Tunnel geological disaster early warning method and system based on kilometer level horizontal hole
By constructing a sensor network and deep learning model in a kilometer-level horizontal borehole tunnel, the problems of limited monitoring range and low early warning accuracy were solved, achieving full-dimensional monitoring and high-precision early warning.
Patent Information
- Application Number
- CN202511110550.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-08
- Publication Date
- 2025-11-07
AI Technical Summary
Existing methods for monitoring and early warning of geological disasters in tunnels with kilometer-level horizontal boreholes suffer from limited monitoring range, poor data coordination, and low early warning accuracy.
Based on geological survey data of tunnel engineering, a sensor network is constructed, including stress sensors, electrical resistivity sensors, osmotic pressure sensors and seismic wave sensors, to collect multi-source monitoring data in real time. Sensitive features of geological hazards are extracted through edge processing modules, and geological hazard assessment is carried out using deep learning network models to determine risk levels and issue early warnings.
It has achieved full-dimensional monitoring coverage of kilometer-level horizontal borehole tunnels, improved data coordination and real-time early warning, enhanced early warning accuracy, and met the requirements for long-distance real-time monitoring.
Smart Images
Figure CN120913345A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of geological disaster early warning, in particular to a tunnel geological disaster early warning method and system based on a kilometer-level horizontal hole. BACKGROUND
[0002] In tunnel engineering construction, especially for kilometer-level horizontal hole related tunnels, due to the complex and changeable geological conditions through which the tunnels pass, the occurrence of geological disasters is often sudden and destructive, seriously threatening the safety of the project and the safety of construction personnel. For example, water and mud inrush disasters can instantly flood the construction area, causing equipment damage and personnel casualties; landslides and rock bursts can cause tunnel structure damage, delay project progress, and increase project cost.
[0003] At present, the monitoring and early warning methods for tunnel geological disasters mainly include the following types:
[0004] Manual inspection method: construction personnel regularly inspect the internal and surrounding geological conditions of the tunnel and determine whether there are disaster hazards based on experience. However, this method is inefficient, highly subjective, difficult to monitor in real time, and for kilometer-level horizontal hole tunnels, some areas are difficult for humans to reach, resulting in monitoring blind spots.
[0005] Single-point sensor monitoring method: stress, strain, displacement, and seepage pressure sensors are arranged in the tunnel to monitor specific physical quantities. However, kilometer-level tunnels have a large length and depth, and a wide range needs to be monitored. This method can only monitor local points, the monitoring range is limited, it is difficult to fully reflect the overall geological conditions of the tunnel, there are monitoring blind spots, and there is a lack of collaborative analysis between different sensor data, which can easily result in false positives or false negatives.
[0006] Traditional data analysis method: simple statistical analysis or threshold judgment is performed on the monitored data, and an early warning is issued when the data exceeds the preset threshold. However, the occurrence of geological disasters is a complex dynamic process, and a single threshold cannot accurately capture the precursor information of disaster occurrence, resulting in low early warning accuracy. SUMMARY
[0007] The present application aims to solve the problems of limited monitoring range, poor data collaboration, and low early warning accuracy in existing kilometer-level horizontal hole tunnel geological disaster monitoring and early warning methods, and proposes a tunnel geological disaster early warning method and system based on a kilometer-level horizontal hole.
[0008] The technical solution adopted by the present application to solve the above technical problems is:
[0009] In a first aspect, the present application provides a tunnel geological disaster early warning method based on a kilometer-level horizontal hole, which comprises:
[0010] Based on the tunnel engineering geological survey data, the monitoring points are determined in the kilometer-level horizontal hole and around the tunnel, and the types and quantities of sensors at each monitoring point are determined, and the corresponding types and quantities of sensors are collaboratively arranged at each monitoring point to form a sensor network, wherein the sensors include stress sensors, electrical sensors, osmotic pressure sensors, acoustic wave sensors and seismic wave sensors;
[0011] The sensor network collects multi-source monitoring data in real time, and the multi-source monitoring data includes stress data, underground water data, seepage flow data, acoustic wave data and adverse geological data;
[0012] After the multi-source monitoring data collected by the data preprocessing module close to the sensor side is preprocessed, the geological disaster sensitive features are extracted, and the geological disaster sensitive features are transmitted to the data processing center;
[0013] The data processing center inputs the geological disaster sensitive features into the pre-trained geological disaster evaluation model to obtain the probability of geological disaster occurrence, determines the risk level according to the probability of geological disaster occurrence, and if the risk level reaches the warning threshold, the corresponding warning information is sent to control the warning module to issue the corresponding warning.
[0014] Further, the preprocessed multi-source monitoring data includes:
[0015] The collected multi-source monitoring data is subjected to noise reduction processing, normalization processing and abnormal value processing.
[0016] Further, the geological disaster sensitive features include stress change rate, seepage flow growth rate, acoustic wave main frequency shift and seismic wave velocity.
[0017] Further, the geological disaster evaluation model adopts a deep learning network model architecture, including an input layer, a hidden layer and an output layer, the input layer is used to receive the geological disaster sensitive features, the hidden layer is composed of multiple neurons, and is used to mine the spatio-temporal correlation between features through multiple nonlinear transformations, and the output layer is used to output the probability of geological disaster occurrence and the risk level.
[0018] Further, the training method of the geological disaster evaluation model includes:
[0019] Obtain the geological disaster cases and historical multi-source monitoring data, generate multi-sensor joint time series data similar to real disaster precursors based on the geological disaster cases and historical multi-source monitoring data and using the generative adversarial network, and train the deep learning network model according to the multi-sensor joint time series data, and add a sensor signal attenuation correction term in the loss function during training.
[0020] Further, the risk level is determined according to the probability of geological disaster occurrence, including:
[0021] If the probability of the geological disaster is in the first preset range, the risk level is low risk; if the probability of the geological disaster is in the second preset range, the risk level is medium risk; and if the probability of the geological disaster is in the third preset range, the risk level is high risk.
[0022] Further, the early warning information includes sound and light early warning information, terminal early warning information and platform early warning information.
[0023] In a second aspect, the present application provides a tunnel geological disaster early warning system based on a kilometer-level horizontal hole, which is used to realize the tunnel geological disaster early warning method based on a kilometer-level horizontal hole as described in the first aspect, and the system comprises:
[0024] A sensor arrangement module is configured to determine monitoring points and types and quantities of sensors of each monitoring point in the kilometer-level horizontal hole and the periphery of the tunnel based on tunnel engineering geological survey data, and to cooperatively arrange sensors of corresponding types and quantities at each monitoring point to form a sensor network, wherein the sensors include stress sensors, electrical method sensors, osmotic pressure sensors, sound wave sensors and seismic wave sensors.
[0025] A sensor network is configured to collect multi-source monitoring data in real time, wherein the multi-source monitoring data includes stress data, underground water data, seepage flow data, sound wave data and adverse geological data.
[0026] A data preprocessing module is arranged close to the sensor side, configured to preprocess the collected multi-source monitoring data, extract geological disaster sensitive features, and transmit the geological disaster sensitive features to a data processing center.
[0027] The data processing center is configured to input the geological disaster sensitive features into a pre-trained geological disaster evaluation model, obtain a probability of the geological disaster, determine a risk level according to the probability of the geological disaster, and issue corresponding early warning information if the risk level reaches an early warning threshold.
[0028] An early warning module is configured to issue corresponding early warning according to the early warning information.
[0029] The beneficial effects of the present application are: the tunnel geological disaster early warning method and system based on a kilometer-level horizontal hole provided by the present application, based on engineering geological survey data, accurately locates potential disaster risk areas in the horizontal hole and around the tunnel, ensures monitoring coverage of key points, and multi-sensor collaborative arrangement can cover the axial and radial full-dimensional risk points of the kilometer-level hole; through unified feature extraction, multi-physical field data are fused, the data fragmentation problem in the traditional method is solved, and the cooperativity of the monitoring data is enhanced; the feature extraction can be completed on the edge device close to the sensor, only the low-dimensional characteristic value is transmitted, the data transmission delay is reduced, the real-time requirement of the kilometer-level long distance is met, and the early warning real-time is improved; the nonlinear disaster mechanism is captured by using the geological disaster evaluation model, the real-time quantitative evaluation of the risk level is realized, compared with the traditional method, the early warning accuracy is improved; during training, a sensor signal attenuation correction term is added to the loss function, the influence of the end data distortion caused by the hole length is eliminated, and the early warning accuracy is further improved. BRIEF DESCRIPTION OF DRAWINGS
[0030] Figure 1 The flowchart of the tunnel geological disaster early warning method based on a kilometer-level horizontal hole provided for the embodiment is shown in the figure;
[0031] Figure 2 The structure diagram of the geological disaster evaluation model provided for the embodiment is shown in the figure;
[0032] Figure 3 The structure diagram of the tunnel geological disaster early warning system based on a kilometer-level horizontal hole provided for the embodiment is shown in the figure. DETAILED DESCRIPTION
[0033] The technical scheme of the present application is applicable to the application scenario of needing to perform geological disaster early warning on a kilometer-level horizontal hole tunnel. Due to the influence of the large length, large burial depth and complex and variable geological conditions of the kilometer-level tunnel, there are problems of limited monitoring range, poor data cooperativity and low early warning accuracy in the current kilometer-level horizontal hole tunnel geological disaster monitoring and early warning.
[0034] Based on this, the technical scheme of the present application is proposed. In the present application, first, targeted network deployment is performed based on geological survey data, a sensor network capable of covering the axial and radial full-dimensional risk points of the kilometer-level hole is constructed, and the monitoring coverage range is ensured; in addition, the sensor network is constructed by multiple types of sensors, which can accurately and comprehensively reflect the rock mass stress state, underground water activity and seepage pressure, rock mass rupture signal and wave speed anomaly, and ensure monitoring coverage of key disaster-causing factors; then, multi-source monitoring data collected by the sensor network are subjected to edge processing preprocessing and feature extraction, so as to compress the long-distance transmission data volume, reduce the data transmission delay, and improve the data transmission real-time; finally, based on the extracted geological disaster sensitive features and based on the geological disaster evaluation model, the dynamic quantization of the geological risk is performed, and the early warning accuracy is improved.
[0035] The technical solutions in the embodiments will be described clearly and completely below in combination with the drawings in the embodiments. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments.
[0036] Figure 1 A flowchart of a tunnel geological disaster early warning method based on a kilometer-level horizontal hole is shown, please refer to Figure 1 The method comprises the following steps:
[0037] Step 1, based on the tunnel engineering geological survey data, the monitoring points in the kilometer-level horizontal hole and the surrounding tunnel are determined, as well as the types and quantities of sensors at each monitoring point, and the corresponding types and quantities of sensors are collaboratively laid at each monitoring point to form a sensor network.
[0038] It can be understood that the geological risk characteristics directly determine the sensor type, density and spatial topological relationship. Based on this, the present embodiment targets the network according to the tunnel engineering geological survey data, and then ensures that the sensor network can comprehensively cover the axial and radial full-dimensional risk points of the kilometer-level hole from the source, while selecting the corresponding types and quantities of sensors for different types of risk points to accurately obtain the corresponding geological risk characteristics according to the corresponding density and sensor type.
[0039] In actual application, the tunnel engineering geological survey data contains lithology distribution, fault position, hydrological partition and other data, and then the network can be targeted according to the tunnel engineering geological survey data. For example, the seismic wave sensor is laid according to the position of the fault fracture zone to monitor the precursor of rock mass instability; the stress sensor is deployed according to the high stress area prediction map to capture stress redistribution; the seepage pressure sensor + electrical method sensor is configured for the water-rich structure area to form an underground water seepage monitoring matrix; the sound wave sensor array is added in the potential rock burst section to listen to the micro-fracture acoustic emission signal.
[0040] In the present embodiment, the sensors include stress sensors, electrical method sensors, seepage pressure sensors, sound wave sensors and seismic wave sensors. Among them, the stress sensor collects the surrounding rock stress fluctuation data in real time. The electrical method sensor synchronously acquires the rock resistivity change to reflect the fracture water migration path, the seepage pressure sensor monitors the pore water pressure gradient to quantify the water inrush driving force, and the sound wave / seismic wave sensor captures the elastic wave propagation anomaly to reflect the rock mass damage degree.
[0041] Step 2, the sensor network collects multi-source monitoring data in real time, and the multi-source monitoring data includes stress data, underground water data, seepage flow data, sound wave data and adverse geological data.
[0042] In practical applications, all sensors are sampled with a unified timestamp to establish a spatiotemporal alignment benchmark for multi-physical field monitoring data, laying a foundation for subsequent fusion analysis.
[0043] Step 3: After pre-processing the collected multi-source monitoring data by the data pre-processing module close to the sensor side, the geological disaster sensitive features are extracted and transmitted to the data processing center.
[0044] It can be understood that, due to the large amount of data collected by the sensor network, direct transmission will result in high delay, especially for kilometer-level tunnels, high delay cannot meet the real-time requirements of early warning. Based on this, the present embodiment pre-processes the collected multi-source monitoring data by the data pre-processing module of the edge device, extracts the geological disaster sensitive features, and then only transmits the low-dimensional geological disaster sensitive features, thereby reducing the data transmission delay. Through edge processing, the transmission bottleneck is eliminated, and the monitoring data delay of kilometer-level holes is compressed from minutes to seconds.
[0045] In the present embodiment, the pre-processing of the collected multi-source monitoring data includes noise reduction processing, normalization processing and outlier processing. Through pre-processing, the key features can be preserved, the construction vibration interference can be filtered out, the dimension difference can be eliminated, and the device drift can be repaired, thereby improving the monitoring data quality and laying a foundation for extracting the geological disaster sensitive features.
[0046] In the present embodiment, feature extraction converts physical signals into model-recognizable disaster indicators, and the extracted geological disaster sensitive features include stress change rate, seepage flow growth rate, sound wave main frequency shift and seismic wave velocity. Among them, the sudden increase of stress change rate indicates the risk of rock burst, the large seepage flow growth rate indicates the water inrush precursor, the sound wave main frequency shift is the rock mass damage sign, and the seismic wave velocity is the crack expansion signal.
[0047] Step 4: The data processing center inputs the geological disaster sensitive features into the pre-trained geological disaster evaluation model to obtain the probability of geological disaster occurrence, determines the risk level according to the probability of geological disaster occurrence, and if the risk level reaches the early warning threshold, the corresponding early warning information is issued to control the early warning module to issue the corresponding early warning.
[0048] In the present embodiment, the geological disaster evaluation model adopts a deep learning network model architecture, please refer to Figure 2 , which includes an input layer, a hidden layer and an output layer. The input layer is used to receive the geological disaster sensitive features, the hidden layer is composed of multiple neurons, and is used to mine the spatiotemporal correlation between features through multiple layers of nonlinear transformation (such as CNN-LSTM hybrid network), and the output layer is used to output the probability of geological disaster occurrence and the risk level.
[0049] The deep learning network model can capture the nonlinear disaster mechanism and realize real-time quantitative evaluation of the risk level, thereby improving the early warning accuracy compared with the traditional method.
[0050] In the embodiment, the training method of the geological disaster evaluation model comprises: obtaining occurred geological disaster cases and historical multi-source monitoring data, generating multi-sensor joint time series data similar to real disaster precursors based on the geological disaster cases and historical multi-source monitoring data and by using the generative adversarial network, and training the deep learning network model according to the multi-sensor joint time series data, and adding a sensor signal attenuation correction term in the loss function during the training.
[0051] In actual application, a data set can be constructed according to the multi-sensor joint time series data and the corresponding geological disaster occurrence, and the data set is divided into a training set and a validation set according to a preset ratio, the deep learning network model is trained by using the training set, and the prediction error of the deep learning network model is verified by using the validation set. When the loss function of the validation set is minimum and the prediction error is less than an error threshold, the geological disaster evaluation model is obtained. The loss function can be a mean square error, a mean absolute error or other error measurement functions suitable for regression tasks, and the model optimization can use an Adam optimizer.
[0052] In the training process, the deep learning network model uses a self-attention mechanism to learn the interaction of stress-seepage-acoustic parameters, thereby capturing the nonlinear disaster mechanism. Meanwhile, the embodiment adds a sensor signal attenuation correction term in the loss function to eliminate the influence of the distortion of the end data caused by the hole length, thereby further improving the early warning accuracy.
[0053] After obtaining the geological disaster sensitive features in the data processing center, the geological disaster sensitive features are input into the pre-trained geological disaster evaluation model, and the probability of the occurrence of the geological disaster is obtained. The risk level is determined according to the probability of the occurrence of the geological disaster. For example, if the probability of the occurrence of the geological disaster is in a first preset range (such as less than or equal to 0.3), the risk level is low risk; if the probability of the occurrence of the geological disaster is in a second preset range (such as greater than 0.3 and less than 0.6), the risk level is medium risk; and if the probability of the occurrence of the geological disaster is in a third preset range (such as greater than or equal to 0.6), the risk level is high risk.
[0054] If the risk level reaches a warning threshold, such as medium risk, corresponding warning information is sent to control the corresponding warning of the warning module, including sending sound and light warning information, terminal warning information and platform warning information, to control the corresponding warning of the sound and light warning module, the terminal warning module and the platform warning module, so that the relevant personnel can take corresponding emergency measures.
[0055] In summary, the tunnel geological disaster early warning method based on a kilometer-level horizontal hole provided in the embodiment is based on engineering geological survey data, accurately locates potential disaster risk areas in the horizontal hole and around the tunnel, ensures monitoring coverage of key points, and multi-sensor collaborative arrangement can cover axial and radial full-dimensional risk points of a kilometer-level hole; through unified feature extraction, fusion of multi-physical field data, the data fragmentation problem in the traditional method is solved, and the cooperativity of monitoring data is enhanced; feature extraction can be completed on the edge device close to the sensor, only low-dimensional feature values are transmitted, data transmission delay is reduced, real-time requirements of a kilometer-level long distance are met, and early warning real-time is improved; the geological disaster evaluation model is used to capture the nonlinear disaster mechanism, realize real-time quantitative evaluation of the risk level, and compared with the traditional method, the early warning accuracy is improved; during training, a sensor signal attenuation correction term is added to the loss function, the influence of end data distortion caused by hole length is eliminated, and the early warning accuracy is further improved.
[0056] Based on the above technical solution, the embodiment further provides a tunnel geological disaster early warning system based on a kilometer-level horizontal hole, which is used to realize the tunnel geological disaster early warning method based on a kilometer-level horizontal hole described in the embodiment, please refer to Figure 3 , the system comprises:
[0057] A sensor arrangement module is configured to determine monitoring points and types and quantities of sensors of each monitoring point based on tunnel engineering geological survey data in a kilometer-level horizontal hole and around the tunnel, collaboratively arrange sensors of corresponding types and quantities at each monitoring point to form a sensor network, and the sensors comprise stress sensors, electrical method sensors, osmotic pressure sensors, acoustic wave sensors and seismic wave sensors.
[0058] A sensor network is configured to collect multi-source monitoring data in real time, and the multi-source monitoring data comprises stress data, underground water data, seepage flow data, acoustic wave data and adverse geological data.
[0059] A data preprocessing module is arranged close to the sensor and configured to preprocess the collected multi-source monitoring data, extract geological disaster sensitive features, and transmit the geological disaster sensitive features to a data processing center.
[0060] The data processing center is configured to input the geological disaster sensitive features into a pre-trained geological disaster evaluation model, obtain a probability of geological disaster occurrence, determine a risk level according to the probability of geological disaster occurrence, and if the risk level reaches an early warning threshold, corresponding early warning information is issued.
[0061] An early warning module is configured to issue corresponding early warning according to the early warning information.
[0062] It can be understood that, since the tunnel geological disaster early warning system based on the kilometer-level horizontal hole described in the embodiment is a system for implementing the tunnel geological disaster early warning method based on the kilometer-level horizontal hole described in the embodiment, for the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, it is described more simply, and the relevant part is referred to the part of the method description.
Claims
1. A tunnel geological disaster early warning method based on a kilometer-level horizontal hole, characterized in that, The method comprises: Based on the tunnel engineering geological survey data, the monitoring points are determined in the kilometer-level horizontal hole and the periphery of the tunnel, and the types and quantities of sensors of each monitoring point are determined, and the corresponding types and quantities of sensors are collaboratively arranged at each monitoring point to form a sensor network, wherein the sensors comprise stress sensors, electrical method sensors, osmotic pressure sensors, acoustic wave sensors and seismic wave sensors; The sensor network collects multi-source monitoring data in real time, wherein the multi-source monitoring data comprises stress data, underground water data, seepage flow data, acoustic wave data and adverse geological data; After the collected multi-source monitoring data is preprocessed by the data preprocessing module close to the sensor side, the geological disaster sensitive features are extracted, and the geological disaster sensitive features are transmitted to the data processing center; The data processing center inputs the geological disaster sensitive features into a pre-trained geological disaster evaluation model to obtain a probability of geological disaster occurrence, determines a risk level according to the probability of geological disaster occurrence, and if the risk level reaches a warning threshold, corresponding warning information is sent to control the warning module to send corresponding warning. 2.The kilometer-level horizontal hole based tunnel geological disaster early warning method according to claim 1, characterized in that, The preprocessing of the collected multi-source monitoring data comprises: The collected multi-source monitoring data is subjected to noise reduction processing, normalization processing and abnormal value processing. 3.The kilometer-level horizontal hole based tunnel geological disaster early warning method according to claim 1, characterized in that, The geological disaster sensitive features comprise stress change rate, seepage flow growth rate, acoustic wave main frequency shift and seismic wave velocity. 4.The kilometer-level horizontal hole based tunnel geological disaster early warning method according to claim 1, characterized in that, The geological disaster evaluation model adopts a deep learning network model architecture, comprising an input layer, a hidden layer and an output layer, the input layer is used to receive the geological disaster sensitive features, the hidden layer is composed of multiple neurons and is used to mine the spatio-temporal correlation between features through multiple layers of nonlinear transformation, and the output layer is used to output the probability of geological disaster occurrence and the risk level.
5. The kilometer-level horizontal hole-based tunnel geological disaster early warning method according to claim 4, characterized in that, The training method of the geological disaster evaluation model comprises: The occurred geological disaster cases and historical multi-source monitoring data are obtained, the multi-sensor joint time series data similar to the real disaster precursor are generated based on the geological disaster cases and the historical multi-source monitoring data and by using the generative adversarial network, the deep learning network model is trained according to the multi-sensor joint time series data, and a sensor signal attenuation correction term is added in the loss function during the training. 6.The kilometer-level horizontal hole based tunnel geological disaster early warning method according to claim 1, characterized in that, The determination of the risk level according to the probability of geological disaster occurrence comprises: If the probability of geological disaster occurrence is in a first preset range, the risk level is low risk; if the probability of geological disaster occurrence is in a second preset range, the risk level is medium risk; and if the probability of geological disaster occurrence is in a third preset range, the risk level is high risk. 7.The kilometer-level horizontal hole based tunnel geological disaster early warning method according to claim 1, characterized in that, The warning information comprises sound and light warning information, terminal warning information and platform warning information.
8. A tunnel geological disaster early warning system based on a kilometer-level horizontal hole, characterized in that, The system for implementing the kilometer-level horizontal hole-based tunnel geological disaster warning method according to any one of claims 1 to 7 comprises: A sensor arrangement module is configured to determine monitoring points in the kilometer-level horizontal hole and the periphery of the tunnel based on the tunnel engineering geological survey data, and determine the types and quantities of sensors of each monitoring point, and collaboratively arrange the corresponding types and quantities of sensors at each monitoring point to form a sensor network, wherein the sensors comprise stress sensors, electrical method sensors, osmotic pressure sensors, acoustic wave sensors and seismic wave sensors; A sensor network is used to collect multi-source monitoring data in real time, which includes stress data, underground water data, seepage flow data, sound wave data and adverse geological data; A data preprocessing module is arranged near the sensor side, which is used to preprocess the collected multi-source monitoring data, extract sensitive features of geological disasters and transmit the sensitive features to a data processing center; The data processing center is used to input the sensitive features of geological disasters into a pre-trained geological disaster assessment model, obtain a probability of occurrence of geological disasters, determine a risk level according to the probability of occurrence of geological disasters, and issue corresponding early warning information if the risk level reaches a warning threshold; An early warning module is used to issue corresponding early warning according to the early warning information.