Karst tunnel collapse intelligent early warning method, system and equipment based on microseismic multi-precursor characteristics

Abstract

The present application relates to the technical field of karst tunnel collapse early warning, and particularly relates to a karst tunnel collapse intelligent early warning method, system and equipment based on microseismic multi-precursor characteristics. The technical scheme comprises multi-scale microseismic monitoring network construction, multi-modal microseismic feature processing, quantum neural network early warning model building, self-adaptive early warning and risk assessment. The present application collects data in all directions through a multi-scale microseismic monitoring network, uses different depth sensors and intelligent transmission modes to make the monitoring accurate and stable, combines multi-modal feature extraction and fusion to mine multi-dimensional information in the time domain, frequency domain and waveform, accurately judges the surrounding rock condition, uses quantum neural network to strengthen early warning analysis by quantum characteristics, and further adjusts the self-adaptive threshold and multi-factor risk assessment to accurately warn and efficiently prevent collapse, thereby effectively ensuring the safety of tunnel construction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of karst tunnel collapse early warning, and in particular to a karst tunnel collapse intelligent early warning method, system and device based on microseismic multi-precursor characteristics. BACKGROUND

[0002] With the rapid development of tunnel engineering construction, collapse, as a common and serious geological disaster in tunnel construction, has always been a key concern in the engineering field. In karst tunnels, due to the special geological structure, such as karst caves and dissolution fissures, the stability of surrounding rock is more complex and variable, and the risk of collapse is further increased.

[0003] Traditional collapse early warning methods rely mainly on deformation monitoring of surrounding rock. However, for hard and brittle surrounding rock, the deformation before collapse is often very small, which makes this deformation monitoring-based early warning method ineffective in many cases. The emergence of microseismic technology provides a new way for collapse early warning. Microseismic is a low-energy elastic wave or stress wave that accompanies the process of crack generation, expansion and penetration in rock mass. By collecting microseismic signals through microseismic equipment, the whole process from micro-fracture to macro-instability of potential collapse body can be tracked, thereby achieving early warning.

[0004] However, there are many technical difficulties in using microseismic multi-precursor characteristics for karst tunnel collapse early warning. One of the prominent problems is the effective processing and comprehensive analysis of microseismic multi-characteristics parameters. Microseismic signals contain a wealth of information, such as source location, magnitude, frequency, waveform characteristics and other characteristic parameters. These parameters are interrelated and have a high degree of non-linear relationship. Current analysis methods cannot fully exploit the internal relationship between these parameters, and cannot accurately extract key feature information closely related to collapse from complex microseismic data. For example, under different karst geological conditions, the variation of microseismic characteristic parameters is different. How to accurately identify these differences and establish an adaptive early warning model is a major challenge.

[0005] In addition, in actual engineering, the mechanical properties of tunnel surrounding rock and construction disturbance factors will affect the microseismic characteristics, resulting in a large amount of noise interference in microseismic data, further increasing the difficulty of accurately extracting effective early warning information. Existing data processing techniques, while removing noise, often fail to preserve the integrity and authenticity of microseismic characteristic parameters, resulting in a significant reduction in the accuracy of early warning.

[0006] In view of the above, the present application proposes a karst tunnel collapse intelligent early warning method, system and device based on microseismic multi-precursor characteristics. SUMMARY

[0007] The purpose of the present application is to solve the problem of low accuracy of karst tunnel collapse early warning in the background art, and propose a karst tunnel collapse intelligent early warning method, system and equipment based on microseismic multi-precursor characteristics.

[0008] In a first aspect, the present application provides a karst tunnel collapse intelligent early warning method based on microseismic multi-precursor characteristics, comprising the following steps:

[0009] Multi-scale microseismic monitoring network construction: a multi-scale monitoring network containing different types of sensors in shallow, middle and deep layers and supplemented with arrays in key positions is constructed according to the geological conditions of the karst tunnel, and data is transmitted through a wireless ad hoc network system with mixed communication of low-power Bluetooth and ZigBee;

[0010] Multi-modal microseismic feature processing: for the collected microseismic signals, time domain feature analysis, short-time Fourier transform frequency domain analysis, Hilbert-Huang transform to construct time-frequency joint feature matrix, and waveform morphology feature mining are performed, and then multi-modal deep belief network is used to deeply fuse the time domain-frequency domain joint features and waveform morphology features;

[0011] Quantum neural network early warning model construction: quantum bits are used to encode the fused microseismic multi-modal features to construct a quantum feature vector, a quantum neural network with a quantum entanglement gate and a quantum rotation gate to construct a hidden layer structure is constructed, the quantum feature vector is used as input and the collapse early warning result is used as output, the model is trained using quantum backpropagation algorithm, and the intelligent early warning function based on quantum computing is achieved;

[0012] Adaptive early warning and risk assessment: Kalman filter algorithm is used to filter and predict the microseismic feature data and dynamically adjust the early warning threshold, and at the same time, the threshold is corrected by a fuzzy logic reasoning system combined with construction progress, support condition and geological condition change factors; the weights of microseismic features, geological intensity index, groundwater level change and construction blasting vibration are determined by analytic hierarchy process, and the tunnel collapse risk is comprehensively evaluated by fuzzy comprehensive evaluation method.

[0013] Optionally, the multi-scale microseismic monitoring network construction specifically comprises the following steps:

[0014] According to the geological prediction model of the karst tunnel, the surrounding rock of the tunnel is divided into three monitoring levels of shallow, middle and deep layers, high-resolution piezoelectric sensors are used in the shallow layer, and the interval is 3-5 meters and is distributed in a ring around the tunnel wall;

[0015] The middle layer uses electromagnetic induction type microseismic sensors arranged in a quincunx shape with an interval of 5-8 meters; the deep layer arranges fiber Bragg grating microseismic sensors, and arranges a monitoring point every 10-15 meters along the tunnel axis;

[0016] At the same time, in the karst development intensive area and the key part of the fault fracture zone, additional micro acceleration sensor array is added;

[0017] A wireless ad hoc network transmission system using low-power Bluetooth and ZigBee hybrid communication technology is constructed, and the communication mode is intelligently switched according to the distance between the sensor node and the data processing center and the signal interference condition to transmit the sensor data to the data processing center.

[0018] The sensitivity of the shallow piezoelectric sensor is as high as (unit: ), and the sensor spacing is set to (m) according to the tunnel radius and the surrounding rock characteristics;

[0019] The spacing of the middle electromagnetic induction type microseismic sensor is (m);

[0020] The spacing of the deep optical fiber grating microseismic sensor is (m).

[0021] Optionally, the multi-modal microseismic feature extraction and fusion specifically includes the following steps: time domain feature analysis is performed on the collected microseismic signals, peak amplitude, duration, rise time and pulse count parameters are calculated, frequency domain analysis is performed using short-time Fourier transform to obtain a frequency spectrum and extract frequency domain features such as dominant frequency, frequency bandwidth and spectral centroid, intrinsic mode functions are obtained by Hilbert-Huang transform decomposition of the microseismic signals, and instantaneous frequency and instantaneous amplitude are calculated to construct a time-frequency joint feature matrix, geometric feature parameters such as concave-convexness, symmetry and steepness of the microseismic waveform are analyzed in depth, and multi-modal deep belief networks are used to fuse the time domain-frequency domain joint features and waveform shape features.

[0022] In time domain feature analysis, the peak amplitude , the duration , wherein is the time when the microseismic signal first exceeds the threshold value determined according to the background noise, is the time when the microseismic signal last exceeds the threshold value, the rise time , wherein is the time when the microseismic signal reaches the peak value, and the pulse count is

[0023]

[0024] , wherein is a function for judging the pulse of the microseismic signal, when the signal slope change rate exceeds , , otherwise , The number of signal sampling points;

[0025] The window function of short-time Fourier transform is The window length is Then:

[0026]

[0027] Wherein, is time, is frequency;

[0028] Main frequency:

[0029]

[0030] Bandwidth:

[0031]

[0032] Wherein, and The energy in the spectrum is greater than a certain threshold The maximum and minimum frequency determined according to the energy distribution of the signal, the spectrum center of gravity:

[0033] .

[0034] Optionally, in the multi-modal microseismic feature extraction and fusion, the intrinsic mode function is obtained by Hilbert-Huang transform, and the instantaneous frequency is:

[0035]

[0036] Wherein, is the intrinsic mode function;

[0037] Instantaneous amplitude:

[0038]

[0039] When the waveform morphology feature is mined, the concave-convex nature The positive and negative judgment of waveform concave-convex nature, if is concave, is convex, symmetry:

[0040]

[0041] Wherein, is the midpoint time of the waveform, is the half-waveform time width, the steepness The change rate of steepness evaluates the steepness, and calculates The slope change near the peak value;

[0042] Multimodal deep belief networks are composed of multiple restricted Boltzmann machines stacked together. The number of visible layer cells in a layered RBM is The number of hidden layer units is The visible layer state vector is The hidden layer state vector is The energy function is:

[0043]

[0044] in, For connection weights, For visible layer bias, For hidden layer bias;

[0045] Parameter update formula:

[0046]

[0047]

[0048]

[0049] in, This is the learning rate.

[0050] Optionally, the construction of the intelligent early warning model based on quantum neural network includes the following steps: using qubits to encode the fused microseismic multimodal features to construct a quantum feature vector, constructing a quantum neural network, whose input layer is the quantum-encoded microseismic feature vector, the hidden layer uses quantum entanglement gates and quantum rotation gates to construct a complex quantum computing structure, the output layer is the collapse early warning result, and the quantum backpropagation algorithm is used to train the quantum neural network;

[0051] In the construction of the intelligent early warning model based on quantum neural networks, the quantum bit encoding formula is as follows:

[0052]

[0053]

[0054] in, , It is a quantum state. These are elements in the microseismic feature vector. This encoding method maps the microseismic feature vector to the quantum state space, and utilizes the superposition and entanglement of quantum states to explore the potential correlation between microseismic features.

[0055] In the construction of the intelligent early warning model based on quantum neural networks, the quantum neural network employs a quantum rotating gate:

[0056]

[0057] The loss function adopts mean square error:

[0058]

[0059] wherein, is the sample number, is the predicted output, is the actual output.

[0060] Optionally, the adaptive early warning and risk assessment specifically comprises the following steps: real-time filtering and prediction of microseismic characteristic data by using Kalman filtering algorithm, dynamic adjustment of the early warning threshold according to the change trend and uncertainty of the predicted value, correction of the early warning threshold by a fuzzy logic reasoning system in combination with the construction progress, support condition and geological condition change factors in the tunnel; construction of a multi-factor karst tunnel collapse risk assessment system, determination of the weights of microseismic characteristics, geological strength index of tunnel surrounding rock, underground water level change and construction blasting vibration factors by using analytic hierarchy process, and comprehensive evaluation of the collapse risk of the tunnel by using fuzzy comprehensive evaluation method.

[0061] In the adaptive early warning and risk assessment, the Kalman filtering algorithm is that the microseismic characteristic data vector is , the state vector is , the measurement matrix is , the state transition matrix is , the process noise covariance matrix is , the measurement noise covariance matrix is , the prediction step is: , ; the update step is: , , When the determinant value of the prediction error covariance matrix is less than , the early warning threshold , wherein is an adjustment coefficient, and is a predicted value change amount.

[0062] Optionally, in the adaptive early warning and risk assessment, when the analytic hierarchy process is used to determine the weight, the judgment matrix is set as , wherein, represents the importance of factor relative to factor , the product of the elements in each row is calculated:

[0063]

[0064] and then​ of nth root , and finally normalized to obtain ;

[0065] In the adaptive early warning and risk assessment, in the fuzzy comprehensive evaluation method, the evaluation set is , the single-factor fuzzy evaluation matrix is , and the fuzzy comprehensive evaluation result is , wherein is a fuzzy synthesis operator, and a weighted average operator is used.

[0066] In a second aspect, the present application provides a karst tunnel collapse intelligent early warning system based on microseismic multi-precursor characteristics, comprising:

[0067] A multi-scale microseismic monitoring module is composed of a plurality of microseismic sensors arranged in layers according to a karst tunnel geological prediction model, wherein high-resolution piezoelectric sensors are arranged in the shallow layer, with a spacing of 3-5 meters and a ring-shaped distribution around the tunnel wall; electromagnetic induction type microseismic sensors are arranged in the middle layer in a quincunx pattern with a spacing of 5-8 meters; optical fiber grating microseismic sensors are arranged in the deep layer, with a monitoring point every 10-15 meters along the tunnel axis, and a micro acceleration sensor array is arranged at key positions in karst development dense areas and fault fracture zones; the module further comprises a wireless ad hoc network transmission submodule using low-power Bluetooth and ZigBee hybrid communication technology, for transmitting data collected by the sensors to a data processing center;

[0068] A data processing and feature extraction module is used to process the collected microseismic signals, including time domain feature analysis of the microseismic signals to obtain peak amplitude, duration, rise time and pulse count, frequency domain analysis using short-time Fourier transform to obtain main frequency, frequency bandwidth, spectral centroid and other frequency domain features, decomposition of the microseismic signal using Hilbert-Huang transform to obtain the intrinsic mode function and calculate its instantaneous frequency and instantaneous amplitude to construct a time-frequency joint feature matrix, and analysis of the concave-convex, symmetry, steepness and other geometric feature parameters of the microseismic waveform, and fusion of the time domain-frequency domain joint features and waveform morphological features through a multi-modal deep belief network;

[0069] An intelligent early warning model module includes a quantum encoding unit and a quantum neural network unit, the quantum encoding unit encodes the fused microseismic multi-modal features using quantum bits to construct a quantum feature vector, and the quantum neural network unit takes the quantum-encoded microseismic feature vector as input, the hidden layer uses quantum entanglement gates and quantum rotation gates to construct a complex quantum computing structure, and the output layer outputs the collapse early warning result, and the module uses a quantum backpropagation algorithm to train the quantum neural network;

[0070] ​The early warning threshold adjustment and risk assessment module comprises a Kalman filtering submodule and a multi-factor risk assessment submodule; the Kalman filtering submodule adopts a Kalman filtering algorithm to perform real-time filtering and prediction on microseismic characteristic data, and dynamically adjusts the early warning threshold according to the change trend and uncertainty of the predicted value, while combining the construction progress, support condition and geological condition change factors in the tunnel to correct the early warning threshold through a fuzzy logic reasoning system; the multi-factor risk assessment submodule constructs a risk assessment system comprising microseismic characteristics, geological strength indexes of the tunnel surrounding rock, underground water level changes and construction blasting vibration factors, determines the weight of each factor by using an analytic hierarchy process, and comprehensively evaluates the tunnel collapse risk by using a fuzzy comprehensive evaluation method.

[0071] In a third aspect, the present application provides a karst tunnel collapse intelligent early warning device based on microseismic multi-precursor characteristics, comprising:

[0072] a memory for storing instructions; wherein the instructions are used to implement the karst tunnel collapse intelligent early warning method based on microseismic multi-precursor characteristics according to the first aspect;

[0073] a processor for executing the instructions in the memory.

[0074] Compared with the prior art, the present application has at least one of the following beneficial technical effects:

[0075] By arranging different types of sensors in the shallow layer, the middle layer and the deep layer of the tunnel surrounding rock, the overall monitoring of the state of the tunnel surrounding rock at different depths is realized, and the comprehensiveness and accuracy of the monitoring are improved;

[0076] The multi-dimensional characteristics can comprehensively reflect the characteristics of the microseismic event, providing rich information for accurately judging the state of the surrounding rock; the multi-modal deep belief network is used to fuse the microseismic characteristics of different modalities, which can learn the internal relationship and complex relationship between the characteristics, avoid the limitations of single characteristic analysis, improve the accuracy of comprehensive judgment of the microseismic event, and further more accurately evaluate the stability of the tunnel surrounding rock;

[0077] The fused microseismic multi-modal characteristics are encoded by using quantum bits, which can more effectively represent the microseismic characteristics and improve the analysis ability of the model for the microseismic signal;

[0078] The real-time filtering and prediction of the microseismic characteristic data based on the Kalman filtering algorithm improves the reliability of the early warning, and can comprehensively and objectively evaluate the tunnel collapse risk.

[0079] The present application collects data in all directions through a multi-scale microseismic monitoring network, uses different depth sensors and intelligent transmission modes, is accurate and stable in monitoring, combines multi-modal feature extraction and fusion, excavates time domain, frequency domain and waveform multidimensional information, accurately judges the surrounding rock condition, uses quantum characteristics of a quantum neural network to strengthen early warning analysis, and then matches adaptive threshold adjustment and multi-factor risk assessment, so as to accurately warn and efficiently prevent collapse, and effectively guarantee the safety of tunnel construction. BRIEF DESCRIPTION OF DRAWINGS

[0080] Figure 1 A flow chart of the intelligent early warning method for karst tunnel collapse based on microseismic multi-precursor characteristics. DETAILED DESCRIPTION

[0081] The technical solutions of the present application will be further described below in combination with the drawings and specific embodiments.

[0082] Embodiment 1

[0083] As shown in the drawings, Figure 1 The present application proposes an intelligent early warning method for karst tunnel collapse based on microseismic multi-precursor characteristics, which includes four steps of multi-scale microseismic monitoring network construction, multi-modal microseismic feature processing, quantum neural network early warning model building, and adaptive early warning and risk assessment. Each step will be described in detail below.

[0084] I. Multi-scale microseismic monitoring network construction: according to the geological conditions of karst tunnels, a multi-scale monitoring network containing different types of sensors in shallow, middle and deep layers and supplemented with arrays in key positions is constructed, and data is transmitted through a wireless ad hoc network system of mixed communication of low-power Bluetooth and ZigBee;

[0085] Specifically, the following steps are included:

[0086] According to the geological prediction model of the karst tunnel, the surrounding rock of the tunnel is divided into three monitoring levels of shallow, middle and deep layers. High-resolution piezoelectric sensors are used in the shallow layer, and are distributed in a ring shape around the tunnel wall with a spacing of 3-5 meters;

[0087] Electromagnetic induction type microseismic sensors are used in the middle layer and arranged in a plum blossom shape with a spacing of 5-8 meters; fiber Bragg grating microseismic sensors are arranged in the deep layer, and a monitoring point is arranged every 10-15 meters along the tunnel axis;

[0088] At the same time, micro acceleration sensor arrays are additionally arranged in key positions in areas with dense karst development and fault fracture zones;

[0089] A wireless ad hoc network transmission system using low-power Bluetooth and ZigBee mixed communication technology is constructed, and according to the distance between the sensor nodes and the data processing center and the signal interference condition, the communication mode is intelligently switched to transmit the sensor data to the data processing center;

[0090] The sensitivity of the shallow piezoelectric sensor is up to (unit: ), and the sensor spacing is According to the tunnel radius and the surrounding rock characteristics, it is set to (m);

[0091] The spacing of the middle electromagnetic induction type microseismic sensor is (m);

[0092] The spacing of the deep fiber grating microseismic sensor is (m).

[0093] By arranging different types of sensors in the shallow, middle and deep layers of the tunnel surrounding rock, the all-around monitoring of the surrounding rock state at different depths of the tunnel is realized. The shallow piezoelectric sensor can sensitively capture the small rock crack changes caused by construction disturbance and shallow karst cave; the middle electromagnetic induction sensor can monitor the stress changes of the surrounding rock in a larger range; the deep fiber grating sensor focuses on the structural changes of the deep rock mass.

[0094] In the intensive karst development area and the key parts of the fault fracture zone, a micro acceleration sensor array is added to supplement the monitoring of microseismic signals under complex geological conditions, further improving the comprehensiveness and accuracy of the monitoring, and not missing any potential hidden dangers that may cause landslides.

[0095] The wireless ad hoc network transmission system using low-power Bluetooth and ZigBee hybrid communication technology can intelligently switch communication modes according to the signal interference and transmission distance in the tunnel. This not only ensures that the sensor data can be stably and efficiently transmitted to the data processing center, but also reduces power consumption, prolongs the service life of the sensor, and ensures the long-term and reliable acquisition of monitoring data.

[0096] II. Multi-modal microseismic feature processing: For the collected microseismic signals, time domain feature analysis, short-time Fourier transform frequency domain analysis, Hilbert-Huang transform to construct time-frequency joint feature matrix, and waveform morphology feature mining are performed, and then multi-modal deep belief network is used to deeply fuse the time-frequency joint features and waveform morphology features. Multi-modal microseismic feature extraction and fusion includes the following steps: time domain feature analysis is performed on the collected microseismic signals, peak amplitude, duration, rise time and pulse count parameters are calculated, short-time Fourier transform is used for frequency domain analysis to obtain frequency spectrum and extract frequency domain features such as main frequency, frequency band width and spectrum center of gravity, Hilbert-Huang transform is used to decompose the microseismic signal to obtain intrinsic mode function and calculate its instantaneous frequency and instantaneous amplitude to construct time-frequency joint feature matrix, the concave-convex, symmetry, steepness and other geometric feature parameters of the microseismic waveform are analyzed in depth, and multi-modal deep belief network is used to fuse the time-frequency joint features and waveform morphology features.

[0097] In time-domain characteristic analysis, peak amplitude Duration ,in, For the first time the microseismic signal exceeds the threshold Time determined based on background noise The rise time is the time when the microseismic signal last exceeds the threshold. ,in, Pulse count is the time it takes for the microseismic signal to reach its peak value.

[0098]

[0099] in, To determine the function of the microseismic signal pulse, when the rate of change of the signal slope exceeds... hour ,otherwise , This represents the number of signal sampling points.

[0100] Short-time Fourier transform, let the window function be... The window length is ,but:

[0101]

[0102] in, It is time. It is frequency;

[0103] Clock speed:

[0104]

[0105] Bandwidth:

[0106]

[0107] in, and The energy in the spectrum is greater than a certain threshold. Based on the maximum and minimum frequencies determined by the signal energy distribution, the centroid of the spectrum is:

[0108]

[0109] The Hilbert-Huang transform yields the eigenmode functions and instantaneous frequencies.

[0110]

[0111] in, These are intrinsic modulo functions;

[0112] Instantaneous amplitude:

[0113]

[0114] Waveform morphology feature mining, concave-convex of the waveform concave-convex, if for concave, for convex, symmetry:

[0115]

[0116] wherein, is the midpoint time of the waveform, is the half-waveform time width, the steepness of the change rate evaluation steepness, calculation in the vicinity of the peak slope change;

[0117] The multi-modal deep belief network is stacked by multiple restricted Boltzmann machines, the number of visible layer units of the first layer RBM is , the number of hidden layer units is , the visible layer state vector is , the hidden layer state vector is , and the energy function is:

[0118]

[0119] wherein, is the connection weight, is the visible layer bias, is the hidden layer bias;

[0120] Parameter update formula:

[0121]

[0122]

[0123]

[0124] wherein, is the learning rate.

[0125] The microseismic signal is subjected to time domain-frequency domain joint feature extraction, including calculation of peak amplitude, duration, rise time, pulse count and other time domain features, and use of short-time Fourier transform and Hilbert-Huang transform to obtain frequency domain features and time-frequency joint feature matrix. Meanwhile, waveform morphology features such as concave-convex, symmetry and steepness are deeply mined. These multi-dimensional features can comprehensively reflect the characteristics of microseismic events, and provide rich information for accurately judging the surrounding rock state.

[0126] The multi-modal deep belief network is used for fusing microseismic features of different modes, which can learn the internal relationship and complex relationship between the features, avoid the limitation of single feature analysis, improve the accuracy of comprehensive judgment of microseismic events, and further accurately evaluate the stability of the tunnel surrounding rock.

[0127] III. Quantum neural network early warning model building: quantum bits are used to encode the fused microseismic multi-modal features to construct a quantum feature vector, a quantum neural network with quantum entanglement gates and quantum rotation gates is constructed to build a hidden layer structure, the quantum feature vector is used as input, and the collapse early warning result is used as output, the quantum back propagation algorithm is used to train the model to achieve intelligent early warning function based on quantum computing; the specific steps of building the intelligent early warning model based on quantum neural network include: using quantum bits to encode the fused microseismic multi-modal features to construct a quantum feature vector, building a quantum neural network, the input layer of which is the quantum encoded microseismic feature vector, the hidden layer adopts quantum entanglement gates and quantum rotation gates to build a complex quantum computing structure, and the output layer is the collapse early warning result, and the quantum back propagation algorithm is used to train the quantum neural network;

[0128] In the construction of the intelligent early warning model based on quantum neural network, the quantum bit encoding formula is:

[0129]

[0130]

[0131] wherein, , is a quantum state, is an element in the microseismic feature vector, which maps the microseismic feature vector to the quantum state space by this encoding method, and uses the superposition and entanglement of quantum states to mine the potential association between microseismic features;

[0132] In the construction of the intelligent early warning model based on quantum neural network, the quantum rotation gate used by the quantum neural network is:

[0133]

[0134] The loss function adopts mean square error:

[0135]

[0136] wherein, is the number of samples, is the predicted output, is the actual output.

[0137] The fused microseismic multi-modal features are encoded by quantum bits. With the superposition and entanglement of quantum states, the potential correlations and complex relationships between microseismic features can be explored, which can effectively represent microseismic features and improve the analysis ability of the model.

[0138] The construction and training of quantum neural networks can accelerate the training speed of the model by utilizing the parallel computing capability of quantum states, making it adapt to new data and environmental changes faster. Meanwhile, the quantum entanglement mechanism helps to enhance the learning ability of the model for the nonlinear relationships between microseismic features, thus more accurately predicting the risk of collapse.

[0139] Visualizing the output results of the model through quantum state tomography helps engineers understand the decision-making process of the early warning model, enabling them to take more targeted preventive measures and improve the scientificity of construction decisions.

[0140] Four, adaptive early warning and risk assessment: using Kalman filtering algorithm to filter and predict microseismic feature data and dynamically adjust the early warning threshold, and combining construction progress, support conditions and geological condition changes to correct the threshold through fuzzy logic reasoning system; using AHP to determine the weight of microseismic features, geological strength index, groundwater level change, construction blasting vibration; using fuzzy comprehensive evaluation method to comprehensively evaluate the risk of tunnel collapse.

[0141] Adaptive early warning and risk assessment specifically includes the following steps: using Kalman filtering algorithm to filter and predict microseismic feature data in real time, dynamically adjusting the early warning threshold according to the trend and uncertainty of the predicted value, and correcting the early warning threshold through fuzzy logic reasoning system by combining tunnel construction progress, support conditions and geological condition changes; building a multi-factor karst tunnel collapse risk assessment system, using AHP to determine the weight of microseismic features, tunnel surrounding rock geological strength index, groundwater level change, construction blasting vibration, and using fuzzy comprehensive evaluation method to comprehensively evaluate the risk of tunnel collapse;

[0142] In adaptive early warning and risk assessment, the microseismic feature data vector is , the state vector is , the measurement matrix is , the state transition matrix is , the process noise covariance matrix is , the measurement noise covariance matrix is , the prediction step is: , ; the update step is: , , When the prediction error covariance matrix The determinant value is less than At that time, the warning threshold ,in To adjust the coefficient, This represents the change in the predicted value.

[0143] In adaptive early warning and risk assessment, when determining weights using the analytic hierarchy process (AHP), a judgment matrix is ​​set. ,in, Indicator Factors Relative factors First calculate the importance. The product of the elements in each row:

[0144]

[0145] Recalculate of Root Finally, for Normalization process is performed to obtain .

[0146] In adaptive early warning and risk assessment, in the fuzzy comprehensive evaluation method, let the evaluation set be... The single-factor fuzzy evaluation matrix is Fuzzy comprehensive evaluation results ,in For fuzzy synthesis operators, a weighted average operator is used.

[0147] The system uses a Kalman filter algorithm to filter and predict microseismic characteristic data in real time, and dynamically adjusts the warning threshold based on the changing trend and uncertainty of the predicted values. This adaptive threshold adjustment mechanism can issue warning signals in a timely and accurate manner based on the actual microseismic activity within the tunnel, avoiding false alarms or missed alarms that may be caused by fixed thresholds, and improving the reliability of the warnings.

[0148] By constructing a multi-factor karst tunnel collapse risk assessment system, in addition to microseismic characteristics, the system comprehensively considers the geological strength indicators of the tunnel surrounding rock, groundwater level changes, and construction blasting vibration factors. The weights of each factor are determined using the analytic hierarchy process (AHP), and a fuzzy comprehensive evaluation method is applied for comprehensive assessment. This system can comprehensively and objectively assess tunnel collapse risks, providing construction personnel with detailed and accurate risk information, enabling them to take reasonable construction adjustment measures based on the risk level and ensure tunnel construction safety.

[0149] Example 2

[0150] This invention provides an intelligent early warning system for karst tunnel collapse based on multiple precursor features of microseismic events, comprising:

[0151] Multi-scale microseismic monitoring module: composed of multiple microseismic sensors arranged in layers according to the karst tunnel geological prediction model, wherein high-resolution piezoelectric sensors are arranged in the shallow layer with a spacing of 3-5 meters and distributed in a ring around the tunnel wall; electromagnetic induction type microseismic sensors are arranged in the middle layer in a quincunx pattern with a spacing of 5-8 meters; fiber Bragg grating microseismic sensors are arranged in the deep layer along the tunnel axis every 10-15 meters, and miniature acceleration sensor arrays are arranged at key positions in karst development intensive areas and fault fracture zones; it also includes a wireless ad hoc network transmission submodule using low-power Bluetooth and ZigBee hybrid communication technology for transmitting sensor collected data to the data processing center;

[0152] Data processing and feature extraction module: used for processing the collected microseismic signals, including time domain feature analysis of microseismic signals to obtain peak amplitude, duration, rise time and pulse count, frequency domain analysis using short-time Fourier transform to obtain main frequency, frequency bandwidth, spectral centroid and other frequency domain features, decomposition of microseismic signals using Hilbert-Huang transform to obtain intrinsic mode functions and calculate their instantaneous frequency and instantaneous amplitude to construct a time-frequency joint feature matrix, and analysis of the concave-convex, symmetry, steepness and other geometric feature parameters of the microseismic waveform, and fusion of time domain-frequency domain joint features and waveform morphological features through a multi-modal deep belief network;

[0153] Intelligent early warning model module: including a quantum encoding unit and a quantum neural network unit, the quantum encoding unit encodes the fused microseismic multi-modal features using quantum bits to construct a quantum feature vector, the quantum neural network unit takes the quantum encoded microseismic feature vector as input, the hidden layer uses quantum entanglement gates and quantum rotation gates to construct a complex quantum computing structure, and the output layer outputs the collapse early warning result, and the module uses quantum backpropagation algorithm to train the quantum neural network;

[0154] Early warning threshold adjustment and risk assessment module: including a Kalman filter submodule and a multi-factor risk assessment submodule, the Kalman filter submodule uses Kalman filter algorithm to perform real-time filtering and prediction on microseismic feature data, adjusts the early warning threshold dynamically according to the trend and uncertainty of the predicted value, and modifies the early warning threshold through a fuzzy logic reasoning system in combination with the construction progress, support condition and geological condition change factors in the tunnel; the multi-factor risk assessment submodule constructs a risk assessment system containing microseismic features, geological strength indicators of tunnel surrounding rock, groundwater level changes and construction blasting vibration factors, determines the weight of each factor using the analytic hierarchy process, and uses fuzzy comprehensive evaluation method to comprehensively evaluate the collapse risk of the tunnel.

[0155] Example 3

[0156] The present embodiment provides a karst tunnel collapse intelligent early warning device based on microseismic multi-precursor features, comprising:

[0157] a memory for storing instructions; wherein the instructions are used to implement the karst tunnel collapse intelligent early warning method based on microseismic multi-precursor characteristics of embodiment 1;

[0158] a processor for executing the instructions in the memory.

[0159] The above specific embodiments are only several optional embodiments of the present application, and based on the technical solutions of the present application and the related inspirations of the above embodiments, the person skilled in the art can make various alternative improvements and combinations on the above specific embodiments.

Claims

1. A method for intelligent early warning of karst tunnel collapse based on multiple precursor features of microseismic events, characterized in that, Includes the following steps: Construction of a multi-scale microseismic monitoring network: Based on the geological conditions of karst tunnels, a multi-scale monitoring network is constructed, which includes different types of sensors in shallow, middle and deep layers and supplementary arrays are added in key parts. Data is transmitted through a wireless self-organizing network system that uses a hybrid communication of Bluetooth Low Energy and ZigBee. Multimodal microseismic feature processing: For the acquired microseismic signals, time-domain feature analysis, short-time Fourier transform frequency-domain analysis, Hilbert-Huang transform to construct time-frequency joint feature matrix and waveform morphology feature mining are performed. Then, a multimodal deep belief network is used to deeply fuse the time-frequency joint features and waveform morphology features. Quantum Neural Network Early Warning Model Construction: Quantum bits are used to encode the fused micro-seismic multimodal features to construct a quantum feature vector. A quantum neural network with a hidden layer structure consisting of quantum entanglement gates and quantum rotation gates is constructed. The quantum feature vector is used as input and the collapse early warning result is used as output. The model is trained using the quantum backpropagation algorithm to achieve an intelligent early warning function based on quantum computing. Adaptive early warning and risk assessment: The Kalman filter algorithm is used to filter and predict microseismic characteristic data and dynamically adjust the early warning threshold accordingly. At the same time, the threshold is corrected by fuzzy logic reasoning system in combination with factors such as construction progress, support conditions and geological condition changes. The weight of multiple factors such as microseismic characteristics, geological strength index, groundwater level changes and construction blasting vibration is determined by the analytic hierarchy process. The fuzzy comprehensive evaluation method is used to comprehensively assess the risk of tunnel collapse. The adaptive early warning and risk assessment specifically includes the following steps: using the Kalman filter algorithm to filter and predict microseismic feature data in real time; dynamically adjusting the early warning threshold based on the changing trend and uncertainty of the predicted values; and correcting the early warning threshold through a fuzzy logic reasoning system in combination with factors such as tunnel construction progress, support conditions, and geological condition changes. A multi-factor karst tunnel collapse risk assessment system is constructed, using the analytic hierarchy process (AHP) to determine the weights of microseismic features, geological strength indicators of the tunnel surrounding rock, groundwater level changes, and construction blasting vibration factors; and using the fuzzy comprehensive evaluation method to comprehensively assess the tunnel collapse risk. In the adaptive early warning and risk assessment, the Kalman filter algorithm is used, assuming the microseismic feature data vector is... The state vector is The measurement matrix is The state transition matrix is The process noise covariance matrix is The measurement noise covariance matrix is Prediction steps: , Update steps: , , When the prediction error covariance matrix The determinant value is less than At that time, the warning threshold ,in To adjust the coefficient, This represents the change in the predicted value.

2. The intelligent early warning method for karst tunnel collapse based on multiple precursor features of microseismic events according to claim 1, characterized in that, The construction of the multi-scale microseismic monitoring network specifically includes the following steps: Based on the geological prediction model of karst tunnels, the surrounding rock of the tunnel is divided into three monitoring layers: shallow, middle and deep. In the shallow layer, high-resolution piezoelectric sensors are used, which are distributed in a ring around the tunnel wall at a spacing of 3-5 meters. Electromagnetic induction microseismic sensors are used in the middle layer, arranged in a quincunx pattern with a spacing of 5-8 meters; fiber optic grating microseismic sensors are used in the deep layer, with a monitoring point placed every 10-15 meters along the tunnel axis. At the same time, additional micro-accelerometer arrays were installed in key areas of densely developed karst and fault fracture zones; A wireless self-organizing network transmission system using hybrid communication technologies of Bluetooth Low Energy and ZigBee is constructed. The communication mode is intelligently switched to transmit sensor data to the data processing center based on the distance between the sensor node and the data processing center and the signal interference. The sensitivity of shallow piezoelectric sensors is as high as ,unit: And sensor spacing Based on tunnel radius and surrounding rock properties set (rice); Spacing of mid-layer electromagnetic induction micro-seismic sensors (rice); Spacing of deep fiber optic grating microseismic sensors (rice).

3. The intelligent early warning method for karst tunnel collapse based on multiple precursor features of microseismic events according to claim 1, characterized in that, The multimodal microseismic feature extraction and fusion specifically includes the following steps: performing time-domain feature analysis on the acquired microseismic signals, calculating peak amplitude, duration, rise time, and pulse count parameters; using short-time Fourier transform for frequency-domain analysis to obtain the spectrum and extract frequency-domain features such as dominant frequency, bandwidth, and spectral centroid; using Hilbert-Huang transform to decompose the microseismic signals to obtain intrinsic mode functions and calculate their instantaneous frequency and instantaneous amplitude to construct a time-frequency joint feature matrix; deeply analyzing the geometric feature parameters such as concavity, symmetry, and steepness of the microseismic waveform; and using a multimodal deep belief network to fuse the time-frequency joint features and waveform morphology features. In time-domain characteristic analysis, peak amplitude Duration ,in, For the first time the microseismic signal exceeds the threshold Time determined based on background noise The rise time is the time when the microseismic signal last exceeds the threshold. ,in, Pulse count is the time it takes for the microseismic signal to reach its peak value. ； in, To determine the function of the microseismic signal pulse, when the rate of change of the signal slope exceeds... hour ,otherwise , This represents the number of signal sampling points. Short-time Fourier transform, let the window function be... The window length is ,but: ； in, It is time. It is frequency; Clock speed: ； Bandwidth: ； in, and The energy in the spectrum is greater than a certain threshold. Based on the maximum and minimum frequencies determined by the signal energy distribution, the centroid of the spectrum is: 。 4. The intelligent early warning method for karst tunnel collapse based on multiple precursor features of microseismic events according to claim 3, characterized in that, In the multimodal microseismic feature extraction and fusion, the Hilbert-Huang transform is used to obtain the eigenmode functions and instantaneous frequencies. ； in, These are intrinsic modulo functions; Instantaneous amplitude: ； When mining waveform morphology features, concavity and convexity The sign of the waveform determines its concavity or convexity. It is concave. Convex, symmetry: ； in, The midpoint time of the waveform. Half-wavelength time width, steepness The rate of change is used to assess the steepness of the slope, and the calculation is performed. The slope variation near the peak; Multimodal deep belief networks are composed of multiple restricted Boltzmann machines stacked together. The number of visible layer cells in a layered RBM is The number of hidden layer units is The visible layer state vector is The hidden layer state vector is The energy function is: ； in, For connection weights, For visible layer bias, For hidden layer bias; Parameter update formula: ； ； ； in, This is the learning rate.

5. The intelligent early warning method for karst tunnel collapse based on multiple precursor features of microseismic events according to claim 1, characterized in that, The construction of an intelligent early warning model based on quantum neural networks includes the following steps: using qubits to encode the fused microseismic multimodal features to construct a quantum feature vector; constructing a quantum neural network, whose input layer is the quantum-encoded microseismic feature vector, whose hidden layer uses quantum entanglement gates and quantum rotation gates to construct a complex quantum computing structure, whose output layer is the collapse early warning result; and using the quantum backpropagation algorithm to train the quantum neural network. In the construction of the intelligent early warning model based on quantum neural networks, the quantum bit encoding formula is as follows: ； ； in, , It is a quantum state. These are elements in the microseismic feature vector. This encoding method maps the microseismic feature vector to the quantum state space, and utilizes the superposition and entanglement of quantum states to explore the potential correlation between microseismic features. In the construction of the intelligent early warning model based on quantum neural networks, the quantum neural network employs a quantum rotating gate: ； The loss function uses mean squared error: ； in, For the sample size, To predict the output, This is the actual output.

6. The intelligent early warning method for karst tunnel collapse based on multiple precursor features of microseismic events according to claim 1, characterized in that, In the adaptive early warning and risk assessment, when determining weights using the analytic hierarchy process, a judgment matrix is ​​set. ,in, Indicator Factors Relative factors First calculate the importance. The product of the elements in each row: ； Recalculate of Root Finally, for Normalization process is performed to obtain ; In the adaptive early warning and risk assessment, the fuzzy comprehensive evaluation method assumes that the evaluation set is... The single-factor fuzzy evaluation matrix is Fuzzy comprehensive evaluation results ,in For fuzzy synthesis operators, a weighted average operator is used.

7. An intelligent early warning system for karst tunnel collapse based on multiple precursor characteristics of microseismic events, characterized in that, include: The multi-scale microseismic monitoring module consists of various microseismic sensors layered according to the geological prediction model of karst tunnels. The shallow layer contains high-resolution piezoelectric sensors, spaced 3-5 meters apart and arranged in a ring around the tunnel wall. The middle layer contains electromagnetic induction microseismic sensors, arranged in a quincunx pattern with a spacing of 5-8 meters. The deep layer contains fiber optic grating microseismic sensors, with a monitoring point every 10-15 meters along the tunnel axis. Miniature accelerometer arrays are also located in areas with dense karst development and key locations in fault fracture zones. The module also includes a wireless self-organizing network transmission submodule using a hybrid communication technology of Bluetooth Low Energy and ZigBee to transmit the data collected by the sensors to the data processing center. The data processing and feature extraction module is used to process the acquired microseismic signals, including performing time-domain feature analysis to obtain peak amplitude, duration, rise time, and pulse count; using short-time Fourier transform for frequency-domain analysis to obtain frequency-domain features such as dominant frequency, bandwidth, and spectral centroid; using Hilbert-Huang transform to decompose the microseismic signals to obtain intrinsic mode functions and calculate their instantaneous frequency and instantaneous amplitude to construct a time-frequency joint feature matrix; and analyzing geometric feature parameters such as concavity, symmetry, and steepness of the microseismic waveform. Finally, it fuses the time-frequency joint features and waveform morphology features through a multimodal deep belief network. The intelligent early warning model module includes a quantum encoding unit and a quantum neural network unit. The quantum encoding unit uses qubits to encode the fused microseismic multimodal features to construct a quantum feature vector. The quantum neural network unit takes the quantum-encoded microseismic feature vector as input. Its hidden layer uses quantum entanglement gates and quantum rotation gates to construct a complex quantum computing structure. The output layer outputs the collapse early warning result. This module uses the quantum backpropagation algorithm to train the quantum neural network. The early warning threshold adjustment and risk assessment module includes a Kalman filtering submodule and a multi-factor risk assessment submodule. The Kalman filtering submodule uses the Kalman filtering algorithm to filter and predict microseismic characteristic data in real time, and dynamically adjusts the early warning threshold based on the changing trend and uncertainty of the predicted values. Simultaneously, it uses a fuzzy logic reasoning system to correct the early warning threshold by considering factors such as tunnel construction progress, support conditions, and changes in geological conditions. The multi-factor risk assessment submodule constructs a risk assessment system that includes microseismic characteristics, geological strength indicators of the tunnel surrounding rock, groundwater level changes, and construction blasting vibration factors. It uses the analytic hierarchy process (AHP) to determine the weights of each factor and employs a fuzzy comprehensive evaluation method to comprehensively assess the tunnel's collapse risk. The Kalman filter algorithm is used to filter and predict microseismic feature data in real time. The warning threshold is dynamically adjusted according to the changing trend and uncertainty of the predicted value. The warning threshold is corrected by fuzzy logic reasoning system in combination with the tunnel construction progress, support conditions and geological condition changes. A multi-factor karst tunnel collapse risk assessment system is constructed. The weights of microseismic features, geological strength index of tunnel surrounding rock, groundwater level change and construction blasting vibration factors are determined by the analytic hierarchy process. The fuzzy comprehensive evaluation method is used to comprehensively assess the tunnel collapse risk. Kalman filtering algorithm, assuming the microseismic feature data vector is... The state vector is The measurement matrix is The state transition matrix is The process noise covariance matrix is The measurement noise covariance matrix is Prediction steps: , Update steps: , , When the prediction error covariance matrix The determinant value is less than At that time, the warning threshold ,in To adjust the coefficient, This represents the change in the predicted value.

8. An intelligent early warning device for karst tunnel collapse based on multiple precursor characteristics of microseismic events, characterized in that, include: A memory for storing instructions; wherein the instructions are used to implement the intelligent early warning method for karst tunnel collapse based on multiple precursor features of microseismic events as described in any one of claims 1-6; A processor for executing instructions in the memory.

Images