A vertical shaft vertical hole intelligent blasting method based on multi-source perception correction

Abstract

This application relates to an intelligent blasting method for vertical boreholes in shafts based on multi-source perception correction, comprising: constructing an intelligent optimization model for blasting parameters; intelligently predicting the number of boreholes, hole spacing, hole depth, charge quantity, and delayed detonation parameters under multi-cycle construction conditions based on multi-source data and deep neural networks, generating multiple blasting schemes and achieving dynamic optimization; perceiving multi-source data such as post-blast images, shaft wall cross-sectional point clouds, blasting vibration timing signals, and advance shaping, and constructing a comprehensive intelligent post-blast evaluation model in conjunction with energy utilization rate, performing multi-dimensional evaluation of block size distribution, over-excavation and under-excavation, shaft wall disturbance, vibration response, and comprehensive quality level; predicting the correction amount of blasting parameters for the next cycle based on the evaluation results, and feeding it back to the intelligent optimization model for blasting parameters, thereby realizing the linkage and self-learning iterative optimization of blasting design, construction implementation, post-blast evaluation, and parameter correction.

Description

Technical Field

[0001] This invention relates to the field of shaft construction technology, and in particular to an intelligent blasting method for vertical blast holes in shafts based on multi-source sensing correction. Background Technology

[0002] In mine shaft or underground engineering construction, drilling and blasting is a commonly used method for rock breaking. Traditional shaft construction often suffers from problems such as low blasting energy utilization, uneven rock block size distribution, shaft collapse, uncontrollable over-excavation and under-excavation, and reduced construction safety due to unreasonable hole placement, improper charge control, reliance on experience in hole spacing design, and underutilization of drilling and charge deviations. Existing shaft cut-out blasting methods typically divide blast holes into cut-out holes, auxiliary holes, and peripheral holes, and achieve orderly superposition of explosive stress waves through delayed detonation control, which improves blasting efficiency and shaft completion quality to some extent. However, existing schemes still mainly rely on static parameter calculations and manual experience corrections, making it difficult to adapt to different rock strata conditions and parameter changes during multi-cycle construction.

[0003] Furthermore, existing methods lack comprehensive utilization of information such as borehole deviation, actual charge deviation, plugging differences, post-blast images, wellbore cross-sectional morphology, blasting vibration, and energy utilization. They typically lack a multi-dimensional evaluation mechanism for post-blast results and a control mechanism to reverse-correct blasting parameters for the next cycle based on these results. Therefore, there is an urgent need for an intelligent vertical borehole blasting method for vertical shafts that further incorporates multi-source sensing, post-blast evaluation, and adaptive parameter correction mechanisms, building upon the foundation of vertical borehole blasting. This would reduce reliance on manual experience and improve the consistency, forming quality, energy utilization, and safety of single-cycle and multi-cycle operations. Summary of the Invention

[0004] To address the aforementioned shortcomings in existing technologies, this invention provides a smart blasting method for vertical boreholes in shafts based on multi-source sensing correction. The improvement lies in that the blasting method includes the following steps: Step S1: Integrate vertical shaft excavation depth, geological conditions, historical blasting effect data and construction constraint information to construct an intelligent optimization model for blasting parameters, and use the intelligent optimization model for blasting parameters to determine the theoretical design parameters for various types of blast holes; Step S2: Based on the theoretical design parameter benchmark values, combined with the construction equipment capacity, shaft cross-sectional boundary, safety control threshold and blasting vibration limit, generate multiple candidate blasting schemes. Each candidate blasting scheme corresponds to different combinations of borehole number, hole spacing, hole depth, charge amount and delayed detonation parameters. Step S3: Input the geological parameters of the current cycle, historical blasting effect data, construction constraint parameters, and structured parameters corresponding to each candidate blasting scheme into the intelligent optimization model of blasting parameters to obtain the block size distribution prediction results, over-excavation and under-excavation prediction results, well wall disturbance prediction results, vibration response prediction results, energy utilization rate prediction results, and comprehensive quality level prediction results corresponding to each candidate blasting scheme. Based on the comprehensive objective function, sort the multiple candidate blasting schemes and select the candidate blasting scheme that meets the preset constraints and has the best comprehensive evaluation as the execution scheme of the current cycle. Step S4: Perform borehole drilling, charging, and plugging according to the current cycle execution plan, and collect the actual drilling construction parameters and charging construction parameters. The actual drilling construction parameters include borehole position deviation, hole depth deviation, and borehole deviation. The charging construction parameters include charging amount deviation, plugging length, and plugging density. Step S5: Using multi-source sensing and monitoring units deployed on the working face, well wall and wellhead of the vertical shaft, collect post-blast image data, well wall cross-sectional point cloud data, blasting vibration timing data and footage formation data after detonation. Combine the design parameters of this cycle, the measured parameters of drilling construction and the charging construction parameters to calculate the energy utilization rate η. Step S6: Input the post-blast image data, wellbore cross-sectional point cloud data, blasting vibration time sequence data, drilling footage data, borehole construction measured parameters, charge construction parameters and energy utilization rate η into the post-blast comprehensive evaluation model based on multi-source heterogeneous data fusion, and output block size score, over-excavation and under-excavation score, wellbore disturbance score, vibration score, energy utilization score and comprehensive quality level. Step S7: Based on the post-blast comprehensive evaluation results, generate the following corrections for the next cycle: number of boreholes ΔN, hole spacing ΔS, hole depth ΔH, charge amount ΔQ, and delayed detonation parameter ΔT. Feed these corrections back to the intelligent optimization model for blasting parameters and the candidate blasting scheme generation process to achieve adaptive correction and dynamic optimization of the blasting parameters for the next cycle.

[0005] Preferably, step S1 further includes: Step S1-1: Determine the reference value for borehole depth. As shown in the following formula: ; Where H c Here, ɛ represents the depth of the cut hole, α represents the over-depth coefficient, and H represents the advance per cycle. f For the depth of auxiliary holes and surrounding holes, H ɛ Let f(d) be the extra-deep length of the cut hole, and f(d) be the cross-sectional correction value. Step S1-2: Determine the reference value for borehole spacing. As shown in the following formula: ; Where Y is the spacing between auxiliary holes, R is the radius of the peripheral hole ring, W is the thickness of the blasting layer around the peripheral holes, and R W The radius of the outermost auxiliary hole ring is given, n is the number of auxiliary hole rings, E is the spacing between the surrounding holes, f is the rock firmness coefficient, and K is the adjustment coefficient. These are then optimized and corrected based on historical cycle block size distribution. Step S1-3: Determine the baseline value for the number of blast holes. As shown in the following formula: ; Where N is the total number of boreholes, q is the unit explosive consumption, and S is the cross-sectional area of ​​the shaft. To maximize the utilization rate of boreholes, This is the charge coefficient; Where D is the mass of the explosive cartridge and D is the diameter of the wellbore. The number of peripheral holes, Correction number for the number of peripheral holes. The number of slotted holes, To assist in determining the number of slotting holes, This is a correction factor for the number of slotted holes. Number of auxiliary holes; Steps S1-4, Baseline values ​​for charge amount As shown in the following formula: ; Q f The charge amount in the auxiliary hole is given by d, the diameter of the auxiliary hole is given by L, and the length of the propellant grain is given by L. Q is the density of the explosive. t Q represents the charge amount in the cut hole, α is the charge coefficient in the cut hole, and Q is the charge amount in the cut hole. z β is the charge amount in the peripheral holes, and β is the charge coefficient in the peripheral holes; Steps S1-5: Determine the baseline values ​​for the delayed detonation parameters. As shown in the following formula: ; ; Where T is the th The detonation delay time of the initiation phase, This serves as the baseline detonation time for the first detonation stage. This is the time interval between adjacent detonation stages. For the number of detonation stages, T W The minimum resistance line or the distance of action of the free surface. This refers to the rock mass fracture propagation velocity or effective stress wave propagation velocity. The dominant frequency of blasting vibration, The delay correction factor is used; and the delay detonation parameters for each detonation section are determined sequentially from the central slotted hole, auxiliary hole to the peripheral hole.

[0006] Preferably, in step S2, the candidate blasting scheme is constructed in the following manner: For the j-th candidate blasting scheme, its parameter vector is represented as: ; in: ; Where, N j S represents the number of blast holes for the j-th candidate blasting scheme. j Let the hole spacing of the j-th candidate blasting scheme be... ， H j Let the depth of the blast hole for the j-th candidate blasting scheme be... ， Q j Let T be the charge quantity for the j-th candidate blasting scheme. j To delay the detonation time of the j-th candidate blasting scheme, the candidate perturbation quantities varying within a preset range include: δN j δS represents the disturbance of the number of boreholes. j δH represents the hole spacing disturbance. j Let δQ be the depth disturbance. j δT represents the charge perturbation. j The delay initiation time disturbance is defined as follows: the candidate disturbance satisfies the constraints of borehole layout boundary, charge safety, vibration control, and equipment construction capability.

[0007] Preferably, step S3 includes: Step S3-1: The intelligent parameter optimization model is trained based on historical multi-cycle samples, and the input feature vector of the j-th candidate blasting scheme is denoted as X. j The latent feature vector h of the candidate solution is extracted through a multi-layer deep neural network. j （L） The calculation is as follows: ; ; Among them, W (l) and b (l) Let be the weight matrix and bias vector of the l-th layer network, respectively, and σ be the non-linear activation function; Let be the initial input feature vector of the j-th candidate blasting scheme. Let be the hidden feature vector of the j-th candidate blasting scheme in the (l+1)th layer of the network. Let be the hidden feature vector of the j-th candidate blasting scheme in the l-th layer of the network. This represents the total number of network layers. Step S3-2, the output of the intelligent parameter optimization model for the j-th candidate blasting scheme includes a scheme score vector and a quality level probability vector, which are expressed as follows: ; ; ; in, These are predicted values, including predicted block size, over-excavation / under-excavation, wellbore disturbance, vibration, and energy utilization rate. j This is the probability distribution vector of the overall quality level. To predict the overall quality level, The weight matrix for the solution scoring output layer. This is the bias vector of the scheme scoring output layer. The weight matrix of the output layer for quality level classification is given. The bias vector for the quality level classification output layer, k is the comprehensive quality level score index, and p j,k For the first The candidate blasting scheme belongs to the first... Predicted probabilities for each comprehensive quality level category; The implicit feature vectors of the candidate schemes are used to output the predicted block size index, predicted over-excavation and under-excavation index, predicted well wall disturbance index, predicted vibration index, predicted energy utilization rate index and predicted comprehensive quality level for each candidate blasting scheme. Step S3-3, when selecting the best blasting scheme from multiple candidate schemes, is based on the comprehensive objective function G. j The ranking of multiple candidate blasting schemes is represented as follows: ; in, Let the block size objective score be the j-th candidate blasting scheme. To score the target points for over- or under-digging. Scoring is given for wellbore disturbance targets. To score the vibration control target, To score the energy utilization rate target, The penalty term for violating the constraints is λ1 to λ5, which are the weight coefficients of each objective term; taking the values ​​that satisfy the constraints and G j The optimal candidate brute-force solution is used as the current loop execution solution, and it is represented as: ; Preferably, in step S5, the multi-source sensing and monitoring unit includes at least an image acquisition unit, a wellbore cross-section scanning unit, a blasting vibration monitoring unit, a borehole trajectory detection unit, a charge plugging detection unit, and a footage and forming detection unit; wherein, the image acquisition unit includes a first industrial camera set at the edge of the bottom working platform facing the blasting area, and a second industrial camera set at the wellbore hanging bracket facing the wellbore surface; the wellbore cross-section scanning unit includes a laser scanner, structured light scanner, or photogrammetry device deployed at the wellhead platform, hanging platform, or temporary wellbore bracket; the blasting vibration monitoring unit includes multiple triaxial vibration sensors.

[0008] Preferably, in step S5, the energy utilization rate η is calculated as shown in the following formula: ; Where η is the energy utilization rate; i represents different types of boreholes, including slotted holes, auxiliary holes, and peripheral holes; Q i The charge per borehole of the same type; n i H represents the number of boreholes of the same type; i V represents the borehole depth of the same type of borehole; V represents the wellbore volume; σ represents the borehole depth of the same type of borehole. c η represents the rock mass strength; s represents the energy conversion ratio coefficient; when η is within the range of 0.3%–1%, the current cycle design parameters are considered reasonable; when η exceeds the range of 0.3%–1%, the parameter correction module is triggered to adjust the number of boreholes, hole spacing, or charge amount for the next cycle.

[0009] Preferably, step S6 includes: Step S6-1, the post-blast comprehensive evaluation model adopts a multimodal multi-task deep neural network, including a structured parameter feature extraction branch and a fusion decision branch; the input features of the structured parameter feature extraction branch include the theoretical design parameters of the various types of boreholes obtained in step S1, and the energy utilization rate η obtained in step S5. Step S6-2: Obtain image feature vectors As shown in the following formula: ; in, img This represents the image feature extraction network; I represents the post-explosion image; Step S6-3: Obtain point cloud feature vectors As shown in the following formula: ; in, pc This represents a point cloud feature extraction network; Add cloud patterns to the cross-section of the well wall; Step S6-4: Obtain the vibration feature vector As shown in the following formula:

[0010] in, vib This represents a vibration feature extraction network; This is a timing signal for blasting vibration; Step S6-5: Obtain the parameter feature vector As shown in the following formula:

[0011] in, tab This represents a structured parametric feature extraction network, x tab The input vector is structured. Step S6-6: The fusion decision branch fuses the image feature vector, point cloud feature vector, vibration feature vector, and parameter feature vector to obtain a comprehensive feature vector F, as shown in the following formula: ; ; ; Among them, α m For the fusion weights of the corresponding modes, f m The corresponding modal feature vector is used; and based on the comprehensive feature vector F, the block size score, over-excavation / under-excavation score, wellbore disturbance score, vibration score, energy utilization score, and comprehensive quality level are output, along with the parameter correction amount for the next cycle. For the first Attention scoring for each modality For attention weight vectors, For the first Feature transformation weight matrix for each modality For the corresponding bias vector, For the first The attention score for each modality, where r is the modality index. Steps S6-7: Based on the comprehensive feature vector F, output the comprehensive score vector and the comprehensive quality level, which are expressed as follows: ; in, The scoring system includes block size score, over-excavation / under-excavation score, wellbore disturbance score, vibration score, and energy utilization score, where p is the probability distribution of the overall quality level. To synthesize the results of the overall quality level identification, The weight matrix of the output layer of the rating regression is... The bias vector of the output layer of the rating regression. The weight matrix of the output layer for quality level classification is given. The bias vector for the quality level classification output layer. For overall quality to belong to the first The predicted probability of each level category.

[0012] Preferably, step S7 includes: Step S7-1, based on the comprehensive feature vector The correction values ​​for the number of boreholes, hole spacing, hole depth, charge amount, and delayed detonation parameters for the next cycle are output as follows: ; And the design parameters for the next cycle satisfy the following formula: ; Where, N next The number of boreholes corrected in the next cycle; S next The hole spacing corrected for the next cycle; H next Hole depth corrected for the next cycle; Q next The corrected charge amount for the next cycle; T next Based on the revised delayed detonation parameters for the next cycle, a new set of candidate blasting schemes is generated. Multiple candidate blasting schemes are then predicted, evaluated, and optimized again. Adjust the weight matrix of the output layer to match the parameters. Correct the bias vector of the output layer for the parameters; Step S7-2, the post-explosion comprehensive evaluation model adopts a multi-task joint training method, and the model loss function is... The loss includes rating regression loss, ranking classification loss, parameter correction loss, and energy utilization constraint loss, as shown in the following formula: ; in: ; ; ; ; ; Among them, L score For the score regression loss, L cls For the classification loss, L rank L represents the loss for ranking the solutions. adj For parameter correction loss, L η The loss is constrained by energy utilization efficiency, where α, β, γ, δ, and ε are all weighting coefficients for the loss term. and s k Let y represent the predicted and actual values ​​of the k-th rating indicator, respectively. c and p c Let Ω represent the true label and predicted probability of the c-th quality level, respectively; let m be the set of ranked candidate sample pairs; and let m be the ranking interval constant. Δp and Δp represent the predicted and actual values ​​of the parameter correction, respectively. To predict energy utilization efficiency, η min and η max Preset reasonable upper and lower limits for energy utilization. Let n be the true label of the nth sample in the cth quality level category. Let N be the predicted probability of the nth sample in the c-th quality class, where n is the nth candidate blasting scheme, and N is the predicted probability of the nth sample in the c-th quality class. s Where C is the total number of samples, G is the total number of comprehensive quality level categories, and G is the total number of samples. i The comprehensive evaluation score of the i-th candidate blasting scheme, G j The comprehensive objective function value of the j-th candidate blasting scheme.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: This application proposes an intelligent blasting method for vertical boreholes in shafts based on multi-source sensing correction. It unifies, fuses, analyzes, and collaboratively utilizes multi-source heterogeneous information, including geological parameters, drilling parameters, charge parameters, post-blast images, shaft wall cross-sectional point clouds, blasting vibration timing signals, and footage data. It constructs an optimization mechanism that runs through the entire process of blasting parameter design, candidate scheme generation, intelligent scheme selection, post-blast comprehensive evaluation, and parameter correction for the next cycle. This overcomes the problems of insufficient adaptability, biased evaluation, and delayed correction caused by existing technologies that mainly rely on manual experience, static calculations, and post-explosive corrections. By establishing an intelligent optimization model for blasting parameters, this invention can combine different geological conditions, construction states, and historical blasting effects to optimize the number of boreholes, hole spacing, and hole depth. This invention intelligently predicts and dynamically optimizes the charge quantity and delayed detonation parameters, improving the rationality and relevance of blasting parameter configuration and the adaptive adjustment capability between different construction cycles. Furthermore, by constructing a post-blast comprehensive evaluation model based on multimodal fusion, it achieves multidimensional quantitative evaluation of block size distribution, over- and under-excavation, well wall disturbance, vibration response, energy utilization rate, and overall quality level, improving the objectivity, accuracy, and comprehensiveness of post-blast effect judgment. Simultaneously, based on the post-blast evaluation results, it generates the correction amount for the next cycle's blasting parameters and feeds it back to the front-end design stage, enabling continuous iterative optimization of blasting parameters. This is beneficial for improving the quality of vertical shaft blasting formation, energy utilization efficiency, and construction safety level, and enhances the stability, consistency, and controllability of the multi-cycle construction process. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the process of a smart blasting method for vertical boreholes in shafts based on multi-source sensing correction according to the present invention. Figure 2 This is a diagram showing the borehole layout of the intelligent blasting method for vertical boreholes in a shaft based on multi-source sensing correction, as provided in an embodiment of the present invention. Figure 3 A cross-sectional view of a vertical shaft for a vertical blast hole, provided by an embodiment of the present invention, for a smart blasting method for vertical shaft blast holes based on multi-source perception correction. Figure 4 A deep neural network structure diagram of the intelligent blasting method for vertical boreholes in shafts based on multi-source perception correction provided in this embodiment of the invention; Figure 5 This diagram illustrates the data interaction between the parameter optimization module and the post-blast evaluation module of the intelligent blasting method for vertical boreholes in shafts based on multi-source perception correction, as provided in an embodiment of the present invention.

[0015] The markings in the diagram are: 1. Cut hole; 2. Auxiliary hole; 3. Peripheral hole. Detailed Implementation

[0016] To better understand this invention, the following description, in conjunction with the accompanying drawings and examples, will further illustrate the invention.

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0018] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0019] like Figures 1 to 5 As shown, this application provides an intelligent blasting method for vertical boreholes in shafts based on multi-source sensing correction. In shaft construction, this embodiment adopts a vertical downward drilling method for blasting. Calculations and analyses are performed using a specific example of shaft construction in a mine. The shaft has a designed diameter of 6.0m, and the boreholes are arranged in a concentric ring pattern. The rock stratum is medium-hard rock (uniaxial compressive strength 80MPa, rock density...). =2600 kg / m³). Emulsion explosives (density) are selected. =1150kg / m³, detonation velocity 4800m / s), with a bore diameter of d=200mm.

[0020] Step S1: Integrate vertical shaft excavation depth, geological conditions, historical blasting effect data and construction constraint information to construct an intelligent optimization model for blasting parameters, and use the intelligent optimization model for blasting parameters to determine the theoretical design parameters for various types of blast holes; Specifically, in step S1, the reference values ​​for hole depth, hole spacing, number of blast holes, charge amount, and delayed detonation parameters of the slotted hole 1, auxiliary hole 2, and surrounding hole 3 are first calculated based on parameters such as shaft cross-sectional dimensions, single-cycle advance, rock firmness coefficient, cross-sectional correction value, borehole radius, minimum resistance line, charge coefficient, explosive density, number of detonation stages, blasting vibration dominance frequency, and rock fracture propagation velocity or effective stress wave propagation velocity. At the same time, the theoretical design parameters are corrected by combining the block size distribution and forming quality of historical cycles to obtain reference design parameters that are more in line with the current geological conditions and construction status.

[0021] Furthermore, step S1 also includes: Step S1-1: Determine the reference value for borehole depth. As shown in the following formula:

[0022] Where H c Here, ɛ represents the depth of the cut hole, α represents the over-depth coefficient, and H represents the advance per cycle. f For the depth of auxiliary holes and surrounding holes, H ɛ Let f(d) be the extra-deep length of the cut hole, and f(d) be the cross-sectional correction value. Step S1-2: Determine the reference value for borehole spacing. As shown in the following formula:

[0023] Where Y is the spacing between auxiliary holes, R is the radius of the peripheral hole ring, W is the thickness of the blasting layer around the peripheral holes, and R W The radius of the outermost auxiliary hole ring is given, n is the number of auxiliary hole rings, E is the spacing between the surrounding holes, f is the rock firmness coefficient, and K is the adjustment coefficient. These are then optimized and corrected based on historical cycle block size distribution. Step S1-3: Determine the baseline value for the number of blast holes. As shown in the following formula:

[0024] Where N is the total number of boreholes, q is the unit explosive consumption, and S is the cross-sectional area of ​​the shaft. To maximize the utilization rate of boreholes, This is the charge coefficient; Where D is the mass of the explosive cartridge and D is the diameter of the wellbore. The number of peripheral holes, Correction number for the number of peripheral holes. The number of slotted holes, To assist in determining the number of slotting holes, This is a correction factor for the number of slotted holes. Number of auxiliary holes; Steps S1-4, Baseline values ​​for charge amount As shown in the following formula:

[0025] Q f The charge amount in the auxiliary hole is given by d, the diameter of the auxiliary hole is given by L, and the length of the propellant grain is given by L. Q is the density of the explosive. t Q represents the charge amount in the cut hole, α is the charge coefficient in the cut hole, and Q is the charge amount in the cut hole. z β is the charge amount in the peripheral holes, and β is the charge coefficient in the peripheral holes; Steps S1-5: Determine the baseline values ​​for the delayed detonation parameters. As shown in the following formula:

[0026]

[0027] Where T is the th The detonation delay time of the initiation phase, This serves as the baseline detonation time for the first detonation stage. This is the time interval between adjacent detonation stages. For the number of detonation stages, T W The minimum resistance line or the distance of action of the free surface. This refers to the rock mass fracture propagation velocity or effective stress wave propagation velocity. The dominant frequency of blasting vibration, The delay correction factor is used; and the delay detonation parameters for each detonation section are determined sequentially from the central slotted hole, auxiliary hole to the peripheral hole.

[0028] Furthermore, the baseline values ​​for the number of slotting holes are determined to be 10, the hole spacing to be 1.2m, the hole depth to be 4.5m, the charge quantity to be 23.15kg / hole, and the detonation delay parameter to be 0ms; the baseline values ​​for the number of auxiliary holes are 60, the hole spacing to be 1.2m, the hole depth to be 4.3m, the charge quantity to be 2.57kg / hole, and the detonation delay parameter to be 25ms; and the baseline values ​​for the number of peripheral holes are 38, the hole spacing to be 1.2m, the hole depth to be 4.3m, the charge quantity to be 2.44kg / hole, and the detonation delay parameter to be 50ms. These baseline values ​​are used as the theoretical design parameters for this cycle and are superimposed with the corresponding corrections from the model output in step S7 to obtain the optimized design parameters for the next cycle.

[0029] Step S2: Based on the theoretical design parameter benchmark values, combined with the construction equipment capacity, shaft cross-sectional boundary, safety control threshold and blasting vibration limit, generate multiple candidate blasting schemes. Each candidate blasting scheme corresponds to different combinations of borehole number, hole spacing, hole depth, charge amount and delayed detonation parameters. Specifically, centered on the baseline values ​​for the number of boreholes, borehole spacing, borehole depth, charge quantity, and delayed detonation parameters determined in step S1, adjustments are made to the number of boreholes, borehole spacing, borehole depth, charge quantity, and delayed detonation parameters within a preset disturbance range, resulting in multiple candidate blasting schemes. Each candidate disturbance in the candidate blasting scheme satisfies the borehole layout boundary constraints, charge safety constraints, vibration control constraints, and equipment construction capability constraints. This method allows for the formation of multiple sets of optional design schemes suitable for the current cycle, while ensuring construction feasibility and safety, providing input for subsequent intelligent optimization. Further, in step S2, the candidate blasting scheme is constructed in the following manner: For the j-th candidate blasting scheme, its parameter vector is represented as:

[0030] in:

[0031] Where, N j S represents the number of blast holes for the j-th candidate blasting scheme. j Let the hole spacing of the j-th candidate blasting scheme be... ， H j Let the depth of the blast hole for the j-th candidate blasting scheme be... ， Q j Let T be the charge quantity for the j-th candidate blasting scheme. j To delay the detonation time of the j-th candidate blasting scheme, the candidate perturbation quantities varying within a preset range include: δN j δS represents the disturbance of the number of boreholes. j δH represents the hole spacing disturbance. j Let δQ be the depth disturbance. j δT represents the charge perturbation. j The delay initiation time disturbance is defined as follows: the candidate disturbance satisfies the constraints of borehole layout boundary, charge safety, vibration control, and equipment construction capability.

[0032] Step S3: Input the geological parameters of the current cycle, historical blasting effect data, construction constraint parameters, and structured parameters corresponding to each candidate blasting scheme into the intelligent optimization model of blasting parameters to obtain the block size distribution prediction results, over-excavation and under-excavation prediction results, well wall disturbance prediction results, vibration response prediction results, energy utilization rate prediction results, and comprehensive quality level prediction results corresponding to each candidate blasting scheme. Based on the comprehensive objective function, sort the multiple candidate blasting schemes and select the candidate blasting scheme that meets the preset constraints and has the best comprehensive evaluation as the execution scheme of the current cycle. Specifically, the intelligent parameter optimization model is trained based on historical multi-cycle samples and can predict the blasting effect of each candidate blasting scheme under the current geological and construction constraints. It outputs prediction results including block size distribution, over- and under-excavation, wellbore disturbance, vibration response, energy utilization rate, and overall quality grade. Based on this, multiple candidate blasting schemes are ranked according to a comprehensive objective function, and the scheme with the optimal comprehensive objective among those satisfying the constraints is selected as the execution scheme for the current cycle.

[0033] Further, step S3 includes: Step S3-1: The intelligent parameter optimization model is trained based on historical multi-cycle samples, and the input feature vector of the j-th candidate blasting scheme is denoted as X. j The latent feature vector h of the candidate solution is extracted through a multi-layer deep neural network. j （L） The calculation is as follows:

[0034]

[0035] Among them, W (l) and b (l) Let be the weight matrix and bias vector of the l-th layer network, respectively, and σ be the non-linear activation function; Let be the initial input feature vector of the j-th candidate blasting scheme; Let be the hidden feature vector of the j-th candidate blasting scheme in the (l+1)th layer of the network; Let be the hidden feature vector of the j-th candidate blasting scheme in the l-th layer of the network; This represents the total number of network layers. Step S3-2, the output of the intelligent parameter optimization model for the j-th candidate blasting scheme includes a scheme score vector and a quality level probability vector, which are expressed as follows:

[0036]

[0037]

[0038] in, These are predicted values, including predicted block size, over-excavation / under-excavation, wellbore disturbance, vibration, and energy utilization rate. j This is the probability distribution vector of the overall quality level. To predict the overall quality level, The weight matrix for the solution scoring output layer. This is the bias vector of the scheme scoring output layer. The weight matrix of the output layer for quality level classification is given. The bias vector for the quality level classification output layer, k is the comprehensive quality level score index, and p j,k For the first The candidate blasting scheme belongs to the first... Predicted probabilities for each comprehensive quality level category; The implicit feature vectors of the candidate blasting schemes are used to output the predicted block size index, predicted over-excavation and under-excavation index, predicted wellbore disturbance index, predicted vibration index, predicted energy utilization rate index, and predicted comprehensive quality level for each candidate blasting scheme. Step S3-3, when selecting the best blasting scheme from multiple candidate schemes, is based on the comprehensive objective function G. j The ranking of multiple candidate blasting schemes is represented as follows:

[0039] in, Let the block size objective score be the j-th candidate blasting scheme. To score the target points for over- or under-digging. Scoring is given for wellbore disturbance targets. To score the vibration control target, To score the energy utilization rate target, The penalty term for violating the constraints is λ1 to λ5, which are the weight coefficients of each objective term; taking the values ​​that satisfy the constraints and G j The optimal candidate brute-force solution is used as the current loop execution solution, and it is represented as:

[0040] Step S4: Perform borehole drilling, charging, and plugging according to the current cycle execution plan, and collect the actual drilling construction parameters and charging construction parameters. The actual drilling construction parameters include borehole position deviation, hole depth deviation, and borehole deviation. The charging construction parameters include charging amount deviation, plugging length, and plugging density. Specifically, the drilling equipment is placed on the bottom platform of the well, and the drilling rig is adjusted to keep the drill rod as perpendicular as possible to the shaft axis to reduce the deviation rate. After drilling, the rock powder in the hole is cleaned to improve the continuity of subsequent charging and energy transfer. Then, the explosives are charged according to the selected optimal blasting scheme, and clay or special stemming clay is used for layered sealing. The sealing length is preferably controlled to be 10% to 15% of the hole depth, and each layer is compacted to improve the sealing density and the hole opening constraint effect. At the same time, actual construction parameters such as borehole position deviation, hole depth deviation, borehole deviation, explosive charge deviation, sealing length, and sealing density are collected and used as input data for subsequent post-blast evaluation and correction.

[0041] Step S5: Using multi-source sensing and monitoring units deployed on the working face, well wall and wellhead of the vertical shaft, collect post-blast image data, well wall cross-sectional point cloud data, blasting vibration timing data and footage formation data after detonation. Combine the design parameters of this cycle, the measured parameters of drilling construction and the charging construction parameters to calculate the energy utilization rate η. Specifically, in step S5, the multi-source sensing and monitoring unit includes at least an image acquisition unit, a wellbore cross-section scanning unit, a blasting vibration monitoring unit, a borehole trajectory detection unit, a charge plugging detection unit, and a footage and forming detection unit; wherein, the image acquisition unit includes a first industrial camera set at the edge of the bottom working platform facing the blasting area, and a second industrial camera set at the wellbore hanging bracket facing the wellbore surface; the wellbore cross-section scanning unit includes a laser scanner, structured light scanner, or photogrammetry device deployed at the wellhead platform, hanging platform, or temporary wellbore bracket; the blasting vibration monitoring unit includes multiple triaxial vibration sensors.

[0042] Further, in step S5, the energy utilization rate η is calculated as shown in the following formula:

[0043] Where η is the energy utilization rate; i represents different types of boreholes, including slotted holes, auxiliary holes, and peripheral holes; Q i The charge per borehole of the same type; n i H represents the number of boreholes of the same type; i V represents the borehole depth of the same type of borehole; V represents the wellbore volume; σ represents the borehole depth of the same type of borehole. cη represents the rock mass strength; s represents the energy conversion ratio coefficient; when η is within the range of 0.3%–1%, the current cycle design parameters are considered reasonable; when η exceeds the range of 0.3%–1%, the parameter correction module is triggered to adjust the number of boreholes, hole spacing, or charge amount for the next cycle. After blasting, post-blast images, wellbore cross-sectional point clouds, blasting vibration timing signals, and advance data are collected by the multi-source sensing monitoring unit. Simultaneously, the energy utilization rate η is calculated by combining the cycle design parameters, actual drilling parameters, and charge parameters. The calculation results show that the energy utilization rate η for this cycle is approximately 0.5%, which is within the preset reasonable range, indicating that the blasting parameter configuration for the current cycle is relatively reasonable, and the blasting energy release is relatively sufficient.

[0044] Step S6: Input the post-blast image data, wellbore cross-sectional point cloud data, blasting vibration time sequence data, drilling footage data, borehole construction measured parameters, charge construction parameters and energy utilization rate η into the post-blast comprehensive evaluation model based on multi-source heterogeneous data fusion, and output block size score, over-excavation and under-excavation score, wellbore disturbance score, vibration score, energy utilization score and comprehensive quality level. Specifically, step S6 includes: Step S6-1, the post-blast comprehensive evaluation model adopts a multimodal multi-task deep neural network, including a structured parameter feature extraction branch and a fusion decision branch; the input features of the structured parameter feature extraction branch include the theoretical design parameters of the various types of boreholes obtained in step S1, and the energy utilization rate η obtained in step S5. Step S6-2: Obtain image feature vectors As shown in the following formula:

[0045] in, img This represents the image feature extraction network; I represents the post-explosion image; Step S6-3: Obtain point cloud feature vectors As shown in the following formula:

[0046] in, pc This represents a point cloud feature extraction network; Add cloud patterns to the cross-section of the well wall; Step S6-4: Obtain the vibration feature vector As shown in the following formula:

[0047] in, vib This represents a vibration feature extraction network; This is a timing signal for blasting vibration; Step S6-5: Obtain the parameter feature vector As shown in the following formula:

[0048] in, tab This represents a structured parametric feature extraction network, x tab The input vector is structured. Step S6-6: The fusion decision branch fuses the image feature vector, point cloud feature vector, vibration feature vector, and parameter feature vector to obtain a comprehensive feature vector F, as shown in the following formula:

[0049]

[0050]

[0051] Where, α m For the fusion weights of the corresponding modes, f m The corresponding modal feature vector is used; and based on the comprehensive feature vector F, the block size score, over-excavation / under-excavation score, wellbore disturbance score, vibration score, energy utilization score, and comprehensive quality level are output, along with the parameter correction amount for the next cycle. For the first Attention scoring for each modality For attention weight vectors, For the first Feature transformation weight matrix for each modality For the corresponding bias vector, For the first The attention score for each modality, where r is the modality index.

[0052] Steps S6-7: Based on the comprehensive feature vector F, output the comprehensive score vector and the comprehensive quality level, which are expressed as follows:

[0053] in, The scoring system includes block size score, over-excavation / under-excavation score, wellbore disturbance score, vibration score, and energy utilization score, where p is the probability distribution of the overall quality level. To synthesize the results of the overall quality level identification, The weight matrix of the output layer of the rating regression is... The bias vector of the output layer of the rating regression. The weight matrix of the output layer for quality level classification is given. The bias vector for the quality level classification output layer. For overall quality to belong to the first The predicted probability of each level category.

[0054] Step S7: Based on the post-blast comprehensive evaluation results, generate the following corrections for the next cycle: number of boreholes ΔN, hole spacing ΔS, hole depth ΔH, charge amount ΔQ, and delayed detonation parameter ΔT. Feed these corrections back to the intelligent optimization model for blasting parameters and the candidate blasting scheme generation process to achieve adaptive correction and dynamic optimization of the blasting parameters for the next cycle.

[0055] Specifically, step S7 includes: Step S7-1, based on the comprehensive feature vector The correction values ​​for the number of boreholes, hole spacing, hole depth, charge amount, and delayed detonation parameters for the next cycle are output as follows:

[0056] And the design parameters for the next cycle satisfy the following formula: Where, N next The number of boreholes corrected in the next cycle; S next The hole spacing corrected for the next cycle; H next Hole depth corrected for the next cycle; Q next The corrected charge amount for the next cycle; T next Based on the revised delayed detonation parameters for the next cycle, a new set of candidate blasting schemes is generated. Multiple candidate blasting schemes are then predicted, evaluated, and optimized again. Adjust the weight matrix of the output layer to match the parameters. Correct the bias vector of the output layer for the parameters; Step S7-2, the post-explosion comprehensive evaluation model adopts a multi-task joint training method, and the model loss function is... The loss includes rating regression loss, ranking classification loss, parameter correction loss, and energy utilization constraint loss, as shown in the following formula:

[0057] in:

[0058]

[0059]

[0060]

[0061]

[0062] Among them, Lscore For the score regression loss, L cls For the classification loss, L rank L represents the loss for ranking the solutions. adj For parameter correction loss, L η The loss is constrained by energy utilization efficiency, where α, β, γ, δ, and ε are all weighting coefficients for the loss term. and s k Let y represent the predicted and actual values ​​of the k-th rating indicator, respectively. c and p c Let Ω represent the true label and predicted probability of the c-th quality level, respectively; let m be the set of ranked candidate sample pairs; and let m be the ranking interval constant. Δp and Δp represent the predicted and actual values ​​of the parameter correction, respectively. To predict energy utilization efficiency, η min and η max Preset reasonable upper and lower limits for energy utilization. Let n be the true label of the nth sample in the cth quality level category. Let N be the predicted probability of the nth sample in the c-th quality class, where n is the nth candidate blasting scheme, and N is the predicted probability of the nth sample in the c-th quality class. s Where C is the total number of samples, G is the total number of comprehensive quality level categories, and G is the total number of samples. i The comprehensive evaluation score of the i-th candidate blasting scheme, G j The comprehensive objective function value of the j-th candidate blasting scheme.

[0063] Furthermore, when the model detects excessively large local block size, excessively high local disturbances in the wellbore, or uneven energy distribution in the surrounding holes, it can output corresponding corrections for hole spacing, charge amount, and delayed detonation parameters for the next cycle, allowing for targeted adjustments to the relevant borehole designs. When the model determines that the overall quality of the current cycle is good, the corresponding corrections can be smaller or remain unchanged. Through iterative optimization across multiple consecutive cycles, the blasting quality, energy utilization rate, and construction stability of the vertical shaft can be gradually improved.

[0064] Furthermore, the post-blast comprehensive evaluation model can employ a multi-task joint training approach. The model loss function includes score regression loss, grade classification loss, scheme ranking loss, parameter correction loss, and energy utilization constraint loss. Specifically, the score regression loss constrains the prediction errors of block size score, over-excavation / under-excavation score, wellbore disturbance score, vibration score, and energy utilization score; the grade classification loss constrains the classification error of the comprehensive quality grade; the scheme ranking loss enhances the ranking ability of candidate blasting schemes; the parameter correction loss constrains the prediction error of the parameter correction amount in the next cycle; and the energy utilization constraint loss constrains the energy utilization rate corresponding to the corrected parameters to be within a preset reasonable range. Through this joint training method, the model's performance on multiple tasks such as post-blast evaluation, scheme ranking, and parameter correction can be improved simultaneously.

[0065] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0066] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0067] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0068] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0069] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of the claims of the present invention pending approval.

Claims

1. A smart blasting method for vertical boreholes in shafts based on multi-source sensing correction, characterized in that, Includes the following steps: Step S1: Integrate vertical shaft excavation depth, geological conditions, historical blasting effect data and construction constraint information to construct an intelligent optimization model for blasting parameters, and use the intelligent optimization model for blasting parameters to determine the theoretical design parameters for various types of blast holes; Step S2: Based on the theoretical design parameter benchmark values, combined with the construction equipment capacity, shaft cross-sectional boundary, safety control threshold and blasting vibration limit, generate multiple candidate blasting schemes. Each candidate blasting scheme corresponds to different combinations of borehole number, hole spacing, hole depth, charge amount and delayed detonation parameters. Step S3: Input the geological parameters of the current cycle, historical blasting effect data, construction constraint parameters, and structured parameters corresponding to each candidate blasting scheme into the intelligent optimization model of blasting parameters to obtain the block size distribution prediction results, over-excavation and under-excavation prediction results, well wall disturbance prediction results, vibration response prediction results, energy utilization rate prediction results, and comprehensive quality level prediction results corresponding to each candidate blasting scheme. Based on the comprehensive objective function, sort the multiple candidate blasting schemes and select the candidate blasting scheme that meets the preset constraints and has the best comprehensive evaluation as the execution scheme of the current cycle. Step S4: Perform borehole drilling, charging, and plugging according to the current cycle execution plan, and collect the actual drilling construction parameters and charging construction parameters. The actual drilling construction parameters include borehole position deviation, hole depth deviation, and borehole deviation. The charging construction parameters include charging amount deviation, plugging length, and plugging density. Step S5: Using multi-source sensing and monitoring units deployed on the working face, well wall and wellhead of the vertical shaft, collect post-blast image data, well wall cross-sectional point cloud data, blasting vibration timing data and footage formation data after detonation. Combine the design parameters of this cycle, the measured parameters of drilling construction and the charging construction parameters to calculate the energy utilization rate η. Step S6: Input the post-blast image data, wellbore cross-sectional point cloud data, blasting vibration time sequence data, drilling footage data, borehole construction measured parameters, charge construction parameters and energy utilization rate η into the post-blast comprehensive evaluation model based on multi-source heterogeneous data fusion, and output block size score, over-excavation and under-excavation score, wellbore disturbance score, vibration score, energy utilization score and comprehensive quality level. Step S7: Based on the post-blast comprehensive evaluation results, generate the following corrections for the next cycle: number of boreholes ΔN, hole spacing ΔS, hole depth ΔH, charge amount ΔQ, and delayed detonation parameter ΔT. Feed these corrections back to the intelligent optimization model for blasting parameters and the candidate blasting scheme generation process to achieve adaptive correction and dynamic optimization of the blasting parameters for the next cycle.

2. The intelligent blasting method for vertical boreholes in shafts based on multi-source sensing correction as described in claim 1, characterized in that, Step S1 includes: Step S1-1: Determine the reference value for borehole depth. As shown in the following formula: ； Where H c Here, ɛ represents the depth of the cut hole, α represents the over-depth coefficient, and H represents the advance per cycle. f For the depth of auxiliary holes and surrounding holes, H ɛ Let f(d) be the extra-deep length of the cut hole, and f(d) be the cross-sectional correction value. Step S1-2: Determine the reference value for borehole spacing. As shown in the following formula: ； Where Y is the spacing between auxiliary holes, R is the radius of the peripheral hole ring, W is the thickness of the blasting layer around the peripheral holes, and R W The radius of the outermost auxiliary hole ring is given, n is the number of auxiliary hole rings, E is the spacing between the surrounding holes, f is the rock firmness coefficient, and K is the adjustment coefficient. These are then optimized and corrected based on historical cycle block size distribution. Step S1-3: Determine the baseline value for the number of blast holes. As shown in the following formula: ； Where N is the total number of boreholes, q is the unit explosive consumption, and S is the cross-sectional area of ​​the shaft. To maximize the utilization rate of boreholes, For the charge coefficient, Where D is the mass of the explosive cartridge and D is the diameter of the wellbore. The number of peripheral holes, Correction number for the number of peripheral holes. The number of slotted holes, To assist in determining the number of slotting holes, This is a correction factor for the number of slotted holes. Number of auxiliary holes; Steps S1-4, Baseline values ​​for charge amount As shown in the following formula: ； Q f The charge amount in the auxiliary hole is given by d, the diameter of the auxiliary hole is given by L, and the length of the propellant grain is given by L. Q is the density of the explosive. t Q represents the charge amount in the cut hole, α is the charge coefficient in the cut hole, and Q is the charge amount in the cut hole. z β is the charge amount in the peripheral holes, and β is the charge coefficient in the peripheral holes; Steps S1-5: Determine the baseline values ​​for the delayed detonation parameters. As shown in the following formula: ； ； Where T is the th The detonation delay time of the initiation phase, This serves as the baseline detonation time for the first detonation stage. This is the time interval between adjacent detonation stages. For the number of detonation stages, T W The minimum resistance line or the distance of action of the free surface. This refers to the rock mass fracture propagation velocity or effective stress wave propagation velocity. The dominant frequency of blasting vibration, The delay correction factor is used; and the delay detonation parameters for each detonation section are determined sequentially from the central slotted hole, auxiliary hole to the peripheral hole.

3. The intelligent blasting method for vertical boreholes in shafts based on multi-source sensing correction as described in claim 1, characterized in that, In step S2, the candidate blasting scheme is constructed as follows: For the j-th candidate blasting scheme, its parameter vector is represented as: ； in: ； Where, N j S represents the number of blast holes for the j-th candidate blasting scheme. j Let the hole spacing of the j-th candidate blasting scheme be... ， H j Let the depth of the blast hole for the j-th candidate blasting scheme be... ， Q j Let T be the charge quantity for the j-th candidate blasting scheme. j To delay the detonation time of the j-th candidate blasting scheme, the candidate perturbation quantities varying within a preset range include: δN j δS represents the disturbance of the number of boreholes. j δH represents the hole spacing disturbance. j Let δQ be the depth disturbance. j δT represents the charge perturbation. j The delay initiation time disturbance is defined as follows: the candidate disturbance satisfies the constraints of borehole layout boundary, charge safety, vibration control, and equipment construction capability.

4. The intelligent blasting method for vertical boreholes in shafts based on multi-source sensing correction as described in claim 1, characterized in that, Step S3 includes: Step S3-1: The intelligent parameter optimization model is trained based on historical multi-cycle samples, and the input feature vector of the j-th candidate blasting scheme is denoted as X. j The latent feature vector h of the candidate solution is extracted through a multi-layer deep neural network. j （L） The calculation is as follows: ； ； Among them, W (l) and b (l) Let be the weight matrix and bias vector of the l-th layer network, respectively, and σ be the non-linear activation function; Let be the initial input feature vector of the j-th candidate blasting scheme. Let be the hidden feature vector of the j-th candidate blasting scheme in the (l+1)th layer of the network; Let be the hidden feature vector of the j-th candidate blasting scheme in the l-th layer of the network. This represents the total number of network layers. Step S3-2, the output of the intelligent parameter optimization model for the j-th candidate blasting scheme includes a scheme score vector and a quality level probability vector, which are expressed as follows: ； ； ； in, These are predicted values, including predicted block size, over-excavation / under-excavation, wellbore disturbance, vibration, and energy utilization rate. j This is the probability distribution vector of the overall quality level. To predict the overall quality level, The weight matrix for the solution scoring output layer. This is the bias vector of the scheme scoring output layer. The weight matrix of the output layer for quality level classification is given. The bias vector for the quality level classification output layer, k is the comprehensive quality level score index, and p j,k For the first The candidate blasting scheme belongs to the first... Predicted probabilities for each comprehensive quality level category; The implicit feature vectors of the candidate schemes are used to output the predicted block size index, predicted over-excavation and under-excavation index, predicted well wall disturbance index, predicted vibration index, predicted energy utilization rate index and predicted comprehensive quality level for each candidate blasting scheme. Step S3-3, when selecting the best blasting scheme from multiple candidate schemes, is based on the comprehensive objective function G. j The ranking of multiple candidate blasting schemes is represented as follows: ； in, Let the block size objective score be the j-th candidate blasting scheme. To score the target points for over- or under-digging. Scoring is given for wellbore disturbance targets. To score the vibration control target, To score the energy utilization rate target, The penalty term for violating the constraints is λ1 to λ5, which are the weight coefficients of each objective term; taking the values ​​that satisfy the constraints and G j The optimal candidate brute-force solution is used as the current loop execution solution, and it is represented as: 。 5. The intelligent blasting method for vertical boreholes in shafts based on multi-source sensing correction according to claim 1, characterized in that, In step S5, the multi-source sensing and monitoring unit includes at least an image acquisition unit, a wellbore cross-section scanning unit, a blasting vibration monitoring unit, a borehole trajectory detection unit, a charge plugging detection unit, and a footage and forming detection unit; wherein, the image acquisition unit includes a first industrial camera set at the edge of the bottom working platform facing the blasting area, and a second industrial camera set at the wellbore hanging bracket facing the wellbore surface; the wellbore cross-section scanning unit includes a laser scanner, structured light scanner, or photogrammetry device deployed at the wellhead platform, hanging platform, or temporary wellbore bracket; the blasting vibration monitoring unit includes multiple triaxial vibration sensors.

6. The intelligent blasting method for vertical boreholes in shafts based on multi-source sensing correction as described in claim 1, characterized in that, In step S5, the energy utilization rate η is calculated as shown in the following formula: ； Where η is the energy utilization rate; i represents different types of boreholes, including slotted holes, auxiliary holes, and peripheral holes; Q i The charge per borehole of the same type; n i H represents the number of boreholes of the same type; i V represents the borehole depth of the same type of borehole; V represents the wellbore volume; σ represents the borehole depth of the same type of borehole. c η represents the rock mass strength; s represents the energy conversion ratio coefficient; when η is within the range of 0.3%–1%, the current cycle design parameters are considered reasonable; when η exceeds the range of 0.3%–1%, the parameter correction module is triggered to adjust the number of boreholes, hole spacing, or charge amount for the next cycle.

7. The intelligent blasting method for vertical boreholes in shafts based on multi-source sensing correction as described in claim 1, characterized in that, Step S6 includes: Step S6-1, the post-blast comprehensive evaluation model adopts a multimodal multi-task deep neural network, including a structured parameter feature extraction branch and a fusion decision branch; the input features of the structured parameter feature extraction branch include the theoretical design parameters of the various types of boreholes obtained in step S1, and the energy utilization rate η obtained in step S5. Step S6-2: Obtain image feature vectors As shown in the following formula: ； in, img This represents the image feature extraction network; I represents the post-explosion image; Step S6-3: Obtain point cloud feature vectors As shown in the following formula: ； in, pc This represents a point cloud feature extraction network; Add cloud patterns to the cross-section of the well wall; Step S6-4: Obtain the vibration feature vector As shown in the following formula: ； in, vib This represents a vibration feature extraction network; This is a timing signal for blasting vibration; Step S6-5: Obtain the parameter feature vector As shown in the following formula: ； in, tab This represents a structured parametric feature extraction network, x tab The input vector is structured. Step S6-6: The fusion decision branch fuses the image feature vector, point cloud feature vector, vibration feature vector, and parameter feature vector to obtain a comprehensive feature vector F, as shown in the following formula: ； ； ； Where, α m For the fusion weights of the corresponding modes, f m The corresponding modal feature vector is used; and based on the comprehensive feature vector F, the block size score, over-excavation and under-excavation score, well wall disturbance score, vibration score, energy utilization score and comprehensive quality level are output, while the parameter correction amount for the next cycle is also output. For the first Attention scoring for each modality For attention weight vectors, For the first Feature transformation weight matrix for each modality For the corresponding bias vector, For the first Attention scores for each modality, where r is the modality index; Steps S6-7: Based on the comprehensive feature vector F, output the comprehensive score vector and the comprehensive quality level, which are expressed as follows: ； in, The scoring system includes block size score, over-excavation / under-excavation score, wellbore disturbance score, vibration score, and energy utilization score, where p is the probability distribution of the overall quality level. To synthesize the results of the overall quality level identification, The weight matrix of the output layer of the rating regression is... The bias vector of the output layer of the rating regression. The weight matrix of the output layer for quality level classification is given. The bias vector for the quality level classification output layer. For overall quality to belong to the first The predicted probability of each level category.

8. The intelligent blasting method for vertical boreholes in shafts based on multi-source sensing correction as described in claim 1, characterized in that, Step S7 includes: Step S7-1, based on the comprehensive feature vector The correction values ​​for the number of boreholes, hole spacing, hole depth, charge amount, and delayed detonation parameters for the next cycle are output as follows: ； And the design parameters for the next cycle satisfy the following formula: ； Where, N next The number of boreholes corrected in the next cycle; S next The hole spacing corrected for the next cycle; H next Hole depth corrected for the next cycle; Q next The corrected charge amount for the next cycle; T next Based on the revised delayed detonation parameters for the next cycle, a new set of candidate blasting schemes is generated. Multiple candidate blasting schemes are then predicted, evaluated, and optimized again. Adjust the weight matrix of the output layer to match the parameters. Correct the bias vector of the output layer for the parameters; Step S7-2, the post-explosion comprehensive evaluation model adopts a multi-task joint training method, and the model loss function is... The loss includes rating regression loss, ranking classification loss, parameter correction loss, and energy utilization constraint loss, as shown in the following formula: ； in: ； ； ； ； ； Among them, L score For the score regression loss, L cls For the classification loss, L rank L represents the loss for ranking the solutions. adj For parameter correction loss, L η The loss is constrained by energy utilization efficiency, where α, β, γ, δ, and ε are all weighting coefficients for the loss term. and s k Let y represent the predicted and actual values ​​of the k-th rating indicator, respectively. c and p c Let Ω represent the true label and predicted probability of the c-th quality level, respectively; let m be the set of ranked candidate sample pairs; and let m be the ranking interval constant. Δp and Δp represent the predicted and actual values ​​of the parameter correction, respectively. To predict energy utilization efficiency, η min and η max Preset reasonable upper and lower limits for energy utilization. Let n be the true label of the nth sample in the cth quality level category. Let N be the predicted probability of the nth sample in the c-th quality class, where n is the nth candidate blasting scheme, and N is the predicted probability of the nth sample in the c-th quality class. s Where C is the total number of samples, G is the total number of comprehensive quality level categories, and G is the total number of samples. i The comprehensive evaluation score of the i-th candidate blasting scheme, G j The comprehensive objective function value of the j-th candidate blasting scheme.

Images