A design method and system for smooth blasting charge schemes in tunnels

Abstract

This invention discloses a method and system for designing tunnel smooth blasting charge schemes, relating to the field of tunnel blasting engineering technology. The method includes: collecting and preprocessing basic parameter data required for tunnel smooth blasting design; calculating initial design parameter values ​​for the smooth blasting charge based on smooth blasting design theory and a pre-set parameter calculation model, generating an initial design scheme; inputting the initial design parameter values, cross-sectional geometric parameter data, surrounding rock condition parameter data, explosive parameter data, and historical blasting effect feedback index values ​​into a pre-trained neural network optimization model, outputting optimized smooth blasting charge design parameter values; automatically updating the peripheral hole layout and charge structure in the design scheme based on the optimized design parameter values, generating and outputting the final tunnel smooth blasting charge design scheme. This invention enables intelligent and adaptive design of tunnel smooth blasting charge parameters, improving engineering applicability and molding quality.

Description

Technical Field

[0001] This invention relates to the field of tunnel blasting engineering technology, and in particular to a design method and system for tunnel smooth blasting charge schemes. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Smooth blasting in tunnels is a core technology for achieving precise contour control in tunnel drilling and blasting excavation. By arranging peripheral blast holes along the excavation contour line, it precisely controls the charge amount, charge structure, and detonation sequence, resulting in a continuous fracture surface after blasting. This achieves a smooth excavation contour, minimal disturbance to the surrounding rock, and controllable over- and under-excavation. It is a key means to ensure construction quality and surrounding rock stability in tunnel engineering projects such as highways, railways, and subways.

[0004] Currently, the design of smooth blasting parameters for tunnels still relies primarily on empirical formulas combined with manual calculations. Designers manually determine key parameters such as the minimum resistance line, borehole spacing, borehole depth, and single-hole charge based on tunnel cross-sectional geometry, surrounding rock grade, Protodyakonov coefficient, and explosive properties, and then repeatedly refine these parameters through on-site blasting trials during construction. This existing design model has significant drawbacks: the surrounding rock conditions exhibit spatial variability, construction conditions are unstable, and the parameters obtained using empirical formulas are general but not optimal, making it difficult to balance forming quality and surrounding rock disturbance control under different lithology and cross-sectional conditions. Furthermore, on-site construction is prone to arbitrarily adjusting borehole spacing and charge amounts based on habit, leading to inconsistencies between the charge structure and the design plan. This can easily result in increased over-excavation, under-excavation residue, decreased half-hole ratio, and even risks to support safety. Therefore, the traditional design-test-correction iterative manual model is not only inefficient and costly, but also highly dependent on personnel experience, lacks adaptability to complex geological conditions, and cannot meet the comprehensive requirements of modern tunnel construction for blasting accuracy, efficiency, and safety.

[0005] In recent years, research on intelligent blasting design has included the use of genetic algorithms and neural networks to optimize blasting parameters, or the introduction of drilling process information and blasting effect feedback data for parameter adjustment. Artificial neural networks, with their nonlinear mapping and self-learning capabilities, have advantages in depicting the complex coupling relationship between "surrounding rock conditions, blasting parameters, and forming effect." Furthermore, automated design systems combining deep learning have been proposed, which can output optimized parameters and assist in generating borehole layout schemes after inputting cross-section and surrounding rock information. This intelligent approach can reduce reliance on experience and improve design consistency. However, existing technologies still have significant shortcomings, lacking a complete intelligent design scheme for smooth blasting in tunnels: First, design schemes relying solely on intelligent optimization are prone to deviating from traditional design rules, resulting in optimized parameters lacking practical engineering applicability; second, the lack of a closed-loop learning mechanism prevents adaptive optimization of key parameters under different construction conditions, still requiring extensive on-site blasting tests and manual adjustments. Summary of the Invention

[0006] To address the problems of existing tunnel smooth blasting parameter design schemes, such as heavy reliance on manual experience, poor parameter adaptability, low design efficiency, and a lack of engineering practicality and closed-loop optimization in intelligent schemes, this invention provides a tunnel smooth blasting charge design method and system. By deeply integrating traditional smooth blasting design principles with deep learning optimization, it achieves intelligent and adaptive design of tunnel smooth blasting charge parameters and constructs a closed-loop optimization mechanism of "design-construction-feedback-relearning," reducing on-site blasting tests and manual adjustments, improving the precision and engineering applicability of smooth blasting design, and optimizing the forming quality and construction efficiency.

[0007] In a first aspect, the present invention provides a design method for a tunnel smooth-surface blasting charge scheme.

[0008] A method for designing a tunnel smooth blasting charge scheme, comprising: Collect and preprocess the basic parameter data required for tunnel smooth blasting design; the basic parameters include tunnel cross-sectional geometric parameters, borehole diameter, surrounding rock condition parameters, and explosive parameters. Based on the smooth blasting design theory and the preset parameter calculation model, the initial design parameter values ​​of the smooth blasting charge are calculated according to the basic parameter data, and the initial design scheme is generated. The initial design parameter values, cross-sectional geometric parameter data, surrounding rock condition parameter data, explosive parameter data, and historical blasting effect feedback index values ​​are input into the pre-trained CNN-BiLSTM-Attention optimization model, and the optimized smooth blasting charge design parameter values ​​are output. The design scheme automatically updates the layout of peripheral holes and the charge structure based on the optimized design parameter values, generating and outputting the final tunnel smooth blasting charge design scheme.

[0009] Further technical solutions also include: Based on the final design scheme, smooth blasting construction of the tunnel was carried out. After construction is completed, actual blasting effect data is collected and correlated with the corresponding basic parameter data and optimized design parameter values, and stored as a sample in the database. Based on newly added samples in the database, the CNN-BiLSTM-Attention optimization model is incrementally updated or periodically retrained.

[0010] In a further technical solution, the cross-sectional geometric parameters include the cross-sectional geometry and dimensions; the surrounding rock condition parameters include the surrounding rock stability level and the Protodyakonov coefficient; and the explosive parameters include the explosive type, density, detonation velocity, and energy. The preprocessing includes: standardizing the units of the basic parameter data, checking for missing values, and normalizing the values ​​to form a standardized parameter data set; The design parameters include the minimum resistance line W, borehole spacing E, borehole depth L, and single-hole charge Q. The calculation process is as follows: Based on the tunnel cross-sectional geometry and dimensions and surrounding rock condition parameters, the minimum resistance line W and the borehole spacing E are determined using empirical formulas or recommended values ​​from specifications. The borehole depth L is determined based on the stability level of the surrounding rock; whereby the borehole depth L is directly proportional to the stability level of the surrounding rock. Based on the empirical correction coefficient K and the volume of the fractured rock mass in a single hole, the charge amount Q in a single hole is determined, i.e., Q = K × V; where V = W × E × L represents the volume of the fractured rock mass in a single hole; K is an empirical correction coefficient, determined based on the stability level of the surrounding rock and the explosive power.

[0011] Further technical solutions, based on the initial design parameters of smooth blasting charges, automatically generate initial design schemes according to preset fixed rules, including peripheral hole arrangement schemes based on borehole spacing E and charge structure schemes based on single-hole charge amount Q. The charging structure scheme is generated based on the principle of decoupled charging and fixed segmented charging rules, namely: when the single-hole charging amount Q is less than the set value, single-segment centralized charging is adopted; when the single-hole charging amount Q is greater than the set value, segmented charging is adopted, and the charging packages of segmented charging are separated by stemming or air gaps.

[0012] A further technical solution is provided, wherein the CNN-BiLSTM-Attention optimization model includes an input layer, a convolutional feature extraction layer, a bidirectional long short-term memory layer, an attention weighting layer, and an output layer, wherein: The tunnel cross-section geometric parameters, surrounding rock condition parameters, explosive parameters, initial design parameters, and historical blasting effect feedback values ​​are used to construct the feature vector of the t-th blasting cycle corresponding to the t-th time step.x t The historical sample sequence is input into the input layer, using the feature vector as the input sample. The input sample feature vector is subjected to one-dimensional convolution operation by a convolutional feature extraction layer, and the extracted local coupling feature representation constitutes the historical sample sequence embedding. The correlation of the historical sample sequence embedding is bidirectionally modeled by a bidirectional long short-term memory layer to extract the latent state sequence. The attention weighting layer calculates the temporal attention weights of the latent state sequence and performs adaptive weighting to extract the attention-weighted context vector. The context vector is input to the output layer, and the optimized minimum resistance line W, borehole spacing E, borehole depth L, and single-hole charge Q are output. Among them, the historical blasting effect feedback indicators include at least the half-hole rate and the average linear over-excavation; the borehole depth L is corrected in the network model optimization, and the number of surrounding holes and hole position arrangement are automatically updated based on the optimized borehole spacing E, and the charge structure is automatically updated based on the optimized single-hole charge Q.

[0013] A further technical solution is that the final tunnel smooth blasting design scheme is output in the form of a design parameter summary table, a borehole layout diagram around the tunnel, and a charge structure diagram. The design parameter summary table is used to display detailed data on the minimum resistance line, borehole spacing, borehole depth, single-hole charge amount, and total charge amount. The borehole layout diagram around the tunnel is used to display the position and spacing of the boreholes on the tunnel outline. The charge structure diagram is used to display the explosive segments and plugging positions inside each borehole. The actual blasting effect data includes at least the contour forming quality index, average linear over-excavation, and half-hole ratio.

[0014] Secondly, the present invention provides a design system for tunnel smooth blasting charge schemes.

[0015] A tunnel smooth blasting charge design system includes: The front-end interface module, as a web-based user interface, is built using the Vue framework. It is used to receive input parameter data, send design requests, and display design results. The backend computing module, as a server application based on the Spring Boot architecture, is used to execute the above-mentioned tunnel smooth blasting charge design method; The model and database module includes a MySQL database and a model management unit. The database is used to store basic parameter data, historical actual blasting effect data, model training samples, and optimization design results. The model management unit is used to store loaded model parameters, perform incremental updates and periodic retraining of the model, and provide a prediction interface.

[0016] Thirdly, the present invention also provides an electronic device, comprising: a memory for storing executable instructions; and a processor for executing the executable instructions stored in the memory to implement the above-mentioned tunnel smooth blasting charge design method.

[0017] Fourthly, the present invention also provides a computer-readable storage medium storing executable instructions for causing a processor to execute the executable instructions to implement the above-described tunnel smooth blasting charge design method.

[0018] Fifthly, the present invention also provides a computer program product, which includes executable instructions stored in a computer-readable storage medium; wherein, when the processor of the electronic device reads the executable instructions from the computer-readable storage medium and executes the executable instructions, the above-mentioned tunnel smooth blasting charge design method is realized.

[0019] The above one or more technical solutions have the following beneficial effects: 1. This invention proposes a design method and system for tunnel smooth blasting charge schemes. It deeply integrates the traditional smooth blasting design principles with deep learning optimization. Initial parameters are calculated based on traditional smooth blasting design rules to avoid pure algorithm optimization from deviating from engineering practice. Then, a neural network optimization model is used to intelligently optimize the initial parameters. The model mines the correlation between "surrounding rock conditions - blasting parameters - blasting effect" in historical data and provides refined parameter adjustments for different lithologies and cross-sectional conditions. This realizes intelligent and adaptive design of tunnel smooth blasting charge parameters, significantly improves the matching degree between design parameters and actual blasting effects, effectively increases the half-hole ratio, controls over-excavation and under-excavation, ensures the smoothness of the tunnel excavation outline, improves the refinement and engineering applicability of smooth blasting design, and optimizes the forming quality and construction efficiency.

[0020] 2. This invention transmits the actual blasting effect data after construction back to the database for incremental updates and periodic retraining of the model, forming a closed-loop optimization mechanism of "design-construction-feedback-relearning". This allows the model to adaptively adjust design parameters to adapt to complex and changing geological conditions, avoiding the drawbacks of traditional methods that select parameters uniformly or rely entirely on experience. Furthermore, for new geological conditions that exceed the initial training range, the applicability of the model can be expanded by feeding back new data, keeping the design scheme in an optimized state, reducing on-site blasting tests and manual adjustments, and improving construction efficiency.

[0021] 3. The Web-based intelligent system of the present invention can automatically complete the entire process of parameter calculation, intelligent optimization, and scheme generation. Designers only need to input basic parameters, without the need for manual trial calculations and repeated adjustments, which greatly shortens the design time. At the same time, it reduces the number of on-site blasting tests and over-excavation rework, reduces the waste of explosives and other materials, and reduces the input of manpower and material resources. The economic benefits are particularly significant in large-scale tunnel projects.

[0022] 4. This invention standardizes the input parameters, and the system stores the parameters and corresponding blasting effect data for each design, forming a standardized and traceable database. This facilitates quality supervision and project review. The database can help enterprises establish their own blasting design knowledge base, realize the digital accumulation of engineering experience, and promote the continuous improvement of the enterprise's tunnel blasting design level through continuous data analysis and model training.

[0023] 5. The Web-based intelligent system of this invention adopts a mature Web architecture of Vue front-end + Spring Boot back-end + MySQL database, and uses PyTorch to deploy neural network models. The technology is universal, the maintenance cost is low, and it is easy to integrate with existing construction management platforms. The Web-based operation interface is user-friendly, and only designers with basic blasting knowledge need to get started. No deep artificial intelligence professional background is required. It can be widely used in tunnel construction sites with computer and Internet access.

[0024] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0025] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0026] Figure 1 This is a flowchart of the tunnel smooth blasting charge design method in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the CNN-BiLSTM-Attention optimization model in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of the architecture of the tunnel smooth blasting charge design system in Embodiment 2 of the present invention; Figure 4 This is the basic parameter input interface for the system's web interface in Embodiment 2 of the present invention; Figure 5 This is the web-based output interface for the propellant loading parameters in Embodiment 2 of the present invention. Figure 6This is the interface for recording historical smooth surface explosion effects on the web side of the system in Embodiment 2 of the present invention.

[0027] Among them, 4-1, surrounding rock condition parameters; 4-2, cross-sectional geometry and size parameters; 4-3, explosive parameters; 4-4, outline; 4-5, borehole diameter; 5-1, borehole position; 5-2, charging parameter table; 5-3, charging structure; 5-4, single-hole charging amount; 6-1, over-excavation and under-excavation index; 6-2, half-hole rate index. Detailed Implementation

[0028] It should be noted that the following detailed descriptions are exemplary and are intended only to describe specific embodiments and to provide further explanation of the invention, and are not intended to limit the scope of exemplary embodiments of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0029] The overall concept proposed in this invention is as follows: Considering the current lack of a complete intelligent design scheme for smooth blasting in tunnels, on the one hand, based on intelligent parameter design, it is necessary to inherit and constrain the principles and rules of classic smooth blasting, such as calculating initial parameters based on cross-sectional geometry, surrounding rock grade, Protodyakonov coefficient, and explosive parameters, and generating the charge structure and surrounding hole layout according to fixed rules; on the other hand, it is necessary to utilize historical blasting effect feedback (such as half-hole ratio, average linear over-excavation, etc.) to form a closed-loop learning mechanism, adaptively optimizing key parameters under different working conditions, thereby reducing on-site blasting tests and repeated adjustments, improving the forming quality, and ensuring construction safety. Therefore, this invention integrates traditional design experience calculations with deep learning optimization (such as CNN-BiLSTM-Attention) to propose a design method and system for tunnel smooth blasting charge schemes. This system enables intelligent updating and automatic generation of schemes for key parameters such as minimum resistance line, hole spacing, hole depth, and single-hole charge quantity, effectively improving the refinement and engineering applicability of smooth blasting design, and further optimizing the forming quality and construction efficiency of tunnel smooth blasting.

[0030] Example 1 This embodiment provides a method for designing a smooth-surface blasting charge scheme for tunnels, such as... Figure 1 As shown, the specific steps include: Step S1: Collect and preprocess the basic parameter data required for the tunnel smooth blasting design; the basic parameters include tunnel cross-sectional geometric parameters, borehole diameter, surrounding rock condition parameters, and explosive parameters. Step S2: Based on the smooth blasting design theory and the preset parameter calculation model, calculate the initial design parameter values ​​of the smooth blasting charge according to the basic parameter data, and generate the initial design scheme. Step S3: Input the initial design parameter values, cross-sectional geometric parameter data, surrounding rock condition parameter data, explosive parameter data, and historical blasting effect feedback index values ​​into the pre-trained CNN-BiLSTM-Attention optimization model, and output the optimized smooth blasting charge design parameter values. Step S4: Automatically update the layout of peripheral holes and the charge structure in the design scheme based on the optimized design parameter values, generate the final tunnel smooth blasting charge design scheme and output it.

[0031] The following content provides a more detailed introduction to the tunnel smooth blasting charge design method proposed in this embodiment.

[0032] In step S1, the initial input parameters required for tunnel blasting design are collected. These parameters include tunnel cross-sectional geometric parameters, borehole diameter (determined based on the drilling equipment used), surrounding rock condition parameters, and explosive parameters. The tunnel cross-sectional geometric parameters include the cross-sectional shape and dimensions, such as the design values ​​for excavation width and height. The surrounding rock condition parameters include the surrounding rock stability level (referred to as the rock stability category determined by geological survey) and the Protodyakonov coefficient (an index characterizing the hardness of rock). The relevant parameters of the selected explosive include the explosive type, density, detonation velocity, and energy. These parameters are obtained from design drawings, geological reports, and equipment specifications, providing basic data for subsequent calculations.

[0033] As a further implementation method, the basic parameter data collected above is subjected to unit unification, missing value verification and numerical normalization to form a standardized parameter data set for traditional calculation and intelligent optimization, so as to ensure the accuracy of subsequent data calculation and data processing.

[0034] In step S2, based on the smooth blasting design theory and the preset parameter calculation model, the initial design parameter values ​​of the smooth blasting charge are calculated using the standardized parameter data set from the previous step. These design parameters include the minimum resistance line W, borehole spacing E, borehole depth L, and single-hole charge Q.

[0035] Furthermore, W is the minimum resistance distance between adjacent boreholes or from a borehole to the free surface, E is the distance between adjacent boreholes, L is the borehole depth, and Q is the amount of explosive charged in each borehole. The specific calculation process is as follows: (1) Based on the tunnel cross-sectional geometry and dimensions, and surrounding rock condition parameters, determine the minimum resistance line W and borehole spacing E using empirical formulas or recommended values ​​from specifications. Specifically, the formulas for calculating the minimum resistance line W and borehole spacing E are as follows: ; ; In the above formula, f is the Protodyakonov coefficient.

[0036] Preferably, the recommended values ​​for the minimum resistance line W and the borehole spacing E can be calculated based on the existing "Technical Specification for Highway Tunnel Construction JTG / T 3660-2020", as follows:

[0037] (2) Determine the borehole depth L based on the recommended values ​​for the surrounding rock grade. The borehole depth L is directly proportional to the stability of the surrounding rock: the more stable the rock, the larger the borehole depth (i.e., the borehole depth) is generally permissible, while L can be appropriately reduced when the surrounding rock is of poor quality. In this embodiment, the design rule for L is as follows:

[0038] (3) Based on the empirical correction coefficient K and the volume of the fractured rock mass V in a single borehole, the charge amount Q in a single borehole is determined, i.e., Q = K × V; where V = W × E × L represents the volume of the fractured rock mass in a single borehole; K is an empirical correction coefficient, determined based on factors such as the rock's Protodyakonov coefficient and the explosive power (i.e., explosive performance), to ensure that the calculated Q can fully fracture the rock mass of volume V without excessive fracturing. In this embodiment, the formula for calculating the empirical correction coefficient K is: ; In the above formula, P represents the work capacity of the explosive, and the unit is mL.

[0039] Preferably, the empirical correction factor K is typically taken in the range of 0.15 to 0.25 kg / m³, with a smaller value used in weak surrounding rock and low-power explosives, and a larger value used in other cases.

[0040] Furthermore, based on the calculated initial parameters W, E, L, and Q, specific borehole charging structure design and borehole grid layout planning are carried out. Specifically, regarding the charging structure, preset fixed design rules are adopted, such as the principle of decoupled charging and fixed segmented charging rules, and a reasonable charging scheme is adaptively generated according to the size of Q. For example, when the borehole diameter is fixed, if Q is small (less than the set value), a single-segment concentrated charging can be used; if Q is large (greater than the set value), segmented charging can be used, that is, the explosive is divided into multiple segments and placed with stemming material or air gaps in between to reduce the blasting intensity. At the same time, in order to reduce the impact on the borehole wall, a decoupled charging method is adopted, that is, a charge slightly smaller than the borehole diameter is selected and an appropriate gap is maintained around the charging segment to further optimize the blasting effect. Regarding the hole pattern layout, firstly, based on the length of the tunnel cross-section's perimeter outline and the set hole spacing E, the number of peripheral smooth blasting holes is calculated to ensure that each hole is distributed as evenly as possible along the outline. Then, the geometric coordinates of each peripheral hole are determined to ensure that the holes are arranged close to the designed outline, thereby achieving the smooth profile required for smooth blasting. It is important to note that the charge structure and hole layout will automatically update as parameters W, E, L, and Q change. Therefore, when these parameters are optimized in subsequent steps, the charge scheme and layout diagram will also be automatically adjusted synchronously, eliminating the need for manual redesign.

[0041] In step S3, the initial design parameters of the smooth blasting charge, the corresponding cross-sectional geometric parameters, surrounding rock condition parameters, explosive parameters, and historical blasting effect feedback indicators are combined as input samples. These sample sequences are then input into a pre-trained CNN-BiLSTM-Attention optimization model, arranged cyclically according to the blasting process. The model analyzes and predicts the data, outputting optimized smooth blasting charge design parameters. Based on these steps, a deep learning model is used to intelligently optimize the initial design parameters, improving the accuracy and adaptability of the blasting design.

[0042] Specifically, this embodiment uses a CNN-BiLSTM-Attention neural network as the optimization model for the charge parameters, and its structure is as follows: Figure 2 As shown, the model includes an input layer, a convolutional feature extraction layer (CNN), a bidirectional long short-term memory layer (BiLSTM), an attention weighting layer, and an output layer. In this embodiment, the aforementioned cross-sectional geometric parameters, surrounding rock condition parameters, explosive parameters, initial design parameters, and historical feedback indicators are organized into a sequence of samples arranged according to the blasting cycle and input into the charge parameter optimization model. The processing procedure of each network layer in the model is as follows: (1) Input layer: The input is a historical sample sequence X∈R of length T. {T×F} The t-th time step corresponds to the feature vector of the t-th round of the blasting cycle (or a round record within an adjacent cycle window). x tThe feature vector is composed of the following input parameters: cross-sectional geometric parameters (cross-sectional shape, size, etc.), surrounding rock condition parameters (surrounding rock grade, Protodyakonov coefficient, etc.), explosive parameters (type, density, detonation velocity, energy, etc.), initial design parameters (initial values ​​of W, E, L, Q calculated by traditional formulas / rules in step S2 above), and post-blast effect feedback indicators (half-pore ratio, average linear over-excavation, etc.).

[0043] Furthermore, the historical data set of the above sequence samples, that is, the sample sequence containing the three types of fields of condition, design, and effect feedback, is denoted as historical blast data, so as to keep it consistent with the definition of the input fields.

[0044] (2) Convolutional Feature Extraction Layer (CNN): For the feature vector at each time step x t One-dimensional convolution is performed on the feature dimension to obtain the local coupling feature representation of this round of blasting cycle. z t In order to give the sliding window of one-dimensional convolution in the feature dimension a clear engineering semantic, the feature vector... x t The parameters are arranged in a fixed order: "section geometry—surrounding rock—explosive—initial design—feedback indicators." The sliding window of the convolution kernel slides across this fixed arrangement, covering adjacent parameter groups (adjacent arrangements of surrounding rock parameter group, explosive parameter group, and WELQ design parameter group). This is used to extract local interaction patterns between multi-source parameters, as well as the synergistic change characteristics of W and E under changes in surrounding rock, and the linkage characteristics of Q with W and E under changes in explosive performance. Finally, after convolution and nonlinear activation, the sequence embedding is obtained. Z =[ z 1 ,…,z T ].

[0045] (3) Bidirectional Long Short-Term Memory (BiLSTM) layer: The sequence output from the convolutional layer is embedded into Z as the input of this layer. The correlation between adjacent bursting cycles is bidirectionally modeled along the time dimension, and the hidden states at each time step are output to form a sequence of hidden states. H= [ h 1 ,…,h T This layer is used to learn the temporal dependencies and long-range correlations between parameters and effects during blasting (the feedback continuity between adjacent cycles under gradually changing surrounding rock conditions or continuous construction conditions), and because each z t Since it already includes information on cross-section, surrounding rock, explosives, and initial design parameters, BiLSTM does not only process feedback indicators, but also performs time-series modeling of the comprehensive feature representation for each round.

[0046] (4) Attention Weighting Layer: To enable the model to adaptively select historical loop information that is more relevant to the current optimization, the hidden state sequence output by BiLSTM is weighted. H= [ h 1 ,…,h T Calculate the temporal attention weights and form the context vector c. The attention calculation process is as follows: ; ; ; In the above formula, Indicates the first t The attention score of a blasting cycle represents the degree of relevance of the historical blasting data of that cycle to the optimization of the current charge parameters. The higher the score, the greater the reference value of the data of that cycle. This represents the transpose of the learnable weight vector in the attention mechanism, which serves as the adaptively updated parameter during model training and is used to fit the feature associations of blasting engineering data. This is a learnable attention weight matrix; For bias terms; The hidden state vector of the t-th blasting cycle output by the BiLSTM layer contains the comprehensive temporal characteristics of the cross section, surrounding rock, explosive, initial design parameters and historical feedback for that cycle. This represents the contribution weight of the t-th historical cycle to the current parameter optimization, thereby achieving adaptive weighted fusion of key historical cycle features.

[0047] (5) Output Layer: The context vector c is input into the fully connected regression layer, and the optimization results are output. The output can be the correction amounts ΔW, ΔE, ΔL, and ΔQ based on the initial design parameters. The optimized parameter set is obtained by W'=W+ΔW, E'=E+ΔE, L'=L+ΔL, and Q'=Q+ΔQ. Subsequently, the number and arrangement of the surrounding holes are automatically updated based on the optimized E and the cross-sectional profile, and the charge structure is automatically updated according to a preset fixed rule based on the optimized Q.

[0048] Based on the above process, the neural network optimization model describes "parameter optimization" as the learning and regression of the "parameter value patterns that can achieve better blasting effects under given conditions" from historical engineering samples. During the training phase, the input sequence X is constructed by extracting T consecutive rounds of historical records (i.e., historical blasting data) from the database; the parameters ultimately adopted in historical projects / after on-site correction (or their correction relative to the initial design parameters) are used as supervision labels Y. A regression loss function is used to train the network, minimizing... It can also superimpose engineering constraint penalties to limit W, E, L, and Q to fall within a reasonable range and meet established rules. After training, during the inference phase, the current section / surrounding rock / explosive parameters, initial design parameters, and feedback sequences from adjacent historical cycles are input, and parameter correction suggestions for the current working condition can be output, realizing adaptive updates to traditional initial design parameters.

[0049] In step S4, the parameter values ​​of the optimized design are output: W, E, L, and Q values. The relevant design charts, surrounding hole layout, and charge structure in the design scheme are automatically updated to generate the final tunnel smooth blasting charge design scheme, forming a complete design scheme for implementation and inspection.

[0050] Specifically, the final tunnel smooth blasting design scheme is output in the form of a design parameter summary table, a borehole layout diagram around the tunnel, and a charge structure diagram. The design parameter summary table lists all the final adopted blasting parameter values, which is used to show detailed data on the minimum resistance line, borehole spacing, borehole depth, single-hole charge amount, and total charge amount. The charge amount calculation results provide a statistical summary of the charge amount for all boreholes. The borehole layout diagram around the tunnel is used to show the position and spacing of the boreholes on the tunnel outline. The charge structure diagram is used to show the location of the explosive segments and plugging points inside each borehole.

[0051] As a further implementation method, the aforementioned tunnel smooth blasting charge design method also includes a model closed-loop update step. Specifically, based on the final generated design scheme, smooth blasting construction of the tunnel is carried out; after construction is completed, actual blasting effect data is collected, and the blasting effect data is recorded and fed back, such as recording the actual measured contour forming quality indicators, average linear over-excavation, half-hole ratio, etc., and comparing them with the design expectations, and associating them with the corresponding basic parameter data and optimized design parameter values, and storing them as samples in the database; using the new samples in the database, the CNN-BiLSTM-Attention optimization model is incrementally updated or periodically retrained to correct the model parameters, so that it can more accurately predict the optimized parameters in subsequent tunnel blasting, and achieve continuous improvement of the design. The whole process forms a closed-loop process of "design-construction-feedback-relearning". The model uses new feedback / new samples for incremental updates or periodic retraining, continuously optimizing and correcting the model parameters, and gradually improving the intelligence level and design quality of tunnel smooth blasting design.

[0052] Example 2 This embodiment provides a tunnel smooth blasting charge design system, such as... Figure 3 As shown, it specifically includes a front-end interface module, a back-end computing module, and a model and database module.

[0053] The front-end interface module serves as a web-based user interface, receiving input parameter data (such as user-inputted tunnel cross-sectional dimensions, surrounding rock type, blasting requirements, and other design parameters) and sending design requests to the back-end calculation module. Preferably, the front-end interface module includes a parameter input unit, which provides input areas for lithological parameters, contour line parameters, explosive parameters, and borehole parameters, and generates the tunnel excavation contour line in real time.

[0054] Furthermore, the front-end interface module is built using the Vue framework and includes an output display unit to implement functions such as parameter input forms, model call triggers, and design result display. Specifically, this output display unit is used to display borehole location distribution maps, charge quantity statistics tables, and charge structure diagrams. It also provides a recording and display area for historical blasting effects, visually displays the optimized charge design scheme results, supports result downloading, and simultaneously transmits historical design and actual blasting effect data back to the database for model updates.

[0055] The backend computing module, as a server application based on the Spring Boot architecture, receives data input from the frontend and executes the tunnel smooth blasting charge design method proposed in Example 1. The backend computing module deploys a CNN-BiLSTM-Attention model based on PyTorch and includes a parameter calculation unit and an intelligent optimization unit: the parameter calculation unit calculates the initial design parameter values ​​of the smooth blasting charge based on basic parameter data, generating an initial design scheme; the intelligent optimization unit integrates a pre-trained CNN-BiLSTM-Attention optimization model, used to optimize and predict the initial design parameter values, generating the final optimized smooth blasting charge design parameter values ​​and design results.

[0056] The model and database module includes a MySQL database and a model management unit: the database is used to store basic parameter data, historical actual blasting effect data, model training samples, optimization design results, etc.; the model management unit is used to store loaded model parameters, perform incremental updates and periodic retraining of the model, and provide prediction interface services.

[0057] Furthermore, the aforementioned system is a web-based system that can be integrated with existing tunnel construction management platforms, supporting online use at construction sites with computer and internet access, and the design results can be downloaded to guide on-site construction.

[0058] Based on the above tunnel smooth blasting charge design system, the blasting charge design is carried out, including the following steps: (1) Basic parameter collection and standardization. For example Figure 4As shown, the basic parameter data or information required for the smooth blasting design of the tunnel are collected and input, including the geometric shape and size parameters of the tunnel excavation section 4-2, the borehole diameter 4-5, the surrounding rock condition parameters 4-1 (including the surrounding rock grade and the Protodyakonov coefficient of the rock) and the explosive parameters 4-3. The above input parameter data are then processed by unifying units, checking for missing values ​​and normalizing values ​​to form a standardized parameter set for traditional calculation and intelligent optimization.

[0059] (2) Traditional rule calculation and initial scheme generation. Based on the smooth blasting design theory and the pre-set parameter calculation model, the initial design parameters are calculated using a standardized parameter set, such as... Figure 5 As shown, based on the surrounding rock grade, Protodyakonov coefficient, and borehole diameter, the minimum resistance line W of the peripheral holes and the spacing E of the peripheral boreholes are determined, thus clarifying the borehole positions 5-1; the borehole depth L is recommended and determined based on the surrounding rock condition parameter 4-1; then, the single-hole charge amount is calculated based on the volumetric charging model and combined with the explosive parameters 4-3 and the empirical correction coefficient K 5-4. Subsequently, an initial scheme is automatically generated according to preset fixed rules. Holes are evenly spaced on the excavation outline 4-4 according to the hole spacing to obtain the peripheral hole layout. Based on the decoupled charging principle and fixed segmented charging rules, the single-hole charge amount 5-4 automatically generates the charging structure 5-3, thus forming the initial peripheral hole layout scheme and charging parameter scheme.

[0060] (3) Intelligent optimization of the CNN-BiLSTM-Attention model. The initial charge parameters, along with the corresponding cross-sectional geometry and size parameters (4-2), surrounding rock condition parameters (4-1), explosive parameters (4-3), and historical blasting effect feedback indicators, are jointly input into the pre-trained CNN-BiLSTM-Attention optimization model for analysis and prediction. For example... Figure 6 As shown, the historical blasting effect feedback indicators include at least the half-hole ratio indicator 6-2 and the average linear over- and under-excavation indicator 6-1. The model outputs an optimized parameter set adapted to the current tunnel conditions, used to correct the initial design parameters and achieve adaptive optimization of the peripheral hole layout and charge quantity control; among them, the number and position of peripheral holes are automatically updated with the optimized borehole position 5-1, and the charge structure 5-3 is automatically updated according to a fixed rule with the optimized single-hole charge quantity 5-4.

[0061] (4) Result Output and Closed-Loop Update. Based on the optimized minimum resistance line W, borehole spacing E, borehole depth L, and single-hole charge Q, the final tunnel smooth blasting design scheme is generated and output to the user interface in the form of charge parameter table 5-2 and charge structure diagram 5-3. At the same time, after the blasting construction is completed, actual effect data is collected, including at least the contour line 4-4 forming quality index, average linear over-excavation and under-excavation index 6-1, half-hole rate index 6-2, etc. The effect data is associated with the corresponding input conditions and the final adopted parameters and stored in the database. The new samples in the database of the system are used to incrementally update or periodically retrain the CNN-BiLSTM-Attention optimization model to achieve continuous optimization and self-learning improvement of subsequent blasting parameters.

[0062] Example 3 This embodiment provides an electronic device, including: a memory for storing executable instructions; and a processor for executing the executable instructions stored in the memory to implement the method provided in this embodiment.

[0063] Example 4 This embodiment also provides a computer-readable storage medium storing executable instructions, which, when executed by a processor, will cause the processor to execute the method described above in this embodiment.

[0064] Example 5 This embodiment provides a computer program product including executable instructions, which are computer instructions; the executable instructions are stored in a computer-readable storage medium. When the processor of an electronic device reads the executable instructions from the computer-readable storage medium and executes the executable instructions, the electronic device performs the method described in this embodiment.

[0065] The steps and methods involved in Embodiments 2 to 5 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0066] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0067] The above description is only a preferred embodiment of the present invention. Although the specific implementation of the present invention has been described in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that, based on the technical solution of the present invention, various modifications or variations that can be made by those skilled in the art without creative effort are still within the scope of protection of the present invention.

Claims

1. A method for designing a tunnel smooth-surface blasting charge scheme, characterized in that, include: Collect and preprocess the basic parameter data required for tunnel smooth blasting design. The basic parameters include tunnel cross-sectional geometric parameters, borehole diameter, surrounding rock condition parameters, and explosive parameters. Based on the smooth blasting design theory and the preset parameter calculation model, the initial design parameter values ​​of the smooth blasting charge are calculated according to the basic parameter data, and the initial design scheme is generated. The initial design parameter values, cross-sectional geometric parameter data, surrounding rock condition parameter data, explosive parameter data, and historical blasting effect feedback index values ​​are input into the pre-trained CNN-BiLSTM-Attention optimization model, and the optimized smooth blasting charge design parameter values ​​are output. The design scheme automatically updates the layout of peripheral holes and the charge structure based on the optimized design parameter values, generating and outputting the final tunnel smooth blasting charge design scheme.

2. The tunnel smooth-surface blasting charge design method as described in claim 1, characterized in that, Also includes: Based on the final design scheme, smooth blasting construction of the tunnel was carried out. After construction is completed, actual blasting effect data is collected and correlated with the corresponding basic parameter data and optimized design parameter values, and stored as a sample in the database. Based on newly added samples in the database, the CNN-BiLSTM-Attention optimization model is incrementally updated or periodically retrained.

3. The tunnel smooth-surface blasting charge design method as described in claim 1, characterized in that, The cross-sectional geometric parameters include the cross-sectional geometry and dimensions; the surrounding rock condition parameters include the surrounding rock stability level and the Protodyakonov coefficient; and the explosive parameters include the explosive type, density, detonation velocity, and energy. The preprocessing includes: standardizing the units of the basic parameter data, checking for missing values, and normalizing the values ​​to form a standardized parameter data set; The design parameters include the minimum resistance line W, borehole spacing E, borehole depth L, and single-hole charge Q. The calculation process is as follows: Based on the tunnel cross-sectional geometry and dimensions and surrounding rock condition parameters, the minimum resistance line W and the borehole spacing E are determined using empirical formulas or recommended values ​​from specifications. The borehole depth L is determined based on the stability level of the surrounding rock; whereby the borehole depth L is directly proportional to the stability level of the surrounding rock. Based on the empirical correction coefficient K and the volume of the fractured rock mass in a single hole, the charge amount Q in a single hole is determined, i.e., Q = K × V; where V = W × E × L represents the volume of the fractured rock mass in a single hole; K is an empirical correction coefficient, determined based on the stability level of the surrounding rock and the explosive power.

4. The tunnel smooth-surface blasting charge design method as described in claim 3, characterized in that, Based on the initial design parameters of the smooth blasting charge, the initial design scheme is automatically generated according to the preset fixed rules, including the peripheral hole arrangement scheme based on the borehole spacing E and the charge structure scheme based on the single hole charge Q. The charging structure scheme is generated based on the principle of decoupled charging and fixed segmented charging rules, namely: when the single-hole charging amount Q is less than the set value, single-segment centralized charging is adopted; when the single-hole charging amount Q is greater than the set value, segmented charging is adopted, and the charging packages of segmented charging are separated by stemming or air gaps.

5. The tunnel smooth-surface blasting charge design method as described in claim 1, characterized in that, The CNN-BiLSTM-Attention optimized model includes an input layer, a convolutional feature extraction layer, a bidirectional long short-term memory layer, an attention weighting layer, and an output layer, wherein: The tunnel cross-section geometric parameters, surrounding rock condition parameters, explosive parameters, initial design parameters, and historical blasting effect feedback values ​​are used to construct the feature vector of the t-th blasting cycle corresponding to the t-th time step. x t The historical sample sequence is input into the input layer, using the feature vector as the input sample. The input sample feature vector is subjected to one-dimensional convolution operation by a convolutional feature extraction layer, and the extracted local coupling feature representation constitutes the historical sample sequence embedding. The correlation of the historical sample sequence embedding is bidirectionally modeled by a bidirectional long short-term memory layer to extract the latent state sequence. The attention weighting layer calculates the temporal attention weights of the latent state sequence and performs adaptive weighting to extract the attention-weighted context vector. The context vector is input to the output layer, and the optimized minimum resistance line W, borehole spacing E, borehole depth L, and single-hole charge Q are output. Among them, the historical blasting effect feedback indicators include at least the half-hole rate and the average linear over-excavation; the borehole depth L is corrected in the network model optimization, and the number of surrounding holes and hole position arrangement are automatically updated based on the optimized borehole spacing E, and the charge structure is automatically updated based on the optimized single-hole charge Q.

6. The tunnel smooth-surface blasting charge design method as described in claim 1, characterized in that, The final tunnel smooth blasting design scheme is output in the form of a design parameter summary table, a borehole layout diagram around the tunnel, and a charge structure diagram. The design parameter summary table is used to display detailed data on the minimum resistance line, borehole spacing, borehole depth, single-hole charge amount, and total charge amount. The borehole layout diagram around the tunnel is used to display the position and spacing of the boreholes on the tunnel outline. The charge structure diagram is used to display the explosive segments and plugging positions inside each borehole. The actual blasting effect data includes at least the contour forming quality index, average linear over-excavation, and half-hole ratio.

7. A tunnel smooth-surface blasting charge design system, characterized in that, include: The front-end interface module, as a web-based user interface, is built using the Vue framework. It is used to receive input parameter data, send design requests, and display design results. The backend computing module, as a server application based on the Spring Boot architecture, is used to execute the tunnel smooth blasting charge design method as described in any one of claims 1-6; The model and database module includes a MySQL database and a model management unit. The database is used to store basic parameter data, historical actual blasting effect data, model training samples, and optimization design results. The model management unit is used to store loaded model parameters, perform incremental updates and periodic retraining of the model, and provide a prediction interface.

8. An electronic device, characterized in that, include: Memory, used to store executable instructions; The processor, when executing executable instructions stored in the memory, implements the tunnel smooth blasting charge design method according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The device stores executable instructions that, when executed by a processor, implement the tunnel smooth blasting charge design method according to any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes executable instructions stored in a computer-readable storage medium; When the processor of the electronic device reads the executable instructions from the computer-readable storage medium and executes the executable instructions, it implements the tunnel smooth blasting charge design method according to any one of claims 1-6.

Images