Method for judging property of underground karst cave

Abstract

The invention relates to a method for judging the property of an underground karst cave, which comprises the following steps of: acquiring a horizontal electric field signal and a vertical magnetic field signal of the karst cave, and inputting the horizontal electric field signal and the vertical magnetic field signal into a forward model to obtain resistivity and polarizability results; the forward model is obtained by training an original forward model by using a training set, and the training set comprises time domain features; the step of training the original forward model through the training set comprises the steps that the time domain features are input into the forward model, prediction data are obtained, a prior constraint condition is set, the constraint condition is converted into a multi-data-volume inversion target function, Taylor expansion is conducted on the multi-data-volume inversion target function, and high-order terms are omitted. And solving first-order and second-order partial derivatives of the multi-data volume inversion target function after Taylor expansion to obtain a data updating model, and iteratively correcting the original forward modeling model by using time domain features and prediction data according to the data updating model. According to the invention, fine detection of the large-depth karst cave and fine distinguishing of karst cave properties are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering geology, and particularly to a method for determining the properties of underground karst caves. Background Art

[0002] Karst caves are widely distributed, with complex and diverse geomorphic forms and fragile ecological environments. During engineering construction, accidents such as foundation sliding, surface subsidence, and tunnel water inrush may be triggered due to the development of karst caves, damaging buildings and hindering the operation and construction of roads and tunnels. The formation of karst caves is the result of long-term corrosion of groundwater in limestone areas, and the distribution of karst caves has characteristics such as complexity, uncertainty, and concealment. Therefore, fine detection of underground karst caves and accurate determination of the properties and spatial distribution patterns of karst caves are important measures to ensure the safety of areas with karst cave development.

[0003] According to the water content, karst caves can be divided into dry karst caves and water karst caves. A water cave refers to a karst cave with perennial underground water flow. A dry cave is a fossil cave that has separated from the free water surface, developed in a place with a relatively high terrain, and has a relatively long development history. The cave is often decorated with various colorful stalactites. Currently, most karst caves have both dry karst caves and water caves. Limestone is the main surrounding rock type in karst cave areas and has a relatively high resistivity. If a karst cave is filled with water, its resistivity will be very low, but if the karst cave is not filled with water, the resistivity value of the dry karst cave will be very high, much higher than that of the surrounding rock limestone. Therefore, to achieve fine monitoring of karst caves, it is necessary to have a strong resolution ability for both high-resistance and low-resistance dry karst caves and water caves.

[0004] Commonly used electrical methods for shallow karst detection include Electrical resistivity tomography, a geophysical method that uses the resistivity difference of underground media for imaging, ground penetrating radar method, and loop source transient electromagnetic method. However, whether it is ground penetrating radar or high-density electrical method, the detection depth is limited, generally less than dozens of meters, and often only effective for a single low-resistance or high-resistance target, unable to achieve fine resolution of both low-resistance and high-resistance targets simultaneously. The loop source transient electromagnetic method has much worse resolution ability for high-resistance targets based on the principle of electromagnetic induction and cannot accurately locate high-resistance dry karst caves. Summary of the Invention

[0005] To solve the problems existing in the above-mentioned prior art, the object of the present invention is to provide a method for determining the properties of underground karst caves. Through the forward simulation of electrical sources with multiple components for dry karst caves and water caves, since water caves and dry karst caves with induced polarization effects will cause the sign reversal phenomenon of the horizontal electric field response, while dry karst caves will not cause the distortion of the vertical magnetic field response. The difference in the occurrence of the sign reversal phenomenon of the horizontal electric field and the vertical magnetic field can be used as the basis for judging the properties of underground karst caves. When only the sign of the horizontal electric field is reversed, the underground cavity is an empty karst cave. When the sign reversal phenomena of the horizontal electric field and the vertical magnetic field occur simultaneously, the underground karst cave is a water cave. Based on the verification of the measured data of the Tonghua karst cave obtained from the above theoretical simulation, it provides a new technical solution for studying the properties of underground karst caves, especially for the detection of karst caves with relatively large burial depths and the discrimination of karst cave properties.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A method for determining the properties of underground karst caves, comprising:

[0008] Obtain the horizontal electric field and vertical magnetic field signals of the karst cave, input the horizontal electric field and vertical magnetic field signals into the forward model, and obtain the resistivity and polarization rate results; the forward model is obtained by training the original forward model using a training set, and the training set includes: time domain features;

[0009] Training the original forward model using the training set includes: inputting the time domain features into the forward model, obtaining prediction data, setting prior constraint conditions, converting the constraint conditions into a multi-data body inversion objective function, performing Taylor expansion on the multi-data body inversion objective function, and omitting high-order terms. Take the first-order and second-order partial derivatives of the Taylor-expanded multi-data body inversion objective function to obtain a data update model. According to the data update model, use the time domain features and the prediction data to iteratively correct the original forward model.

[0010] Optionally, obtaining the horizontal electric field and vertical magnetic field signals of the karst cave includes:

[0011] Obtain the depth of the karst cave, and according to the depth of the karst cave, determine the grounding wire source and the offset distance for arranging a multi-point observation device around the karst cave at the same time;

[0012] Based on the multi-point observation device, conduct forward simulation, calculate the abnormal responses caused by karst caves of different sizes, burial depths, and shapes, establish a curve set, and infer the acquisition time range from the time when the abnormal response is caused. Adopt a timed sampling method to obtain the horizontal electric field and vertical magnetic field signals at different time periods.

[0013] Optionally, obtaining the time domain features includes:

[0014] Remove the background noise of the original horizontal electric field and vertical magnetic field signals and increase the signal intensity; the background noise includes: environmental noise and interference generated by the equipment itself.

[0015] Normalize and segment the signal with increased intensity to obtain signal characteristics in different time periods, and extract the time-domain characteristics using gradient changes.

[0016] Optionally, after extracting the time-domain characteristics, it includes:

[0017] Compare and analyze the time-domain characteristics with the curve set. If negative values appear in the data curve, analyze whether it is an irregular change caused by noise or a regular change caused by underground karst caves. If the time-domain characteristics are regular changes, they are used to identify the potential properties of karst caves.

[0018] Optionally, setting the prior constraint conditions includes:

[0019] P a (m) = φ(m) + as(m)

[0020] Where, P a (m) is the total objective function, a is the regularization factor, φ(m) is the sum of squares of the difference between the observed data and the predicted data, and s(m) is the stabilizer.

[0021] Optionally, the expression of the multi-data body inversion objective function is:

[0022] P α (m) = ||W1[d obs1 -F1(m)]|| 2 +||W2[d obs2 -F2(m)]| 2 +…+||W n [d obsn -F n (m)]|| 2 +α||m - m ref || 2

[0023] Where, d obsn is the measured data of different field quantities or observation area responses, F n (m) is the response function, W n is the weight coefficient matrix of the measured data, m ref is the prior model, α is the regularization factor, W1 and W2 are the weight coefficient matrices at the first and second iterations of the measured data, and m is the model corresponding to the measured data.

[0024] Optionally, obtaining the data update model includes:

[0025]

[0026] Among them, J is the sensitivity matrix, m k is the iterative model, n is the number of iterations, and W n is the weight coefficient matrix at the nth iteration of the measured data, α is the regularization factor, and T is the transpose operation.

[0027] Optionally, the method further includes: presenting the resistivity and polarizability results in the form of a chart or a report, and showing the determined karst cave properties and their geological characteristics.

[0028] The beneficial effects of the present invention are as follows:

[0029] The present invention updates the model according to data, uses time-domain features and predicted data to iteratively correct the original forward model, and uses the model that meets the accuracy requirements to simulate the actual geological conditions, realizing the fine detection of large-depth karst caves, the fine differentiation of karst cave properties, and also realizing the effectiveness of the determination of karst cave properties and the high efficiency of specific implementation. Description of the Drawings

[0030] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0031] Figure 1 is a flowchart of a method for determining the properties of underground karst caves according to an embodiment of the present invention;

[0032] Figure 2 is a schematic diagram of the resistivity and polarizability results according to an embodiment of the present invention;

[0033] Figure 3 is the response curve caused by a dry karst cave according to an embodiment of the present invention; (a) is the horizontal electric field caused by a dry karst cave, and (b) is the vertical magnetic field caused by a dry karst cave;

[0034] Figure 4 is the response curve caused by a water karst cave according to an embodiment of the present invention; (a) is the horizontal electric field caused by a water karst cave, and (b) is the vertical magnetic field caused by a water karst cave;

[0035] Figure 5 is the layout diagram of the survey line according to an embodiment of the present invention;

[0036] Figure 6 is the response curve of a typical measuring point on the survey line L18 according to an embodiment of the present invention; (a) is the response curve at the measuring point of 10 m, and (b) is the response curve at the measuring point of 200 m;

[0037] Figure 7 The typical measuring point response curve of the measuring line L12 in the embodiment of the present invention; (a) is the response curve at the measuring point of 20 m, (b) is the response curve at the measuring point of 70 m; (c) is the response curve at the measuring point of 250 m;

[0038] Figure 8 The multi-channel response curve on the L18 measuring line in the embodiment of the present invention; (a) is the horizontal electric field, (b) is the vertical magnetic field;

[0039] Figure 9 The schematic diagram of the resistivity inversion result of L18 in the embodiment of the present invention;

[0040] Figure 10 The multi-channel response curve on the L12 measuring line in the embodiment of the present invention; (a) is the horizontal electric field, (b) is the vertical magnetic field;

[0041] Figure 11 The schematic diagram of the resistivity inversion result of L12 in the embodiment of the present invention. Specific implementation manners

[0042] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0043] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0044] As Figure 1 shown, this embodiment discloses a method for determining the nature of an underground karst cave, including: acquiring the horizontal electric field and vertical magnetic field signals of the karst cave, inputting the horizontal electric field and vertical magnetic field signals into a forward model to obtain resistivity and polarizability results; the forward model is obtained by training an original forward model using a training set, and the training set includes: time-domain features; training the original forward model using the training set includes: inputting the time-domain features into the forward model to obtain prediction data, setting prior constraint conditions, converting the constraint conditions into a multi-data body inversion objective function, performing Taylor expansion on the multi-data body inversion objective function, and omitting high-order terms, taking the first-order and second-order partial derivatives of the Taylor-expanded multi-data body inversion objective function to obtain a data update model, and iteratively correcting the original forward model according to the data update model using the time-domain features and prediction data.

[0045] Furthermore, obtaining the horizontal electric field and vertical magnetic field signals of the karst cave includes: obtaining the depth of the karst cave, and based on the depth of the karst cave, determining the grounding wire source and offset distance for simultaneously arranging multi-point observation devices around the karst cave; conducting forward modeling simulations based on the multi-point observation devices, calculating the abnormal responses caused by karst caves of different sizes, burial depths, and shapes, establishing a curve set, and inferring the acquisition time range from the time when the abnormal response occurs. Adopt the method of timed sampling to obtain the horizontal electric field and vertical magnetic field signals at different time periods.

[0046] Specifically, collect the horizontal electric field and vertical magnetic field signals of the karst cave:

[0047] In this step, accurately collecting the electromagnetic signals of the underground karst cave is the basis for determining its properties. The following is a detailed description of the collection process:

[0048] 1. Equipment layout:

[0049] Grounding wire source: Analyze the approximate depth of the detection target based on the results of previous geological surveys, select a suitable grounding wire source and offset distance to ensure a wide coverage range within the target area.

[0050] Transient electromagnetic multi-component observation device: Simultaneously arrange multi-point observation devices around the karst cave to capture the electric and magnetic field signals in the equatorial direction. The horizontal electric field should be strictly parallel to the emission source, and the receiving electrode spacing should be arranged strictly according to the design to receive the electrodes. The magnetic field is collected using magnetic probes, which should be completely perpendicular to the ground surface.

[0051] 2. Signal acquisition:

[0052] Obtain the curve set through forward modeling: Calculate the responses caused by karst caves of different sizes, burial depths, and shapes, and establish a curve set.

[0053] Time range selection: Based on the depth of the target karst cave and the geological environment, conduct forward modeling simulations, and infer the acquisition time range from the time when the abnormal response occurs. Generally, signals with a lower fundamental frequency are more suitable for deep exploration, while signals with a higher fundamental frequency can provide higher resolution.

[0054] Data acquisition strategy: Adopt the method of timed sampling to ensure obtaining sufficient signal data at different time periods for comprehensive analysis.

[0055] 3. Signal processing:

[0056] Noise filtering: After signal acquisition, use digital filtering technology to remove background noise, including environmental noise and interference generated by the equipment itself.

[0057] Signal enhancement: Through signal enhancement algorithms, improve the intensity and clarity of the useful signals for subsequent analysis.

[0058] 4. Data recording and storage:

[0059] Real-time monitoring: During the signal acquisition process, the quality and integrity of the data are monitored in real time to ensure that important data is not lost.

[0060] Data storage: The collected electric and magnetic field signals are saved into a data storage system in an appropriate format and structure for subsequent analysis and processing.

[0061] Furthermore, obtaining time-domain features includes: removing the background noise of the original horizontal electric field and vertical magnetic field signals and increasing the signal intensity; the background noise includes environmental noise and interference generated by the equipment itself; normalizing and segmenting the signal with increased intensity to obtain signal features in different time periods, and extracting time-domain features using gradient changes.

[0062] Furthermore, after extracting the time-domain features, it includes: comparing and analyzing the time-domain features with a curve set. If negative values appear in the data curve, analyze whether it is an irregular change caused by noise or a regular change caused by an underground cave. If the time-domain features are regular changes, they are used to identify the nature of potential caves.

[0063] Specifically, processing the horizontal electric field and vertical magnetic field signals includes:

[0064] After completing the signal acquisition, the next step is to deeply analyze the horizontal electric field and vertical magnetic field signals to determine the nature of the underground cave. The following is a detailed description of this step:

[0065] 1. Data preprocessing:

[0066] Signal normalization: Normalize the collected electric and magnetic field signals to eliminate the influence of signal intensity at different positions and ensure the comparability of analysis results.

[0067] Data segmentation: Segment the continuous signal data for easy comparison and analysis of signal features in different time periods.

[0068] 2. Feature extraction:

[0069] Time-domain feature extraction: When there are no geological bodies such as caves, the observed response signal is monotonic. Therefore, the gradient of the observed data is positive. When affected by target bodies such as caves, the calculated response gradient shows negative values. Use gradient changes to extract the time-domain features of the signal, such as the maximum value, minimum value, mean value, variance, etc. of the signal, and pay special attention to the features of sign reversal. Compare and analyze with the curve set of the previous forward simulation. When negative values appear in the data curve, analyze the error bars of the data to determine the nature of the negative values, whether it is an irregular change caused by noise or a regular change caused by an underground cave. When the data shows regular changes, it can be used to identify the nature of potential caves.

[0070] 3. Determination of the properties of karst caves:

[0071] Joint data inversion: Inversion is performed to obtain the resistivity and polarizability characteristics of subsurface geological bodies, and information such as the burial depth and distribution space of karst caves is obtained.

[0072] Furthermore, the prior constraint conditions are set as follows:

[0073] P a (m) = φ(m) + as(m)

[0074] where P a (m) is the total objective function, a is the regularization factor, φ(m) is the sum of squares of the difference between the observed data and the predicted data, and s(m) is the stabilizer.

[0075] Furthermore, the expression of the multi-data body inversion objective function is:

[0076] P α (m) = ||W1[d obs1 - F1(m)]|| 2 + ||W2[d obs2 - F2(m)]|| 2 + … + ||W n [d obsn - F n (m)]|| 2 + α||m - m ref || 2

[0077] where d obsn is the measured data of different field quantities or the response of the observation area, F n (m) is the response function, W n is the weight coefficient matrix of the measured data, m ref is the prior model, α is the regularization factor, W1 and W2 are the weight coefficient matrices at the first and second iterations of the measured data, and m is the model corresponding to the measured data.

[0078] Furthermore, in the process of taking the first and second partial derivatives of the multi-data body inversion objective function after Taylor expansion, the derivative of Δm in the multi-data body inversion objective function after Taylor expansion is taken:

[0079]

[0080] where Δm is the model parameter residual, and Pα(m k ) is the objective function.

[0081] Furthermore, obtaining the data update model includes:

[0082]

[0083] Among them, J is the sensitivity matrix, m k is the iterative model, n is the number of iterations, W n is the weight coefficient matrix at the nth iteration of the measured data, α is the regularization factor, and T is the transpose operation.

[0084] Furthermore, this method also includes: presenting the resistivity and polarizability results in the form of charts or reports to show the determined karst cave properties and their geological characteristics.

[0085] Specifically, for the inversion problem, the non-uniqueness of the solution stems from the finiteness of the observed data and the errors in the observed data. Extending the inversion of a data volume to the simultaneous inversion of multiple data volumes and obtaining a geophysical model that satisfies multiple data volumes improves the reliability of the results. In principle, these data volumes can be any type of data and can be used to interpret the same model.

[0086] A comprehensive study on the joint inversion of transient electromagnetic multi-component data of a pair of grounded wire sources. To improve the stability and non-uniqueness problems, a regularization inversion method is introduced to enhance the stability of the inversion process by adding prior constraint conditions and reduce the non-uniqueness of the inversion results.

[0087] P a (m) = φ(m) + as(m)

[0088] In the formula, P a (m) is the total objective function: a is the regularization factor; φ(m) is the sum of squares of the difference between the observed data and the predicted data (data objective function); s(m) is the stabilizer (model constraint objective function).

[0089] The multi-data volume inversion objective function can be expressed as:

[0090] P α (m) = ||W1[d obs1 - F1(m)]|| 2 + ||W2[d obs2 - F2(m)]|| 2 + … + ||W n [d obsn - F n (m)]|| 2 + α||m - m ref || 2

[0091] In the formula, d obsn is the measured data of different field quantities or the response of the observation area, F n (m) is the response function, W nWeight coefficient matrix of measured data, m ref is the prior model.

[0092] To linearize the non - linear objective function, perform Taylor expansion on the objective function and omit the high - order terms:

[0093]

[0094] In the formula, m k is the k - th iteration value of the model.

[0095] Take the derivative of Δm to obtain the inversion iteration update formula:

[0096]

[0097] Take the first - order and second - order partial derivatives of the objective function to finally obtain the data update formula:

[0098]

[0099] In the formula, J is the sensitivity matrix. Using the data update formula m k+1 , through iteration, continuously correct the forward model m, and finally use the model that meets the accuracy requirements (the sum of squares or residuals between the observed data and the predicted data can be used to judge the model accuracy) to simulate the actual geological conditions.

[0100] Joint inversion of multiple data volumes can generally improve the reliability of the results. The provided multi - data volumes contain complementary information. The obtained resistivity and polarizability results are as Figure 2 shown.

[0101] Uncertainty analysis: Evaluate the uncertainty of the judgment results and analyze possible error sources. Through multiple tests and cross - validations, improve the reliability of the judgment.

[0102] Result visualization: Present the analysis results in the form of charts or reports, clearly showing the nature of the karst caves determined and their possible geological characteristics.

[0103] In this embodiment, forward simulations of the responses of horizontal electric fields and vertical magnetic fields of transient electromagnetic methods with electrical sources for karst caves of different natures are also carried out. Through response characteristics and RMS analysis, it is found that the horizontal electric field is sensitive to both low - resistivity and high - polarization water - filled karst caves and high - resistivity dry karst caves, while the vertical magnetic field is only sensitive to low - resistivity and high - polarization water - filled karst caves, as Figure 3 (a) - (b) shown. Low - resistivity and high - polarization water - filled karst caves and high - resistivity dry karst caves will cause distortions and even sign - reversal phenomena in the responses of vertical magnetic fields and horizontal electric fields, while dry karst caves will only cause sign - reversal phenomena in the responses of horizontal electric fields. This difference can be used as a basis for judging the water content of underground karst caves.

[0104] For high-resistivity dry karst caves, the presence of high-resistivity layers causes the sign reversal of the horizontal electric field. This is because charge accumulation occurs at the low-high resistivity electrical interface, leading to the distortion and sign reversal of the electric field response. The vertical magnetic field does not show obvious distortion and sign reversal. Whether it is a low-resistivity, high-polarization water-filled karst cave or a high-resistivity dry karst cave, the sign reversal of the horizontal electric field will occur. The sign reversal of the vertical magnetic field is only caused by low-resistivity, high-polarization water-filled karst caves.

[0105] When a low-resistivity, polarized water-filled karst cave exists, sign reversals occur in both the time derivatives of the horizontal electric field and the vertical magnetic field. Compared with the vertical magnetic field component, the sign reversal of the horizontal electric field occurs before 1e-3 s ( Figure 4 a), while the sign reversal of the vertical magnetic field occurs around 1e-1 s ( Figure 4 b), indicating that the low-resistivity, polarized layer has a greater impact on the horizontal electric field.

[0106] The study area is located on the northeastern margin of the North China Craton, where the strata are relatively complete, mainly including the Archean, Proterozoic, Paleozoic, Mesozoic, and Cenozoic. The Archean rock groups include the Anshan Group, and the Proterozoic includes the Jian Group, Laoling Group, and Sinian System. The Paleozoic includes the Cambrian, Ordovician, Carboniferous, and Permian. The Mesozoic includes the Triassic, Jurassic, and Cretaceous. The Cenozoic includes the Neogene and Quaternary.

[0107] Tectonically, it belongs to the North China stratigraphic region, where carbonate rock strata from the Middle Proterozoic to the Lower Paleozoic are distributed. Among them, the thickness of the Middle Ordovician Majiagou Formation and the Sinian Badaojiang Formation is relatively large and the purity is relatively high, providing a good material basis for the development of karst caves.

[0108] The Tonghua karst cave is developed in the Ordovician Majiagou limestone strata, which are shallow marine carbonate rocks. Due to the large thickness, pure texture, and strong rigidity of the limestone, it is conducive to the expansion of joints and fractures and good water permeability, so large and long karst caves can be developed. The formation of karst caves is greatly affected by geological structures. The structural forms of rock strata, such as folds, inclinations, or horizontal states, have obvious effects on cave development. Tectonically, the Yunxia karst cave is located in the Hunjiang fault-fold belt. It is mainly composed of the NE-trending Hunjiang syncline in the Paleozoic in the central part of the Tonghua area, including the NE-trending fault zones on both sides. Structurally, it belongs to a tensional fault zone. Groundwater follows this tensional fault zone to develop karst and form underground river pipelines. The Hunjiang fault-fold belt is also cut into pieces by other structures, such as the NNE-trending faults. Therefore, the intersection of structures is dissolved, expanded, and collapsed to form solution pools. With the occurrence of neotectonic movements, the crust rises and the groundwater level drops, which is conducive to the continuous progress of groundwater dissolution and mechanical erosion of underground water flow. The karstification dominated by horizontal dissolution is transformed into that dominated by vertical dissolution and collapse, and the solution pool is further expanded into a cave hall.

[0109] Figure 5The distribution of the emission source and the measured survey lines is given. The length of the emission source represented in red is 600 m, and the offsets between the two survey lines and the emission source are 380 m and 500 m respectively.

[0110] As Figure 6 (a)-(b) shows that there is no sign reversal in the horizontal electric field and vertical magnetic field at the 10 m measurement point, while at the 200 m measurement point, a sign reversal occurs in the horizontal electric field between 1 ms and 10 ms, but there is no such phenomenon in the vertical magnetic field.

[0111] As Figure 7 (a)-(c) shows the response curves of typical measurement points on survey line L12. Three situations occur in the measured data of survey line 12: there is no sign reversal in the horizontal electric field and vertical magnetic field at the 70 m measurement point; at the 20 m measurement point, a sign reversal occurs in the horizontal electric field, but there is no such phenomenon in the vertical magnetic field; at the 250 m measurement point, sign reversals occur in both the horizontal electric field and the vertical magnetic field.

[0112] To better compare this response difference between different measurement points, Figure 8 (a)-(b) gives the multi-trace curves of survey line L18. In the multi-trace curves of the horizontal electric field, after 2.5 ms, sign reversals occur within the measurement point range of 40 m to 270 m. However, no sign reversal is found in the multi-trace map of the vertical magnetic field.

[0113] Inversion is performed on the vertical magnetic field data without sign reversal. The inversion results are as Figure 9 shown. In the shallow part, there are yellow and orange low-resistivity layers, corresponding to the Quaternary and weathered sedimentary layers; the second layer is a high-resistivity layer with a large thickness and high electrical characteristics in light blue, inferred to be the reflection of the Ordovician Majiagou Formation limestone. The relatively local dark blue high-resistivity anomaly should be related to karst development and has weak water content; the deep green relatively medium-high resistivity layer in the range of 10 - 170 m is inferred to be the lower Ordovician strata and Cambrian strata. The dark blue high-resistivity area in the figure corresponds to the known dry karst cave.

[0114] In Figure 10 (a)-(b)'s multi-trace map, there are two sign reversals in the horizontal electric field, located within the ranges of measurement points 10 - 30 m and 120 - 270 m respectively. The times of the two sign reversals are also different. The sign reversal in the range of 10 - 30 m occurs after 2.5 ms, and the sign reversal in the range of 120 - 270 m occurs at 1.25 ms, occurring earlier. And there is only 1 sign reversal in the multi-trace map of the vertical magnetic field, occurring within the range of 120 - 270 m, corresponding to the horizontal electric field. But there is no sign reversal in the range of 10 - 30 m.

[0115] The inversion results are asFigure 11 As shown, the shallow part is a low-resistivity layer, corresponding to the Quaternary system and weathered sedimentary layer; the second layer is a medium-high resistivity layer with large thickness in light blue and green, inferred to be the reflection of the Ordovician Majiagou Formation limestone. A dark high-resistivity anomaly appears in the measuring point range of 10 - 30 m at the elevation of 550 m, which should be related to the karst development and has weak water-bearing property; below the elevation of 500 m, a relatively low-resistivity anomaly in the range of 110 - 270 m in the deep part is inferred to be a water-bearing karst cave in the lower strata of the Ordovician system. The low-resistivity area in the figure corresponds to the known water-bearing karst cave.

[0116] The embodiments described above are only descriptions of the preferred embodiments of the present invention, and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for determining the properties of an underground cave, characterized in that: include: Acquire horizontal electric field and vertical magnetic field signals of the cave, input the horizontal electric field and vertical magnetic field signals into the forward model, and obtain resistivity and polarizability results; The forward model is obtained by training the original forward model using a training set, and the training set includes: time domain features; Using the training set to train the original forward model includes: inputting the time domain features into the forward model, obtaining predicted data, setting a priori constraints, converting the constraints into a multi-volume inversion objective function, performing Taylor expansion on the multi-volume inversion objective function and ignoring higher-order terms, calculating the first-order and second-order partial derivatives of the Taylor-expanded multi-volume inversion objective function to obtain a data update model, and iteratively correcting the original forward model based on the data update model using the time domain features and the predicted data.

2. The method for determining the properties of underground caves according to claim 1, characterized in that: Acquiring the horizontal electric field and vertical magnetic field signals of the cave includes: Obtaining the depth of the cave, and determining the grounding wire source and offset distance according to the depth of the cave, so as to simultaneously arrange multiple observation devices around the cave; Based on the multi-point observation device, forward simulation is carried out to calculate the abnormal responses caused by caves of different sizes, burial depths and shapes, establish a curve set, and infer the acquisition time range from the time when the abnormal response is caused. The horizontal electric field and vertical magnetic field signals in different time periods are obtained by timed sampling.

3. The method for determining the properties of underground caves according to claim 1, characterized in that: Acquiring the time domain feature includes: Remove the background noise of the original horizontal electric field and vertical magnetic field signals and improve the signal strength; the background noise includes: environmental noise and interference generated by the device itself; The signal with enhanced intensity is normalized and data segmented to obtain signal features in different time periods, and the time domain features are extracted using gradient changes.

4. The method for determining the properties of underground caves according to claim 2, characterized in that: After extracting the time domain features, the following steps are included: The time domain features are compared and analyzed with the curve set. If a negative value appears on the data curve, it is analyzed whether it is an irregular change caused by noise or a regular change caused by underground caves. If the time domain features are regular changes, they are used to identify potential cave properties.

5. The method for determining the properties of underground caves according to claim 1, characterized in that: The prior constraints are set as follows: P a (m)=φ(m)+as(m) Among them, P a (m) is the overall objective function, a is the regularization factor, φ(m) is the sum of squares of the differences between the observed data and the predicted data, and s(m) is the stabilizer.

6. The method for determining the properties of underground caves according to claim 1, characterized in that: The expression of the multi-volume inversion objective function is: P α (m)=||W1[d obs1 -F1(m)]|| 2 +||W2[d obs2 -F2(m)]|| 2 +…+||W n [d obsn -F n (m)]|| 2 +α||m-m ref || 2 Among them, d obsn is the measured data of different field quantities or observation area responses, F n (m) is the response function, W n The weight coefficient matrix of the measured data, m ref is the prior model, α is the regularization factor, W1 and W2 are the weight coefficient matrices of the first and second iterations of the measured data, and m is the model corresponding to the measured data.

7. The method for determining the properties of underground caves according to claim 1, characterized in that: Acquiring the data update model includes: Among them, J is the sensitivity matrix, m k is the iterative model, n is the number of iterations, W n is the weight coefficient matrix of the measured data at the nth iteration, α is the regularization factor, and T is the transposition operation.

8. The method for determining the properties of underground caves according to claim 1, characterized in that: The method also includes: presenting the resistivity and polarizability results in the form of a chart or report to display the determined cave properties and geological characteristics.

Images