A method for monitoring ground subsidence layering based on force sensors

Abstract

The present invention relates to the technical field of engineering geological survey, and in particular to a method for monitoring ground settlement layering based on force sensors. Force sensors are deployed at the optimal monitoring deployment sites of each settlement monitoring layer to obtain compressive stress sensor readings, and the multidimensional settlement cells constructed by settlement crossing analysis and compressive stress sensor readings are used to perform isosurface interpolation of the inverse settlement distance to obtain the settlement vector applied by the compressive stress between the upper and lower settlement monitoring layers; the settlement vector is inverted and inferred to obtain the relative settlement displacement amplitude of each settlement monitoring layer, and the trend reconstruction and quantitative projection embedding of the relative displacement between each layer of settlement monitoring layers are performed based on the linear correlation of the relative settlement displacement amplitude to obtain the ground settlement monitoring amount of the target layered monitoring area. The present invention can use force sensors to perform settlement layering, settlement vector monitoring and analysis, and settlement trend quantitative calculation on the target area, thereby accurately outputting the monitoring results of settlement at different levels of the ground.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering geological survey, and in particular to a ground subsidence stratification monitoring method based on a force sensor. Background Art

[0002] Land subsidence is the gradual decrease in ground surface elevation caused by factors such as overexploitation of water resources, soil compression, or underground engineering activities. It is a widespread phenomenon in urban construction, transportation, mineral development, and water conservancy projects. Timely and accurate monitoring of land subsidence is crucial to ensuring the safety of engineering structures and the stability of the geological environment. Common land subsidence monitoring technologies currently include leveling, GNSS monitoring, InSAR technology, and laser ranging. These technologies are primarily used for macroscopic monitoring and are suitable for large-scale, overall settlement analysis. However, at the microscopic level, particularly for monitoring stratified settlement within soil layers, they cannot accurately measure settlement at different depths, hindering the accurate identification of key layers and their deformation characteristics during the subsidence process. Furthermore, traditional methods struggle to infer the impact of stress application on the settlement vector and trend of the underlying soil layer. Furthermore, the calculation of settlement levels requires manual effort, which can be subject to significant human error, leading to inaccurate monitoring results. Summary of the Invention

[0003] The present invention overcomes the deficiencies of the prior art and provides a method for monitoring ground subsidence stratification based on a force sensor.

[0004] In order to achieve the above object, the technical solution adopted by the present invention is:

[0005] A first aspect of the present invention provides a method for monitoring ground subsidence stratification based on a force sensor, comprising the following steps:

[0006] A feature change inference network is constructed to estimate several historical spatiotemporal rock layer sample maps within the target layered monitoring area, generating a directed edge map of rock layer subsidence. The directed edge map is then used to infer the edge probability of rock layer feature change by forward and reverse message transfer to determine the subsidence monitoring level of the target layered monitoring area.

[0007] Extract the ground rock structure and historical compression coefficient of each settlement monitoring layer in the target layered monitoring area, and make reasonable deployment decisions based on the ground rock structure and historical compression coefficient to avoid areas prone to relative settlement. Optimal monitoring deployment sites are obtained and added to the force sensor deployment plan.

[0008] Force sensors are deployed at the optimal monitoring deployment sites of each settlement monitoring layer to obtain compressive stress sensor readings. The settlement crosses the multidimensional settlement cell constructed by analyzing the compressive stress sensor readings, and the isosurface interpolation of the inverse settlement distance is performed to obtain the settlement vector exerted by the compressive stress between the upper and lower settlement monitoring layers.

[0009] The settlement vector is inverted and inferred to obtain the relative settlement displacement amplitude of each settlement monitoring layer. The relative settlement displacement amplitude is linearly correlated to perform trend reconstruction and quantitative projection embedding of the relative displacement between each settlement monitoring layer to obtain the ground settlement monitoring amount of the target layered monitoring area.

[0010] More specifically, the method of constructing a feature change inference network to estimate several historical spatiotemporal rock layer sample maps within the target layered monitoring area, generating a directed edge map of rock layer subsidence, and using the directed edge map to infer the edge probability of rock layer feature change by forward and reverse message transfer to determine the subsidence monitoring level of the target layered monitoring area specifically includes the following steps:

[0011] Obtain the target layered monitoring area and monitoring tasks for ground subsidence, and define the termination monitoring point for ground subsidence based on the monitoring tasks;

[0012] By monitoring the central control log, several historical spatiotemporal rock layer sample maps of the target layer monitoring area where the ground subsided to the termination monitoring point within the preset time period were extracted. The local binary pattern algorithm was introduced to calculate the characteristics of each historical spatiotemporal rock layer sample map, and the LBP rock layer characteristic value of each historical spatiotemporal rock layer sample map was obtained.

[0013] A characteristic change inference network is constructed. Based on the LBP rock layer characteristic values, the conditional probability of the rock layer characteristics shifting from a certain historical spatiotemporal rock layer sample map to another historical spatiotemporal rock layer sample map along with the sedimentation time series is estimated in the characteristic change inference network, and the joint probability of rock layer characteristic changes between each historical spatiotemporal rock layer sample map is obtained.

[0014] Define the historical spatiotemporal rock layer sample graph as a random variable node, construct a factor node for each random variable node based on the joint probability of rock layer characteristic changes, and connect each random variable node based on the factor node structure to build a directed edge graph of rock layer settlement;

[0015] The lithologic characteristics and settlement elements of the target layer monitoring area are obtained through big data. Based on the tectonic characteristics and settlement elements of the strata, a message is forwardly transmitted from the random variable node to the factor node along the stratum characteristic change edge in the directed edge graph. After the forward transmission is completed, the message is reversely transmitted from the factor node to the random variable node along the stratum characteristic change edge.

[0016] Repeat the above steps to continuously update all messages of each historical spatiotemporal rock layer sample graph by forward and backward propagation of messages along the edge in the directed edge graph until the maximum iteration frequency is reached, generating multiple forward edge probability inference messages and multiple reverse edge probability inference messages;

[0017] The edge restoration is calculated by combining multiple forward edge probability inference messages and multiple reverse edge probability inference messages to obtain the edge probability of rock layer characteristic change, and the settlement monitoring level of the target layer monitoring area is determined based on the edge probability of rock layer characteristic change.

[0018] More specifically, the method of extracting the ground rock structure and historical compression coefficient of each settlement monitoring layer in the target layered monitoring area, avoiding areas prone to relatively high settlement based on the ground rock structure and historical compression coefficient, and reasonably determining the deployment sites to obtain the optimal monitoring deployment sites and add them to the force sensor deployment plan specifically includes the following steps:

[0019] Obtaining basic model information of the force sensor to be used for ground subsidence monitoring in the target layered monitoring area, and retrieving monitoring specification parameters of the force sensor from the big data network based on the basic model information;

[0020] Based on big data, we obtain settlement cases in the target layered monitoring area, and extract the ground rock structure and historical compression coefficient of each settlement monitoring layer in the target layered monitoring area through the settlement cases.

[0021] Establish a deployment sampling area for each settlement monitoring layer according to the ground rock structure, divide each deployment sampling area into several sub-sampling areas according to monitoring specification parameters, and independently sample the deployment sites of force sensors in each sub-sampling area of ​​each deployment sampling area to obtain N deployment site samples of force sensors randomly provided in each sub-sampling area;

[0022] The inter-layer compression characteristic difference between each sub-sampling area is calculated based on the historical compression coefficient. If the inter-layer compression characteristic difference is greater than the preset inter-layer compression characteristic difference threshold, the sub-sampling area corresponding to the compression characteristic difference is extracted and marked as a relatively prone to subsidence area, thereby generating a pattern of relatively prone to subsidence areas between each subsidence monitoring layer.

[0023] According to the pattern of the area prone to relative settlement, the settlement orientation vector is preset, and the layout orientation of the force sensor is estimated using N layout site samples with the settlement orientation vector as the avoidance reference benchmark to obtain the expected layout estimated orientation of each settlement monitoring level;

[0024] Preset the minimum deployment mean square error, obtain the simulated monitoring variance and simulated monitoring cost of the force sensor simulated deployment at the expected deployment estimated position on each layer, and calculate the deployment mean square error of each deployment site sample based on the simulated monitoring variance and simulated monitoring cost;

[0025] Only the deployment site samples corresponding to the unique deployment mean square error less than the minimum deployment mean square error are extracted and calibrated as the optimal monitoring deployment site. The optimal monitoring deployment site is added to the force sensor deployment plan.

[0026] More specifically, force sensors are deployed at the optimal monitoring deployment sites of each settlement monitoring level to obtain compressive stress sensor readings, and the settlement crosses the multidimensional settlement cell constructed by analyzing the compressive stress sensor readings to perform isosurface interpolation of the inverse settlement distance to obtain the settlement vector exerted by the compressive stress between the upper and lower settlement monitoring levels. Specifically, the steps include:

[0027] By deploying and installing force sensors at the optimal monitoring deployment sites of each settlement monitoring layer, the ground settlement in the target layered monitoring area is monitored to obtain the compressive stress sensor readings of each settlement monitoring layer in the target layered monitoring area;

[0028] Construct a multidimensional compressive stress density space at the settlement monitoring level, obtain the contour threshold of the multidimensional compressive stress density space, assign the contour value of the compressive stress density according to the compressive stress sensor readings to construct scalar corner points, and form a multidimensional settlement unit cell of the compressive stress density based on the scalar corner points;

[0029] If the contour value of one or more scalar corner points on the multidimensional settlement cell crosses below the contour threshold of the multidimensional compressive stress density space, it means that the multidimensional settlement cell is crossed by settlement and is marked as a multidimensional crossing settlement cell. The scalar corner points on the boundary of the multidimensional crossing settlement cell that cross the isosurface are stripped off and defined as settlement crossing scalar corner points.

[0030] A reverse settlement distance weight function is preset for the time series settlement change of the compressive stress sensor readings, and the distance from each compressive stress sensor reading to the settlement monitoring layer when the force sensor does not produce a reading is calculated to obtain the reverse settlement distance. The reverse settlement distance is calculated using the reverse settlement distance weight function to obtain the reverse settlement distance weight value;

[0031] The settlement crossing intersection point and the compressive stress normal vector are determined by linear intersection interpolation of the anti-settlement distance weight value. Based on the settlement crossing intersection point and the compressive stress normal vector, the equivalent grid vertices are solved and connected to determine the settlement vector of the next settlement monitoring layer that applies the soil compressive stress of the previous settlement monitoring layer.

[0032] More specifically, the method of determining the settlement crossing intersection point and the compressive stress normal vector by linear interpolation of the anti-settlement distance weight value, solving and connecting the equivalent grid vertices based on the settlement crossing intersection point and the compressive stress normal vector to determine the settlement vector and settlement trend of the soil compressive stress applied by the previous settlement monitoring level to the next settlement monitoring level, specifically includes the following steps:

[0033] Performing linear intersection interpolation on one or more settlement crossing scalar corner points using the anti-settlement distance weight value to determine the settlement crossing intersection point and obtain the compressive stress normal vector of the settlement crossing intersection point in the multi-dimensional compressive stress density space;

[0034] For each multidimensional crossing settlement cell, a crossing equivalent vertex is defined, a quadratic error function is constructed based on the settlement crossing intersection point and the compressive stress normal vector, and the quadratic error function is solved to generate an equivalent mesh vertex of the multidimensional crossing settlement cell;

[0035] If there are crossing equivalue vertices in cells adjacent to the current multi-dimensional crossing settlement cell, the equivalued grid vertices are connected to obtain the settlement compressive stress isosurface map of the settlement monitoring level. The settlement compressive stress isosurface map is analyzed to determine the settlement vector of the next settlement monitoring level that applies the soil compressive stress of the previous settlement monitoring level.

[0036] More specifically, the inversion and inference of settlement vectors is performed to obtain the relative settlement displacement amplitude of each settlement monitoring layer, and the trend reconstruction and quantitative projection embedding of the relative displacement between each settlement monitoring layer are performed based on the linear correlation of the relative settlement displacement amplitude to obtain the ground settlement monitoring amount of the target layered monitoring area, which specifically includes the following steps:

[0037] The belief displacement prior distribution of soil settlement in the target layer monitoring area is obtained through the ground rock structure, stratum lithology characteristics and settlement factors in the big data network, and the displacement inversion model is constructed based on the prior node pattern and parameters of the belief displacement prior distribution;

[0038] The maximum likelihood method is introduced to calculate the settlement vectors between the settlement monitoring levels to obtain the settlement likelihood function of each settlement monitoring level. The displacement inversion model is used to infer the displacement posterior of the settlement monitoring level under the condition of the settlement vector corresponding to the settlement likelihood function to obtain the relative settlement displacement amplitude of each settlement monitoring level.

[0039] Obtaining the response timing step lengths when the force sensors of adjacent settlement monitoring layers monitor and output compressive stress sensing readings, and presetting the reconstruction error threshold of the relative displacement between adjacent settlement monitoring layers based on the response timing step lengths;

[0040] The least squares neighbor linear correlation method is introduced to calculate the weight between the relative settlement displacement amplitude of each settlement monitoring layer and the relative settlement displacement amplitude of the adjacent settlement monitoring layer until the reconstruction error of the neighbor linear correlation is less than the relative displacement reconstruction error threshold, and the global correlation weight matrix is ​​obtained;

[0041] Construct a low-dimensional settlement space and use the global correlation weight matrix to obtain K settlement trend eigenvalues ​​and the settlement trend eigenvectors corresponding to each settlement trend eigenvalue;

[0042] Based on K settlement trend eigenvalues, the settlement trend eigenvector is embedded into the low-dimensional settlement space to perform quantitative projection expression of the relative displacement between each settlement monitoring layer, and finally the ground settlement monitoring amount of the target layered monitoring area is obtained.

[0043] The second aspect of the present invention provides a ground subsidence stratification monitoring system based on a force sensor, the ground subsidence stratification monitoring system includes a memory and a processor, the memory stores a ground subsidence stratification monitoring method program based on a force sensor, when the ground subsidence stratification monitoring method program is executed by the processor, any one of the steps of the ground subsidence stratification monitoring method is implemented.

[0044] The present invention solves the technical defects existing in the background technology, and the beneficial technical effects of the present invention are:

[0045] A feature change inference network is constructed to estimate several historical spatiotemporal rock strata sample maps within the target stratified monitoring area, generating a directed edge map of rock strata settlement. The directed edge map is then used to infer the edge probability of rock strata feature change through forward and backward message transfer to determine the settlement monitoring levels within the target stratified monitoring area. The ground rock strata structure and historical compression coefficient of each settlement monitoring level within the target stratified monitoring area are extracted. Based on the ground rock strata structure and historical compression coefficient, reasonable deployment sites are determined to avoid areas prone to relative settlement. The optimal monitoring deployment sites are obtained and added to the force sensor deployment plan. Force sensors are deployed at the optimal monitoring deployment sites at each settlement monitoring level to obtain compressive stress sensor readings. The settlement traverses the multidimensional settlement cells constructed from the compressive stress sensor readings, and isosurface interpolation of inverse settlement distances is performed to obtain the settlement vectors imposed by the compressive stress between the upper and lower settlement monitoring levels. The settlement vectors are inverted to obtain the relative settlement displacement amplitudes of each settlement monitoring level. The relative settlement displacement amplitudes are linearly correlated to perform trend reconstruction and quantitative projection embedding of the relative displacements between the settlement monitoring levels to obtain the ground settlement monitoring volume within the target stratified monitoring area. The present invention can utilize force sensors to perform settlement stratification, settlement vector monitoring and analysis, and quantitative calculation of settlement trends in the target area, thereby achieving a layered settlement monitoring effect, accurately outputting the settlement degree of different layers of the ground, and significantly improving the reliability and credibility of the ground settlement monitoring results. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, without paying any creative work, they can also obtain drawings of other embodiments based on these drawings.

[0047] Figure 1 A first method flow chart of a method for monitoring ground subsidence stratification based on a force sensor is shown;

[0048] Figure 2A second method flow chart of a method for monitoring ground subsidence stratification based on a force sensor is shown;

[0049] Figure 3 The system framework diagram of a ground subsidence stratification monitoring system based on a force sensor is shown. DETAILED DESCRIPTION

[0050] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features therein can be combined with each other.

[0051] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0052] The first aspect of the present invention provides a method for monitoring ground subsidence stratification based on a force sensor, such as Figure 1 As shown, the following steps are included:

[0053] A feature change inference network is constructed to estimate several historical spatiotemporal rock layer sample maps within the target layered monitoring area, generating a directed edge map of rock layer subsidence. The directed edge map is then used to infer the edge probability of rock layer feature change by forward and reverse message transfer to determine the subsidence monitoring level of the target layered monitoring area.

[0054] Extract the ground rock structure and historical compression coefficient of each settlement monitoring layer in the target layered monitoring area, and make reasonable deployment decisions based on the ground rock structure and historical compression coefficient to avoid areas prone to relative settlement. Optimal monitoring deployment sites are obtained and added to the force sensor deployment plan.

[0055] Force sensors are deployed at the optimal monitoring deployment sites of each settlement monitoring layer to obtain compressive stress sensor readings. The settlement crosses the multidimensional settlement cell constructed by analyzing the compressive stress sensor readings, and the isosurface interpolation of the inverse settlement distance is performed to obtain the settlement vector exerted by the compressive stress between the upper and lower settlement monitoring layers.

[0056] The settlement vector is inverted and inferred to obtain the relative settlement displacement amplitude of each settlement monitoring layer. The relative settlement displacement amplitude is linearly correlated to perform trend reconstruction and quantitative projection embedding of the relative displacement between each settlement monitoring layer to obtain the ground settlement monitoring amount of the target layered monitoring area.

[0057] More specifically, the method of constructing a feature change inference network to estimate several historical spatiotemporal rock layer sample maps within the target layered monitoring area, generating a directed edge map of rock layer subsidence, and using the directed edge map to infer the edge probability of rock layer feature change by forward and reverse message transfer to determine the subsidence monitoring level of the target layered monitoring area specifically includes the following steps:

[0058] Obtain the target layered monitoring area and monitoring tasks for ground subsidence, and define the termination monitoring point for ground subsidence based on the monitoring tasks;

[0059] By monitoring the central control log, several historical spatiotemporal rock layer sample maps of the target layer monitoring area where the ground subsided to the termination monitoring point within the preset time period were extracted. The local binary pattern algorithm was introduced to calculate the characteristics of each historical spatiotemporal rock layer sample map, and the LBP rock layer characteristic value of each historical spatiotemporal rock layer sample map was obtained.

[0060] A characteristic change inference network is constructed. Based on the LBP rock layer characteristic values, the conditional probability of the rock layer characteristics shifting from a certain historical spatiotemporal rock layer sample map to another historical spatiotemporal rock layer sample map along with the sedimentation time series is estimated in the characteristic change inference network, and the joint probability of rock layer characteristic changes between each historical spatiotemporal rock layer sample map is obtained.

[0061] Define the historical spatiotemporal rock layer sample graph as a random variable node, construct a factor node for each random variable node based on the joint probability of rock layer characteristic changes, and connect each random variable node based on the factor node structure to build a directed edge graph of rock layer settlement;

[0062] The lithologic characteristics and settlement elements of the target layer monitoring area are obtained through big data. Based on the tectonic characteristics and settlement elements of the strata, a message is forwardly transmitted from the random variable node to the factor node along the stratum characteristic change edge in the directed edge graph. After the forward transmission is completed, the message is reversely transmitted from the factor node to the random variable node along the stratum characteristic change edge.

[0063] Repeat the above steps to continuously update all messages of each historical spatiotemporal rock layer sample graph by forward and backward propagation of messages along the edge in the directed edge graph until the maximum iteration frequency is reached, generating multiple forward edge probability inference messages and multiple reverse edge probability inference messages;

[0064] The edge restoration is calculated by combining multiple forward edge probability inference messages and multiple reverse edge probability inference messages to obtain the edge probability of rock layer characteristic change, and the settlement monitoring level of the target layer monitoring area is determined based on the edge probability of rock layer characteristic change.

[0065] It should be noted that traditional monitoring of ground subsidence usually uses tools such as inclinometers, settlement markers, or liquid settlement meters. However, ground subsidence is often caused by the different compressibility of different soil layers, changes in groundwater levels, construction disturbances, etc. These traditional monitoring tools are difficult to achieve stratified monitoring of rock and soil, making it impossible to accurately determine the depth and cause of subsidence and to promptly detect abnormal subsidence development, resulting in settlement warning errors and missed intervention opportunities. Therefore, precise stratified monitoring of rock and soil is crucial. To this end, this method first extracts the characteristics of historical spatiotemporal rock layer samples in the target stratified monitoring area to obtain the LBP rock layer characteristic value of each sample. The LBP rock layer characteristic value characterizes the category distinction labels of different geological rock layers contained in the target stratified monitoring area, which can provide a basis for defining geological monitoring strata with different rock layer distributions. Since the historical spatiotemporal rock layer sample map is a series of historical data, the corresponding calculated LBP rock layer characteristic value presents a characteristic value of the historical changes between different types of rock layers. Therefore, the LBP rock layer characteristic value can be used to further infer the joint probability of rock layer characteristic changes between each historical spatiotemporal rock layer sample map in the characteristic change inference network, thereby revealing the law of settlement characteristic changes between different rock and soil layers. Then, the factor node constructed based on the joint probability of rock layer characteristic changes is connected to each historical spatiotemporal rock layer sample map, so that the directed factor distribution of the edges of each rock and soil layer in the target stratified monitoring area can be clarified, and the dependence relationship between spatiotemporal rock layer samples under historical settlement conditions can be clearly expressed, providing a credible structural basis for the transmission and inference of settlement information at the edge of the rock and soil layer.

[0066] It should be noted that settlement factors include Quaternary loose sediments, groundwater distribution, and artificial fill. Because the lithologic characteristics and settlement factors of different rock layers directly or indirectly contribute to their settlement, they are key information transmission media for rock layer edge changes. Therefore, this method, based on the tectonic characteristics of the strata and settlement factors, transmits messages forward and backward from random variable nodes to factor nodes along the rock layer characteristic change edges in a directed edge graph. The forward message transfer informs neighboring factor nodes of their own edge change recognition probability, while the backward message transfer integrates the messages from all neighboring factor nodes based on their internal joint probability function structure to inform a random variable node of the edge change recognition probability it should tend to. Thus, through the message exchange between factor nodes, each random variable node gradually approaches the edge change probability under historical settlement conditions, thereby characterizing the random evolution law of settlement edges between rock layers. Finally, multiple forward edge probability inference messages and multiple reverse edge probability inference messages are combined to restore and calculate the edge settlement of each rock and soil layer. This method can infer the appropriate settlement monitoring layer within the target layer monitoring area, which is easy and excellent for monitoring settlement behavior. This method can infer the historical characteristic change data of different rock and soil layer samples in the target area to establish a reasonable monitoring layer, significantly improving the accuracy of determining the depth and cause of settlement, effectively optimizing the settlement assessment performance of ground foundation structures, and making ground settlement monitoring more accurate and reliable.

[0067] More specifically, the method of extracting the ground rock structure and historical compression coefficient of each settlement monitoring layer in the target layered monitoring area, avoiding areas prone to relatively high settlement based on the ground rock structure and historical compression coefficient, and reasonably determining the deployment sites to obtain the optimal monitoring deployment sites and add them to the force sensor deployment plan specifically includes the following steps:

[0068] Obtaining basic model information of the force sensor to be used for ground subsidence monitoring in the target layered monitoring area, and retrieving monitoring specification parameters of the force sensor from the big data network based on the basic model information;

[0069] Based on big data, we obtain settlement cases in the target layered monitoring area, and extract the ground rock structure and historical compression coefficient of each settlement monitoring layer in the target layered monitoring area through the settlement cases.

[0070] Establish a deployment sampling area for each settlement monitoring layer according to the ground rock structure, divide each deployment sampling area into several sub-sampling areas according to monitoring specification parameters, and independently sample the deployment sites of force sensors in each sub-sampling area of ​​each deployment sampling area to obtain N deployment site samples of force sensors randomly provided in each sub-sampling area;

[0071] The inter-layer compression characteristic difference between each sub-sampling area is calculated based on the historical compression coefficient. If the inter-layer compression characteristic difference is greater than the preset inter-layer compression characteristic difference threshold, the sub-sampling area corresponding to the compression characteristic difference is extracted and marked as a relatively prone to subsidence area, thereby generating a pattern of relatively prone to subsidence areas between each subsidence monitoring layer.

[0072] According to the pattern of the area prone to relative settlement, the settlement orientation vector is preset, and the layout orientation of the force sensor is estimated using N layout site samples with the settlement orientation vector as the avoidance reference benchmark to obtain the expected layout estimated orientation of each settlement monitoring level;

[0073] Preset the minimum deployment mean square error, obtain the simulated monitoring variance and simulated monitoring cost of the force sensor simulated deployment at the expected deployment estimated position on each layer, and calculate the deployment mean square error of each deployment site sample based on the simulated monitoring variance and simulated monitoring cost;

[0074] Only the deployment site samples corresponding to the unique deployment mean square error less than the minimum deployment mean square error are extracted and calibrated as the optimal monitoring deployment site. The optimal monitoring deployment site is added to the force sensor deployment plan.

[0075] It should be noted that compared to traditional monitoring tools, force sensors can continuously collect minute changes in stress, strain, and relative displacement in real time, thereby obtaining more comprehensive information on structural behavior. They can also be connected to data acquisition systems (DAS) and cloud platforms, enabling unattended automated monitoring and significantly reducing the maintenance costs of traditional equipment. However, if force sensors are haphazardly installed and deployed at each settlement monitoring level, this can lead to local or global linear and nonlinear deviations in the force sensor monitoring data, resulting in significant errors in the ground subsidence estimates based on these force sensor data, seriously impacting geological hazard assessments and urban construction project planning. Therefore, the precise and reasonable deployment of force sensors at each settlement monitoring level is crucial. To address this issue, this method independently samples deployment sites within the deployment sampling area of ​​each settlement monitoring level to obtain deployment site samples. Due to the different compression coefficient ranges between different settlement monitoring levels under the premise of different actual monitoring and governance needs, some local areas between adjacent settlement monitoring levels may have settlement monitoring deviations restricted by interlayer compression characteristics. To this end, this method calculates the interlayer compression characteristic difference between each sub-sampling area. If the interlayer compression characteristic difference is greater than the preset interlayer compression characteristic difference threshold, it means that a local position on the settlement monitoring level is more prone to compression of interlayer relative displacement. The interlayer compression characteristics of the sub-sampling area corresponding to the local position will seriously interfere with the monitoring performance of the force sensor. Therefore, it is necessary to significantly mark these areas prone to relative settlement, so as to clarify the pattern of areas prone to relative settlement that are extremely likely to reduce the monitoring performance of the force sensor between all settlement monitoring levels, and further provide a hierarchical local chain basis for eliminating unreasonable layout and installation positions.

[0076] It should be noted that when the pattern of areas prone to relative settlement is generated, it is clear that the unreasonable points at each layer where force sensors need to be installed should be avoided. Therefore, this method presets the settlement orientation vector based on the pattern of areas prone to relative settlement. This settlement orientation vector serves as a guide for the avoidance direction. Compared with the traditional installation planning method of randomly selecting the orientation, it can significantly improve the reliability and stability of force sensors in accurately monitoring settlement data between adjacent layers. Finally, the deployment mean square error of the force sensor at the expected estimated orientation of each layer is simulated. This deployment mean square error expresses the performance index of the force sensor in simulating ground settlement monitoring at the expected estimated orientation. If the deployment mean square error is less than the minimum deployment mean square error, it means that the force sensor has the best simulated monitoring performance at the expected estimated orientation, and therefore it is selected as the optimal monitoring deployment location for the force sensor. This method can be used to sample and arrange site samples based on the ground rock structure and historical compression coefficient of each settlement monitoring layer in the target layered monitoring area and avoid screening, thereby making more reasonable, accurate and stable layout and installation site decisions for force sensors, maximizing the monitoring performance of force sensors for different settlement monitoring layers, and significantly improving the accuracy of ground settlement monitoring results.

[0077] More specifically, force sensors are deployed at the optimal monitoring deployment sites of each settlement monitoring level to obtain compressive stress sensor readings, and the settlement crosses the multidimensional settlement cell constructed by analyzing the compressive stress sensor readings to perform isosurface interpolation of the anti-settlement distance, and obtain the settlement vector exerted by the compressive stress between the upper and lower settlement monitoring levels, such as Figure 2 As shown, the specific steps include:

[0078] S202: performing sensor monitoring on the ground settlement in the target layered monitoring area by deploying and installing force sensors at the optimal monitoring deployment sites of each settlement monitoring layer, so as to obtain compressive stress sensor readings of each settlement monitoring layer in the target layered monitoring area;

[0079] S204: constructing a multidimensional compressive stress density space of the settlement monitoring level, obtaining contour thresholds of the multidimensional compressive stress density space, assigning contour values ​​of the compressive stress density according to the compressive stress sensor readings to construct scalar corner points, and forming a multidimensional settlement cell of the compressive stress density based on the scalar corner points;

[0080] S206: If the contour line values ​​of one or more scalar corner points on the multidimensional settlement cell cross below the contour line threshold of the multidimensional compressive stress density space, it means that the multidimensional settlement cell is crossed by settlement and is marked as a multidimensional crossing settlement cell. The scalar corner points on the boundary of the multidimensional crossing settlement cell that cross the isosurface are stripped off and defined as settlement crossing scalar corner points.

[0081] S208: Preset an anti-settlement distance weight function for the time series settlement change of the compressive stress sensor reading, calculate the distance of each compressive stress sensor reading to the settlement monitoring level when the force sensor does not produce a reading, obtain the anti-settlement distance, and calculate the anti-settlement distance using the anti-settlement distance weight function to obtain an anti-settlement distance weight value;

[0082] S210: Determine the settlement crossing intersection point and the compressive stress normal vector using the linear intersection interpolation of the anti-settlement distance weight value, solve and connect the equivalent grid vertices based on the settlement crossing intersection point and the compressive stress normal vector to determine the settlement vector of the soil compressive stress applied by the previous settlement monitoring level to the next settlement monitoring level.

[0083] It should be noted that the settlement vector includes the direction, range, and situation of the settlement. When the rock and soil layer settles, its surface may collapse or stagger, which may cause the relative displacement to have peaks and valleys or faults. Each layer of force sensors monitors the vertical stress changes between the upper and lower soil layers. When the upper soil sinks, the settlement range and direction of the peaks and valleys or faults will generate additional compressive stress on the nodes of the lower layer, causing the force sensor readings to change. Therefore, when obtaining the force sensor readings of each settlement monitoring layer, this method first spatializes the multidimensional compressive stress density of each settlement monitoring layer. Because the upper soil layer may not have settlement peaks and valleys or faults before the force sensor monitors the reading, there is an isoline benchmark at this time, which can be used as the isoline threshold of the multidimensional compressive stress density space. When the compressive stress reading is generated, the soil sinks, causing peaks, valleys, and faults in the surface due to the compressive stress applied by a certain settlement vector. This results in a fluctuating expression of equal values ​​compared to before settlement. Therefore, scalar corner points are constructed by assigning the corresponding compressive stress density contour values ​​according to the compressive stress sensor readings, thereby establishing a multidimensional settlement cell that reflects the fluctuations when settlement occurs to visualize the settlement condition. Among them, if the contour values ​​of one or more scalar corner points on the multidimensional settlement cell cross below the contour threshold of the multidimensional compressive stress density space, it means that the soil at that location has sunk below the surface when the compressive stress reading was not monitored, showing a sign of settlement crossing compared to the initial surface state, indicating that the soil surface at that location has formed an undulating fluctuation with a certain settlement vector. Therefore, it is marked as a multidimensional crossing settlement cell, and the settlement crossing scalar corner points on its boundary that pass below the isosurface are obtained.

[0084] It should be noted that there may be one or more settlement crossing scalar corner points below the isosurface. These settlement crossing scalar corner points are important source data for isovalue expression. The crossing of settlement crossing scalar corner points causes the applied compressive stress to gradually generate and change, and there is a close connection between the two. Therefore, this method calculates the reverse settlement distance of the settlement monitoring layer from each compressive stress sensor reading to the time when the force sensor does not generate a reading. In short, the reverse settlement distance is the settlement distance between the initial soil layer and the soil layer fluctuation when the force sensor gradually generates a compressive stress reading. The reverse settlement distance weight value can reflect the applied weight of each compressive stress when settlement fluctuation occurs. Then, according to the reverse settlement distance weight value, the settlement crossing scalar corner points are linearly interpolated one by one to determine the settlement crossing intersection point and the compressive stress normal vector, thereby further drawing the isosurface map when settlement fluctuation occurs in the soil layer. The isosurface map can clearly and completely determine the settlement vector of the compressive stress applied to the next layer of soil when each settlement monitoring layer settles. This method can accurately and efficiently depict the vectors and trends of different layers of soil that exert compressive stress on the next layer of soil when settlement occurs in the form of isosurface maps, thereby visualizing the current status of settlement development in ground stratification monitoring, making the subsequent concrete analysis and calculation of settlement degree more accurate and reliable, and greatly improving the credibility of monitoring results.

[0085] More specifically, the method of determining the settlement crossing intersection point and the compressive stress normal vector by using the linear intersection interpolation of the anti-settlement distance weight value, solving and connecting the equivalent grid vertices based on the settlement crossing intersection point and the compressive stress normal vector to determine the settlement vector of the soil compressive stress applied by the previous settlement monitoring level to the next settlement monitoring level specifically includes the following steps:

[0086] Performing linear intersection interpolation on one or more settlement crossing scalar corner points using the anti-settlement distance weight value to determine the settlement crossing intersection point and obtain the compressive stress normal vector of the settlement crossing intersection point in the multi-dimensional compressive stress density space;

[0087] For each multidimensional crossing settlement cell, a crossing equivalent vertex is defined, a quadratic error function is constructed based on the settlement crossing intersection point and the compressive stress normal vector, and the quadratic error function is solved to generate an equivalent mesh vertex of the multidimensional crossing settlement cell;

[0088] If there are crossing equivalue vertices in cells adjacent to the current multi-dimensional crossing settlement cell, the equivalued grid vertices are connected to obtain the settlement compressive stress isosurface map of the settlement monitoring level. The settlement compressive stress isosurface map is analyzed to determine the settlement vector of the next settlement monitoring level that applies the soil compressive stress of the previous settlement monitoring level.

[0089] It should be noted that by linearly interpolating one or more settlement crossing scalar corner points using the weighted values ​​of the inverse settlement distance, the applied compressive stress weight can be assigned to the multidimensional settlement cells where settlement fluctuations occur. This allows the multidimensional settlement cells corresponding to the settlement fluctuations that form peaks, valleys, or faults to define the corresponding level of fluctuations as the compressive stress readings change over time. This allows the isosurfaces of settlement fluctuations to have more accurate compressive stress expression values, improving the visualization accuracy of the isosurfaces. Because the mesh vertices of the isosurfaces are no longer restricted to the edges of the multidimensional settlement crossing cells, but can drift freely to the optimal position, the quadratic error function constructed from the compressive stress normal vectors of the settlement crossing intersections after interpolation and the corresponding solution can better restore complex or sharp shapes (such as corners), making the isosurface description of different peaks, valleys, and faults formed by settlement more flexible and adaptable, and maximizing the detailed expression of settlement vectors and trends. If a cell adjacent to the current multi-dimensional crossing settlement cell has a crossing equivalence vertex, it means that the adjacent cell also has a crossing state due to a certain settlement vector and trend compared to the initial soil layer. Therefore, it is included in the equivalent category of settlement fluctuation and is jointly expressed, ultimately forming a settlement compressive stress isosurface map of the settlement monitoring layer. This method can visualize the fluctuation of compressive stress exerted on the next soil layer by settlement of each settlement monitoring layer, thereby more accurately revealing the settlement vector of each rock and soil layer and providing a more reliable analysis basis for subsequent calculation of the global settlement degree.

[0090] More specifically, the inversion and inference of settlement vectors is performed to obtain the relative settlement displacement amplitude of each settlement monitoring layer, and the trend reconstruction and quantitative projection embedding of the relative displacement between each settlement monitoring layer are performed based on the linear correlation of the relative settlement displacement amplitude to obtain the ground settlement monitoring amount of the target layered monitoring area, which specifically includes the following steps:

[0091] The belief displacement prior distribution of soil settlement in the target layer monitoring area is obtained through the ground rock structure, stratum lithology characteristics and settlement factors in the big data network, and the displacement inversion model is constructed based on the prior node pattern and parameters of the belief displacement prior distribution;

[0092] The maximum likelihood method is introduced to calculate the settlement vectors between the settlement monitoring levels to obtain the settlement likelihood function of each settlement monitoring level. The displacement inversion model is used to infer the displacement posterior of the settlement monitoring level under the condition of the settlement vector corresponding to the settlement likelihood function to obtain the relative settlement displacement amplitude of each settlement monitoring level.

[0093] Obtaining the response timing step lengths when the force sensors of adjacent settlement monitoring layers monitor and output compressive stress sensing readings, and presetting the reconstruction error threshold of the relative displacement between adjacent settlement monitoring layers based on the response timing step lengths;

[0094] The least squares neighbor linear correlation method is introduced to calculate the weight between the relative settlement displacement amplitude of each settlement monitoring layer and the relative settlement displacement amplitude of the adjacent settlement monitoring layer until the reconstruction error of the neighbor linear correlation is less than the relative displacement reconstruction error threshold, and the global correlation weight matrix is ​​obtained;

[0095] Construct a low-dimensional settlement space and use the global correlation weight matrix to obtain K settlement trend eigenvalues ​​and the settlement trend eigenvectors corresponding to each settlement trend eigenvalue;

[0096] Based on K settlement trend eigenvalues, the settlement trend eigenvector is embedded into the low-dimensional settlement space to perform quantitative projection expression of the relative displacement between each settlement monitoring layer, and finally the ground settlement monitoring amount of the target layered monitoring area is obtained.

[0097] It should be noted that the settlement vector that applies compressive stress to the force sensor when each settlement monitoring layer settles is the basic structure of settlement monitoring, which describes the magnitude and amplitude of the development of rock and soil settlement in each layer. However, since the settlement vector contains different multidimensional settlement indicators, it is difficult to quantify these multidimensional settlement indicators into accurate settlement monitoring values ​​using traditional methods. In response to this, this method uses the displacement inversion model constructed by the prior distribution of the belief displacement of the soil layer settlement in the target layered monitoring area to perform a global inversion inference on the settlement vector of each monitoring layer, thereby clarifying the specific amplitude of the relative settlement displacement caused by settlement between adjacent settlement monitoring layers. The actual soil settlement is inferred by the prior distribution of historical data. Compared with traditional methods, the calculation of relative displacement can be made more accurate and saturated, ensuring that the subsequent calculation of settlement degree is more stable and reliable. Because the monitoring response of the force sensor will have a certain time series step when relative settlement displacement occurs between each layer of soil under the condition of the settlement vector, this is the trend connection error of the multidimensional indicators in the settlement vector. Therefore, the reconstruction error threshold of the relative displacement between adjacent settlement monitoring layers is preset based on the response time series step. The least squares method is introduced to linearly associate neighbor points to reconstruct the weight of the relative settlement displacement amplitude between each layer. The local geometric structure of the data point within the settlement vector is extracted and encoded. For example, if a relative settlement displacement amplitude is slightly to the left of its neighbors in the settlement vector, this information is stored in the corresponding weight, thus obtaining a global correlation weight matrix for each multidimensional indicator in the settlement vector to describe the relative settlement displacement trend. Finally, the settlement trend eigenvalue and the corresponding settlement trend eigenvector are solved and further embedded in the low-dimensional settlement space for the projection expression of the settlement trend, so that the ground settlement monitoring amount in the target layer monitoring area can be accurately quantified. On the one hand, this method can replace the tedious steps of traditional manual monitoring and calculation of ground subsidence, save time and effort, reduce manual intervention errors and unnecessary calculation processes, and improve the efficiency of layered monitoring of ground subsidence; on the other hand, it uses the multidimensional indicators covered in the inverted settlement vector of the prior distribution to infer the relative settlement amplitude of each monitored soil layer, and discovers and reconstructs the settlement trend expressed in the multidimensional indicators to quantify the accurate ground subsidence results, thereby improving the monitoring accuracy of specific rock and soil settlement between different layers, optimizing the quality of monitoring output, and improving the credibility of monitoring results.

[0098] The second aspect of the present invention provides a ground subsidence stratification monitoring system based on a force sensor, such as Figure 3 As shown, the ground subsidence stratification monitoring system includes a memory 31 and a processor 32. The memory 31 stores a ground subsidence stratification monitoring method program based on a force sensor. When the ground subsidence stratification monitoring method program is executed by the processor 32, any one of the steps of the ground subsidence stratification monitoring method is implemented.

[0099] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for monitoring ground subsidence stratification based on force sensors, characterized in that: The following steps are involved: A feature change inference network is constructed to estimate several historical spatiotemporal rock layer sample maps within the target layered monitoring area, generating a directed edge map of rock layer subsidence. The directed edge map is then used to infer the edge probability of rock layer feature change by forward and reverse message transfer to determine the subsidence monitoring level of the target layered monitoring area. Extract the ground rock structure and historical compression coefficient of each settlement monitoring layer in the target layered monitoring area, and make reasonable deployment decisions based on the ground rock structure and historical compression coefficient to avoid areas prone to relative settlement. Optimal monitoring deployment sites are obtained and added to the force sensor deployment plan. Force sensors are deployed at the optimal monitoring deployment sites of each settlement monitoring layer to obtain compressive stress sensor readings. The settlement crosses the multidimensional settlement cell constructed by analyzing the compressive stress sensor readings, and the isosurface interpolation of the inverse settlement distance is performed to obtain the settlement vector exerted by the compressive stress between the upper and lower settlement monitoring layers. The settlement vector is inverted and inferred to obtain the relative settlement displacement amplitude of each settlement monitoring layer. The relative settlement displacement amplitude is linearly correlated with the neighbors to perform trend reconstruction and quantitative projection embedding of the relative displacement between each settlement monitoring layer to obtain the ground settlement monitoring amount of the target layered monitoring area.

2. A method for monitoring ground subsidence stratification based on a force sensor according to claim 1, characterized in that: The method of constructing a feature change inference network to estimate several historical spatiotemporal rock layer sample maps within the target layered monitoring area, generating a directed edge map of rock layer subsidence, and using the directed edge map to infer the edge probability of rock layer feature change by forward and reverse message transmission to determine the subsidence monitoring level of the target layered monitoring area specifically includes the following steps: Obtain the target layered monitoring area and monitoring tasks for ground subsidence, and define the termination monitoring point for ground subsidence based on the monitoring tasks; By monitoring the central control log, several historical spatiotemporal rock layer sample maps of the target layer monitoring area where the ground subsided to the termination monitoring point within the preset time period were extracted. The local binary pattern algorithm was introduced to calculate the characteristics of each historical spatiotemporal rock layer sample map, and the LBP rock layer characteristic value of each historical spatiotemporal rock layer sample map was obtained. A characteristic change inference network is constructed. Based on the LBP rock layer characteristic values, the conditional probability of the rock layer characteristics shifting from a certain historical spatiotemporal rock layer sample map to another historical spatiotemporal rock layer sample map along with the sedimentation time series is estimated in the characteristic change inference network, and the joint probability of rock layer characteristic changes between each historical spatiotemporal rock layer sample map is obtained. Define the historical spatiotemporal rock layer sample graph as a random variable node, construct a factor node for each random variable node based on the joint probability of rock layer characteristic changes, and connect each random variable node based on the factor node structure to build a directed edge graph of rock layer settlement; The lithologic characteristics and settlement elements of the target layer monitoring area are obtained through big data. Based on the tectonic characteristics and settlement elements of the strata, a message is forwardly transmitted from the random variable node to the factor node along the stratum characteristic change edge in the directed edge graph. After the forward transmission is completed, the message is reversely transmitted from the factor node to the random variable node along the stratum characteristic change edge. Repeat the above steps to continuously update all messages of each historical spatiotemporal rock layer sample graph by forward and backward propagation of messages along the edge in the directed edge graph until the maximum iteration frequency is reached, generating multiple forward edge probability inference messages and multiple reverse edge probability inference messages; The edge restoration is calculated by combining multiple forward edge probability inference messages and multiple reverse edge probability inference messages to obtain the edge probability of rock layer characteristic change, and the settlement monitoring level of the target layer monitoring area is determined based on the edge probability of rock layer characteristic change.

3. The method for monitoring ground subsidence stratification based on a force sensor according to claim 1, characterized in that: The method comprises extracting the ground rock structure and historical compression coefficient of each settlement monitoring layer in the target layered monitoring area, avoiding areas prone to relatively high settlement based on the ground rock structure and historical compression coefficient, and reasonably determining the deployment sites to obtain the optimal monitoring deployment sites and add them to the force sensor deployment plan. Specifically, the method comprises the following steps: Obtaining basic model information of the force sensor to be used for ground subsidence monitoring in the target layered monitoring area, and retrieving monitoring specification parameters of the force sensor from the big data network based on the basic model information; Based on big data, we obtain settlement cases in the target layered monitoring area, and extract the ground rock structure and historical compression coefficient of each settlement monitoring layer in the target layered monitoring area through the settlement cases. Establish a deployment sampling area for each settlement monitoring layer according to the ground rock structure, divide each deployment sampling area into several sub-sampling areas according to monitoring specification parameters, and independently sample the deployment sites of force sensors in each sub-sampling area of ​​each deployment sampling area to obtain N deployment site samples of force sensors randomly provided in each sub-sampling area; The inter-layer compression characteristic difference between each sub-sampling area is calculated based on the historical compression coefficient. If the inter-layer compression characteristic difference is greater than the preset inter-layer compression characteristic difference threshold, the sub-sampling area corresponding to the compression characteristic difference is extracted and marked as a relatively prone to subsidence area, thereby generating a pattern of relatively prone to subsidence areas between each subsidence monitoring layer. According to the pattern of the area prone to relative settlement, the settlement orientation vector is preset, and the layout orientation of the force sensor is estimated using N layout site samples with the settlement orientation vector as the avoidance reference benchmark to obtain the expected layout estimated orientation of each settlement monitoring level; Preset the minimum deployment mean square error, obtain the simulated monitoring variance and simulated monitoring cost of the force sensor simulated deployment at the expected deployment estimated position on each layer, and calculate the deployment mean square error of each deployment site sample based on the simulated monitoring variance and simulated monitoring cost; Only the deployment site samples corresponding to the unique deployment mean square error less than the minimum deployment mean square error are extracted and calibrated as the optimal monitoring deployment site. The optimal monitoring deployment site is added to the force sensor deployment plan.

4. The method for monitoring ground subsidence stratification based on a force sensor according to claim 1, characterized in that: The method comprises the following steps: deploying force sensors at the optimal monitoring deployment sites of each settlement monitoring layer to obtain compressive stress sensor readings; performing isosurface interpolation of the inverse settlement distance on the multidimensional settlement cells constructed by analyzing the compressive stress sensor readings; and obtaining the settlement vector exerted by the compressive stress between the upper and lower settlement monitoring layers. By deploying and installing force sensors at the optimal monitoring deployment sites of each settlement monitoring layer, the ground settlement in the target layered monitoring area is monitored to obtain the compressive stress sensor readings of each settlement monitoring layer in the target layered monitoring area; Construct a multidimensional compressive stress density space at the settlement monitoring level, obtain the contour threshold of the multidimensional compressive stress density space, assign the contour value of the compressive stress density according to the compressive stress sensor readings to construct scalar corner points, and form a multidimensional settlement unit cell of the compressive stress density based on the scalar corner points; If the contour value of one or more scalar corner points on the multidimensional settlement cell crosses below the contour threshold of the multidimensional compressive stress density space, it means that the multidimensional settlement cell is crossed by settlement and is marked as a multidimensional crossing settlement cell. The scalar corner points on the boundary of the multidimensional crossing settlement cell that cross the isosurface are stripped off and defined as settlement crossing scalar corner points. A reverse settlement distance weight function is preset for the time series settlement change of the compressive stress sensor readings, and the distance from each compressive stress sensor reading to the settlement monitoring layer when the force sensor does not produce a reading is calculated to obtain the reverse settlement distance. The reverse settlement distance is calculated using the reverse settlement distance weight function to obtain the reverse settlement distance weight value; The settlement crossing intersection point and the compressive stress normal vector are determined by linear intersection interpolation of the anti-settlement distance weight value. Based on the settlement crossing intersection point and the compressive stress normal vector, the equivalent grid vertices are solved and connected to determine the settlement vector of the next settlement monitoring layer that applies the soil compressive stress of the previous settlement monitoring layer.

5. The method for monitoring ground subsidence stratification based on a force sensor according to claim 4, characterized in that: The method uses linear intersection interpolation of anti-settlement distance weight values ​​to determine the settlement crossing intersection point and the compressive stress normal vector, solves and connects the equivalent grid vertices based on the settlement crossing intersection point and the compressive stress normal vector, and determines the settlement vector and settlement trend of the soil layer compressive stress applied by the previous settlement monitoring level to the next settlement monitoring level. Specifically, the method includes the following steps: Performing linear intersection interpolation on one or more settlement crossing scalar corner points using the anti-settlement distance weight value to determine the settlement crossing intersection point and obtain the compressive stress normal vector of the settlement crossing intersection point in the multi-dimensional compressive stress density space; For each multidimensional crossing settlement cell, a crossing equivalent vertex is defined, a quadratic error function is constructed based on the settlement crossing intersection point and the compressive stress normal vector, and the quadratic error function is solved to generate an equivalent mesh vertex of the multidimensional crossing settlement cell; If there are crossing equivalue vertices in cells adjacent to the current multi-dimensional crossing settlement cell, the equivalued grid vertices are connected to obtain the settlement compressive stress isosurface map of the settlement monitoring level. The settlement compressive stress isosurface map is analyzed to determine the settlement vector of the next settlement monitoring level that applies the soil compressive stress of the previous settlement monitoring level.

6. The method for monitoring ground subsidence stratification based on a force sensor according to claim 1, characterized in that: The inversion and inference of the settlement vector is used to obtain the relative settlement displacement amplitude of each settlement monitoring layer, and the relative settlement displacement amplitude is linearly correlated with the neighbors to perform trend reconstruction and quantitative projection embedding of the relative displacement between each settlement monitoring layer to obtain the ground settlement monitoring amount of the target layered monitoring area, which specifically includes the following steps: The belief displacement prior distribution of soil settlement in the target layer monitoring area is obtained through the ground rock structure, stratum lithology characteristics and settlement factors in the big data network, and the displacement inversion model is constructed based on the prior node pattern and parameters of the belief displacement prior distribution; The maximum likelihood method is introduced to calculate the settlement vectors between the settlement monitoring levels to obtain the settlement likelihood function of each settlement monitoring level. The displacement inversion model is used to infer the displacement posterior of the settlement monitoring level under the condition of the settlement vector corresponding to the settlement likelihood function to obtain the relative settlement displacement amplitude of each settlement monitoring level. Obtaining the response timing step lengths when the force sensors of adjacent settlement monitoring layers monitor and output compressive stress sensing readings, and presetting the reconstruction error threshold of the relative displacement between adjacent settlement monitoring layers based on the response timing step lengths; The least squares method is introduced to calculate the weight between the relative settlement displacement amplitude of each settlement monitoring layer and the relative settlement displacement amplitude of the adjacent settlement monitoring layer, until the reconstruction error of the neighbor linear correlation is less than the relative displacement reconstruction error threshold, and the global correlation weight matrix is ​​obtained. Construct a low-dimensional settlement space and use the global correlation weight matrix to obtain K settlement trend eigenvalues ​​and the settlement trend eigenvectors corresponding to each settlement trend eigenvalue; Based on K settlement trend eigenvalues, the settlement trend eigenvector is embedded into the low-dimensional settlement space to perform quantitative projection expression of the relative displacement between each settlement monitoring layer, and finally the ground settlement monitoring amount of the target layered monitoring area is obtained.

7. A ground subsidence stratification monitoring system based on force sensors, characterized in that: The ground subsidence stratification monitoring system includes a memory and a processor. The memory stores a ground subsidence stratification monitoring method program based on a force sensor. When the ground subsidence stratification monitoring method program is executed by the processor, the ground subsidence stratification monitoring method steps as described in any one of claims 1 to 6 are implemented.

Images