A method for blasting partition based on structural characteristics of rock mass

Abstract

The application provides a rock mass structure feature-based blastability zoning method, and belongs to the technical field of mine blasting, and comprises the following steps: collecting samples from a mine mining area for testing, obtaining rock physical and mechanical properties and rock mass structure information of the mine mining area, and inversely calculating spatial distribution of a concealed structure surface according to a depth prediction engine and a machine learning model to extract spatial distribution parameters of the structure surface; performing quantification and normalization processing on the extracted spatial distribution parameters of the structure surface; constructing a blasting simulation model based on a PFC discrete element method, dynamically adjusting joint filling parameters, taking the spatial distribution parameters of the structure surface as variables, simulating a blasting process, and outputting block size distribution parameters; calculating influence weights of the structure surface parameters on the block size distribution through a random forest regression algorithm; redefining a comprehensive sensitivity index according to the weights, performing 2.5m*2.5m grid division on a blasting area, calculating blastability scores in combination with geological structure constraints, obtaining a final zoning criterion table, and dividing a practical mining area into three-level blastability zones.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mine blasting, in particular, especially relates to a blastability zoning method based on rock mass structure characteristics. BACKGROUND

[0002] In the field of mine blasting engineering, the rock mass blastability refers to the degree of difficulty of rock mass breaking under the action of explosive explosion, which directly determines the blasting design parameters, explosive unit consumption and final fragmentation distribution. The structure surface such as joint, fracture and fault widely developed in the rock mass is the primary factor affecting the blastability: the occurrence of the structure surface controls the expansion direction of the blasting fissure, the density and spacing determine the initial size of the rock block, and the type and thickness of the filling material significantly weaken the propagation efficiency of the explosive stress wave. Therefore, quantitatively characterizing the structure surface and mapping the results to the blasting zoning are the basis for realizing "precision blasting".

[0003] At present, the empirical analogy method or the Protodyakonov coefficient method is generally used for blasting zoning at home and abroad. The former relies on the on-site experience of engineers, and gives qualitative conclusions such as "easy to explode" and "difficult to explode" according to the rock description and simple joint statistics; the latter converts the Protodyakonov coefficient through the uniaxial compressive strength of rock, and then divides the grades according to the specification table in combination with the joint spacing. 、 In recent years, some researches try to import the joint data obtained by three-dimensional laser scanning or borehole television into the finite element / discrete element model, predict the fragmentation distribution through numerical simulation, and indirectly evaluate the blastability by indicators such as

[0004] However, the existing technologies generally have the following defects:

[0005] (1) The structure surface parameter collection accuracy is insufficient, the surface investigation is disconnected with the deep drilling data, the concealed structure surface is difficult to identify, resulting in incomplete three-dimensional geological model;

[0006] (2) The numerical model has poor coupling with dynamic geological conditions, the mechanical behavior of joint filling material in the blasting process is complex, and the existing model mainly uses fixed parameters, which cannot be corrected in real time;

[0007] (3) There is lack of adaptive zoning regulation mechanism for structure surface sensitivity, the experience weight or single indicator is difficult to reflect the coupling effect of occurrence, spacing, filling characteristics and other factors, resulting in uneven distribution of blasting energy, excessive fragmentation, overbreak or underbreak and other problems.

[0008] The above defects seriously restrict the further improvement of blasting efficiency and safety, therefore, an technical scheme which can quantize the structure surface characteristics, dynamically optimize the blasting parameters and realize intelligent zoning is urgently needed. SUMMARY

[0009] According to the technical problems proposed above, a rock mass structure feature-based explosibility zoning method is provided.The present application realizes accurate release of blasting energy through high-precision geological data acquisition, multi-physical field coupling modeling, parameter sensitivity dynamic optimization and intelligent zoning regulation.

[0010] The technical means adopted by the present application are as follows:

[0011] A rock mass structure feature-based explosibility zoning method comprises:

[0012] S1, samples are collected from a mining area for testing to obtain rock physical and mechanical properties and rock mass structure information of the mining area, and a depth prediction engine and a machine learning model are used to inverse spatial distribution of a concealed structure surface to extract spatial distribution parameters of the structure surface, including occurrence parameters, joint spacing and filling coefficients;

[0013] S2, the extracted spatial distribution parameters of the structure surface are quantized and normalized;

[0014] S3, a blasting simulation model is constructed based on a PFC (Particle Flow Code) discrete element method, the joint spacing and the filling coefficients are dynamically adjusted, the occurrence parameters, the joint spacing and the filling coefficients are used as variables, a blasting process is simulated, and block size distribution parameters are output;

[0015] S4, a random forest regression algorithm is used to calculate weights of influences of the spatial distribution parameters of the structure surface on the block size distribution parameters;

[0016] S5, according to the weights, a comprehensive sensitivity index is defined, a 2.5m*2.5m grid division is performed on a blasting area, an explosibility score is calculated in combination with geological structure constraints, a final zoning criterion table is obtained, and a three-level explosibility zoning of an actual mining area is divided.

[0017] Further, step S1 specifically comprises:

[0018] S11, structure surface measurement is performed, surface joint parameters including occurrence, density, opening, and filling thickness are obtained through unmanned aerial vehicle photogrammetry, a surface scanning library is established, deep fracture parameters including tendency, inclination, and filling distribution are verified through borehole imaging, a borehole verification library is established, and a three-dimensional joint topology library is established based on the surface scanning library and the borehole verification library;

[0019] S12, lithology measurement is performed, rock density and tensile strength are obtained through a Brazilian split test under static load, inversion parameter combinations are generated through Latin hypercube sampling, inversion errors for identifying positions of concealed structure surfaces are calculated by comparing measured data and inversion prediction results.

[0020] Further, the inversion error of identifying the position of the concealed structural plane is controlled by a machine learning model confidence threshold, and the inversion error is set to be less than or equal to 0.3 m. When the inversion error is greater than 0.3 m, reacquisition is triggered.

[0021] Further, in step S2, specifically comprising:

[0022] S21, calculate the filling coefficient of the structural plane, and the calculation formula is as follows:

[0023]

[0024] Among them, represents the filling coefficient of the structural plane; represents the thickness of the filling; represents the joint opening degree of the measuring point;

[0025] S22, count the vertical spacing of adjacent structural planes, and calculate the average joint spacing in different regions;

[0026] S23, normalize the occurrence parameters including the dip direction angle and the dip angle , and convert them into spherical coordinate system, and the formula is as follows:

[0027]

[0028]

[0029]

[0030] Among them, , , represents the normalized spherical coordinate, reflects the size of the dip angle and the occurrence deviation degree; , is the dip direction angle corresponding to the principal stress direction.

[0031] Further, in step S3, specifically comprising:

[0032] S31, select the occurrence parameters, joint spacing and filling coefficient as variables, use the dynamic damage model for the rock matrix, use the shear model for the joint element, embed the concealed structural plane, and simulate the blasting process in the open-pit mine area by the discrete element method;

[0033] S32, use the joint element and the non-continuous deformation analysis to dynamically adjust the joint spacing and the filling coefficient, configure the mine blasting structural plane simulation scheme, and obtain the block size distribution parameters including the value, the non-uniformity coefficient by the simulation results.

[0034] Further, step S4 specifically comprises:

[0035] S41, constructing a data set by a random forest algorithm, and establishing a random forest regression model, as follows:

[0036]

[0037] wherein, represents the random forest regression model, represents the input factor occurrence parameter, represents the input factor joint spacing, represents the input factor filling coefficient, represents the output factor value, represents the output factor uneven coefficient , represents the sample number;

[0038] S42, training the random forest regression model, and calculating the benchmark error, as follows:

[0039]

[0040] wherein, represents the benchmark error; represents the sample number; represents the output factor value of the i-th sample, represents the predicted value of the i-th sample on the output factor value , represents the output factor uneven coefficient of the i-th sample, represents the predicted value of the i-th sample on the output factor uneven coefficient , represents the predicted value of the i-th sample on the output factor uneven coefficient ;

[0041] S43, based on the calculated benchmark error, calculating the error increase value, as follows:

[0042]

[0043] wherein, represents the error increase value; represents the model error after using the shuffled features ;

[0044] ​​S44, based on the calculated error increase value, the weight of the structural plane spatial distribution parameter on the block size distribution parameter is calculated, and the formula is as follows:

[0045]

[0046] Wherein, Indicates the weight; Indicates the total number of decision trees, Indicates the error increase calculated in the th tree.

[0047] Further, step S5 specifically includes:

[0048] S51, the weight of the structural plane spatial distribution parameter on the block size distribution parameter calculated in step S44 is normalized, and the formula is as follows:

[0049]

[0050] Wherein, Indicates the normalized weight, Indicates the sum of the weights of the three input factors;

[0051] S52, calculate the sensitivity index, the formula is as follows:

[0052]

[0053] Wherein, Indicates the sensitivity index, Indicates the normalized value of the occurrence parameter, Indicates the normalized value of the joint spacing, Indicates the normalized value of the filling coefficient, Indicates the weight coefficient of the occurrence parameter, Indicates the weight coefficient of the joint spacing, Indicates the weight coefficient of the filling coefficient;

[0054] S53, according to the calculated sensitivity index, define the comprehensive sensitivity index , the calculation formula is as follows:

[0055]

[0056] Wherein, Indicates the minimum value of the sensitivity index, Indicates the maximum value of the sensitivity index;

[0057] S54, according to the sensitivity index and the geological structure constraint, divide into three levels, specifically:

[0058] Will Divide into difficult explosion area; divide Divide into moderate explosion area; divide Divide into easy explosion area.

[0059] Compared with the prior art, the present application has the following advantages:

[0060] 1. The rock mass structure feature-based explosibility partition method provided by the present application can improve the blasting energy distribution accuracy, improve the block uniformity, and reduce the large block rate compared with the existing partition method. Through dynamic weight optimization and adaptive partition, the blasting energy consumption is reduced by 15-20%.

[0061] 2. The rock mass structure feature-based explosibility partition method provided by the present application obtains integrated surface- underground joint fissure data through unmanned aerial vehicle photogrammetry coupling drilling imaging technology, and then inverses through an XGBoost+LSTM hybrid machine learning model to realize three-dimensional high-precision quantitative expression of rock mass structure features, solves the problems of disconnection of deep and shallow data and omission of concealed structure in the traditional method, and provides real and reliable geological input for subsequent blasting design.

[0062] 3. The rock mass structure feature-based explosibility partition method provided by the present application couples the spherical coordinate system normalization algorithm with joint spacing and filling coefficient, and for the first time, maps the three core parameters of occurrence parameter, joint spacing and filling coefficient to dimensionless space, eliminates the interference of dimensional difference on weight calculation, realizes the comparability and superposition of structure surface parameters, and significantly improves the objectivity and accuracy of subsequent sensitivity analysis and weight distribution.

[0063] 4. The rock mass structure feature-based explosibility partition method provided by the present application constructs a “dynamic damage + shear slip” coupling model through a PFC discrete element platform, and automatically adjusts the joint shear strength according to the real-time filling coefficient during simulation, truly reproduces the energy attenuation and fissure expansion process when the explosion stress wave meets the structure surface, and greatly improves the reliability of blasting block size prediction.

[0064] 5. The rock mass structure feature-based explosibility partition method provided by the present application trains the typical working conditions and simulation data through a random forest regression model, quantifies the weight of the influence of the occurrence parameter, joint spacing and filling coefficient on the block size distribution through the feature scrambling method, establishes a nonlinear and renewable weight system, breaks through the limitation of fixed and unchanged empirical weight, and realizes the dynamic optimization of parameter sensitivity.

[0065] 6. The rock mass structure feature-based explosibility zoning method provided by the application superimposes the weight and the normalization parameter through the comprehensive sensitivity index, and combines with the geological structure constraint to perform grid-based explosibility scoring on the blasting area, and finally divides it into three levels of zoning; field industrial test shows that the zoning method significantly improves the blasting energy utilization rate and the slope forming quality.

[0066] 7. The rock mass structure feature-based explosibility zoning method provided by the application guarantees real-time updating of the model input data through the "error threshold-triggered resampling" closed-loop control mechanism, synchronizes the zoning result with the field geological changes, and truly realizes the intelligentization and self-adaptive regulation and control of the rock mass explosibility zoning, thereby providing a popular and replicable technical paradigm for safe and efficient mining of the open-pit mine.

[0067] Based on the above reasons, the application can be widely popularized in the field of mine blasting. BRIEF DESCRIPTION OF DRAWINGS

[0068] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0069] Figure 1 The method flowchart of the application. DETAILED DESCRIPTION

[0070] In order to enable the personnel in the technical field to better understand the application scheme, the following will combine the drawings in the embodiments of the application to clearly and completely describe the technical solutions in the embodiments of the application. Obviously, the described embodiments are only some of the embodiments of the application, but not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor should belong to the protection scope of the application.

[0071] It should be noted that the terms "include" and "have" and any variations thereof in the specification and claims of the application and the above drawings are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to only those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to the process, method, product or device.

[0072] As shown in Figure 1 The application provides a rock mass structure feature-based explosibility zoning method, which comprises:

[0073] S1, collect samples from the mining area for testing to obtain the petrophysical and mechanical properties of the mining area and rock mass structure information, and inversely predict the spatial distribution of the concealed structure surface based on a depth prediction engine (XGBoost) and a machine learning model (LSTM), and extract the spatial distribution parameters of the structure surface, including the occurrence parameters, joint spacing, and filling coefficient;

[0074] S2, quantize and normalize the extracted spatial distribution parameters of the structure surface;

[0075] S3, construct a blasting simulation model based on the PFC discrete element method, dynamically adjust the joint spacing and filling coefficient, and take the occurrence parameters, joint spacing, and filling coefficient as variables to simulate the blasting process and output the block size distribution parameters;

[0076] S4, calculate the weight of the influence of the spatial distribution parameters of the structure surface on the block size distribution parameters through a random forest regression algorithm;

[0077] S5, define a comprehensive sensitivity index according to the weight, divide the blasting area into a 2.5m x 2.5m grid, calculate the blastability score in combination with the geological structure constraint, obtain the final partition criterion table, and divide the actual mining area into three blastability partitions.

[0078] In specific implementation, as a preferred embodiment of the present application, step S1 specifically includes:

[0079] S11, perform structure surface measurement, obtain surface joint parameters including occurrence, density, opening, and filler thickness through unmanned aerial vehicle photogrammetry, establish a surface scanning library, verify deep fracture parameters including tendency, inclination, and filler distribution through borehole imaging, establish a borehole verification library, and establish a three-dimensional joint topology library based on the surface scanning library and the borehole verification library;

[0080] S12, perform lithology measurement, perform a Brazilian split test on the rock under static load to test the rock density and tensile strength, generate an inversion parameter combination through Latin hypercube sampling, compare the measured data with the inversion prediction results, and set an inversion error for identifying the position of the concealed structure surface.

[0081] In the present embodiment, a standard reference (such as a 20cm scale) is placed during shooting to facilitate later quantification of fracture size, a 3D point cloud model is generated through multi-angle photos to quantify fracture density, FDTD inversion is performed using GPRMax software, the XGBoost model parameters are set as: maximum tree depth 8, tree number 200, learning rate 0.05; the LSTM model unit number = 64, dropout rate 0.2, training round 300, the inversion error for identifying the position of the concealed structure surface (error ≤0.3 m) is controlled through the confidence threshold of the machine learning model, and when the error is greater than 0.3m, re-collection is triggered. Measured directly using digital calipers, recording range 0.5-15mm. Filler thickness Measured using laser rangefinder in combination with probe method, filler mainly clay, thickness range 1-8mm.

[0082] In specific implementation, as a preferred embodiment of the present application, in step S2, specifically includes:

[0083] S21, calculate the filling coefficient of the structural plane, the calculation formula is as follows:

[0084]

[0085] Wherein, represents the filling coefficient of the structural plane; represents the filler thickness; represents the joint opening degree of the measurement point; in this embodiment, the measured filling coefficient ranges from 0.1 to 0.8.

[0086] S22, count the vertical spacing of adjacent structural planes, and calculate the average joint spacing in different areas;

[0087] S23, normalize the occurrence parameters including the dip angle and the dip angle , and convert them into spherical coordinate system, the formula is as follows:

[0088]

[0089]

[0090]

[0091] Wherein, , , represents the normalized spherical coordinate, reflects the dip angle size and occurrence deviation; , is the dip angle corresponding to the principal stress direction.

[0092] In this embodiment, step S2 combines the correlation analysis of various indexes of the structural plane, and the selection of the basic index data of the structural plane specifically includes: normalizing the geometric parameters, occurrence characteristics, filling characteristics, rock mass strength and other parameters of the joint, calculating the filling coefficient of each structural plane from the joint opening width and the filler thickness, calculating the spacing of each joint according to the obtained joint information of the ore rock, and normalizing the occurrence parameters dip angle ( °) and dip angle ( After normalization, spherical coordinates are obtained, with the principal stress direction corresponding to the rock movement direction. Specific parameters are shown in the table below:

[0093]

[0094] In this embodiment, the dataset consists of 5 sets of simulated experimental data, and the input features are normalized to the [0,1] interval.

[0095] In a specific implementation, as a preferred embodiment of the present invention, step S3 specifically includes:

[0096] S31. Select the attitude parameters, joint spacing and filling coefficient as variables, adopt the dynamic damage model for the rock matrix, adopt the shear model for the joint elements, embed the hidden structural plane, and simulate the explosion process in the open-pit mine blasting using the discrete element method.

[0097] S32. Using joint elements and discontinuous deformation analysis, the joint spacing and filling coefficient are dynamically adjusted to configure a simulation scheme for mine blasting structural surfaces. The block size distribution parameters are then obtained from the simulation results, including... Value, coefficient of uniformity .

[0098] In this embodiment, PFC numerical simulation software is used to simulate mine bench blasting. Based on the standardized samples in the data table of step S2, several sets of joint units with different attitudes and joint spacings are embedded. Typical parameter combinations are selected to configure the blasting simulation scheme. Among them, the rock matrix parameters are: density Elastic modulus 50 GPa. Joint parameters: shear stiffness 1 GPa / m, normal stiffness 2 GPa / m, friction angle... Set different fill factors The joint shear strength is dynamically adjusted based on the filling coefficient. (Shear strength decreases by 30% at this time). Simulation results: Value (80% based on ore size), uniformity coefficient .

[0099] In a specific implementation, as a preferred embodiment of the present invention, step S4 specifically includes:

[0100] S41. Construct a dataset using the random forest algorithm and establish a random forest regression model, as follows:

[0101]

[0102] in, This represents a random forest regression model. Indicates the input factor, attitude parameter. Indicates the input factor joint spacing. denotes the input factor filling coefficient, denotes the output factor value, denotes the output factor uneven coefficient , denotes the sample number;

[0103] S42, train the random forest regression model and calculate the benchmark error, the formula is as follows:

[0104]

[0105] wherein, denotes the benchmark error; denotes the sample number; denotes the output factor value of the i-th sample, denotes the predicted value of the i-th sample on the output factor value denotes the predicted value of the i-th sample on the output factor uneven coefficient value denotes the output factor uneven coefficient of the i-th sample, denotes the predicted value of the i-th sample on the output factor uneven coefficient

[0106] S43, based on the calculated benchmark error, calculate the error increase value, the formula is as follows:

[0107]

[0108] wherein, denotes the error increase value; denotes the model error after using the shuffled features

[0109] S44, based on the calculated error increase value, calculate the weight of the influence of the structural plane spatial distribution parameter on the block size distribution parameter, the formula is as follows:

[0110]

[0111] wherein, denotes the weight; denotes the total number of decision trees, denotes the error increase amount calculated in the i-th tree.

[0112] ​​​​​​In this embodiment, the feature shuffling method is implemented using the `permutation_importance` function of Scikit-learn. The Scikit-learn library is used, with the number of decision trees T=100, the maximum depth 10, and the MSE baseline error set to 0.18. The weights for the attitude parameter are determined to be 0.45, the joint spacing to be 0.35, and the filling coefficient to be 0.20.

[0113] In a specific implementation, as a preferred embodiment of the present invention, step S5 specifically includes:

[0114] S51. Normalize the weights of the influence of the spatial distribution parameters of the structural surfaces calculated in step S44 on the block size distribution parameters, as shown in the following formula:

[0115]

[0116] in, This represents the weights after normalization. This represents the summation of the weights of the three input factors;

[0117] S52. Calculate the sensitivity index using the following formula:

[0118]

[0119] in, Indicates the sensitivity index. This represents the normalized value of the attitude parameter. The normalized value representing the joint spacing. This represents the normalized value of the fill factor. The weighting coefficients representing the attitude parameters. The weighting coefficient represents the joint spacing. This represents the weighting factor of the fill factor;

[0120] S53. Based on the calculated sensitivity index, define the comprehensive sensitivity index. The calculation formula is as follows:

[0121]

[0122] in, This represents the minimum value of the sensitivity index. This indicates the maximum value of the sensitivity index;

[0123] S54. Based on the sensitivity index and geological structural constraints (fault distance), the area is divided into three levels, as follows:

[0124] Will Divide into low-explosive zones; Divide into medium-explosive zones; The area is divided into an explosive area.

[0125] In the embodiment, the charge quantity of the difficult explosive area (S>0.7) is increased by 20%, the standard hole net parameters are adopted in the medium explosive area (0.4≤S≤0.7), and the charge quantity of the easy explosive area (S<0.4) is reduced by 15%.

[0126] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for blasting partitioning based on structural features of rock mass, characterized in that, The method comprises the following steps: S1, collecting samples from the mining area for testing to obtain the petrophysical and mechanical properties of the mining area and the information of rock mass structure, and inversing the spatial distribution of hidden structural planes according to the depth prediction engine and machine learning model to extract the spatial distribution parameters of the structural planes, including the occurrence parameters, joint spacing and filling coefficient; S2, quantifying and normalizing the extracted spatial distribution parameters of the structural planes; S3, constructing a blasting simulation model based on the PFC discrete element method, dynamically adjusting the joint spacing and filling coefficient, taking the occurrence parameters, joint spacing and filling coefficient as variables, simulating the blasting process and outputting the block size distribution parameters; S4, calculating the weight of the influence of the spatial distribution parameters of the structural planes on the block size distribution parameters through the random forest regression algorithm; S5, defining a comprehensive sensitivity index according to the weight, dividing the blasting area into a 2.5m×2.5m grid, calculating the blastability score in combination with the geological structure constraint to obtain a final partition criterion table, and dividing the actual mining area into three-level blastability partitions.

2. The method of claim 1, wherein, Step S1 specifically comprises: S11, performing structural plane measurement, obtaining surface joint parameters including occurrence, density, opening, and filling thickness through unmanned aerial vehicle photogrammetry to establish a surface scanning library, verifying deep fracture parameters including inclination, dip angle, and filling distribution through borehole imaging to establish a borehole verification library, and establishing a three-dimensional joint topology library based on the surface scanning library and the borehole verification library; S12, performing lithology measurement, performing Brazilian splitting test on rocks under static load to obtain rock density and tensile strength, generating inversion parameter combinations through Latin hypercube sampling, comparing the measured data with the inversion prediction results, and setting an inversion error for identifying the position of hidden structural planes.

3. The method of claim 2, wherein, The inversion error for identifying the position of hidden structural planes is controlled through a machine learning model confidence threshold, and when the inversion error is greater than 0.3m, reacquisition is triggered.

4. The method of claim 1, wherein, In step S2, specifically comprising: S21, calculating the filling coefficient of the structural plane, and the calculation formula is as follows: wherein, represents a structure plane filling coefficient; represents a filling thickness; represents a joint opening degree at a measurement point; S22, calculating the average joint spacing in different regions by counting the vertical spacing of adjacent structural planes; S23, normalize the occurrence parameters including the dip angle and the dip angle and convert them into the spherical coordinate system, as follows: wherein , , denotes the normalized spherical coordinates, reflecting the inclination size and deviation degree of occurrence; , is the inclination angle corresponding to the principal stress direction.

5. The method of claim 1, wherein the method further comprises: Step S3 specifically comprises: S31, selecting the occurrence parameters, joint spacing and filling coefficient as variables, using a dynamic damage model for the rock matrix and a shear model for the joint unit, embedding hidden structural planes, and simulating the blasting process in the open-pit mining area through the discrete element method; S32, using joint units and non-continuous deformation analysis to dynamically adjust joint spacing and filling coefficient, configuring mine blasting structural plane simulation scheme, and obtaining block size distribution parameters including value, non-uniformity coefficient .

6. The method of claim 1, wherein, Step S4 specifically comprises: S41, constructing a data set through the random forest algorithm and establishing a random forest regression model, as follows: wherein, represents a random forest regression model, represents an input factor occurrence parameter, represents an input factor joint spacing, represents an input factor fill factor, represents an output factor value, represents an output factor non-uniformity factor , represents a number of samples; S42, training the random forest regression model and calculating the benchmark error, as follows: wherein, represents the reference error; represents the number of samples; represents the output factor value of the th sample; represents the predicted value of the th sample at the output factor value represents the output factor non-uniformity coefficient represents the predicted value of the th sample at the output factor non-uniformity coefficient ;​​​​​ S43, calculating the error increase value based on the calculated benchmark error, as follows: wherein, represents an error increase value; represents the model error after using the scrambled features ​ S44, calculating the weight of the influence of the spatial distribution parameters of the structural planes on the block size distribution parameters based on the calculated error increase value, as follows: wherein, represents a weight; represents the total number of decision trees, represents the error increase calculated in the th tree.

7. The method of claim 1, wherein, Step S5 specifically comprises: S51, normalizing the weight of the influence of the spatial distribution parameters of the structural planes on the block size distribution parameters calculated in step S44, as follows: wherein, denotes the normalized weight, denotes the summation of the three input factor weights; S52, calculating the sensitivity index, as follows: wherein, represents a sensitivity index, represents a normalized value of the occurrence parameter, represents a normalized value of the joint spacing, represents a normalized value of the filling coefficient, represents a weight coefficient of the occurrence parameter, represents a weight coefficient of the joint spacing, represents a weight coefficient of the filling coefficient; S53. Defining a composite sensitivity index from the calculated sensitivity indices The formula is as follows: wherein denotes the minimum value of the sensitivity index, denotes the maximum value of the sensitivity index; S54, dividing the three-level partitions according to the sensitivity index and the geological structure constraint, specifically as follows: Divide into non-hazardous zone; divide into moderate hazard zone; divide into hazardous zone. Divide into non-hazardous zone; divide into moderate hazard zone; divide into hazardous zone. Divide into non-hazardous zone; divide into moderate hazard zone; divide into hazardous zone. Divide into non-hazardous zone; divide into moderate hazard

Images