Crushed surrounding rock mountain tunnel construction monitoring method and related equipment

Abstract

The invention discloses a broken surrounding rock mountain tunnel construction monitoring method and related equipment, and relates to the technical field of tunnel construction. Optimizing the state data by adopting a Kalman filtering algorithm to obtain optimized state data; generating fragmentation degree information and spatial distribution information based on the optimization state data; and a reinforcement learning algorithm is adopted, and a construction strategy is adjusted based on the crushing degree information and the spatial distribution information.

Description

Technical Field

[0001] The present application relates to the technical field of tunnel construction, and in particular, to a construction monitoring method and related equipment for a mountain tunnel with broken surrounding rock. Background Art

[0002] During the tunnel construction process, the fragmentation of the surrounding rock is one of the key factors affecting construction safety and progress, especially prominent under complex geological conditions. Traditional monitoring methods for surrounding rock fragmentation mainly rely on manual measurement or simple sensor data collection. These methods often cannot reflect the dynamic changes of the surrounding rock in real time and comprehensively, and there are lags in data processing and analysis, making it difficult to respond to changes in factors such as stress and strain during construction in a timely manner.

[0003] The existing technologies mostly stay in the static evaluation stage, lacking the ability of dynamic monitoring of the construction environment and automatic adjustment of construction strategies. This results in difficulty in timely adjusting the construction plan to cope with potential risks during the actual construction process when the fragmentation situation of the surrounding rock changes, thus increasing the probability of construction accidents. Therefore, there is an urgent need for a construction monitoring method for a mountain tunnel with broken surrounding rock to solve the above problems. Summary of the Invention

[0004] A series of simplified concepts are introduced in the Summary of the Invention section, which will be further elaborated in the Detailed Description section. The Summary of the Invention section of the present application does not mean to attempt to define the key features and essential technical features of the claimed technical solution, nor does it mean to attempt to determine the protection scope of the claimed technical solution.

[0005] In a first aspect, the present application provides a construction monitoring method for a mountain tunnel with broken surrounding rock, including:

[0006] Obtaining the state data of the surrounding rock;

[0007] Optimizing the above state data by using the Kalman filter algorithm to obtain optimized state data;

[0008] Generating fragmentation degree information and spatial distribution information based on the above optimized state data;

[0009] Adjusting the construction strategy by using the reinforcement learning algorithm based on the above fragmentation degree information and the above spatial distribution information.

[0010] In some embodiments, the optimizing the above state data by using the Kalman filter algorithm to obtain optimized state data includes:

[0011] Constructing a state vector based on the above state data;

[0012] Predict the above state vector and estimated error covariance matrix using the state transition matrix to obtain the prior state estimate and the prior estimated error covariance matrix;

[0013] Based on the new observation data, determine the filtering model for the next moment;

[0014] Based on the observation matrix of the above filtering model, the above prior estimated error covariance matrix, and the observation noise covariance matrix, calculate the Kalman gain matrix;

[0015] Based on the above Kalman gain matrix and the observation residual, correct the above prior state estimate to obtain the posterior state estimate;

[0016] Use the above posterior state estimate as the optimized state data.

[0017] In some embodiments, based on the new observation data, determining the filtering model for the next moment includes:

[0018] Based on the prediction error of the new observation data by the filtering model, calculate the model probability P i (t), expressed as:

[0019]

[0020] where η is the normalization factor; σ i is the noise standard deviation of the i-th model; H i is the observation matrix of the i-th model; P i (t) is the probability that the i-th model is selected at time t; y(t + 1) is the above new observation data; is the prior state estimate of the i-th model; P i (t + 1) is the probability that the i-th model is selected at time t + 1;

[0021] Based on the model probability, select the model i with the highest probability * as the filtering model at time t + 1.

[0022] In some embodiments, it further includes:

[0023] Based on the above Kalman gain matrix, update the above prior estimated error covariance matrix to obtain the posterior estimated error covariance matrix for updating the state at the next moment.

[0024] In some embodiments, based on the above optimized state data, generating the fragmentation degree information includes:

[0025] The above fragmentation degree information is determined by the following formula:

[0026]

[0027] Among them, F 破率 (t) is the fragmentation degree information, which is the fragmentation degree of the surrounding rock at time t; σ(x, t) is the stress at position x at time t; ∈(x, t) is the strain at position x at time t; θ(x, t) is the plasticity index; is the steepness of the non-linear function; φ is the starting point of the non-linear function; L is the total length of the tunnel.

[0028] In some embodiments, based on the above optimization status data, spatial distribution information is generated, including:

[0029] Based on the above optimization status data, a convolutional neural network is used to extract the characteristics of the spatial distribution of the surrounding rock to obtain spatial characteristics;

[0030] Based on the above spatial characteristics, a recurrent neural network is used to capture the time variation to obtain the dynamic characteristics of the surrounding rock fragmentation;

[0031] Based on the above dynamic characteristics of the surrounding rock fragmentation, the spatial distribution of the surrounding rock fragmentation is determined to generate spatial distribution information.

[0032] In some embodiments, the above-mentioned reinforcement learning algorithm is used to adjust the construction strategy based on the above fragmentation degree information and the above spatial distribution information, including:

[0033] Define the action space and state space of the construction strategy;

[0034] Based on the predicted above fragmentation degree information and the above spatial distribution information, evaluate the current construction strategy;

[0035] Optimize the construction parameters through the reinforcement learning algorithm to minimize the risk of surrounding rock fragmentation;

[0036] Real-time feedback the optimization result and adjust the above construction strategy.

[0037] In a second aspect, the present application proposes a construction monitoring device for a mountain tunnel with fractured surrounding rock, including:

[0038] A surrounding rock data acquisition module for acquiring the status data of the surrounding rock;

[0039] A status data optimization module for optimizing the above status data by using the Kalman filtering algorithm to obtain optimized status data;

[0040] A surrounding rock information generation module for generating fragmentation degree information and spatial distribution information based on the above optimized status data;

[0041] A construction strategy adjustment module for adjusting the construction strategy by using the reinforcement learning algorithm based on the above fragmentation degree information and the above spatial distribution information.

[0042] In a third aspect, an electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor is configured to implement the steps of the monitoring method for constructing a fractured surrounding rock mountain tunnel according to any one of the first aspects when executing the computer program stored in the memory.

[0043] In a fourth aspect, the present application further provides a computer-readable storage medium, on which a computer program is stored. The computer program, when executed by a processor, implements the monitoring method for constructing a fractured surrounding rock mountain tunnel according to any one of the first aspects.

[0044] In summary, through the integration of the Kalman filter algorithm, deep learning technology, and reinforcement learning algorithm, the present application can monitor and dynamically manage the fracture condition of the surrounding rock during tunnel construction in real time and with high precision. Specifically, the Kalman filter algorithm optimizes the surrounding rock state data collected by multi-source sensors, effectively removing noise and errors and improving the accuracy of the data. Subsequently, a deep learning model combining a convolutional neural network and a recurrent neural network is used to generate detailed information on the fracture degree and spatial distribution of the surrounding rock, comprehensively reflecting the fracture state of the surrounding rock under different time and space conditions. Finally, the reinforcement learning algorithm is used to intelligently adjust the construction strategy according to the real-time monitoring data, such as optimizing the blasting force and excavation speed, to automatically optimize the construction parameters, thereby effectively reducing the risk of surrounding rock fracture and improving construction safety and efficiency. In addition, the system has a high degree of adaptability and intelligent decision-making functions, can respond in real time to changes in the construction environment, significantly enhance the robustness and flexibility of the tunnel construction process, reduce human intervention and operation errors, and improve the construction quality and progress control level. Generally speaking, the present application provides an efficient and intelligent tunnel construction monitoring and management solution, which not only helps to reduce construction accidents and ensure the safety of construction personnel, but also improves construction efficiency and quality, has broad application prospects and significant economic and social benefits, and promotes the development of tunnel construction technology towards intelligence and modernization. Description of the Drawings

[0045] By reading the detailed description of the preferred embodiments below, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to limit this specification. Moreover, throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0046] Figure 1 It is a schematic flowchart of the monitoring method for constructing a fractured surrounding rock mountain tunnel provided by an embodiment of the present application;

[0047] Figure 2 It is a schematic structural diagram of the monitoring device for constructing a fractured surrounding rock mountain tunnel provided by an embodiment of the present application;

[0048] Figure 3 This is a schematic structural diagram of the electronic equipment for construction monitoring of a broken surrounding rock mountain tunnel provided by an embodiment of the present application. Specific embodiments

[0049] The terms "first", "second", "third", "fourth", etc. (if any) in the specification, claims and above-mentioned drawings of the present application are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that such used data can be interchanged under appropriate circumstances so that the embodiments described herein can be implemented in an order different from that illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices. The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments.

[0050] Please refer to Figure 1 , which is a schematic flow diagram of a method for construction monitoring of a broken surrounding rock mountain tunnel provided by an embodiment of the present application, and specifically may include:

[0051] S110. Obtain the state data of the surrounding rock;

[0052] Exemplarily, during the tunnel construction process, obtaining the state data of the surrounding rock is the basis and key link of the entire monitoring method. In order to achieve real-time and comprehensive monitoring of the dynamic changes of the surrounding rock, the present application adopts a distributed sensor network to collect the physical quantity data inside the tunnel in real time through a variety of sensor devices. These sensors include strain gauges, pressure sensors, displacement sensors, temperature sensors, humidity sensors, etc., which are respectively used to measure key parameters such as stress, strain, displacement, temperature and humidity of the surrounding rock during the construction process. Strain gauges and pressure sensors can reflect the internal stress distribution and deformation of the surrounding rock under external forces such as blasting and excavation, and the displacement sensor is used to monitor the displacement of the surrounding rock to timely detect potential displacement anomalies. In addition, temperature and humidity sensors can capture the changes in environmental conditions, and these factors have a significant impact on the mechanical properties and fragmentation behavior of the surrounding rock.

[0053] To ensure the accuracy and real-time nature of data acquisition, the sensor network adopts a hierarchical architecture, which is divided into basic sensors, middle-layer sensors, and high-layer sensors. The basic sensors are mainly deployed at key positions close to the construction area and are responsible for collecting high-frequency stress and strain data; the middle-layer sensors cover the overall structure and monitor the overall deformation and displacement of the surrounding rock; the high-layer sensors are used for environmental monitoring and record environmental parameters such as temperature and humidity in real time. Each layer of sensor nodes has preliminary data processing capabilities, such as pre-filtering and preliminary analysis, and can perform preliminary data processing and filtering locally, reducing the amount of data transmitted and improving the data transmission efficiency. At the same time, the sensors transmit the collected data to the central processing system in real time through wireless communication technology to ensure the real-time nature and integrity of the data. This distributed data acquisition and processing architecture not only improves the coverage and accuracy of data acquisition but also enhances the adaptability and robustness of the system in complex construction environments, providing a reliable data basis for subsequent state data optimization and fragmentation degree assessment.

[0054] S120. Optimize the above-mentioned state data using the Kalman filtering algorithm to obtain optimized state data;

[0055] Exemplarily, after obtaining the state data of the surrounding rock, the next step is to optimize these data using the Kalman filtering algorithm to obtain a more accurate state estimate. Since the actually collected sensor data often contains noise and errors, the Kalman filtering algorithm can effectively remove noise, reduce errors, and optimize the state estimate of the surrounding rock by recursively combining the system's prediction model and the observed data. The Kalman filter is an optimal estimation method based on the minimum mean square error. It combines the system state estimate with the observed data and dynamically adjusts the weights of the prediction and observed data in the final estimate by calculating the Kalman gain matrix, thereby improving the accuracy of the data.

[0056] S130. Generate fragmentation degree information and spatial distribution information based on the above-mentioned optimized state data;

[0057] Exemplarily, based on the surrounding rock state data optimized by the Kalman filtering algorithm, the next step is to generate the fragmentation degree information and spatial distribution information of the surrounding rock. This process mainly relies on the comprehensive evaluation of multiple physical quantities, including data such as stress, strain, temperature, and humidity, which provide the necessary inputs for calculating the fragmentation degree of the surrounding rock. Based on the optimized state data, the finite element method is used to model the mechanical behavior of the surrounding rock, and the physical quantity distribution of the surrounding rock at different positions and times is calculated through numerical simulation. Stress and strain are the main factors determining the fragmentation of the surrounding rock. Through the spatial distribution of these data, it is possible to judge whether the surrounding rock has reached the critical point of fragmentation. The fragmentation degree information is generated by constructing a multi-dimensional fragmentation evaluation model. This model combines multiple factors such as stress, strain, and plastic deformation, and is corrected through a non-linear correction function to reflect the dynamic evolution process of fragmentation.

[0058] Secondly, the generation of the spatial distribution information adopts the combined application of CNN (Convolutional Neural Network) and RNN (Recurrent Neural Network). The optimized state data first extracts the spatial features of the surrounding rock through CNN. These features include the distribution patterns of stress and strain at various positions in the tunnel. The convolutional layer of CNN can identify the local change trends of the surrounding rock at different spatial positions and generate high-level spatial feature representations. Subsequently, these spatial features are input into the RNN to capture the dynamic changes of the surrounding rock fragmentation over time. RNN can process time series data, identify the temporal correlation and evolution law in the surrounding rock fragmentation process, and generate the dynamic features of the surrounding rock fragmentation. Finally, by combining the spatial features extracted by CNN and the temporal dynamic features captured by RNN, the system can accurately generate the spatial distribution information of the surrounding rock fragmentation, comprehensively reflecting the fragmentation state of the surrounding rock under different time and space conditions.

[0059] S140. Adopt the reinforcement learning algorithm to adjust the construction strategy based on the above fragmentation degree information and the above spatial distribution information.

[0060] Exemplarily, after obtaining the fragmentation degree information and spatial distribution information, the next step is to use the reinforcement learning algorithm to optimize the construction strategy based on this information. Reinforcement learning is a machine learning method that learns the optimal strategy through interaction with the environment. In tunnel construction, the optimization of the construction strategy can be regarded as a reinforcement learning problem, where the construction personnel select appropriate construction measures according to the fragmentation degree and spatial distribution information of the surrounding rock. The reinforcement learning algorithm gradually learns the optimal construction strategy that can reduce the risk of surrounding rock fragmentation by evaluating the effects of current construction decisions.

[0061] In summary, in this application, the Kalman filter algorithm is used to optimize multi-source sensor data, eliminate noise and errors, and obtain a more accurate estimation of the surrounding rock state, thereby providing reliable input for subsequent fragmentation assessment and spatial distribution prediction. By combining deep learning and reinforcement learning algorithms, the construction plan can be dynamically adjusted according to real-time data, and potential surrounding rock fragmentation risks can be warned in a timely manner, thus greatly improving the safety and efficiency of tunnel construction. At the same time, based on the spatial covariance matrix and Bayesian data fusion technology, this application effectively improves the accuracy of the surrounding rock fragmentation spatial distribution prediction, provides more comprehensive monitoring data for construction personnel, and optimizes the construction management and decision-making process.

[0062] In some examples, the above-mentioned Kalman filter algorithm is used to optimize the above-mentioned state data to obtain optimized state data, including:

[0063] Construct a state vector based on the above state data;

[0064] Use the state transition matrix to predict the above state vector and the estimated error covariance matrix to obtain a priori state estimation and a priori estimated error covariance matrix;

[0065] Based on the new observation data, determine the filtering model for the next moment;

[0066] Based on the observation matrix of the above filtering model, the above a priori estimated error covariance matrix, and the observation noise covariance matrix, calculate the Kalman gain matrix;

[0067] Based on the above Kalman gain matrix and the observation residual, correct the above a priori state estimation to obtain a posteriori state estimation;

[0068] Use the above a posteriori state estimation as the optimized state data.

[0069] Exemplarily, a state vector is constructed based on the obtained original state data, and the vector covers key parameters such as the stress, strain, displacement, temperature, and humidity of the surrounding rock, comprehensively reflecting the dynamic state of the surrounding rock during the tunnel construction process. Next, the state transition matrix is used to predict the state vector and its estimated error covariance matrix to generate a priori state estimation and a priori estimated error covariance matrix. This step pre-predicts the state at the next moment through the system's dynamic model, providing a basis for subsequent observation updates.

[0070] Based on new observation data, determine the filtering model for the next moment. The selection of the filtering model usually adopts a multi-model method that combines EKF (Extended Kalman Filter) and UKF (Unscented Kalman Filter) to adapt to different types of dynamic changes during tunnel construction. Based on the observation matrix, prior estimation error covariance matrix, and observation noise covariance matrix of the selected filtering model, calculate the Kalman gain matrix. The Kalman gain matrix plays a crucial role in this process, which determines the weight allocation of the predicted value and the actual observation value in the final state estimation. By combining the Kalman gain matrix with the observation residual, correct the prior state estimation, and finally obtain the posterior state estimation. This posterior state estimation is not only closer to the true surrounding rock state but also effectively reduces the uncertainty introduced by sensor noise and measurement errors. Finally, use the obtained posterior state estimation as the optimized state data to provide high-precision and reliable data support for subsequent fragmentation degree evaluation and construction strategy adjustment.

[0071] In addition, to further improve the adaptability and accuracy of the Kalman filtering algorithm, this application introduces a dynamic adjustment mechanism for the adaptive noise covariance matrix. According to the real-time changes in the construction environment, adaptively adjust the process noise covariance matrix and the observation noise covariance matrix, so that the filter can more accurately reflect the uncertainty of the system and measurement. This adaptive adjustment not only improves the robustness of the filter in different construction stages and environmental conditions but also enhances the real-time response ability of the system in the complex tunnel construction environment. Through the above optimization steps, the Kalman filtering algorithm can effectively remove the noise and errors in the sensor data, provide high-precision optimized state data, and ensure more accurate and reliable surrounding rock monitoring during tunnel construction.

[0072] The system model is used to describe how the physical state of the surrounding rock changes over time and how these states are observed by sensors, specifically including:

[0073] Define the state vector, denoted as:

[0074] x(t) = [p(t), v(t), σ(t), e(t), T(t), H(t)] T

[0075] Among them, \(x(t)\) is the state vector, representing the state of the system at time \(t\); \(p(t)\) is the position, which reflects the specific position of the sensor inside the tunnel and records the spatial distribution of sensors at different positions; \(v(t)\) is the velocity. The movement, extrusion, deformation, etc. of the surrounding rock are dynamic. As a dynamic variable, the velocity can reflect the displacement velocity during the construction process; \(\sigma(t)\) is the stress, which reflects the internal reaction of the tunnel surrounding rock when subjected to external forces; \(\epsilon(t)\) is the stress, which can intuitively reflect the deformation degree of the surrounding rock; \(T(t)\) is the temperature. Temperature changes will affect the strength and elastic modulus of the surrounding rock. When the temperature is high, the brittleness of the rock will decrease, resulting in it being prone to plastic deformation or fragmentation. Temperature also affects the thermal expansion and stress distribution of materials; \(H(t)\) is the humidity. Humidity changes have a particularly significant impact on certain types of surrounding rocks (such as clay rocks, sandstones, etc.). An increase in humidity may cause the surrounding rock to expand, soften or rupture, affecting the mechanical properties of the surrounding rock.

[0076] The method adopted in this application is implemented based on the state equation and the observation equation. The state equation describes the dynamic changes of the system and is set according to the dynamic model, expressed as;

[0077] \(x(t + 1)=A\times x(t)+B\times u(t)+w(t)\)

[0078] Among them, \(x(t + 1)\) is the state vector at time \(t + 1\); \(A\) is the state transition matrix, which describes how the system transfers from the previous time \(t\) to the current time \(t + 1\); \(B\) is the control input matrix, which describes how the external control input \(u(t)\) affects the system state and can be an externally applied force or other control parameters; \(w(t)\) is the process noise, which reflects the changes in the system during the state transition process.

[0079] The observation equation describes the relationship between the observed value \(y(t)\) measured by the sensor and the true state, expressed as;

[0080] \(y(t)=H(t)\times x(t)+v(t)\)

[0081] Among them, \(y(t)\) is the observed value of the sensor at time \(t\), usually including measurement data such as stress, strain, pressure, etc.; \(H\) is the observation matrix, which represents how to obtain the observed value \(y(t)\) from the true state \(x(t)\) of the system; \(v(t)\) is the observation noise, which represents the measurement error of the sensor.

[0082] In the embodiment of this application, the system model connects the internal dynamics and external observations of the system through the state equation and the observation equation, which is the basis for the filtering algorithm to perform state estimation. The Kalman filter predicts and updates the state of the system in a recursive manner. Whenever new observation data arrives, the Kalman filter updates according to the difference between the predicted value and the observed value to optimize the system state estimation.

[0083] The state equation is used to predict the state of the system at the next moment, and the specific steps include:

[0084] State prediction, expressed as:

[0085] x - (t + 1) = A × x(t) + B × u(t)

[0086] where x - (t + 1) is the prior estimate of the state at the next moment, that is, the estimate without fusing new observation data; x(t) is the state vector at the current moment.

[0087] Covariance prediction, expressed as:

[0088] P - (t + 1) = A × P(t) × A T + Q(t)

[0089] where P(t) is the estimated error covariance matrix at the current moment; P - (t + 1) is the prior estimated error covariance matrix at the next moment, that is, the predicted estimated error covariance matrix, representing the uncertainty of the state prediction at the next moment; Q(t) is the process noise covariance matrix.

[0090] After the prediction step, the prior estimate value x - (t + 1) of the state at the next moment and its uncertainty description P - (t + 1) are corrected based on the new observation data.

[0091] The observation equation is used to update the state estimate of the system, combine the new observation data to correct the prediction result, and the specific steps include:

[0092] Obtain new observation data. When the moment t + 1 arrives, the sensor will generate a set of new observation values y(t + 1), and this observation includes data such as strain, pressure, displacement, and temperature measured by the sensor.

[0093] At this time, based on the filtering model (EKF, UKF or different physical models), the state is estimated, and according to the prediction error of each model for the current observation value, the model probability P i (t) is dynamically calculated, expressed as:

[0094]

[0095] where η is the normalization factor; σ i is the noise standard deviation of the i-th model; H i is the observation matrix of the i-th model, describing the mapping relationship from the state vector to the observed quantity; P i$(t)$ is the probability that the $i$-th model is selected at time $t$, which is used to quantify the credibility of the filtering model that is currently most likely to be applicable to the system state; $P$ i $(t + 1)$ is the probability that the $i$-th model is selected at time $t + 1$. The probability of each model is updated through a recursive formula to reflect the impact of new observed data on the model applicability.

[0096] Based on the model probabilities, select the model $i$ with the highest probability * as the filtering model at time $t + 1$, denoted as:

[0097]

[0098] Dynamically select the most suitable filtering model according to real-time observed data to improve the adaptability and estimation accuracy of the system in different dynamic environments. When facing complex conditions such as non-linearity and noise variation, it can automatically adjust the filtering strategy to enhance the robustness of the system.

[0099] Selected model $i$ * After that, use the observation matrix of this model and the observed value $y(t + 1)$ to update the state estimate, denoted as:

[0100]

[0101] Among them, $K(t + 1)$ is the Kalman gain matrix, which is used to adjust the weights of the predicted state and the observed data, and determines the contribution ratio of the predicted value and the actual observed value in the target state estimation. The higher the gain, the more dependent on the observed data, and the lower the gain, the more dependent on the predicted value; is the observation matrix corresponding to the selected model $i$ * , which maps the system state vector $x(t)$ to the observation space and is used to calculate the predicted observed value; $R(t + 1)$ is the observation noise covariance matrix at time $t + 1$, which describes the characteristics of the sensor measurement data error.

[0102] Based on the Kalman gain matrix and the observation residual, correct the prior state estimate $x$ - $(t + 1)$ to obtain the posterior state estimate $x(t + 1)$, denoted as:

[0103]

[0104] In the formula, is the observation residual, which represents the difference between the actual observed value and the predicted observed value; this expression feeds back the residual of the sensor's true measurement into the state, so as to obtain a posterior estimate $x(t + 1)$ closer to the real situation, that is, the state estimate vector after sequential filtering.

[0105] Based on the Kalman gain matrix, update the state estimation error covariance matrix, denoted as:

[0106]

[0107] Among them, P(t + 1) is the posterior state estimation error covariance matrix, which represents the uncertainty of the system state estimation. Through the state update step, this matrix reflects the new uncertainty level after combining the observation data; is the identity matrix.

[0108] In some instances, it also includes:

[0109] In the data fusion process, the spatial covariance matrix and Bayesian fusion are used to combine the spatial distribution information of multiple sensors to improve the accuracy and reliability of target state estimation, expressed as:

[0110] x fused (t + 1) = S(t + 1) × [S(t + 1) + R(t + 1)] - 1 × x(t + 1)

[0111] S = C·D·C T

[0112] Among them, S(t + 1) is the spatial covariance matrix, which describes the spatial correlation of multiple sensors; C is the sensor position matrix, which records the coordinate positions of all sensors in space; D is the distance correlation function matrix, which describes the function of the distance correlation between sensors; x(t + 1) is the state estimation vector after time series filtering, which is the state estimation at time t + 1 obtained through time series filtering, and the time series filtering is Kalman filtering; x fused (t + 1) is the fused state estimation vector, which is the target state estimation obtained through Bayesian fusion by combining the spatial covariance matrix and the observation noise covariance.

[0113] After completing the above steps, x fused (t + 1) is the optimal state estimation obtained after multi-model selection, time series Kalman filtering, adaptive noise adjustment, and spatial Bayesian fusion at time t + 1. x fused (t + 1) contains the best estimates and fusion results of each state component for the actual measured values of the sensors, that is, the output of the target data.

[0114] In some instances, it also includes:

[0115] In the Kalman filter algorithm, adaptive noise adjustment is used to dynamically update the process noise covariance matrix Q(t) and the observation noise covariance matrix R(t). This enables the filter to adjust the noise level according to the latest data residuals, making it more flexible and accurate when facing changing construction environments. The data residuals are the differences between the predictions and the actual observations.

[0116] The formula for adaptive noise adjustment, expressed as:

[0117] Q(t + 1) = α × Q(t) + (1 - α)·Cov(w(t))

[0118] R(t + 1) = β × R(t) + (1 - β)·Cov(v(t))

[0119] Where, Q(t + 1) is the updated process noise covariance matrix at the next time t + 1; Q(t) is the process noise covariance matrix at the current time t; R(t + 1) is the updated measurement noise covariance matrix at the next time t + 1; R(t) is the measurement noise covariance matrix at the current time t; α and β are smoothing coefficients used to control the weights of old and new information. The larger the smoothing coefficient, the more the old and new information is combined, and the slower the covariance changes. The smaller the smoothing coefficient, the faster the newly calculated covariance changes, and the more responsive the filter is to new information; Cov(w(t)) is the covariance estimate of the process noise; Conv(v(t)) is the covariance estimate of the measurement noise.

[0120] In the above content, for different types of dynamic changes during tunnel construction, the extended Kalman filter is used to process linear and weakly nonlinear systems, while the unscented Kalman filter is used for strongly nonlinear systems. By dynamically switching the filtering model, the adaptability and accuracy of filtering are improved. According to the real-time changes in the construction environment, the process noise covariance matrix Q and the measurement noise covariance matrix R are adaptively adjusted to more accurately reflect the uncertainty of the system and measurement. The spatial correlation between sensors is introduced, and by constructing the spatial covariance matrix S, the correlation between the measurement values of different sensors is described, enhancing the accuracy of data fusion. Using Bayesian theory, the observed values from different sensors are probabilistically fused, comprehensively considering the measurement accuracy and correlation of each sensor to improve the reliability of the overall data estimation.

[0121] In some examples, based on the above optimized state data, fragmentation degree information is generated, including:

[0122] Exemplarily, the optimized real-time sensor data is used to preliminarily evaluate the fragmentation degree of the tunnel surrounding rock. Since the tunnel surrounding rock is a complex inhomogeneous material and its fragmentation degree is affected by factors such as stress, strain, and plastic deformation, this step adopts a fragmentation evaluation model based on the finite element method.

[0123] Simulating the mechanical behavior of surrounding rock by the finite element method can describe in detail the changes in physical quantities such as stress and strain that the surrounding rock undergoes during tunnel excavation. These physical quantities provide the basis for calculating the subsequent degree of fragmentation. With the help of sensors deployed inside the tunnel, stress, strain, pressure and other data are obtained in real time. Based on these data, combined with the physical properties and structure of the tunnel surrounding rock, the degree of fragmentation of the surrounding rock is evaluated. In order to accurately evaluate the degree of fragmentation of the surrounding rock, a multi-dimensional fragmentation evaluation model is proposed. This model synthesizes factors such as stress, strain and plastic deformation, and non-linearly corrects the degree of fragmentation through the Sigmoid function.

[0124] The optimized target data is input into the fragmentation evaluation model, and the degree of fragmentation is output for evaluating the fragmentation situation of the surrounding rock. The expression is:

[0125]

[0126] where F 破碎 (t) is the fragmentation degree information, indicating the fragmentation degree of the surrounding rock at time t; σ(x, t) is the stress at position x at time t; ∈(x, t) is the strain at position x at time t; θ(x, t) is the plastic index, which is deduced from the mutual relationship between stress σ(t) and strain ∈(t), obtained based on the Von Mises criterion and the plastic flow criterion, and characterizes the plastic deformation degree of the surrounding rock. The larger the value, the stronger the plastic deformation of the surrounding rock; in the Sigmoid function and φ control the steepness and starting point of this non-linear function; L is the total length of the tunnel, and the integration range covers the influence of the entire tunnel surrounding rock.

[0127] In the above content, the stress term σ(x, t) and the strain term ∈(x, t) are determined based on the output of the optimized target data. The stress term σ(x, t) describes the change of the stress field of the tunnel surrounding rock with the excavation progress. During tunnel excavation, the surrounding rock will be affected by blasting, support and self-weight and other aspects, generating stress. If the stress is too large, it may cause the surrounding rock to break. The strain term ∈(x, t) of the surrounding rock reflects the deformation degree of the surrounding rock under the action of stress. Larger strain usually accompanies a greater possibility of damage. The plastic deformation behavior of the surrounding rock when it is subjected to external forces. During tunnel excavation, the surrounding rock first undergoes elastic deformation. When the stress exceeds a certain critical value, plastic deformation will occur, resulting in the rupture of the surrounding rock. When the strain ∈(x, t) is large, the non-linear correction part of the fragmentation degree information is realized through the Sigmoid function.

[0128] The form of the Sigmoid function is:

[0129]

[0130] The function approaches zero when ∈(x, t) is small, indicating that the surrounding rock has not been broken; while when ∈(x, t) increases, the value of the function increases rapidly, indicating that the degree of fragmentation of the surrounding rock also increases accordingly. The parameters and φ control the steepness and starting point of this non-linear function.

[0131] Through this model, the degree of fragmentation of the surrounding rock can be dynamically evaluated during the tunnel construction process; in practical applications, the specific steps for verifying and applying this model include:

[0132] Real-time stress and strain data of the surrounding rock are obtained through sensors installed in the tunnel. The data collected by the sensors are input into the above-mentioned fragmentation evaluation model to calculate the degree of fragmentation of the surrounding rock at different positions and time points. The parameters in the model are adjusted through experimental data or historical data to improve the prediction accuracy of the model. By monitoring the dynamic changes in the degree of fragmentation of the tunnel surrounding rock, construction personnel can timely adjust the construction method and excavation strategy, thereby reducing the occurrence of destructive fragmentation and ensuring the safety of tunnel construction.

[0133] This step can realize the dynamic evaluation of the degree of fragmentation of the surrounding rock by constructing a finite element model and combining real-time sensor data. The proposed degree of fragmentation information, through the comprehensive consideration of multiple physical quantities and combined with a non-linear correction function, enables the model to more accurately reflect the actual fragmentation situation, providing strong support to optimize the tunnel construction process and ensure construction safety.

[0134] In some instances, based on the optimized state data, spatial distribution information is generated, including:

[0135] Based on the optimized state data, a CNN is used to extract the characteristics of the spatial distribution of the surrounding rock to obtain spatial characteristics;

[0136] Based on the spatial characteristics, an RNN is used to capture the time variation to obtain the dynamic characteristics of the surrounding rock fragmentation;

[0137] Based on the dynamic characteristics of the surrounding rock fragmentation, the spatial distribution of the surrounding rock fragmentation is determined to generate spatial distribution information.

[0138] Exemplarily, after obtaining the sensor data and optimizing it through Kalman filtering, the optimized data is used to predict the spatial distribution of the surrounding rock fragmentation and is dynamically updated through deep learning algorithms. The ultimate goal is to timely adjust the construction plan based on these prediction results to reduce the risks brought by the surrounding rock fragmentation.

[0139] First, physical quantities such as strain, stress, temperature, and humidity of surrounding rock at different positions were collected through a sensor network, and Kalman filtering was used to denoise and optimize these data. Therefore, the optimized data should be a more accurate estimation of the surrounding rock state. The optimized data will be used as the input data in step 130, and a combined model of CNN and RNN will be used to extract the spatial features and dynamic characteristics of the time series of the surrounding rock, respectively, so as to predict the distribution of broken spaces.

[0140] First, use CNN to extract the features of the spatial distribution of the surrounding rock. The convolutional layer can extract features of local regions from the input data (optimized sensor data), helping to identify the spatial change trends of physical quantities such as stress and strain, expressed as:

[0141] Z = CNN(X) = ReLU(W 1 *X + b 1 )

[0142] where X is the input sensor data, including various physical quantities of the surrounding rock; W 1 is the weight matrix of the convolutional kernel, used to extract spatial features; b 1 is the bias term of the convolutional layer; * is the convolutional operation, usually a two-dimensional convolution, used to scan the input data spatially; ReLU(·) is the activation function, usually used to increase the non-linear expression ability of the model; Z is the output after the convolutional operation, representing the feature representation of the input data in space.

[0143] Use RNN to capture the dynamic changes of the data over time, especially the evolution process of the surrounding rock fragmentation over time. RNN can process time series data, thus revealing the time correlation in the fragmentation evolution process, expressed as:

[0144] h(t) = RNN(Z(tt), h(t - 1)) = tanh(W 2 ×Z(t) + U 2 ×h(t - 1) + -b 2 )

[0145] where Z(t) is the spatial feature at time t output from CNN; h(t - 1) is the hidden state at the previous moment, representing the memory of the fragmentation situation at the previous moment; W 2 is the weight matrix in RNN, used to process the hidden state of the input features; U 2 is the weight matrix in RNN, used to process the hidden state at the previous moment; b 2 is the bias term; tanh(·) is the activation function, increasing the non-linear expression ability of the model; h(t) is the hidden state at time t, representing the dynamic information of the fragmentation.

[0146] After being processed by CNN and RNN, the output is the spatial distribution of surrounding rock fragmentation, expressed as:

[0147] Y(t) = W 3 ×h(t) + b 3

[0148] Where Y(t) is the target prediction result, representing the spatial distribution of the surrounding rock fragmentation degree at time t; W 3 is the weight matrix of the output layer, used to map the RNN hidden state h(t) to the predicted value of the fragmentation distribution; b 3 is the bias term of the output layer.

[0149] Through repeated training and prediction, this model can update the spatial distribution of surrounding rock fragmentation in real time according to the input sensor data. When new sensor data arrives, CNN and RNN will continuously update the prediction results of fragmentation based on this data. This process is dynamically updated, that is, as new data enters, the model will continuously adjust and optimize the prediction results.

[0150] In some instances, the above-mentioned reinforcement learning algorithm is adopted to adjust the construction strategy based on the above-mentioned fragmentation degree information and the above-mentioned spatial distribution information, including:

[0151] Define the action space and state space of the construction strategy;

[0152] Evaluate the current construction strategy based on the predicted above-mentioned fragmentation degree information and the above-mentioned spatial distribution information;

[0153] Optimize the construction parameters through the reinforcement learning algorithm to minimize the risk of surrounding rock fragmentation;

[0154] Provide real-time feedback on the optimization results and adjust the above-mentioned construction strategy.

[0155] Exemplarily, it is first necessary to define the state space, action space and reward function of reinforcement learning. The state space consists of the fragmentation degree information and the spatial distribution information, comprehensively reflecting the current fragmentation condition and spatial distribution characteristics of the surrounding rock. These information include key parameters such as stress, strain, temperature, humidity, etc. of the surrounding rock at different positions and time points. The action space covers all adjustable construction parameters, such as increasing or decreasing the blasting intensity, adjusting the excavation speed, and configuring the support equipment, etc. The design of the reward function is crucial. It is based on multiple indicators such as the risk of surrounding rock fragmentation, construction efficiency and environmental factors, aiming to maximize the long-term return, that is, to improve the overall construction effect by reducing the fragmentation risk and optimizing the construction progress.

[0156] Reinforcement learning algorithms interact with the environment continuously, evaluate the effects of each action in a specific state, and update the policy to select the optimal action. For example, when it is detected that the degree of surrounding rock fragmentation in a certain area has increased significantly, the algorithm may choose to reduce the blasting intensity or slow down the excavation speed to avoid further damage; conversely, when the surrounding rock state is stable or improved, the construction progress may be accelerated to improve efficiency. Through this adaptive policy adjustment, the system can respond to changes in the construction environment in real time, significantly reduce construction risks, and improve construction safety and efficiency. In addition, reinforcement learning algorithms have a high degree of self-optimization ability, and can continuously learn and improve construction strategies in complex and changeable construction environments to ensure the robustness and flexibility of the tunnel construction process.

[0157] The state space represents the "state" of the agent at each moment, usually including various information related to the construction process, such as the current severity of surrounding rock fragmentation, construction parameters, etc. The state vector s(t) can be defined as the parameters including the surrounding rock fragmentation situation and the construction environment, expressed as:

[0158]

[0159] where, F 破碎 (t) is the overall fragmentation degree, representing the severity of surrounding rock fragmentation at time t; Y(t) is the spatial distribution of surrounding rock fragmentation, representing the fragmentation situation of surrounding rock in space; construction parameters include blasting intensity, excavation speed, equipment configuration, etc.

[0160] The action space represents the behaviors that the agent can take. In tunnel construction, common actions include adjusting the blasting intensity a 爆破 (t), adjusting the excavation speed a 开挖 (t), changing the construction path a 路径 (t), etc., expressed as:

[0161]

[0162] The reward function is the most crucial part in reinforcement learning, which is used to measure the effect of each action. According to the surrounding rock fragmentation situation, a reward function is designed with the goal of maximizing the long-term return, that is, reducing the risk of surrounding rock fragmentation.

[0163] The reward function r(t) is expressed as:

[0164] r(t) = -ι × R 破碎 (t) - κ × R 施工 (t) + γ × R 环境 ((t)

[0165] Among them, \(l\) is the weight factor of the fragmentation risk, which determines the impact of the surrounding rock fragmentation degree on the reward. A larger \(\iota\) value means that the risk of surrounding rock fragmentation will have a greater impact on the reward function, and the goal is to minimize the risk of surrounding rock fragmentation; \(\kappa\) is the weight factor of the construction risk, which controls the impact of construction parameters during the construction process on the reward. A larger \(\kappa\) value indicates that the construction parameters have a greater impact on the reward, and the system will tend to avoid the risks caused by excessive construction; \(\gamma\) is the weight factor of environmental factors, which measures the impact of environmental factors on the reward. A larger \(\gamma\) value means that environmental factors have a greater impact on the optimization of the construction plan, and the system needs to adjust the construction strategy according to external factors such as temperature and humidity.

[0166] R 破碎 (t) is the fragmentation risk calculated based on the fragmentation degree \(F\) 破碎 (t) and the spatial distribution \(Y(t)\). The goal is to minimize the risk of surrounding rock fragmentation, which is expressed as:

[0167]

[0168] Among them, \(Y(t)\) i is the fragmentation degree at the \(i\)-th position in space; \(\lambda\) is the weight factor, which controls the impact of the spatial distribution on the risk assessment. A larger \(\lambda\) value indicates that the spatial distribution has a greater impact on the fragmentation risk.

[0169] R 施工 (t) is the risk based on construction parameters. For example, too fast excavation speed may lead to a greater fragmentation risk, and too large blasting force will cause vibrations and affect the surrounding environment, which is expressed as:

[0170] R 施工 (t)=\(\gamma\) 1 ×|a 开挖 (t)|+\(\gamma\) 2 ×|a 爆破 (t)|

[0171] Among them, \(\gamma\) 1 and \(\gamma\) 2 are the weight factors for adjusting the impact of excavation speed and blasting force on the construction risk.

[0172] R 环境 (t) is the risk considering environmental factors, such as the impact of humidity, temperature, etc. on the surrounding rock fragmentation, which is expressed as:

[0173] R 环境 (t)=\(\eta\times(T(t)+H(t))\)

[0174] Among them, \(T(t)\) is the temperature; \(H(t)\) is the humidity; \(\eta\) is the weight factor for adjusting the impact of environmental factors.

[0175] Q-learning continuously optimizes the strategy by recursively updating the Q-value (the expected return of the state-action pair). The Q-value update formula is expressed as:

[0176]

[0177] where Q(s(t), a(t)) is the Q-value of taking action a(t) in state s(t), representing the expected return; α l is the learning rate, which determines the influence of new experience on the Q-value; r(t) is the reward function, measuring the effect of the current action; max a(t+1) Q(s(t + 1), a(t + 1)) is the Q-value of the optimal action selected in the next state s(t + 1); γ z is the discount factor, representing the importance of future returns.

[0178] By continuously updating the Q-value, the construction plan optimization system can learn the optimal actions to take in different states. For example, when the fragmentation degree F 破碎 (t) of the surrounding rock is relatively high, the system may choose to reduce the blasting intensity or slow down the excavation speed to avoid more serious fragmentation; while when F 破碎 (t) is relatively low, the system may accelerate the construction progress. During the construction process, the construction strategy will be continuously adjusted according to real-time feedback. If the surrounding rock in a certain area is severely fragmented, the Q-learning system will choose to reduce the excavation speed or adjust the blasting intensity according to this state, so as to avoid further damage to the surrounding rock.

[0179] Please refer to Figure 2 , which is a schematic structural diagram of a construction monitoring device for a mountain tunnel with fractured surrounding rock provided by an embodiment of the present application, including:

[0180] A surrounding rock data acquisition module 21 for acquiring the state data of the surrounding rock;

[0181] A state data optimization module 22 for optimizing the above state data by using the Kalman filter algorithm to obtain optimized state data;

[0182] A surrounding rock information generation module 23 for generating fragmentation degree information and spatial distribution information based on the above optimized state data;

[0183] A construction strategy adjustment module 24 for adjusting the construction strategy by using the reinforcement learning algorithm based on the above fragmentation degree information and the above spatial distribution information.

[0184] Please refer to Figure 3, The embodiment of the present application also provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored on the memory 310 and executable on the processor. When the processor 320 executes the computer program 311, the steps of any of the above methods for monitoring the construction of a fractured surrounding rock mountain tunnel are implemented.

[0185] Since the electronic device introduced in this embodiment is the device used to implement a device for monitoring the construction of a fractured surrounding rock mountain tunnel in the embodiment of the present application, based on the method introduced in the embodiment of the present application, those skilled in the art can understand the specific implementation manners and various forms of change of the electronic device in this embodiment. Therefore, the specific implementation of how this electronic device implements the method in the embodiment of the present application will not be described in detail here. As long as the device used by those skilled in the art to implement the method in the embodiment of the present application belongs to the scope protected by the present application.

[0186] In the specific implementation process, when the computer program 311 is executed by the processor, any implementation manner in the corresponding embodiment of the first aspect can be realized.

[0187] It should be noted that in the above embodiments, the descriptions of each embodiment have their own emphases. For the parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0188] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-readable program code.

[0189] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0190] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to operate in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction device that implements the functions specified in one or more of the processes Figure 1 or boxes. Figure 1 The functions specified in one or more of the boxes.

[0191] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, such that a series of operational steps are performed on the computer or other programmable apparatus to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one or more of the processes Figure 1 or boxes. Figure 1 The functions specified in one or more of the boxes.

[0192] Embodiments of the present application also provide a computer program product, which includes computer software instructions that, when running on a processing device, cause the processing device to execute Figure 1 the process of a method for monitoring the construction of a fractured surrounding rock mountain tunnel in a corresponding embodiment.

[0193] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions according to the embodiments of the present application are wholly or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from a website, computer, server, or data center to another website, computer, server, or data center in a wired manner (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or a wireless manner (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be stored by a computer or a data storage device such as a server or data center that includes one or more integrated available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)).

[0194] Those skilled in the art can clearly understand that for the sake of convenience and brevity of description, the specific working processes of the systems, apparatuses, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0195] In several embodiments provided by the present application, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of devices or units can be in electrical, mechanical, or other forms.

[0196] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0197] In addition, each functional unit in various embodiments of the present application can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0198] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in various embodiments of the present application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories, random access memories, magnetic disks, or optical discs that can store program codes.

[0199] The above embodiments are only used to illustrate the technical solutions of the present application and are not intended to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features. And these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of various embodiments of the present application.

[0200] Although the preferred embodiments of the present specification have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concepts. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications falling within the scope of the present specification.

[0201] Obviously, those skilled in the art can make various changes and deformations to the present specification without departing from the spirit and scope of the present specification. Thus, if these modifications and deformations of the present specification fall within the scope of the claims of the present specification and their equivalent technologies, the present specification is also intended to include these modifications and deformations.

Claims

1. A method for monitoring construction of a broken rock mountain tunnel, characterized in that: The method comprises: Obtaining state data of surrounding rock; The state data is optimized by using a Kalman filter algorithm to obtain optimized state data; Based on the optimized state data, generating fragmentation degree information and spatial distribution information; A reinforcement learning algorithm is used to adjust the construction strategy based on the fragmentation degree information and the spatial distribution information.

2. The method for monitoring construction of a broken rock mountain tunnel according to claim 1 is characterized in that: The step of optimizing the state data by using a Kalman filter algorithm to obtain optimized state data includes: constructing a state vector based on the state data; Using a state transfer matrix to predict the state vector and the estimation error covariance matrix, to obtain a priori state estimation and a priori estimation error covariance matrix; Based on the new observation data, determine the filtering model for the next moment; Calculating a Kalman gain matrix based on the observation matrix of the filtering model, the prior estimation error covariance matrix and the observation noise covariance matrix; Based on the Kalman gain matrix and the observation residual, the prior state estimate is corrected to obtain a posterior state estimate; The a posteriori state estimate is used as optimized state data.

3. The method for monitoring construction of a broken rock mountain tunnel according to claim 2 is characterized in that: Based on the new observation data, determine the filtering model for the next moment, including: Based on the prediction error of the filtering model for the new observation data, the model probability P is calculated. i (t), expressed as: Among them, η is the normalization factor; σ i is the noise standard deviation of the ith model; H i is the observation matrix of the i-th model; P i (t) is the probability of the i-th model being selected at time t; y(t+1) is the new observation data; is the prior state estimate of the ith model; P i (t+1) is the probability of the i-th model being selected at time i+1; Based on the model probability, the model i* with the highest probability is selected as the filtering model at time i+1.

4. The method for monitoring construction of a broken rock mountain tunnel according to claim 2 is characterized in that: Also includes: Based on the Kalman gain matrix, the prior estimation error covariance matrix is ​​updated to obtain the posterior estimation error covariance matrix, which is used to update the state at the next moment.

5. The method for monitoring construction of a broken rock mountain tunnel according to claim 1, characterized in that: The generating of fragmentation degree information based on the optimization state data comprises: The fragmentation degree information is determined by the following formula: Among them, F 破碎 (t) is the degree of crushing information, the degree of crushing of the surrounding rock at time t; σ(x, t) is the stress at position x at time t; ∈(x, t) is the strain at position x at time t; θ(x, t) is the plasticity index; is the steepness of the nonlinear function; φ is the starting point of the nonlinear function; L is the total length of the tunnel.

6. The method for monitoring construction of a broken rock mountain tunnel according to claim 1, characterized in that: The generating of spatial distribution information based on the optimization state data includes: Based on the optimized state data, a convolutional neural network is used to extract the characteristics of the spatial distribution of the surrounding rock to obtain spatial characteristics; Based on the spatial characteristics, a recurrent neural network is used to capture the temporal changes and obtain the dynamic characteristics of surrounding rock crushing; Based on the dynamic characteristics of surrounding rock crushing, the spatial distribution of surrounding rock crushing is determined to generate spatial distribution information.

7. The method for monitoring construction of a broken rock mountain tunnel according to claim 1, characterized in that: The adopting of a reinforcement learning algorithm to adjust the construction strategy based on the fragmentation degree information and the spatial distribution information includes: Define the action space and state space of the construction strategy; Based on the predicted fragmentation degree information and the spatial distribution information, evaluating the current construction strategy; Optimize construction parameters through reinforcement learning algorithms to minimize the risk of surrounding rock crushing; The optimization results are fed back in real time to adjust the construction strategy.

8. A monitoring device for construction of a broken rock mountain tunnel, characterized in that: include: A surrounding rock data acquisition module is used to acquire state data of surrounding rocks; A state data optimization module, used to optimize the state data using a Kalman filter algorithm to obtain optimized state data; A surrounding rock information generation module, used to generate fragmentation degree information and spatial distribution information based on the optimization state data; The construction strategy adjustment module is used to adjust the construction strategy based on the fragmentation degree information and the spatial distribution information by using a reinforcement learning algorithm.

9. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is used to implement the steps of the method for monitoring construction of a broken rock mountain tunnel as described in any one of claims 1 to 7 when executing the computer program stored in the memory.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for monitoring construction of a broken rock mountain tunnel as described in any one of claims 1 to 7 is implemented.

Images