A distributed cloud-edge-end collaborative fusion prediction and early warning system for karst collapse

Abstract

The present invention discloses a distributed cloud-edge-end collaborative fusion prediction and early warning system for karst collapse, in which a polymer measuring tube is laid below the critical depth of stability of a karst soil cave, a flexible conductive film array is arranged on the surface of the polymer measuring tube, and distributed optical fibers are respectively embedded and encapsulated in the guide grooves of the neutral axis on both sides of the polymer measuring tube. The edge resistance change rate dynamic signal processing module collects the dynamic signal of the resistance change rate of the flexible conductive film array on the surface of the measuring tube, and the edge distributed optical fiber acoustic sensor demodulation module collects the distributed optical fiber acoustic vibration signal. The edge multivariate data fusion module fuses the dynamic resistance signal and the acoustic vibration signal based on spatial positioning. The cloud-based prediction and early warning module identifies the activation or activation stop behavior of the karst soil cave development based on the fusion signal, and predicts the activation rate and development depth of the soil cave development. When the soil cave development depth will reach the soil cave critical depth or the soil cave boundary will develop to the control position, a warning signal is issued to achieve early warning of karst collapse disasters.

Description

Technical Field

[0001] The invention relates to a geological disaster monitoring and early warning system and method, and in particular to a karst collapse prediction and early warning system. Background Art

[0002] Karst collapse is a common geological disaster phenomenon, which refers to the formation of underground cavities due to long-term dissolution of groundwater in areas with soluble rocks such as limestone. When the upper covering layer loses support, the ground collapses. The main causes of karst collapse include over-exploitation of groundwater, vibration of traffic loads, and foundation instability and soil disturbance caused by engineering construction (including underground engineering construction, excessive loads on ground buildings, or dense high-rise buildings). Karst collapse is mostly caused by the development of soil caves in the covering layer, which leads to the overall destruction of the soil from the top of the soil cave to the surface. The development of soil caves is hidden in space, and karst collapse is sudden in time. The signs of ground deformation before collapse are often not obvious, which can cause serious casualties and property losses. When using traditional geophysical methods such as precision leveling, displacement meters, subsidence markers or lidar to monitor karst collapse, it is difficult to determine the target area due to the small surface deformation caused by the development of underground soil caves. However, it is difficult to achieve automated real-time monitoring and unmanned remote sensing using geological radar, high-density point method, cross-hole seismic CT and ultrasonic methods. Therefore, real-time monitoring and accurate prediction and early warning of karst collapse have always been a problem.

[0003] Chinese Patent 202111008923.8 discloses a distributed flexible edge intelligent early warning and forecasting system and method for karst ground collapse, the system includes a polymer measuring tube, a flexible conductive film, a signal processing module, a multivariate data fusion and an early warning and forecasting module; the polymer measuring tube is laid below the critical depth of the karst soil cave by horizontal directional drilling, and a plurality of flexible conductive films and metal electrodes on both sides thereof are arranged on the surface; the metal electrodes are connected to the signal processing module through wires; the signal processing module obtains the resistance signal of each flexible conductive film on the measuring tube and calculates the bending and shear deformation of the measuring tube, and estimates the deformation field distribution of the soil around the measuring tube and the boundary of the soil cave through the multivariate Taylor series expansion Kalman filtering algorithm in the multivariate data fusion module, and predicts the soil cave depth and boundary development rate through the Bayesian optimization random forest Kalman filtering algorithm in the early warning and forecasting module. When the predicted soil cave development depth reaches the critical depth of the soil cave, the karst ground collapse early warning signal is triggered. The system and method only consider the prediction of ground collapse using the static resistance signal measured by the flexible conductive film. In this scheme, the flexible conductive film is more effective when the formation is greatly deformed, but it is not easy to monitor the data when the formation is only slightly deformed in the early stage of karst soil cave development. The limitation of single data affects the reliability of the prediction results, and also affects the effectiveness of the identification of karst collapse disaster precursors and early warning. Summary of the invention

[0004] Purpose of the invention: In view of the above-mentioned existing technologies, a distributed cloud-edge-end collaborative fusion prediction and early warning system for karst collapse is proposed, which is suitable for situations where slight and large deformations occur in the karst overlying strata, and improves the effectiveness and reliability of karst collapse disaster precursor identification and early warning.

[0005] Technical solution: A distributed cloud-edge-end collaborative fusion prediction and early warning system for karst collapse, including: polymer measuring tube, flexible conductive film array, optical fiber, edge resistance change rate dynamic signal processing module, edge distributed optical fiber acoustic sensor demodulation module, edge multivariate data fusion module, cloud prediction and early warning module;

[0006] Guide grooves are provided on the central axis of both sides of the polymer measuring tube, and optical fibers are respectively embedded in the guide grooves and then packaged, and the flexible conductive film array is arranged on the surface of the polymer measuring tube; the polymer measuring tube is horizontally laid below the critical depth of soil hole stability, and each flexible conductive film is connected to the edge resistance change rate dynamic signal processing module through a wire, and the end of the optical fiber is connected to the edge distributed optical fiber acoustic sensor demodulation module;

[0007] The edge resistance change rate dynamic signal processing module is used to collect the resistance change rate dynamic signal of the flexible conductive film array on the surface of the measuring tube; the edge distributed optical fiber acoustic sensing demodulation module is used to collect the acoustic vibration signal of the optical fiber packaged in the guide grooves on both sides of the measuring tube;

[0008] The edge multivariate data fusion module is used to fuse the received dynamic signal of resistance change rate with the acoustic vibration signal based on spatial positioning, and upload the fused signal to the cloud prediction and warning module;

[0009] The cloud-based prediction and warning module is used to identify the activation or activation stop behavior of karst soil cave development based on the fusion signal, and predict the soil cave development activation rate and development depth based on the fusion signal. When it is predicted that the soil cave development depth will reach the soil cave critical depth or the soil cave boundary will develop to the control position within the warning time, the cloud-based prediction and warning module issues a warning signal.

[0010] Furthermore, a plurality of monitoring sections are arranged on the polymer measuring tube, each section is equidistantly spaced, and a group of flexible conductive films are arranged on each monitoring section, and the flexible conductive films on each monitoring section together form the flexible conductive film array; wherein, a group of flexible conductive films is composed of conductive films respectively coated on the top and bottom surfaces of the polymer measuring tube along the axial direction, and conductive films respectively coated on both sides of the polymer measuring tube obliquely; the conductive films on the top and bottom surfaces are used to sense the tensile and compressive dynamic strains of the measuring tube to characterize the development time and position of the soil hole, and the oblique conductive films on the two side surfaces are used to sense the shear dynamic strains of the measuring tube to characterize the boundaries on both sides of the soil hole development; each electrode in each group of flexible conductive films is connected to the edge resistance change rate dynamic signal processing module through a wire.

[0011] Furthermore, the edge multivariate data fusion module performs fusion based on spatial positioning, specifically including: for any monitoring section, the resistance signal data of each flexible conductive film is fused through Kalman filtering to obtain the fused conductive film data of the monitoring section, that is, data one; the acoustic vibration signal data of each optical fiber channel measured by the edge distributed optical fiber acoustic sensor demodulation module is pre-processed, and the data is denoised by wavelet denoising to obtain data two; the data one corresponding to the monitoring section is fused with the data two of the optical fiber channel adjacent to the monitoring section, and the data fusion adopts the classical particle filter equation after Taylor series decomposition.

[0012] Furthermore, the cloud-based prediction and early warning module predicts the activation rate and development depth of the soil hole development based on the fusion signal, specifically including: using the moving average method to divide the displacement of the soil hole top plate into trend items and random items according to the time series, using the particle filter algorithm to predict the displacement of the trend item, and using the XGBoost algorithm with the hyperparameters of the model optimized by the sparrow algorithm to predict the displacement of the random item; the displacement of the trend item and the displacement of the random item together constitute the predicted value of the cumulative displacement of the soil hole top plate, and finally predicting the activation rate and development depth of the soil hole development based on the time series data fused by the multivariate data;

[0013] The cloud-based prediction and early warning module is used to identify the activation or activation stop behavior of karst soil cave development, specifically including: judging the activation or activation stop behavior of soil cave development according to the predicted cumulative displacement value of the soil cave roof; when the predicted cumulative displacement value of the soil cave roof shows continuous peak segments within a certain period of time and the displacement remains greater than 0, it is judged that the soil cave is activated at this time; when the predicted cumulative displacement value of the soil cave roof does not fluctuate significantly with the increase of time or shows a significant decrease compared with the soil cave development activation segment in the previous period of time and approaches 0, it is judged that the soil cave development activation has stopped.

[0014] Furthermore, when the cloud-based prediction and early warning module identifies that the karst soil cave has developed and activated, the edge end is controlled to increase the signal collection frequency; when the cloud-based prediction and early warning module identifies that the karst soil cave has stopped being activated, the edge end is controlled to reduce the signal collection frequency.

[0015] Furthermore, the polymer measuring tube is a PE round tube with an outer diameter greater than 60 mm, and the length of the polymer measuring tube is greater than 1.5 m; each section of the flexible conductive film is greater than 40 mm in length and greater than 10 mm in width.

[0016] Furthermore, the flexible conductive film is a graphene-based flexible conductive film, and zinc foil is used as a reducing agent to reduce graphene oxide by substitution reaction, and the reduced graphene oxide composite material is prepared and then added to a polyurethane matrix to prepare the graphene-based flexible conductive film.

[0017] Beneficial effects: The prediction and early warning system provided by the present invention fuses the dynamic signal of the resistance change rate measured by the flexible conductive film with the acoustic vibration signal measured by the optical fiber. The cloud-based prediction and early warning module identifies the activation or activation stop behavior of the karst soil cave development based on the uploaded fusion signal, and predicts the activation rate and development depth of the soil cave development, so as to dynamically track and warn of progressive collapse and destruction, and ultimately achieve early warning of karst collapse disasters. Among them, the edge distributed optical fiber acoustic sensing demodulation module senses the collapse behavior of the soil block on the top of the soil cave through the acoustic vibration signal, and the sensing data is used to identify whether the soil cave is activated, and the activation rate and development depth of the soil cave are predicted through the acoustic vibration effect generated by the slight deformation of the overlying strata of the soil cave; and the dynamic information of the resistance of the flexible conductive film can effectively verify the activation behavior of the soil cave development, and the fusion of the dynamic signal of the resistance change rate and the acoustic vibration signal of the optical fiber improves the reliability of identifying the activation behavior of the soil cave development.

[0018] Specifically, multiple monitoring sections are set on the polymer measuring tube, and each section is equidistantly spaced. Four flexible conductive films are arranged on each monitoring section, wherein the top and bottom surfaces are coated with a group of conductive films along the axial direction to sense the tensile and compressive strains of the measuring tube to characterize the development time and position of the soil hole, and the two sides are obliquely coated with flexible conductive films to sense the shear strain of the measuring tube to characterize the boundaries of the two sides of the soil hole development. The prior art 202111008923.8 mentioned in the background technology uses both oblique and circumferentially coated conductive films on the sides. In the experiment, the resistance change rate signal measured by the annular conductive film is insensitive to the development time and position of the soil hole or the boundary characterization signals on both sides of the soil hole. At the same time, it is found that in a specific application scenario, when the soil hole expands and the top plate of the soil hole continues to develop upward, the measuring tube produces settlement deformation, but due to the small annular strain generated on both sides, compared with the axial coating on the top and bottom surfaces and the oblique coating on both sides of the conductive film, the resistance change rate amplitude measured by the side annularly coated conductive film is not obvious, and the sensing accuracy is insufficient. Therefore, due to the poor accuracy of its own data, the accuracy of the results after data fusion is also affected, which may increase the monitoring error. The present invention omits the side annularly coated conductive film with insignificant effect on the basis of the above reasons, thereby reducing the monitoring error.

[0019] During the development and activation of soil caves, the overlying soil layer develops local catastrophic events such as soil block collapse and dynamic changes in water (gas) pressure. Therefore, the present invention introduces acoustic vibration signals, dynamically senses the acoustic vibration signals along the measuring pipe through distributed optical fibers, and performs multivariate data fusion with the dynamic signals of the resistance change rate. While considering both stratum deformation and internal acoustic vibration characteristics, the limitations of a single data source are overcome, the reliability of the sensing results is improved, and early identification of the precursors of soil cave collapse disasters is achieved. In addition, the optical fiber is encapsulated in the guide groove along the neutral axis on both sides of the measuring pipe. Since the axial strain of the neutral axis on both sides of the measuring pipe is close to 0, even if the soil cave continues to develop and causes the measuring pipe to produce a large bending deformation, the encapsulated optical fiber will not be broken; and the optical fiber and the measuring pipe are coupled as one through encapsulation, and the measuring pipe expands the contact range between the optical fiber and the stratum, thereby improving the sensitivity of acoustic vibration signal perception. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic diagram of the three-dimensional structure of the flexible conductive film and optical fiber arrangement in the system of the present invention;

[0021] Figure 2 It is a schematic diagram of the cross-sectional structure of the flexible conductive film and optical fiber arrangement in the system of the present invention;

[0022] Figure 3 It is a schematic diagram of the underground arrangement of the measuring pipes and the system structure in the system of the present invention;

[0023] Figure 4 is a resistance change rate-strain correlation curve of the flexible conductive film used in the system of the present invention;

[0024] Figure 5 It is a data curve measured by the distributed acoustic sensor demodulation module in the system of the present invention;

[0025] Figure 6 It is a comparison chart of the dynamic signal of the resistance change rate of a part of the cross section and the curve generated by the acoustic vibration signal data of each channel of the optical fiber in the same coordinate system;

[0026] Figure 7 It is a flow chart of data fusion performed by the multivariate data fusion module in the system of the present invention;

[0027] Figure 8 It is a flow chart of the multivariate Taylor series expansion particle filtering algorithm in the multivariate data fusion module of the system of the present invention;

[0028] Fig. 9 It is a flow chart of the SSA-XGBoost particle filter prediction algorithm in the early warning and forecasting module of the system of the present invention;

[0029] Fig.10 It is a data curve chart of the dynamic signal of the cross-section resistance change rate and the acoustic vibration signal data and their fusion. DETAILED DESCRIPTION

[0030] The present invention will be further explained below in conjunction with the accompanying drawings.

[0031] A distributed cloud-edge-end collaborative fusion prediction and early warning system for karst collapse, the specific application scenario and the intelligent early warning and forecasting system structure adopted by the present invention are as follows Figure 3 As shown in the figure, near the potential karst collapse target area to be monitored, the polymer measuring tube is horizontally laid below the critical depth of the soil cave, and the tube body is kept parallel to the ground. When the soil cave develops and activates, the soil blocks on the soil cave arch gradually collapse, and the soil cave roof continues to develop upward. An air-facing area is gradually formed below the measuring tube and the load is applied above. At this time, the bottom of the measuring tube will gradually approach the soil cave arch and produce bending deformation; at the same time, there is a relative deformation trend between the soil cave boundary, the soil cave roof and the soil on both sides, which causes shear deformation of the measuring tube.

[0032] like Figure 1 , Figure 2 As shown, guide grooves 5 are respectively provided on the central axis of both sides of the polymer measuring tube 1, and an optical fiber 3 is respectively embedded in the guide grooves 5, and the optical fiber 3 is a common single-mode communication optical fiber. Multiple monitoring sections are set on the polymer measuring tube 1, and each section is equally spaced, and a group of flexible conductive films are arranged on each monitoring section. Among them, a group of flexible conductive films is composed of conductive films 4 coated on the top and bottom surfaces of the polymer measuring tube 1 along the axial direction, and conductive films 2 obliquely coated on both sides of the polymer measuring tube 1 at an angle of 45°, and the conductive films 2 obliquely coated on both sides of the polymer measuring tube 1 at an angle of 45° just cross the surface of the guide groove 5. When the measuring tube is bent or sheared, the resistance change rate of the flexible conductive film and the optical fiber acoustic vibration signal change. The conductive film 4 on the top and bottom surfaces is used to sense the tensile and compressive strains of the measuring tube to characterize the time and position of the soil hole development, and the conductive films 2 obliquely coated on both sides at an angle of 45° are used to sense the shear strain of the measuring tube to characterize the boundaries on both sides of the soil hole development. The polymer measuring tube 1, flexible conductive films 2 and 4, and optical fiber 3 belong to the end-side equipment in this system.

[0033] The edge devices in this system include an edge resistance change rate dynamic signal processing module, an edge distributed fiber optic acoustic sensing demodulation module, and an edge multivariate data fusion module. Each flexible conductive film is connected to the edge resistance change rate dynamic signal processing module through electrodes on both sides. In this embodiment, one edge resistance change rate dynamic signal processing module can be connected to multiple groups of flexible conductive films. The optical fibers 3 on both sides of the polymer measuring tube 1 are connected to the edge distributed fiber optic acoustic sensing demodulation module. A cloud prediction and early warning module is deployed on the cloud of this system. Specifically, the edge resistance change rate dynamic signal processing module obtains the resistance change rate dynamic signal of the flexible conductive films 2 and 4 on the measuring tube and uploads it to the edge multivariate data fusion module; the distributed acoustic sensing demodulation module obtains the acoustic vibration signal of the optical fiber 3 on the measuring tube and uploads it to the edge multivariate data fusion module. The multivariate data fusion module fuses the received dynamic signal of resistance change rate with the fiber acoustic vibration signal to identify the development and activation behavior of soil caves, and predicts the development activation rate and development depth of soil caves. When it is predicted that the development depth of soil caves will reach the critical depth of soil caves or the boundary of soil caves will develop to the control position within the warning time, the cloud prediction and warning module issues a warning signal to restrict human activities within the scope of ground collapse, and realize the early warning of karst collapse disasters. Among them, the cloud prediction and warning module is realized by artificial intelligence algorithm, so as to realize the "cloud-edge-end" collaborative architecture of the system of the present invention.

[0034] In the above system, the edge multivariate data fusion module fuses the dynamic signal of the resistance change rate of the conductive film with the optical fiber acoustic vibration signal through the multivariate Taylor series expansion particle filtering algorithm. Figure 7 As shown, data fusion specifically includes: for any monitoring section, the dynamic signal data of the resistance change rate of the axial conductive film 4 and the oblique conductive film 2 are fused through Kalman filtering to obtain the fused conductive film data of the monitoring section, that is, data 1; the acoustic vibration signal data of each channel of the optical fiber 3 measured by the edge distributed optical fiber acoustic sensor demodulation module are pre-processed (de-trending and de-averaging), and then the data are denoised through wavelet denoising to obtain data 2; the data 1 of the monitoring section and the data 2 of the optical fiber channel adjacent to the monitoring section are subjected to multivariate data fusion through the particle filter equation after Taylor series decomposition. Figure 8 It is a flow chart of the multivariate Taylor series expansion particle filtering algorithm in the multivariate data fusion module of the system of the present invention.

[0035] like Fig. 9As shown in the figure, the activation rate and development depth of soil holes are predicted by the cloud prediction and early warning module, which specifically includes: using the moving average method to divide the displacement of the soil hole top plate into trend items and random items according to the time series, using the particle filter algorithm to predict the displacement of the trend item, and using the XGBoost algorithm after the sparrow algorithm (SSA) optimizes the model hyperparameters to predict the displacement of the random item under the consideration of external influencing factors (such as rainfall, external disturbance, collapse deformation); the trend item and the random item together constitute the cumulative prediction value of the displacement of the soil hole top plate, and finally predict the activation rate and development depth of the soil hole according to the time series data of the multivariate data fusion. The measured data are used to construct the training set and the validation set, the training set is used to train the model in the cloud, and the validation set is used to verify the model accuracy.

[0036] In this embodiment, the polymer measuring tube 1 is a PVC round tube with an outer diameter greater than 60 mm, the length of the polymer measuring tube is greater than 1.5 m, the length of each section of the flexible conductive film is greater than 40 mm, and the width is greater than 10 mm. According to the actual needs of the on-site prediction and early warning target area, multiple sections of the above-mentioned polymer measuring tube 1 can also be connected by electric fusion. The flexible conductive film is a graphene-based flexible conductive film. Zinc foil is used as a reducing agent, and graphene oxide is reduced by substitution reaction. After the reduced graphene oxide composite material is prepared, it is added to the polyurethane matrix to prepare the graphene-based flexible conductive film. The additive silicon-based three-component modified polysiloxane conductive coating is used, and the relative deformation of the flexible conductive film can reach 70%-90%, which can ensure that the flexible conductive film can still work effectively after the measuring tube undergoes a large deformation in the event of a sudden ground collapse.

[0037] The edge resistance change rate dynamic signal processing module used in the present invention includes a signal acquisition storage module, an amplifier chip, a signal conditioning module, and a temperature compensation chip. The amplifier chip is an AD620 chip, and the temperature compensation chip is a DS18B20 temperature measurement chip. The signal conditioning module is used to measure the resistance value of the flexible conductive film, and the temperature compensation chip is used to measure the ambient temperature of the pipe and perform temperature compensation on the resistance measurement value of the flexible conductive film. The signal acquisition storage module adopts a miniature low-power structure, and each signal acquisition storage module is connected to a number of conductive films, and is installed one by one inside the measuring pipe after waterproof packaging. The edge distributed optical fiber acoustic sensing demodulation module is also arranged in the ground monitoring base station where the measuring pipe is located after waterproof treatment. Figure 4 5 is the resistance change rate-strain correlation curve of the flexible conductive film. The distributed acoustic sensing demodulation module can effectively sense the developmental activation behaviors such as soil collapse inside the soil cave. 5 is the data curve when the distributed acoustic sensing demodulation module collects the acoustic vibration signal.

[0038] Before the collapse occurs and during the development and activation of the soil cave, the soil above the soil cave roof undergoes progressive destruction similar to the surrounding rock of the underground cave roof, with tension and shear coupling deformation as the main deformation. The soil below the measuring pipe gradually decreases, and the buried measuring pipe can be regarded as an elastic foundation beam. The spring stiffness of the soil below it continues to decrease, and the measuring pipe gradually undergoes settlement deformation, causing the flexible conductive film array on the surface of the PE pipe to undergo dynamic deformation and change the resistance change rate. At the same time, the optical fiber recognizes abnormal acoustic vibration signals such as soil blocks falling in the soil cave, and the dynamic signal of the resistance change rate is fused with the acoustic vibration signal in the edge multivariate data fusion module, and the fused signal is uploaded to the cloud. Figure 6 This is a comparison chart of the dynamic signal of the resistance change rate of some sections and the curve generated by the acoustic vibration signal data of multiple channels of the optical fiber in the same coordinate system. Fig.10 It is a curve diagram of the dynamic signal of the resistance change rate and the acoustic vibration signal data of a section and the data obtained after fusion, wherein D1 is the acoustic vibration signal data; C1 is the dynamic signal data of the resistance change rate; and F1 is the fused data.

[0039] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A distributed cloud-edge-end collaborative fusion prediction and early warning system for karst collapse, characterized in that: include: Polymer measuring tube, flexible conductive film array, optical fiber, edge resistance change rate dynamic signal processing module, edge distributed optical fiber acoustic sensor demodulation module, edge multivariate data fusion module, cloud prediction and early warning module; Guide grooves are provided on the central axis of both sides of the polymer measuring tube, and optical fibers are respectively embedded in the guide grooves and then packaged, and the flexible conductive film array is arranged on the surface of the polymer measuring tube; the polymer measuring tube is horizontally laid below the critical depth of soil hole stability, and each flexible conductive film is connected to the edge resistance change rate dynamic signal processing module through a wire, and the end of the optical fiber is connected to the edge distributed optical fiber acoustic sensor demodulation module; The edge resistance change rate dynamic signal processing module is used to collect the resistance change rate dynamic signal of the flexible conductive film array on the surface of the measuring tube; The edge distributed optical fiber acoustic sensing demodulation module is used to collect acoustic vibration signals of the optical fibers packaged in the guide grooves on the surfaces of both sides of the measuring tube; The edge multivariate data fusion module is used to fuse the received dynamic signal of resistance change rate with the acoustic vibration signal based on spatial positioning, and upload the fused signal to the cloud prediction and warning module; The cloud-based prediction and warning module is used to identify the activation or activation stop behavior of karst soil cave development based on the fusion signal, and predict the soil cave development activation rate and development depth based on the fusion signal. When it is predicted that the soil cave development depth will reach the soil cave critical depth or the soil cave boundary will develop to the control position within the warning time, the cloud-based prediction and warning module issues a warning signal; The cloud-based prediction and early warning module predicts the activation rate and development depth of the soil hole development based on the fusion signal, specifically including: using the moving average method to divide the displacement of the soil hole top plate into trend items and random items according to the time series, using the particle filter algorithm to predict the displacement of the trend item, and using the XGBoost algorithm after the sparrow algorithm optimizes the model hyperparameters to predict the displacement of the random item; the displacement of the trend item and the displacement of the random item together constitute the predicted value of the cumulative displacement of the soil hole top plate, and finally predicting the activation rate and development depth of the soil hole development based on the time series data fused by the multivariate data; The cloud-based prediction and early warning module is used to identify the activation or activation stop behavior of karst soil cave development, specifically including: judging the activation or activation stop behavior of soil cave development according to the predicted cumulative displacement value of the soil cave roof; when the predicted cumulative displacement value of the soil cave roof shows continuous peak segments within a certain period of time and the displacement remains greater than 0, it is judged that the soil cave is activated at this time; when the predicted cumulative displacement value of the soil cave roof does not fluctuate significantly with the increase of time or shows a significant decrease compared with the soil cave development activation segment in the previous period of time and approaches 0, it is judged that the soil cave development activation has stopped.

2. The karst collapse distributed cloud-edge-end collaborative fusion prediction and early warning system according to claim 1 is characterized in that: A plurality of monitoring sections are arranged on the polymer measuring tube, and each section is equally spaced. A group of flexible conductive films is arranged on each monitoring section, and the flexible conductive films on each monitoring section together form the flexible conductive film array; wherein, a group of flexible conductive films is composed of conductive films respectively coated on the top surface and the bottom surface of the polymer measuring tube along the axial direction, and conductive films respectively coated on both sides of the polymer measuring tube obliquely; the conductive films on the top surface and the bottom surface are used to sense the tensile and compressive dynamic strains of the measuring tube to characterize the development time and position of the soil hole, and the oblique conductive films on the two side surfaces are used to sense the shear dynamic strains of the measuring tube to characterize the boundaries on both sides of the soil hole development; each electrode in each group of flexible conductive films is connected to the edge resistance change rate dynamic signal processing module through a wire.

3. The distributed cloud-edge-end collaborative fusion prediction and early warning system for karst collapse according to claim 2 is characterized in that: The edge multivariate data fusion module performs fusion based on spatial positioning, specifically including: for any monitoring section, the resistance signal data of each flexible conductive film is fused through Kalman filtering to obtain the fused conductive film data of the monitoring section, that is, data one; the acoustic vibration signal data of each optical fiber channel measured by the edge distributed optical fiber acoustic sensor demodulation module is pre-processed, and the data is denoised by wavelet denoising to obtain data two; the data one corresponding to the monitoring section is multivariately fused with the data two of the optical fiber channel adjacent to the monitoring section, and the data fusion adopts the classical particle filter equation after Taylor series decomposition.

4. The karst collapse distributed cloud-edge-end collaborative fusion prediction and early warning system according to claim 2 or 3 is characterized in that: When the cloud-based prediction and early warning module identifies that the karst soil cave has developed and activated, the edge end is controlled to increase the signal collection frequency; when the cloud-based prediction and early warning module identifies that the karst soil cave has stopped being activated, the edge end is controlled to reduce the signal collection frequency.

5. The distributed cloud-edge-end collaborative fusion prediction and early warning system for karst collapse according to claim 2 or 3 is characterized in that: The polymer measuring tube is a PE round tube with an outer diameter greater than 60 mm, and the length of the polymer measuring tube is greater than 1.5 m; the length of each section of the flexible conductive film is greater than 40 mm and the width is greater than 10 mm.

6. The distributed cloud-edge-end collaborative fusion prediction and early warning system for karst collapse according to claim 2 or 3 is characterized in that: The flexible conductive film is a graphene-based flexible conductive film. Zinc foil is used as a reducing agent, and graphene oxide is reduced by a substitution reaction. The reduced graphene oxide composite material is prepared and then added to a polyurethane matrix to prepare the graphene-based flexible conductive film.

Images