Blasting zoning method based on geophysical exploration technology
By combining multi-source data joint inversion and dynamic response parameter acquisition with intelligent geomechanical modeling and machine learning, the problems of single data and reliance on experience in blasting design are solved, achieving high-precision blasting zoning and self-optimization, thus improving the applicability and safety of blasting design.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- UNIV OF SCI & TECH LIAONING
- Filing Date
- 2026-02-11
- Publication Date
- 2026-05-08
AI Technical Summary
Existing technologies in blasting design suffer from problems such as limited data and distorted models, disconnect between dynamic and static elements, and reliance on experience, resulting in poor blasting effects and a lack of self-optimization capabilities, as well as poor applicability.
By employing multi-source data acquisition and joint inversion, combined with dynamic response parameter acquisition, and through intelligent geomechanical modeling and machine learning mapping models, a high-resolution three-dimensional geomechanical model is generated to achieve intelligent zoning and design, and to establish a feedback optimization closed loop for self-evolution.
It achieves high-precision blasting zoning, improves the accuracy and reliability of blasting design, adapts to different geological conditions, has self-optimization capabilities, and significantly improves the safety and efficiency of engineering construction.
Smart Images
Figure CN121994091A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of geophysical exploration technology, and more specifically, to a method for explosive zoning based on geophysical exploration technology. Background Technology
[0002] Blasting operations are a common construction method in engineering projects, widely used in mining, road construction, site leveling, and other engineering scenarios. However, different geological conditions have a significant impact on blasting results. Inappropriate blasting design may lead to poor blasting effects, which not only increases the difficulty and cost of subsequent construction but may also cause safety hazards such as flying rocks and seismic wave damage.
[0003] Common geophysical exploration methods include seismic exploration, electrical resistivity tomography (ERT), and gravity exploration. Seismic exploration utilizes the propagation characteristics of artificially generated seismic waves in the subsurface medium, such as reflection and refraction, to infer subsurface geological structures and rock properties. ERT studies the differences in the electrical properties of the subsurface medium, such as resistivity and dielectric constant, to detect geological structures. Gravity exploration analyzes the density distribution of subsurface materials based on changes in the Earth's gravitational field, thereby inferring geological structures.
[0004] The existing technology has the following technical defects, specifically:
[0005] 1. Limited Data and Distorted Models: Traditional methods rely on single geophysical data for independent inversion, lacking collaborative constraints and verification between different types of data. This leads to severe ambiguity in the inversion process, resulting in static rock mass models with low spatial resolution and blurred structural boundaries, failing to accurately depict complex geological structures, such as the difficulty in accurately distinguishing fracture zones from aquifers.
[0006] 2. Disconnect between dynamic and static loads and inaccurate predictions: Existing technologies completely ignore the nonlinear response of rock masses under dynamic blasting loads, relying solely on static physical parameters for zonal predictions, leading to a severe disconnect between mechanical judgments and actual conditions. Due to the lack of quantitative characterization of the dynamic fracturing behavior of rock masses, the distribution of blasting energy is mismatched with the rock mass's blast resistance, easily resulting in problems such as over-fracture or under-blasting.
[0007] 3. Reliance on Experience and Rigidity: From estimating key mechanical parameters to designing the final blasting scheme, the entire process relies heavily on the personal experience of engineers, lacking standardized and quantitative intelligent decision support. Furthermore, the established models are rigid and cannot be self-verified and iteratively optimized by collecting blasting effect data, resulting in poor technology reusability and inconsistent accuracy across different projects. Summary of the Invention
[0008] The purpose of this invention is to provide a blasting zoning method based on geophysical exploration technology to solve the problems mentioned in the background art.
[0009] To achieve the above objectives, the present invention aims to provide a blasting zoning method based on geophysical exploration technology, including: S1, multi-source data acquisition and joint inversion: synchronously acquiring multi-source data of the target area, and using a cross-gradient joint inversion algorithm to perform collaborative inversion on the multi-source data to generate a three-dimensional model of static rock mass physical parameters.
[0010] S2. Acquisition of dynamic response parameters: A small-yield calibration blast is carried out in a typical zone of the target area, and a vibration monitoring network is deployed to collect blast vibration signals. Based on the attenuation characteristics and full waveform information of the vibration signals, the three-dimensional spatial distribution of dynamic damage factors characterizing the dynamic fracture behavior of the rock mass is calculated by full waveform inversion technology.
[0011] S3. Intelligent Geomechanical Modeling: The three-dimensional model of static rock mass physical parameters is integrated with the dynamic damage factor. The integrated multi-dimensional feature parameters are used as input to a pre-trained machine learning mapping model. The output is a high-resolution three-dimensional geomechanical model of the target area, which directly includes key mechanical parameters such as compressive strength and elastic modulus.
[0012] S4. Intelligent Zoning and Design: Based on the three-dimensional geomechanical model, combined with the preset multi-index fusion zoning rules based on mechanical parameters and dynamic response parameters, the blastability level is divided, and corresponding blasting hole network parameters and charge quantity suggestions are automatically generated for each level.
[0013] S5. Feedback Optimization Closed Loop: After blasting, blasting effect evaluation data is collected, and the blasting effect evaluation data is compared with the zoning prediction effect to generate a feedback signal. Based on the feedback signal, the machine learning mapping model is incrementally learned and the parameters are fine-tuned to realize the self-evolution of the zoning model.
[0014] As a further improvement to this technical solution, the multi-source data of the target area includes at least seismic wave exploration data, DC resistivity data, and induced polarization data.
[0015] As a further improvement to this technical solution, the specific implementation method for generating a three-dimensional model of static rock mass physical parameters is as follows: Based on the seismic wave exploration data and the DC resistivity method and induced polarization method data, a cross-gradient joint inversion algorithm is used for processing to discretize the target area into a three-dimensional grid model and construct forward simulation systems for the seismic wave field and the stable current field respectively. A unified objective function is established, which includes the L2 norm fitting term of the seismic wave travel time residual, the L2 norm fitting term of the resistivity observation data, and the cross-gradient structure constraint term. The wave velocity value and resistivity value of each three-dimensional grid node are iteratively updated based on the least squares optimization algorithm. When the objective function converges to a preset threshold, the iteration stops, and a three-dimensional model of static rock mass physical parameters is generated.
[0016] As a further improvement to this technical solution, the three-dimensional spatial distribution of the dynamic damage factor is specifically implemented as follows: using the wave velocity distribution in the three-dimensional model of the static rock mass physical parameters as the initial model, the vibration time history curves of each measuring point recorded by the vibration monitoring network in the blasting test as the observation data, and using the full waveform inversion technology, the wave velocity parameters of the initial model are continuously adjusted through an iterative optimization algorithm to minimize the residual between the synthetic seismic record obtained by forward modeling based on the current model and the field measured vibration record. In each iteration, the dynamic damage factor is calculated based on the update amount of the wave velocity model. When the full waveform inversion iteration converges, the distribution model of the dynamic damage factor in the three-dimensional space of the target area is output.
[0017] As a further improvement to this technical solution, the pre-trained machine learning mapping model is specifically implemented as follows: Core sampling points are systematically arranged within the target area. The true values of uniaxial compressive strength, elastic modulus, and cohesion mechanical parameters are obtained through indoor rock mechanics tests on the core samples. Simultaneously, multi-source fusion feature parameters corresponding to the spatial location of each sampling point are extracted. These parameters include at least P-wave velocity, S-wave velocity, and resistivity parameters obtained from the three-dimensional model of static rock mass physical parameters, and dynamic damage factor values obtained from the three-dimensional spatial distribution of dynamic damage factors. The multi-source fusion feature parameters are used as input feature vectors, and the true values of the mechanical parameters are used as supervised learning targets. The gradient boosting decision tree algorithm is employed for model training. The optimal hyperparameter combination, including the maximum tree depth, learning rate, and subsampling ratio, is determined through grid search and cross-validation to obtain a training model capable of establishing a nonlinear mapping relationship from geophysical features to rock mechanical properties.
[0018] As a further improvement to this technical solution, the specific method for classifying explosiveness levels is as follows:
[0019] Multiple discrimination indicators for each three-dimensional grid node are extracted from the three-dimensional geomechanical model, including static mechanical indicators, dynamic response indicators, and native structural indicators. Then, based on the preset multi-indicator fusion partitioning rules, logical judgment and level classification are performed. The levels include at least difficult-to-blast areas, easy-to-blast areas, and tectonic influence areas. Finally, a three-dimensional partitioning model with different blastability level labels is generated that corresponds to the three-dimensional geomechanical model space.
[0020] As a further improvement to this technical solution, the automatic generation of corresponding blasting hole network parameters and charge quantity suggestions is specifically implemented as follows: the three-dimensional partition model containing different blasting level labels is coupled with a built-in expert knowledge base. This knowledge base pre-stores the quantitative mapping relationship between different blasting levels and blasting design parameters. Based on this mapping relationship, the system automatically generates a customized blasting design scheme for each partition unit and outputs a digital blasting design drawing containing hole position coordinates, drilling depth, hole network density, explosive type, and charge quantity distribution.
[0021] As a further improvement to this technical solution, the blasting effect evaluation data includes at least blasting block size distribution data, blasting morphology and integrity data, blasting harmful effect monitoring data, and comprehensive economic and technical indicator data.
[0022] As a further improvement to this technical solution, the specific method for generating the feedback signal is as follows:
[0023] The actual observed values in the blasting effect evaluation data, including the average block size and large block rate in the blasting block size distribution data, and the base rate and over-excavation amount in the blasting morphology data, are compared with the expected effect values predicted based on the three-dimensional partitioning model and the generated blasting design parameters. The difference between each observed value and the predicted value is calculated using a preset loss function, which is the weighted sum of squares of each difference item. The calculated total loss function value and the spatial location information of each partition in the corresponding three-dimensional partitioning model together constitute the feedback signal.
[0024] Compared with the prior art, the beneficial effects of the present invention are reflected in:
[0025] 1. Multi-source fusion and dynamic-static synergy: This invention utilizes a cross-gradient joint inversion algorithm to deeply fuse structural information from seismic waves and resistivity data, effectively overcoming multiple solutions and generating a high-resolution static physical model with clear geological significance. Furthermore, through calibrated blasting and full-waveform inversion techniques, it achieves for the first time the precise quantification of the three-dimensional spatial distribution of "dynamic damage factors," constructing a complete rock mass characterization system of "static structure + dynamic response."
[0026] 2. Intelligent Zoning and Quantitative Design: Based on a high-precision three-dimensional geomechanical model, this invention applies intelligent zoning rules that integrate multiple indicators to achieve the scientific division of "difficult-to-blast zones," "easy-to-blast zones," and other similar areas. Subsequently, by calling the built-in expert knowledge base, the system can automatically generate quantitative hole mesh parameters and charge quantity suggestions for different zones, transforming blasting design from "experience-based art" to "data science," significantly improving the accuracy and reliability of the design scheme.
[0027] 3. Closed-Loop Optimization and Broad Adaptability: This invention innovatively establishes a closed-loop process of "implementation-evaluation-feedback-optimization." By comparing the predicted values with the block size, vibration, and other effect data after each blast, feedback signals are generated to drive the machine learning model for incremental learning. This enables the entire system to possess continuous self-evolution capabilities, not only improving the zoning accuracy of this project in the long term but also greatly enhancing the technology's strong adaptability and reusability under different geological conditions and engineering scenarios. Attached Figure Description
[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0029] Figure 1 This is a schematic diagram of the implementation steps of the method of the present invention. Detailed Implementation
[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0031] Example: Please refer to Figure 1 As shown, a blasting zoning method based on geophysical exploration technology is provided, including: S1, multi-source data acquisition and joint inversion: multi-source data of the target area are acquired simultaneously, and the multi-source data are collaboratively inverted using a cross-gradient joint inversion algorithm to generate a three-dimensional model of static rock mass physical parameters. This model simultaneously includes wave velocity distribution and resistivity distribution information to eliminate the ambiguity of a single geophysical interpretation.
[0032] In one specific embodiment, the multi-source data of the target area includes at least seismic wave exploration data, DC resistivity data, and induced polarization data.
[0033] Seismic wave exploration data includes at least shot point coordinates and elevation information, receiver point coordinates and elevation information, observation system data, seismic gathers, first arrival time data, S-wave arrival time data, and wavefield characteristic data.
[0034] DC resistivity data: should include at least electrode coordinates and elevation information, device type and parameters, power supply current intensity, potential difference between measuring electrodes, and apparent resistivity data.
[0035] Excitation polarization method data: including at least the primary field potential difference, secondary field decay curve / charge rate data, polarizability / charge rate, decay curve characteristic parameters, etc.
[0036] In one specific embodiment, the method for generating a three-dimensional model of static rock mass physical parameters is as follows: based on the seismic wave exploration data and the DC resistivity method and induced polarization method data, a cross-gradient joint inversion algorithm is used to process the data, discretize the target area into a three-dimensional grid model, and construct forward simulation systems for the seismic wave field and the stable current field respectively. A unified objective function is established, which includes the L2 norm fitting term of the seismic wave travel time residual, the L2 norm fitting term of the resistivity observation data, and the cross-gradient structure constraint term. The wave velocity value and resistivity value of each three-dimensional grid node are iteratively updated based on the least squares optimization algorithm. When the objective function converges to a preset threshold, the iteration stops, and a three-dimensional model of static rock mass physical parameters is generated.
[0037] The cross-gradient joint inversion algorithm and the least squares optimization algorithm are existing technologies and will not be elaborated on here. The cross-gradient structural constraint term is implemented by calculating and minimizing the cross product modulus of the spatial gradient vector of the wave velocity model and the spatial gradient vector of the resistivity model, thereby forcing the structural boundaries of the two physical parameter models to maintain consistency in three-dimensional space.
[0038] The specific implementation method of constructing the forward modeling system for the seismic wave field and the steady current field is as follows: Based on the discretized three-dimensional mesh model, corresponding mathematical and physical equation solution systems are established respectively; for the forward modeling of the seismic wave field, the acoustic wave equation or elastic wave equation is used as the governing equation, spatial discretization is performed by the finite difference method or finite element method, and absorbing boundary conditions are set to eliminate artificial boundary reflections. The complete wave field of the seismic wave propagating from the shot point to the receiver point is simulated by iterative calculation in the time domain; for the forward modeling of the steady current field, the Poisson equation is used as the governing equation, the finite element method is used to handle complex terrain and physical property boundaries, and the potential distribution formed in the underground space after the current is injected by the power supply electrode is calculated by solving a large sparse linear equation system; both forward modeling systems improve the solution efficiency through parallel computing technology and provide an accurate basis for calculating the Jacobian matrix or sensitivity matrix for subsequent joint inversion.
[0039] The L2 norm fitting term of the seismic wave travel time residuals is used to measure the degree of matching between the seismic wave velocity prediction model and the measured data. Its calculation process is as follows: First, based on the current iterative three-dimensional wave velocity model, the theoretical travel time of seismic waves from each shot point to each receiver point is calculated using ray tracing or forward modeling of the wave equation, generating a theoretical travel time dataset. Then, this theoretical travel time dataset is compared point-by-point with the first arrival time dataset of seismic waves acquired through actual field observations, and the residual (i.e., the difference between the observed value and the theoretical value) is calculated for each data point. Finally, the L2 norm (i.e., the square root of the sum of the squares of all residuals) is taken for this residual vector; this value is the value of this fitting term. The smaller the value of this term, the more consistent the current wave velocity model's explanation of seismic wave propagation phenomena is with the actual situation.
[0040] The L2 norm fitting term of the resistivity observation data is used to evaluate the ability of the resistivity model to reproduce actual electrical exploration observation data. The calculation process is as follows: Based on the current iteration of the three-dimensional resistivity model, the theoretical potential difference between the measuring electrodes under a specific electrode arrangement is predicted through stable current field forward modeling (usually using the finite element method), and the theoretical apparent resistivity value is further calculated to generate a theoretical apparent resistivity dataset. Next, this theoretical dataset is compared point-by-point with the apparent resistivity observation dataset obtained from actual field measurements to obtain the data residual for each measurement point. Finally, the L2 norm of this residual vector is calculated as the value of this term. By minimizing this term, the inverted resistivity model is driven to reproduce the actual electrical exploration observation results as accurately as possible.
[0041] The cross-gradient structural constraint term is the core of achieving structural coupling between seismic and electrical resistivity data. It does not directly fit the data but instead imposes a geometric structural constraint. Its calculation method is as follows: at each element node of the 3D mesh model, the spatial gradient vectors of the current wave velocity model and resistivity model are calculated respectively. Then, the cross product (or vector product) of these two gradient vectors is calculated, and its magnitude is taken. This magnitude is physically proportional to the sine of the angle between the two gradient vectors. This term is zero when the directions of property changes (i.e., structural boundaries) of the two models are completely consistent or completely opposite. Finally, the magnitude values at all nodes in the entire model space are summed or the L2 norm is taken to obtain the value of the cross-gradient constraint term. Minimizing this term means that the structural boundaries of the wave velocity model and the resistivity model are forced to be spatially aligned during the inversion process, thereby effectively reducing the ambiguity of geophysical interpretation and yielding a geologically more reasonable model.
[0042] The preset threshold is set by professionals, for example, based on two objective criteria: first, the data fitting residual decreases to a range comparable to the error level of the observed data, indicating that the model has fully explained the effective signal; second, the change in the objective function value or model parameter update between consecutive iterations is less than a minimum value, indicating that the inversion process has reached a stable convergence state.
[0043] S2. Acquisition of dynamic response parameters: A small-yield calibration blast is carried out in a typical zone of the target area, and a vibration monitoring network is deployed to collect blast vibration signals. Based on the attenuation characteristics and full waveform information of the vibration signals, the three-dimensional spatial distribution of dynamic damage factors characterizing the dynamic fracture behavior of the rock mass is calculated by full waveform inversion technology.
[0044] The full waveform inversion technique is an existing technology and will not be described in detail here.
[0045] In one specific embodiment, the three-dimensional spatial distribution of the dynamic damage factor is specifically implemented as follows: using the wave velocity distribution in the three-dimensional model of the static rock mass physical parameters as the initial model, the vibration time history curves of each measuring point recorded by the vibration monitoring network in the blasting test as the observation data, and employing full waveform inversion technology, the wave velocity parameters of the initial model are continuously adjusted through an iterative optimization algorithm to minimize the residual between the synthetic seismic record obtained from the forward simulation based on the current model and the field measured vibration record. In each iteration, the dynamic damage factor is calculated based on the update amount of the wave velocity model. The specific calculation formula is as follows: ,in The P-wave velocity in the current iteration. The initial P-wave velocity is used as the initial value. When the full waveform inversion iteration converges, the distribution model of the dynamic damage factor in the three-dimensional space of the target area is output. This model quantitatively characterizes the spatial heterogeneity of the damage and fragmentation degree inside the rock mass under the action of blasting dynamic load.
[0046] S3. Intelligent Geomechanical Modeling: The three-dimensional model of static rock mass physical parameters is integrated with the dynamic damage factor. The integrated multi-dimensional feature parameters are used as input to a pre-trained machine learning mapping model. The output is a high-resolution three-dimensional geomechanical model of the target area, which directly includes key mechanical parameters such as compressive strength and elastic modulus.
[0047] In one specific embodiment, the pre-trained machine learning mapping model is implemented as follows: Core sampling points are systematically arranged within the target area. The true values of uniaxial compressive strength, elastic modulus, and cohesion mechanical parameters are obtained through indoor rock mechanics tests on the core samples. Simultaneously, multi-source fusion feature parameters corresponding to the spatial location of each sampling point are extracted. These parameters include at least P-wave velocity, S-wave velocity, and resistivity parameters obtained from the three-dimensional model of static rock mass physical parameters, and dynamic damage factor values obtained from the three-dimensional spatial distribution of dynamic damage factors. The multi-source fusion feature parameters are used as input feature vectors, and the true values of the mechanical parameters are used as supervised learning targets. A gradient boosting decision tree algorithm is employed for model training. The optimal hyperparameter combination, including the maximum tree depth, learning rate, and subsampling ratio, is determined through grid search and cross-validation to obtain a training model capable of establishing a nonlinear mapping relationship from geophysical features to rock mechanical properties.
[0048] The gradient boosting decision tree algorithm is existing technology and will not be described in detail here.
[0049] S4. Intelligent Zoning and Design: Based on the three-dimensional geomechanical model, combined with the preset multi-index fusion zoning rules based on mechanical parameters and dynamic response parameters, the blastability level is divided, and corresponding blasting hole network parameters and charge quantity suggestions are automatically generated for each level.
[0050] In one specific embodiment, the method for classifying blastability levels is as follows: Multiple discrimination indicators for each three-dimensional grid node are extracted from the three-dimensional geomechanical model, including static mechanical indicators, dynamic response indicators, and native structural indicators. Then, based on preset multi-indicator fusion partitioning rules, logical judgment and level classification are performed. The levels include at least difficult-to-blast zones, easily-blasted zones, and tectonic-affected zones. If a region simultaneously satisfies a P-wave velocity higher than a first threshold, a uniaxial compressive strength higher than a second threshold, and a dynamic damage factor lower than a third threshold, then the region is classified as a "difficult-to-blast zone." If a region simultaneously satisfies a P-wave velocity lower than a fourth threshold, a uniaxial compressive strength lower than a fifth threshold, and a dynamic damage factor higher than a sixth threshold, then it is classified as an "easily-blasted zone." If a region satisfies a resistivity lower than a seventh threshold and a dynamic damage factor higher than an eighth threshold, then regardless of its wave velocity, it is classified as a "tectonic-affected zone." Finally, a three-dimensional partitioning model corresponding to the three-dimensional geomechanical model space and containing different blastability level labels is generated. This directly guides the design of differentiated blasting parameters.
[0051] In one specific embodiment, the automatic generation of corresponding blasting hole network parameters and charge quantity suggestions is implemented as follows: the three-dimensional partition model containing different blastability level labels is coupled with a built-in expert knowledge base. This knowledge base pre-stores quantitative mapping relationships between different blastability levels and blasting design parameters. Based on this mapping relationship, the system automatically generates customized blasting design schemes for each partition unit: for "difficult-to-blast zones," smaller hole spacing and row spacing, higher explosive consumption per unit, and high-power explosive types are automatically calculated and recommended; for "easy-to-blast zones," larger hole spacing and row spacing, and lower explosive consumption per unit are automatically calculated and recommended; for "tectonic influence zones," in addition to adjusting the charge quantity, the structural boundaries are automatically identified in the three-dimensional model, and instructions for setting vibration damping holes or adopting segmented charge structures are generated; a digital blasting design drawing containing hole position coordinates, drilling depth, hole network density, explosive type, and charge quantity distribution is output.
[0052] S5. Feedback Optimization Closed Loop: After blasting, blasting effect evaluation data is collected, and the blasting effect evaluation data is compared with the zoning prediction effect to generate a feedback signal. Based on the feedback signal, the machine learning mapping model is incrementally learned and the parameters are fine-tuned to realize the self-evolution of the zoning model.
[0053] In one specific embodiment, the blasting effect evaluation data includes at least blasting block size distribution data, blasting morphology and integrity data, blasting harmful effect monitoring data, and comprehensive economic and technical indicator data.
[0054] In one specific embodiment, the generation of the feedback signal is implemented by comparing the actual observed values in the blasting effect evaluation data, including the average block size and large block rate in the blasting block size distribution data, and the base rate and over-excavation amount in the blasting morphology data, with the expected effect values predicted based on the three-dimensional partitioning model and the generated blasting design parameters. A preset loss function is used to calculate the difference between each observed value and the predicted value, where the loss function is the weighted sum of squares of each difference item. The calculated total loss function value and the corresponding spatial location information of each partition in the three-dimensional partitioning model together constitute the feedback signal. This feedback signal, as a supervisory signal, is input into the pre-trained machine learning mapping model, driving it to fine-tune the model weight parameters through online learning to narrow the gap between prediction and actual measurement, thereby achieving adaptive optimization of the partitioning prediction model.
[0055] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.
Claims
1. A method for explosive zoning based on geophysical exploration technology, characterized in that, include: S1. Multi-source data acquisition and joint inversion: Multi-source data of the target area are acquired simultaneously, and the multi-source data are jointly inverted using the cross-gradient joint inversion algorithm to generate a three-dimensional model of static rock mass physical parameters. S2. Acquisition of dynamic response parameters: A small-yield calibration blast is carried out in a typical zone of the target area, and a vibration monitoring network is deployed to collect blast vibration signals. Based on the attenuation characteristics and full waveform information of the vibration signals, the three-dimensional spatial distribution of dynamic damage factors characterizing the dynamic fracture behavior of the rock mass is calculated by full waveform inversion technology. S3. Intelligent Geomechanical Modeling: The three-dimensional model of static rock mass physical parameters is integrated with the dynamic damage factor. The integrated multi-dimensional feature parameters are used as input to a pre-trained machine learning mapping model. The output is a high-resolution three-dimensional geomechanical model of the target area, which directly includes key mechanical parameters such as compressive strength and elastic modulus. S4. Intelligent Zoning and Design: Based on the three-dimensional geomechanical model, combined with the preset multi-index fusion zoning rules based on mechanical parameters and dynamic response parameters, the blasting level is divided, and corresponding blasting hole network parameters and charge quantity suggestions are automatically generated for each level. S5. Feedback Optimization Closed Loop: After blasting, blasting effect evaluation data is collected, and the blasting effect evaluation data is compared with the zoning prediction effect to generate a feedback signal. Based on the feedback signal, the machine learning mapping model is incrementally learned and the parameters are fine-tuned to realize the self-evolution of the zoning model.
2. The explosive zoning method based on geophysical exploration technology according to claim 1, characterized in that, The multi-source data for the target area includes at least seismic wave exploration data, DC resistivity data, and induced polarization data.
3. The explosive zoning method based on geophysical exploration technology according to claim 2, characterized in that, The specific method for generating the three-dimensional model of static rock mass physical parameters is as follows: Based on the seismic wave exploration data and the DC resistivity method and induced polarization method data, a cross-gradient joint inversion algorithm is used to process the data. The target area is discretized into a three-dimensional grid model, and forward modeling simulation systems for the seismic wave field and the steady current field are constructed respectively. A unified objective function is established, which includes the L2 norm fitting term of the seismic wave travel time residual, the L2 norm fitting term of the resistivity observation data, and the cross-gradient structure constraint term. The wave velocity value and resistivity value of each three-dimensional grid node are iteratively updated based on the least squares optimization algorithm. The iteration stops when the objective function converges to a preset threshold, generating a three-dimensional model of static rock mass physical parameters.
4. The explosive zoning method based on geophysical exploration technology according to claim 3, characterized in that, The three-dimensional spatial distribution of the dynamic damage factor is specifically implemented as follows: Using the wave velocity distribution in the three-dimensional model of the static rock mass physical parameters as the initial model, the vibration time history curves of each measuring point recorded by the vibration monitoring network in the blasting test are used as observation data. The wave velocity parameters of the initial model are continuously adjusted by the full waveform inversion technology and the iterative optimization algorithm to minimize the residual between the synthetic seismic record obtained by forward modeling based on the current model and the field measured vibration record. In each iteration, the dynamic damage factor is calculated according to the update amount of the wave velocity model. When the full waveform inversion iteration converges, the distribution model of the dynamic damage factor in the three-dimensional space of the target area is output.
5. The explosive zoning method based on geophysical exploration technology according to claim 4, characterized in that, The pre-trained machine learning mapping model is specifically implemented as follows: Core sampling points are systematically arranged within the target area. Indoor rock mechanics tests are conducted on the core samples to obtain the true values of uniaxial compressive strength, elastic modulus, and cohesion. Simultaneously, multi-source fusion feature parameters corresponding to the spatial location of each sampling point are extracted. These parameters include at least P-wave velocity, S-wave velocity, and resistivity parameters obtained from a three-dimensional model of static rock mass physical parameters, and dynamic damage factor values obtained from the three-dimensional spatial distribution of dynamic damage factors. The multi-source fusion feature parameters are used as input feature vectors, and the true values of the mechanical parameters are used as supervised learning targets. A gradient boosting decision tree algorithm is employed for model training. The optimal hyperparameter combination, including the maximum tree depth, learning rate, and subsampling ratio, is determined through grid search and cross-validation to obtain a training model capable of establishing a nonlinear mapping relationship from geophysical features to rock mechanical properties.
6. The explosive zoning method based on geophysical exploration technology according to claim 5, characterized in that, The specific method for classifying explosiveness levels is as follows: Multiple discrimination indicators for each three-dimensional grid node are extracted from the three-dimensional geomechanical model, including static mechanical indicators, dynamic response indicators, and native structural indicators. Then, based on the preset multi-indicator fusion partitioning rules, logical judgment and level classification are performed. The levels include at least difficult-to-blast areas, easy-to-blast areas, and tectonic influence areas. Finally, a three-dimensional partitioning model with different blastability level labels is generated that corresponds to the three-dimensional geomechanical model space.
7. The explosive zoning method based on geophysical exploration technology according to claim 6, characterized in that, The specific method for automatically generating the corresponding blasting hole network parameters and charge quantity suggestions is as follows: The three-dimensional partition model containing different blastability level labels is coupled with a built-in expert knowledge base. This knowledge base pre-stores quantitative mapping relationships between different blastability levels and blasting design parameters. Based on this mapping relationship, the system automatically generates customized blasting design schemes for each partition unit and outputs a digital blasting design drawing containing hole position coordinates, drilling depth, hole mesh density, explosive type, and charge distribution.
8. The explosive zoning method based on geophysical exploration technology according to claim 7, characterized in that, The blasting effect evaluation data includes at least the blasting block size distribution data, blasting morphology and integrity data, blasting harmful effect monitoring data, and comprehensive economic and technical indicator data.
9. A method for explosive zoning based on geophysical exploration technology according to claim 8, characterized in that, The specific method for generating the feedback signal is as follows: The actual observed values in the blasting effect evaluation data, including the average block size and large block rate in the blasting block size distribution data, and the base rate and over-excavation amount in the blasting morphology data, are compared with the expected effect values predicted based on the three-dimensional partitioning model and the generated blasting design parameters. The difference between each observed value and the predicted value is calculated using a preset loss function, which is the weighted sum of squares of each difference item. The calculated total loss function value and the spatial location information of each partition in the corresponding three-dimensional partitioning model together constitute the feedback signal.