A real-time prediction system and method for blasting rock mass damage during tunnel drilling and blasting construction.
By acquiring and processing multi-source data, and combining hybrid-driven damage prediction and parameter optimization, real-time and accurate prediction and closed-loop control of rock mass damage in tunnel drilling and blasting construction have been achieved. This solves the problems of high safety risks and rough control of over-excavation and under-excavation caused by reliance on manual experience in existing technologies, and improves construction safety and efficiency.
Patent Information
- Application Number
- CN202611101193.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-23
- Publication Date
- 2026-11-17
AI Technical Summary
In existing tunnel drilling and blasting construction, the prediction of blasted rock mass damage relies on human experience, resulting in high construction safety risks, crude control of over-excavation and under-excavation, and difficulty in quantifying dynamic damage evolution by traditional single sensor monitoring, making it difficult to achieve accurate damage prediction and closed-loop control.
A multi-source data acquisition module is used to collect blasting vibration signals, acoustic emission signals of crack propagation inside the rock mass, and dynamic image data of surface cracks in real time. Wavelet noise reduction and crack geometric parameter identification are performed through a data intelligent processing module. Combined with a hybrid drive damage prediction module and a dynamic optimization module for blasting parameters, real-time damage prediction and parameter optimization are achieved, and automatic charging machine and detonator controller are linked to execute commands.
It significantly improved the accuracy of rock mass damage prediction, reduced over-excavation rate, saved support materials, shortened construction period, improved construction safety and efficiency, and realized intelligent and refined tunnel construction.
Smart Images

Figure CN123116775A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of tunnel monitoring technology, and in particular to a real-time prediction system and method for blasting rock mass damage during tunnel drilling and blasting construction. 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] Currently, the prediction and control of rock mass damage during tunnel drilling and blasting construction mainly rely on manual experience to set blasting parameters and assess the extent of surrounding rock damage through single sensors or post-drilling inspections. Therefore, construction personnel need to frequently enter the excavation face to measure over-excavation and under-excavation, and adjust subsequent blasting plans based on experience.
[0004] However, relying on manual experience to judge rock mass damage and adjust blasting parameters with a lag in related technologies not only increases construction safety risks (such as collapse and rock bursts) but also makes it difficult to accurately quantify dynamic damage evolution, resulting in large fluctuations in over-excavation rates and persistently high support costs. Especially in complex geological conditions, experience-driven extensive control can easily lead to excessive damage to the surrounding rock, reducing the stability of the tunnel structure, while excessive blasting vibration may threaten the safety of adjacent structures.
[0005] In recent years, with the development of multi-source sensing technology, physical-data hybrid modeling, and real-time feedback control technology, especially the engineering application of high-precision digital electronic detonators, edge computing terminals, and reinforcement learning algorithms, new approaches have been provided for real-time monitoring and closed-loop control of blasting damage. The intelligent transformation of tunnel drilling and blasting construction has become an industry trend. If multimodal data acquisition, hybrid-driven damage prediction, and adaptive optimization technologies can be integrated to build an integrated "perception-decision-execution" system, it will significantly improve damage control accuracy, reduce over-excavation costs, and decrease the risk of manual intervention, thus promoting the evolution of tunnel construction towards digitalization and intelligence.
[0006] However, current intelligent algorithms for tunnel damage prediction still have the following shortcomings: First, existing machine learning models are mostly trained offline and rely on static inference, lacking real-time data closure, resulting in prediction results lagging significantly behind the construction process and failing to support timely adjustments to blasting parameters. Second, traditional models rely on single vibration or acoustic data sources, making it difficult to effectively integrate multimodal information, resulting in limited damage identification accuracy and insufficient automated extraction capabilities for fracture geometric parameters. Third, existing models have poor generalization ability; when faced with different lithologies such as granite and shale, the damage threshold is fixed, lacking a geological adaptive correction mechanism, leading to inaccurate predictions under complex geological conditions. Furthermore, machine learning outputs are mostly abstract numerical values, lacking spatial integration and visual interaction with BIM models, making it difficult for on-site construction personnel to intuitively understand risk distribution. Finally, current technology remains at the prediction level, failing to form intelligent linkages with execution equipment such as automatic charging machines and electronic detonators, failing to achieve closed-loop control of "perception-decision-execution," with manual intervention still dominating, limiting further improvements in construction efficiency and safety. Summary of the Invention
[0007] To address the shortcomings of existing technologies, the purpose of this invention is to provide a real-time prediction system and method for rock mass damage during tunnel drilling and blasting construction. This system can dynamically optimize blasting parameters to balance excavation efficiency and surrounding rock protection, achieve real-time and accurate prediction of rock mass damage, and improve the safety and economy of tunnel drilling and blasting construction.
[0008] To achieve the above objectives, the present invention is implemented through the following technical solution: The first aspect of this invention provides a real-time prediction system for blasted rock mass damage during tunnel drilling and blasting construction, comprising: The multi-source data acquisition module is used to acquire blasting vibration signals, acoustic emission signals of crack propagation inside the rock mass, and dynamic image data of surface cracks in real time. The data intelligent processing module is used to perform wavelet noise reduction on vibration signals, identify crack geometric parameters based on surface crack dynamic image data using crack identification model, and convert acoustic emission signals into crack density time series data to generate a spatiotemporally aligned damage feature matrix. The hybrid-driven damage prediction module is used to predict damage based on the fracture geometric parameters and damage feature matrix using a damage prediction model, obtain damage prediction results, and adaptively determine damage thresholds for different lithologies. The damage prediction results include rock mass damage depth and dynamic damage index. The damage prediction model consists of a material constitutive model, a regression model, and a deep learning model. The dynamic optimization module for blasting parameters is used to generate adjustment plans based on damage prediction results, optimize multi-objective functions through reinforcement learning algorithms, and link the automatic charging machine and detonator controller to execute commands.
[0009] Furthermore, the data intelligence processing module includes: The signal noise reduction module is used to filter out environmental noise in vibration signals based on wavelet transform algorithm, extract effective frequency band features, and calculate vibration energy distribution spectrum; The image recognition module is used to identify the geometric parameters of the cracks through the crack recognition model constructed by the YOLO-v5 model, output a structured crack feature table, and associate it with the coordinates of the blasting hole. The acoustic emission analysis module is used to perform short-time Fourier transform on the acoustic emission signal, calculate the crack density growth rate and the cumulative value of time-series energy, and generate the crack evolution time-series curve.
[0010] Furthermore, the hybrid-driven damage prediction module includes: The physical model calculation module is used to simulate the dynamic damage evolution of rock mass under explosive loading based on the HJC constitutive equation, and introduces the stress wave superposition effect parameter. A data-driven prediction module is used to predict damage depth through an XGBoost regression model and to predict the dynamic curve of the damage index through an LSTM network. The threshold determination module is used to dynamically adjust the critical damage threshold based on the initial integrity coefficient of the rock mass and output the over-excavation risk level.
[0011] Furthermore, the dynamic optimization module for blasting parameters includes: The parameter response library module is used to store the mapping rule base and supports the fusion of fuzzy logic reasoning and expert experience. A reinforcement learning decision-making module is used to train an agent using the PPO algorithm, generate multi-objective collaborative optimization schemes, and calculate parameters in real time to adjust confidence levels. The execution control module is used to send charge density commands to the automatic charger and write millisecond-level delay detonation timing sequences to the digital electronic detonator, and synchronously feed back the execution status to the three-dimensional visualization and early warning module.
[0012] Furthermore, it also includes a 3D visualization and early warning module, which is used to overlay damage cloud maps onto the tunnel BIM model, mark high-risk areas in real time, display parameter optimization suggestions and damage evolution trend warnings through an AR interface, and generate construction logs and risk reports.
[0013] Furthermore, the 3D visualization and early warning module includes: The damage cloud map fusion module is used to map the damage index value into a heat map, overlay it onto the corresponding blasting area of the tunnel BIM model, and mark the damage depth. The AR interactive module is used to display the 3D calibration box of high-risk areas, optimize parameter floating prompts, and support gesture control to switch perspectives or retrieve historical damage data. The risk report generation module is used to automatically export damage distribution statistics tables, support cost comparison analysis charts, and construction logs.
[0014] The second aspect of this invention provides a method for real-time prediction of blasted rock mass damage during tunnel drilling and blasting construction, comprising the following steps: Real-time acquisition of blasting vibration signals, acoustic emission signals of crack propagation inside the rock mass, and dynamic image data of surface cracks; The vibration signal is subjected to wavelet noise reduction processing. Based on the crack identification model, the surface crack dynamic image data is used to identify the crack geometric parameters. The acoustic emission signal is converted into crack density time series data to generate a spatiotemporally aligned damage feature matrix. Damage prediction is performed using a damage prediction model to predict the geometric parameters of the fracture and the damage feature matrix, and the damage prediction results are obtained. The damage threshold for different lithologies is adaptively determined. The damage prediction results include the rock mass damage depth and dynamic damage index. The damage prediction model consists of a material constitutive model, a regression model and a deep learning model. An adjustment plan is generated based on the damage prediction results. The multi-objective function is optimized through reinforcement learning algorithm, and the automatic loading machine and detonator controller are linked to execute the instructions.
[0015] A third aspect of the present invention provides a computer-readable storage medium storing a computer program adapted to be loaded by a processor and to execute steps in the real-time prediction method for blasted rock mass damage during tunnel drilling and blasting construction as described in the first aspect of the present invention.
[0016] A fourth aspect of the present invention provides a computer device comprising: Memory, processor, and computer programs stored in memory and capable of running on the processor.
[0017] When the processor executes the program, it implements the steps in the real-time prediction method for blasted rock mass damage during tunnel drilling and blasting construction as described in the first aspect of the present invention.
[0018] A fifth aspect of the present invention provides a computer program product or computer program comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps of the real-time prediction method for blasted rock mass damage in tunnel drilling and blasting construction as described in the first aspect of the present invention.
[0019] The above one or more technical solutions have the following beneficial effects: This invention discloses a real-time prediction system and method for blasting rock mass damage during tunnel drilling and blasting construction. The system utilizes a multi-source data acquisition module to collect blasting vibration signals, acoustic emission signals from crack propagation within the rock mass, and dynamic image data of surface cracks in real time. A data intelligent processing module performs wavelet noise reduction on the vibration signals and identifies crack geometric parameters, generating a spatiotemporally aligned damage feature matrix. A hybrid-driven damage prediction module combines the HJC constitutive equation with machine learning algorithms to output the dynamic damage index (DI) and damage depth (D) of the rock mass, and adaptively determines damage thresholds for different lithologies. A dynamic blasting parameter optimization module generates adjustment schemes for charge quantity, hole spacing, and detonation sequence, and coordinates the automatic charging machine and electronic detonators to execute feedback commands. A 3D visualization and early warning module integrates the damage cloud map with the tunnel BIM model, marking high-risk areas in real time and displaying parameter optimization suggestions and safety warning information. This significantly reduces reliance on manual experience and improves the accuracy of rock mass damage prediction and construction safety. This solves the problems in related technologies, such as the lag in rock mass damage prediction caused by relying on manual experience to adjust blasting parameters, the extensive control of over-excavation and under-excavation, and the difficulty in quantifying dynamic damage evolution by traditional single-sensor monitoring, thus achieving synergistic optimization of construction efficiency, cost, and safety.
[0020] This invention significantly shortens the data acquisition and parameter optimization execution cycle, solving the problem of long processing times required for traditional manual adjustments. Furthermore, by integrating multi-source information from vibration, acoustic emission, and images, it improves damage prediction accuracy to over 90%. This invention dynamically corrects model parameters using RQD values, adapting to complex geological conditions such as granite, shale, and fault fracture zones. The system reduces over-excavation, saves support materials, and shortens the single-cycle construction period. Through this method and system, tunnel drilling and blasting construction can leapfrog towards intelligent and precise development, providing a reusable technological benchmark for the industry.
[0021] 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
[0022] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0023] Figure 1 This is a schematic diagram of the overall framework of the real-time prediction system for blasted rock mass damage during tunnel drilling and blasting construction in Embodiment 1 of the present invention; Figure 2This is a schematic diagram of the specific structure of the real-time prediction system for blasted rock mass damage during tunnel drilling and blasting construction in Embodiment 1 of the present invention; Figure 3 This is a flowchart of the real-time prediction method for blasted rock mass damage during tunnel drilling and blasting construction in Embodiment 2 of the present invention; Figure 4 This is a schematic diagram of the computer device structure according to Embodiment 4 of the present invention; The system includes: 1. Memory; 2. Processor; 3. Communication interface; 10. Real-time prediction system for blasted rock mass damage during tunnel drilling and blasting construction; 100. Multi-source data acquisition module; 101. Sensor deployment module; 102. High-speed imaging module; 103. Data synchronization module; 200. Intelligent data processing module; 201. Signal noise reduction module; 202. Image recognition module; 203. Acoustic emission analysis module; 300. Hybrid-driven damage prediction module; 301. Physical model calculation module; 302. Data-driven prediction module; 303. Threshold determination module; 400. Dynamic optimization module for blasting parameters; 401. Parameter response library module; 402. Reinforcement learning decision-making module; 403. Execution control module; 500. 3D visualization and early warning module; 501. Damage cloud map fusion module; 502. AR interaction module; 503. Risk report generation module. Detailed Implementation
[0024] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration 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.
[0025] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. 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. The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.
[0026] Example 1: Embodiment 1 of this invention provides a real-time prediction system 10 for blasted rock mass damage in tunnel drilling and blasting construction. This system addresses the problems in related technologies, such as delayed rock mass damage prediction due to reliance on manual experience, inefficient control of over-excavation and under-excavation, and the difficulty in quantifying dynamic damage evolution using traditional single-sensor monitoring. It achieves synergistic optimization of construction safety, efficiency, and cost. Figure 1 and Figure 2 As shown, it includes a multi-source data acquisition module 100, a data intelligent processing module 200, a hybrid-driven damage prediction module 300, a dynamic optimization module for blasting parameters 400, and a three-dimensional visualization and early warning module 500.
[0027] In one specific implementation, a multi-source data acquisition module is used to acquire in real time blasting vibration signals, acoustic emission signals of crack propagation inside the rock mass, and dynamic image data of surface cracks.
[0028] Specifically, the multi-source data acquisition module is used to deploy a triaxial vibration sensor array (frequency band 0~200Hz) at preset intervals (5 meters) along the tunnel outline and within the borehole, along with a broadband acoustic emission probe (20kHz~1MHz) pre-embedded in the borehole. Simultaneously, the explosion-proof high-speed camera captures dynamic evolution images of surface cracks in the rock mass at the moment of blasting. By synchronously acquiring blasting vibration signals, acoustic emission signals of crack propagation within the rock mass, and dynamic images of surface cracks in real time, multi-modal data acquisition significantly improves the comprehensiveness and timeliness of damage monitoring, providing high-precision input for subsequent intelligent analysis.
[0029] The multi-source data acquisition module includes a sensor deployment module 101, a high-speed imaging module 102, and a data synchronization module 103.
[0030] The sensor deployment module automatically plans the placement of vibration sensor arrays and acoustic emission probes based on the tunnel cross-section dimensions and lithological characteristics. A triaxial vibration sensor array is deployed along the tunnel outline at preset intervals (5 meters), and broadband acoustic emission probes (20kHz~1MHz) are installed inside the borehole to collect blasting vibration velocity and crack propagation acoustic signals in real time. The device status is also verified in real time via a wireless transmission network. Figure 2 As shown, the sensor deployment module 101 has a built-in tunnel BIM model interface, which can import design parameters, automatically generate the optimal sensor deployment scheme, and communicate with sensor nodes through the LoRa wireless network to monitor the online status of the equipment in real time.
[0031] The high-speed imaging module is used to control explosion-proof camera equipment to capture dynamic images of cracks at a frame rate of ≥1000fps during blasting, and is equipped with a ring-shaped supplementary lighting unit to adapt to the low-light environment of tunnels.
[0032] The data synchronization module is used to align vibration signals, acoustic emission signals and image sequences through timestamps, generate spatiotemporally consistent multi-source datasets, and compress and transmit them to edge computing terminals.
[0033] In one specific implementation, the data intelligent processing module is used to perform wavelet noise reduction processing on the vibration signal, identify the geometric parameters of the crack based on the crack identification model using dynamic image data of the surface crack, and convert the acoustic emission signal into crack density time series data to generate a spatiotemporally aligned damage feature matrix.
[0034] Specifically, the data intelligence processing module 200 is deployed on the edge computing terminal and supports real-time processing of multi-source data. Its core structured processing is to uniformly encode multi-source heterogeneous data (vibration time-domain waveforms, acoustic emission radio frequency domain signals and image pixel arrays) into a standardized digital matrix.
[0035] The data intelligent processing module includes a signal denoising module 201, an image recognition module 202, and an acoustic emission analysis module 203. The signal denoising module 201 employs a parallel computing architecture; the image recognition module 202 uses a pre-trained YOLO-v5 model, achieving a single-frame image crack recognition accuracy of ≥95%. This embodiment utilizes the data intelligent processing module 200 to process multi-source data in real time, significantly improving the efficiency and accuracy of damage feature analysis and providing reliable input for subsequent hybrid-driven prediction.
[0036] First, time-frequency domain features are extracted from the vibration signal after wavelet denoising, including the energy proportion of each 10Hz subband in the 0-200Hz frequency band, the attenuation coefficient of the main oscillation frequency, and the vector synthesis energy time history of the triaxial vibration. After STFT transformation of the acoustic emission signal, the burst event count rate, cumulative energy release curve, and spectral centroid shift in the high-frequency band (20kHz-1MHz) are calculated. Simultaneously, after YOLO-v5 recognition of the crack image, geometric motion parameters such as the total crack length, fractal dimension, dominant orientation angle, and propagation velocity are output. Then, all the above feature sequences are interpolated to the same time series using a unified time reference. The data is processed in columns (sampling interval 10ms), and spatial location labels are assigned based on the spatial coordinates of each sensor (tunnel outline mileage and radial depth). The resulting damage feature matrix is a four-dimensional tensor data structure with dimensions of [time step × spatial location point × feature dimension × data source identifier]. Each spatial location point corresponds to a spatial grid, and each feature dimension contains 28 standardized feature values in three categories: vibration, acoustic emission, and image. This matrix fully depicts the multi-physics evolution of rock mass damage in the spatiotemporal domain during blasting and can be directly input into subsequent hybrid-driven damage prediction models for damage quantification analysis.
[0037] More specifically: The signal denoising module is used to perform multi-scale decomposition of the original vibration signal based on the wavelet transform algorithm, filter out environmental noise (such as mechanical vibration and human activity interference) from the vibration signal, extract effective frequency band features, and calculate the vibration energy distribution spectrum. Through a parallel computing architecture, the single-channel signal processing time is ≤10ms, ensuring real-time performance. In this embodiment, effective frequency band features from 0 to 200Hz are extracted.
[0038] The image recognition module is used to analyze and identify the geometric parameters of cracks frame by frame in high-speed camera footage using a crack recognition model constructed from a pre-trained YOLO-v5 model. It outputs a structured crack feature table and associates the recognition results with the coordinates of the blasting hole. The crack geometric parameters include crack length, width and orientation.
[0039] In this embodiment, the YOLO-v5 model used is not an existing structure, but rather incorporates three key improvements tailored to the characteristics of high-speed dynamic crack images after tunnel blasting: 1. A deformable convolutional module is embedded in the feature extraction backbone network. By introducing learnable offsets to the convolutional kernels, the accuracy of micro-crack recognition is significantly improved. 2. A small target detection layer specifically for fine cracks is added to the detection head, reducing the minimum recognizable crack width of the original YOLO-v5 from 3mm to 1mm. 3. A temporal correlation mechanism is introduced, fusing features from multiple consecutive frames of images through a lightweight convolutional long short-term memory network. The recognition results of the previous frame constrain the crack tracking of the current frame, suppressing false detections and missed detections caused by dust interference.
[0040] The acoustic emission analysis module is used to perform short-time Fourier transform (STFT) on the acoustic emission signal, calculate the crack density growth rate and the time-series energy accumulation value, and generate the crack evolution time-series curve.
[0041] In actual execution, the acoustic emission analysis module 203 performs short-time Fourier transform on the acoustic emission signal in the 20kHz~1MHz frequency band, calculates the crack density growth rate and the cumulative value of time-series energy, generates the crack evolution time-series curve by combining the peak detection algorithm, and eliminates signal jitter by Kalman filtering.
[0042] In one specific implementation, a hybrid-driven damage prediction module is used to predict damage based on fracture geometric parameters and damage feature matrices using a damage prediction model, obtaining damage prediction results and adaptively determining damage thresholds for different lithologies. The damage prediction results include rock mass damage depth and dynamic damage index. The damage prediction model consists of a material constitutive model, a regression model, and a deep learning model. The inputs are the initial integrity coefficient of the rock mass, stress wave superposition effect parameters, and the real-time damage feature matrix. In this embodiment, the material constitutive model uses the HJC (Holmquist-Johnson-Cook) constitutive equation; the regression model uses the XGBoost regression model; and the deep learning model uses the LSTM model.
[0043] In actual implementation, the hybrid-driven damage prediction module 300 employs a physics-data joint training strategy, combining physical mechanisms with a data-driven model for damage prediction: the physical model calculation module 301 provides the initial damage field as prior knowledge, the data-driven prediction module 302 updates the model weights online using measured data, and the threshold determination module 303 dynamically corrects the warning threshold based on the initial rock mass integrity coefficient (RQD value), forming an intelligent prediction system that adapts to complex geological conditions. This embodiment, by integrating explosion mechanics mechanisms and machine learning algorithms through the hybrid-driven damage prediction module 300, significantly improves the accuracy and generalization of damage prediction under complex geological conditions, providing a scientific basis for dynamically optimizing blasting parameters.
[0044] In this embodiment, the initial rock mass integrity coefficient is derived from geological survey reports or borehole core experiments, and is a known static parameter. It is transmitted to the hybrid-driven damage prediction module as part of the damage feature matrix in the output of the data intelligent processing module. The derivation process of the stress wave superposition effect parameter is as follows: based on the detonation sequence of each borehole and the rock mass wave velocity, the time difference of the stress wave reaching each spatial point is calculated, and the correction coefficient is dynamically calculated using empirical formulas. The stress wave superposition effect parameter is an intermediate derivation parameter output by the physical model calculation module, used to correct the initial damage field.
[0045] The hybrid-driven damage prediction module includes a physical model calculation module 301, a data-driven prediction module 302, and a threshold determination module 303.
[0046] The physical model calculation module is used to simulate the dynamic damage evolution of rock mass under explosive loading based on the HJC constitutive equation, and introduces the stress wave superposition effect parameter. In this embodiment, the inputs to the HJC constitutive equation are rock mass density, elastic modulus, and dynamic compressive strength. The finite element solver simulates the stress wave superposition effect of porous blasting, and a correction factor is introduced to accurately quantify the energy dissipation and crack propagation path in jointed rock mass.
[0047] The physical model calculation module first constructs the material parameter set required for the HJC constitutive equation based on static parameters such as rock mass density, elastic modulus, and compressive strength obtained from geological surveys and laboratory tests, combined with damage thresholds derived from empirical formulas. Next, it imports dynamic parameters from the blasting design files, including charge quantity, borehole spacing, detonation sequence, and borehole coordinates, and calculates and derives the explosion load time history for each borehole using the Heinrich formula. Then, it establishes a three-dimensional finite difference mesh using tunnel geometry data from the BIM model and a defined computational domain, applies material parameters and explosion loads to the model for explicit dynamic solving, and calls the HJC constitutive subroutine at each time step to calculate and output the initial elements. Damage field; subsequently, the module performs stress wave interferometry analysis in conjunction with the borehole detonation sequence, and compares it with measured vibration waveform data (if available) from the multi-source data acquisition module. The stress wave superposition correction factor is derived through empirical formulas and dynamic calculation based on measured feedback to correct the initial damage field. Finally, key indicators are extracted from the corrected damage field, including the damage depth calculated along the contour normal by a search algorithm and the damage index calculated by volume weighted average. These results are verified and calibrated with measured data such as borehole cameras, and the final damage depth, damage index, and calculation confidence level are output, thus completing the closed-loop calculation process from multi-source data input to quantitative damage assessment.
[0048] A data-driven prediction module is used to predict damage depth D using an XGBoost regression model and the dynamic curve of the damage index DI using an LSTM network. The input features of the XGBoost regression model are peak velocity (PPV), initial rock mass integrity coefficient (RQD), and crack density growth rate. The model is trained based on over 500 sets of field blasting test data, and 5-fold cross-validation is used to ensure generalization.
[0049] The data-driven prediction module 302 in this embodiment has made three core improvements on the basis of traditional methods: a cascaded network structure with cross-fusion of features is constructed, and the damage depth prediction value output by XGBoost is used as an additional temporal feature input to the LSTM network; a dynamic feature weight allocation mechanism is introduced, which automatically adjusts the contribution weights of PPV, RQD and crack density in different lithologies through the attention layer; and an online incremental learning module is designed to continuously update the model parameters using the measured data after each blasting cycle. This module takes the preprocessed results output by the data intelligent processing module as direct input—including the PPV peak value extracted by wavelet denoising, the RQD value provided by geological data, and the crack density obtained by fusing acoustic emission analysis and image recognition. After processing such as outlier filtering, normalization, and time-series alignment, the XGBoost regression model (trained based on historical blasting data) first calculates and outputs a scalar prediction value of the damage depth D, which directly quantifies the longitudinal extension of the physical range of rock mass damage. Then, the LSTM network receives the fused features including the PPV time-series sequence, crack density evolution, static RQD, and the aforementioned damage depth D prediction value. Through its time-series modeling capabilities, it outputs a dynamic curve of the damage index DI. This curve, through its peak value, area under the curve, and stabilization time, fully depicts the dynamic evolution of rock mass damage from occurrence and development to stabilization, thus providing directly quantifiable data for blasting parameter optimization and support decisions.
[0050] The threshold determination module is used to dynamically adjust the critical damage threshold based on the initial rock mass integrity coefficient (RQD) and output the over-excavation risk level. In this embodiment, the over-excavation risk level is set to three levels: low, medium, and high.
[0051] In this embodiment, the threshold determination module 303 has a built-in lithology adaptive rule base: Hard rock conditions (RQD>75%): Set the critical damage threshold DI=0.8, and the over-excavation risk level is "low"; Medium-grade rock mass (50%≤RQD≤75%): Set DI=0.7, risk level is "medium"; Fractured rock mass (RQD<50%): Set DI=0.6, risk level is "high".
[0052] In actual implementation, the threshold determination module 303 receives ground-penetrating radar scanning data in real time, dynamically corrects the early warning threshold, and pushes risk level prompts through the AR interface.
[0053] In one specific implementation, the dynamic optimization module for blasting parameters generates adjustment schemes based on damage prediction results. It optimizes a multi-objective function using a reinforcement learning algorithm and coordinates the automatic charging machine and detonator controller to execute commands, achieving closed-loop control. This embodiment's adjustment scheme includes adjustments to the charge quantity, hole spacing, and electronic detonator initiation sequence. In actual execution, such as... Figure 2 As shown, the dynamic optimization module for blasting parameters 400 achieves real-time communication with hardware devices through ROS (Robot Operating System). The parameter response library module 401 has more than 2,000 historical blasting case rules built in, the reinforcement learning decision module 402 updates the policy network weights once per cycle, and the execution control module 403 supports batch command issuance from the charging machine and detonators.
[0054] The dynamic optimization module for blasting parameters includes a parameter response library module 401, a reinforcement learning decision module 402, and an execution control module 403. The parameter response library module 401 receives rock mass damage depth and dynamic damage index output from the hybrid-driven damage prediction module as core inputs, and retrieves expert experience rules from the historical case library to generate a preliminary set of blasting parameter adjustment suggestions through fuzzy logic reasoning. This suggestion is then input as part of the status input, along with real-time vibration and fracture characteristic data from the data intelligence processing module, to the reinforcement learning decision module 402. This module performs collaborative optimization of multiple objectives such as charge quantity, hole spacing, and delay based on the PPO algorithm framework, calculating and outputting a final optimization scheme with confidence assessment. This scheme is ultimately passed to the execution control module 403, which precisely parses it into specific charge density commands and millisecond-level detonation timing sequences, driving the automatic charging machine and digital electronic detonators to execute them respectively. Simultaneously, the execution status is fed back to the 3D visualization and early warning module in real time, thus forming a closed-loop control link from intelligent decision-making to precise execution. In actual implementation, this embodiment uses the dynamic optimization module 400 for blasting parameters to convert the predicted rock damage index (DI) and depth (D) into operable blasting parameter adjustment instructions, which significantly reduces the over-excavation rate and improves construction efficiency.
[0055] More specifically: The parameter response library module is used to store the mapping rule base and supports the fusion of fuzzy logic reasoning and expert experience.
[0056] This embodiment's mapping rule base includes mapping rules between damage prediction results and parameters such as charge quantity, hole spacing, and delay. The parameter response library module 401 presets multiple adjustment rules based on historical blasting cases and expert experience. For example, when the damage index (DI) is greater than 0.6, the charge quantity is reduced by 15% and the hole spacing is reduced by 20%; or when the rock integrity coefficient (RQD) is less than 50% and the damage depth (D) exceeds 1.5 meters, the detonation sequence is adjusted to a per-hole micro-delay. Furthermore, the parameter response library module 401 supports dynamic updates to the rule base. When the deviation between the on-site detection results and the predicted values exceeds 10%, a manual review mechanism is automatically invoked, where engineers correct the rule weights and reload the code to ensure the timeliness and accuracy of the rule base.
[0057] The reinforcement learning decision-making module is used to train agents through the PPO algorithm, generate multi-objective collaborative optimization schemes, and calculate parameters and adjust confidence in real time.
[0058] In this embodiment, the reinforcement learning decision-making module trains the agent based on historical blasting data and real-time monitoring information. It generates parameter optimization schemes with the goal of minimizing the damage index (DI) and over-excavation, and calculates the confidence level of parameter adjustments in real time. For example, when well-developed rock joints are detected in a certain area, the algorithm automatically reduces the charge density and shortens the hole spacing, while optimizing the detonation sequence to reduce the stress wave superposition effect. After the model training is completed, the reliability of the adjustment scheme is verified using actual construction data to ensure that the confidence level is not less than 90%.
[0059] In this embodiment, DI is a dynamic damage quantification index output by the hybrid-driven damage prediction module. The physical model calculation module of this module simulates the evolution of rock damage field under explosive load based on the HJC constitutive equation and outputs the initial damage field. The data-driven prediction module uses an LSTM network to combine real-time acquired vibration signals, acoustic emission signals and fracture image features to perform time-series prediction and dynamic correction of the damage field, and finally fuses them to obtain a DI value in the range of 0 to 1. The closer the DI is to 1, the more severe the rock damage is, and the closer it is to 0, the more intact the rock is.
[0060] It should be noted that this embodiment balances the damage index and over-excavation amount based on the Pareto optimal multi-objective optimization strategy. Specifically, in the reinforcement learning decision module, when training the agent through the PPO algorithm, DI and over-excavation amount are regarded as two mutually restraining optimization objectives, and a weighted summation method is introduced to construct a comprehensive reward function. The weight coefficients are dynamically adjusted according to the project priority. During the exploration process, the agent gradually learns how to find the optimal compromise between the two. At the same time, constraints are set to prevent any index from exceeding the safety threshold, thereby ensuring that the optimization scheme achieves a reasonable balance between the surrounding rock stability and construction efficiency.
[0061] In addition, the calculation basis for the parameter adjustment confidence level in this embodiment is dynamically generated by the reinforcement learning decision module based on the matching degree between historical decision experiences and the current input state, which includes three aspects: first, the consistency degree between the current input data and the distribution of training samples; second, the prediction variance of the policy network output by the PPO algorithm in this decision, where a smaller variance indicates that the model is more certain about the current decision; third, the success rate statistics of this type of optimization scheme in historical implementations, where a good achieved effect increases the current confidence level, and the opposite reduces it.
[0062] An execution control module, configured to send a charging density instruction to the automatic charging machine, write millisecond-level delay detonation timing into the digital electronic detonator, and synchronously feed back the execution status to the 3D visualization and early warning module.
[0063] The execution control module 403 sends a charging density instruction to the automatic charging machine via the industrial Ethernet protocol, and writes millisecond-level detonation timing into the digital electronic detonator. After the execution is completed, the module transmits the charging completion status and the detonator ready signal back to the 3D visualization module in real time, forming a complete closed-loop control process.
[0064] In a specific implementation, the 3D visualization and early warning module is configured to superimpose the damage cloud map onto the tunnel BIM model, mark high-risk areas in real time, display parameter optimization suggestions and damage evolution trend early warnings through the AR interface, and simultaneously generate construction logs and risk reports. In the actual execution process, the 3D visualization and early warning module 500 renders the damage cloud map through the Unity3D engine, the AR interaction module 502 supports multi-person collaborative operation, the risk report generation module 503 automatically generates reports in PDF / Excel formats based on templates, and synchronizes the reports to the cloud management platform via the 5G network. In this embodiment, the 3D visualization and early warning module 500 dynamically integrates the rock mass damage prediction result with the 3D tunnel model, which significantly improves the efficiency of construction personnel in understanding the damage distribution, and enhances the intuitiveness of risk early warning, thereby reducing the over-excavation risk and ensuring construction safety.
[0065] The 3D visualization and early warning module includes a damage cloud map fusion module 501, an AR interaction module 502, and a risk report generation module 503.
[0066] The damage cloud map fusion module is configured to map the damage index values to a heat map, superimpose the heat map onto the corresponding blasting area of the tunnel BIM model, and mark the damage depth.
[0067] The damage cloud map fusion module 501 marks areas with DI>0.6 as red early warning, 0.4<DI≤0.6 as yellow prompt, and DI≤0.4 as green safe area according to the DI value and D value output by the hybrid-driven damage prediction module, and simultaneously displays the damage depth distribution in the form of contour lines. The module supports comparison with the designed section model, automatically calculates the volume of over-excavation and under-excavation areas, and updates the result to the visualization interface in real time.
[0068] The AR interaction module is used to display 3D calibration boxes of high-risk areas and floating prompts for optimized parameters through head-mounted devices, and supports gesture control to switch perspectives or retrieve historical damage data.
[0069] In actual operation, the AR interactive module 502 projects damage heatmaps, parameter adjustment schemes, and historical data comparisons onto the field of vision of construction personnel. For example, when a sudden increase in the DI value of a certain area is detected, a red warning box automatically pops up on the AR interface, suggesting "shortening the hole spacing to 0.8 meters." The module supports gesture control to zoom the view or retrieve adjacent cycle damage data, facilitating rapid on-site decision-making.
[0070] The risk report generation module is used to automatically export damage distribution statistics (over-excavation rate, DI mean), support cost comparison analysis chart, and construction log.
[0071] The risk report generation module 503 automatically generates PDF / Excel reports based on damage data and parameter adjustment records collected in each cycle. These reports include over-excavation rate statistics, support cost comparisons, and safety risk ratings. The reports are synchronized to a cloud management platform via a 5G network, supporting multi-terminal access and historical data review.
[0072] It should be noted that the color thresholds of the heatmap, AR prompts and report templates mentioned above can be customized by technical personnel according to the actual needs of the project, and no specific limitations are made here.
[0073] To better illustrate the superiority of the system in this embodiment, the working principle of the embodiment of this application will be described in detail below with a specific example.
[0074] First, this embodiment checks whether the vibration sensor array, acoustic emission probe, and explosion-proof high-speed camera of the multi-source data acquisition module are working properly, verifies the timestamp accuracy of the data synchronization module, and sets the sensor deployment plan according to the tunnel cross-section dimensions. Real-time data acquisition is initiated through the edge computing terminal, combined with manual inspection to confirm the equipment status, ensuring the synchronous capture of blasting vibration signals, acoustic emission signals of crack propagation inside the rock mass, and crack images.
[0075] Next, in this embodiment, the original vibration signal is subjected to wavelet noise reduction processing through the data intelligent processing module to extract the energy distribution spectrum of the effective frequency band from 0 to 200 Hz. At the same time, the YOLO-v5 model is used to identify the crack length, width and orientation angle in the high-speed camera footage to generate a spatiotemporally aligned damage feature matrix.
[0076] Secondly, in this embodiment, the initial rock mass integrity coefficient (RQD value), stress wave superposition parameters, and damage feature matrix are input into the hybrid-driven damage prediction module. The damage evolution under explosive loading is simulated through the HJC constitutive equation, and the damage depth D is predicted by combining the XGBoost regression model. The LSTM network outputs the dynamic damage index DI curve. When a region with DI > 0.6 is detected, the module automatically marks it as a high-risk area.
[0077] Furthermore, in this embodiment, based on the damage prediction results, an optimization scheme is generated through a reinforcement learning algorithm (PPO), and the control module is linked to send a charge density command to the automatic charge machine and write the hole-by-hole micro-differential detonation sequence to the digital electronic detonator.
[0078] Finally, in this embodiment, a 3D visualization module overlays the damage cloud map onto the tunnel BIM model. The AR interface marks high-risk areas in real time and automatically generates a construction report including over-excavation comparisons and support cost savings rates. After a single cycle task is completed, the edge computing terminal cache is cleared to prepare for the next round of data collection and optimization.
[0079] In summary, this embodiment, based on multi-source data fusion, physical-data hybrid modeling, and real-time feedback control technology, achieves second-level prediction of rock mass damage and dynamic optimization of blasting parameters, completely replacing the traditional manual experience-based adjustment method, reducing over-excavation rate, reducing support costs, and standardizing the operation process so that construction personnel can quickly master it without complicated training, significantly improving the safety, economy, and intelligence level of tunnel drilling and blasting construction.
[0080] Example 2: Embodiment 2 of this invention provides a real-time prediction method for blasted rock mass damage during tunnel drilling and blasting construction, comprising the following steps: real-time acquisition of blasting vibration signals, acoustic emission signals of crack propagation inside the rock mass, and dynamic image data of surface cracks; wavelet denoising processing of the vibration signals; identification of crack geometric parameters using the dynamic image data of surface cracks based on a crack identification model; conversion of acoustic emission signals into crack density time-series data; generation of a spatiotemporally aligned damage feature matrix; damage prediction using a damage prediction model on the crack geometric parameters and damage feature matrix to obtain damage prediction results; adaptive determination of damage thresholds for different lithologies; wherein the damage prediction results include rock mass damage depth and dynamic damage index; the damage prediction model consists of a material constitutive model, a regression model, and a deep learning model; generation of adjustment schemes based on the damage prediction results; optimization of multi-objective functions using reinforcement learning algorithms; and linkage of automatic charging machine and detonator controller to execute commands.
[0081] like Figure 3 As shown, the specific steps include: S301: According to the preset sensor deployment plan, deploy vibration sensor arrays and acoustic emission probes in the tunnel outline and borehole, and simultaneously start the explosion-proof high-speed camera equipment to perform synchronous acquisition of blasting vibration signals, acoustic emission signals of crack propagation inside the rock mass, and dynamic images of surface cracks.
[0082] S302: Perform wavelet noise reduction on the acquired vibration signal, extract the effective frequency band features from 0 to 200 Hz, and identify the crack geometric parameters through the YOLO-v5 model to generate a spatiotemporally aligned damage feature matrix.
[0083] S303: Based on the HJC constitutive equation and the XGBoost / LSTM hybrid driving model, it takes the initial rock mass integrity coefficient (RQD), stress wave superposition parameters and real-time damage feature matrix as input, outputs the dynamic damage index (DI) and damage depth (D) of the rock mass, and adaptively determines the damage threshold of different lithologies.
[0084] S304: Based on the damage prediction results, the charge amount, hole spacing and electronic detonator initiation sequence are optimized through reinforcement learning algorithm (PPO) to generate a multi-objective collaborative adjustment scheme, and the automatic charger and detonator controller are linked to execute feedback commands.
[0085] S305: Integrates damage cloud maps with tunnel BIM models, marks high-risk areas in real time, displays damage evolution trends, parameter optimization suggestions and safety warning information through AR interface, and generates over-excavation comparison analysis reports and construction logs. S306: Based on the damage control effect of real-time feedback, the parameters of the hybrid driving model are dynamically updated to achieve closed-loop optimization of blasting damage prediction and construction technology.
[0086] It should be noted that the explanation of the above-mentioned embodiment of the real-time prediction system for blasted rock mass damage in tunnel drilling and blasting construction also applies to the real-time prediction and feedback method of this embodiment, and will not be repeated here.
[0087] The real-time prediction system for blasting rock mass damage in tunnel drilling and blasting construction proposed in this application can acquire blasting vibration signals, acoustic emission signals of crack propagation inside the rock mass, and dynamic image data of surface cracks in real time through a multi-source data acquisition module. The vibration signals are then processed by a data intelligent processing module to perform wavelet noise reduction and identify crack geometric parameters, generating a spatiotemporally aligned damage feature matrix. The hybrid-driven damage prediction module combines the HJC constitutive equation and machine learning algorithm to output the dynamic damage index (DI) and damage depth (D) of the rock mass and adaptively determine the damage threshold for different lithologies. The dynamic optimization module for blasting parameters generates adjustment schemes for charge quantity, hole spacing, and detonation sequence, and links the automatic charging machine and electronic detonators to execute feedback commands. Finally, the 3D visualization module integrates the damage cloud map with the tunnel BIM model, marks high-risk areas in real time, and displays optimization suggestions, thereby significantly reducing reliance on manual experience and improving the accuracy of rock mass damage prediction and construction safety. This solves the problems in related technologies, such as the lag in rock mass damage prediction caused by relying on manual experience to adjust blasting parameters, the extensive control of over-excavation and under-excavation, and the difficulty in quantifying dynamic damage evolution by traditional single-sensor monitoring, thus achieving synergistic optimization of construction efficiency, cost, and safety.
[0088] Example 3: Embodiment 3 of the present invention provides a computer-readable storage medium storing a computer program adapted to be loaded by a processor and executed the steps of the real-time prediction method for blasted rock mass damage in tunnel drilling and blasting construction as described in Embodiment 1 of the present invention.
[0089] Example 4: Embodiment 4 of the present invention provides a computer device, the device comprising: Memory 1, processor 2, and computer program stored on memory 1 and capable of running on processor 2.
[0090] When processor 2 executes the program, it implements the real-time prediction and feedback method for the degree of damage to blasted rock mass during tunnel drilling and blasting construction provided in the above embodiments.
[0091] Furthermore, electronic devices also include: Communication interface 3 is used for communication between memory 1 and processor 2.
[0092] Memory 1 is used to store computer programs that can run on processor 2.
[0093] Memory 1 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.
[0094] If memory 1, processor 2, and communication interface 3 are implemented independently, then communication interface 3, memory 1, and processor 2 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 4 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.
[0095] Optionally, in a specific implementation, if the memory 1, processor 2, and communication interface 3 are integrated on a single chip, then the memory 1, processor 2, and communication interface 3 can communicate with each other through an internal interface.
[0096] Processor 2 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.
[0097] Example 5: Embodiment 5 of the present invention provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps in the real-time prediction method for blasted rock mass damage during tunnel drilling and blasting construction as described in Embodiment 1 of the present invention.
[0098] The steps and methods involved in Examples 2, 3, 4 and 5 above correspond to those in Example 1. For specific implementation methods, please refer to the relevant description section of Example 1.
[0099] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0100] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium can be any available medium that a computer can access or a data processing device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium, an optical medium, or a semiconductor medium, etc.
[0101] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A real-time prediction system for blasted rock mass damage during tunnel drilling and blasting construction, characterized in that, include: The multi-source data acquisition module is used to acquire blasting vibration signals, acoustic emission signals of crack propagation inside the rock mass, and dynamic image data of surface cracks in real time. The data intelligent processing module is used to perform wavelet noise reduction on vibration signals, identify crack geometric parameters based on surface crack dynamic image data using crack identification model, and convert acoustic emission signals into crack density time series data to generate a spatiotemporally aligned damage feature matrix. The hybrid-driven damage prediction module is used to predict damage based on the fracture geometric parameters and damage feature matrix using a damage prediction model, obtain damage prediction results, and adaptively determine damage thresholds for different lithologies. The damage prediction results include rock mass damage depth and dynamic damage index. The damage prediction model consists of a material constitutive model, a regression model, and a deep learning model. The dynamic optimization module for blasting parameters is used to generate adjustment plans based on damage prediction results, optimize multi-objective functions through reinforcement learning algorithms, and link the automatic charging machine and detonator controller to execute commands.
2. The real-time prediction system for blasted rock mass damage during tunnel drilling and blasting construction as described in claim 1, characterized in that, The data intelligence processing module includes: The signal noise reduction module is used to filter out environmental noise in vibration signals based on wavelet transform algorithm, extract effective frequency band features, and calculate vibration energy distribution spectrum; The image recognition module is used to identify the geometric parameters of the cracks through the crack recognition model constructed by the YOLO-v5 model, output a structured crack feature table, and associate it with the coordinates of the blasting hole. The acoustic emission analysis module is used to perform short-time Fourier transform on the acoustic emission signal, calculate the crack density growth rate and the cumulative value of time-series energy, and generate the crack evolution time-series curve.
3. The real-time prediction system for blasted rock mass damage during tunnel drilling and blasting construction as described in claim 1, characterized in that, The hybrid-driven damage prediction module includes: The physical model calculation module is used to simulate the dynamic damage evolution of rock mass under explosive loading based on the HJC constitutive equation, and introduces the stress wave superposition effect parameter. A data-driven prediction module is used to predict damage depth through an XGBoost regression model and to predict the dynamic curve of the damage index through an LSTM network. The threshold determination module is used to dynamically adjust the critical damage threshold based on the initial integrity coefficient of the rock mass and output the over-excavation risk level.
4. The real-time prediction system for blasted rock mass damage during tunnel drilling and blasting construction as described in claim 1, characterized in that, The dynamic optimization module for blasting parameters includes: The parameter response library module is used to store the mapping rule base and supports the fusion of fuzzy logic reasoning and expert experience. A reinforcement learning decision-making module is used to train an agent using the PPO algorithm, generate multi-objective collaborative optimization schemes, and calculate parameters in real time to adjust confidence levels. The execution control module is used to send charge density commands to the automatic charger and write millisecond-level delay detonation timing sequences to the digital electronic detonator, and synchronously feed back the execution status to the three-dimensional visualization and early warning module.
5. The real-time prediction system for blasted rock mass damage during tunnel drilling and blasting construction as described in claim 1, characterized in that, It also includes a 3D visualization and early warning module, which overlays damage cloud maps onto the tunnel BIM model, marks high-risk areas in real time, displays parameter optimization suggestions and damage evolution trend warnings through an AR interface, and generates construction logs and risk reports.
6. The real-time prediction system for blasted rock mass damage during tunnel drilling and blasting construction as described in claim 5, characterized in that, The 3D visualization and early warning module includes: The damage cloud map fusion module is used to map the damage index value into a heat map, overlay it onto the corresponding blasting area of the tunnel BIM model, and mark the damage depth. The AR interactive module is used to display the 3D calibration box of high-risk areas, optimize parameter floating prompts, and support gesture control to switch perspectives or retrieve historical damage data. The risk report generation module is used to automatically export damage distribution statistics tables, support cost comparison analysis charts, and construction logs.
7. A method for real-time prediction of blasted rock mass damage during tunnel drilling and blasting construction, characterized in that, Includes the following steps: Real-time acquisition of blasting vibration signals, acoustic emission signals of crack propagation inside the rock mass, and dynamic image data of surface cracks; The vibration signal is subjected to wavelet noise reduction processing. Based on the crack identification model, the surface crack dynamic image data is used to identify the crack geometric parameters. The acoustic emission signal is converted into crack density time series data to generate a spatiotemporally aligned damage feature matrix. Damage prediction is performed using a damage prediction model to predict the geometric parameters of the fracture and the damage feature matrix, and the damage prediction results are obtained. The damage threshold for different lithologies is adaptively determined. The damage prediction results include the rock mass damage depth and dynamic damage index. The damage prediction model consists of a material constitutive model, a regression model and a deep learning model. An adjustment plan is generated based on the damage prediction results. The multi-objective function is optimized through reinforcement learning algorithm, and the automatic loading machine and detonator controller are linked to execute the instructions.
8. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the real-time prediction method for blasted rock mass damage during tunnel drilling and blasting construction as described in claim 7.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed as described in claim 7, which is a method for real-time prediction of blasted rock mass damage during tunnel drilling and blasting.
10. A computer device, characterized in that, include: Memory, processor, and computer programs stored in memory and capable of running on the processor. When the processor executes the program, it implements the real-time prediction method for blasted rock mass damage during tunnel drilling and blasting construction as described in claim 7.
