Tunnel surrounding rock damage risk assessment method and system based on multi-scale feature fusion

Abstract

The present invention relates to the field of data processing technology, and specifically to a method and system for assessing the damage risk of tunnel surrounding rocks based on multi-scale feature fusion. The method comprises the following steps: obtaining data on cross-sectional geometric parameters and geomechanical parameters of tunnel surrounding rocks; optimizing the data; constructing a damage state prediction model; establishing a risk assessment system; obtaining optimized construction parameters; establishing a model system for predicting damage states, assessing damage risks, and optimizing construction parameters; and implementing an assessment of the damage risk of tunnel surrounding rocks through the model system. The present invention achieves accurate damage risk assessment and parameter optimization by integrating geomechanical parameters, historical and real-time monitoring data, and combining the RHT constitutive model with deep reinforcement learning, significantly improving the reliability of damage risk assessment, and providing intelligent decision support for tunnel excavation construction.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a tunnel surrounding rock damage risk assessment method and system based on multi-scale feature fusion. Background Art

[0002] Tunnel excavation projects occupy a core position in mining and underground engineering construction. Construction efficiency and safety are directly related to the overall benefits of the project. In this process, surrounding rock damage risk assessment is crucial, involving construction parameter selection, blasting effect control and construction safety assurance. However, scientific and accurate surrounding rock damage risk assessment is still a technical problem that needs to be overcome in the field of tunnel excavation projects.

[0003] Traditional blasting design mostly relies on the experience of engineers. It may work under simple geological conditions, but it shows limitations under complex geological conditions. Although existing technologies try to use data analysis and machine learning algorithms to optimize parameters and assess risks, they mostly focus on the analysis of a single data source and ignore the inherent connection between geomechanical parameters, blasting design data and real-time monitoring data. In addition, these methods have obvious deficiencies in predicting the expansion of rock cracks, assessing the dynamics of risks, and data fusion, resulting in limited accuracy of assessment results and difficulty in effectively supporting construction decisions.

[0004] In response to the above problems, the present invention proposes a tunnel surrounding rock damage risk assessment method based on multi-scale feature fusion. This method constructs a comprehensive data set by integrating geomechanical parameters, blasting design data and real-time monitoring data. On this basis, the RHT constitutive model is used to construct a tunnel surrounding rock damage state prediction model, and multi-scale feature extraction and fusion are achieved through physically driven feature engineering. A deep reinforcement learning algorithm is further introduced to optimize risk assessment and construction parameters, and dynamic corrections are made based on real-time monitoring data. This method not only improves the accuracy and reliability of surrounding rock risk assessment, but also provides intelligent decision-making support for tunnel excavation construction. Compared with traditional methods, the present invention has more advantages under complex geological conditions, and provides strong support for construction parameter optimization, blasting efficiency improvement and construction safety assurance. Summary of the invention

[0005] In view of the defects in the prior art, the present invention provides a tunnel surrounding rock damage risk assessment method and system based on multi-scale feature fusion.

[0006] In the first aspect, the present invention provides a method for assessing the risk of tunnel surrounding rock damage based on multi-scale feature fusion, which includes the following steps: obtaining data on cross-sectional geometric parameters and geomechanical parameters of the tunnel surrounding rock; optimizing the data to obtain optimization results; constructing a damage state prediction model based on the optimization results; establishing a risk assessment system based on the damage state prediction model; obtaining optimized construction parameters based on the risk assessment system; establishing a model system for predicting damage states, assessing damage risks and optimizing construction parameters based on the optimized construction parameters; and implementing an assessment of the damage risk of the tunnel surrounding rock through the model system. The present invention improves the accuracy and reliability of data by comprehensively acquiring and optimizing the data of cross-sectional geometric parameters and geomechanical parameters of tunnel surrounding rocks, thereby constructing a more accurate damage state prediction model; through the damage state prediction model, the damage state of tunnel surrounding rocks can be accurately predicted, and then a scientific and reasonable risk assessment system is established, and a quantitative assessment of the damage risk of tunnel surrounding rocks is achieved; based on the risk assessment results, optimized construction parameters are obtained, which effectively reduces the risks in the construction process; based on the optimized construction parameters, a model system integrating prediction, assessment and optimization is established, which realizes dynamic monitoring and precise control of the damage risk of tunnel surrounding rocks.

[0007] Optionally, the acquisition of data on the cross-sectional geometric parameters and geomechanical parameters of the tunnel surrounding rock includes: dividing the cross-section of the tunnel surrounding rock into multiple regions; collecting data on the cross-sectional geometric parameters according to the multiple regions; collecting data on geomechanical parameters based on the data on the cross-sectional geometric parameters, wherein the data on geomechanical parameters include data on elastic modulus, Poisson's ratio, compressive strength and tensile strength of the rock mass. The present invention achieves accurate collection of cross-sectional geometric parameter data by meticulously dividing the tunnel surrounding rock cross-section into multiple regions; based on the geometric parameter data, further collecting data on geomechanical parameters, including key indicator data such as elastic modulus, Poisson's ratio, compressive strength and tensile strength of the rock mass, provides comprehensive data support for damage risk assessment; by systematically collecting and organizing these data, a rich database is established, providing a solid foundation for data optimization processing, model construction and risk assessment.

[0008] Optionally, the optimization processing of the data to obtain the optimization processing result includes: integrating the data of the cross-sectional geometric parameters and geomechanical parameters of the tunnel surrounding rock; performing data cleaning and standardization processing on the data; performing formatting processing on the data; performing feature engineering processing on the data, and the feature engineering processing includes high-dimensional feature enhancement and intelligent dimensionality reduction processing; through the integration processing, the data cleaning and standardization processing, the formatting processing and the feature engineering processing, the optimization processing result is obtained. The present invention realizes the comprehensive integration of data by integrating the data of the cross-sectional geometric parameters and geomechanical parameters of the tunnel surrounding rock, and provides a complete data set for damage analysis; through data cleaning and standardization processing, outliers and noise in the data are effectively eliminated to ensure the accuracy and consistency of the data; through formatting processing, the data is converted into a feature matrix format to improve the readability and ease of use of the data; through feature engineering processing, including high-dimensional feature enhancement and intelligent dimensionality reduction processing, the high-dimensional feature information of the data is strengthened and the dimension of the data is reduced, which lays a solid foundation for the construction of the damage state prediction model and the establishment of the risk assessment system.

[0009] Optionally, the formatting process includes: constructing a characteristic matrix of the cross-section area, the characteristic matrix including a geometric characteristic matrix, a physical characteristic matrix and a dynamic characteristic matrix; and constructing a correlation matrix of the cross-section area and the cracks according to the characteristic matrix. The present invention comprehensively and systematically integrates a variety of key characteristic information of the tunnel surrounding rock cross-section by constructing a characteristic matrix including a geometric characteristic matrix, a physical characteristic matrix and a dynamic characteristic matrix, providing more detailed data support for data analysis and model construction; further constructing a correlation matrix of the cross-section area and the cracks through the characteristic matrix, revealing the intrinsic connection between the cross-section area and the cracks, and providing a new perspective for understanding the damage mechanism of the tunnel surrounding rock.

[0010] Optionally, the construction of the damage state prediction model based on the optimization processing result includes: based on the optimization processing result, using the damage constitutive model, constructing a damage state time evolution model; based on the damage state time evolution model, using real-time monitoring data to perform dynamic damage correction to obtain correction results; using the correction results, updating the damage state time evolution model to obtain a damage state prediction model. The present invention constructs a damage state time evolution model by combining the damage constitutive model with the optimization processing result, and realizes an accurate description of the evolution of the damage state of the tunnel surrounding rock over time; based on the damage state time evolution model, using real-time monitoring data to perform dynamic damage correction, effectively integrating real-time monitoring information and data, and improving the real-time and accuracy of the model; updating the damage state time evolution model through the correction results, obtains a more accurate damage state prediction model, which can more accurately predict the future damage state of the tunnel surrounding rock.

[0011] Optionally, the risk assessment system is established based on the damage state prediction model, including: integrating multi-scale features based on the damage state prediction model, wherein the multi-scale features include local features, global features and time series features; outputting risk scores based on the damage state prediction model; establishing a risk distribution prediction model based on the damage state prediction model; and establishing a risk assessment system through the multi-scale features, the risk scores and the risk distribution prediction model. The present invention comprehensively and deeply understands the damage state of the tunnel surrounding rock by integrating multi-scale features such as local features, global features and time series features, and provides more abundant and accurate information for risk assessment; outputting risk scores based on the damage state prediction model realizes the quantitative assessment of the damage risk of the tunnel surrounding rock, and provides a scientific basis for the optimization of construction parameters; by establishing a risk distribution prediction model, the spatial distribution of the damage risk of the tunnel surrounding rock is predicted, which provides strong support for risk control and safety management during the construction process; through the risk assessment system, not only the accuracy and comprehensiveness of risk assessment are improved, but also more effective tools and methods are provided for the stability analysis and construction management of the tunnel surrounding rock, which has important engineering practical significance.

[0012] Optionally, the risk score is calculated by combining the normalized damage degree, deformation rate and surrounding rock stability score with dynamic weights. The present invention achieves flexible adjustment of the importance of different damage indicators by introducing dynamic weights, so that the risk score can better reflect the actual condition of the surrounding rock of the current tunnel; by normalizing the damage degree, deformation rate and surrounding rock stability scores, the influence of the dimensions between different indicators is eliminated, and the comparability and accuracy of the scores are improved; the risk score calculated by combining the normalized scores and dynamic weights can comprehensively and objectively reflect the damage risk of the surrounding rock of the tunnel, providing a more reliable basis for construction decisions.

[0013] Optionally, the step of obtaining optimized construction parameters based on the risk assessment system includes: constructing a deep reinforcement learning framework based on the risk assessment system; optimizing the dynamic adjustment mechanism of parameters based on the deep reinforcement learning framework in combination with real-time monitoring data to obtain optimization results; evaluating the optimization effect of construction parameters based on the optimization results to obtain evaluation results; and obtaining optimized construction parameters through the evaluation results. The present invention makes full use of the rich information provided by the risk assessment system by constructing a deep reinforcement learning framework to achieve intelligent optimization of construction parameters; combining real-time monitoring data and optimizing the dynamic adjustment mechanism of parameters using the deep reinforcement learning framework can respond to changes in the state of the surrounding rock of the tunnel in real time to ensure that the construction parameters are in a better state; through the evaluation of the optimization effect, not only the effectiveness of the optimization strategy is verified, but also a scientific basis is provided for the optimization and adjustment of construction parameters.

[0014] Optionally, the optimized construction parameters include explosive unit consumption, hole spacing, row spacing, delay interval and blast hole number parameters. The present invention achieves a significant improvement in blasting effect by accurately optimizing explosive unit consumption, while reducing explosive consumption and saving costs; by reasonably adjusting hole spacing and row spacing, the uniformity and efficiency of blasting are ensured, excessive damage and unexploded areas are reduced, and construction quality is improved; by optimizing the delay interval, effective control of blasting vibration is achieved, the stability of tunnel surrounding rock is protected, and safety risks are reduced; by accurately calculating the number of blast holes, the accuracy and efficiency of blasting operations are ensured, and construction efficiency is improved.

[0015] In the second aspect, the tunnel surrounding rock damage risk assessment system based on multi-scale feature fusion provided by the present invention comprises an input device, a processor, an output device and a memory, wherein the input device, the processor, the output device and the memory are interconnected, wherein the memory is used to store a computer program, the computer program comprises a program instruction, the processor is configured to call the program instruction, the input device comprises a data acquisition module, the data acquisition module comprises a geological parameter data acquisition unit, a monitoring data acquisition unit and a construction parameter data acquisition unit; the processor comprises an intelligent analysis module, the intelligent analysis module comprises a damage prediction unit, a risk assessment unit and a parameter optimization unit; the system uses the tunnel surrounding rock damage risk assessment method based on multi-scale feature fusion. The present invention realizes a comprehensive perception of the tunnel surrounding rock state by integrating multi-source data acquisition modules such as geological parameters, monitoring data and construction parameters, and provides a rich and accurate data basis for damage analysis; through the intelligent analysis module in the processor, the intelligent assessment of the tunnel surrounding rock damage risk and the optimization of construction parameters are realized, and the accuracy of the assessment and the construction efficiency are improved; by comprehensively considering local features, global features and time series features, the comprehensiveness and scientificity of the risk assessment are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flow chart of a tunnel surrounding rock damage risk assessment method based on multi-scale feature fusion according to an embodiment of the present invention;

[0017] Figure 2 A cross-sectional area division diagram of an embodiment of the present invention;

[0018] Figure 3 The blasting hole arrangement and damage cloud map of the embodiment of the present invention;

[0019] Figure 4 It is a structural schematic diagram of a tunnel surrounding rock damage risk assessment system based on multi-scale feature fusion according to an embodiment of the present invention;

[0020] Figure 5This is an operation flow chart of a tunnel surrounding rock damage risk assessment system based on multi-scale feature fusion according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The specific embodiments of the present invention will be described in detail below. It should be noted that the embodiments described herein are only for illustration and are not intended to limit the present invention. In the following description, a large number of specific details are set forth in order to provide a thorough understanding of the present invention. However, it is obvious to those of ordinary skill in the art that these specific details do not need to be adopted to implement the present invention. In other examples, in order to avoid confusing the present invention, known circuits, software or methods are not specifically described.

[0022] Throughout the specification, references to "one embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in conjunction with the embodiment or example is included in at least one embodiment of the present invention. Therefore, the phrases "in one embodiment," "in an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily all refer to the same embodiment or example. In addition, particular features, structures, or characteristics may be combined in one or more embodiments or examples in any suitable combination and / or subcombination. In addition, it should be understood by those of ordinary skill in the art that the figures provided herein are for illustrative purposes and that the figures are not necessarily drawn to scale.

[0023] See also Figure 1 The embodiment of the present invention provides a method for assessing the risk of tunnel surrounding rock damage based on multi-scale feature fusion, the method comprising the following steps:

[0024] S1. Obtain data on cross-sectional geometric parameters and geomechanical parameters of tunnel surrounding rock.

[0025] Among them, S1 includes the following steps:

[0026] S11. Divide the cross-sectional area of ​​the tunnel surrounding rock.

[0027] In one embodiment, based on the geometric shape and structural features of the cross section, it is divided into five areas, including the left top, the right top, the left side wall and the left bottom plate, the right side wall and the right bottom plate, and the middle groove area. Figure 2 As shown, number 1 is the left top, number 2 is the right top, number 3 is the left side wall and the left bottom plate, number 4 is the right side wall and the right bottom plate, and number 5 is the middle groove area.

[0028] The implementation effects of this embodiment are as follows: first, it is conducive to the formulation of different blasting strategies; second, it provides a solid foundation for crack expansion prediction, risk assessment and support design; third, it helps to comprehensively consider the differences in each area from the perspectives of space, mechanics and construction, and provide precise support for physical modeling and optimization decisions.

[0029] S12. Obtain data on cross-sectional geometric parameters of the tunnel surrounding rock.

[0030] In one embodiment, based on the cross-sectional area division result of step S11, a laser rangefinder is used to accurately measure the coordinates of the boundary points of each area, and key geometric dimensions such as width, height, curvature, etc. are recorded, while three-dimensional scanning technology is used to obtain more detailed geometric data.

[0031] S13. Obtain data on geomechanical parameters of tunnel surrounding rock.

[0032] In one embodiment, by comprehensively collecting static and dynamic geomechanical parameter data, the mechanical properties of the rock mass are fully understood, providing a scientific basis for risk assessment, support design and blasting parameter optimization of tunnel excavation.

[0033] Specifically, firstly, static parameter data of rocks, such as elastic modulus, Poisson's ratio, compressive strength and tensile strength, are obtained through laboratory tests.

[0034] Furthermore, strain gauges and ground stress probes are used to measure the ground stress distribution on site to clarify the direction and magnitude of the stress.

[0035] Furthermore, a field survey is conducted to record in detail the geometric characteristics of the rock mass fractures, including the inclination, direction and length of the fractures.

[0036] Furthermore, monitoring equipment is installed to collect real-time data on dynamic parameters such as surrounding rock deformation, support stress, and blasting vibration.

[0037] S2. Optimize the data to obtain optimization results.

[0038] Among them, S2 includes the following steps:

[0039] S21. Integrate the data.

[0040] In one embodiment, the relationship between the crack and the regional space is established based on the geometric characteristics and mechanical characteristics of the crack.

[0041] Specifically, an influence evaluation function is introduced, and the form of the influence evaluation function is as follows:

[0042] ,

[0043] in, is the influence evaluation function, is the crack length, is the distance from the crack to the region boundary, is the angle between the crack and the roadway direction, is the crack opening width, is the characteristic size, , , is the weight coefficient.

[0044] Furthermore, a multi-dimensional crack threshold judgment condition is constructed, as follows:

[0045] ,

[0046] in, is the measured length of the crack, is the critical length threshold, is the shortest vertical distance from the crack to the boundary of a region, To affect the distance threshold, is the angle between the crack direction and the roadway axis, is the critical angle at which the crack direction affects the regional stability. The weight calculation expression of the influence is as follows:

[0047] ,

[0048] in, is the weight of the influencing factor, for The weight of the influencing factor, is the normalized impact factor function, is the number of influencing factors.

[0049] Furthermore, the data are integrated and quantified through a regional crack density model, which satisfies the following expression:

[0050] ,

[0051] in, is the regional crack density, For the The length of the crack, is the area of ​​the region, is the number of cracks.

[0052] Furthermore, the crack parameters are subjected to time series evolution using a time series evolution function. The form of the time series evolution function is as follows:

[0053] ,

[0054] in, is the time series evolution function of the crack parameters, is the initial parameter, is the parameter variation, is the attenuation coefficient, is the time variable.

[0055] S22. Perform data cleaning and standardization on the data.

[0056] In one embodiment, statistical methods are used to detect outliers in the data, the statistical methods comprising box plots.

[0057] Furthermore, the detected extreme values ​​are removed to ensure the accuracy and consistency of the data.

[0058] Furthermore, each variable in the data is checked to determine which data points are missing. For missing values ​​of a single variable, linear interpolation is used to fill in the missing values. The specific formula of the linear interpolation method is as follows:

[0059] ,

[0060] in, are the missing values ​​that need to be interpolated. , are adjacent known data points, , is the time or space index of adjacent data points, A temporal or spatial index for missing data.

[0061] S23. Format the data.

[0062] In one embodiment, in order to effectively input the geological data and fracture characteristics of the cross-section area into the subsequent model, the data needs to be formatted. The expression is:

[0063] ,

[0064] in, For the The characteristic matrix of the cross-sectional area, is the geometric property, For physical properties, The geometric characteristics, the physical characteristics and the dynamic characteristics can be represented by a geometric characteristic matrix, a physical characteristic matrix and a dynamic characteristic matrix respectively.

[0065] Specifically, the geometric property matrix satisfies the following relationship:

[0066] ,

[0067] in, For the The geometric property matrix of the cross-sectional area, , , are the coordinates of the region center, , , is the area size, , , is the regional attitude angle.

[0068] The physical property matrix satisfies the following relationship:

[0069] ,

[0070] in, For the The physical property matrix of the cross-section area, is the elastic modulus, is Poisson's ratio, is the compressive strength, is the tensile strength, , The principal stress state.

[0071] The dynamic characteristic matrix satisfies the following relationship:

[0072] ,

[0073] in, For the The dynamic characteristic matrix of the cross-section area, is the displacement function, is the speed function, is the acceleration function, is the external force, is the dynamic pressure, To capture the effects of blasting and environmental changes on the thermodynamic properties of rock mass.

[0074] Furthermore, based on the characteristic matrix of the cross-sectional area, a correlation matrix of the cross-sectional area and the cracks is constructed.

[0075] Specifically, first, the correlation distance between the cross-section area and the crack is obtained, and the formula of the correlation distance is as follows:

[0076] ,

[0077] in, For the Section area and The correlation distance of the cracks, is the coordinate of the crack center point, The coordinates of the center or boundary point of the region.

[0078] Furthermore, based on the association distance, an association matrix is ​​constructed, and the form of the association matrix is ​​as follows:

[0079] ,

[0080] in, is the correlation matrix, is the total number of cross-sectional areas, Represents the total number of cracks.

[0081] Furthermore, the correlation matrix is ​​accurately solved by using a discriminant formula for determining whether the crack has a significant impact on the cross-section area. The discriminant formula is as follows:

[0082] ,

[0083] S24. Perform feature engineering on the data.

[0084] In one embodiment, by constructing a physics-driven feature extraction framework, the material damage evolution law is integrated into feature engineering; through multi-scale feature fusion and feature optimization strategies, intelligent extraction and dimensionality reduction optimization of high-dimensional features are achieved, thereby improving the physical consistency and predictive ability of the model.

[0085] Specifically, the characteristics of this process involve damage variables, equivalent stress intensity at the damage stage, equivalent stress intensity at the failure surface, strain rate enhancement factor, and pressure parameters under compaction. The damage variables satisfy the following relationship:

[0086] ,

[0087] in, is the damage variable, is the plastic strain increment, is the failure plastic strain. The equivalent stress strength in the damage stage satisfies the following relationship:

[0088] ,

[0089] in, is the equivalent stress intensity at the current damage stage, is the stress intensity when the material fails, is the residual equivalent stress strength. The equivalent stress strength of the failure surface satisfies the following relationship:

[0090] ,

[0091] in, is the equivalent stress strength of the failure surface, is the equivalent stress strength of the quasi-static failure surface compression meridian for Angular factor, is the strain rate dynamic enhancement factor, is the angle between the crack direction and the roadway axis, For pressure, is the strain rate. The strain rate enhancement factor is a dynamic response characteristic, which satisfies the following relationship:

[0092] ,

[0093] in, is the strain rate enhancement factor, is the compressive strain rate, is the dynamic strength parameter, is the reference compressive strain rate, is the reference tensile strain rate, and are the compressive strain rate exponent and the tensile strain rate exponent, respectively. Represents pressure. The pressure in the compacted state satisfies the following relationship:

[0094] ,

[0095] in, is the pressure under compaction, is the material compression ratio, is the initial strain, , , , is the empirical coefficient.

[0096] Furthermore, the contribution of each feature to the damage and dynamic enhancement factor is quantified through physical gradient sensitivity analysis. The expression of physical gradient sensitivity analysis is as follows:

[0097] ,

[0098] in, Features The importance rating of , is the target physical property, is the physical sensitivity of the feature, is the second norm, indicating the sensitivity.

[0099] Furthermore, the features are iteratively optimized through gradient updating and dynamic screening. The expression is as follows:

[0100] ,

[0101] in, For the After round optimization, is the learning rate, is the target loss based on the current feature, The bitwise multiplication is used to combine the gradient update of the feature with the importance screening. Keep only importance scores Exceeding the threshold The expression of gradient update and dynamic screening combines the gradient update of features with importance screening, achieving better dynamic optimization effect.

[0102] S3. Based on the optimization processing results, a damage status prediction model is constructed.

[0103] In one embodiment, a dynamic prediction system is created by combining the RHT constitutive model with real-time monitoring data. The system integrates physical-driven and data-driven methods to quantitatively evaluate and dynamically correct the degree of rock damage to ensure construction safety and parameter optimization. The RHT constitutive model is also called a damage constitutive model.

[0104] Specifically, in order to accurately evaluate the damage state of each area of ​​the rock mass, a rock mass damage evolution model based on the RHT constitutive model is constructed. This model integrates static parameters and dynamic monitoring data, focusing on the influence of plastic damage and strain rate effect on damage evolution. Among them, based on the plastic strain increment and failure plastic strain, the damage variable is defined is in the range of [0,1].

[0105] Furthermore, by combining the physical model of rock damage evolution with data-driven technology, a damage state prediction and optimization system is constructed. This system can not only accurately evaluate the damage degree of the cross-section area, but also optimize the support scheme through multi-source data. Among them, based on the stress tensor, strain tensor and time variable, the regional damage degree is defined to evaluate the regional damage. The regional damage degree satisfies the following expression:

[0106] ,

[0107] in, is the regional damage degree, is the evaluation function of regional damage, is the stress tensor, is the strain tensor, is the time variable.

[0108] Furthermore, based on the data and dynamic monitoring, a time evolution model of the damage state is constructed, and the time evolution model satisfies the following expression:

[0109] ,

[0110] in, for The damage status at the moment, for The damage status at the moment, is the stress state at the current moment, is the strain state at the current moment, are model parameters, is the evaluation function of regional damage.

[0111] Furthermore, by combining real-time monitoring data with regional characteristics, the dynamic changes of the damage state during the construction of the section are adjusted and optimized. This method achieves real-time response to complex construction environments by building a correction model, evaluating the correction confidence, and dynamically adjusting the warning threshold.

[0112] Specifically, in order to ensure that the model output is consistent with the actual construction situation, the damage state needs to be corrected, and the predicted damage variables are dynamically adjusted using real-time monitoring data to ensure consistency with the actual working conditions. The expression of damage correction is as follows:

[0113] ,

[0114] in, is the corrected damage variable, is the damage variable predicted based on the model, is the correction coefficient for dynamic adjustment, is the gradient change of the monitoring data.

[0115] Furthermore, the time evolution model of the damage state is updated using the modified damage variables to obtain a damage state prediction model.

[0116] S4. Establish a risk assessment system based on the damage status prediction model.

[0117] In one embodiment, based on the damage state prediction model, an evaluation system of multi-scale feature fusion, risk score output and distribution warning is established for the damage and risk characteristics during the cross-section construction process. The evaluation system ensures construction safety and optimizes support strategies by quantifying the correlation between damage characteristics and geological parameters, risk scoring and grading, as well as risk distribution prediction and warning design.

[0118] Specifically, S4 includes the following steps:

[0119] S41. Based on the damage state prediction model, multi-scale features are integrated.

[0120] In this embodiment, based on the damage state prediction model, the local, global and temporal features are adaptively weighted and fused using a feature extraction network, which can be expressed as:

[0121] ,

[0122] in, is a multi-scale fusion feature. is a local feature, For the overall characteristics, is the time series feature, is the adaptive weight.

[0123] Furthermore, to quantify the importance of each feature, the input feature set is optimized:

[0124] ,

[0125] in, Features The weight of Features and Features The relevance of Features The importance rating of is the total number of features.

[0126] S42. Based on the multi-scale features, output a risk score.

[0127] In this embodiment, the damage status of the cross-section area is quantitatively evaluated by outputting a risk score.

[0128] Specifically, considering the dynamic nature of the relationship between the features, the weights are dynamically adjusted to construct a risk scoring model, which satisfies the following expression:

[0129] ,

[0130] in, For the comprehensive risk score, , The dynamic weight reflects the changes in the importance of features at different times. is the normalized damage degree, is the normalized deformation rate, is the normalized surrounding rock stability score.

[0131] S43. Based on the risk score, establish a risk distribution prediction model and a dynamic early warning system.

[0132] In this embodiment, based on the real-time monitoring data of the cross-section area and historical risk analysis, a risk distribution prediction model and a dynamic early warning system are established to help the construction team identify high-risk areas in a timely manner and formulate response measures.

[0133] Specifically, the spatial interpolation method is used in combination with dynamic monitoring data to generate a cross-sectional risk distribution map. The relevant expression is as follows:

[0134] ,

[0135] in, The risk score of the spatial coordinates on the time axis, is the risk scoring model, is the degree of damage, is the strain rate, For the damaged area, For time, , is the coordinate point.

[0136] Furthermore, the damage change trend is predicted by using a time series model. The time series model satisfies the following expression:

[0137]

[0138] in, for The risk score at the moment, for The risk score at the moment, is a time series forecasting model, for The damage change rate at time, for The strain rate at time, is time. This time series model can also be called a risk distribution prediction model.

[0139] Furthermore, a dynamic early warning system is established using the above risk distribution prediction model.

[0140] S5. Obtain optimized construction parameters based on the risk assessment system.

[0141] In one embodiment, a deep reinforcement learning framework is used to take the section damage state prediction and risk assessment results obtained in the early stage as the optimization basis, and the blasting parameters are intelligently and dynamically optimized to improve the construction efficiency and reduce the surrounding rock damage while ensuring the construction safety. The blasting parameters include charge amount, hole spacing, row spacing and delay interval. Figure 3As shown, according to the blasting parameters, the blastholes are arranged and their damage cloud map is obtained. Through real-time data feedback and dynamic adjustment mechanism, the construction parameters can be flexibly adjusted according to the real-time changes of the engineering environment to achieve intelligent control and optimized decision-making of the construction process.

[0142] Specifically, based on the prediction of the section damage state and the risk assessment results, a deep reinforcement learning framework is constructed, which satisfies the following expression:

[0143] ,

[0144] in, It is a deep reinforcement learning framework. is the regional damage degree distribution matrix, is the current stress state, Current deformation state, is the risk score distribution.

[0145] Furthermore, for the above deep reinforcement learning framework, its action space is defined as:

[0146] ,

[0147] in, is the action space, is the unit charge parameter set, is the hole spacing parameter set, is the row spacing parameter set, is the delay interval, is the number of blast holes.

[0148] Furthermore, based on the action space, a reward function is designed, and the expression of the reward function is as follows:

[0149] ,

[0150] in, is the reward function, is the overall damage degree, For construction efficiency, is the comprehensive risk value, , , is the weight coefficient, is the preset risk threshold.

[0151] Furthermore, according to the real-time monitoring data, the dynamic parameter adjustment mechanism is optimized to achieve flexible response to changes in the construction environment. In the optimization process of the dynamic parameter adjustment mechanism, the construction parameters can be updated using the policy gradient method. The policy gradient method satisfies the following expression:

[0152] ,

[0153] in, for The construction parameters at the moment, for The construction parameters at the moment, Based on the current state and actions The policy gradient of is the step length control, It is the angle between the crack and the direction of the tunnel.

[0154] Furthermore, a dynamic adjustment mechanism is introduced to adjust the step size. The dynamic adjustment mechanism satisfies the following expression:

[0155] ,

[0156] in, is the current step length, is the initial step length, is the attenuation coefficient, Score the current risk, Score the target risk. near Time, step length It decreases exponentially, ensuring that the optimization process is smoother when approaching the optimal solution.

[0157] Furthermore, in order to ensure the effectiveness and practicality of the optimization results, the following evaluation method is used to evaluate the optimization results.

[0158] Specifically, the parameter sensitivity analysis method is used to quantify the influence of each parameter on the optimization target. The parameter sensitivity analysis method satisfies the following expression:

[0159] ,

[0160] in, is the parameter sensitivity, The objective function parameter The partial guide, is the parameter value, is the objective function.

[0161] Furthermore, an evaluation index of the optimization effect is designed, and the evaluation index satisfies the following expression:

[0162] ,

[0163] in, As the evaluation index of the optimization effect, , , Represent the improvement of damage degree, efficiency and risk respectively, , , They represent the damage degree, efficiency and risk before optimization respectively. , , is the weight coefficient.

[0164] Furthermore, according to the above evaluation index of the optimization effect, combined with the set threshold, the optimization result is evaluated to obtain the optimized construction parameters.

[0165] S6. Based on the optimized construction parameters, a model system is established for predicting damage status, assessing damage risk and optimizing construction parameters.

[0166] In one embodiment, based on the optimized construction parameters, models for predicting damage status, assessing damage risk and optimizing construction parameters are trained and verified; the purpose of the training and verification is to establish a model system with physical consistency and engineering applicability through data preprocessing, deep learning model training and multi-scenario engineering verification, to ensure that the model can accurately predict damage status, assess risks and optimize construction parameters in actual construction.

[0167] Specifically, sub-datasets are constructed according to different geological conditions to ensure that the model can learn and adapt to diverse and complex geological environments. Through classification and refinement, the model's ability to understand rock mass characteristics and surrounding rock conditions is improved.

[0168] Specifically, the data is classified by lithology, that is, the data is divided into hard rock, soft rock, and broken zone, etc., to construct a lithology data set:

[0169] ,

[0170] in, is the lithology dataset, For the The lithology of the data, is the total number of data.

[0171] Furthermore, the data are classified according to the surrounding rock stability, and the following surrounding rock stability data sets are obtained:

[0172] ,

[0173] in, is the surrounding rock stability data set, For the The surrounding rock stability of the data, is the total number of data, It is the grade of surrounding rock stability.

[0174] Furthermore, the continuity of the time series data is checked.

[0175] Specifically, identify and remove unreasonable monitoring data jumps to ensure data smoothness and improve the accuracy of the model's prediction of time changes. The following time series continuity index is used to verify time series continuity. The time series continuity index satisfies the following expression:

[0176] ,

[0177] in, is the temporal continuity indicator, is the total number of time series data points, For the The observed value at a time point, For the The observed value at a time point, is the change between two adjacent time points, is the maximum allowed change threshold.

[0178] Furthermore, through phased training, multi-objective optimization and integration of construction process constraints, the performance of the model in damage prediction, risk assessment and construction parameter optimization is improved to ensure the physical consistency and engineering applicability of the model.

[0179] Specifically, the physical model pre-training expression is as follows:

[0180] ,

[0181] in, is the result of physical model pre-training. The target values ​​generated for the physical model, is the predicted value of the deep learning model, is the regularization term of the model parameters, is the regularization strength, , is the model parameter.

[0182] Furthermore, the engineering data is fine-tuned, and the expression is:

[0183] ,

[0184] in, This is the result after fine-tuning the engineering data. is the observation data in actual engineering. is the loss of pre-training the physical model, In order to control the degree of integration between the physical model and the measured data, is the predicted value of the deep learning model, , is the model parameter.

[0185] Furthermore, through multi-scenario verification and performance evaluation, the adaptability and reliability of the model in actual construction are fully tested. The focus is on evaluating the prediction ability of the model under different geological conditions and construction scenarios, as well as its guiding role in construction risks and efficiency. Among them, the evaluation indicator of construction parameter sensitivity is involved. The construction parameter sensitivity satisfies the following expression:

[0186] ,

[0187] in, is the sensitivity of construction parameters, is the construction parameter, is the model output, is the partial derivative of the output with respect to the construction parameters, is the relative rate of change factor.

[0188] S7. Through the model system, the risk of damage to the surrounding rock of the tunnel is evaluated.

[0189] In one embodiment, the model system established in step S6 realizes real-time assessment of the damage risk of the tunnel surrounding rock. By inputting real-time geological and construction data, the model system can quickly calculate the current damage risk and provide early warnings and suggestions. This will help to timely discover potential safety hazards and take appropriate measures to prevent and deal with them.

[0190] See also Figure 4 , Figure 4 The structure diagram of the tunnel surrounding rock damage risk assessment system based on multi-scale feature fusion in an embodiment of the present invention is shown in FIG. The system includes an input device, a processor, an output device and a memory, wherein the input device, the processor, the output device and the memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, the processor is configured to call the program instructions, the system uses the tunnel surrounding rock damage risk assessment method based on multi-scale feature fusion, and the specific process of the system operation is as follows: Figure 5 shown.

[0191] In this embodiment, the input device includes a data acquisition module, which is responsible for collecting various types of construction and environmental data.

[0192] Specifically, the data acquisition module includes a geological parameter data acquisition unit, a monitoring data acquisition unit and a construction parameter data acquisition unit;

[0193] The geological parameter data acquisition unit is used to collect static data of rock mass physical and mechanical parameters and ground stress state parameters, wherein the rock mass physical and mechanical parameters include elastic modulus and compressive strength;

[0194] The monitoring data acquisition unit is used to obtain dynamic monitoring data such as stress, strain, displacement, etc. in real time;

[0195] The construction parameter data acquisition unit is used to obtain key data in the construction process such as blasting parameters and support parameters.

[0196] The processor includes an intelligent analysis module, and the intelligent analysis module is used for data processing and optimization.

[0197] Specifically, the intelligent analysis module includes a damage prediction unit, a risk assessment unit and a parameter optimization unit;

[0198] The damage prediction unit evaluates the damage state of the section based on the RHT constitutive model;

[0199] The risk assessment unit outputs a regionalized risk score by fusing multi-scale features;

[0200] The parameter optimization unit uses deep reinforcement learning to achieve dynamic optimization of construction parameters;

[0201] The output device includes a display for displaying the processing result of the processor.

[0202] The memory adopts a high-speed solid-state hard disk, which has the characteristics of fast reading and writing speed, large capacity and high reliability. It is mainly used to store data input by the input device and the result data processed by the processor, and can meet the needs of large data storage.

[0203] In a specific embodiment, a monitoring system is deployed based on the data acquisition module.

[0204] Specifically, the monitoring systems deployed include stress monitoring systems, deformation monitoring systems and environmental monitoring systems to fully cover the key monitoring needs of the cross-section construction area; the stress monitoring system captures the stress changes of the rock mass in real time by deploying hole stress gauges along the periphery of the cross-section, deploying anchor dynamometers at key points of the support system, and installing ground stress monitoring devices in reserved observation holes; the deformation monitoring system deploys multi-point displacement gauges around the cross-section, deploys optical fiber strain gauges in key areas for real-time strain monitoring, and uses convergence strain gauges to monitor the overall deformation state of the cross-section. The environmental monitoring system deploys crack monitors in crack development areas, monitors groundwater levels and water pressure in real time, and uses vibration monitoring devices to evaluate the impact of blasting vibration on the surrounding environment, thereby achieving comprehensive dynamic monitoring of the construction process.

[0205] In summary, the tunnel surrounding rock damage risk assessment method based on multi-scale feature fusion provided by the present invention realizes accurate damage risk assessment and parameter optimization by integrating geomechanical parameters, historical and real-time monitoring data, and combining the RHT constitutive model with deep reinforcement learning, which significantly improves the reliability of damage risk assessment and provides intelligent decision-making support for tunnel excavation construction.

[0206] The tunnel surrounding rock damage risk assessment system based on multi-scale feature fusion provided by the present invention realizes a comprehensive perception of the tunnel surrounding rock status by integrating multi-source data acquisition modules such as geological parameters, monitoring data and construction parameters, and provides a rich and accurate data basis for damage analysis; through the intelligent analysis module in the processor, it realizes intelligent assessment of tunnel surrounding rock damage risk and optimization of construction parameters, thereby improving the accuracy of assessment and construction efficiency; by comprehensively considering local characteristics, global characteristics and time series characteristics, it improves the comprehensiveness and scientificity of risk assessment.

[0207] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A tunnel surrounding rock damage risk assessment method based on multi-scale feature fusion, characterized in that: The method comprises the following steps: Obtain data on cross-sectional geometric parameters and geomechanical parameters of tunnel surrounding rock; Performing optimization processing on the data to obtain optimization processing results; Based on the optimization processing results, a damage state prediction model is constructed; According to the damage state prediction model, multi-scale features are integrated, wherein the multi-scale features include local features, global features and time series features; Based on the multi-scale features, a risk score is output, wherein the risk score is calculated by combining the normalized damage degree, deformation rate and surrounding rock stability score with a dynamic weight; Based on the risk score, establishing a risk distribution prediction model; Establishing a risk assessment system through the multi-scale features, the risk score and the risk distribution prediction model; Obtaining optimized construction parameters based on the risk assessment system; Based on the optimized construction parameters, a model system is established for predicting damage status, assessing damage risk and optimizing construction parameters; The model system is used to evaluate the damage risk of tunnel surrounding rocks.

2. According to the method for risk assessment of tunnel surrounding rock damage based on multi-scale feature fusion according to claim 1, it is characterized in that: The data for obtaining the cross-sectional geometric parameters and geomechanical parameters of the tunnel surrounding rock include: Divide the cross section of the tunnel surrounding rock into multiple areas; According to the plurality of regions, data of cross-sectional geometric parameters are collected; Based on the data of the cross-section geometric parameters, data of geomechanical parameters are collected, and the data of the geomechanical parameters include data of elastic modulus, Poisson's ratio, compressive strength and tensile strength of the rock mass.

3. The method for risk assessment of tunnel surrounding rock damage based on multi-scale feature fusion according to claim 1 is characterized in that: The optimizing the data to obtain the optimizing result comprises: Integrate and process the data of the cross-section geometric parameters and geomechanical parameters of the tunnel surrounding rock; Performing data cleaning and standardization on the data; Formatting the data; Performing feature engineering processing on the data, wherein the feature engineering processing includes high-dimensional feature enhancement and intelligent dimensionality reduction processing; Through the integration processing, the data cleaning and standardization processing, the formatting processing and the feature engineering processing, an optimization processing result is obtained.

4. The method for risk assessment of tunnel surrounding rock damage based on multi-scale feature fusion according to claim 3 is characterized in that: The formatting process includes: Constructing a characteristic matrix of the cross-section region, wherein the characteristic matrix includes a geometric characteristic matrix, a physical characteristic matrix, and a dynamic characteristic matrix; Based on the characteristic matrix, a correlation matrix of cross-section areas and cracks is constructed.

5. The method for risk assessment of tunnel surrounding rock damage based on multi-scale feature fusion according to claim 1 is characterized in that: The constructing of a damage state prediction model based on the optimization processing result includes: Based on the optimization processing results, a damage state time evolution model is constructed using a damage constitutive model; Based on the damage state time evolution model, dynamic damage correction is performed using real-time monitoring data to obtain correction results; The correction result is used to update the damage state time evolution model to obtain a damage state prediction model.

6. The method for risk assessment of tunnel surrounding rock damage based on multi-scale feature fusion according to claim 1 is characterized in that: According to the risk assessment system, obtaining optimized construction parameters includes: Based on the risk assessment system, a deep reinforcement learning framework is constructed; According to the deep reinforcement learning framework, combined with real-time monitoring data, the dynamic adjustment mechanism of parameters is optimized to obtain optimization results; Based on the optimization results, the optimization effect of the construction parameters is evaluated to obtain an evaluation result; The optimized construction parameters are obtained through the evaluation results.

7. The method for risk assessment of tunnel surrounding rock damage based on multi-scale feature fusion according to claim 6 is characterized in that: The optimized construction parameters include explosive consumption per unit, hole spacing, row spacing, delay interval and blast hole quantity parameters.

8. A tunnel surrounding rock damage risk assessment system based on multi-scale feature fusion, the system using a tunnel surrounding rock damage risk assessment method based on multi-scale feature fusion according to any one of claims 1 to 7, characterized in that: The system includes an input device, a processor, an output device, and a memory; The input device includes a data acquisition module, and the data acquisition module includes a geological parameter data acquisition unit, a monitoring data acquisition unit and a construction parameter data acquisition unit; The processor includes an intelligent analysis module, which includes a damage prediction unit, a risk assessment unit and a parameter optimization unit; The input device, the processor, the output device and the memory are connected to each other, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions.

Images