Underground cavern construction method generation method and system based on multi-source data fusion

Abstract

The invention provides an underground cavern construction method generation method and system based on multi-source data fusion, and the method comprises the steps: collecting geological exploration, construction machinery operation and environment monitoring data, carrying out the time-space alignment processing, generating a standardized monitoring data set, carrying out the multi-source feature extraction of the standardized monitoring data set, and carrying out the multi-source feature extraction of the standardized monitoring data set; obtaining a construction state feature set and a geological risk feature set, inputting a reinforcement learning model for multi-target parameter optimization to obtain a preliminary construction method parameter set, inputting the preliminary construction method parameter set into a digital twin simulation system for simulating mechanical response, and iteratively optimizing parameters when a preset safety threshold is exceeded to obtain a preliminary construction method parameter set; the optimized construction method parameter set is matched and verified with a preset construction specification knowledge base, manual correction is carried out when illegal items exist, finally, full-period risk prediction is carried out on the corrected construction method parameter set, dynamic regulation and control are carried out according to a construction risk distribution map, a final construction method instruction set is generated, and scientific and intelligent construction method generation is achieved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method and system for generating an underground cavern construction method based on multi-source data fusion. Background Art

[0002] In the field of underground cavern construction, traditional construction methods such as drilling and blasting, shield tunneling, and layered excavation and support have long dominated. These methods are highly dependent on the experience and judgment of construction personnel. When faced with complex and changing geological conditions, they often show problems such as poor geological adaptability, prominent safety hazards, and low process coordination efficiency. With the expansion of project scale and the increasing complexity of geological conditions, the limitations of traditional construction methods have become increasingly obvious, and it is difficult to meet the comprehensive requirements of modern underground cavern construction for safety, efficiency, and economy.

[0003] On the other hand, although the advanced forecasting and stability monitoring technology for complex geological conditions can provide certain risk warnings, the above technology has the problem of delayed data updates. In actual construction, geological conditions may change at any time, and the delayed risk warning information cannot guide the dynamic adjustment of the construction plan in time, which makes the construction face potential safety risks. Summary of the invention

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a method for generating an underground cavern construction method based on multi-source data fusion, the method comprising: Collecting geological exploration data, construction machinery operation data and environmental monitoring data, performing spatiotemporal alignment processing on the geological exploration data, construction machinery operation data and environmental monitoring data, and generating a standardized monitoring data set; Performing multi-source feature extraction on the standardized monitoring data set to obtain a construction status feature set and a geological risk feature set, inputting the construction status feature set and the geological risk feature set into a reinforcement learning model to perform multi-objective parameter optimization processing to generate a preliminary construction method parameter set; Inputting the preliminary construction method parameter set into the digital twin simulation system for mechanical response simulation processing, and triggering the iterative optimization processing of the construction method parameters to generate an optimized construction method parameter set when the result of the mechanical response simulation processing exceeds a preset safety threshold; Perform compliance matching processing on the optimized construction method parameter set and the preset construction specification knowledge base to generate a construction method compliance verification result, and when there are violations in the construction method compliance verification result, trigger manual correction processing to generate a corrected construction method parameter set; The revised construction method parameter set is subjected to full-cycle risk prediction processing to generate a construction risk distribution map, and the revised construction method parameter set is subjected to dynamic control processing according to the construction risk distribution map to generate a final construction method instruction set.

[0005] On the other hand, an embodiment of the present invention also provides an underground cavern construction method generation system based on multi-source data fusion, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0006] Based on the above aspects, the embodiment of the present invention realizes the full-chain intelligent decision-making of underground cavern construction method generation through multi-source data fusion and intelligent algorithm collaborative drive. First, the discrete geological, mechanical and environmental data are converted into a standardized data set with spatiotemporal consistency through time-space alignment processing, and then the multi-dimensional features are optimized for multi-objective parameters by using a reinforcement learning model, breaking through the experience-dependent limitations of traditional method parameter determination, and realizing the dynamic adaptation of construction status and geological risks. Further, a virtual mechanical response model is constructed through a digital twin simulation system, and the reliability and safety of method parameters are significantly improved by combining the safety threshold triggering iterative optimization mechanism; on this basis, a preset construction specification knowledge base is introduced for compliance verification, which not only ensures the standardization of method parameters, but also retains the channel for expert experience integration through the manual correction mechanism; finally, through full-cycle risk prediction and dynamic regulation, the construction risk distribution map is converted into an executable method instruction set, realizing the full-process intelligent decision-making closed loop from data collection to method generation. As a result, not only the efficiency and scientificity of construction method generation are improved, but also a construction method generation system with autonomous optimization capabilities is constructed through the deep coupling of multimodal data fusion and intelligent algorithms. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 It is a schematic diagram of the execution flow of the method for generating an underground cavern construction method based on multi-source data fusion provided in an embodiment of the present invention.

[0008] Figure 2 It is a schematic diagram of exemplary hardware and software components of an underground cavern construction method generation system based on multi-source data fusion provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0009] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1It is a flow chart of a method for generating an underground cavern construction method based on multi-source data fusion provided by an embodiment of the present invention. The method for generating an underground cavern construction method based on multi-source data fusion is introduced in detail below.

[0010] This embodiment takes a water diversion tunnel project of a pumped storage power station as an example to elaborate on the specific implementation process of the underground cavern construction method generation method based on multi-source data fusion.

[0011] Step S110: Collect geological exploration data, construction machinery operation data and environmental monitoring data, perform spatiotemporal alignment processing on the geological exploration data, construction machinery operation data and environmental monitoring data, and generate a standardized monitoring data set.

[0012] In this embodiment, in order to comprehensively and accurately obtain data related to underground cavern construction, it is necessary to collect multi-source data. In detail, for geological exploration data, a SIR-4000 geological radar can be used, with an antenna frequency of 100MHz, and a horizontal survey line is arranged every 20m along the axis of the cave to record parameters such as dielectric constant and wave velocity, with a detection accuracy of ±2%. At the same time, drilling core data and ground stress measurement data can also be collected. During the collection of construction machinery operation data, a three-axis vibration sensor can be installed on the three-arm rock drilling trolley, for example, with a range of ±50g and a sampling frequency of 1kHz, to obtain drill bit vibration frequency data; at the same time, a Beidou high-precision positioning terminal can be installed, with a horizontal error of ≤2cm and an elevation error of ≤5cm, to obtain the positioning data of the rock drilling trolley, and also record the load data of the slag truck. Environmental monitoring data is collected by deploying convergence meters (range 0-50mm, resolution 0.01mm) and piezometers (range 0-2MPa, accuracy 0.1%FS), with a set of monitoring points arranged every 5m, to obtain convergence deformation data in the cave, groundwater level change data and air composition monitoring data.

[0013] Among them, the geological exploration data mainly includes borehole core data, geological radar scanning data and ground stress measurement data. The borehole core data obtains information such as the physical and mechanical properties and structural characteristics of the rock by analyzing the core taken out of the borehole, providing a basis for judging the strength and stability of the rock. The geological radar scanning data uses geological radar technology to detect underground geological structures, and can identify the location and scale of geological structures such as faults and folds, as well as changes in the stratigraphic interface. The ground stress measurement data obtains the stress state of the underground rock mass, including the magnitude and direction of the ground stress, through professional ground stress measurement equipment.

[0014] The construction machinery operation data covers the drilling rig positioning data, drill bit vibration frequency data and slag truck load data. The drilling rig positioning data is obtained through the Beidou high-precision positioning terminal installed on the drilling rig, which can track the position and movement trajectory of the drilling rig in real time and reflect the progress of the construction. The drill bit vibration frequency data is collected through a three-axis vibration sensor, which is closely related to the hardness and drilling efficiency of the rock, and can be used to judge the working status of the drill bit and the physical properties of the rock. The slag truck load data records the weight of rock waste transported by the slag truck each time, which indirectly reflects the progress and efficiency of excavation.

[0015] The environmental monitoring data include convergence and deformation data in the cave, groundwater level change data and air composition monitoring data. The convergence and deformation data in the cave are monitored in real time by a convergence meter, which is an important indicator for measuring the stability of the cave. The groundwater level change data is obtained by a piezometer, which reflects the dynamic changes of the groundwater level and has an important impact on the stability of the cave and construction safety. The air composition monitoring data uses air quality monitoring equipment to monitor the oxygen content, carbon monoxide content, dust concentration and other air components in the cave in real time to ensure the health and safety of construction workers.

[0016] Wherein, step S110 may include: Step S111: establishing a conversion mapping relationship between a geological exploration coordinate system and a construction machinery coordinate system, and performing continuous geological interface generation processing on discrete exploration points in the geological exploration data to obtain a three-dimensional geological model data set.

[0017] In this embodiment, the data collected by the geological radar uses the WGS84 coordinate system, while the construction requires the construction coordinate system (the origin is the center point of the cave entrance), so the seven-parameter method can be used for coordinate system conversion, and the translation parameters ΔX=123.456m, ΔY=45.678m, ΔZ=-3.210m. For the discrete exploration points in the collected geological exploration data, the spherical semivariogram model (nugget value 0.1, base value 1.8, range 25m) can be used for Kriging interpolation, and the fault zone area is encrypted (grid size 0.3m×0.3m) to generate a 0.5m×0.5m gridded three-dimensional geological model data set with an error of ≤3cm.

[0018] Step S112: Dynamically calibrate the position coordinate data in the construction machinery operation data to generate a construction machinery real-time positioning data set, and spatially overlay the construction machinery real-time positioning data set with the three-dimensional geological model data set to generate a machinery-geology spatial association data set.

[0019] In this embodiment, the position coordinate data is obtained by the Beidou high-precision positioning terminal installed on the drilling rig, and the above position coordinate data is dynamically calibrated to generate a real-time positioning data set of the construction machinery. Then, the real-time positioning data set of the construction machinery is spatially superimposed with the three-dimensional geological model data set generated previously, for example, the specific position of the drilling rig in the three-dimensional geological model is determined, thereby generating a mechanical-geological spatial association data set, which can clearly observe the working position of the construction machinery under different geological conditions.

[0020] Step S113: performing time stamp synchronization processing on the environmental monitoring data to generate a synchronized environmental monitoring data sequence, and performing time axis alignment processing on the synchronized environmental monitoring data sequence and the mechanical-geological spatial correlation data set to generate a spatiotemporal correlation monitoring data set.

[0021] In this embodiment, since the environmental monitoring data is collected at different time points, in order to make the environmental monitoring data have temporal consistency, the environmental monitoring data is time-stamped and synchronized to generate a synchronized environmental monitoring data sequence. Then, the synchronized environmental monitoring data sequence is aligned with the mechanical-geological spatial correlation data set, for example, the convergence deformation data in the hole at a certain moment is associated with the position information of the drilling rig in the geological model at the same moment, and a time-space correlation monitoring data set is generated, which is convenient for comprehensive analysis of the time-space changes during the construction process.

[0022] Step S114: performing abnormal data cleaning on the spatiotemporal correlation monitoring data set, and performing missing value repair processing on the cleaned data using preset data integrity verification rules to generate the standardized monitoring data set.

[0023] In this embodiment, the isolation forest algorithm (contamination=0.05) can be used to detect outliers for the spatiotemporal correlation monitoring data set. For example, some data that obviously deviate from the normal range are considered outliers and are removed. For missing values ​​in the cleaned data, the KNN interpolation (k=5, distance metric=Euclidean distance) method is used to repair them. Finally, the cleaned and repaired data is converted into JSON format, and the fields include {"timestamp":"2023-08-20T14:30:00Z","location":[x,y,z],"rock_RQD":75,"drill_speed":2.4m / min}, and the ApacheKafka cluster (3 nodes, throughput ≥100,000 / second) is deployed to realize real-time data stream processing and generate standardized monitoring data sets.

[0024] Step S120: Perform multi-source feature extraction on the standardized monitoring data set to obtain a construction status feature set and a geological risk feature set, and input the construction status feature set and the geological risk feature set into a reinforcement learning model for multi-objective parameter optimization processing to generate a preliminary construction method parameter set.

[0025] In this embodiment, multi-source feature extraction can be performed on the standardized monitoring data set generated above. For example, geological risk features such as rock mass strength and geological structure are extracted from geological exploration data, and construction status features such as rock drilling speed and convergence deformation rate in the cave are extracted from construction machinery operation data and environmental monitoring data, thereby obtaining a construction status feature set and a geological risk feature set, which are then input into the reinforcement learning model for multi-objective parameter optimization processing.

[0026] For example, step S120 may include: Step S121: construct a state space set of the reinforcement learning model, wherein the state space set includes rock strength characteristics, cavern span characteristics and groundwater flow characteristics, the rock strength characteristics are extracted from the geological risk feature set, and the cavern span characteristics and groundwater flow characteristics are extracted from the construction status feature set.

[0027] In this embodiment, the state space set of the reinforcement learning model includes multiple features. Among them, the rock mass strength feature is extracted from the geological risk feature set. For example, the uniaxial compressive strength of the rock mass is obtained by comprehensive analysis of the drilling core data and the ground stress measurement data. The value range is between 20-250MPa. The uniaxial compressive strength of the rock mass in this project is 65MPa. The cavern span feature is obtained from the construction design information. The cavern span of this project is 18m. The groundwater flow feature is extracted from the environmental monitoring data. The groundwater flow of this project is 50L / min. Thus, the above features and other possible features can be combined into a multidimensional state space set.

[0028] Step S122: define an action space set of the reinforcement learning model, wherein the action space set includes blasting hole spacing parameters, charge density parameters and anchor spacing parameters, the blasting hole spacing parameters and charge density parameters control the dynamic excavation intensity, and the anchor spacing parameters and concrete spraying layer thickness parameters control the support structure configuration.

[0029] In this embodiment, the action space set of the reinforcement learning model can contain multiple parameters. For example, the blasting hole spacing parameter ranges from 0.8 to 1.2 m, and the charge density parameter ranges from 0.3 to 0.6 kg / m. These two parameters jointly control the dynamic excavation intensity. The anchor spacing parameter ranges from 1.0 to 1.5 m, and the concrete spraying layer thickness parameter is determined according to actual conditions. These two parameters control the support structure configuration, and the output action space is a multidimensional parameter vector, which is used to determine the specific parameters of the construction method.

[0030] Step S123: Construct a multi-objective reward function of the reinforcement learning model, wherein the multi-objective reward function consists of a construction efficiency reward item, a safety risk penalty item and a cost control reward item, wherein the construction efficiency reward item is calculated based on the single-cycle footage speed and the machinery utilization rate, the safety risk penalty item is calculated based on the surrounding rock deformation rate and the support structure stress peak value, and the cost control reward item is calculated based on the material consumption rate and the energy utilization efficiency.

[0031] In this embodiment, the multi-objective reward function of the reinforcement learning model is a function that comprehensively considers construction efficiency, safety risks and cost control. The construction efficiency reward item is calculated based on the single-cycle footage speed and the machine utilization rate. For example, the faster the single-cycle footage speed and the higher the machine utilization rate, the greater the value of the construction efficiency reward item. The safety risk penalty item is calculated based on the surrounding rock deformation rate and the peak stress of the support structure. If the surrounding rock deformation rate exceeds a certain threshold or the peak stress of the support structure is too large, the value of the safety risk penalty item will increase. The cost control reward item is calculated based on the material consumption rate and energy utilization efficiency. The lower the material consumption rate and the higher the energy utilization efficiency, the greater the value of the cost control reward item. By comprehensively considering the above three aspects, a reasonable multi-objective reward function is constructed to guide the reinforcement learning model to find the optimal construction method parameters.

[0032] Step S124: adopting the proximal strategy optimization algorithm to perform strategy network training processing on the reinforcement learning model, inputting the state space set and the action space set into the strategy network of the reinforcement learning model to generate a blasting parameter gradient set and a support parameter gradient set.

[0033] In this embodiment, the proximal policy optimization algorithm (PPO algorithm, clip_epsilon=0.2, learning rate α=0.0003, discount factor γ=0.99) can be used to train the policy network of the reinforcement learning model, and the state space set and action space set constructed previously are input into the policy network of the reinforcement learning model. Through continuous learning and adjustment, the blasting parameter gradient set and the support parameter gradient set are generated. The above blasting parameter gradient set and support parameter gradient set reflect how to adjust the blasting parameters and support parameters to obtain better rewards.

[0034] Step S125: The convergence of the generated parameter gradient is verified through the value function network of the reinforcement learning model. When the output error of the value function network is less than the preset threshold, the preliminary construction method parameter set including the optimal blasting hole spacing parameters, line charge density parameters and support parameters is output.

[0035] In this embodiment, the value function network of the reinforcement learning model is used to verify the convergence of the generated parameter gradient. By calculating the output error of the value function network, when the output error is less than the preset threshold, it means that the reinforcement learning model has converged to a better solution. At this time, the output includes the optimal blasting hole spacing parameter (such as 1.05m in this project), line charge density parameter (such as 0.45kg / m in this project) and support parameter (such as anchor bolt spacing of 1.2m) The preliminary construction method parameter set.

[0036] Step S126: When the parameter combination generated by the reinforcement learning model exceeds the execution capability of the construction machinery, the collaborative optimization processing of the reinforcement learning model and the preset emergency method library is triggered, and the conflicting parameters are dynamically corrected to generate an updated construction method parameter set.

[0037] In this embodiment, if the parameter combination generated by the reinforcement learning model exceeds the execution capability of the construction machinery, for example, the blasting hole spacing is too small, the drilling rig cannot operate according to the parameters. At this time, the collaborative optimization processing of the reinforcement learning model and the preset emergency method library is triggered. According to the preset scheme in the emergency method library, the conflicting parameters are dynamically corrected, such as appropriately increasing the blasting hole spacing, and an updated construction method parameter set is generated to ensure that the construction can proceed smoothly.

[0038] Step S130: Input the preliminary construction method parameter set into the digital twin simulation system for mechanical response simulation processing. When the result of the mechanical response simulation processing exceeds a preset safety threshold, the iterative optimization processing of the construction method parameters is triggered to generate an optimized construction method parameter set.

[0039] In this embodiment, the construction efficiency indicators included in the multi-objective parameter optimization process are the single-cycle footage speed and the machine utilization rate. The single-cycle footage speed refers to the forward distance of the cave excavation in one construction cycle, which directly reflects the progress of the construction. In this project, the single-cycle footage speed is improved by optimizing the blasting parameters and construction technology, thereby speeding up the construction progress of the entire project. The machine utilization rate refers to the ratio of the actual working time of the construction machinery to the total available time. Improving the machine utilization rate can give full play to the efficiency of the construction machinery and reduce the construction cost. For example, the operation sequence and time of the rock drilling rig and the slag truck can be reasonably arranged to reduce the idle time of the equipment and improve the machine utilization rate.

[0040] The safety risk indicators are the surrounding rock deformation rate and the peak stress of the support structure. The surrounding rock deformation rate reflects the deformation of the surrounding rock of the cavern during the construction process. When the surrounding rock deformation rate exceeds a certain threshold, it may mean that the cavern is at risk of instability. By real-time monitoring of the surrounding rock deformation rate, timely adjustment of construction methods and support measures can be made to ensure the safety of the cavern. The peak stress of the support structure refers to the maximum stress value generated by the support structure (such as anchor rods, concrete spraying, etc.) when it is subjected to the pressure of the surrounding rock. When the peak stress of the support structure exceeds its design strength, the support structure may be damaged, affecting the stability of the cavern. Therefore, in the process of multi-objective parameter optimization, it is necessary to control the surrounding rock deformation rate and the peak stress of the support structure to ensure construction safety.

[0041] The cost control indicators are material consumption rate and energy efficiency. Material consumption rate refers to the ratio of the actual amount of material consumed during the construction process to the theoretical amount of material calculated. Reducing the material consumption rate can reduce material waste and reduce construction costs. For example, optimizing blasting parameters can reduce the amount of explosives used, and rationally designing support structures can reduce the amount of anchors and concrete used. Energy efficiency refers to the effective use of energy during the construction process. Improving energy efficiency can reduce energy consumption costs. For example, using energy-saving construction equipment, optimizing equipment operating parameters, and improving energy efficiency.

[0042] The digital twin simulation system includes a surrounding rock constitutive model library, a support structure mechanical model library and a construction machinery behavior model library. The surrounding rock constitutive model library adopts the Moore-Coulomb criterion and the strain softening model. The Moore-Coulomb criterion is a commonly used yield criterion for geotechnical materials. It takes into account the internal friction angle and cohesion of the rock, and can better describe the yield and failure behavior of the rock under stress. The strain softening model further takes into account the characteristic that the strength of the rock decreases with the increase of strain after failure, and more accurately simulates the actual mechanical behavior of the rock. The support structure mechanical model library includes an anchor group support effect model and a concrete spray layer collaborative deformation model. The anchor group support effect model takes into account the interaction and collaborative work between the anchors, and can accurately calculate the reinforcement effect of the anchor group on the surrounding rock. The concrete spray layer collaborative deformation model simulates the interaction between the concrete spray layer and the surrounding rock, and takes into account the deformation and bearing capacity of the spray layer. The construction machinery behavior model library simulates the operation behavior of construction machinery under different working conditions, and provides accurate mechanical parameters for the simulation of the construction process.

[0043] In this project, the digital twin simulation system played an important role in the optimization and verification of the construction method parameters. By inputting the preliminary construction method parameter set into the digital twin simulation system, the mechanical response simulation is performed using the surrounding rock constitutive model library, the support structure mechanical model library, and the construction machinery behavior model library. For example, the stress redistribution of the surrounding rock, the expansion of the plastic zone, and the deformation and stress state of the support structure when subjected to the surrounding rock pressure during the blasting process are simulated. The safety and feasibility of the construction method are evaluated based on the simulation results. When the simulation results exceed the preset safety threshold, the iterative optimization of the construction method parameters is triggered to ensure that the construction method meets the project requirements.

[0044] In this embodiment, the preliminary construction method parameter set generated previously is input into the digital twin simulation system for mechanical response simulation processing to evaluate the safety and feasibility of the construction method.

[0045] Step S131: input the blasting hole spacing parameters and charge density parameters in the preliminary construction method parameter set into the blasting effect simulation module of the virtual cavern three-dimensional grid model, and generate a surrounding rock stress redistribution data set after excavation based on the initial geostress field data loaded in the virtual cavern three-dimensional grid model.

[0046] In this embodiment, the geological model (FBX format, number of faces ≥ 100,000) can be imported into Unity3D, and the FLAC3D solver (grid size 0.5m×0.5m×0.5m) is integrated to build a virtual cavern 3D grid model. Then, the blasting hole spacing parameter (such as 1.05m) and charge density parameter (such as 0.45kg / m) in the preliminary construction method parameter set are input into the blasting effect simulation module of the virtual cavern 3D grid model, and the initial geostress field data (such as horizontal geostress σ 1 =25MPa, vertical ground stress σ 3 =8MPa), and by simulating the blasting process, a data set of surrounding rock stress redistribution after excavation is generated, which reflects the change of surrounding rock stress after blasting.

[0047] Step S132: input the surrounding rock stress redistribution data set into the plastic zone expansion calculation module, determine the surrounding rock plastic deformation range according to the rock yield criterion, and generate the plastic zone expansion depth and distribution morphology data set.

[0048] In this embodiment, the previously generated surrounding rock stress redistribution data set can be input into the plastic zone expansion calculation module, and the Moore-Coulomb criterion can be used to determine whether the surrounding rock has entered a plastic state. The plastic deformation range of the surrounding rock is determined based on the judgment result, and a plastic zone expansion depth and distribution morphology data set is generated. This set can intuitively display the changes in the plastic zone of the surrounding rock after blasting.

[0049] Step S133: Input the anchor spacing parameters and concrete spraying layer thickness parameters in the preliminary construction method parameter set into the support structure bearing capacity simulation module, calculate the deformation of the support structure in combination with the plastic zone expansion depth data, and generate the anchor axial force distribution data set and the spraying layer stress peak data set.

[0050] In this embodiment, the anchor spacing parameter (such as 1.2m) and the concrete spraying layer thickness parameter in the preliminary construction method parameter set can be input into the support structure bearing capacity simulation module, combined with the previously generated plastic zone expansion depth data, the deformation of the support structure when subjected to surrounding rock pressure is calculated, and the anchor axial force distribution data set and the spraying layer stress peak data set are generated. The above anchor axial force distribution data set and the spraying layer stress peak data set can evaluate the bearing capacity of the support structure.

[0051] Step S134: Input the surrounding rock stress redistribution data set, the plastic zone extension depth data set, the anchor axial force distribution data set and the spray layer stress peak data set into the comprehensive safety assessment module, perform multi-index fusion analysis through the preset deformation threshold judgment rules and support failure criteria, and generate a surrounding rock stability grade assessment result set.

[0052] In this embodiment, the surrounding rock stress redistribution data set, the plastic zone expansion depth data set, the anchor axial force distribution data set and the spray layer stress peak data set can be input into the comprehensive safety assessment module, and a multi-indicator fusion analysis is performed according to the preset deformation threshold judgment rules (such as the top arch displacement prediction value > 50mm or the side wall convergence > 30mm is judged as unstable) and support failure criteria (such as the anchor axial force exceeds the design value) to generate a surrounding rock stability level assessment result set, which can judge the stability level of the surrounding rock under the construction method.

[0053] Step S135: When the top arch displacement assessment value or the side wall convergence assessment value in the surrounding rock stability grade assessment result set exceeds the preset safety threshold, the parameter iterative optimization process is triggered, and the optimization weights of the blasting parameters and the support parameters are adjusted according to the distribution characteristics of the stress exceeding area, and an intermediate optimization parameter set including the hole spacing correction parameter, the charge density correction parameter and the support strength correction parameter is generated.

[0054] In this embodiment, if the top arch displacement assessment value or the side wall convergence assessment value in the surrounding rock stability level assessment result set exceeds the preset safety threshold, for example, the top arch displacement prediction value reaches 52mm (exceeding the threshold of 50mm), the parameter iteration optimization process is triggered. According to the distribution characteristics of the stress exceeding standard area, such as stress concentration at the top of the cavern, the optimization weight of the support parameter is appropriately increased, and the optimization weight of the blasting parameter is reduced, and an intermediate optimization parameter set including hole spacing correction parameters (such as increasing the blasting hole spacing to 1.1m), charge density correction parameters (such as reducing the charge density to 0.4kg / m) and support strength correction parameters (such as reducing the anchor spacing to 1.1m) is generated.

[0055] Step S136: re-input the intermediate optimization parameter set into the blasting effect simulation module and the support structure bearing capacity simulation module for secondary simulation verification. If all indicators in the surrounding rock stability grade assessment result set generated by the secondary simulation are lower than the preset safety threshold, the intermediate optimization parameter set is marked as the optimized construction method parameter set.

[0056] In this embodiment, the intermediate optimization parameter set is re-input into the blasting effect simulation module and the support structure bearing capacity simulation module for secondary simulation verification. If all indicators in the surrounding rock stability grade assessment result set generated by the secondary simulation are lower than the preset safety threshold, such as the predicted value of the top arch displacement is 48 mm (lower than the threshold of 50 mm) and the side wall convergence is 28 mm (lower than the threshold of 30 mm), then the intermediate optimization parameter set is marked as the optimized construction method parameter set, indicating that the parameter set meets the safety requirements.

[0057] Step S137: If there are still indicators exceeding the safety threshold after the second simulation, the intermediate optimization parameter set is iteratively corrected for multiple rounds using a gradient descent algorithm until the safety threshold requirement is met and the final optimization result is output.

[0058] In this embodiment, if there are still indicators exceeding the safety threshold after the second simulation, for example, the side wall converges to 32mm (exceeding the threshold of 30mm), the gradient descent algorithm is used to perform multiple rounds of iterative corrections on the intermediate optimization parameter set. The blasting parameters and support parameters are continuously adjusted, such as further increasing the blasting hole spacing, reducing the charge density, and increasing the support strength, until all indicators in the surrounding rock stability level assessment result set are lower than the preset safety threshold, and the final optimization result is output, that is, the optimized construction method parameter set.

[0059] Step S140: Compliance matching is performed on the optimized construction method parameter set and the preset construction specification knowledge base to generate a construction method compliance verification result. When there are violations in the construction method compliance verification result, manual correction is triggered to generate a corrected construction method parameter set.

[0060] In this embodiment, the optimized construction method parameter set is matched with the preset construction specification knowledge base for compliance, so as to ensure that the construction method complies with the relevant specification requirements.

[0061] Step S141: parsing the clause constraints in the preset construction specification knowledge base to generate a specification parameter threshold set and a construction process requirement set.

[0062] In this embodiment, various relevant specifications can be structured and analyzed to extract key clauses, such as "Class IV surrounding rock anchor spacing ≤ 1.2m", etc., and the above-mentioned clause constraints can be analyzed into a specification parameter threshold set and a construction process requirement set.

[0063] Step S142: extract key parameters from the optimized construction method parameter set to generate a blasting control parameter subset, a support strength parameter subset, and an excavation sequence parameter subset.

[0064] In this embodiment, for the optimized construction method parameter set, it is necessary to extract the parameters that are crucial to the construction compliance verification. From a technical point of view, different types of construction operations correspond to different key parameters. Classifying and extracting the above parameters will help to conduct targeted matching verification with the specifications later. In specific implementation, the data structure of the optimized construction method parameter set is analyzed and classified according to the meaning and function of the parameters. For example, for parameters related to blasting control, such as blasting hole spacing, charge density, etc., they are combined into a blasting control parameter subset; for parameters related to support strength, such as anchor spacing, concrete spray layer thickness, etc., they are extracted to form a support strength parameter subset; and for the sequence of the construction process, such as the excavation sequence of different tunnel sections, the excavation sequence parameter subset is formed. Taking this project as an example, the optimized construction method parameter set includes specific parameters such as blasting hole spacing of 1.1m, charge density of 0.4kg / m, anchor spacing of 1.1m, and concrete spray layer thickness. After extraction, the blasting control parameter subset includes blasting hole spacing of 1.1m and charge density of 0.4kg / m, and the support strength parameter subset includes anchor spacing of 1.1m and corresponding concrete spray layer thickness.

[0065] Step S143: Compare the blasting control parameter subset with the blasting safety standards in the standard parameter threshold set item by item to generate a blasting compliance check result set.

[0066] In this embodiment, the blasting control parameter subset is compared with the blasting safety standard in the standard parameter threshold set to ensure that the blasting construction complies with the safety specification. From a technical principle, the blasting safety standard in the specification is formulated based on a large number of engineering practices and safety studies. By comparison, it can be found whether the actual construction parameters are within the safety range. In specific implementation, each parameter in the blasting control parameter subset is compared one by one with the corresponding standard in the specification. For example, the specification stipulates that the blasting hole spacing should be between 0.8-1.2m and the charge density should be between 0.3-0.6kg / m. For the 1.1m blasting hole spacing and 0.4kg / m charge density in the blasting control parameter subset in this project, they are compared with the standard respectively. If the parameter is within the standard range, it is marked as compliant; if it exceeds the range, it is marked as illegal. Finally, the comparison results of each parameter are combined into a blasting compliance check result set, which clearly reflects whether the blasting construction parameters meet the standard requirements.

[0067] Step S144: Perform strength verification processing on the support strength parameter subset and the support design requirements in the specification parameter threshold set to generate a support compliance check result set.

[0068] In this embodiment, the strength verification of the support strength parameter subset is to ensure that the support structure can effectively support the surrounding rock and ensure construction safety. From a technical point of view, the support design requirements are formulated according to the geological conditions of the surrounding rock and the principles of engineering mechanics. By verifying the support strength parameter subset, it is possible to evaluate whether the actual bearing capacity of the support structure meets the requirements. In specific implementation, according to the support design requirements in the specification, it may involve the calculation and comparison of the tensile strength of the anchor rod, the compressive strength of the concrete spray layer, etc. Taking the support strength parameter subset of this project as an example, it includes an anchor rod spacing of 1.1m and a certain thickness of concrete spray layer. According to the specification, for Class IV surrounding rock, the anchoring force of the anchor rod needs to meet certain numerical requirements, and the thickness and strength of the concrete spray layer also have corresponding standards. By performing mechanical calculation and analysis on the above parameters, it is determined whether the specification requirements are met. If it is met, it is marked as compliant; if it is not met, it is marked as illegal, and finally a support compliance inspection result set is formed.

[0069] Step S145: topologically sort and verify the excavation sequence parameter subset and the process logic rules in the construction process requirement set to generate a process compliance check result set.

[0070] In this embodiment, the excavation sequence parameter subset is topologically sorted and verified to ensure the rationality and logic of the construction process. From a technical point of view, the process logic rules in the construction process requirement set are formulated based on multiple factors such as construction safety, efficiency and quality. Through topological sorting verification, it can be found whether there are conflicts or unreasonable aspects in the excavation sequence. In specific implementation, the excavation sequence parameter subset is regarded as a directed graph, in which each excavation step is a node in the graph, and the order relationship between the steps is a directed edge. Then the directed graph is topologically sorted to check whether a reasonable sorting result can be obtained. For example, in this project, the excavation sequence may stipulate that the top of the cavern is excavated first, and then the side wall is excavated. If the order in the excavation sequence parameter subset does not meet the rule, a conflict will be found in the topological sorting process. The result of the topological sorting is judged. If it meets the process logic rules, it is marked as compliant; otherwise, it is marked as illegal, and finally a process compliance check result set is generated.

[0071] Step S146: Perform comprehensive judgment processing on the blasting compliance inspection result set, the support compliance inspection result set and the process compliance inspection result set to generate the construction method compliance verification result.

[0072] In this embodiment, the three compliance check result sets are comprehensively judged in order to comprehensively evaluate the compliance of the construction method. From a technical point of view, the compliance of the construction method does not only depend on one aspect, but needs to comprehensively consider multiple links such as blasting, support and process. In specific implementation, all marked results in the blasting compliance check result set, the support compliance check result set and the process compliance check result set are summarized and analyzed. If all parameters in the three sets are marked as compliant, the construction method is judged to be compliant; as long as one parameter is marked as illegal, the construction method is judged to have illegal items. For example, in this project, if all parameters in the blasting compliance check result set are compliant, the anchor spacing in the support compliance check result set meets the requirements, but the concrete spraying strength does not meet the specifications, and there is a conflict in the excavation order in the process compliance check result set, then it is comprehensively judged that there are illegal items in the construction method, and the corresponding construction method compliance verification results are generated.

[0073] Step S147: When there are violations in the construction method compliance verification result, a manual correction process is triggered to generate a corrected construction method parameter set.

[0074] In this embodiment, when there are violations in the construction method compliance verification results, manual correction is required. From a technical point of view, manual correction can combine the experience and professional knowledge of engineers to adjust the parameters that do not meet the specifications to ensure that the construction method meets the requirements. In specific implementation, a three-dimensional visualization tool is developed based on WebGL, allowing engineers to interact with the tool. For example, for the problem that the strength of the concrete spray layer in the support strength parameter subset does not meet the specifications, engineers can use the tool to drag and adjust the thickness of the concrete spray layer or modify parameters such as the mix ratio of the concrete; for the problem that there is a conflict in the excavation sequence in the process compliance check result set, engineers can re-plan the excavation sequence. During the adjustment process, engineers can view the adjusted effect in real time. At the same time, the manually corrected data is added to the reinforcement learning experience pool, and the online learning rate η=0.01 is set. The model is incrementally updated every 24 hours so that the model can learn the experience of manual correction and improve the accuracy of the subsequent generation of construction method parameters. After manual correction, a corrected construction method parameter set is generated.

[0075] Step S150: Perform full-cycle risk prediction processing on the revised construction method parameter set to generate a construction risk distribution map, and dynamically control the revised construction method parameter set according to the construction risk distribution map to generate a final construction method instruction set.

[0076] In this embodiment, the full-cycle risk prediction of the revised construction method parameter set is to identify the risks that may occur during the construction process in advance so that corresponding measures can be taken to prevent them. By generating a construction risk distribution map, the risk situation of different construction stages and different areas can be intuitively understood, so as to dynamically adjust the revised construction method parameter set to ensure safe and efficient construction.

[0077] Step S151: Establish construction stage division rules to divide the construction period into preparation stage, excavation stage, support stage and finishing stage, and perform risk factor identification processing on the preparation stage, excavation stage, support stage and finishing stage to generate a stage risk feature set.

[0078] In this embodiment, the construction phase division rules are established based on the different characteristics and tasks of the construction process. From a technical point of view, different construction phases face different types of risks, and dividing the construction cycle helps to identify risks more accurately. In specific implementation, according to the construction process and time sequence, the construction cycle is clearly divided into the preparation phase, excavation phase, support phase and finishing phase. In the preparation phase, risk factors may include improper equipment commissioning, insufficient material supply, etc.; risk factors in the excavation phase may include unstable surrounding rock and water inrush caused by blasting; risk factors in the support phase may be loose installation of the support structure; risk factors in the finishing phase may be inadequate cleaning work, etc. By analyzing the construction activities and environmental factors of each stage, the corresponding risk factors are identified, and the above risk factors are combined into a stage risk feature set. For example, in the preparation phase of this project, it is identified that abnormal vibration sensor data during equipment commissioning may cause the drill bit vibration frequency to be unable to be accurately obtained, which is a risk factor; in the excavation phase, according to geological conditions and blasting parameters, the risk factor of water inrush in the fault fracture zone is identified.

[0079] Step S152: performing Monte Carlo simulation processing on the stage risk feature set to generate a risk occurrence probability distribution data set and a risk impact assessment result set for each construction stage.

[0080] The specific sub-steps are described in detail below: Step S1521: extract the probability distribution type parameters and distribution parameter set of each risk factor from the stage risk feature set to generate risk factor random sampling configuration data In this step, it is necessary to conduct an in-depth analysis of each risk factor in the stage risk feature set. From a technical point of view, different risk factors often follow different probability distribution types, which are determined based on their inherent physical properties, historical data statistics, and related engineering experience.

[0081] For risk factors of geological parameters, such as the uniaxial compressive strength of rock mass, according to a large amount of geological exploration data and previous experience of similar projects, they usually obey normal distribution. Through statistical analysis of the drilling core data of this project, its mean value μ=65MPa and standard deviation σ=8MPa are determined. The mean and standard deviation here are the distribution parameters of the risk factor obeying normal distribution.

[0082] For some construction parameter risk factors, such as the adjustment range of blasting hole spacing, their values ​​are random within a certain range and the probability of each value occurring is relatively uniform, so they may obey uniform distribution. Assuming that the range of the adjustment range of blasting hole spacing is between -0.1m and 0.1m, then this range is the distribution parameter of the risk factor obeying uniform distribution.

[0083] The probability distribution type (such as normal distribution, uniform distribution, etc.) and the corresponding distribution parameters (such as mean, standard deviation, value range, etc.) of each risk factor are sorted and summarized to generate random sampling configuration data of risk factors, which will serve as the basis for random sampling in subsequent Monte Carlo simulations.

[0084] Step S1522: Setting the sampling times parameter of the Monte Carlo simulator according to the risk factor random sampling configuration data, and performing multiple rounds of independent random sampling to generate a risk scenario simulation data set According to the random sampling configuration data of the risk factor, the sampling times of the Monte Carlo simulator are set. The determination of the sampling times needs to comprehensively consider the accuracy of the simulation results and the consumption of computing resources. Generally speaking, the more sampling times, the closer the simulation results are to the actual situation, but the computing time and resource requirements will also increase accordingly. In this embodiment, considering the complexity of the project and the requirements for the accuracy of the results, the sampling times are set to 100,000 times.

[0085] The Monte Carlo simulator performs independent random sampling based on the probability distribution type and distribution parameters of each risk factor. Each sampling will obtain a set of risk factor value combinations, which represent a possible risk scenario. For example, in a certain sampling, the uniaxial compressive strength of the rock mass was drawn to 68MPa, and the adjustment range of the blasting hole spacing was drawn to 0.05m.

[0086] Such independent random sampling is repeated 100,000 times, and the risk scenario data obtained from each sampling is recorded and organized to generate a risk scenario simulation data set, which contains a large number of different risk scenario instances and covers various possible risk factor combinations.

[0087] Step S1523: Perform risk trigger condition detection on each risk scenario instance in the risk scenario simulation data set, and count the risk occurrence frequency data set of each construction stage. After obtaining the risk scenario simulation data set, it is necessary to detect the risk trigger conditions for each risk scenario instance. The risk trigger conditions are pre-set according to the actual situation of the project and relevant safety standards. For example, for water inrush risk, when the simulated groundwater level rises to a certain threshold (such as exceeding 2m from the bottom of the cavern), the risk scenario is judged to trigger the water inrush risk; for rock burst risk, when the stress state of the rock mass meets the specific rock burst criterion (such as energy release rate ≥ 5×10³J / m³), the risk scenario is judged to trigger the rock burst risk.

[0088] For each construction stage, all risk scenario instances in the risk scenario simulation data set are checked one by one, and the frequency of each risk is counted. For example, in the 100,000 simulations of the excavation stage, the water inrush risk was triggered 2,000 times, the rock burst risk was triggered 1,500 times, etc. The frequency of occurrence of different risks in each construction stage is sorted out to form a data set of risk occurrence frequency in each construction stage.

[0089] Step S1524: Calculate the risk occurrence probability distribution data set for each construction stage based on the risk occurrence frequency data set Based on the risk frequency data set, the probability of occurrence of each risk in each construction stage is calculated. The probability of risk occurrence is calculated by dividing the frequency of occurrence of the risk by the total number of samplings. For example, in the excavation stage, the risk of water inrush occurred 2,000 times, and the total number of samplings was 100,000 times, so the probability of water inrush risk was 2,000 ÷ 100,000 = 2%.

[0090] This calculation is performed for all risks in each construction stage to obtain the probability of occurrence of each risk. The probability of occurrence of different risks in each construction stage is sorted and summarized to form a data set of risk occurrence probability distribution for each construction stage, which shows the probability of occurrence of different risks in each construction stage.

[0091] Step S1525: Perform quantitative calculation of the impact degree of each risk scenario instance in the risk scenario simulation data set to generate a risk impact value sequence for each construction stage. For each risk scenario instance in the risk scenario simulation data set, its risk impact needs to be quantified. The quantification of the risk impact requires comprehensive consideration of multiple factors, such as the impact of risk events on construction progress, cost, quality, and safety.

[0092] Taking the risk of water inrush as an example, when the simulated water inrush risk occurs, the number of days of delay in construction progress, additional costs (such as investment in drainage equipment, costs of repairing damaged facilities, etc.) and potential impact on construction quality and safety are quantitatively calculated based on factors such as the amount of water inrush, duration, and impact on construction equipment and personnel. The above-mentioned influencing factors are comprehensively evaluated to obtain a quantitative risk impact value.

[0093] Such a quantitative calculation of the impact degree is performed on each risk scenario instance in the risk scenario simulation data set, and the risk impact value corresponding to each risk scenario instance is recorded to form a risk impact value sequence for each construction stage.

[0094] Step S1526: Perform statistical distribution fitting processing on the risk impact value sequence to generate a risk impact assessment result set for each construction stage After obtaining the risk impact value sequence of each construction stage, it is necessary to perform statistical distribution fitting on the risk impact value sequence. The purpose of statistical distribution fitting is to find a suitable probability distribution to describe the distribution of risk impact values, so as to better evaluate the impact of risks.

[0095] Common probability distribution types include normal distribution, lognormal distribution, Weibull distribution, etc. By statistically analyzing the risk impact value sequence, a suitable fitting method (such as maximum likelihood estimation) is used to determine the probability distribution type and its distribution parameters that best suit the sequence. For example, after analysis, it is found that the water inrush risk impact value sequence of a certain construction stage conforms to the lognormal distribution, and its distribution parameters such as mean and standard deviation are obtained through fitting calculation.

[0096] Such statistical distribution fitting processing is performed on all risk impact value sequences of each construction stage, and the probability distribution type and distribution parameters corresponding to each risk are sorted and summarized to form a set of risk impact degree assessment results for each construction stage, which can be used to predict the probability of occurrence of different risk impact degrees.

[0097] Step S153: Matrix fusion processing is performed on the risk occurrence probability distribution data set and the risk impact assessment result set to generate a three-dimensional risk heat map data set.

[0098] In this embodiment, the risk occurrence probability distribution data set and the risk impact assessment result set are matrix-fused in order to comprehensively display the probability and impact of the risk. From a technical point of view, through matrix fusion, information of two different dimensions can be integrated into a three-dimensional space to form a risk heat map, which intuitively reflects the risk status of different regions and different construction stages. In specific implementation, the risk occurrence probability distribution data set and the risk impact assessment result set are regarded as two matrices, and the elements of each matrix represent the probability of risk occurrence and the impact degree of different regions or different construction stages. The two matrices are fused through a certain algorithm, for example, the risk occurrence probability is used as one dimension, the risk impact degree is used as another dimension, and the construction stage or region is used as the third dimension to generate a three-dimensional risk heat map data set. In this project, an intuitive three-dimensional graph can be obtained through matrix fusion, in which the depth of color represents the level of risk, and different coordinate positions represent different construction stages and regions.

[0099] Step S154: performing spatial interpolation processing on the three-dimensional risk heat map data set to generate a continuous risk distribution surface, and performing time axis mapping processing on the continuous risk distribution surface and the construction schedule data to generate the construction risk distribution map.

[0100] In this embodiment, the three-dimensional risk heat map data set is spatially interpolated in order to obtain a continuous risk distribution surface in order to more accurately describe the distribution of risks. From a technical point of view, since the three-dimensional risk heat map data set is discrete, spatial interpolation can be used to perform interpolation calculations between discrete data points to obtain a continuous risk distribution. In specific implementation, a suitable interpolation method, such as Kriging interpolation, is used to process the three-dimensional risk heat map data set to generate a continuous risk distribution surface. The continuous risk distribution surface is then mapped to the construction schedule data on the time axis to associate the risk distribution with the construction time. For example, in this project, the construction schedule specifies the start and end time of each construction stage. The risk information of different construction stages on the continuous risk distribution surface is mapped to the corresponding time points to generate a construction risk distribution map, which can clearly show the risk situation at different times and in different areas.

[0101] Step S155: performing risk area identification processing on the construction risk distribution map to generate a key monitoring area coordinate set and a risk level identification set.

[0102] In this embodiment, the risk area identification of the construction risk distribution map is to determine the areas that need to be monitored and the corresponding risk levels. From a technical point of view, the construction risk distribution map intuitively shows the distribution of risks. By analyzing the map, the areas with higher risks can be identified. In specific implementation, the risk values ​​in the construction risk distribution map are compared according to the preset risk threshold. For example, the risk value is divided into three levels: low, medium and high. The low risk threshold is set to a risk value less than 10%, the medium risk threshold is 10%-20%, and the risk threshold is greater than 20%. When the risk value of a certain area in the map exceeds the corresponding threshold, the area is identified as a key monitoring area. The coordinate information of the key monitoring area is recorded to generate a key monitoring area coordinate set. At the same time, the corresponding risk level is determined according to the risk value of the area, and a risk level identification set is generated. In this project, by analyzing the construction risk distribution map, it is identified that the risk value of the fault fracture zone area is higher, and its coordinates are recorded in the key monitoring area coordinate set and marked as risk level.

[0103] Step S156: adjusting the construction machinery operation path planning data according to the key monitoring area coordinate set to generate an avoidance path solution set and a deceleration control parameter set.

[0104] In this embodiment, the operation path planning data of the construction machinery is adjusted according to the key monitoring area coordinate set to avoid the construction machinery from entering the risk area and ensure construction safety. From a technical point of view, the operation path planning of the construction machinery is based on the construction site and the construction task. When the key monitoring area is identified, the original path planning needs to be adjusted. In specific implementation, the operation path of the construction machinery is replanned according to the key monitoring area coordinate set in combination with the topography of the construction site and the performance of the construction machinery. For example, for a rock drilling rig, if the key monitoring area is on its original planned path, a path that bypasses the area can be planned to generate an avoidance path solution set. At the same time, in order to further ensure safety, when approaching the key monitoring area, the construction machinery needs to be decelerated. According to the risk level of the key monitoring area and the type of construction machinery, the deceleration control parameters are determined and the deceleration control parameter set is generated. In this project, when the rock drilling rig approaches the key monitoring area of ​​the fault fracture zone, a path that bypasses the area is planned, and the running speed of the rock drilling rig is reduced to 50% of the original speed, which is recorded in the deceleration control parameter set.

[0105] Step S157: adjusting the real-time monitoring frequency parameters according to the risk level identification set to generate an enhanced monitoring scheme set and an early warning threshold adjustment parameter set.

[0106] In this embodiment, the real-time monitoring frequency parameters are adjusted according to the risk level identification set in order to strengthen the monitoring of risk areas and timely discover risk hazards. From a technical point of view, areas with different risk levels require different monitoring frequencies, and risk areas require more frequent monitoring. In specific implementation, the real-time monitoring frequency parameters are adjusted according to the risk level in the risk level identification set. For example, for risk areas, the monitoring frequency is increased to twice the original; for medium-risk areas, the monitoring frequency is increased to 1.5 times the original; for low-risk areas, the original monitoring frequency is maintained. At the same time, the warning threshold is adjusted according to the risk level, and for risk areas, the warning threshold is lowered to detect risks earlier. The adjusted monitoring frequency and warning threshold are recorded, and an enhanced monitoring scheme set and a warning threshold adjustment parameter set are generated respectively. In this project, for the risk area of ​​the fault fracture zone, the monitoring frequency of convergence deformation in the cave is increased from once per hour to once every half an hour, and the warning threshold of convergence deformation is reduced from 5mm to 3mm, which are recorded in the enhanced monitoring scheme set and the warning threshold adjustment parameter set.

[0107] Step S158: performing instruction encoding processing on the avoidance path solution set, deceleration control parameter set, enhanced monitoring solution set and warning threshold adjustment parameter set to generate the final construction method instruction set.

[0108] In this embodiment, the avoidance path scheme set, the deceleration control parameter set, the enhanced monitoring scheme set and the warning threshold adjustment parameter set are subjected to instruction encoding processing in order to convert the above information into instructions that can be recognized and executed by the construction equipment and the monitoring system. From a technical point of view, different construction equipment and monitoring systems have different instruction formats and communication protocols. The above information can be uniformly converted into appropriate instructions through instruction encoding. In specific implementation, according to the requirements of the construction equipment and the monitoring system, the corresponding encoding algorithm is used to encode the information in the above set. For example, for the avoidance path scheme set, the path information is encoded into a series of coordinate points and motion instructions; for the deceleration control parameter set, the deceleration ratio and speed value are encoded into a specific digital signal. The encoded instructions are combined together to generate the final construction method instruction set. In this project, through instruction encoding processing, the avoidance path, deceleration control parameters, convergence deformation monitoring frequency and warning threshold of the rock drilling trolley are converted into instructions that can be sent to the rock drilling trolley control system and monitoring equipment through the MQTT protocol to form the final construction method instruction set.

[0109] Through the above steps, this embodiment realizes the full-chain intelligent decision-making of underground cavern construction method generation through multi-source data fusion and intelligent algorithm collaborative drive. First, through time-space alignment processing, discrete geological, mechanical and environmental data are converted into standardized data sets with time-space consistency, and then the reinforcement learning model is used to optimize the multi-dimensional features for multi-objective parameters, breaking through the experience-dependent limitations of traditional method parameter determination, and realizing the dynamic adaptation of construction status and geological risks. Further, a virtual mechanical response model is constructed through the digital twin simulation system, and the safety threshold triggers the iterative optimization mechanism, which significantly improves the reliability and safety of the method parameters; on this basis, a preset construction specification knowledge base is introduced for compliance verification, which not only ensures the standardization of the method parameters, but also retains the channel for the integration of expert experience through the manual correction mechanism; finally, through full-cycle risk prediction and dynamic regulation, the construction risk distribution map is converted into an executable method instruction set, realizing the full-process intelligent decision-making closed loop from data collection to method generation. As a result, not only the efficiency and scientificity of construction method generation are improved, but also a construction method generation system with autonomous optimization capabilities is constructed through the deep coupling of multimodal data fusion and intelligent algorithms.

[0110] Figure 2 A schematic diagram of exemplary hardware and software components of a system 100 for generating an underground cavern construction method based on multi-source data fusion that can implement the concept of the present invention is shown in some embodiments of the present invention. For example, the processor 120 can be used in the system 100 for generating an underground cavern construction method based on multi-source data fusion and used to perform the functions of the present invention.

[0111] The underground cavern construction method generation system 100 based on multi-source data fusion can be a general server or a special-purpose server, both of which can be used to implement the underground cavern construction method generation method based on multi-source data fusion of the present invention. Although the present invention only shows one server, for convenience, the functions described in the present invention can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0112] For example, the underground cavern construction method generation system 100 based on multi-source data fusion may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and storage media 140 in different forms, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the underground cavern construction method generation system 100 based on multi-source data fusion may also include program instructions stored in ROM, RAM, or other types of non-temporary storage media, or any combination thereof. The method of the present invention can be implemented according to the above-mentioned program instructions. The underground cavern construction method generation system 100 based on multi-source data fusion also includes an input / output (I / O) interface 150 between the computer and other input / output devices.

[0113] For ease of explanation, only one processor is described in the underground cavern construction method generation system 100 based on multi-source data fusion. However, it should be noted that the underground cavern construction method generation system 100 based on multi-source data fusion in the present invention may also include multiple processors, so the steps performed by one processor described in the present invention may also be performed jointly or individually by multiple processors. For example, if the processor of the underground cavern construction method generation system 100 based on multi-source data fusion executes step A and step B, it should be understood that step A and step B may also be executed jointly by two different processors or individually in one processor. For example, the first processor executes step A, the second processor executes step B, or the first processor and the second processor execute steps A and B together.

[0114] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer executable instructions are preset. When a processor executes the computer executable instructions, the above-mentioned method for generating an underground cavern construction method based on multi-source data fusion is implemented.

[0115] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, various features are sometimes combined into one embodiment, drawing or description thereof.

Claims

1. A method for generating underground cavern construction methods based on multi-source data fusion, characterized in that: The method comprises: Collecting geological exploration data, construction machinery operation data and environmental monitoring data, performing spatiotemporal alignment processing on the geological exploration data, construction machinery operation data and environmental monitoring data, and generating a standardized monitoring data set; Performing multi-source feature extraction on the standardized monitoring data set to obtain a construction status feature set and a geological risk feature set, inputting the construction status feature set and the geological risk feature set into a reinforcement learning model to perform multi-objective parameter optimization processing to generate a preliminary construction method parameter set; Inputting the preliminary construction method parameter set into the digital twin simulation system for mechanical response simulation processing, and triggering the iterative optimization processing of the construction method parameters to generate an optimized construction method parameter set when the result of the mechanical response simulation processing exceeds a preset safety threshold; Perform compliance matching processing on the optimized construction method parameter set and the preset construction specification knowledge base to generate a construction method compliance verification result, and when there are violations in the construction method compliance verification result, trigger manual correction processing to generate a corrected construction method parameter set; The revised construction method parameter set is subjected to full-cycle risk prediction processing to generate a construction risk distribution map, and the revised construction method parameter set is subjected to dynamic control processing according to the construction risk distribution map to generate a final construction method instruction set.

2. The method for generating underground cavern construction methods based on multi-source data fusion according to claim 1, characterized in that: The step of performing spatiotemporal alignment processing on the geological exploration data, construction machinery operation data and environmental monitoring data to generate a standardized monitoring data set includes: Establishing a conversion mapping relationship between a geological exploration coordinate system and a construction machinery coordinate system, and performing continuous geological interface generation processing on discrete exploration points in the geological exploration data to obtain a three-dimensional geological model data set; Performing dynamic trajectory calibration processing on the position coordinate data in the construction machinery operation data to generate a construction machinery real-time positioning data set, and performing spatial superposition processing on the construction machinery real-time positioning data set and the three-dimensional geological model data set to generate a machinery-geology spatial association data set; Performing time stamp synchronization processing on the environmental monitoring data to generate a synchronized environmental monitoring data sequence, and performing time axis alignment processing on the synchronized environmental monitoring data sequence and the mechanical-geological spatial correlation data set to generate a spatiotemporal correlation monitoring data set; The spatiotemporal correlation monitoring data set is cleaned for abnormal data, and the cleaned data is repaired for missing values ​​using preset data integrity verification rules to generate the standardized monitoring data set.

3. The method for generating underground cavern construction methods based on multi-source data fusion according to claim 1, characterized in that: The step of inputting the construction status feature set and the geological risk feature set into a reinforcement learning model for multi-objective parameter optimization processing to generate a preliminary construction method parameter set includes: Constructing a state space set of the reinforcement learning model, wherein the state space set includes rock mass strength characteristics, cavern span characteristics, and groundwater flow characteristics, the rock mass strength characteristics are extracted from the geological risk characteristic set, and the cavern span characteristics and groundwater flow characteristics are extracted from the construction state characteristic set; Defining an action space set of the reinforcement learning model, wherein the action space set includes a blasting hole spacing parameter, a charge density parameter, and an anchor spacing parameter, the blasting hole spacing parameter and the charge density parameter control the dynamic excavation intensity, and the anchor spacing parameter and the concrete spraying layer thickness parameter control the support structure configuration; Constructing a multi-objective reward function of the reinforcement learning model, the multi-objective reward function consists of a construction efficiency reward item, a safety risk penalty item, and a cost control reward item, wherein the construction efficiency reward item is calculated based on the single-cycle footage speed and the machine utilization rate, the safety risk penalty item is calculated based on the surrounding rock deformation rate and the support structure stress peak value, and the cost control reward item is calculated based on the material consumption rate and the energy utilization efficiency; The reinforcement learning model is trained with a proximal strategy optimization algorithm, and the state space set and the action space set are input into the strategy network of the reinforcement learning model to generate a blasting parameter gradient set and a support parameter gradient set; The convergence of the generated parameter gradient is verified by the value function network of the reinforcement learning model. When the output error of the value function network is less than a preset threshold, the preliminary construction method parameter set including the optimal blasting hole spacing parameter, line charge density parameter and support parameter is output; When the parameter combination generated by the reinforcement learning model exceeds the execution capability of the construction machinery, the collaborative optimization processing of the reinforcement learning model and the preset emergency method library is triggered, and the conflicting parameters are dynamically corrected to generate an updated construction method parameter set.

4. The method for generating underground cavern construction methods based on multi-source data fusion according to claim 1, characterized in that: The preliminary construction method parameter set is input into the digital twin simulation system for mechanical response simulation processing, and when the result of the mechanical response simulation processing exceeds a preset safety threshold, the iterative optimization processing of the construction method parameters is triggered to generate an optimized construction method parameter set, including: Inputting the blasting hole spacing parameter and the charge density parameter in the preliminary construction method parameter set into the blasting effect simulation module of the virtual cavern three-dimensional grid model, and generating a surrounding rock stress redistribution data set after excavation based on the initial geostress field data loaded in the virtual cavern three-dimensional grid model; Inputting the surrounding rock stress redistribution data set into the plastic zone expansion calculation module, judging the surrounding rock plastic deformation range according to the rock mass yield criterion, and generating the plastic zone expansion depth and distribution form data set; Input the anchor spacing parameter and the concrete spraying layer thickness parameter in the preliminary construction method parameter set into the support structure bearing capacity simulation module, calculate the support structure deformation in combination with the plastic zone extension depth data, and generate the anchor axial force distribution data set and the spraying layer stress peak data set; The surrounding rock stress redistribution data set, the plastic zone extension depth data set, the anchor axial force distribution data set and the spray layer stress peak data set are input into the comprehensive safety assessment module, and a multi-index fusion analysis is performed through the preset deformation threshold judgment rule and support failure criterion to generate a surrounding rock stability grade assessment result set; When the top arch displacement assessment value or the side wall convergence assessment value in the surrounding rock stability grade assessment result set exceeds the preset safety threshold, the parameter iterative optimization process is triggered, and the optimization weights of the blasting parameters and the support parameters are adjusted according to the distribution characteristics of the stress exceeding area, and an intermediate optimization parameter set including the hole spacing correction parameter, the charge density correction parameter and the support strength correction parameter is generated; The intermediate optimization parameter set is re-input into the blasting effect simulation module and the support structure bearing capacity simulation module for secondary simulation verification. If all indicators in the surrounding rock stability grade assessment result set generated by the secondary simulation are lower than the preset safety threshold, the intermediate optimization parameter set is marked as the optimized construction method parameter set; If there are still indicators exceeding the safety threshold after the second simulation, the gradient descent algorithm is used to perform multiple rounds of iterative corrections on the intermediate optimization parameter set until the safety threshold requirement is met and the final optimization result is output.

5. The method for generating underground cavern construction methods based on multi-source data fusion according to claim 1, characterized in that: The step of performing compliance matching processing on the optimized construction method parameter set and the preset construction specification knowledge base to generate the construction method compliance verification result includes: Parsing the clause constraints in the preset construction specification knowledge base to generate a specification parameter threshold set and a construction process requirement set; Performing key parameter extraction processing on the optimized construction method parameter set to generate a blasting control parameter subset, a support strength parameter subset, and an excavation sequence parameter subset; Comparing the blasting control parameter subset with the blasting safety standards in the standard parameter threshold set item by item to generate a blasting compliance inspection result set; Perform strength verification processing on the support strength parameter subset and the support design requirements in the specification parameter threshold set to generate a support compliance check result set; Performing topological sorting verification processing on the excavation sequence parameter subset and the process logic rules in the construction process requirement set to generate a process compliance check result set; The blasting compliance inspection result set, the support compliance inspection result set and the process compliance inspection result set are comprehensively judged and processed to generate the construction method compliance verification result.

6. The method for generating underground cavern construction methods based on multi-source data fusion according to claim 1, characterized in that: The step of performing full-cycle risk prediction processing on the modified construction method parameter set to generate a construction risk distribution map includes: Establishing construction phase division rules to divide the construction period into a preparation phase, an excavation phase, a support phase and a finishing phase, and performing risk factor identification processing on the preparation phase, the excavation phase, the support phase and the finishing phase to generate a phase risk feature set; Performing Monte Carlo simulation on the risk feature set of the stage to generate a risk occurrence probability distribution data set and a risk impact assessment result set for each construction stage; Performing matrix fusion processing on the risk occurrence probability distribution data set and the risk impact assessment result set to generate a three-dimensional risk heat map data set; The three-dimensional risk heat map data set is spatially interpolated to generate a continuous risk distribution surface, and the continuous risk distribution surface is time-axis mapped with the construction schedule data to generate the construction risk distribution map.

7. The method for generating underground cavern construction methods based on multi-source data fusion according to claim 6, characterized in that: The Monte Carlo simulation process is performed on the stage risk feature set to generate a risk occurrence probability distribution data set and a risk impact assessment result set for each construction stage, including: Extracting probability distribution type parameters and distribution parameter sets of each risk factor from the risk feature set of the stage, and generating random sampling configuration data of the risk factor; Setting the sampling number parameter of the Monte Carlo simulator according to the risk factor random sampling configuration data, and performing multiple rounds of independent random sampling to generate a risk scenario simulation data set; Performing risk trigger condition detection processing on each risk scenario instance in the risk scenario simulation data set, and counting the risk occurrence frequency data set of each construction stage; Calculate the risk occurrence probability distribution data set for each construction stage according to the risk occurrence frequency data set; Performing quantitative calculation processing on the impact degree of each risk scenario instance in the risk scenario simulation data set to generate a risk impact value sequence for each construction stage; The risk impact value sequence is subjected to statistical distribution fitting processing to generate a set of risk impact assessment results for each construction stage.

8. The method for generating underground cavern construction methods based on multi-source data fusion according to claim 1, characterized in that: The dynamically regulating and processing the modified construction method parameter set according to the construction risk distribution map to generate a final construction method instruction set includes: Performing risk area identification processing on the construction risk distribution map to generate a key monitoring area coordinate set and a risk level identification set; Adjust the construction machinery operation path planning data according to the key monitoring area coordinate set to generate an avoidance path solution set and a deceleration control parameter set; Adjusting the real-time monitoring frequency parameters according to the risk level identification set to generate an enhanced monitoring scheme set and an early warning threshold adjustment parameter set; The avoidance path solution set, deceleration control parameter set, enhanced monitoring solution set and warning threshold adjustment parameter set are processed by instruction encoding to generate the final construction method instruction set.

9. The method for generating underground cavern construction methods based on multi-source data fusion according to claim 1, characterized in that: The geological exploration data include drilling core data, geological radar scanning data and ground stress measurement data; the construction machinery operation data include drilling trolley positioning data, drill bit vibration frequency data and slag truck load data; the environmental monitoring data include tunnel convergence deformation data, groundwater level change data and air composition monitoring data; The multi-objective parameter optimization process includes construction efficiency indicators such as single cycle footage speed and machine utilization rate, safety risk indicators such as surrounding rock deformation rate and support structure stress peak, and cost control indicators such as material consumption rate and energy efficiency. The digital twin simulation system includes a surrounding rock constitutive model library, a support structure mechanical model library and a construction machinery behavior model library. The surrounding rock constitutive model library adopts the Moore-Coulomb criterion and the strain softening model, and the support structure mechanical model library includes an anchor group support effect model and a concrete spray layer collaborative deformation model.

10. A system for generating underground cavern construction methods based on multi-source data fusion, characterized in that: It includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the underground cavern construction method generation method based on multi-source data fusion as described in any one of claims 1 to 9.

Images