily determined to exist, further detection and verification measures were taken. After the detection was completed, computer graphics processing technology was used to plot and analyze the detection results to obtain geological data for the key locations. The further measures included using a rotary core drill to obtain rock core samples to determine the type and nature of adverse geological structures. For the detection of joints and fissures, automatic joint and fissure recorders were used for measurement; for the detection of karst caves and weak interlayers, advance horizontal drilling methods were used.

[0102] The face structure data acquisition system includes an image acquisition device 303 and an image recognition and processing system. The image acquisition device is used to capture images of the face 302 to obtain images of the face area.

[0103] The image recognition and processing system is used for image-based 3D reconstruction. It reconstructs the 3D structure of the tunnel face scene, performs texture mapping and post-processing, and finally obtains a high-resolution global tunnel face surface image. The global tunnel face surface is then processed to accurately describe the morphological characteristics of the surrounding rock and obtain key structural data of the tunnel face. Specifically, motion reconstruction is performed on the input continuous tunnel face area map to obtain the pose of each image at the time of capture, and a sparse feature point cloud of the entire tunnel face is constructed. Based on the recovered camera pose, the acquired depth map is depth-projected to restore the dense point cloud information of the scene, obtaining a highly visualized 3D point cloud. Surface reconstruction and texture mapping are performed on the dense point cloud to obtain a 3D surface model with the texture features of the original image. Orthogonal mapping is performed on the 3D surface model to obtain a 2D global image of the tunnel face, and post-processing is performed to determine the image scale information, ultimately obtaining a high-resolution global tunnel face surface image. The entire tunnel face surface is then processed to accurately describe the morphological characteristics of the surrounding rock and obtain key structural data of the tunnel face. This includes: image preprocessing, edge detection and image segmentation of the preprocessed images, constructing a denoising algorithm for tunnel face joint feature data, training tunnel face joint samples to identify general features of the tunnel face, image transmission via local area network to the cloud, and image storage as either local or cloud object storage. In cloud processing, the Caffe visualization toolbox is used to extract layer-by-layer features and attach feature histograms for each layer to accurately describe the morphological characteristics of the surrounding rock and obtain key structural data of the tunnel face. The fiber optic monitoring system includes multiple fiber optic sensors 304, symmetrically installed around the tunnel face 302, used to monitor changes in the mechanical properties and safety stability of the rock mass near the tunnel face during blasting to obtain data on the stress state of the surrounding rock. The specific operation is as follows: Drill holes in the rock mass surrounding the working face 302; install fiber optic sensors 304 and embed them into the surrounding rock to be monitored; connect the signal receiving device of the fiber optic sensor 304 to ensure that the signal receiving device can receive and record the reflection spectrum signal of the fiber optic sensor 304 in real time; set the threshold of the early warning device, and when the monitored surrounding rock deformation or stress exceeds the preset threshold, the early warning device will issue an early warning signal; process and analyze the monitored data, including demodulating, analyzing and processing the reflection spectrum signal to obtain real-time monitoring data of the surrounding rock deformation and stress state; and take corresponding support measures based on the real-time monitoring data to ensure the stability of the surrounding rock.

[0104] The data fusion system uses data acquired from seismic wave systems, tunnel face structure data acquisition systems, and fiber optic monitoring systems, combined with geological data from key locations obtained through advanced drilling. It employs a BP neural network algorithm for data processing to obtain geological structural indicators of the surrounding rock grade. Based on these indicators, the system determines the surrounding rock grade and adjusts and optimizes the tunnel blasting construction plan accordingly. Specifically, it includes:

[0105] Data processing: The collected geological data of the construction site, geological data of key locations, key structural data of the tunnel face, and stress state data of the surrounding rock are preprocessed. The preprocessing includes data cleaning, missing value handling, and outlier handling, and the data is normalized.

[0106] Constructing a BP neural network model: Using the processed data, a BP neural network model is constructed. The geological data of the construction site, the geological data of key locations, the key structural data of the working face, and the stress state data of the surrounding rock are used as input variables, and the integrity of the surrounding rock and the hardness of the rock are used as output variables. The appropriate neural network structure is selected according to the scale and complexity of the data.

[0107] Training the model: The BP neural network model is trained using the training dataset. During the training process, the parameters of the neural network need to be adjusted, including the learning rate, number of iterations, and weight matrix, in order to obtain the optimal model performance.

[0108] Model validation and optimization: The trained model is validated and optimized using a validation dataset. If there is a large error between the model's prediction results and the actual results, the model needs to be adjusted and optimized.

[0109] Application Model: Based on the rock integrity and rock hardness obtained from the trained BP neural network model, the basic quality index formula of BQ rock mass is used to determine the surrounding rock grade, and the tunnel blasting construction plan is adjusted and optimized according to the surrounding rock grade.

[0110] BQ = 100 + 3Rc + 250Kv

[0111] Where Rc is the uniaxial saturated compressive strength of the rock, and Kv is the integrity coefficient of the rock mass.

[0112] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-source data fusion method for determining the surrounding rock grade of tunnels, characterized in that: Includes the following steps: S1. Select multiple locations along the tunnel's forward route to drill holes and install a seismic wave system. Utilize the reflection characteristics of seismic waves to obtain geological data of the construction site. The geological data of the construction site includes interfaces of geological and lithological changes, structural fracture zones, karst, and karst development zones. S2. Based on the geological data of the construction site, select key locations and conduct advanced drilling to obtain geological data of key locations, including the locations of joints, fissures, karst caves, and weak interlayer geological structures. S3. Take images of the working face and obtain key structural data of the working face through image recognition algorithms; S4. Install fiber optic sensors around the working face to monitor the changes in the mechanical properties and safety and stability of the rock mass near the working face during blasting, and obtain data on the stress state of the surrounding rock. S5. Using the geological data of the construction site, the geological data of key locations, the key structural data of the tunnel face, and the stress state data of the surrounding rock, a BP neural network model is constructed to obtain the geological structure index of the surrounding rock grade. Based on the geological structure index of the surrounding rock grade, the surrounding rock grade is judged, and the tunnel blasting construction plan is adjusted and optimized according to the surrounding rock grade. In step S3, images of the tunnel face are captured, and key structural data of the tunnel face are obtained through image recognition algorithms, including the following steps: S31. Use an image acquisition device to capture images of the working face to obtain images of the working face area; S32. Based on the image-based 3D reconstruction method, a high-resolution global working face surface image is obtained by reconstructing the 3D structure of the working face scene, performing texture mapping and post-processing: Motion recovery structure is performed on the input continuous working face region map to obtain the pose of each image at the time of capture, and a sparse feature point cloud of the entire working face is constructed; based on the recovered camera pose, the obtained depth map is depth-projected to restore the dense point cloud information of the scene, obtaining a highly visualized 3D point cloud; surface reconstruction and texture mapping are performed on the dense point cloud to obtain a 3D surface model with the texture features of the original image; orthogonal mapping is performed on the 3D surface model to obtain a 2D global working face image, and post-processing is performed to determine the image scale information, ultimately obtaining a high-resolution global working face surface image. S33. Perform image preprocessing, edge detection and image segmentation on the preprocessed image, construct a denoising algorithm for tunnel face joint feature data, train the tunnel face joint samples to identify general features of the tunnel face, transmit the image via local area network to the cloud, and store the image as a local or cloud object storage. In cloud processing, use the Caffe visualization toolbox to extract layer-by-layer features and attach feature histograms of each layer to accurately describe the morphological features of the surrounding rock and obtain key structural data of the tunnel face. In step S5, a BP neural network model is constructed using the geological data of the construction site, the geological data of key locations, the key structural data of the tunnel face, and the stress state data of the surrounding rock. Geological structural indicators of the surrounding rock grade are obtained. Based on these indicators, the surrounding rock grade is determined, and the tunnel blasting construction plan is adjusted and optimized according to the surrounding rock grade. This includes the following steps: S51. Data Processing: Preprocess the collected geological data of the construction site, geological data of key locations, key structural data of the tunnel face, and stress state data of the surrounding rock. The preprocessing includes data cleaning, missing value handling, and outlier handling, and the data is normalized. S52. Construct a BP neural network model: Construct a BP neural network model using the processed data. Take the geological data of the construction site, the geological data of key locations, the key structural data of the working face, and the stress state data of the surrounding rock as input variables, and take the integrity of the surrounding rock and the hardness of the rock as output variables. Select an appropriate neural network structure according to the scale and complexity of the data. S53. Training the model: Use the training dataset to train the BP neural network model. During the training process, it is necessary to adjust the parameters of the neural network, including the learning rate, number of iterations, and weight matrix, in order to obtain the optimal model performance. S54. Model Validation and Optimization: Use the validation dataset to validate and optimize the trained model. If there is a large error between the model's prediction results and the actual results, the model needs to be adjusted and optimized. S55. Application Model: Based on the rock integrity and rock hardness obtained from the trained BP neural network model, the BQ rock mass basic quality index formula is used to determine the surrounding rock grade, and the tunnel blasting construction plan is adjusted and optimized according to the surrounding rock grade. BQ = 100 + 3Rc + 250Kv; Where Rc is the uniaxial saturated compressive strength of the rock, and Kv is the integrity coefficient of the rock mass.

2. The multi-source data fusion method for determining the surrounding rock grade of tunnels according to claim 1, characterized in that: In step S1, multiple locations are selected along the tunnel's forward path for drilling and the installation of a seismic wave system. The reflection characteristics of seismic waves are used to obtain geological data of the construction site. This includes the following steps: S11. Install seismic wave receiving equipment on the ground, select a suitable seismic source location at the top of the tunnel to set up seismic wave transmitting equipment, and set the excitation parameters of the seismic waves. At the same time, the layout of the seismic survey lines should take into account the location of the detection target, geological conditions, and the depth and resolution requirements of the detection. The survey lines are laid along the direction of the pipeline, and the spacing between the survey lines is set evenly. S12. Induce seismic waves, which then propagate through the Earth's crust and are reflected at different geological interfaces; S13. Use seismic wave receiving equipment to record the reflected signals of seismic waves, process and analyze the signals to determine the location and shape of geological interfaces, and then infer the location, size and shape of geological structures.

3. The multi-source data fusion method for determining the surrounding rock grade of tunnels according to claim 2, characterized in that: In step S2, based on the geological data of the construction site, key locations are selected, and advanced drilling is conducted to obtain geological data for these key locations. These key locations include the locations of joints, fissures, karst caves, and weak interlayer geological structures. The process includes the following steps: S21. During the drilling process, pay attention to the vibration of the drill rod, the color and flow rate of the flushing fluid, the sound of the impactor, the drilling speed and its changes, rock powder, and stuck drill, in order to make a preliminary judgment on whether there are adverse geological structures at the detection location. S22. If it is initially determined that there are unfavorable geological structures, further measures shall be taken for detection and verification. S23. After the exploration is completed, computer graphics processing technology is used to draw and analyze the exploration results in order to obtain geological data of key locations; In step S22, a further measure is to use a rotary coring drill to drill and obtain rock core samples, thereby determining the type and nature of the adverse geological structure. Specifically, for the detection of joints and fissures, an automatic joint and fissure recorder is used for measurement, and for the detection of karst caves and weak interlayers, an advanced horizontal drilling method is used for detection.

4. The multi-source data fusion method for determining the surrounding rock grade of tunnels according to claim 1, characterized in that: In step S4, fiber optic sensors are installed around the tunnel face to monitor the changes in the mechanical properties and safety stability of the rock mass near the tunnel face during blasting, and to obtain data on the stress state of the surrounding rock. This includes the following steps: S41. Drill holes in the rock mass surrounding the working face; S42. Install fiber optic sensors and embed them into the surrounding rock to be monitored; S43. Connect the signal receiving device of the fiber optic sensors to ensure that the signal receiving device can receive and record the reflection spectrum signal of the fiber optic sensors in real time; S44. Set the threshold of the early warning device. When the detected deformation or stress of the surrounding rock exceeds the preset threshold, the early warning device will issue an early warning signal; S45. Process and analyze the monitored data, including demodulating, analyzing and processing the reflection spectrum signal to obtain real-time monitoring data of the deformation and stress state of the surrounding rock; S46. Take corresponding support measures based on the real-time monitoring data to ensure the stability of the surrounding rock.

5. A multi-source data fusion system for determining the surrounding rock grade of tunnels, characterized in that: include: The seismic wave system involves drilling at multiple locations around the tunnel's forward route and installing the seismic wave system. By utilizing the reflection characteristics of seismic waves, geological data of the construction site can be obtained. Based on the geological data of the construction site, key locations are selected for advanced drilling to obtain geological data of those key locations. A face structure data acquisition system includes an image acquisition device and an image recognition and processing system. The image acquisition device is used to capture images of the face to obtain images of the face area. The image recognition and processing system is used to: reconstruct the three-dimensional structure of the face scene based on an image 3D reconstruction method, perform texture mapping and post-processing, and finally obtain a high-resolution global face surface image. The global face surface is then processed to accurately describe the morphological characteristics of the surrounding rock and obtain key structural data of the face. The fiber optic monitoring system includes multiple fiber optic sensors symmetrically installed around the working face to monitor changes in the mechanical properties and safety and stability of the rock mass near the working face during blasting, in order to obtain data on the stress state of the surrounding rock. The data fusion system is used to process data based on seismic wave systems, face structure data acquisition systems, and fiber optic monitoring systems, combined with key location geological data obtained from advanced drilling, using a BP neural network algorithm to obtain geological structure indicators of the surrounding rock grade. Based on these indicators, the surrounding rock grade is determined, and the tunnel blasting construction plan is adjusted and optimized according to the surrounding rock grade. Among them, the image-based 3D reconstruction method reconstructs the 3D structure of the tunnel face scene, performs texture mapping and post-processing, and finally obtains a high-resolution global tunnel face surface image. This includes: performing motion reconstruction on the input continuous tunnel face region map to obtain the pose at the time of each image capture, and constructing a sparse feature point cloud of the entire tunnel face; performing depth projection on the acquired depth map based on the recovered camera pose to restore the dense point cloud information of the scene and obtain a highly visualized 3D point cloud; performing surface reconstruction and texture mapping on the dense point cloud to obtain a 3D surface model with the texture features of the original image; and obtaining a 2D image by orthogonally mapping the 3D surface model. The global image of the tunnel face is obtained and post-processed to determine the image scale information, ultimately resulting in a high-resolution global image of the tunnel face surface. The global tunnel face surface is then pre-processed, with edge detection and image segmentation performed on the pre-processed image. A denoising algorithm for tunnel face joint feature data is constructed and trained on tunnel face joint samples to identify general features of the tunnel face. Image transmission is achieved via a local area network connected to the cloud, and images are stored either locally or in cloud object storage. In cloud processing, the Caffe visualization toolbox is used to extract layer-by-layer features and attach feature histograms for each layer to accurately describe the morphological characteristics of the surrounding rock and obtain key structural data of the tunnel face. Based on data acquired from seismic wave systems, tunnel face structure data acquisition systems, and fiber optic monitoring systems, combined with key location geological data supplemented by advanced drilling, a BP neural network algorithm is used for data processing to obtain geological structural indicators of the surrounding rock grade. Based on these indicators, the surrounding rock grade is determined, and the tunnel blasting construction plan is adjusted and optimized according to the surrounding rock grade, including: Data processing: The collected geological data of the construction site, geological data of key locations, key structural data of the tunnel face, and stress state data of the surrounding rock are preprocessed. The preprocessing includes data cleaning, missing value handling, and outlier handling, and the data is normalized. Constructing a BP neural network model: Using the processed data, a BP neural network model is constructed. The geological data of the construction site, the geological data of key locations, the key structural data of the working face, and the stress state data of the surrounding rock are used as input variables, and the integrity of the surrounding rock and the hardness of the rock are used as output variables. The appropriate neural network structure is selected according to the scale and complexity of the data. Training the model: The BP neural network model is trained using the training dataset. During the training process, the parameters of the neural network need to be adjusted, including the learning rate, number of iterations, and weight matrix, in order to obtain the optimal model performance. Model validation and optimization: The trained model is validated and optimized using a validation dataset. If there is a large error between the model's prediction results and the actual results, the model needs to be adjusted and optimized. Application Model: Based on the rock integrity and rock hardness obtained from the trained BP neural network model, the basic quality index formula of BQ rock mass is used to determine the surrounding rock grade, and the tunnel blasting construction plan is adjusted and optimized according to the surrounding rock grade. BQ = 100 + 3Rc + 250Kv; Where Rc is the uniaxial saturated compressive strength of the rock, and Kv is the integrity coefficient of the rock mass.

6. The multi-source data fusion system for determining the surrounding rock grade of tunnels according to claim 5, characterized in that: By drilling at multiple locations along the tunnel's path and installing seismic wave systems, the geological data of the construction site is obtained by utilizing the reflection characteristics of seismic waves, including: Seismic wave receiving equipment is installed on the ground, and seismic wave transmitting equipment is set at a suitable source location at the top of the tunnel. The excitation parameters of the seismic waves are also set. Meanwhile, the layout of the seismic survey lines should take into account the location of the target and the geological conditions, as well as the depth and resolution requirements of the detection. The survey lines are laid along the direction of the pipeline, and the spacing between the survey lines is set evenly. This generates seismic waves, which propagate through the Earth's crust and are reflected at different geological interfaces. By using seismic wave receiving equipment to record the reflected signals of seismic waves, and processing and analyzing the signals, the location and shape of geological interfaces can be determined, thereby inferring the location, size and shape of geological structures.

Images