Mining disturbance and rock mass weathering degree evaluation method based on rock core length intelligent identification

Abstract

The invention discloses a mining disturbance and rock mass weathering degree evaluation method based on rock core length intelligent identification, and belongs to the technical field of mine safety monitoring and rock mass engineering detection, and the method comprises the steps: constructing a multi-layer neural network model, identifying a single rock core section from a rock core image before mining, and determining the rock mass weathering degree of the single rock core section; calculating the average length of the rock core before mining in different depth intervals of each drill hole; geometric parameter information of a slope surface structural plane after mining is measured, a discrete fracture network model is constructed, and the average length of a virtual drilling rock core after mining is obtained through a virtual drilling method; and converting the average length of the rock core into fracture density, and constructing a fracture density change rate to quantitatively evaluate the influence of mining disturbance and weathering on the rock mass. According to the method, the influence of engineering disturbance and weathering is effectively separated from the total fracture development through the actual and virtual comparison thought, quantitative evaluation of the rock mass damage source is achieved, and the result is more accurate.

Description

Technical Field

[0001] This invention belongs to the field of mine safety monitoring and rock mass engineering detection technology, and specifically discloses a method for evaluating mining disturbance and rock mass weathering degree based on intelligent identification of rock core length. Background Technology

[0002] China's rapid economic development has greatly promoted infrastructure construction and energy development, with many projects involving geotechnical engineering. However, due to the complexity of natural rock mass structures and the disturbance caused by slope excavation, mine slope stability has always been a challenge in the field of rock mechanics, making timely and accurate monitoring of regional changes in mine slopes crucial. With the increasing depletion of shallow mineral resources, open-pit mining activities are continuously expanding into deeper areas with more complex geological conditions. In this process, the engineering rock mass not only suffers from direct mechanical disturbances caused by mining activities (such as blasting, unloading, and stress redistribution), but is also exposed to the climatic environment for extended periods, experiencing continuous weathering effects (physical, chemical, and biological). These two processes—intense artificial disturbance and slow natural degradation—often couple and intensify each other, profoundly altering the structural integrity, mechanical properties, and hydraulic characteristics of the rock mass, thus posing a severe challenge to slope stability.

[0003] Traditional rock mass quality assessment systems, such as RMR (Rock Mechanics Classification), Q system (Barton Rock Mass Quality Classification), or GSI (Geological Strength Index), are primarily based on static descriptions of rock mass structure (e.g., number of fracture groups, spacing, continuity, etc.) and rock block strength. These methods are effective for describing rock masses in a relatively stable state, but they struggle to dynamically and quantitatively characterize the rock mass damage evolution process caused by engineering disturbances and continuous weathering. The average core length, calculated by dividing the total length of all core segments extracted from a borehole of a certain length by the number of core segments, reflects the average size of continuous rock blocks within the rock mass and can provide information about the degree of fracture and joint development within the rock mass.

[0004] Therefore, there is an urgent need for an efficient, objective, and quantitative method to quantitatively assess the degree of damage to rock mass structure caused by mining disturbance and weathering, so as to provide a scientific basis for mine safety design and disaster prevention and control. Summary of the Invention

[0005] The present invention aims to solve the technical problem of the difficulty in efficiently and quantitatively separating and evaluating the degree of damage to rock mass structure caused by mining disturbance and weathering in the existing technology.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A method for evaluating mining disturbance and rock mass weathering degree based on intelligent core length identification includes the following steps:

[0008] S1. Construct a multi-layer neural network model to identify individual core segments from the core images of each borehole before mining and calculate the average length of the core segments before mining in different depth ranges of each borehole.

[0009] S2. Measure the geometric parameters of the exposed structural surface on the slope after mining, construct a discrete fracture network model, arrange the same virtual borehole trajectory in the discrete fracture network model corresponding to the actual borehole location in S1, obtain the virtual core segment generated after the virtual borehole trajectory is cut by discrete fracture, and calculate the average length of the virtual core after mining in different depth ranges of each borehole.

[0010] S3. Based on the inverse relationship between the average core length and fracture density, the average core length before mining and the average virtual core length after mining are converted into the original fracture density before mining, respectively. Simulated fracture density after mining Construct the fracture density change rate (%) as an evaluation indicator:

[0011] ;

[0012] The fracture density variation rate at different depths of each borehole was calculated and used to evaluate the degree of damage and deterioration at various locations in the rock mass.

[0013] In a specific embodiment, in S1, the multilayer neural network model includes a single-row core identification module and a core segment identification module. The single-row core identification module is used to identify all single-row core targets from the core image, and the core segment identification module is used to identify each core segment of the single-row core from the single-row core targets and output the pixel size of each core segment.

[0014] In a specific embodiment, the process of constructing the discrete fracture network model in S2 is as follows: using the statistical distribution of the geometric parameters of the structural surface as a constraint, a structural model of the region to be analyzed is constructed in three-dimensional space, and a large number of discrete fractures are randomly generated by the Monte Carlo method to obtain a discrete fracture network model that conforms to the actual statistical law.

[0015] In a specific embodiment, in step S2, the intersection points of the virtual borehole trajectory and all fracture surfaces in the discrete fracture network model are calculated using a spatial analytical geometry method, thereby obtaining the virtual core segment generated after being cut by the discrete fractures.

[0016] Compared with the prior art, the present invention has the following advantages:

[0017] (1) Quantification and precision: By comparing the "actual" and "virtual" approaches, the effects of engineering disturbance and weathering are effectively separated from the development of total fractures, thus achieving a quantitative evaluation of the rock mass damage source and resulting in more precise results.

[0018] (2) High efficiency and automation: The use of deep learning technology to replace traditional manual core logging greatly improves the efficiency and objectivity of core information extraction and is suitable for the rapid processing of massive data.

[0019] (3) Intuitive and practical: The final evaluation index of crack density change rate has a clear physical meaning, providing an intuitive and reliable basis for engineering design and safety decision-making.

[0020] (4) Foresight and scalability: This method deeply integrates artificial intelligence, 3D modeling and geotechnical engineering professional analysis, providing a new technical path for intelligent geotechnical engineering exploration and monitoring. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the method of the present invention. Detailed Implementation

[0022] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and should not be construed as limiting the scope of the invention.

[0023] This embodiment provides a method for evaluating mining disturbance and rock mass weathering degree based on intelligent core length identification, such as... Figure 1 As shown, it includes the following steps:

[0024] S1. A deep learning algorithm is used to construct and train a multi-layer neural network model to automatically identify and segment individual core segments from pre-mining core images of each borehole. The trained multi-layer neural network model is then used to batch process pre-mining core images taken on-site to identify the length and number of all core segments at different depth intervals in each borehole, thereby calculating the average core length at different depth intervals for each borehole. Specifically:

[0025] S101. Data Preprocessing: Standardize the file names of the core images from each borehole before mining and extract information to establish a mapping relationship between the photos and their spatial locations. The extracted information includes the borehole number and the start and end depths.

[0026] S102. A deep learning algorithm is used to construct and train a multi-layer neural network model for automatically identifying and segmenting individual core segments from pre-mining borehole core images. The trained multi-layer neural network model is then used to batch process pre-mining borehole core images taken on-site to identify the length and number of all core segments at different depth intervals in each borehole core. This multi-layer neural network model is existing technology, specifically disclosed in Chinese invention patent "An Intelligent Method for Identifying RQD from Borehole Core Photographs" (Patent No. ZL202011137412.1, Authorization Announcement Date 2023.07.28). This application directly references the multi-layer neural network model disclosed in that patent and makes appropriate improvements to the model for identifying core segments and their pixel sizes.

[0027] The multi-layer neural network model includes a single-row core identification module and a core segment identification module. The single-row core identification module uses the `core_band_identification` sub-model, and the core segment identification module uses the `core_segment_identification` sub-model. The pre-processed core images are input into the trained multi-layer neural network model, which outputs the pixel dimensions of each core segment in each image, providing a basis for subsequent core length cataloging. The construction and training process of the multi-layer neural network model is as follows:

[0028] S1021. Create a rock core image dataset; use the labeling tool LabelMe to label a single row of rock cores in the rock core image to obtain data labels as learning samples for the Mask R-CNN deep learning network, and randomly divide the learning samples into training set and test set at a ratio of 3:1; the labeling of a single row of rock cores is to connect continuous points to form a closed polygon, mark the edge contour of the target rock core, and convert it into a labeling file of a set format;

[0029] S1022. Using transfer learning, the training samples are input into the Mask R-CNN deep learning network for training to obtain a pre-trained model. During the training process, the characteristic of transfer learning—that it can transfer a learner trained on a large number of data samples to a domain with only a small number of data samples—is utilized without causing a significant performance degradation. The mask_rcnn_R_50_FPN_3x pre-trained model trained on the COCO dataset is used. Initial parameters are set for the pre-trained model, and the learning rate is lowered. The training set is input into the Mask-RCNN network. After forward propagation, the prediction results of the single-row core location are obtained, and the results are compared with the data labels to obtain the validation loss value. Backpropagation is then performed, and the gradient of the Mask R-CNN deep learning network is updated using the mini-batch gradient descent method until the loss value reaches a preset threshold or the number of iterations reaches a preset value, thus completing the training of the core_band_identification sub-model for identifying single-row cores.

[0030] S1023. Input the test set into the trained core_band_identification sub-model, use the region proposal network to generate N predicted bounding boxes in the core image, calculate the probability value of the predicted bounding box belonging to the target category, arrange the probability values ​​from largest to smallest, take the predicted bounding box with the largest probability value as the benchmark, calculate the intersection-union ratio of the remaining predicted bounding boxes with it, if it is greater than a set threshold, remove this predicted bounding box, repeat the calculation to identify all targets in the image, realize the iterative-traversal process to suppress redundancy, and each target corresponds to a predicted bounding box, that is, obtain the anchor box. The core_band_identification sub-model identifies the boundary of the target from each anchor box and fills it with color blocks to accurately locate the target, that is, obtain the mask;

[0031] S1024. Using the acquired single-row core image, continue to label all core segments using LabelMe; repeat S1022 and S1023 to obtain the core_segment_identification sub-model, which is used to identify core segments from the single-row core image and output their pixel size;

[0032] S103. Length Calculation and Statistics: Based on the scale in the core image, the pixel size is converted into the actual physical length to obtain the length of each core segment at different starting and ending burial depths in the core image; the length and total number of each core segment are counted according to the borehole depth range to calculate the average length of the core segments in each depth range before mining; specifically in this embodiment, the core segment length information of a certain borehole depth range is shown in Table 1, and the average length of the core segments in different depth ranges before mining of this borehole is shown in Table 2.

[0033] Table 1. Initial data on the average length of some core samples from borehole z431

[0034]

[0035] Table 2. Average length processing data of borehole Z431 core.

[0036]

[0037] S2. Geometric parameters of the exposed structural surfaces on the slope surface after mining were measured in the field. A discrete fracture network (DFN) model was constructed using the Monte Carlo method to characterize the spatial statistical distribution characteristics of the rock mass structural surfaces after mining. Virtual borehole trajectories, identical to those in the actual borehole locations in S1, were arranged in this DFN model. The drilling process was simulated through geometric calculations to obtain virtual core segments generated after being cut by discrete fractures within the DFN model. The average length of the virtual core segments at different depth intervals of each borehole after mining was calculated. Specifically:

[0038] S201. Obtain geometric parameter information of structural surfaces through slope surface measurement, including dip direction, dip angle, trace length, spacing, linear density, volume density, number of structural surface groups and quantity.

[0039] S202. Using the statistical distribution of structural surface geometric parameters as constraints, a structural model of the area to be analyzed is constructed in three-dimensional space using MATLAB. A large number of discrete fractures are randomly generated using the Monte Carlo method to simulate the impact of mining disturbances and secondary weathering on the rock mass structure, resulting in a DFN model that conforms to actual statistical laws. The statistical distribution of geometric parameters includes the mean, variance, and probability density function. The use of the Monte Carlo method to randomly generate discrete fractures enables the construction of a network model that reflects the characteristics of a fracture system under real geological conditions. Its core advantage lies in replacing deterministic measurements with statistical laws.

[0040] S203. In the DFN model space, at the location corresponding to the actual borehole in S1, arrange virtual borehole trajectories with the same location, direction, and depth as the actual borehole. Calculate the intersection points of the virtual borehole trajectories with all fracture surfaces in the DFN model using spatial analytical geometry. The intersection points divide the borehole trajectory into multiple virtual core segments. Based on the same depth range as in S1, count the length and number of each virtual core segment in each borehole, and calculate the average length of the virtual cores in different depth ranges of each borehole after mining.

[0041] S3. Based on the inverse relationship between the average core length and fracture density, the average core length before mining obtained in S1 and the virtual core length after mining obtained in S2 are converted into the original fracture density before mining for each depth interval of the corresponding borehole. Simulated fracture density after mining All units are lines / m; construct the fracture density change rate. (%) as an evaluation indicator:

[0042] ;

[0043] The corresponding depth ranges of each borehole and Substituting into the above formula, the fracture density change rate at different depth intervals of each borehole is calculated. This rate is used to quantitatively characterize the increase in fracture development caused by mining disturbance and secondary weathering, in order to evaluate the degree of damage and deterioration at each location of the rock mass, and thus evaluate the impact of mining disturbance and weathering on the rock mass. Specifically, a larger fracture density change rate indicates a greater impact from mining disturbance and weathering at that borehole, i.e., a greater degree of rock mass damage. Table 3 shows the fracture density changes before and after mining in some boreholes in this embodiment:

[0044] Table 3. Changes in average core length before and after drilling in some boreholes.

[0045]

[0046] As shown in Table 3, the fissure density change rate of the rock mass at all borehole locations in this embodiment was greater than 1 after mining, indicating that the number of fissures in the slope rock mass increased under the disturbance of mining and continuous weathering. At the same time, there were significant differences in the fissure density change rate corresponding to different borehole locations, showing obvious spatial differences in the degree of impact on the rock mass. For example, the fissure density change rate at the rock mass location corresponding to the depth range of borehole zk676 (75.6~80.95m) was 6.77%, while the fissure density change rate at the rock mass location corresponding to the depth range of borehole zk634 (128.1~134m) reached 1445.28%, the latter being more than 200 times that of the former. Therefore, it is evident that the degree of impact of mining disturbance and weathering on different locations of the slope rock mass varies significantly, and the corresponding degree of rock mass damage also differs markedly. The fracture density change rate index used in this embodiment can quantitatively assess the degree of damage at different locations in the rock mass. The assessment results can provide a scientific and reliable quantitative basis for subsequent safety design, slope stability control and mine disaster early warning.

[0047] The above description represents a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the technical scope disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for evaluating mining disturbance and rock mass weathering degree based on intelligent core length identification, characterized in that, include: S1. Construct a multi-layer neural network model to identify individual core segments from the core images of each borehole before mining and calculate the average length of the core segments before mining in different depth ranges of each borehole. S2. Measure the geometric parameters of the exposed structural surface on the slope after mining, construct a discrete fracture network model, arrange the same virtual borehole trajectory in the discrete fracture network model corresponding to the actual borehole location in S1, obtain the virtual core segment generated after the virtual borehole trajectory is cut by discrete fracture, and calculate the average length of the virtual core after mining in different depth ranges of each borehole. S3. Convert the average length of the pre-mining core and the average length of the virtual core after mining to the original fracture density before mining. Simulated fracture density after mining Construct the fracture density change rate : ； The rate of change of fracture density at different depths in each borehole is calculated to evaluate the degree of damage at various locations in the rock mass.

2. The method for evaluating mining disturbance and rock mass weathering degree based on intelligent core length identification according to claim 1, characterized in that, In S1, the multi-layer neural network model includes a single-row core identification module and a core segment identification module. The single-row core identification module is used to identify all single-row core targets from the core image, and the core segment identification module is used to identify each core segment of the single-row core from the single-row core targets and output the pixel size of each core segment.

3. The method for evaluating mining disturbance and rock mass weathering degree based on intelligent core length identification according to claim 2, characterized in that, The multi-layer neural network model is built based on the Mask-RCNN deep learning network.

4. The method for evaluating mining disturbance and rock mass weathering degree based on intelligent core length identification according to claim 1, characterized in that, In S2, the process of constructing the discrete fracture network model is as follows: using the statistical distribution of the geometric parameters of the structural surface as a constraint, a structural model of the region to be analyzed is constructed in three-dimensional space, and a large number of discrete fractures are randomly generated to obtain a discrete fracture network model that conforms to the actual statistical law.

5. The method for evaluating mining disturbance and rock mass weathering degree based on intelligent core length identification according to claim 4, characterized in that, A large number of discrete fractures were randomly generated using the Monte Carlo method.

6. The method for evaluating mining disturbance and rock mass weathering degree based on intelligent core length identification according to claim 1, characterized in that, In step S2, the intersection points of the virtual borehole trajectory and all fracture surfaces in the discrete fracture network model are calculated using spatial analytical geometry methods, thereby obtaining the virtual core segment generated after being cut by the discrete fractures.

7. The method for evaluating mining disturbance and rock mass weathering degree based on intelligent core length identification according to claim 1, characterized in that, In step S3, based on the inverse relationship between the average core length and the fracture density, the average core length before mining and the average virtual core length after mining are respectively converted into the original fracture density before mining. Simulated fracture density after mining .

Images