An optimized method for detecting leakage position of foundation pit based on electrical method

Abstract

This invention proposes an optimized method for detecting leakage locations in foundation pits based on electrical resistivity tomography (EDT), addressing the current technical problems of low detection efficiency, long processing time, increased detection frequency due to unpredictable future pit conditions, and low accuracy. The invention includes: acquiring data sequences of voltage, water content, dielectric constant, and pore water pressure within a unit area of ​​the foundation pit, extracting features, constructing a state vector for the current time period within that unit area, and inputting this state vector into a trained classifier to determine whether the corresponding state category of a certain area of ​​the foundation pit is leakage. If leakage is identified, sensors are deployed at increased density in that area; otherwise, the future state is predicted using a network, and the data is repeatedly fed into the trained classifier for cyclical prediction and classification. This invention achieves rapid determination of leakage locations in foundation pits, offering high detection efficiency, high accuracy, and the ability to predict future pit conditions.

Description

Technical Field

[0001] This invention relates to the technical field of detecting leakage in foundation pits, and in particular to an optimized method for detecting the location of leakage in foundation pits based on electrical resistivity tomography. Background Technology

[0002] Foundation pit leakage is a typical engineering problem caused by construction defects such as those in the water-stopping system. It is particularly prevalent in areas with abundant groundwater and high water levels. Since the water-stopping system is a concealed component, leakage in a foundation pit cannot be detected by visual inspection alone. If water seepage is visible to the naked eye, it indicates that the leakage is irreversible. Therefore, it is necessary to inspect the interior of the foundation pit to allow for proactive measures.

[0003] For example, Chinese invention patent CN110006595B, with an authorization announcement date of April 2, 2021, discloses a method for detecting planar leakage in the retaining structure of ultra-deep foundation pits. This method includes: utilizing the principle of electroosmosis to perform highly sensitive measurements of the weak ion movement at the leakage point to detect leakage in complex underground structures. Even slight leakage will cause changes in the electric field of the entire stratum due to ion movement. This change can be detected by a developed multi-channel, multi-sensor, high-precision measurement system, which can pinpoint the location of the electric field anomaly and thus locate the leakage point. This invention achieves the detection efficiency of multiple sensors with a single sensor and plans its path, enabling the detection of leakage in the retaining structure at various points.

[0004] While the aforementioned patents can also detect leakage inside the foundation pit, the effective detection range of the sensors is limited due to the large area of ​​the foundation pit. Therefore, the more accurate the detection of leakage in the entire foundation pit, the more sensors need to be deployed, resulting in low testing efficiency and long testing time. In addition, they can only detect the current state of the foundation pit and cannot predict the future state of the foundation pit based on the current state, which leads to an increase in the number of tests required and higher costs. Summary of the Invention

[0005] This invention proposes an optimized method for detecting leakage locations in foundation pits based on electrical resistivity tomography, which solves the technical problems of low detection efficiency, long detection time, and increased detection frequency and cost due to the inability to predict future foundation pit conditions.

[0006] To achieve the above objectives, the technical solution of the present invention is an optimized method for detecting the location of seepage in a foundation pit based on electrical resistivity tomography, the optimized method comprising the following steps:

[0007] Step 1: The interior of the foundation pit includes the inner wall and the bottom of the foundation pit. The interior of the foundation pit is evenly divided into several unit areas, and sensors are placed at the center of each unit area to collect information.

[0008] Step 2: The sensor includes an electrode sensor. Using the electrode sensor, the voltage of a unit area inside the foundation pit is collected at a preset sampling frequency to obtain the voltage data sequence of the unit area inside the foundation pit within the current time period.

[0009] Based on the voltage data sequence of a unit area inside the foundation pit, features are extracted to construct the state vector of the unit area inside the foundation pit for the current time period.

[0010] Step 3: Repeat Step 2 to obtain the state vector of each unit area inside the foundation pit for the current time period. Input the state vector of each unit area inside the foundation pit for the current time period into the trained binary classification linear classifier SVM to obtain the foundation pit state category corresponding to the state vector of each unit area inside the foundation pit for the current time period. Determine whether the foundation pit state category of each unit area inside the foundation pit for the current time period is leakage. If the state category of the unit area is leakage, set the unit area as the leakage area. In the leakage area, increase the density of sensor placement. Based on the original sensor placement points, divide the leakage area evenly into several rectangular grids. The original sensor placement points are set at the center of the leakage area, and the newly placed sensors are placed at the center of each of the several rectangular grids. Iterate and search until the leakage location is found.

[0011] If the state category of the unit area is not leakage, the future state vector of the unit area inside the foundation pit is predicted according to the prediction network. Then, the future state vector of the unit area is fed into the trained binary classification linear classifier SVM for cyclic prediction and classification judgment until the end.

[0012] Step 4: After identifying the leakage area, set a leakage threshold based on the development and changes in the construction status. Determine whether the leakage area is a serious leak based on the leakage threshold. If so, repair the seriously leaking area. After repair, continue the analysis of Steps 1 to 3.

[0013] Furthermore, the optimization method includes the following steps:

[0014] Step 1: The interior of the foundation pit includes the inner wall and the bottom of the foundation pit. The interior of the foundation pit is evenly divided into several unit areas, and sensors are placed at the center of each unit area to collect information.

[0015] Step 2: The sensor includes an electrode sensor and a moisture content sensor; using the electrode sensor, the voltage of a unit area inside the foundation pit is collected at a preset sampling frequency to obtain the voltage data sequence of the unit area inside the foundation pit within the current time period; using the moisture content sensor, the moisture content of a unit area inside the foundation pit is collected at a preset sampling frequency to obtain the moisture content data sequence of the unit area inside the foundation pit within the current time period.

[0016] Based on the voltage data sequence and water content data sequence of a unit area inside the foundation pit, features are extracted to construct the state vector of the unit area inside the foundation pit for the current time period.

[0017] Furthermore, the optimization method includes the following steps:

[0018] Step 1: The interior of the foundation pit includes the inner wall and the bottom of the foundation pit. The interior of the foundation pit is evenly divided into several unit areas, and sensors are placed at the center of each unit area to collect information.

[0019] Step 2: The sensors include an electrode sensor, a moisture content sensor, and a dielectric constant sensor. Using the electrode sensor, the voltage of a unit area inside the foundation pit is collected at a preset sampling frequency to obtain a voltage data sequence for the unit area inside the foundation pit within the current time period. Using the moisture content sensor, the moisture content of a unit area inside the foundation pit is collected at a preset sampling frequency to obtain a moisture content data sequence for the unit area inside the foundation pit within the current time period. Using the dielectric constant sensor, the dielectric constant of a unit area inside the foundation pit is collected at a preset sampling frequency to obtain a dielectric constant data sequence for the unit area inside the foundation pit within the current time period.

[0020] Based on the voltage data sequence, water content data sequence, and dielectric constant data sequence of a unit area inside the foundation pit, features are extracted to construct the state vector of the unit area inside the foundation pit for the current time period.

[0021] Furthermore, the optimization method includes the following steps:

[0022] Step 1: The interior of the foundation pit includes the outer side of the foundation pit retaining structure and the bottom of the foundation pit. The interior of the foundation pit is evenly divided into several unit areas, and sensors are placed at the center of each unit area to collect information.

[0023] Step 2: The sensors include an electrode sensor, a moisture content sensor, a dielectric constant sensor, and a pore water pressure sensor. The electrode sensor is used to collect the voltage of a unit area inside the foundation pit at a preset sampling frequency to obtain a voltage data sequence for the unit area inside the foundation pit within the current time period. The moisture content sensor is used to collect the moisture content of a unit area inside the foundation pit at a preset sampling frequency to obtain a moisture content data sequence for the unit area inside the foundation pit within the current time period. The dielectric constant sensor is used to collect the dielectric constant of a unit area inside the foundation pit at a preset sampling frequency to obtain a dielectric constant data sequence for the unit area inside the foundation pit within the current time period. The pore water pressure sensor is used to collect the pore water pressure of a unit area inside the foundation pit at a preset sampling frequency to obtain a pore water pressure data sequence for the unit area inside the foundation pit within the current time period.

[0024] Based on the voltage data sequence, water content data sequence, dielectric constant data sequence, and pore water pressure data sequence of a unit area inside the foundation pit, features are extracted to construct the state vector of the unit area inside the foundation pit for the current time period.

[0025] Furthermore, the method for extracting features includes:

[0026] Obtain the standard deviation and range of voltage data series, water content data series, dielectric constant data series, and pore water pressure data series within a unit area of ​​the foundation pit;

[0027] Based on the standard deviation and range of the voltage data sequence of the unit area inside the foundation pit, a first characteristic voltage data sequence of the unit area inside the foundation pit is obtained; based on the standard deviation and range of the water content data sequence of the unit area inside the foundation pit, a second characteristic water content data sequence of the unit area inside the foundation pit is obtained; based on the standard deviation and range of the dielectric constant data sequence of the unit area inside the foundation pit, a third characteristic dielectric constant data sequence of the unit area inside the foundation pit is obtained; based on the standard deviation and range of the pore water pressure data sequence of the unit area inside the foundation pit, a fourth characteristic pore water pressure data sequence of the unit area inside the foundation pit is obtained.

[0028] Furthermore, the method for constructing the state vector of a unit area inside the foundation pit for the current time period includes:

[0029]

[0030] Among them: (F t -F t-1() represents the voltage difference between the previous and current acquisition times in the first characteristic voltage data sequence of a unit area inside the foundation pit. For voltage characteristic values; (W) t -W t-1 () represents the difference in water content between the previous and current acquisition times in the second characteristic water content data sequence of a unit area inside the foundation pit. The characteristic value of moisture content; (Y) t -Y t-1 The third characteristic dielectric constant of a unit region inside the foundation pit is the difference in dielectric constant between the previous acquisition time and the current acquisition time in the data sequence. The characteristic value of dielectric constant; (L) t -L t-1 The fourth characteristic of the pore water pressure data sequence within a unit area inside the foundation pit is the difference between the previous and current acquisition times. X represents the characteristic value of pore water pressure; X is the state vector of a unit area inside the foundation pit during the current time period.

[0031] Furthermore, the leakage threshold classifies the leakage into minor leakage and severe leakage. In the case of minor leakage, measures are taken to increase the density of sensor arrangement; in the case of severe leakage, sealing measures are taken before increasing the density of sensor arrangement.

[0032] Furthermore, the method for obtaining the state category of the foundation pit corresponding to the state vector of all areas inside the foundation pit in the current time period is as follows:

[0033] Obtain the state vectors of each sample and the label type of each sample state vector from the training of the linear classifier SVM for binary classification.

[0034] Obtain the state vector of each sample labeled as leakage; the parameters contained in the sample state vector each include a voltage characteristic value, a water content characteristic value, a dielectric constant characteristic value, and a pore water pressure characteristic value. Combine all parameters to obtain the degree of fluctuation of the sample state.

[0035] Furthermore, the leakage threshold is the ratio of at least one vector feature value in the state vector of the unit area inside the foundation pit in the current time period to the corresponding feature value in the state vector of the unit area inside the foundation pit in the previous time period. The leakage threshold at different stages is adjusted according to the construction status.

[0036] Furthermore, the prediction network employs a binary classification linear classifier (SVM). The future state vector of the non-leaking area inside the foundation pit is predicted based on the binary classification linear classifier (SVM), and the future state vector of the non-leaking area is then fed back into the binary classification linear classifier (SVM).

[0037] The present invention has at least the following beneficial effects: It uniformly divides the foundation pit area into several regions, and places sensors at the center of each region to collect information. The sparse placement of sensors effectively improves detection efficiency while reducing detection costs. After identifying the leakage area through a classifier, sensors are densely deployed in that area to effectively determine the range. Iterative search can quickly and accurately determine the leakage location of the foundation pit, reducing detection time. Furthermore, a predictive network can be used to predict the future state of non-permeable areas in the current time period, and this prediction and classification process is repeated until construction is completed. The present invention achieves the effects of setting up a point to radiate across an area, quickly determining the leakage location of the foundation pit, high detection efficiency, short detection time, and predicting the future state of the foundation pit, thus reducing the number of detections and costs. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 A schematic diagram of a method provided by the present invention.

[0040] Figure 2 This is a schematic diagram of the interior of the foundation pit of the present invention.

[0041] Figure 3 This is a schematic diagram showing the division of the foundation pit into several unit regions according to the present invention.

[0042] Figure 4 This is a schematic diagram showing the positions of the newly arranged sensors within several rectangular grids in this invention.

[0043] In the diagram: 1 represents the inner wall of the pit, 2 represents the bottom of the pit, 3 represents the unit area, and 4 represents the position of the newly deployed sensors within several rectangular grids. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] like Figure 1 As shown in Example 1, an optimized method for detecting the location of foundation pit leakage based on electrical resistivity tomography includes the following steps:

[0046] Step 1: The interior of the foundation pit includes the inner wall and the bottom of the foundation pit. The interior of the foundation pit is evenly divided into several unit areas, and sensors are placed at the center of each unit area to collect information.

[0047] Step 2: The sensor includes an electrode sensor. Using the electrode sensor, the voltage of a unit area inside the foundation pit is collected at a preset sampling frequency to obtain the voltage data sequence of the unit area inside the foundation pit within the current time period.

[0048] Based on the voltage data sequence of a unit area inside the foundation pit, features are extracted to construct the state vector of the unit area inside the foundation pit for the current time period.

[0049] Step 3: Repeat Step 2 to obtain the state vector of each unit area inside the foundation pit for the current time period. Input the state vector of each unit area inside the foundation pit for the current time period into the trained binary classification linear classifier SVM to obtain the foundation pit state category corresponding to the state vector of each unit area inside the foundation pit for the current time period. Determine whether the foundation pit state category of each unit area inside the foundation pit for the current time period is leakage. If the state category of the unit area is leakage, set the unit area as the leakage area. In the leakage area, increase the density of sensor placement. Based on the original sensor placement points, divide the leakage area evenly into several rectangular grids. The original sensor placement points are set at the center of the leakage area, and the newly placed sensors are placed at the center of each of the several rectangular grids. Iterate and search until the leakage location is found.

[0050] If the state category of the unit area is not leakage, the future state vector of the unit area inside the foundation pit is predicted according to the prediction network. Then, the future state vector of the unit area is fed into the trained binary classification linear classifier SVM for cyclic prediction and classification judgment until the end.

[0051] Step 4: After identifying the leakage area, set a leakage threshold based on the development and changes in the construction status. Determine whether the leakage area is a serious leak based on the leakage threshold. If so, repair the seriously leaking area. After repair, continue the analysis of Steps 1 to 3.

[0052] Furthermore, the principle of collecting the voltage inside the foundation pit used in this invention is to measure the natural voltage or excitation voltage inside the foundation pit. When leakage occurs inside the foundation pit, it will cause the voltage inside the foundation pit to be abnormal. Therefore, electrode sensors are arranged in various areas inside the foundation pit to observe whether there is leakage in the current area.

[0053] like Figure 2As shown, the pit is further divided into several uniform regions. Voltage sensors, moisture content sensors, and dielectric constant sensors are placed at the center of each region. The information collected by each sensor is processed and analyzed. The resulting state feature vector is then used to identify leakage areas using a binary linear classifier (SVM). Sensors are then deployed at increased density, and several rectangular grids are used to separately divide the leakage areas. Information is collected and processed again until the specific location of the leakage point is found. The remaining leakage areas are then used to predict future information using an SVM predictive network to obtain future state feature vectors. This process of searching for leakage areas using a binary linear classifier (SVM) continues until the problem is solved.

[0054] like Figure 4 As shown, further, in the leakage area, the density of sensors is increased. Based on the original sensor placement points, the leakage area is evenly divided into several rectangular grids. The original sensor placement points are set at the center of the leakage area, and the newly placed sensors are placed at the center of each of the several rectangular grids. The search is iterated continuously until the leakage location is found.

[0055] like Figure 3 As shown, further, the entire pit is first divided into larger areas on an average basis for information collection and analysis. Once it is determined that an anomaly has occurred in a certain area, this area is then divided into several rectangular grids on an even basis. The original sensor placement points are set at the center, and other sensors are placed at the center of each grid for searching. Since the rectangular grid areas are relatively large, they are further divided evenly and the search is iterated until the specific leakage location is found.

[0056] Furthermore, the standard deviation and range of the voltage data sequence for a unit area inside the foundation pit are obtained;

[0057] Based on the standard deviation and range of the voltage data sequence of a unit area inside the foundation pit, the first characteristic voltage data sequence of a unit area inside the foundation pit is obtained.

[0058] Furthermore, the method for constructing the state vector of a unit area inside the foundation pit for the current time period includes:

[0059]

[0060] Among them: (F t -F t-1 () represents the voltage difference between the previous and current acquisition times in the first characteristic voltage data sequence of a unit area inside the foundation pit. X represents the voltage characteristic value; X is the state vector of a unit area inside the pit during the current time period.

[0061] Furthermore, based on the leakage threshold, the leakage is divided into minor leakage and severe leakage. In the case of minor leakage, the density of the sensors is increased. In the case of severe leakage, sealing measures are taken before the density of the sensors is increased.

[0062] Furthermore, the method for obtaining the state category of the foundation pit corresponding to the state vector of all areas inside the foundation pit in the current time period is as follows:

[0063] Obtain the state vectors of each sample and the label type of each sample state vector from the training of the linear classifier SVM for binary classification.

[0064] Obtain the state vector of each sample labeled as leakage; the parameters contained in the sample state vector include a voltage feature value, and the degree of fluctuation of the sample state is obtained based on the voltage feature value.

[0065] Furthermore, the advantage of the SVM classifier is that it can perform linear or nonlinear classification using hyperplanes and kernel functions, resulting in relatively accurate output. The training process of the SVM classifier is as follows: 1000 sample state vectors are obtained from three aspects during the foundation pit leakage cycle: indoor foundation pit model experiments, on-site foundation pit measurements, and numerical simulations of foundation pit leakage. The entire lifecycle refers to the process from the start of excavation of a new foundation pit to the cessation of construction due to leakage. The proportion of specific foundation pit data information obtained from these three aspects to the total number of sample states can be adjusted according to the implementer and the specific implementation. Then, the SVM classifier is trained using these 1000 sample state vectors, with 90% of the vectors used as training samples. The remaining 10% is used as test samples. The SVM classifier model is built and trained using the training samples. The values ​​of the classifier parameters are changed, and the parameter values ​​corresponding to the best or worst classifier performance are calculated. At this point, the classifier parameter training is complete. Then, the remaining 10% of test samples are fed into the trained classifier to test the classification effect, determine whether the classification is correct, and whether the accuracy meets the usage requirements. If the accuracy is low, the classifier parameters are further modified until the accuracy meets the usage requirements. The structure and specific training process of the SVM classifier are well-known technologies and will not be described in detail here.

[0066] Furthermore, the leakage threshold is the ratio of at least one vector feature value in the state vector of the unit area inside the foundation pit in the current time period to the corresponding feature value in the state vector of the unit area inside the foundation pit in the previous time period. The leakage threshold at different stages is adjusted according to the construction status.

[0067] Furthermore, as construction progresses, the construction personnel adjust the leakage threshold according to the construction status. Different construction stages present different conditions for foundation pit leakage, so the leakage threshold also varies.

[0068] Furthermore, the prediction network employs a binary classification linear classifier (SVM). The future state vector of the non-leaking area inside the foundation pit is predicted based on the binary classification linear classifier (SVM), and the future state vector of the non-leaking area is then fed back into the binary classification linear classifier (SVM).

[0069] Example 2, an optimized method for detecting the location of foundation pit leakage based on electrical resistivity tomography, differs from Example 1 in that the optimized method includes the following steps:

[0070] Step 1: The interior of the foundation pit includes the inner wall and the bottom of the foundation pit. The interior of the foundation pit is evenly divided into several unit areas, and sensors are placed at the center of each unit area to collect information.

[0071] Step 2: The sensor includes an electrode sensor and a moisture content sensor; using the electrode sensor, the voltage of a unit area inside the foundation pit is collected at a preset sampling frequency to obtain the voltage data sequence of the unit area inside the foundation pit within the current time period; using the moisture content sensor, the moisture content of a unit area inside the foundation pit is collected at a preset sampling frequency to obtain the moisture content data sequence of the unit area inside the foundation pit within the current time period.

[0072] Based on the voltage data sequence and water content data sequence of a unit area inside the foundation pit, features are extracted to construct the state vector of the unit area inside the foundation pit for the current time period.

[0073] Furthermore, when leakage occurs in the foundation pit, the moisture content data inside the pit will be abnormal. Therefore, moisture content sensors are used to detect whether there is leakage inside the foundation pit. Moisture content sensors are placed in various areas inside the foundation pit to observe whether there is leakage in the current area. The moisture content sensor adopts a TDR probe sensor, which has high detection sensitivity and low cost.

[0074] Furthermore, the method for extracting features includes:

[0075] Obtain the standard deviation and range of the voltage data sequence and the standard deviation and range of the moisture content data sequence of the unit area inside the foundation pit;

[0076] Based on the standard deviation and range of the voltage data sequence of the unit area inside the foundation pit, a first characteristic voltage data sequence of the unit area inside the foundation pit is obtained; based on the standard deviation and range of the water content data sequence of the unit area inside the foundation pit, a second characteristic water content data sequence of the unit area inside the foundation pit is obtained.

[0077] Furthermore, the method for constructing the state vector of a unit area inside the foundation pit for the current time period includes:

[0078]

[0079] Among them: (F t -F t-1 () represents the voltage difference between the previous and current acquisition times in the first characteristic voltage data sequence of a unit area inside the foundation pit. For voltage characteristic values; (W) t -W t-1 () represents the difference in water content between the previous and current acquisition times in the second characteristic water content data sequence of a unit area inside the foundation pit. X represents the water content characteristic value; X is the state vector of a unit area inside the foundation pit during the current time period.

[0080] Furthermore, the method for obtaining the state category of the foundation pit corresponding to the state vector of all areas inside the foundation pit in the current time period is as follows:

[0081] Obtain the state vectors of each sample and the label type of each sample state vector from the training of the linear classifier SVM for binary classification.

[0082] Obtain the state vector of each sample labeled as "leakage"; the parameters contained in the sample state vector each contain a voltage feature value and a moisture content feature value. Combine all parameters to obtain the degree of fluctuation of the sample state.

[0083] Example 3, an optimized method for detecting the location of foundation pit leakage based on electrical resistivity tomography, differs from Examples 1 and 2 in that the optimized method includes the following steps:

[0084] Step 1: The interior of the foundation pit includes the inner wall and the bottom of the foundation pit. The interior of the foundation pit is evenly divided into several unit areas, and sensors are placed at the center of each unit area to collect information.

[0085] Step 2: The sensors include an electrode sensor, a moisture content sensor, and a dielectric constant sensor. Using the electrode sensor, the voltage of a unit area inside the foundation pit is collected at a preset sampling frequency to obtain a voltage data sequence for the unit area inside the foundation pit within the current time period. Using the moisture content sensor, the moisture content of a unit area inside the foundation pit is collected at a preset sampling frequency to obtain a moisture content data sequence for the unit area inside the foundation pit within the current time period. Using the dielectric constant sensor, the dielectric constant of a unit area inside the foundation pit is collected at a preset sampling frequency to obtain a dielectric constant data sequence for the unit area inside the foundation pit within the current time period.

[0086] Based on the voltage data sequence, water content data sequence, and dielectric constant data sequence of a unit area inside the foundation pit, features are extracted to construct the state vector of the unit area inside the foundation pit for the current time period.

[0087] Furthermore, a dielectric constant sensor is used to detect leakage in the foundation pit. Water has a high dielectric constant, and its polar molecules can alter the overall dielectric constant of the material. Therefore, the dielectric constant of soil mainly depends on its water content. Existing experimental studies have shown a certain relationship between the dielectric constant of soil and its volumetric water content. Generally, the higher the volumetric water content, the higher the dielectric constant. When the dielectric constant is the same, soil with higher porosity has a higher volumetric water content. When the volumetric water content is constant, soil with lower porosity has a higher dielectric constant. If the porosity is too high, the volumetric water content will be overestimated; if the porosity is too low, the dielectric constant will be overestimated.

[0088] Furthermore, in this embodiment, the dielectric constant and water content are combined to avoid the situation where the soil at the site of the foundation pit is relatively soft and the soil porosity is too high, which would cause the collected information to affect the judgment. The dielectric constant, water content and electrodes are used together to verify whether there is leakage inside the foundation pit. The multi-level and multi-angle analysis makes the analysis results more three-dimensional and improves the authenticity of the analysis results.

[0089] Furthermore, the method for extracting features includes:

[0090] Obtain the standard deviation and range of the voltage data sequence, the standard deviation and range of the water content data sequence, and the standard deviation and range of the dielectric constant data sequence of the unit area inside the foundation pit;

[0091] Based on the standard deviation and range of the voltage data sequence of the unit area inside the foundation pit, a first characteristic voltage data sequence of the unit area inside the foundation pit is obtained; based on the standard deviation and range of the water content data sequence of the unit area inside the foundation pit, a second characteristic water content data sequence of the unit area inside the foundation pit is obtained; based on the standard deviation and range of the dielectric constant data sequence of the unit area inside the foundation pit, a third characteristic dielectric constant data sequence of the unit area inside the foundation pit is obtained.

[0092] Furthermore, the method for constructing the state vector of a unit area inside the foundation pit for the current time period includes:

[0093]

[0094] Among them: (F t -F t-1() represents the voltage difference between the previous and current acquisition times in the first characteristic voltage data sequence of a unit area inside the foundation pit. For voltage characteristic values; (W) t -W t-1 () represents the difference in water content between the previous and current acquisition times in the second characteristic water content data sequence of a unit area inside the foundation pit. The characteristic value of moisture content; (Y) t -Y t-1 The third characteristic dielectric constant of a unit region inside the foundation pit is the difference in dielectric constant between the previous acquisition time and the current acquisition time in the data sequence. X is the characteristic value of dielectric constant; X is the state vector of a unit area inside the foundation pit during the current time period.

[0095] Furthermore, the method for obtaining the state category of the foundation pit corresponding to the state vector of all areas inside the foundation pit in the current time period is as follows:

[0096] Obtain the state vectors of each sample and the label type of each sample state vector from the training of the linear classifier SVM for binary classification.

[0097] Obtain the state vector of each sample labeled as leakage; the parameters contained in the sample state vector each contain a voltage characteristic value, a moisture content characteristic value, and a dielectric constant characteristic value. Combine all parameters to obtain the degree of fluctuation of the sample state.

[0098] Example 4, an optimized method for detecting the location of foundation pit leakage based on electrical resistivity tomography, differs from Examples 1, 2 and 3 in that the optimized method includes the following steps:

[0099] Step 1: The interior of the foundation pit includes the outer side of the foundation pit retaining structure and the bottom of the foundation pit. The interior of the foundation pit is evenly divided into several unit areas, and sensors are placed at the center of each unit area to collect information.

[0100] Step 2: The sensors include an electrode sensor, a moisture content sensor, a dielectric constant sensor, and a pore water pressure sensor. The electrode sensor is used to collect the voltage of a unit area inside the foundation pit at a preset sampling frequency to obtain a voltage data sequence for the unit area inside the foundation pit within the current time period. The moisture content sensor is used to collect the moisture content of a unit area inside the foundation pit at a preset sampling frequency to obtain a moisture content data sequence for the unit area inside the foundation pit within the current time period. The dielectric constant sensor is used to collect the dielectric constant of a unit area inside the foundation pit at a preset sampling frequency to obtain a dielectric constant data sequence for the unit area inside the foundation pit within the current time period. The pore water pressure sensor is used to collect the pore water pressure of a unit area inside the foundation pit at a preset sampling frequency to obtain a pore water pressure data sequence for the unit area inside the foundation pit within the current time period.

[0101] Based on the voltage data sequence, water content data sequence, dielectric constant data sequence, and pore water pressure data sequence of a unit area inside the foundation pit, features are extracted to construct the state vector of the unit area inside the foundation pit for the current time period.

[0102] Furthermore, a pore water pressure sensor is used to detect leakage within the foundation pit. For highly permeable soil under no-flow conditions, the pore water pressure is approximately equal to the hydrostatic pressure under no-flow conditions. When water flow occurs, the pore water pressure changes, indicating whether leakage has occurred within the foundation pit. Therefore, pore water pressure data, voltage data, moisture content data, and dielectric constant data are combined to corroborate whether leakage has occurred in the foundation pit.

[0103] Furthermore, the method for extracting features includes:

[0104] Obtain the standard deviation and range of voltage data series, water content data series, dielectric constant data series, and pore water pressure data series within a unit area of ​​the foundation pit;

[0105] Based on the standard deviation and range of the voltage data sequence of the unit area inside the foundation pit, a first characteristic voltage data sequence of the unit area inside the foundation pit is obtained; based on the standard deviation and range of the water content data sequence of the unit area inside the foundation pit, a second characteristic water content data sequence of the unit area inside the foundation pit is obtained; based on the standard deviation and range of the dielectric constant data sequence of the unit area inside the foundation pit, a third characteristic dielectric constant data sequence of the unit area inside the foundation pit is obtained; based on the standard deviation and range of the pore water pressure data sequence of the unit area inside the foundation pit, a fourth characteristic pore water pressure data sequence of the unit area inside the foundation pit is obtained.

[0106] Furthermore, the method for constructing the state vector of a unit area inside the foundation pit for the current time period includes:

[0107]

[0108] Among them: (F t -F t-1 () represents the voltage difference between the previous and current acquisition times in the first characteristic voltage data sequence of a unit area inside the foundation pit. For voltage characteristic values; (W) t -W t-1() represents the difference in water content between the previous and current acquisition times in the second characteristic water content data sequence of a unit area inside the foundation pit. The characteristic value of moisture content; (Y) t -Y t-1 The third characteristic dielectric constant of a unit region inside the foundation pit is the difference in dielectric constant between the previous acquisition time and the current acquisition time in the data sequence. The characteristic value of dielectric constant; (L) t -L t-1 The fourth characteristic of the pore water pressure data sequence within a unit area inside the foundation pit is the difference between the previous and current acquisition times. X represents the characteristic value of pore water pressure; X is the state vector of a unit area inside the foundation pit during the current time period.

[0109] Furthermore, the method for obtaining the state category of the foundation pit corresponding to the state vector of all areas inside the foundation pit in the current time period is as follows:

[0110] Obtain the state vectors of each sample and the label type of each sample state vector from the training of the linear classifier SVM for binary classification.

[0111] Obtain the state vector of each sample labeled as leakage; the parameters contained in the sample state vector each include a voltage characteristic value, a water content characteristic value, a dielectric constant characteristic value, and a pore water pressure characteristic value. Combine all parameters to obtain the degree of fluctuation of the sample state.

[0112] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An optimized method for detecting the location of a foundation pit leakage based on electrical method, characterized in that, The optimization method comprises the following steps: Step one: the inside of the foundation pit includes the inner wall of the foundation pit and the bottom of the foundation pit, the inside of the foundation pit is evenly divided into a plurality of unit areas, and a sensor is arranged at the center position of each unit area to collect information; Step two: the sensor comprises an electrode sensor, the electrode sensor is used to collect the voltage of the unit area inside the foundation pit at a preset sampling frequency, and a voltage data sequence of the unit area inside the foundation pit in a current time period is obtained; According to the voltage data sequence of the unit area inside the foundation pit, a feature is extracted, and a state vector of the unit area inside the foundation pit in the current time period is constructed; Step three: repeating step two, obtaining the state vector of each unit area inside the foundation pit in the current time period, inputting the state vectors of all unit areas inside the foundation pit in the current time period into the trained binary classification linear classifier SVM, obtaining the state category corresponding to the state vector of each unit area inside the foundation pit in the current time period, and respectively judging whether the state category of each unit area inside the foundation pit in the current time period is leakage; if the state category of the unit area is leakage, the unit area is set as a leakage area, the density of the sensor is increased in the leakage area, the leakage area is evenly divided into a plurality of rectangular grids according to the original sensor arrangement points, the original sensor arrangement points are arranged at the center positions of the leakage area, the newly arranged sensors are arranged at the center positions of each grid of the plurality of rectangular grids, and iterative search is continuously performed until the leakage position is found; If the state category of the unit area is not leakage, the future state vector of the unit area is predicted according to the prediction network, and the future state vector of the unit area is input into the trained binary classification linear classifier SVM for cyclic prediction and classification judgment until the end; Step four: after the leakage area is found, a leakage threshold is set according to the development and change of the construction state, whether the leakage area is serious leakage is judged according to the leakage threshold, if yes, the serious leakage area is repaired, and the analysis of steps one to three is continued after the repair; The method for constructing the state vector of the unit area inside the foundation pit in the current time period comprises: wherein: is the voltage difference value between the last acquisition time and the current acquisition time in the first characteristic voltage data sequence of the unit area inside the foundation pit, is the voltage characteristic value; X is the state vector of the unit area inside the foundation pit in the current time period.

2. The optimization method for detecting the leakage position of a foundation pit based on electrical method according to claim 1, characterized in that, The optimization method comprises the following steps: Step two: the sensor comprises an electrode sensor and a water content sensor; the electrode sensor is used to collect the voltage of the unit area inside the foundation pit at a preset sampling frequency, and a voltage data sequence of the unit area inside the foundation pit in a current time period is obtained; the water content sensor is used to collect the water content of the unit area inside the foundation pit at a preset sampling frequency, and a water content data sequence of the unit area inside the foundation pit in the current time period is obtained; According to the voltage data sequence of the unit area inside the foundation pit and the water content data sequence of the unit area inside the foundation pit, a feature is extracted, and a state vector of the unit area inside the foundation pit in the current time period is constructed.

3. The optimization method for detecting the leakage position of a foundation pit based on electrical method according to claim 1, characterized in that, The optimization method comprises the following steps: Step two: the sensor includes an electrode sensor, a water content sensor, and a dielectric constant sensor; using the electrode sensor, the voltage of the unit area inside the foundation pit is collected at a preset sampling frequency to obtain a voltage data sequence of the unit area inside the foundation pit in the current time period; using the water content sensor, the water content of the unit area inside the foundation pit is collected at a preset sampling frequency to obtain a water content data sequence of the unit area inside the foundation pit in the current time period; using the dielectric constant sensor, the dielectric constant of the unit area inside the foundation pit is collected at a preset sampling frequency to obtain a dielectric constant data sequence of the unit area inside the foundation pit in the current time period; According to the voltage data sequence of the unit area inside the foundation pit, the water content data sequence of the unit area inside the foundation pit, and the dielectric constant data sequence of the unit area inside the foundation pit, the features are extracted to construct a state vector of the unit area inside the foundation pit in the current time period.

4. The optimization method for detecting the leakage position of a foundation pit based on electrical method according to claim 1, characterized in that, The optimization method includes the following steps: Step two: the sensor includes an electrode sensor, a water content sensor, a dielectric constant sensor, and a pore water pressure sensor; using the electrode sensor, the voltage of the unit area inside the foundation pit is collected at a preset sampling frequency to obtain a voltage data sequence of the unit area inside the foundation pit in the current time period; using the water content sensor, the water content of the unit area inside the foundation pit is collected at a preset sampling frequency to obtain a water content data sequence of the unit area inside the foundation pit in the current time period; using the dielectric constant sensor, the dielectric constant of the unit area inside the foundation pit is collected at a preset sampling frequency to obtain a dielectric constant data sequence of the unit area inside the foundation pit in the current time period; using the pore water pressure sensor, the pore water pressure of the unit area inside the foundation pit is collected at a preset sampling frequency to obtain a pore water pressure data sequence of the unit area inside the foundation pit in the current time period; According to the voltage data sequence of the unit area inside the foundation pit, the water content data sequence of the unit area inside the foundation pit, the dielectric constant data sequence of the unit area inside the foundation pit, and the pore water pressure data sequence of the unit area inside the foundation pit, the features are extracted to construct a state vector of the unit area inside the foundation pit in the current time period.

5. The optimization method for detecting the leakage position of a foundation pit based on electrical method according to claim 4, characterized in that, The method for extracting features includes: Obtaining the standard deviation and range of the voltage data sequence of the unit area inside the foundation pit, the standard deviation and range of the water content data sequence of the unit area inside the foundation pit, the standard deviation and range of the dielectric constant data sequence of the unit area inside the foundation pit, and the standard deviation and range of the pore water pressure data sequence of the unit area inside the foundation pit; According to the standard deviation and range of the voltage data sequence of the unit area inside the foundation pit, a first characteristic voltage data sequence of the unit area inside the foundation pit is obtained; according to the standard deviation and range of the water content data sequence of the unit area inside the foundation pit, a second characteristic water content data sequence of the unit area inside the foundation pit is obtained; according to the standard deviation and range of the dielectric constant data sequence of the unit area inside the foundation pit, a third characteristic dielectric constant data sequence of the unit area inside the foundation pit is obtained; and according to the standard deviation and range of the pore water pressure data sequence of the unit area inside the foundation pit, a fourth characteristic pore water pressure data sequence of the unit area inside the foundation pit is obtained.

6. The optimization method for detecting the leakage position of a foundation pit based on electrical method according to claim 5, characterized in that, The method for obtaining the state vector of the unit region in the foundation pit at the current time period comprises the following steps: wherein: is a voltage difference value between the last acquisition time and the current acquisition time in the first characteristic voltage data sequence of the unit area inside the foundation pit, is a voltage characteristic value; is a moisture content difference value between the last acquisition time and the current acquisition time in the second characteristic moisture content data sequence of the unit area inside the foundation pit, is a moisture content characteristic value; is a dielectric constant difference value between the last acquisition time and the current acquisition time in the third characteristic dielectric constant data sequence of the unit area inside the foundation pit, is a dielectric constant characteristic value; is a pore water pressure difference value between the last acquisition time and the current acquisition time in the fourth characteristic pore water pressure data sequence of the unit area inside the foundation pit, is a pore water pressure characteristic value; and X is a state vector of the unit area inside the foundation pit at the current time period.

7. The optimization method for detecting leakage position of foundation pit based on electrical method according to claim 2, characterized in that, The leakage threshold value divides the leakage into slight leakage and serious leakage, and the slight leakage adopts the measure of increasing the density of the arranged sensors; and the serious leakage adopts the measure of plugging and then increasing the density of the arranged sensors.

8. The optimization method for detecting leakage position of foundation pit based on electrical method according to claim 4, characterized in that, The method for obtaining the foundation pit state category corresponding to the state vector of all regions in the foundation pit at the current time period comprises the following steps: Obtain each sample state vector of the linear classifier SVM of the training binary classification and the label type of each sample state vector; Obtain each sample state vector of the label type of the leakage; the parameters contained in the sample state vector all contain a voltage characteristic value, a water content characteristic value, a dielectric constant characteristic value and a pore water pressure characteristic value, and the fluctuation degree of the sample state is obtained by combining all the parameters.

9. The optimization method for detecting leakage position of foundation pit based on electrical method according to claim 5, characterized in that, The leakage threshold value is the ratio of at least one vector characteristic value in the state vector of the unit region in the foundation pit at the current time period to the corresponding characteristic value in the state vector of the unit region in the foundation pit at the previous time period, and the leakage threshold value at different stages is adjusted according to the construction state.

10. The optimized method for detecting the leakage position of a foundation pit based on electrical method according to any one of claims 1-6, characterized in that, The prediction network adopts the linear classifier SVM of the binary classification, and the future state vector of the un-leaked region in the foundation pit is obtained by the linear classifier SVM of the binary classification, and the future state vector of the un-leaked region is input into the linear classifier SVM of the binary classification again.

Images