An intelligent detection system and method for underground karst hazards based on electroseismic effect

Abstract

This invention discloses an intelligent detection system and method for underground karst hazards based on the electroseismic effect. It employs an electromagnetic emission module to transmit a single electromagnetic pulse signal to the detection area, and an electromagnetic acquisition module to receive the electromagnetic pulse signal and feed it back to an electromagnetic acquisition base station. Simultaneously, a seismic acquisition module receives seismic wave signals and feeds them back to a seismic acquisition base station. The intelligent control host synchronizes with both base stations, enabling simultaneous acquisition of two types of data with only a single electromagnetic pulse excitation, eliminating the need for additional excitation equipment. This ensures data acquisition under the same spatiotemporal conditions, providing accurate data support for subsequent non-destructive, efficient, and high-precision identification of urban underground hazards. During data processing, a self-constructed model is used to determine the presence of anomalies, eliminating the need for manual intervention and achieving good accuracy, thus providing real-time data support for urban safety management.

Description

Technical Field

[0001] This invention relates to the field of geophysical exploration, specifically to an intelligent detection system and method for underground karst hazards based on electroseismic effects. Background Technology

[0002] Underground karst hazards pose a significant threat to public safety and road construction. Traditional detection methods (such as ground-penetrating radar and shallow seismic surveys) suffer from significant limitations: ground-penetrating radar attenuates rapidly in moist clay layers, limiting its detection depth; shallow seismic surveys require artificial seismic sources, making them difficult and costly to implement in underground environments. Furthermore, these methods require on-site operation and data processing by specialized personnel, hindering rapid and intelligent surveys and early warning systems. Additionally, these methods necessitate separate data collection using different equipment, leading to complex operational procedures and the inability to simultaneously acquire two different types of data, thus hindering efficient and accurate identification of underground hazards.

[0003] Therefore, the research direction of this invention is to provide a new detection system and method that can simultaneously collect seismic and electromagnetic geophysical parameters, achieve non-destructive, high-efficiency, and high-precision identification of urban underground hazards, and automatically identify and issue early warnings after data processing, so as to provide real-time data support for urban safety management. Summary of the Invention

[0004] To address the problems existing in the prior art, this invention provides an intelligent detection system and method for underground karst hazards based on the electro-seismic effect. Utilizing the inverse effect of the seismoelectric effect, it emits an electromagnetic pulse while simultaneously collecting seismic and electromagnetic geophysical parameters, completing one excitation and two acquisitions. This enables non-destructive, high-efficiency, and high-precision identification of urban underground hazards. Furthermore, the system automatically identifies and issues warnings after processing the collected data through a constructed model, providing real-time data support for urban safety management.

[0005] To achieve the above objectives, the technical solution adopted by this invention is: an intelligent detection system for underground karst hazards based on the electroseismic effect, comprising a mobile detection platform and a control center. The mobile detection platform is equipped with an intelligent control host, a remote communication module, an electromagnetic transmission module, an electromagnetic acquisition module, a seismic acquisition module, a battery and power control system, and a motion control system. The electromagnetic transmission module is installed in the middle of the mobile detection platform and is used to generate electromagnetic pulse signals downwards from the platform. Multiple electromagnetic acquisition modules are installed on the mobile detection platform and are at the same horizontal level as the electromagnetic transmission module. These modules are used to receive electromagnetic signals returned from underground after the electromagnetic pulse signals are transmitted and send them to the electromagnetic acquisition base station. The seismic acquisition module includes... The system includes multiple seismic acquisition units deployed at the bottom of the mobile detection platform, used to receive seismic wave signals generated underground after electromagnetic pulse signal transmission and send them to the seismic acquisition base station; the intelligent control host is connected to the seismic acquisition base station and the electromagnetic acquisition base station, and receives control commands from the control center through a remote communication module, controls the movement of the mobile detection platform through the motion control system (i.e., the motor and drive wheels), controls the excitation of the electromagnetic transmission module, and receives data feedback from the seismic acquisition base station and the electromagnetic acquisition base station. After analysis and processing through the built-in model, it automatically identifies underground anomalies and feeds them back to the control center through the remote communication module to generate a three-dimensional anomaly distribution map; the battery and power control system is used to supply power to the above modules.

[0006] Furthermore, there are five electromagnetic acquisition modules in total, one of which is located in the center of the electromagnetic emission module, and the other four are evenly distributed around the electromagnetic emission module. The electromagnetic acquisition module is an electromagnetic expandable telescopic coil, and the four electromagnetic expandable telescopic coils on the periphery extend out of the moving detection platform during detection and retract into the moving detection platform when not detecting, respectively, through the first electric push rod.

[0007] Furthermore, the seismic acquisition unit includes a second electrically driven actuator and a seismic detector. The fixed end of the second electrically driven actuator is connected to the lower part of the mobile detection platform, and the seismic detector is mounted on the telescopic end of the second electrically driven actuator. When not conducting detection, the second electrically driven actuator is in the retracted state, and the seismic detector is not in contact with the ground. When conducting detection, the second electrically driven actuator is in the extended state, and the seismic detector is in close contact with the ground. This arrangement allows the seismic acquisition units to be arrayed under the mobile detection platform, improving the accuracy of receiving seismic data.

[0008] Furthermore, the mobile detection platform is equipped with lidar, acoustic radar, a GPS positioning module, an electronic gyroscope, and an IMU (Inertial Measurement Unit). The lidar and acoustic radar are used to monitor for obstacles around the mobile detection platform and report this information to the intelligent control host. After processing, the intelligent control host confirms the current position and attitude of the mobile detection platform through the GPS positioning module, electronic gyroscope, and IMU, and then controls the mobile detection platform to take obstacle avoidance measures. The mobile detection platform can be remotely controlled by a control center or it can use the aforementioned automatic obstacle avoidance driving method. The latter method is preferred because it eliminates the need for remote control by personnel, saving manpower and overcoming the potential delays that may occur with remote control.

[0009] Furthermore, the mobile detection platform is equipped with a BeiDou clock module, which is used to synchronize the electromagnetic acquisition base station, the seismic acquisition base station and the intelligent control host to ensure that the timestamps of the acquired data are consistent.

[0010] Furthermore, the mobile detection platform is equipped with a touch control screen, which is used to display the system status in real time, receive input commands from on-site operators during on-site operation, and visually display the real-time processing results or alarm information of the current detection point.

[0011] The working method of the above-mentioned intelligent detection system for underground karst hazards based on electroseismic effect includes the following steps:

[0012] Step 1: Deployment of the detection system and transient electromagnetic excitation: Plan the driving route and detection points on the ground where urban karst hazard detection is required. The intelligent control host controls the mobile detection platform to drive along the planned driving route. When it reaches the first detection point, it stops. Then, the intelligent control host receives the control command from the control center through the remote communication module, which causes the electromagnetic emission module to excite an electromagnetic pulse signal.

[0013] Step 2, Transient Electromagnetic Reception: The electromagnetic acquisition module collects the electromagnetic signals that return from underground after the electromagnetic pulse signal is transmitted and sends them to the electromagnetic acquisition base station.

[0014] Step 3: Seismic wave reception: When the electromagnetic pulse propagates underground to the karst area, it can act on the mobile ions in the double electric layer of the solid-liquid interface in the karst area. The ions drag the pore fluid through the viscous force, thereby generating physical force on the surrounding rock mass and exciting seismic waves. At this time, the seismic acquisition module collects the seismic wave signal returning from underground and sends it to the seismic acquisition base station.

[0015] Step 4: Data Acquisition in the Detection Area: The intelligent control host acquires data from the seismic acquisition base station and the electromagnetic acquisition base station, completes the data acquisition of the first detection point, and then makes the mobile detection platform continue to travel along the planned route to each subsequent detection point, and repeats steps one to three respectively, thereby completing the data acquisition of all detection points in the detection area.

[0016] Step 5: Data Processing and Identification of Urban Karst Hazards: The intelligent control host performs noise reduction and gain restoration on the electromagnetic and seismic wave data acquired at each detection point, and extracts time-domain, frequency-domain, and time-frequency joint features. The extracted features are input into the constructed anomaly identification model. The model outputs whether there are underground geological anomalies at each detection point. After the identification of all detection points is completed, the identification results are sent to the control center through the remote communication module to generate a three-dimensional distribution map of underground anomalies in the detection area.

[0017] Furthermore, the anomaly identification model constructed in step five is specifically as follows:

[0018] A. Collect historical exploration samples containing electromagnetic data E(t) and seismic wave data S(t) to form a dataset. ,in For the electromagnetic and seismic wave data sequence of the i-th sample, the label is... It indicates whether an anomaly exists (1 indicates existence, 0 indicates non-existence); and preprocesses the electromagnetic data and seismic wave data respectively.

[0019] B. Feature extraction from the dataset:

[0020] Electromagnetic signal characteristics: using exponential fitting Extract the decay time constant Calculate the power spectral density Extracting the main frequency: The time-frequency distribution is obtained through continuous wavelet transform. .

[0021] Seismic wave signal characteristics: extraction of first arrival time t0 and maximum amplitude S max Root mean square amplitude: Calculate the spectral envelope and extract the center frequency f. c and bandwidth B S The spectrum is obtained through short-time Fourier transform: .

[0022] C. Combine the electromagnetic signal features and seismic wave signal features extracted in step B into a fused feature vector: Principal component analysis (PCA) was used to reduce the dimensionality of the features: ,in, To fuse feature vectors; Wpca The transformation matrix of principal component analysis retains 95% of the variance information.

[0023] D. Construct a deep learning-based classification model, which includes an input layer, hidden layers, and an output layer, and determine the cross-entropy loss function to reduce the dimensionality of the feature vector F. pca After training the classification model with the data, an anomaly identification model is obtained.

[0024] Furthermore, the received electromagnetic signals carry the electrical structure of the underground medium, and analysis of these signals can determine whether cavities and water bodies exist underground; the received seismic wave signals carry information on the porosity, permeability, and fluid saturation of the underground medium, and analysis of these signals can determine the density and wave impedance interface of the underground medium. The combination of these two methods ultimately enables the identification of underground anomalies.

[0025] The core innovative principle of this invention is as follows: Research has shown that when a transient electromagnetic emission source emits a differential electromagnetic pulse into the ground, not only can the returned electromagnetic pulse signal be received at the surface, but also the returned seismic wave signal. Further research reveals that during the propagation of the electromagnetic pulse underground, it acts on mobile ions in the electric double layer at the solid-liquid interface of the karst region. These ions drag the pore fluid through viscous forces, thereby generating body force on the solid framework and exciting seismic waves. These seismic waves carry information about the porosity, permeability, and fluid saturation of the medium below the excitation point. The specific relationship between the electromagnetic pulse and the seismic wave can be derived through the following formula:

[0026] A transient electromagnetic pulse is emitted into the ground from an electromagnetic source. The current waveform is a rapidly turning-off differential pulse with a turn-off time τ_off = 4 μs. The magnetic moment of the transmitting coil is M(t) = N·I(t)·A, where N is the number of turns, A is the coil area, and I is the current.

[0027] The eddy current electric field induced by the changing magnetic moment in the underground conductive medium is called E_induced. The strength of this electric field is related to... The correlation is significant, but it decreases with increasing depth and medium conductivity. Within the double electric layer at the solid-liquid interface in karst regions, the induced eddy current electric field E_induced exerts a Coulomb force on mobile ions q in the diffusion layer. Ion motion is mainly dominated by the viscous force γ, and its velocity satisfies γ·(dx / dt)≈q·E_induced. The ion motion drags the pore fluid through viscosity, generating electroosmotic flow.

[0028] The motion of the fluid then generates stress F_seismic on the solid skeleton. This stress, as the source term for exciting elastic waves (seismic waves), satisfies the wave equation:

[0029] .

[0030] The solution to the wave equation characterizes the propagation of seismic waves, with P-wave and S-wave velocities of V_p = √[(λ + 2μ) / ρ] and V_s = √(μ / ρ), respectively. By analyzing the characteristics of the received seismic wave signals, the elastic parameters (λ, μ, ρ) of the medium can be retrieved, thereby determining whether geological anomalies such as karst cavities exist.

[0031] Based on the above research, the technical solution of this application is that only one electromagnetic pulse excitation is needed to obtain the electromagnetic signal reflected from underground and the reverse effect of the electro-vibration effect generated underground due to the propagation of the electromagnetic pulse, thereby generating seismic waves. The electromagnetic signal carries the electrical structure of the underground medium, and its analysis can determine whether there are cavities and water bodies underground. The received seismic wave signal carries information on the porosity, permeability and fluid saturation of the underground medium, and its analysis can determine the density and wave impedance interface of the underground medium. The combination of the two ultimately realizes the judgment of underground anomalies.

[0032] Compared with the prior art, the present invention has the following advantages:

[0033] 1. This invention, through research, discovers that by using an electromagnetic transmission module to emit an electromagnetic pulse signal into the underground of the detection area, and an electromagnetic acquisition module to receive the electromagnetic pulse signal transmitted from underground and feed it back to the electromagnetic acquisition base station, and by using a seismic acquisition module to receive the seismic wave signal generated underground after the electromagnetic pulse signal is emitted and feed it back to the seismic acquisition base station, the intelligent control host can synchronize with the two base stations to ensure that the timestamps of the data collected by both are consistent. In this way, only one electromagnetic pulse excitation is required to receive two types of detection data (i.e., electromagnetic data and seismic data). Moreover, the two types of data carry different geological information, and no additional excitation equipment is required. This process of simultaneously collecting two types of data ensures that the data is collected under the same spatiotemporal conditions, providing accurate data support for subsequent non-destructive, high-efficiency, and high-precision identification of urban underground hazards.

[0034] 2. This invention employs a self-constructed anomaly identification model during data processing. It can analyze two types of data separately and comprehensively determine whether an anomaly exists. This not only eliminates the need for human intervention but also achieves good identification accuracy, thereby providing real-time data support for urban safety management. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of the overall external structure of the present invention.

[0036] Figure 2 yes Figure 1 Internal structure diagram.

[0037] Figure 3 yes Figure 1A schematic diagram of the electromagnetic acquisition module.

[0038] Figure 4 yes Figure 1 A schematic diagram of the structure of the seismic acquisition module.

[0039] In the diagram, 1. LiDAR, 2. Touch control screen, 3. Vehicle lights, 4. Electromagnetic emission module, 5. Acoustic radar, 6. Seismic acquisition module, 7. Remote communication module, 8. GPS positioning module, 9. Battery and power control system, 10. Beidou clock module, 11. Electronic gyroscope, 12. Electromagnetic acquisition base station, 13. Seismic acquisition base station, 14. Intelligent control host, 15. Electromagnetic acquisition module, 16. First electric actuator, 17. Seismic acquisition unit, 18. Second electric actuator, 19. Seismic detector, 20. Motion control system. Detailed Implementation

[0040] The present invention will be further described below.

[0041] like Figure 1 and Figure 2 As shown, an intelligent detection system for underground karst hazards based on electroseismic effects includes a mobile detection platform and a control center. The mobile detection platform is equipped with an intelligent control host 14, a remote communication module 7, an electromagnetic transmission module 4, an electromagnetic acquisition module 15, a seismic acquisition module 6, and a battery and power control system 9. The electromagnetic transmission module 4 is installed in the middle of the mobile detection platform and is used to generate electromagnetic pulse signals downward from the mobile detection platform. The electromagnetic pulse emission source adopts a SiCMOSFET transmitter, and the off-time is controlled at 4μs to generate a higher di / dt, thereby enhancing the excited electric field strength. Moreover, the pulse waveform it excites adopts a differential pulse, which uses its harmonic components to excite broadband seismic waves. The pulse transmission power is 100 watts to 2 kilowatts, which can be switched in multiple levels. The transmission frequency is 100 Hz to 10 kHz.

[0042] like Figure 3As shown, there are multiple electromagnetic acquisition modules 15 mounted on the mobile detection platform, and these modules are on the same horizontal plane as the electromagnetic transmission module 4. They are used to receive electromagnetic signals returned from underground after the electromagnetic pulse signal is transmitted and send them to the electromagnetic acquisition base station 12. The seismic acquisition module 6 includes multiple seismic acquisition units 17 deployed at the bottom of the mobile detection platform. These units are used to receive seismic wave signals generated underground after the electromagnetic pulse signal is transmitted and send them to the seismic acquisition base station 13. The mobile detection platform is equipped with a Beidou clock module 10, which is used to synchronize the electromagnetic acquisition base station 12, the seismic acquisition base station 13, and the intelligent control host 14 to ensure that the timestamps of the acquired data are consistent. The mobile detection platform is equipped with a touch control screen 2, which is used to display the system status in real time, receive input commands from on-site operators during on-site operations, and visually display the real-time processing results or alarm information of the current detection point. The intelligent control host 14 is connected to the seismic acquisition base station 13 and the electromagnetic acquisition base station 12. The intelligent control host 14 receives control commands from the control center through the remote communication module, controls the movement of the mobile detection platform through the motion control system 20 (i.e., the motor and drive wheels), controls the excitation of the electromagnetic emission module 4, and receives data fed back from the seismic acquisition base station 13 and the electromagnetic acquisition base station 12. After analysis and processing by the built-in model, it automatically identifies underground anomalies and feeds them back to the control center through the remote communication module to generate a three-dimensional anomaly distribution map. The battery and power control system 9 is used to supply power to the above modules.

[0043] As an improvement of the present invention, there are five electromagnetic acquisition modules 15 in total, one of which is located at the center of the electromagnetic emission module 4, and the other four are evenly distributed around the electromagnetic emission module 4. The electromagnetic acquisition module 15 is an electromagnetic expandable telescopic coil, and the four electromagnetic expandable telescopic coils on the periphery extend out of the moving detection platform during detection and retract into the moving detection platform when not detecting, respectively, via the first electric push rod 16. The standard mode of the electromagnetic expandable telescopic coil is designed as 0.5m × 0.5m, and a modular interface is added so that the size of the coil can be manually expanded as needed.

[0044] like Figure 4As shown, the seismic acquisition units 17 are arranged in an array under the mobile detection platform. Each seismic acquisition unit 17 includes a second electric actuator 18 and a seismic detector 19. The fixed end of the second electric actuator 18 is connected to the lower part of the mobile detection platform, and the seismic detector 19 is mounted on the telescopic end of the second electric actuator 18. When not detecting, the second electric actuator 18 is in the retracted state, and the seismic detector 19 is not in contact with the ground. When detecting, the second electric actuator 18 is in the extended state, and the seismic detector 19 is in close contact with the ground. Since the earthquakes generated by electromagnetic pulses are relatively weak, the seismic detector 19 uses a three-component MEMS accelerometer. Range: ±2g or ±5g. Noise density: <100µg / √Hz. Bandwidth: DC (0Hz to 800Hz). This effectively ensures the accuracy of receiving seismic wave signals.

[0045] As another improvement of the present invention, the mobile detection platform is equipped with a lidar 1, an acoustic radar 5, a GPS positioning module 8, an electronic gyroscope 11, and an IMU inertial measurement unit. The lidar 1 and the acoustic radar 5 are used to monitor whether there are obstacles around the mobile detection platform and feed the feedback to the intelligent control host 14. After processing, the intelligent control host 14 confirms the current position and attitude of the mobile detection platform through the GPS positioning module 8, the electronic gyroscope 11, and the IMU inertial measurement unit, and controls the mobile detection platform to take obstacle avoidance measures.

[0046] The working method of the above-mentioned intelligent detection system for underground karst hazards based on electroseismic effect includes the following steps:

[0047] Step 1, Deployment of the detection system and transient electromagnetic excitation: Plan the driving route and detection points on the ground where urban karst hazard detection is required. The intelligent control host 14 controls the mobile detection platform to drive along the planned driving route. When it reaches the first detection point, it stops. Then, the intelligent control host 14 receives the control command from the control center through the remote communication module 7, so that the electromagnetic emission module 4 excites the electromagnetic pulse signal.

[0048] Step 2, Transient electromagnetic reception: The electromagnetic acquisition module 15 acquires the electromagnetic signal that returns from underground after the electromagnetic pulse signal is transmitted and sends it to the electromagnetic acquisition base station 12;

[0049] Step 3: Seismic wave reception: When the electromagnetic pulse propagates underground to the karst area, it can act on the mobile ions in the double electric layer of the solid-liquid interface in the karst area. The ions drag the pore fluid through the viscous force, thereby generating physical force on the surrounding rock mass and exciting seismic waves. At this time, the seismic acquisition module 6 collects the seismic wave signal returning from underground and sends it to the seismic acquisition base station 13.

[0050] Step 4: Data Acquisition in the Detection Area: The intelligent control host 14 acquires data from the seismic acquisition base station 13 and the electromagnetic acquisition base station 12, completing the data acquisition of the first detection point. Then, the mobile detection platform continues to travel along the planned route to each subsequent detection point, and steps one to three are repeated to complete the data acquisition of all detection points in the detection area. The received electromagnetic signals carry the electrical structure of the underground medium, and analysis of these signals can determine whether there are cavities or water bodies underground. The received seismic wave signals carry information on the porosity, permeability, and fluid saturation of the underground medium, and analysis of these signals can determine the density and wave impedance interface of the underground medium. The combination of these two factors ultimately enables the identification of underground anomalies.

[0051] Step 5: Data Processing and Identification of Urban Karst Hazards: The intelligent control host 14 performs denoising and gain restoration on the electromagnetic and seismic wave data acquired at each detection point, and extracts time-domain, frequency-domain, and time-frequency joint features to construct an anomaly identification model, specifically:

[0052] A. Collect historical exploration samples containing electromagnetic data E(t) and seismic wave data S(t) to form a dataset. ,in For the electromagnetic and seismic wave data sequence of the i-th sample, the label is... This indicates whether an anomaly exists (1 indicates presence, 0 indicates absence); and preprocessing is performed separately on the electromagnetic data and seismic wave data, specifically:

[0053] Electromagnetic signal preprocessing , where μ E σ E These represent the mean and standard deviation of the electromagnetic signal, respectively; seismic wave signal preprocessing: , where T is the signal duration.

[0054] B. Feature extraction from the dataset:

[0055] Electromagnetic signal characteristics: using exponential fitting Extract the decay time constant Calculate the power spectral density Extracting the main frequency: The time-frequency distribution is obtained through continuous wavelet transform. Where E0 is the initial amplitude of the electromagnetic signal; F{·} is the Fourier transform operator; denoted as the mother wavelet function; a represents the wavelet scaling parameter system; and b represents the wavelet translation parameter.

[0056] Seismic wave signal characteristics: extraction of first arrival time t0 and maximum amplitude S max Root mean square amplitude: Calculate the spectral envelope and extract the center frequency f. cand bandwidth B S The spectrum is obtained through short-time Fourier transform: ; where w is the window function.

[0057] C. Combine the electromagnetic signal features and seismic wave signal features extracted in step B into a fused feature vector: Principal component analysis (PCA) was used to reduce the dimensionality of the features: ,in, To fuse feature vectors; W pca The transformation matrix of principal component analysis retains 95% of the variance information.

[0058] D. Construct a classification model based on deep learning, which includes an input layer, a hidden layer, and an output layer;

[0059] Input layer: Receives the dimensionality-reduced feature vector F pca ;

[0060] Hidden layer: Employs a two-layer fully connected network;

[0061]

[0062]

[0063] Output layer: Outputs classification probabilities using the softmax function;

[0064]

[0065] The cross-entropy loss function is determined as follows:

[0066]

[0067] In the above formulas, W i Let b be the weight vector. i Let y be the bias vector, ReLU be the linear unit activation function, N be the total number of training samples, and y be the bias vector. i For real labels, L represents the cross-entropy loss value, used to predict probabilities.

[0068] The dimensionality-reduced feature vector F pca The model was trained using the Adam optimizer classification model and then tested and verified to obtain an anomaly identification model.

[0069] The extracted features are input into the constructed anomaly identification model. The model outputs whether there are underground geological anomalies at each detection point. After the identification of all detection points is completed, the identification results are sent to the control center through the remote communication module 7 to generate a three-dimensional distribution map of underground anomalies in the detection area.

[0070] The above-mentioned anomaly identification model can add newly collected and manually verified data to the training set, and perform incremental learning on the model periodically to update the model. The specific formula is as follows:

[0071]

[0072] In the formula, The old loss function value; λ represents the new loss function value; λ is the regularization hyperparameter.

[0073] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A working method for an intelligent detection system for underground karst hazards based on electroseismic effects, characterized in that, Includes the following steps: Step 1: Deployment of the detection system and transient electromagnetic excitation: Plan the driving route and detection points on the ground where urban karst hazard detection is required. The intelligent control host controls the mobile detection platform to drive along the planned driving route. When it reaches the first detection point, it stops. Then, the intelligent control host receives the control command from the control center through the remote communication module, which causes the electromagnetic emission module to excite the electromagnetic pulse signal. Step 2, Transient Electromagnetic Reception: The electromagnetic acquisition module acquires the electromagnetic signals that return from underground after the electromagnetic pulse signal is transmitted and sends them to the electromagnetic acquisition base station; the electromagnetic signals carry the electrical structure of the underground medium, and analysis of them can determine whether there are cavities and water bodies underground; Step 3: Seismic Wave Reception: When the electromagnetic pulse propagates underground to the karst region, it acts on the mobile ions in the double electric layer at the solid-liquid interface of the karst region. The ions drag the pore fluid through viscous forces, thereby generating body force on the surrounding rock mass and exciting seismic waves. At this time, the seismic acquisition module collects the seismic wave signals returning from underground and sends them to the seismic acquisition base station. The seismic wave signals carry information on the porosity, permeability, and fluid saturation of the underground medium. By analyzing them, the density and wave impedance interface of the underground medium can be determined. Step 4: Data Acquisition in the Detection Area: The intelligent control host acquires data from the seismic acquisition base station and the electromagnetic acquisition base station, completes the data acquisition of the first detection point, and then makes the mobile detection platform continue to travel along the planned route to each subsequent detection point, and repeats steps one to three respectively, thereby completing the data acquisition of all detection points in the detection area. Step 5: Data Processing and Identification of Urban Karst Hazards: The intelligent control host performs denoising and gain restoration on the electromagnetic and seismic wave data acquired at each detection point, and extracts time-domain, frequency-domain, and time-frequency joint features. These extracted features are input into the constructed anomaly identification model. The model outputs whether geological anomalies exist at each detection point. After identifying all detection points, the identification results are sent to the control center via a remote communication module, generating a 3D distribution map of underground anomalies in the detection area. The specific construction process of the anomaly identification model is as follows: A. Collect historical detection samples containing electromagnetic data E(t) and seismic wave data S(t) to form a dataset; and preprocess the electromagnetic data and seismic wave data respectively; B. Feature extraction from the dataset: Electromagnetic signal characteristics: The decay time constant is extracted by exponential fitting, the power spectral density is calculated, the dominant frequency is extracted, and the time-frequency distribution is obtained by continuous wavelet transform; Seismic wave signal characteristics: first arrival time, maximum amplitude, and root mean square amplitude are extracted; spectral envelope is calculated; center frequency and bandwidth are extracted; and the spectrum is obtained through short-time Fourier transform. C. Combine the electromagnetic signal features and seismic wave signal features extracted in step B into a fused feature vector, and use principal component analysis to reduce the dimensionality of the features: ,in, To fuse feature vectors; W PCA is the transformation matrix for principal component analysis. D. Construct a deep learning-based classification model, which includes an input layer, hidden layers, and an output layer, and determine the cross-entropy loss function to reduce the dimensionality of the feature vector F. pca After training the classification model with the data, an anomaly identification model is obtained.

2. The working method according to claim 1, characterized in that, The intelligent detection system for underground karst hazards based on the electroseismic effect includes a mobile detection platform and a control center. The mobile detection platform is equipped with an intelligent control host, a remote communication module, an electromagnetic transmission module, an electromagnetic acquisition module, a seismic acquisition module, a battery and power control system, and a motion control system. The electromagnetic transmission module, located in the middle of the mobile detection platform, is used to generate electromagnetic pulse signals downwards from the platform. Multiple electromagnetic acquisition modules are mounted on the mobile detection platform at the same horizontal level as the electromagnetic transmission module. These modules receive the electromagnetic signals returning from underground after the electromagnetic pulse signals are transmitted and transmit them to the electromagnetic acquisition base station. The seismic acquisition module includes modules deployed on the mobile platform... Multiple seismic acquisition units at the bottom of the detection platform are used to receive seismic wave signals generated underground after electromagnetic pulse signal transmission and send them to the seismic acquisition base station. The intelligent control host is connected to the seismic acquisition base station and the electromagnetic acquisition base station. The intelligent control host receives control commands from the control center through a remote communication module, controls the movement of the mobile detection platform through the motion control system, controls the activation of the electromagnetic transmission module, and receives data feedback from the seismic acquisition base station and the electromagnetic acquisition base station. After analysis and processing by the built-in model, it automatically identifies underground anomalies and feeds them back to the control center through the remote communication module to generate a three-dimensional anomaly distribution map. The battery and power control system is used to supply power to the above modules.

3. The working method according to claim 2, characterized in that, There are five electromagnetic acquisition modules in total. One of them is located in the center of the electromagnetic emission module, and the other four are evenly distributed around the electromagnetic emission module. The electromagnetic acquisition module is an electromagnetic expandable telescopic coil. The four electromagnetic expandable telescopic coils on the periphery extend out of the mobile detection platform during detection and retract into the mobile detection platform when not detecting, respectively, through the first electric push rod.

4. The working method according to claim 2, characterized in that, The seismic acquisition unit includes a second electric actuator and a seismic detector. The fixed end of the second electric actuator is connected to the lower part of the mobile detection platform. The seismic detector is installed on the telescopic end of the second electric actuator. When no detection is being performed, the second electric actuator is in the retracted state, and the seismic detector is not in contact with the ground. When detection is being performed, the second electric actuator is in the extended state, and the seismic detector is in close contact with the ground.

5. The working method according to claim 2, characterized in that, The mobile detection platform is equipped with lidar, acoustic radar, GPS positioning module, electronic gyroscope, and IMU inertial measurement unit. The lidar and acoustic radar are used to monitor whether there are obstacles around the mobile detection platform and feed the feedback to the intelligent control host. After processing, the intelligent control host confirms the current position and attitude of the mobile detection platform through the GPS positioning module, electronic gyroscope, and IMU inertial measurement unit, and controls the mobile detection platform to take obstacle avoidance measures.

6. The working method according to claim 2, characterized in that, The mobile detection platform is equipped with a Beidou clock module, which is used to synchronize the electromagnetic acquisition base station, the seismic acquisition base station and the intelligent control host to ensure that the timestamps of the acquired data are consistent.

7. The working method according to claim 2, characterized in that, The mobile detection platform is equipped with a touch control screen.

Images