A foundation pit settlement space inference method based on an improved space-time kriging model

Abstract

The application provides a foundation pit settlement space inference method based on an improved space-time Kriging model. The method specifically comprises the following steps: S1. Collecting data to establish a multi-dimensional and staged database structure; S2. Segmenting the construction period by using the collected data, and establishing a corresponding space-time covariance function for each stage; S3. Segmenting and splicing the space-time covariance functions of multiple stages to establish a space-time joint covariance structure with geological disturbance perception capability; S4. Using the defined space-time covariance function joint covariance matrix and covariance vector; S5. Defining the settlement space inference to predict the settlement value. The application can effectively deal with the settlement mutation caused by the construction stage and geological section, has higher prediction accuracy and stability, realizes the unified modeling of different stages in the whole construction process, and significantly improves the adaptability of the model to time sequence evolution and physical field response.

Description

TECHNICAL FIELD

[0001] The application relates to the combined application of civil engineering monitoring and statistical techniques, is suitable for the foundation pit system of any building structure, infrastructure, urban rail transit, etc., and belongs to the technical field of underground engineering, and particularly relates to a foundation pit settlement space reasoning method based on an improved space-time Kriging model. BACKGROUND

[0002] With the continuous deepening of the development and utilization of urban underground space, foundation pit engineering has gradually become an important part of urban infrastructure construction, and is widely used in subway stations, underground pipe corridors, large commercial bodies and other projects. However, during the excavation of the foundation pit, problems such as soil disturbance, underground water level change and stress redistribution may occur, which may cause the surface and surrounding buildings (structures) to settle, and in severe cases, may even cause structural damage. Therefore, how to accurately monitor and effectively predict the settlement of the foundation pit has become a key issue to ensure the safety of urban underground engineering. However, in actual construction scenarios, the layout of monitoring points is often limited by topography, construction conditions and budget, and it is often impossible to achieve full coverage of the entire foundation pit area, which affects the timeliness of risk assessment and emergency response.

[0003] In order to make up for the sparsity of monitoring data in space, spatial interpolation technology has become a common method for predicting the settlement values of unmeasured points, and Kriging interpolation method is widely used due to its solid theoretical foundation and high prediction accuracy. This method is based on the assumption of second-order stationarity of data, uses the covariance function to describe the correlation between data at different positions, and allocates interpolation weights accordingly. However, in complex foundation pit engineering environments, the actual settlement process often exhibits high non-stationarity and strong spatio-temporal coupling characteristics. This makes the traditional Kriging method have the following defects when solving actual engineering problems:

[0004] (1) The traditional Kriging method relies on the stationarity assumption, and the constructed covariance model is difficult to accurately fit the actual data distribution, so the interpolation result is obviously biased and difficult to adapt to complex settlement scenarios. The Kriging interpolation method is based on the assumption of second-order stationarity of data, which requires the mean of the variable to be constant and the covariance to be related only to the distance. However, actual foundation pit settlement is affected by multiple factors such as geological structure, construction disturbance and underground water change, and often exhibits non-stationarity and local mutation. The traditional method is difficult to capture such complex patterns, resulting in a decrease in interpolation accuracy, especially in the edge of the foundation pit, the structure fracture zone and the local disturbance strong area, the error is more prominent. In addition, the construction process of the empirical variogram relied on by the Kriging model is also sensitive to the initial parameters, which further restricts the stability and reliability of its engineering application.

[0005] (2) The traditional Kriging method is a static method, which does not consider the time-varying characteristics of the settlement process. In the actual construction process, the settlement of the foundation pit is a dynamic process, and the settlement value at each point is constantly changing with time. The traditional Kriging method does not consider the time-varying characteristics of the settlement process, which may lead to inaccurate prediction of the settlement value at each point.

[0006] (2) Covariance modeling relies on experience and is difficult to introduce geological and disturbance information. The covariance function in the traditional Kriging model is usually obtained by empirical fitting of the variogram function, which is difficult to flexibly introduce prior knowledge such as geological zoning, soil properties, groundwater level or construction disturbance. This modeling method lacks physical mechanism support, is easily affected by initial samples, is difficult to adapt to complex and changeable construction environment, and limits its generalization ability and stability in engineering practice.

[0007] Although existing research attempts to introduce time factors into the settlement prediction model, most other spatial reasoning methods for foundation pit settlement ignore the influence of time evolution on spatial distribution and the problem of insufficient time and space decoupling modeling when solving practical engineering problems; Most methods still model time and space as independent variables, without fully considering the interaction between them, and without depicting their interaction, the operation steps at different time periods (such as support structure construction, layered excavation, etc.) during foundation pit construction will have a significant impact on the spatial distribution of settlement. Lack of modeling of the time and space coupling relationship limits the response ability of the prediction model to dynamic evolution.

[0008] Therefore, it is urgent to propose an improved Kriging reasoning model that can fully capture the spatio-temporal dynamic coupling relationship in the settlement process and has the ability to adapt to non-stationary and nonlinear distribution. The model should not only improve the interpolation accuracy, but also have good engineering applicability and computational efficiency, so as to meet the high-frequency and real-time monitoring needs in large-scale complex foundation pit engineering. SUMMARY

[0009] To solve the problems that the Kriging interpolation method in the prior art cannot handle non-stationary settlement data, the covariance modeling cannot integrate geological disturbance information, and other spatial reasoning methods for foundation pit settlement ignore the spatio-temporal coupling effect, the present application provides a foundation pit settlement spatial reasoning method based on an improved spatio-temporal Kriging model, which establishes an improved Kriging model suitable for non-stationary scenarios, can dynamically adjust the covariance structure to cope with complex geological conditions and settlement mutation behavior, and constructs a Kriging interpolation method that integrates time evolution information, realizing spatio-temporal collaborative modeling of settlement data.

[0010] In order to achieve the above technical purposes, the present application provides a foundation pit settlement spatial reasoning method based on an improved spatio-temporal Kriging model, which specifically comprises the following steps:

[0011] S1. Systematic data collection is performed to establish a multi-dimensional, phased database structure, and dynamic monitoring data and static prior information are uniformly managed to provide complete and structured data support for subsequent phased covariance modeling and spatio-temporal reasoning; the collected data includes monitoring data of foundation pit settlement, construction process data, geological and hydrological information;

[0012] S2. Utilize the collected monitoring data, construction process data, geological and hydrological information to segment the construction period, and establish a corresponding spatio-temporal covariance function for each stage , the spatio-temporal covariance function is specifically:

[0013] ①

[0014] Among them, represents the covariance between any measuring point at any time and any measuring point at any time in the kth stage; is the total variance of the settlement data, reflecting the overall fluctuation level; is the spatial correlation function of any measuring point and any measuring point , considering the spatial position and geological conditions; is the time correlation function of any time and any time , considering the time difference and construction disturbance; is the geological parameter vector; is the disturbance state;

[0015] S3. Segment and splice the spatio-temporal covariance functions of multiple stages to establish a spatio-temporal joint covariance structure with geological disturbance perception ability:

[0016] ②

[0017] Among them, represents the covariance between any measuring point at any time and any measuring point at any time in the kth disturbance stage; is the kth disturbance stage;

[0018] S4. After establishing the spatio-temporal joint covariance structure with geological disturbance perception ability, use the defined spatio-temporal covariance function to construct the joint covariance matrix and covariance vector;

[0019] S5. Define the settlement space reasoning, that is, predict the settlement value at a certain location and time , and the prediction formula is:

[0020] ③

[0021] wherein, is the settlement value of the measuring point at time i; n is the number of possible combinations of all measuring points at all times,

[0022] n = measuring point number * time number;

[0023] is the interpolation weight to be solved, and needs to satisfy the following unbiased constraint:

[0024] ④

[0025] Based on the classical Kriging theory, the variance of the estimation error is minimized, and the optimal weighted combination is solved under the unbiased condition, and combined with the covariance matrix and the covariance vector constructed in the S4 step, formula ③ can be rewritten as:

[0026] ⑤.

[0027] The further technical scheme of the application: in addition to providing the predicted value, the method can also simultaneously evaluate the confidence of the prediction and give the variance estimation of the prediction error:

[0028] ⑥

[0029] wherein, is the covariance of the prediction point itself; the second term is the "explainable part" of the prediction point to the existing observation point information, and the smaller the residual variance is obtained after subtraction; the smaller the value is, the more reliable the prediction of the model to the point is.

[0030] The further technical scheme of the application: the foundation pit settlement monitoring data in the S1 step includes the settlement records of the total station, the settlement meter and the inclinometer at the real-time monitoring point; the construction process data includes the construction log, the excavation depth, the installation time and sequence of the supporting structure, the starting and duration of the dewatering operation; the geological and hydrological information includes the soil layer distribution, the soil mechanical parameter and the underground water level change; the above data have time stamps and spatial coordinates, and can simultaneously describe the integrity of the spatial position and the time evolution.

[0031] The further technical scheme of the application: the process of establishing a multi-dimensional and staged database structure in the S1 step includes: firstly, the monitoring data is indexed according to the time sequence and the monitoring point position to form a space-time matrix; secondly, the construction events and the disturbance factors are coded as stage labels, and are corresponded with the corresponding time intervals to be introduced into the model as control variables; finally, the static information geological and hydrological parameters are bound with the spatial coordinates of the monitoring points to form a comprehensive data set containing "position-time-attribute".

[0032] A further technical solution of the present invention: the construction period is segmented in step S2, as follows:

[0033] Assuming the construction period is divided into: Each disturbance stage is denoted as:

[0034]

[0035] Each stage Corresponding to a time interval This indicates that the nature of the disturbances is roughly the same during this stage;

[0036] For any point in time The stage label function is defined as follows:

[0037] .

[0038] A further technical solution of the present invention: In step S2, the spatial correlation function adopts a Gaussian kernel function, and an adjustment term for geological conditions is added, specifically in the form of:

[0039] ⑦

[0040] in, Point and The Euclidean distance between them; This is a spatial scale parameter that controls the rate attenuation of spatial influences. It is a geological weighting factor used to adjust for the impact of differences in geological conditions on the correlation; among which, The expression is:

[0041] ⑧

[0042] in, This is the geological influence adjustment coefficient; and Points , Geological attribute vector; denominator Used to normalize the differences in geological parameters between different measuring points.

[0043] A further technical solution of the present invention: In step S2, the time correlation function adopts an exponentially decaying kernel function combined with a construction disturbance sensitivity mechanism, in the form of:

[0044] 9

[0045] in, denotes the time difference between two time points; is a time scale parameter, controlling the decay speed of time dependence; is a disturbance phase weighting term, judging whether different construction phases are crossed; the definition of the disturbance weighting term is:

[0046] ⑩

[0047] wherein, is a disturbance influence weight; denotes an indicator function, if is different from the phase where is located, then it is 1, otherwise it is 0.

[0048] Further technical solutions of the present application: the covariance matrix between the monitoring points in the S4 step is defined as:

[0049] ⑪

[0050] The covariance vector between the to-be-predicted point and each monitoring point is defined as:

[0051] ⑫

[0052] Advantages of the present application:

[0053] (1) The present application introduces geological zoning, groundwater level, construction disturbance and other prior knowledge to participate in the construction of the covariance function. By constructing a spatio-temporal joint covariance function containing geological difference factors and disturbance perception mechanism, the real correlation between monitoring points can be accurately described, and the sudden changes in settlement caused by construction phases, geological sections, etc. can be effectively dealt with. The interpolation result has higher prediction accuracy and stability in areas with strong geological heterogeneity and frequent disturbance, and the model's interpretability and robustness are improved. The covariance structure can be dynamically adjusted to cope with complex geological conditions and settlement mutation behavior.

[0054] (2) In the present application, the construction process is unified modeled by the phase covariance function construction and weight fusion mechanism. It has accurate spatial reasoning ability at key nodes such as support, dewatering, excavation, etc., significantly improving the adaptability of the model to time evolution and physical field response.

[0055] (3) The present application adopts a modular covariance modeling structure and an adjustable parameter system, and the model is easy to integrate with actual geological data and construction logs. At the same time, the model prediction result supports error estimation and risk analysis, and the output result is visualized clearly, which facilitates the engineering personnel to quickly identify the settlement abnormal area, and assists decision-making and safety management.

[0056] The application improves the traditional Kriging method by introducing a spatio-temporal covariance structure improvement term, combines multi-scale geological information and engineering disturbance factors, and can realize high-precision spatial reasoning of foundation pit settlement in complex environment, solves the problems that the Kriging interpolation method in the prior art cannot process non-stationary settlement data, the covariance modeling cannot fuse geological disturbance information, and the spatio-temporal coupling effect is ignored, and the like. The method takes into account the calculation efficiency and reasoning accuracy, can provide strong data support in real-time monitoring and risk assessment of construction, and improves the accuracy of settlement prediction of positions without monitoring points in the foundation pit construction process. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 The flow framework diagram of the application is shown in the figure.

[0058] Figure 2 The multi-stage modeling and covariance fusion schematic diagram in the application is shown in the figure.

[0059] Figure 3 The improved spatio-temporal covariance structure schematic diagram in the application is shown in the figure. DETAILED DESCRIPTION

[0060] The application will be further described below in combination with specific embodiments. The foundation pit settlement spatial reasoning method based on the improved spatio-temporal Kriging model in the embodiments mainly aims to solve the problems that the Kriging interpolation method in the prior art cannot process non-stationary settlement data, ignores spatio-temporal coupling effect, and covariance modeling cannot fuse geological disturbance information.

[0061] The flow of the foundation pit settlement spatial reasoning method based on the improved spatio-temporal Kriging model provided by the embodiments is shown in the figure. Figure 1 The method specifically includes the following steps:

[0062] S1. Systematic data acquisition is performed, a multi-dimensional and staged database structure is established, dynamic monitoring data and static prior information are uniformly managed, complete and structured data support is provided for subsequent staged covariance modeling and spatio-temporal reasoning; the collected data includes monitoring data of foundation pit settlement, construction process data, geological and hydrological information; the monitoring data of foundation pit settlement includes settlement records of total station, settlement gauge and inclinometer at real-time monitoring points; the construction process data includes construction log, excavation depth, installation time and sequence of supporting structure, and starting and duration of dewatering operation; the geological and hydrological information includes soil layer distribution, soil mechanical parameters and underground water level change; the above data has time stamp and spatial coordinates, and can describe the integrity of spatial position and time evolution.

[0063] The process of establishing a multi-dimensional and staged database structure in the S1 step includes: first, indexing the monitoring data according to the time sequence and the position of the monitoring point to form a space-time matrix; second, encoding the construction events and disturbance factors as stage labels and corresponding to the time interval as control variables introduced into the model; and finally, binding the static information geological and hydrological parameters with the spatial coordinates of the monitoring point to form a comprehensive data set containing "position-time-attribute".

[0064] S2. Using the collected monitoring data, construction process data, geological and hydrological information, the construction period is segmented and processed, and a set of corresponding space-time covariance functions is established for each stage . In the S2 step, a disturbance-aware staged covariance modeling strategy is proposed, which identifies the construction stages and establishes the covariance function in segments to accurately control the time sequence change and disturbance response. As shown in the accompanying drawings, the specific process of establishing the corresponding space-time covariance function for each stage is as follows: Figure 2

[0065] S201. The specific process of segmenting the construction period using the collected monitoring data, construction process data, geological and hydrological information is as follows

[0066] Suppose the construction period is divided into disturbance stages, denoted as:

[0067]

[0068] Where each stage corresponds to a time interval , indicating that the disturbance properties are roughly consistent within the stage, such as "first support construction period" or "second layer soil excavation period". For any time point , define its stage label function as:

[0069]

[0070] The stage label will serve as a "control variable" for covariance function modeling, enabling the model to dynamically switch between different time periods;

[0071] S202. In the problem of foundation pit settlement prediction, the settlement value is not only affected by the spatial position of the monitoring point, but also evolves with the construction process, thus having significant space-time coupling characteristics. The traditional Kriging method is based on the assumption of stationarity, only considering the covariance between spatial positions, ignoring the effects of time factors and construction disturbances, resulting in poor prediction results when the settlement pattern changes dramatically or the geological conditions are complex. To overcome this problem, the embodiment establishes a set of corresponding space-time covariance functions for each stage ​, a corresponding spatiotemporal covariance function is established , which comprehensively considers four factors: spatial distance, time difference, geological condition, and construction disturbance, and realizes more refined modeling of the correlation of settlement values. The spatiotemporal covariance function is denoted as:

[0072] ①

[0073] where, denotes the covariance between any measuring point at any time and any measuring point at any time ; is the total variance of settlement data, reflecting the overall fluctuation level; is the spatial correlation function of any measuring point and any measuring point , considering spatial position and geological condition; is the time correlation function of any time and any time , considering time difference and construction disturbance; is the geological parameter vector, such as soil type, porosity, elastic modulus, etc.; is the disturbance state, for example, the current construction stage (supporting, dewatering, excavation, etc.);

[0074] The spatial correlation function adopts the Gaussian kernel function and adds a geological condition adjustment term, and the specific form is:

[0075] ⑦

[0076] where, denotes the Euclidean distance between points and ; is the spatial scale parameter, controlling the decay speed of spatial influence; is the geological weighting factor, used to adjust the influence of the difference in geological conditions on correlation; where, the expression of

[0077] ⑧

[0078] where, is the geological influence adjustment coefficient; and are the geological attribute vectors of points and ; the denominator is used to normalize the difference in geological parameters between different measuring points.

[0079] The time-dependent function adopts an exponential decay kernel combined with a construction disturbance sensitive mechanism, which is expressed as:

[0080] ⑨

[0081] where, denotes the time difference between two time points; is the time scale parameter, controlling the decay speed of time dependence; is the disturbance phase weighting term, judging whether different construction phases are crossed; the definition of disturbance weighting term is:

[0082] ⑩

[0083] where, is the disturbance impact weight; denotes the indicator function, if is different from the phase of , it is 1, otherwise 0.

[0084] Through this structure, the model can perceive the settlement mutation caused by "phase switching", which has obvious advantages in dealing with scenarios such as "excavation starts-settlement mutation". In general, the spatiotemporal covariance function significantly improves the modeling ability of the model for non-stationarity and spatiotemporal coupling by introducing the geological difference term and the disturbance state term . It is not only suitable for interpolation in static scenarios, but also suitable for settlement prediction in dynamic construction process, with good engineering adaptability and physical interpretability.

[0085] S3. To realize the continuous modeling of the model for the disturbance state of the actual time point , the spatiotemporal covariance functions of multiple phases are segmented and spliced to establish a spatiotemporal joint covariance structure with geological disturbance perception ability:

[0086] ②

[0087] where, denotes the covariance between any measurement point at any time and any measurement point at any time ; is the kth disturbance phase.

[0088] This mechanism ensures smooth transition of the covariance structure and does not appear "mutation" at the phase boundary, which helps to improve the continuity of interpolation and physical consistency.

[0089] S4. After establishing a spatiotemporal joint covariance structure with geological disturbance sensing capability, the joint covariance matrix and covariance vector are constructed using a predefined spatiotemporal covariance function; the covariance matrix between the monitoring points is defined as:

[0090] ⑪

[0091] The covariance vector between the point to be predicted and each monitoring point is defined as:

[0092] ⑫.

[0093] S5. The core objective of this invention is to achieve a specific location. and time Above, predict its settlement value Even if no monitoring instruments were ever installed at this point, nor was it used for training with historical data; therefore, this invention, based on classical Kriging theory, constructs an interpolation inference framework applicable to non-stationary spatiotemporal fields. Its core lies in minimizing the variance of the estimation error and solving for the optimal weighted combination under unbiased conditions, as shown in the appendix. Figure 3 As shown. Specifically, the prediction target is:

[0094] ③

[0095] in, Let be the settlement value corresponding to measuring point i at time i; n is the number of possible combinations of all measuring points at all times.

[0096] n = number of measurement points × number of time points;

[0097] To determine the interpolation weights, the following unbiased constraints must be satisfied:

[0098] ④

[0099] Based on classical Kriging theory, to minimize the variance of the estimation error, and to solve for the optimal weighted combination under the condition of unbiasedness, combined with the covariance matrix and covariance vector constructed in step S4, equation ③ can be rewritten as:

[0100] ⑤.

[0101] The meaning of the above formula is that the predicted value is obtained by weighted average of the present observation values, and the weight is determined by the covariance relationship between the points and the prediction point. The closer the spatial distance, the more similar the geological disturbance, and the closer the time, the greater the weight.

[0102] The method in the embodiment not only provides the predicted value, but also can simultaneously evaluate the confidence of the prediction, and give the variance estimation of the prediction error:

[0103] ⑥

[0104] wherein, is the covariance of the prediction point itself; the second term is the 'explainable part' of the prediction point to the existing observation point information, and the residual variance is obtained after subtraction; the smaller the value, the more reliable the prediction of the model to the point. In actual engineering, the prediction variance can be used as a risk assessment index to identify the 'prediction blind area' or 'high uncertainty area' to assist the construction personnel in focusing on monitoring.

[0105] Through the structure, the model can perceive the settlement mutation caused by'stage switching', and has obvious advantages in processing scenes similar to 'excavation starts - settlement mutation'.

[0106] Overall, the spatiotemporal covariance function in the application significantly improves the modeling ability of the model to non-stationarity and spatiotemporal coupling relationship by introducing the geological difference term and the disturbance state term It is not only suitable for interpolation in static scenes, but also suitable for settlement prediction in dynamic construction process, and has good engineering adaptability and physical interpretability.

[0107] The application can effectively deal with the settlement mutation caused by the construction stage and the geological section, and the interpolation result has higher prediction accuracy and stability in the area with strong geological heterogeneity and frequent disturbance, realizes unified modeling of different stages in the whole construction process, significantly improves the adaptability of the model to time evolution and physical field response, and the model is convenient for fusion with actual geological data and construction log; at the same time, the model prediction result supports error estimation and risk analysis, and the output result is visualized clearly, which facilitates the engineering personnel to quickly identify the settlement abnormal area, assists decision-making and safety management.

Claims

1. A spatial reasoning method for foundation pit settlement based on an improved spatiotemporal kriging model, characterized in that: Specifically, the following steps are included: S1. Conduct systematic data collection and establish a multi-dimensional, phased database structure to uniformly manage dynamic monitoring data and static prior information, providing complete and structured data support for subsequent phased covariance modeling and spatiotemporal inference; the collected data includes monitoring data of foundation pit settlement, construction process data, and geological and hydrological information; S2. Using the collected monitoring data, construction process data, and geological and hydrological information, the construction cycle is segmented, and a corresponding spatiotemporal covariance function is established for each stage. The spatiotemporal covariance function is specifically: ① in, Represents any measurement point within the k-th stage. At any time With any measuring point At any time Covariance between them; The total variance of the settlement data reflects the overall level of fluctuation. For any measuring point With any measuring point Spatial correlation functions, taking into account spatial location and geological conditions; For any time With any time Time-dependent functions, taking into account time differences and construction disturbances; A vector of geological parameters; It is in a disturbed state; S3. Segment and spatiotemporal covariance functions from multiple stages are piecewise concatenated to establish a spatiotemporal joint covariance structure with geological disturbance sensing capabilities: ② in, Indicates any measuring point during the entire construction period At any time With any measuring point At any time Covariance between them; This is the k-th perturbation stage; S4. After establishing a spatiotemporal joint covariance structure with geological disturbance sensing capability, the joint covariance matrix and covariance vector are constructed using the predefined spatiotemporal covariance function. S5. Define settlement spatial reasoning, that is, at a specific location and time Above, predict its settlement value The prediction formula is: ③ in, Let be the settlement value corresponding to measuring point i at time i; n is the number of possible combinations of all measuring points at all times. n = number of measurement points × number of time points; To determine the interpolation weights, the following unbiased constraints must be satisfied: ④ Based on classical Kriging theory, to minimize the variance of the estimation error, and to solve for the optimal weighted combination under the condition of unbiasedness, combined with the covariance matrix and covariance vector constructed in step S4, equation ③ can be rewritten as: ⑤； in, Let be the covariance vector between the point to be predicted and each monitoring point; Let be the covariance matrix between the observation points.

2. The spatial reasoning method for foundation pit settlement based on an improved spatiotemporal kriging model according to claim 1, characterized in that, In addition to providing predicted values, the method can also simultaneously assess the confidence level of the predictions and provide an estimate of the variance of the prediction error. ⑥ in, The first term is the covariance of the predicted point itself; the second term is the interpretable part of the predicted point based on the information of the existing observation points, which is subtracted to obtain the residual variance; the smaller this value is, the more reliable the model's prediction for that point is.

3. A spatial reasoning method for foundation pit settlement based on an improved spatiotemporal kriging model according to claim 1 or 2, characterized in that: In step S1, the settlement monitoring data of the foundation pit includes settlement records from total station, settlement gauge, and inclinometer at real-time monitoring points; the construction process data includes construction log, excavation depth, installation time and sequence of support structure, and initiation and duration of dewatering operations; the geological and hydrological information includes soil layer distribution, soil mechanical parameters, and groundwater level changes; the above data have timestamps and spatial coordinates, and can simultaneously describe the integrity of spatial location and temporal evolution.

4. A spatial reasoning method for foundation pit settlement based on an improved spatiotemporal kriging model according to claim 1 or 2, characterized in that, The process of establishing a multi-dimensional, phased database structure in step S1 includes: First, indexing the monitoring data according to the time series and the location of the monitoring points to form a spatial-temporal matrix; second, encoding construction events and disturbance factors as phase labels and corresponding them to the corresponding time intervals, and introducing them into the model as control variables; finally, binding static geological and hydrological parameters with the spatial coordinates of the monitoring points to form a comprehensive dataset containing "location-time-attribute".

5. A spatial reasoning method for foundation pit settlement based on an improved spatiotemporal kriging model according to claim 1 or 2, characterized in that, In step S2, the construction period is divided into segments, as detailed below: Assuming the construction period is divided into: Each disturbance stage is denoted as: ； Each stage Corresponding to a time interval This indicates that the nature of the disturbances is roughly the same during this stage; For any point in time The stage label function is defined as follows: 。 6. A spatial reasoning method for foundation pit settlement based on an improved spatiotemporal kriging model according to claim 1 or 2, characterized in that, In step S2, the spatial correlation function uses a Gaussian kernel function and includes an adjustment term for geological conditions, specifically in the following form: ⑦ in, Point and The Euclidean distance between them; This is a spatial scale parameter that controls the rate attenuation of spatial influences. It is a geological weighting factor used to adjust for the impact of differences in geological conditions on the correlation; among which, The expression is: ⑧ in, This is the geological influence adjustment coefficient; and Points , Geological attribute vector; denominator Used to normalize the differences in geological parameters between different measuring points.

7. A spatial reasoning method for foundation pit settlement based on an improved spatiotemporal kriging model according to claim 1 or 2, characterized in that, In step S2, the time-related function uses an exponentially decaying kernel function combined with a construction disturbance sensitivity mechanism, in the form of: ⑨ in, This represents the time difference between two points in time. It is a time scale parameter that controls the rate of decay of time-dependent properties; This is a disturbance stage weighting term used to determine whether different construction stages have been crossed; the definition of the disturbance weighting term is: ⑩ in, The weights are affected by the disturbance; Indicates an indicator function, if and If the stage is different, the value is 1; otherwise, the value is 0.

8. A spatial reasoning method for foundation pit settlement based on an improved spatiotemporal kriging model according to claim 1 or 2, characterized in that, The covariance matrix between monitoring points in step S4 is defined as follows: ⑪ The covariance vector between the point to be predicted and each monitoring point is defined as: ⑫。

Images