Construction land approval risk early warning method and system

Abstract

The invention relates to the technical field of risk early warning, in particular to a construction land approval risk early warning method and system. The method comprises the following steps: acquiring and analyzing geological structure data of a construction land to obtain geological profile hierarchical data; then, analyzing the viscosity state of the soil, simulating soil layer expansion building displacement, and performing micro-change expansion displacement accumulation treatment; then, the building expansion cracking risk is predicted according to the accumulated data; and finally, based on expansion cracking risk prediction, identifying risk prefactors of the construction land, and feeding back the risk prefactors to the terminal to realize construction land approval risk early warning. The risk examination and approval early warning technology is optimized, so that the risk examination and approval early warning technology is more perfect.

Description

Technical Field

[0001] The present invention relates to the field of risk warning technology, and in particular to a construction land approval risk warning method and system. Background Art

[0002] The geological risks faced during land development are becoming increasingly complex, especially due to the significant impact of geological hazards such as soil swelling on building safety. Soil swelling, a phenomenon common in arid or semi-arid regions, refers to the volumetric expansion of soil due to environmental factors such as changes in moisture and temperature. This phenomenon can cause cracks, deformation, and even collapse in buildings, posing significant risks to resident safety, economic losses, and social stability. Previous construction land approval processes often failed to fully assess the geological risks of the land, particularly the potential dangers of special geological conditions such as soil swelling, which can easily lead to serious expansion and cracking problems in approved buildings. To improve the scientific and accurate nature of approvals and reduce the occurrence of subsequent accidents, in-depth analysis of the geological structure and soil properties of the construction land is essential. However, traditional risk warning methods for construction land approval suffer from low accuracy in predicting the risk of building displacement and cracking caused by soil swelling, resulting in an inability to accurately assess the risk factors between natural geographical risks and construction conflicts. Summary of the Invention

[0003] Based on this, it is necessary to provide a construction land approval risk warning method and system to solve at least one of the above technical problems.

[0004] To achieve the above-mentioned purpose, a construction land approval risk early warning method is provided, the method comprising the following steps: Step S1: Acquire the geological structure data of the construction land uploaded by the client; perform geological profile layering analysis on the geological structure data of the construction land to obtain geological profile layering data of the construction land; Step S2: performing soil layer viscosity state analysis on the construction land geological profile layer data to obtain soil layer viscosity state data; performing soil layer expansion building displacement simulation and quantification based on the soil layer viscosity state data to obtain soil layer expansion building displacement quantification data; performing micro-variation expansion displacement accumulation processing on the soil layer expansion building displacement quantification data to obtain micro-variation expansion displacement accumulation data; Step S3: Predicting the risk of building expansion cracking based on the cumulative data of expansion displacement micro-changes to obtain building expansion cracking risk prediction data; Step S4: Based on the building expansion and cracking risk prediction data, the construction land risk pre-factor approval is identified to obtain the construction land risk pre-factor data; the construction land risk pre-factor data is fed back to the terminal to execute the construction land approval risk warning.

[0005] Preferably, step S1 includes the following steps: Step S11: Acquire the geological structure data of the construction land uploaded by the client; Step S12: performing data cleaning on the geological structure data of the construction land to obtain the geological structure cleaning data of the construction land; Step S13: Performing geological profile layering analysis on the cleaned geological structure data of the construction land to obtain geological profile layering data of the construction land.

[0006] Preferably, step S2 includes the following steps: Step S21: Obtain construction land development plan; Step S22: extracting the construction foundation excavation structure of the construction land development plan to obtain foundation excavation structure data; Step S23: performing soil layer viscosity state analysis on the construction land geological profile layer data to obtain soil layer viscosity state data; Step S24: performing soil expansion and building displacement simulation and quantification on the foundation excavation structure data according to the soil layer viscosity state data to obtain soil expansion and building displacement quantification data; Step S25: performing micro-change expansion displacement accumulation processing on the quantified data of soil expansion building displacement to obtain micro-change expansion displacement accumulation data.

[0007] Preferably, step S24 includes the following steps: Step S241: simulating the volume change of soil stratified water absorption and expansion per unit time on the soil stratified viscosity state data to obtain the volume change data of soil stratified water absorption and expansion; Step S242: analyzing the nonlinear exponential growth relationship of the expansion pressure based on the volume change data of the soil stratified water absorption expansion to obtain the exponential growth relationship of the soil expansion pressure; Step S243: performing vertical-lateral load pressure difference fluctuation calculation on the foundation excavation structure data according to the soil expansion pressure exponential growth relationship to generate the foundation vertical-lateral load pressure fluctuation difference; Step S244: deriving the foundation jacking anisotropic deflection angle from the foundation excavation structure data according to the foundation vertical-lateral load pressure fluctuation difference to obtain the foundation jacking direction deflection angle; Step S245: simulating and quantifying the displacement of the building due to soil expansion according to the deflection angle of the foundation support direction, and obtaining quantitative data of the displacement of the building due to soil expansion.

[0008] Preferably, step S243 includes the following steps: Perform geometric boundary discretization processing on the foundation excavation structure data to obtain the foundation geometric boundary grid data; perform vertical-lateral structural stiffness analysis on the foundation geometric boundary grid data to obtain the vertical-lateral structural stiffness; Performing exponential growth curve conversion on the exponential growth relationship of soil swelling pressure to obtain an exponential growth curve of soil swelling pressure; Based on the secant method, the slope variance between the growth inflection points of the soil expansion pressure exponential growth curve is calculated to obtain the inflection point growth slope variance; According to the inflection point growth slope variance, the growth iteration deviation geometric calculation is performed on the soil expansion pressure exponential growth curve to obtain the growth iteration deviation geometric data; Based on the growth iterative deviation geometric data, the vertical-lateral structural stiffness is calculated by the second-order differential fluctuation of the vertical-lateral pressure load to obtain the vertical-lateral pressure load differential fluctuation data; The vertical-lateral pressure load differential fluctuation data is subjected to vertical-lateral load pressure difference fluctuation calculation to generate the foundation vertical-lateral load pressure fluctuation difference.

[0009] Preferably, step S244 includes the following steps: The vertical longitudinal supporting force and horizontal transverse supporting force vector fields are analyzed for the vertical-lateral load pressure fluctuation difference of the foundation, and the vertical longitudinal supporting force vector field and horizontal transverse supporting force vector field are obtained respectively; The vertical longitudinal jacking force vector field is fitted with the longitudinal jacking force peak phase continuity to obtain the longitudinal jacking force peak phase continuity data; The horizontal lateral jacking force vector field is analyzed for the lateral jacking force angle component offset to obtain the lateral jacking force angle component offset; The vector field divergence-curl decomposition data are obtained by performing vector field divergence-curl decomposition processing based on the longitudinal top force peak phase continuous data and the lateral top force angle component offset; The anisotropic deflection angle of foundation jacking is derived from the foundation excavation structure data according to the vector field divergence-curl decomposition data, and the deflection angle of foundation jacking direction is obtained.

[0010] Preferably, step S25 includes the following steps: Step S251: Analyze the displacement jump values between the interfaces of each soil layer on the quantitative data of the soil expansion building displacement to obtain the displacement jump values between the interfaces of the soil layers; Step S252: performing an offset direction component analysis on the quantitative data of building displacement caused by soil expansion based on the displacement jump value between soil layer interfaces to obtain the offset direction component of the building displacement; Step S253: performing displacement acceleration geometric increment identification based on the displacement jump value between the soil layer interfaces and the building displacement direction component to obtain displacement acceleration geometric increment data; Step S254: performing a progressive periodic change integration process on the offset acceleration geometric increment data to obtain offset acceleration progressive integral data; Step S255: Performing slight-change expansion displacement accumulation processing according to the offset acceleration progressive integral data to obtain slight-change expansion displacement accumulation data.

[0011] Preferably, step S3 includes the following steps: Step S31: normalizing the cumulative data of the slight change in expansion displacement to obtain the normalized cumulative data of the slight change in expansion displacement; Step S32: performing convergence constraint training on the expansion displacement slight change accumulated normalized data based on the deep Q network to obtain expansion displacement slight change convergence constraint data; Step S33: Predicting the risk of building expansion cracking based on the expansion displacement micro-change convergence constraint data to obtain building expansion cracking risk prediction data.

[0012] Preferably, step S32 includes the following steps: Step S321: extract random samples from the cumulative normalized data of the slight change in expansion displacement to obtain a random sample of the slight change in expansion displacement; Step S322: performing multi-step learning processing on the cumulative random samples of the slight change in the expansion displacement to obtain sample multi-step learning convergence data; Step S333: Based on the deep Q network, convergence constraint training is performed on the sample multi-step learning convergence data to obtain expansion displacement slight change convergence constraint data.

[0013] Preferably, the present invention further provides a construction land approval risk warning system for executing the construction land approval risk warning method described above, the construction land approval risk warning system comprising: The geological profile layered analysis module is used to obtain the geological structure data of the construction land uploaded by the client; perform geological profile layered analysis on the geological structure data of the construction land to obtain geological profile layered data of the construction land; The expansion displacement accumulation module is used to analyze the soil layer viscosity state of the construction land geological profile layer data to obtain soil layer viscosity state data; simulate and quantify the soil layer expansion building displacement based on the soil layer viscosity state data to obtain soil layer expansion building displacement quantitative data; and perform micro-variation expansion displacement accumulation processing on the soil layer expansion building displacement quantitative data to obtain micro-variation expansion displacement accumulation data; A cracking risk prediction module is used to predict the risk of building expansion cracking based on the cumulative data of micro-changes in expansion displacement, and obtain building expansion cracking risk prediction data; The risk pre-factor identification module is used to identify the risk pre-factors of construction land based on the building expansion and cracking risk prediction data, and obtain the risk pre-factor data of construction land; the risk pre-factor data of construction land is fed back to the terminal to implement the risk warning of construction land approval.

[0014] The beneficial effect of the present invention is that, by acquiring and analyzing the geological structure data of construction land, the geological conditions of the land can be accurately understood, laying the foundation for subsequent risk assessment. Geological profile layer analysis can identify changes in soil layers, thereby better understanding the land's bearing capacity and stability. The implementation of this step can provide scientific data support for further analysis of the soil's viscosity and expansion characteristics, avoiding the bias of human judgment and improving the scientific nature and accuracy of the land development process. Soil layer viscosity analysis and soil expansion building displacement simulation quantification can provide a deep understanding of the physical changes of soil layers under different conditions, particularly the soil expansion phenomenon. These analytical data provide a quantitative basis for building stability and help predict the displacement impact of soil expansion on buildings. In addition, the cumulative processing of micro-variable expansion displacements further refines the prediction of expansion risks, allowing risk assessments to promptly identify potential hazards and take necessary preventive measures. Predicting the risk of building expansion cracking based on the cumulative micro-variable expansion displacement data can identify building cracking problems caused by soil expansion in advance. By simulating the response of buildings under different expansion displacements, it is possible to effectively predict which areas of buildings are most susceptible to expansive soils. This prediction can provide construction units with targeted protective measures to avoid structural safety issues in buildings during use, thereby ensuring the long-term stability of construction projects and the safety of residents. Identifying risk pre-factors for construction land based on building expansion and cracking risk prediction data can identify potential risks of construction land in advance, especially those related to soil expansion. The identification of these risk pre-factors not only provides valuable decision-making basis for construction units, but also provides more scientific and detailed risk assessment reports for approval authorities, ensuring the scientificity and accuracy of approval decisions. By timely feeding back risk pre-factor data to the terminal and executing risk warnings, it is possible to effectively reduce uncertainty in the construction process and reduce project delays and economic losses caused by unforeseen geological problems. Therefore, the present invention is an optimization of a traditional construction land approval risk warning method, solving the problem that the traditional construction land approval risk warning method has low accuracy in predicting the risk of building displacement and cracking caused by soil layer expansion, resulting in the inability to accurately approve the risk factors between natural geographical risks and construction conflicts. It improves the accuracy of predicting the risk of building displacement and cracking caused by soil layer expansion, and improves the accuracy of approval warnings for risk factors between natural geographical risks and construction conflicts. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A schematic diagram of the steps of a risk warning method for construction land approval; Figure 2 for Figure 1 Detailed implementation steps of step S2 in FIG. Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG.

[0016] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0017] See also Figures 1 to 3 A construction land approval risk warning method, the method comprising the following steps: Step S1: Acquire the geological structure data of the construction land uploaded by the client; perform geological profile layering analysis on the geological structure data of the construction land to obtain geological profile layering data of the construction land; Step S2: performing soil layer viscosity state analysis on the construction land geological profile layer data to obtain soil layer viscosity state data; performing soil layer expansion building displacement simulation and quantification based on the soil layer viscosity state data to obtain soil layer expansion building displacement quantification data; performing micro-variation expansion displacement accumulation processing on the soil layer expansion building displacement quantification data to obtain micro-variation expansion displacement accumulation data; Step S3: Predicting the risk of building expansion cracking based on the cumulative data of expansion displacement micro-changes to obtain building expansion cracking risk prediction data; Step S4: Based on the building expansion and cracking risk prediction data, the construction land risk pre-factor approval is identified to obtain the construction land risk pre-factor data; the construction land risk pre-factor data is fed back to the terminal to execute the construction land approval risk warning.

[0018] In the embodiment of the present invention, reference Figure 1 The above is a schematic flow chart of the steps of a construction land approval risk warning method of the present invention. In this example, the construction land approval risk warning method includes the following steps: Step S1: Acquire the geological structure data of the construction land uploaded by the client; perform geological profile layering analysis on the geological structure data of the construction land to obtain geological profile layering data of the construction land; In an embodiment of the present invention, the geological structure data of the construction land collected by the client is first uploaded to the API interface specified by the server through the POST method in the HTTP protocol. The server builds a data receiving module based on the Flask framework and adopts the JSON format for structured data transmission. The data fields include geological indicators such as layer number, layer burial depth, layer thickness, main material composition, porosity, water content, natural density, plasticity index, liquid limit, etc. The received data first passes through a data cleaning process implemented by the Pandas library. Sample rows with missing values are filled using the bidirectional linear interpolation method. For fields with outliers, outliers are eliminated based on the IQR (interquartile range) method, and the upper and lower thresholds are set to Q1-1.5IQR and Q3-1.5IQR, respectively. The cleaned data is input into the geological profile stratification analysis module, and the k-means clustering algorithm is used to automatically segment the stratigraphic data. The characteristic vectors of the geological profile are extracted according to the main components of the burial depth change and material composition change. The burial depth increment is used as the feature basis, and the initial number of cluster centers k is set to 5 for clustering. Each cluster represents a stratification result. After the clustering is completed, the stratification label generated is merged with the original data to obtain the stratification data of the geological profile of the construction land. The output fields include the stratification number, the upper and lower boundary depths of the stratification, the main component description and the key mechanical indicators, providing a standardized and hierarchical geological data foundation for the subsequent steps.

[0019] Step S2: performing soil layer viscosity state analysis on the construction land geological profile layer data to obtain soil layer viscosity state data; performing soil layer expansion building displacement simulation and quantification based on the soil layer viscosity state data to obtain soil layer expansion building displacement quantification data; performing micro-variation expansion displacement accumulation processing on the soil layer expansion building displacement quantification data to obtain micro-variation expansion displacement accumulation data; In this embodiment of the present invention, the construction land geological profile layer data output from step S1 is first imported into the viscosity state analysis module. Using the logistic regression method, the module differentiates and models the relationship between the plasticity index and liquid limit for different layers. By setting three classification criteria of 30%, 50%, and 70% of the liquid limit, the soil state is divided into low viscosity, medium viscosity, and high viscosity, and further subdivided based on the plasticity index range. Each layer of data is normalized according to the liquid limit and plasticity index and used as input for the logistic regression model. The corresponding soil layer viscosity state label is output and aggregated into soil layer viscosity state data. Then, based on the viscosity state of each layer, a single-layer unit model is constructed using the finite element discrete simulation method. The initial unit side length is set to 0.5 meters. Standard water absorption expansion stress conditions are applied. The vertical and lateral displacement responses of each layer under the standard water absorption expansion state are simulated. Displacement distributions are derived using the linear elastic expansion assumption. The stress transmission changes after the upper load of each layer are superimposed and analyzed to obtain preliminary quantitative data on building displacement due to soil layer expansion. Next, the quantified vertical and horizontal displacement results of each layer are input into the micro-change accumulation processing module. Using the incremental accumulation method, the cumulative displacement changes are transferred upward layer by layer, with each 0.1-meter burial depth as the differential unit, to obtain the cumulative data of the micro-change of expansion displacement. The output includes detailed parameters such as the micro-change displacement increment of each layer and the quantified value of the cumulative displacement.

[0020] Step S3: Predicting the risk of building expansion cracking based on the cumulative data of expansion displacement micro-changes to obtain building expansion cracking risk prediction data; In this embodiment of the present invention, the cumulative data of micro-changes in expansion displacement obtained in step S2 is normalized, with the normalization interval set to [0, 1]. Using the minimum-maximum normalization method, the cumulative micro-change displacements of each layer are linearly stretched with the maximum and minimum displacement amplitudes of the entire field as the bounds. The normalized data serves as the input samples. Next, a building expansion cracking risk prediction model is constructed based on the Deep Q-Network (DQN) in deep reinforcement learning. The input dimension is set to the length of the layered displacement cumulative sequence. The hidden layer uses a three-layer fully connected neural network with 128, 64, and 32 nodes per layer. The activation function is the Reluctant Unit (ReLU) function. The output layer is configured to predict the cracking risk level, which is divided into four levels: no risk, slight risk, moderate risk, and high risk. During training, an experience replay mechanism is used, with a batch size of 32 and a target network update frequency of every 500 steps. An ε-greedy strategy is used for action selection, with an initial ε value of 1 and linear decay to 0.1. The training data used randomly sampled, slightly modified, cumulatively normalized data samples, and was subjected to multi-step learning (N-step learning, N=3). Each step was updated with the Bellman equation target value. Through convergence-constrained training, the final output was building expansion cracking risk prediction data for each construction land sample. The data included risk level, risk confidence, and key expansion displacement triggering layers.

[0021] Step S4: Based on the building expansion and cracking risk prediction data, the construction land risk pre-factor approval is identified to obtain the construction land risk pre-factor data; the construction land risk pre-factor data is fed back to the terminal to execute the construction land approval risk warning.

[0022] In this embodiment of the present invention, based on the building expansion cracking risk prediction data output in step S3, a risk pre-factor identification rule base is first established. This rule base is based on engineering geological standards and empirical statistics of cracking cases. Rules include hard indicators such as cumulative expansion displacement greater than 3 cm, soil liquid limit greater than 50%, foundation burial depth less than 2 meters, and adjacent groundwater level less than 1 meter. The building expansion cracking risk prediction data is then matched against the rule base in sequence, using a rule decision tree approach for item-by-item comparison. If any of the rule conditions are met, a construction land risk pre-factor is determined to exist. Each successfully matched risk item is labeled according to its risk level, and the corresponding trigger parameters and layer information are recorded. Ultimately, construction land risk pre-factor data is generated. The data format includes risk factor type, trigger parameter value, triggering soil layer number, and potential consequence description. After identification, the construction land risk pre-factor data is pushed to a terminal device or approval management platform via the MQTT protocol asynchronous push mechanism. Upon receiving the data, the terminal platform automatically triggers the approval process risk warning module based on built-in rules, providing real-time notifications and recording of relevant risk warning information.

[0023] In another embodiment, based on an in-depth analysis of the building expansion and cracking risk prediction data, the system first creates a risk factor identification table, including 20 antecedent factors such as soil expansibility, groundwater level changes, uneven load distribution, and improper foundation treatment. The system traces the factors of high-risk points in the prediction data (risk probability > 70%), and determines the contribution of each factor through correlation analysis. The contribution calculation adopts the multivariate linear regression method, with the risk probability as the dependent variable and the antecedent factors as the independent variables. Factors with an absolute value of the regression coefficient greater than 0.3 are identified as major risk factors; the system quantitatively scores the major risk factors, with a scoring range of 0-100 points. The higher the score, the greater the risk. The scoring takes into account the severity of the factor (weight 0.6) and the scope of influence (weight 0.4). The system converts the scoring results into risk levels: low risk (0-30 points), medium risk (31-70 points), high risk (31-70 points), and low risk (0-30 points). (71-100 points); the system generates a heat map of risk factor distribution, using red to represent high-risk areas, yellow to represent medium-risk areas, and green to represent low-risk areas. The heat map resolution is set to 5 meters × 5 meters; the system generates a construction land risk pre-factor data file based on the risk factor analysis results. The file format is structured JSON, which contains information such as a list of risk factors, risk level scores, geographic location coordinates, and recommended treatment measures; the system pushes risk pre-factor data to terminal devices through secure encrypted channels. Terminal devices include approval personnel workstations, mobile approval terminals, and risk monitoring centers. Push is done in real-time and scheduled ways. Real-time push is sent immediately after the risk assessment is completed, and scheduled push sends a daily risk report at a fixed time every day (9 am); after receiving the risk data, the terminal automatically triggers the risk warning mechanism, and uses sound and light alarms to remind approval personnel to pay attention to high-risk items.

[0024] Step S1 includes the following steps: Step S11: Acquire the geological structure data of the construction land uploaded by the client; Step S12: performing data cleaning on the geological structure data of the construction land to obtain the geological structure cleaning data of the construction land; Step S13: Performing geological profile layering analysis on the cleaned geological structure data of the construction land to obtain geological profile layering data of the construction land.

[0025] In an embodiment of the present invention, a dedicated geological data acquisition system installed on a client is used to upload geological structural data of construction land via the HTTPS security protocol set within the local area network. The server is configured with Apache HTTP Server version 2.4 as a receiving platform and uses the POST method to receive data. The uploaded data is formatted as a standardized JSON data packet. The fields within the data packet include specific parameters such as geological point number, sampling time, stratum name, burial depth starting point, burial depth ending point, lithology description, natural water content (%), porosity, dry density (g / cm³), liquid limit (%), and plasticity index. The client acquisition terminal uses a deep-hole geological drilling method with a set frequency of acquisition every 10 meters. The maximum sampling depth is set to 50 meters. The single data upload volume does not exceed 10MB. If it exceeds 10MB, it is uploaded in batches. The SHA-256 digest algorithm is used to verify data integrity during the upload process. Each data packet is accompanied by a digest field. After receiving the geological data, the server performs preliminary data parsing using the requests library in Python 3.8. The data is then written to a MongoDB 5.0 database in JSON format. All geological fields in the database table structure definition are set to non-null to ensure complete data storage. The entire data upload process is handled by an Nginx reverse proxy and load balancing server. A single server supports up to 200 concurrent upload requests per second, ensuring efficient reception of large amounts of construction land geological data. After receiving the construction land geological data uploaded by the client, the server first uses the Pandas 1.3.5 library to clean the raw geological data in the database. This cleaning process is performed in a step-by-step manner. In the first step, a logical consistency check is performed on the depth start and end fields to ensure that the depth start is less than the depth end in each record. Any anomalies are detected and the record is discarded. In the second step, outliers were detected using the boxplot quartile method for the five core numerical fields: natural water content, porosity, dry density, liquid limit, and plasticity index. The lower limit was set at the first quartile minus 1.5 times the interquartile range, and the upper limit was set at the third quartile plus 1.5 times the interquartile range. Records outside these limits were marked as outliers. Outliers were handled using the local mean imputation method, which replaces the outlier record with the mean of the normal records in the stratum where the outlier was located. In the third step, null value detection was performed on the formation name and lithology description fields. If a gap existed, it was directly filled with the formation name and lithology description of adjacent sampling points within the same depth interval. Records with no adjacent samples were discarded. In the fourth step, all numerical fields were normalized to the range [0, 1]. The normalization method used minimum value minimization: each value was subtracted from the minimum value of the field and then divided by the difference between the maximum and minimum values.After cleaning, the processed data is rewritten to MongoDB, and a cleaning completion flag is added to the table record. A flag value of 1 indicates that the data has been cleaned, resulting in cleaned geological structure data for construction land. This provides a standardized, complete, and anomaly-free data foundation for subsequent steps. After cleaning the geological structure data for construction land, the cleaned data is analyzed by geological profile stratification using a K-means clustering method initialized with K-means++. First, the main feature fields in the cleaned data, including the starting depth, ending depth, liquid limit, plasticity index, natural moisture content, and dry density, are extracted as feature vectors to establish a standardized feature matrix. The initial layer number k is set to 5. Based on the recommendations of geological engineering standards, the drilling depth range is evenly divided into five layers for preliminary hypothesis generation. A K-means++ initialization strategy is used to ensure a more uniform initial centroid distribution. When performing cluster training, the maximum number of iterations is set to 300, and the convergence threshold is a centroid change of less than 0.001. After clustering, each data sample is assigned to a corresponding cluster label, with each label representing a geological layer. The data within the same label is then sorted from smallest to largest based on the starting depth, and the starting depth, ending depth, and main lithology type of each layer are calculated. If the standard deviation of the liquid limit within a cluster exceeds the set threshold of 5%, a subdivision is performed, and the data within the cluster is re-clustered using a secondary k-means clustering algorithm to split new layers and ensure the consistency of geological characteristics within the layers. The resulting stratified geological profile data for the construction land includes the layer number, starting and ending depths of the layer, the main lithology, and the average value of the main physical indicators. The mean silhouette coefficient in the clustering results is recorded, and a silhouette coefficient greater than 0.6 is used as a verification standard for the rationality of the stratification. The stratified data is finally output as a structured table, saved to the MongoDB database, marked as "parsed complete," and used as the input data source for subsequent soil viscosity state analysis.

[0026] Step S2 includes the following steps: Step S21: Obtain construction land development plan; Step S22: extracting the construction foundation excavation structure of the construction land development plan to obtain foundation excavation structure data; Step S23: performing soil layer viscosity state analysis on the construction land geological profile layer data to obtain soil layer viscosity state data; Step S24: performing soil expansion and building displacement simulation and quantification on the foundation excavation structure data according to the soil layer viscosity state data to obtain soil expansion and building displacement quantification data; Step S25: performing micro-change expansion displacement accumulation processing on the quantified data of soil expansion building displacement to obtain micro-change expansion displacement accumulation data.

[0027] As an example of the present invention, refer to Figure 2As shown, in this example, step S2 includes: Step S21: Obtain construction land development plan; In an embodiment of the present invention, the development planning data of construction land is obtained through an interface call method, and the development planning data is uniformly stored in parallel in CAD format and standard XML format. First, the file is received, and the CAD format development planning drawings are parsed using the DWG direct parsing library to extract layer information, wherein the building foundation outline layer is marked as "foundation", the road planning layer is marked as "road", the greening planning layer is marked as "green space", and the basement structure is marked as "basement". The coordinates of the parsed layer content are normalized, and all coordinates are uniformly converted into a local Cartesian coordinate system with the center point of the plot as the origin. Subsequently, for the XML format planning text data, the DOM tree traversal method is used to parse the nodes one by one to extract key attribute values, including the plot number, the above-ground building area, the underground development area, the excavation depth, and the excavation boundary coordinate point sequence. For example, if a plot is numbered D101, with a planned above-ground building area of 3,500 square meters, two underground floors, an excavation depth of 8 meters, and boundary point coordinates of [(0,0), (50,0), (50,30), (0,30)], then the corresponding attributes are fully recorded. All extracted data is stored using standardized fields, including: plot number, building foundation coordinates, excavation depth, number of basement floors, and road location coordinates. This ultimately achieves standardized acquisition of construction land development planning data, providing foundational data for subsequent excavation structure extraction.

[0028] Step S22: extracting the construction foundation excavation structure of the construction land development plan to obtain foundation excavation structure data; In this embodiment of the present invention, construction foundation excavation structure extraction is performed based on standardized construction land development planning data. First, a convex hull algorithm (Graham Scan method) is used to generate a minimum enclosed region based on the building foundation outline coordinate point sequence to ensure that the extracted region is connected and free of intersections. Next, the foundation excavation profile of each building unit is calculated based on the number of basement floors and excavation depth data. During this process, if the number of basement floors is greater than one, the excavation depth is accumulated at a standard of 4 meters per floor, i.e., 8 meters for the second underground floor and 12 meters for the third underground floor. If the excavation depth is already specified in the planning data, it is directly used. Then, for each building foundation area, a 3D excavation profile is constructed along the excavation depth direction (i.e., the negative Z-axis direction), generating foundation excavation structure data including the starting ground surface elevation, the ending excavation elevation, and the plane coordinates of the excavation boundary. For example, consider a building foundation with coordinates of [(0,0), (40,0), (40,20), (0,20)]. Assuming the ground surface elevation is 5 meters and the excavation depth is 8 meters, the ending excavation elevation is 5 minus 8, which equals negative 3 meters. The excavation volume is calculated by multiplying the contour area by the excavation depth. All foundation excavation data is output in a standard format, with fields including: plot number, excavation contour coordinate array, excavation start and end elevations, and excavation volume, for subsequent analysis and overlay with geological profiles.

[0029] Step S23: performing soil layer viscosity state analysis on the construction land geological profile layer data to obtain soil layer viscosity state data; In this embodiment of the present invention, based on geological profile data obtained from construction land layers, soil cohesion analysis is performed on each layer using a static rule-based classification method. First, for each geological layer, three indicators are extracted: liquid limit (%), plasticity index, and natural moisture content (%). These data are derived from the original test results in the geological survey report. According to the "Code for Design of Building Foundations" (GB 50007-2011), soils with a liquid limit greater than 50 and a plasticity index greater than 25 are defined as highly plastic soils; soils with a liquid limit between 30 and 50 and a plasticity index between 10 and 25 are defined as moderately plastic soils; and soils with a liquid limit less than 30 and a plasticity index less than 10 are defined as low plastic soils. Natural moisture content exceeding 90% of the liquid limit is marked as a special state of high fluidity. In practice, the liquid limit, plasticity index, and natural moisture content data for each layer are read one by one, matched and judged according to the aforementioned classification criteria, and ultimately each layer is assigned a cohesion state label. For example, if a layer has a liquid limit of 52%, a plasticity index of 28%, and a natural moisture content of 47%, the cohesive state is classified as highly plastic. After determining the cohesive state for all layers, a new data table is created. The table structure includes: layer number, start and end depths, liquid limit, plasticity index, natural moisture content, and cohesive state classification. All soil layer cohesive state data is archived and stored, providing quantitative input for subsequent soil expansion simulation and risk prediction.

[0030] Step S24: performing soil expansion building displacement simulation and quantification on the foundation excavation structure data according to the soil layer viscosity state data to obtain soil expansion building displacement quantification data; In this embodiment of the present invention, soil expansion and building displacement simulation and quantification are performed on foundation excavation structure data based on extracted soil layer viscosity data. The specific method is as follows: First, the excavation structure's excavation profile is spatially superimposed with geological profile data. Based on the spatial coordinate system, each depth segment is matched to the corresponding soil layer below the excavation profile. Each matching unit is a soil layer segment at a specific depth within the coverage of the building excavation profile. Subsequently, an expansion deformation rate parameter is set based on the viscosity classification results of each soil layer. The specific setting standards are: 2% for high-plasticity soil, 1% for medium-plasticity soil, and 0.5% for low-plasticity soil. For each unit, displacement simulation is performed according to the following logic: Using each grid cell of the excavation profile (with a grid size of 1 meter x 1 meter) as the basic calculation unit, information on all soil layers within the excavation depth is extracted, and the vertical cumulative expansion displacement of each layer is calculated. The specific calculation formula is: the expansion displacement of each layer is equal to the thickness of that layer multiplied by the corresponding expansion deformation rate. The expansion displacement values of each layer are then accumulated to obtain the total expansion displacement value for the corresponding grid cell. For example, if a grid cell is located 2 meters below high-plasticity soil, 3 meters below medium-plasticity soil, and 3 meters below low-plasticity soil, the expansion displacements are 2 × 2%, 3 × 1%, and 3 × 0.5%, respectively. The cumulative result is 0.04 m + 0.03 m + 0.015 m, for a total displacement of 0.085 m. After all grid cells have completed the displacement calculation, a complete excavation profile expansion displacement quantification dataset is formed. The data structure includes the grid number, location coordinates, and corresponding expansion displacement value. The final output is a standard raster data format or a structured table data format for subsequent micro-change accumulation analysis.

[0031] Step S25: performing micro-change expansion displacement accumulation processing on the quantified data of soil expansion building displacement to obtain micro-change expansion displacement accumulation data.

[0032] In an embodiment of the present invention, based on the quantitative data of soil expansion building displacement obtained, micro-variable expansion displacement accumulation processing is performed. The specific processing flow is as follows: First, all expansion displacement grid units in the same excavation area are sorted by spatial position, and processed in sequence from the center to the edge of the foundation pit, with a radial step length of 1 meter. Subsequently, on each radial path, the method of adjacent differential accumulation is adopted, that is, every time a step length of 1 meter is advanced, the difference in the expansion displacement value between the position and the previous position is recorded. If the difference value is greater than 0.002 meters, it is regarded as a local micro-variable occurrence and accumulated into the micro-variable displacement. The basic logic of differential accumulation is: the expansion displacement value of the current position is subtracted from the expansion displacement value of the previous position. If the result is positive and exceeds 0.002 meters, the difference is added to the current accumulated micro-variable; otherwise, the accumulated value remains unchanged. For example, if the displacement value sequence on a radial path is 0.080m, 0.082m, 0.084m, and 0.083m, then the first step difference is 0.002m, with a cumulative value of 0m. The second step difference is 0.002m, with a cumulative value of 0m. The third step difference is -0.001m, with a cumulative value of 0m, and the total cumulative micro-change is 0m. If the displacement value sequence on a path is 0.070m, 0.074m, 0.078m, and 0.083m, then the successive differences are 0.004m, 0.004m, and 0.005m, respectively, with cumulative increases of 0.004m, 0.004m, and 0.005m, respectively, for a total cumulative micro-change of 0.013m. By statistically analyzing the micro-variables of all radial paths, a cumulative micro-change diagram of the overall expansion of the excavation area can be generated. The data structure includes radial number, cumulative micro-change displacement, and radial start and end coordinates. All cumulative data on micro-changes in expansion displacement are archived and output according to unified standards, serving as an important source of input data for subsequent expansion risk prediction.

[0033] Step S24 includes the following steps: Step S241: simulating the volume change of soil stratified water absorption and expansion per unit time on the soil stratified viscosity state data to obtain the volume change data of soil stratified water absorption and expansion; Step S242: analyzing the nonlinear exponential growth relationship of the expansion pressure based on the volume change data of the soil stratified water absorption expansion to obtain the exponential growth relationship of the soil expansion pressure; Step S243: performing vertical-lateral load pressure difference fluctuation calculation on the foundation excavation structure data according to the soil expansion pressure exponential growth relationship to generate the foundation vertical-lateral load pressure fluctuation difference; Step S244: deriving the foundation jacking anisotropic deflection angle from the foundation excavation structure data according to the foundation vertical-lateral load pressure fluctuation difference to obtain the foundation jacking direction deflection angle; Step S245: simulating and quantifying the displacement of the building due to soil expansion according to the deflection angle of the foundation support direction, and obtaining quantitative data of the displacement of the building due to soil expansion.

[0034] In one embodiment of the present invention, soil layered viscosity data is used to simulate the volume change due to water absorption and expansion per unit time. The specific operation method is as follows: First, based on the soil layered viscosity data, the basic physical parameters of each soil layer are extracted, including natural water content, liquid limit, plastic limit, specific gravity, and initial porosity. For each soil layer, the volume of the water absorption and expansion experimental unit is set to 100 cubic centimeters, the initial water content is used as the starting point, and the water increase per unit time is set to 5% per hour. According to experimental standards, a linear relationship is established between the change in porosity and volume change caused by water absorption per unit time. The specific operation is to use the hourly increase in water content to infer the expansion of the pore volume between soil particles, and then to infer the change in the total sample volume. Taking a set of high-plasticity clay data as an example, the initial porosity is 0.45. After absorbing 5% water, the porosity increases to 0.4725, corresponding to a volume expansion ratio of 4.94%. According to this method, combined with the water absorption time series, a water absorption and expansion volume change curve is established for each soil layer. The unit time water absorption expansion data of all soil layers are stored in segments with a time step of 1 hour. The data structure includes layer number, time node, and volume change rate, and finally forms a soil stratification water absorption expansion volume change data set, and the output format is a standard structured table. Based on the soil stratification water absorption expansion volume change data obtained in step S241, the nonlinear exponential growth relationship of expansion pressure is analyzed. The specific operation method is as follows: first, the volume expansion rate data at different time nodes of each soil layer are sorted in chronological order, and the corresponding expansion stress growth sample is established. The expansion stress growth sample is established based on the following physical assumptions: the stress increment caused by unit volume soil expansion is nonlinearly positively correlated with the volume expansion rate, and the growth rate accelerates exponentially with time. According to the sorted volume change rate sequence, the stepwise fitting method is used to preliminarily fit the pressure increment of each time node, and then the exponential growth relationship is extracted by the logarithmic linear regression method. For example, if the soil expansion pressure increment in a particular layer is 2 kPa after 1 hour, 4.5 kPa after 2 hours, and 9.8 kPa after 3 hours, the exponential growth trend obtained by fitting indicates an exponential growth of 1.5 over time and a base growth coefficient of 2 kPa. All soil layers are processed in the same manner, ultimately forming a dataset of soil expansion pressure exponential growth relationships. The data structure includes layer number, exponential growth coefficient, base growth coefficient, and fitting error. All data is stored in a structured table format for subsequent load and pressure fluctuation calculations. Based on the soil expansion pressure exponential growth relationship extracted in step S242, vertical-lateral load pressure difference fluctuation calculations are performed on the foundation excavation structure data. The specific operation method is as follows: First, a three-dimensional load profile grid is established for the foundation excavation area. The grid size is set to 1 meter × 1 meter × 1 meter. Each grid cell corresponds to an initial static load state. The initial static load includes vertical load (self-weight) and horizontal lateral load (soil pressure).The exponential growth relationship for the swelling pressure of each soil layer was then applied to the corresponding grid cell, with swelling pressure increments superimposed vertically and laterally. The time step was set to 1 hour, and the calculation was cumulative for 8 hours. Within each time step, the fluctuation difference calculation was performed according to the following logic: first, in the vertical direction, the difference between the swelling pressure increment and the initial vertical load was calculated; in the lateral direction, the difference between the swelling pressure increment and the initial horizontal load was calculated. Subsequently, the lateral difference was subtracted from the vertical difference to obtain the load-pressure fluctuation difference. For example, if the initial vertical load of a cell is 150 kPa and the initial horizontal load is 50 kPa, and the cumulative vertical swelling pressure increment over 8 hours is 30 kPa and the cumulative horizontal swelling pressure increment is 10 kPa, then the vertical load becomes 180 kPa and the horizontal load becomes 60 kPa. The final load-pressure fluctuation difference is (180 - 150) - (60 - 50) = 30 kPa - 10 kPa = 20 kPa. Fluctuation difference calculations are performed on all grid cells separately, and a complete foundation vertical-lateral load pressure fluctuation difference data set is finally generated. The data structure includes grid number, time node, vertical load change, horizontal load change, and load pressure fluctuation difference. The data is output as a standard structured table for subsequent displacement analysis.

[0035] The anisotropic deflection angle of the foundation support is derived from the foundation excavation structure data based on the vertical-lateral load pressure fluctuation difference of the foundation. The specific operation method is as follows: First, the foundation vertical-lateral load pressure fluctuation difference data obtained in step S243 is associated one-to-one with the corresponding relationship of the foundation excavation structure grid. The vertical and lateral load changes of each grid cell are extracted according to its location and the size of the load fluctuation difference. According to the moment equilibrium principle in soil mechanics, a local coordinate system is set for each cell, with the X axis as the lateral horizontal axis and the Z axis as the vertical axis. The direction of the local support force is derived based on the load change as the basic quantity. The specific deduction logic is that the vertical load change value and the lateral load change value form a load vector, and the angle between the load vector and the vertical axis is defined as the support deflection angle. A threshold is set. If the lateral load change is less than 5% of the vertical load change, the deflection angle is considered to be zero, that is, the support direction of the cell is considered to be basically vertical and upward. If the lateral load change exceeds 5% of the vertical load change, the inverse tangent function is used to derive the angle value. That is, the lateral load change is used as the opposite side and the vertical load change as the adjacent side, and the inverse tangent value is calculated to obtain the angle size. Taking a specific unit as an example, the vertical load change is 30kPa, the lateral load change is 6kPa, and the lateral change accounts for 20%. According to the derivation logic, the inverse tangent is calculated, and the jacking deflection angle is approximately 11.31 degrees. The jacking deflection angle is derived for all grid units separately. The derivation results are recorded with parameters such as layer number, grid number, and deflection angle (unit: degrees). Finally, a foundation jacking direction deflection angle dataset is generated and output as a standard structured table. Based on the foundation jacking direction deflection angle obtained in step S244, the soil expansion building displacement simulation and quantification are performed. The specific operation method is as follows: First, based on the three-dimensional grid of the foundation excavation structure, for each grid unit, a micro-unit displacement model is established according to its corresponding jacking deflection angle and expansion pressure increment. The specific simulation logic is to regard the expansion pressure increment as a local external force and the deflection angle as the direction of action. It is assumed that the unit volume of soil produces a small displacement under the action of the external force in each time step. The displacement direction is consistent with the deflection angle direction, and the displacement size is proportional to the expansion pressure increment. The unit displacement is calibrated using an empirical constant, and the unit volume soil displacement corresponding to each increase of 1kPa in expansion pressure is set to 0.002 mm. Taking a unit as an example, if the expansion pressure increment is 20kPa and the top support deflection angle is 15 degrees, the total displacement is 20 times 0.002 mm, which is 0.04 mm. The displacement of 0.04 mm is then decomposed into vertical and horizontal components according to the deflection angle. The vertical component is 0.04 times cosine 15 degrees, and the horizontal component is 0.04 times sine 15 degrees, corresponding to the actual displacement in the vertical and horizontal directions respectively. All grid cells are used to simulate the displacement of soil expansion buildings according to this method. During the simulation process, the time step superposition method is used to accumulate the displacement changes every hour until the specified simulation time (for example, 8 hours).Finally, the simulation results of all units are summarized to form the quantitative data of soil expansion building displacement. The data structure includes grid number, time node, vertical displacement, horizontal displacement, composite displacement, and deflection angle, and is uniformly output in a structured table format for subsequent risk assessment processing.

[0036] Step S243 includes the following steps: Perform geometric boundary discretization processing on the foundation excavation structure data to obtain the foundation geometric boundary grid data; perform vertical-lateral structural stiffness analysis on the foundation geometric boundary grid data to obtain the vertical-lateral structural stiffness; Performing exponential growth curve conversion on the exponential growth relationship of soil swelling pressure to obtain an exponential growth curve of soil swelling pressure; Based on the secant method, the slope variance between the growth inflection points of the soil expansion pressure exponential growth curve is calculated to obtain the inflection point growth slope variance; According to the inflection point growth slope variance, the growth iteration deviation geometric calculation is performed on the soil expansion pressure exponential growth curve to obtain the growth iteration deviation geometric data; Based on the growth iterative deviation geometric data, the vertical-lateral structural stiffness is calculated by the second-order differential fluctuation of the vertical-lateral pressure load to obtain the vertical-lateral pressure load differential fluctuation data; The vertical-lateral pressure load differential fluctuation data is subjected to vertical-lateral load pressure difference fluctuation calculation to generate the foundation vertical-lateral load pressure fluctuation difference.

[0037] In one embodiment of the present invention, geometric boundary discretization is performed on foundation excavation structure data. The specific operation method is as follows: First, the two-dimensional plane boundary vector data of the foundation excavation area is read. The data format is a closed polygon composed of multiple continuous straight lines and curves. A uniform boundary sampling method based on Laplace feature preservation is used to sample nodes on the boundary curve at equal arc length intervals. The sampling interval is set to 0.5 meters, and each node is marked as a boundary control point. Subsequently, the control points are meshed using the Delaunay triangulation algorithm to generate a continuous, non-overlapping triangular mesh. The maximum side length of the triangular mesh must not exceed 1 meter. For boundary segments with sharp local curvature changes (curvature radius less than 2 meters), the node density is increased and the sampling interval is shortened to 0.25 meters to ensure that the mesh accurately approximates the actual geometric shape. After the mesh is triangulated, the topological consistency of the mesh is checked to remove hanging nodes and unusually sharp triangles (internal angles less than 20 degrees). The final foundation geometric boundary mesh data is generated, including node coordinates, cell numbers, and cell adjacency. A vertical-lateral structural stiffness analysis was performed on the foundation geometric boundary mesh data. The specific operation method is as follows: Based on the foundation mesh data, each triangular element is regarded as a local rigid element, and the connection stiffness between each element node is defined. The vertical stiffness and lateral stiffness are independently assigned. The vertical stiffness is set according to the elastic modulus of the soil. The elastic modulus of typical clay soil is set to 2MPa. The element height is taken as 0.5 meters. According to the vertical linear elastic mechanics relationship, the vertical stiffness is equal to the elastic modulus multiplied by the element thickness. The lateral stiffness takes into account the shear modulus and sets the Poisson's ratio to 0.3. The shear modulus is equal to the elastic modulus divided by twice the bracket plus the Poisson's ratio. The shear modulus value is taken as approximately 0.769MPa. The shear area is used to approximate the lateral load area, and the element lateral stiffness is calculated. The vertical and lateral stiffness parameters of all elements are calculated one by one, and the overall stiffness matrix of the mesh is constructed. The matrix elements represent the vertical and lateral stiffness values between node pairs, respectively, forming a complete vertical-lateral structural stiffness data table. The exponential growth relationship of soil swelling pressure was converted to an exponential growth curve. The specific operation method is as follows: First, read the discrete data points of soil swelling pressure growth over time, with a total of no fewer than 20 data points, and the data format is time-swelling pressure pairs. Using the standard natural logarithm transformation method, the swelling pressure values are logarithmically transformed and a scatter plot is plotted with time as the horizontal axis and the logarithmic swelling pressure as the vertical axis. Then, a univariate linear regression is performed on the transformed data using the least squares fitting algorithm to extract the slope and intercept of the fitted line. The original pressure growth form is restored by inverse logarithmic transformation to generate the soil swelling pressure exponential growth curve. The curve is expressed as an exponential function with time as the independent variable and swelling pressure as the dependent variable. The curve is interpolated with an hourly step size to ensure continuity, and finally a standardized soil swelling pressure exponential growth curve data is generated.

[0038] The slope variance between growth inflection points of the soil expansive pressure exponential growth curve is calculated using the secant method. The specific method is as follows: First, the soil expansive pressure exponential growth curve is divided into pairs of adjacent data points with a time step of one hour. Each pair of adjacent points is connected by a straight line, and the secant slope between the two points is calculated. The secant slope is defined as the swelling pressure value of the subsequent point minus the swelling pressure value of the previous point, divided by the time of the subsequent point minus the time of the previous point. After obtaining the secant slope sequence for the entire curve, inflection points are detected based on the increase or decrease trend. The location where the rate of change of the secant slope exceeds 10% is defined as the growth inflection point, and the inflection point index is recorded. For all secant slopes within the interval between adjacent inflection points, the slope variance of that interval is calculated. The slope variance is defined as the sum of the squares of the differences between each secant slope value and the mean of the secant slopes in the interval divided by the number of secants. For example, if a specific inflection point interval contains six secant lines with slope values of 0.3, 0.32, 0.35, 0.31, 0.29, and 0.34, the mean is calculated and the square of the difference between each secant slope and the mean is calculated. The sum is then divided by 6 to obtain the slope variance for that interval. All inflection point growth slope variance values are compiled into a list and output along with the corresponding inflection point interval number to form an inflection point growth slope variance dataset for subsequent calculations of foundation load fluctuation characteristics. Based on the inflection point growth slope variance, the growth iteration deviation of the soil expansion pressure exponential growth curve is proportionally calculated. The specific operation method is as follows: First, based on the extracted inflection point growth slope variance data, the soil expansion pressure exponential growth curve is divided into several inflection point intervals, each corresponding to a slope variance value. For each inflection point interval, a baseline deviation growth factor is set to 1.0. A geometric weighting correction is applied based on the slope variance, defining that a larger slope variance corresponds to a higher deviation growth rate. Specifically, a standard deviation threshold is set at 0.01. When the interval slope variance exceeds this threshold, the square root of the ratio of the variance to the threshold is used as the geometric growth factor. For example, for an inflection point interval with a slope variance of 0.04, the growth factor is 2.0. Using the growth factor as the base, the deviation at each time point is recursively calculated, with the time step as the unit time. The deviation increases geometrically and is added to the original value of the expansion pressure curve to form the pressure curve after the growth iterative deviation. Finally, for each time point in the entire time series, the corresponding growth iterative deviation value is output, forming the growth iterative deviation geometric data. The data format is time-deviation pairs, with the deviation value in kilopascals. The vertical-lateral pressure load second-order differential fluctuation calculation of the vertical-lateral structural stiffness is performed based on the growth iteration deviation geometric data. The specific operation method is: first, the vertical stiffness and lateral stiffness data of each grid unit node of the foundation are used as the basis, and for each unit node, the growth iteration deviation under the corresponding time step is superimposed, and the growth iteration deviation is equated with the new pressure load per unit area.Then, for each node, the pressure load values for three consecutive time steps are recorded, respectively as the load at the previous moment, the load at the intermediate moment, and the load at the next moment. A second-order difference operation is performed in time series, that is, the load value at the next moment minus twice the load value at the intermediate moment plus the load value at the previous moment, to obtain the second-order pressure fluctuation value of the node at the current time step. This process is performed synchronously for all grid nodes, resulting in a table of second-order differential fluctuation values corresponding to the node number. Considering that the vertical and lateral directions are operated independently, the second-order pressure differential fluctuation data in the vertical direction and the second-order pressure differential fluctuation data in the lateral direction are generated respectively. All data are output in the format of node number, time step, and second-order differential fluctuation quantity, and the unit is uniformly kilopascals. A vertical-lateral pressure differential fluctuation calculation is performed on the vertical-lateral pressure load differential fluctuation data to generate the foundation vertical-lateral pressure fluctuation difference. The specific operation method is as follows: First, the second-order differential fluctuations in the vertical and lateral directions for each node at each time step are summarized. For each node, the absolute difference between the vertical and lateral fluctuation values is taken as the load pressure differential fluctuation value. Subsequently, to reflect the overall foundation load fluctuation trend, a weighted average is performed based on the spatial distribution of the nodes. The weight coefficient is set according to the unit area percentage of the node, with nodes with larger area percentages being given a higher weight. The specific processing flow is as follows: at each moment, the load pressure differential fluctuation values of all nodes are multiplied by the node area weight, the sum is then divided by the total area weight to obtain the overall foundation load pressure fluctuation difference value at that moment. This process is repeated at each time step, forming a time-series foundation vertical-lateral pressure fluctuation difference dataset. The data format is time-load pressure differential fluctuation pairs, with units of kilopascals. Finally, the data is compiled into a continuous curve for use in the subsequent foundation deformation derivation step.

[0039] Step S244 includes the following steps: The vertical longitudinal supporting force and horizontal transverse supporting force vector fields are analyzed for the vertical-lateral load pressure fluctuation difference of the foundation, and the vertical longitudinal supporting force vector field and horizontal transverse supporting force vector field are obtained respectively; The vertical longitudinal jacking force vector field is fitted with the longitudinal jacking force peak phase continuity to obtain the longitudinal jacking force peak phase continuity data; The horizontal lateral jacking force vector field is analyzed for the lateral jacking force angle component offset to obtain the lateral jacking force angle component offset; The vector field divergence-curl decomposition data are obtained by performing vector field divergence-curl decomposition processing based on the longitudinal top force peak phase continuous data and the lateral top force angle component offset; The anisotropic deflection angle of foundation jacking is derived from the foundation excavation structure data according to the vector field divergence-curl decomposition data, and the deflection angle of foundation jacking direction is obtained.

[0040] In an embodiment of the present invention, the vertical longitudinal support force and horizontal transverse support force vector field analysis of the vertical-lateral load pressure fluctuation difference of the foundation are performed. The specific operation method is as follows: first, the vertical load pressure fluctuation difference and the lateral load pressure fluctuation difference of each node of the foundation at each time step are extracted separately, and the load fluctuation difference is set to be proportional to the support force size per unit area. The proportional coefficient is set so that 1 square meter of area corresponds to 1 kPa load and produces 1 kN support force. Based on this proportional coefficient, the vertical load fluctuation difference is converted into a vertical support force vector at each node, with the vertical upward direction as the positive direction. At the same time, the lateral load fluctuation difference is converted into a horizontal support force vector. The vector is expanded along the surface plane, with the east direction as the starting point of 0°, and the direction angle is calculated counterclockwise. The support force size and direction of each node are coordinateized, and two sets of three-dimensional data are constructed: one set is the node X and Y coordinates plus the vertical support force size, and the other set is the node X and Y coordinates plus the horizontal support force size and direction angle. Through standardization, the force values are normalized to between 0 and 1 to facilitate subsequent unified processing. The standardization is set based on the maximum support force in the entire area. Finally, the vertical support force vector field data and the horizontal support force vector field data are generated respectively. The data format is a four-tuple of node number, coordinates, support force magnitude, and direction angle. The vertical support force vector field is fitted with the longitudinal support force peak phase continuity. The specific operation method is as follows: first, all nodes in the vertical support force vector field with support force values greater than the overall mean plus one standard deviation are selected as peak nodes to ensure that the local maximum support force area is selected. Then, the nodes are sorted according to the X-axis coordinates. Based on the spatial position distribution of the nodes in the X-axis direction, the least squares method is used for fitting. The fitting curve adopts the form of a cubic spline curve to ensure that the curve has first-order continuity and second-order continuity between each node. During the fitting process, the error tolerance is set to ±5% for each fitting segment, that is, the error between the fitted value and the actual value of each peak node shall not exceed 5% of the actual support force value. If the error tolerance is exceeded, the weights of the fitting nodes are automatically adjusted to be closer to the actual measured values, ensuring a continuous and smooth peak range. After fitting, the phase information corresponding to each time step of the fitting curve is extracted. Phase information is defined as the normalized position of the fitting curve in the node spatial sequence, with a value range of 0 to 1. This generates continuous phase data of the longitudinal support force peak. The data format is a triplet of node number, phase value, and fitted support force magnitude. The horizontal support force angular component offset is analyzed for the horizontal support force vector field. The specific operation method is as follows: first, the horizontal support force value and corresponding azimuth angle of each node in the horizontal support force vector field are extracted. The horizontal support force vector field is then divided into a spatial grid, with each grid cell being 50 meters by 50 meters. All support force vectors within each grid cell are then synthesized. The synthesis method decomposes each vector into its components in the X and Y directions. The X-direction force component is the support force value multiplied by the cosine of the azimuth angle, and the Y-direction force component is the support force value multiplied by the sine of the azimuth angle.The X- and Y-direction force components of all nodes within the cell are then summed to obtain the overall composite force component of the grid cell. Based on the overall force components, the composite jacking force angle for the cell is calculated. This angle is the angle between the inverse tangent of the X- and Y-direction composite forces, ranging from -180° to +180°. This angle is then compared with the theoretical principal direction of the grid center (set to true north, i.e., 90°) to determine the offset between the two. The offset is expressed as an absolute value. Finally, the center coordinates, lateral composite jacking force magnitude, and angular component force offset of each grid cell are output to form lateral jacking force and angular component force offset data. The data format is a four-tuple consisting of grid number, center coordinates, composite jacking force magnitude, and angular component force offset. A vector field divergence-curl decomposition process is performed based on the phase-continuous data of the longitudinal jacking force peak value and the lateral jacking force angular component offset. The specific operation method is as follows: First, the phase-continuous data of the longitudinal jacking force peak value and the lateral jacking force angular component offset data are node-concatenated. Based on the spatial coordinates, the two data sets are reorganized in the same coordinate system to form a six-tuple consisting of node number, node spatial coordinate, vertical jacking force value, phase continuity parameter, horizontal jacking force value, and angular offset. Then, for each node, all nodes within a 20-meter radius are used as the local neighborhood, and the finite difference method is used to extract the vertical and horizontal component change rates of the nodes within the neighborhood. The vertical component change rate is defined as the difference between the vertical jacking force values of adjacent nodes divided by the distance between the two nodes, and the horizontal component change rate is defined as the difference between the horizontal jacking force values of adjacent nodes multiplied by the cosine angle difference divided by the distance between the two nodes. Local continuity is ensured by taking a weighted average of all the change rate data within the local neighborhood, with weights set according to the inverse square of the node distance, with closer nodes receiving a larger weight. According to the divergence calculation principle, the local divergence of the node is defined as the sum of the horizontal and vertical change rates, and the curl is defined as the change rate of the horizontal component in the vertical direction minus the change rate of the vertical component in the horizontal direction, and finally the decomposition processing of the node divergence and curl is completed. After the processing is completed, the divergence value, curl value and original node attributes of each node are output to form vector field divergence-curl decomposition data. The data format is node number, X coordinate, Y coordinate, divergence value, curl value, vertical jacking force value, and horizontal jacking force value. Example: Based on the vector field divergence-curl decomposition data, the foundation jacking anisotropic deflection angle of the foundation excavation structure data is derived. The specific operation method is as follows: first, each structural node in the foundation excavation structure data is spatially paired with the corresponding vector field divergence-curl decomposition data. The matching principle is that the node is considered to be a corresponding node when the spatial distance between the nodes is less than 5 meters. For each structural node, its corresponding divergence value and curl value are extracted, and the area with positive divergence value is defined as the jacking concentration area, and the area with positive curl value is defined as the jacking deflection area. According to the relationship between divergence and curl, the vector superposition method is used to deduce the anisotropic deflection direction.The specific derivation method is: if the divergence value is greater than the rotation value, the divergence direction is the main direction and the rotation direction is the secondary direction, and the combined support direction is formed after superposition; if the rotation value is greater than the divergence value, the rotation direction is the main direction and the divergence direction is the secondary direction, and the combined support direction is formed after superposition. The calculation of the support direction adopts the following rules: taking the positive direction of the divergence as the reference, if the rotation is positive, the comprehensive direction deviates from the divergence direction by 15 degrees clockwise; if the rotation is negative, the comprehensive direction deviates from the divergence direction by 15 degrees counterclockwise. The rotation angle is 15 degrees as an empirical value, which can be adjusted according to actual observations in the area, but this embodiment uses 15 degrees as a fixed value. Finally, based on the node coordinates and the derived support direction angle, the foundation support direction deflection angle data is generated. The data format is the structural node number, X coordinate, Y coordinate, and support deflection angle (values within the range of 0 to 360 degrees).

[0041] Step S25 includes the following steps: Step S251: Analyze the displacement jump values between the interfaces of each soil layer on the quantitative data of the soil expansion building displacement to obtain the displacement jump values between the interfaces of the soil layers; Step S252: performing an offset direction component analysis on the quantitative data of building displacement caused by soil expansion based on the displacement jump value between soil layer interfaces to obtain the offset direction component of the building displacement; Step S253: performing displacement acceleration geometric increment identification based on the displacement jump value between the soil layer interfaces and the building displacement direction component to obtain displacement acceleration geometric increment data; Step S254: performing a progressive periodic change integration process on the offset acceleration geometric increment data to obtain offset acceleration progressive integral data; Step S255: Performing slight-change expansion displacement accumulation processing according to the offset acceleration progressive integral data to obtain slight-change expansion displacement accumulation data.

[0042] In this embodiment of the present invention, based on the layered structure of the soil profile, the building base to the ground surface is vertically divided into several continuous soil layers. The thickness of each soil layer is determined based on the engineering geological survey report. In this embodiment, the layer thickness is 3 meters, and the total profile thickness is set to 30 meters, with a total of 10 layers. The soil layer number and vertical displacement value corresponding to each node in the quantified displacement data of the soil expansion building are extracted to form a triplet of node number, soil layer number, and vertical displacement. For each vertical node sequence, the displacement jump value at the interface between the upper and lower soil layers is calculated, with the adjacent upper and lower soil layer nodes grouped together. The displacement jump value is defined as the vertical displacement of the upper node minus the vertical displacement of the lower node. To avoid errors caused by local abnormal disturbances, a five-point moving average is applied to each group of jump values. This process takes the average of the jump values of the current node and the two nodes above and below it to smooth out local fluctuations. The final displacement jump value data for each soil layer interface is obtained. The data format is node number, upper layer number, lower layer number, and jump value (unit: millimeter). Based on the displacement jump values between soil layer interfaces, the directional components of the building displacement quantification data due to soil expansion are analyzed. The specific processing flow is as follows: First, the directional components of the node corresponding to the jump value at each interface are extracted. Based on the magnitude of the jump value, the jump behavior at the interface is divided into two cases: a positive jump value is defined as an upward displacement, and a negative jump value is defined as a downward settlement displacement. Combined with the horizontal components of the original quantified displacement data, the horizontal displacement in the X and Y directions of each node is extracted. The sign of the displacement determines the displacement to the east, west, north, or south, respectively. Finally, the vertical offset is integrated with the two horizontal offset components to obtain the complete directional component data of the building displacement. The specific data structure includes the node number, the X-direction offset (mm), the Y-direction offset (mm), and the Z-direction offset (mm). The Z-direction offset is the jump value, and the X and Y-direction offsets are the horizontal components of the original quantified displacement data. Based on the displacement jump value between the soil layer interface and the component of the building displacement offset direction, the displacement acceleration is identified in proportional increments. The specific processing flow is as follows: First, based on the displacement data of the continuous time series sampling in the same vertical node sequence, displacement data is collected every 1 hour to form a displacement time series, and the sampling period is fixed at 1 hour. For each node, the displacement velocity, i.e., the displacement change rate, is defined as the difference between the current time t and the previous time t-1 divided by the time difference. The displacement acceleration is then defined as the difference between the current velocity and the previous velocity divided by the time difference. The specific formula is: displacement velocity equals the current displacement minus the previous displacement divided by the time difference; displacement acceleration equals the current velocity minus the previous velocity divided by the time difference.A geometric increment identification method is then applied to the acceleration sequence of each node. This method calculates the ratio of acceleration values at three consecutive moments. If any consecutive acceleration ratio exceeds a preset threshold (set to 1.5 in this example) and the increasing trend persists for more than three times, it is identified as a geometric increment phenomenon. Finally, the node number, time period, and acceleration increment value that meet the requirements are extracted to form offset acceleration geometric increment data. The data format includes the node number, start time, end time, and acceleration increase factor.

[0043] The geometrically incremented offset acceleration data is subjected to a progressive periodic integration process. The specific operation is as follows: First, the corresponding time series acceleration data is extracted using each node number in the geometrically incremented offset acceleration data as an index. The basic sampling period is defined as 1 hour, with each 24-hour period as a complete period window. Based on the acceleration data within a fixed period window, a cumulative integration process is used. Specifically, within each period window, the hourly acceleration value is multiplied by the time interval and then accumulated to obtain the displacement change within that period. The specific calculation formula is: the progressive integral within a period window is equal to the sum of the acceleration of each time step multiplied by the time interval, with the time interval fixed at 1 hour. To reflect the progressive change of acceleration, the acceleration value at each time step is corrected by first-order linear interpolation during the integration process. The interpolation method is to take the average of the acceleration of the current time step and the acceleration of the previous time step as the acceleration value used in the current step. After the cumulative integration method is used, the progressive integral of the offset acceleration for each node within each period window is obtained. The data output format includes the node number, cycle number, cycle start time, cycle end time, and cycle cumulative displacement increment (in millimeters). In this embodiment, for actual foundation monitoring data, the sampling range is set to 90 consecutive days, with a total of 90 integration cycles. The integration results of each cycle are stored separately to form a complete progressive integration data set. Micro-change expansion displacement accumulation processing is performed based on the progressive integration data of offset acceleration. The specific operation process is as follows: First, using the progressive integration data of offset acceleration as the input, the continuous cycle cumulative displacement increment data sequence for each node number is extracted. Because the actual soil expansion deformation process exhibits micro-accumulation characteristics, a micro-change identification threshold is set at 0.5 mm. That is, when the difference between two consecutive cycle displacement increments is less than or equal to 0.5 mm, it is considered a micro-change. Based on this criterion, a cycle-by-cycle sliding judgment is performed on the cycle integration sequence of each node, and data segments that meet the micro-change criteria are marked. Then, all data segments marked as micro-change are accumulated and summed, defining the micro-change cumulative expansion displacement. Specifically, the cycle integrals that continuously meet the micro-change criteria are directly accumulated. To ensure the continuity of the accumulated data, during periods with abrupt changes (i.e., periods where the micro-change condition is not met), the accumulated value is reset to zero and a new accumulation cycle begins. The final output is the cumulative expansion displacement micro-change data, which includes the node number, the cumulative start cycle number, the cumulative end cycle number, and the total accumulated displacement (in millimeters). In this example, the actual sampled data was extracted for node N-1001. Its cumulative period of continuous micro-change conditions lasted from the 15th to the 40th cycle, corresponding to a cumulative expansion displacement of 28 mm, which is fully recorded in the output data set.

[0044] Step S3 includes the following steps: Step S31: normalizing the cumulative data of the slight change in expansion displacement to obtain the normalized cumulative data of the slight change in expansion displacement; Step S32: performing convergence constraint training on the expansion displacement slight change accumulated normalized data based on the deep Q network to obtain expansion displacement slight change convergence constraint data; Step S33: Predicting the risk of building expansion cracking based on the expansion displacement micro-change convergence constraint data to obtain building expansion cracking risk prediction data.

[0045] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes: Step S31: normalizing the cumulative data of the slight change in expansion displacement to obtain the normalized cumulative data of the slight change in expansion displacement; In this embodiment of the present invention, the maximum and minimum parameters required for normalization are first determined based on the cumulative displacement of each node in the cumulative expansion displacement data. The maximum value is defined as the maximum cumulative displacement of all nodes in the dataset, and the minimum value is defined as the minimum cumulative displacement of all nodes. Based on the minimum-maximum normalization method, the following normalization calculation formula is used: the normalized value is equal to the original displacement minus the minimum value, divided by the difference between the maximum and minimum values. This processing method ensures that all normalized cumulative expansion displacement data fall between 0 and 1. In a specific experiment, the maximum displacement of the collected original expansion displacement data was 34.8 mm, and the minimum displacement was 1.2 mm. During the normalization process, all node data were processed according to the above formula. After processing, the normalized results were saved as a new dataset. The data fields include the node number, the original displacement, and the normalized displacement, and the format is standardized to retain six significant digits after the decimal point.

[0046] Step S32: performing convergence constraint training on the expansion displacement slight change accumulated normalized data based on the deep Q network to obtain expansion displacement slight change convergence constraint data; In an embodiment of the present invention, a state space is first constructed based on the normalized cumulative data of the expanded displacement micro-change, and each element of the state space corresponds to the change trend of the normalized displacement of a node at different time steps. The action space is defined as whether a risk judgment label is applied, and the labels are divided into two categories: convergence and expansion. The reward function is set to reward +1 when the convergence label hits the true convergence trend, and punish -1 when it is misjudged. A deep Q network is used for training, and the input layer receives the normalized displacement sequence of the current node as input. The hidden layer is set to 3 layers, and the number of nodes in each layer is 128, 64, and 32 respectively. The activation function uses the ReLU function. The experience replay mechanism is used in the training process, the batch size of each training round is 64, the Adam optimizer is used, the initial learning rate is set to 0.001, and the total number of training rounds is set to 500 rounds. During the training process, the model parameters are saved every 50 rounds for easy use in the subsequent reasoning stage. The output format of convergence constraint data consists of three parts: node number, normalized displacement trend label, and final judgment category. The normalized displacement trend label is marked as a string, such as "convergence" or "extension", and the final judgment category is represented by a Boolean value.

[0047] Step S33: Predicting the risk of building expansion cracking based on the expansion displacement micro-change convergence constraint data to obtain building expansion cracking risk prediction data.

[0048] In this embodiment of the present invention, convergence constraint data is first linked to the spatial distribution data of the building's geographic unit, using the node number as an index. The building foundation area corresponding to each node is located. A cracking risk criterion is set based on the final classification of each node. If more than 30% of the nodes in a building foundation area are classified as expansion, the area is marked as presenting an expansion cracking risk. In practice, the judgment results of all nodes in each building foundation area are first counted, and the proportion of expansion nodes is calculated and compared with a set threshold. If the proportion exceeds the threshold, a building expansion cracking risk prediction result is output. The prediction data format includes five items: building number, total number of nodes in the area, number of expansion nodes, expansion ratio, and whether there is a cracking risk. The data is uniformly stored in a table format. For example, in an actual building foundation area analysis, there were 150 nodes in the building area numbered B-2005, of which 52 were classified as expansion, with an expansion ratio of 34.67%, exceeding the set 30% threshold. Therefore, the final output result indicated the presence of an expansion cracking risk.

[0049] Step S32 includes the following steps: Step S321: extract random samples from the cumulative normalized data of the slight change in expansion displacement to obtain a random sample of the slight change in expansion displacement; Step S322: performing multi-step learning processing on the cumulative random samples of the slight change in the expansion displacement to obtain sample multi-step learning convergence data; Step S333: Based on the deep Q network, convergence constraint training is performed on the sample multi-step learning convergence data to obtain expansion displacement slight change convergence constraint data.

[0050] In this embodiment of the present invention, a sample selection ratio parameter is first determined based on the complete normalized dataset of incremental displacement changes. A random sampling ratio of 30% is set, meaning that 30% of the node data in the entire dataset is randomly selected as training samples. A fixed random seed of 2025 is used to ensure reproducibility of the selection process. The random selection method uses uniformly distributed sampling, meaning that each node has an equal probability of being selected, without any stratification or weight adjustment. After the selection is complete, a random sample set of incremental displacement changes is formed. Each sample contains the node number and the corresponding normalized displacement sequence information. For example, in a practical application, the original dataset contains 10,000 nodes. A 30% sampling ratio is used to obtain 3,000 node samples. The sample data is stored in a standard CSV format, with the fields arranged in the following order: node number, normalized displacement at time step t0, normalized displacement at time step t1, normalized displacement at time step t2, and so on, up to the normalized displacement at time step tN. All data are stored to six decimal places. Multi-step learning is performed on random samples of cumulative dilation displacement micro-changes. The specific operation is as follows: Based on the extracted random sample set, a time series multi-step learning method is used for processing. The normalized displacement data sequence of each node is reorganized using a time-step sliding window method. The window width is set to 5, that is, the normalized displacement values of five consecutive time steps are taken as an input sample, and the corresponding predicted target is the normalized displacement change of the subsequent time step. In the specific implementation, each node sequence is first sliced, all available sliding window subsequences are extracted, and the corresponding target change is recorded. In this implementation, the node sequence length is 100 time steps, and the sliding window step size is set to 1, that is, it slides backward one time step at a time, generating 96 sets of input and output samples. All generated input and output samples are uniquely identified by node number and time index number, ultimately forming a sample multi-step learning convergence dataset. The data format includes the normalized displacement values of the five time steps within the input window and the corresponding predicted target normalized displacement change. All data are organized into batches, each containing 1000 samples, and are uniformly normalized to maintain a consistent numerical scale. First, the sample multi-step learning convergence data was divided into a training set and a validation set with a ratio of 8:2. The training input was a normalized displacement sequence of 5 time steps, and the output was a class label indicating whether the convergence trend was detected. After the input data was normalized, it was fed into a deep Q-network for training. The input layer was set to 5 nodes, and the hidden layer adopted a two-layer fully connected structure. The first hidden layer had 128 nodes, the second hidden layer had 64 nodes, and the output layer had 2 nodes, corresponding to the convergence and non-convergence categories, respectively. The mean squared error loss function was used for training, and the Adam algorithm was used as the optimizer. The initial learning rate was set to 0.0005, and the total number of training rounds was set to 300. The gradient was updated using a batch size of 64 data in each round. To enhance the stability of training, a target network mechanism was introduced, where the parameters of the training main network were synchronously updated to the target network every 10 rounds.During training, an experience replay pool is used to buffer samples. The pool capacity is set to 10,000 samples, and 64 samples are randomly selected from the pool for training each time. After the convergence constraint training is completed, the expansion displacement micro-variation convergence constraint data is output. The data structure contains three parts: node number, predicted category label, and predicted probability value. Each data point retains the normalized displacement sequence characteristics for subsequent risk inference analysis.

[0051] The present invention also provides a construction land approval risk warning system for executing the construction land approval risk warning method described above. The construction land approval risk warning system includes: The geological profile layered analysis module is used to obtain the geological structure data of the construction land uploaded by the client; perform geological profile layered analysis on the geological structure data of the construction land to obtain geological profile layered data of the construction land; The expansion displacement accumulation module is used to analyze the soil layer viscosity state of the construction land geological profile layer data to obtain soil layer viscosity state data; simulate and quantify the soil layer expansion building displacement based on the soil layer viscosity state data to obtain soil layer expansion building displacement quantitative data; and perform micro-variation expansion displacement accumulation processing on the soil layer expansion building displacement quantitative data to obtain micro-variation expansion displacement accumulation data; A cracking risk prediction module is used to predict the risk of building expansion cracking based on the cumulative data of micro-changes in expansion displacement to obtain building expansion cracking risk prediction data; The risk pre-factor identification module is used to identify the risk pre-factors of construction land based on the building expansion and cracking risk prediction data, and obtain the risk pre-factor data of construction land; the risk pre-factor data of construction land is fed back to the terminal to implement the risk warning of construction land approval.

[0052] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A construction land approval risk early warning method, characterized in that: The following steps are involved: Step S1: Acquire the geological structure data of the construction land uploaded by the client; perform geological profile layering analysis on the geological structure data of the construction land to obtain geological profile layering data of the construction land; Step S2: performing soil layer viscosity state analysis on the construction land geological profile layer data to obtain soil layer viscosity state data; performing soil layer expansion building displacement simulation and quantification based on the soil layer viscosity state data to obtain soil layer expansion building displacement quantification data; performing micro-variation expansion displacement accumulation processing on the soil layer expansion building displacement quantification data to obtain micro-variation expansion displacement accumulation data; Step S3: Predicting the risk of building expansion cracking based on the cumulative data of expansion displacement micro-changes to obtain building expansion cracking risk prediction data; Step S4: Identifying construction land risk preconditions for approval based on the building expansion and cracking risk prediction data to obtain construction land risk precondition data; Feedback the data of pre-risk factors of construction land to the terminal to implement risk warning for construction land approval.

2. The construction land approval risk early warning method according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: Acquire the geological structure data of the construction land uploaded by the client; Step S12: performing data cleaning on the geological structure data of the construction land to obtain the geological structure cleaning data of the construction land; Step S13: Performing geological profile layering analysis on the cleaned geological structure data of the construction land to obtain geological profile layering data of the construction land.

3. The construction land approval risk early warning method according to claim 1 is characterized in that: Step S2 includes the following steps: Step S21: Obtain construction land development plan; Step S22: extracting the construction foundation excavation structure of the construction land development plan to obtain foundation excavation structure data; Step S23: performing soil layer viscosity state analysis on the construction land geological profile layer data to obtain soil layer viscosity state data; Step S24: performing soil expansion building displacement simulation and quantification on the foundation excavation structure data according to the soil layer viscosity state data to obtain soil expansion building displacement quantification data; Step S25: performing micro-change expansion displacement accumulation processing on the quantified data of soil expansion building displacement to obtain micro-change expansion displacement accumulation data.

4. The construction land approval risk early warning method according to claim 3 is characterized in that: Step S24 includes the following steps: Step S241: simulating the volume change of soil stratified water absorption and expansion per unit time on the soil stratified viscosity state data to obtain the volume change data of soil stratified water absorption and expansion; Step S242: analyzing the nonlinear exponential growth relationship of the expansion pressure based on the volume change data of the soil stratified water absorption expansion to obtain the exponential growth relationship of the soil expansion pressure; Step S243: performing vertical-lateral load pressure difference fluctuation calculation on the foundation excavation structure data according to the soil expansion pressure exponential growth relationship to generate the foundation vertical-lateral load pressure fluctuation difference; Step S244: deriving the foundation jacking anisotropic deflection angle from the foundation excavation structure data according to the foundation vertical-lateral load pressure fluctuation difference to obtain the foundation jacking direction deflection angle; Step S245: simulating and quantifying the displacement of the building due to soil expansion according to the deflection angle of the foundation support direction, and obtaining quantitative data of the displacement of the building due to soil expansion.

5. The construction land approval risk early warning method according to claim 4 is characterized in that: Step S243 includes the following steps: Perform geometric boundary discretization processing on the foundation excavation structure data to obtain the foundation geometric boundary grid data; perform vertical-lateral structural stiffness analysis on the foundation geometric boundary grid data to obtain the vertical-lateral structural stiffness; Performing exponential growth curve conversion on the exponential growth relationship of soil swelling pressure to obtain an exponential growth curve of soil swelling pressure; Based on the secant method, the slope variance between the growth inflection points of the soil expansion pressure exponential growth curve is calculated to obtain the inflection point growth slope variance; According to the inflection point growth slope variance, the growth iteration deviation geometric calculation is performed on the soil expansion pressure exponential growth curve to obtain the growth iteration deviation geometric data; Based on the growth iterative deviation geometric data, the vertical-lateral structural stiffness is calculated by the second-order differential fluctuation of the vertical-lateral pressure load to obtain the vertical-lateral pressure load differential fluctuation data; The vertical-lateral pressure load differential fluctuation data is subjected to vertical-lateral load pressure difference fluctuation calculation to generate the foundation vertical-lateral load pressure fluctuation difference.

6. The construction land approval risk early warning method according to claim 4 is characterized in that: Step S244 includes the following steps: The vertical longitudinal supporting force and horizontal transverse supporting force vector fields are analyzed for the vertical-lateral load pressure fluctuation difference of the foundation, and the vertical longitudinal supporting force vector field and horizontal transverse supporting force vector field are obtained respectively; The vertical longitudinal jacking force vector field is fitted with the longitudinal jacking force peak phase continuity to obtain the longitudinal jacking force peak phase continuity data; The horizontal lateral jacking force vector field is analyzed for the lateral jacking force angle component offset to obtain the lateral jacking force angle component offset; The vector field divergence-curl decomposition data are obtained by performing vector field divergence-curl decomposition processing based on the longitudinal top force peak phase continuous data and the lateral top force angle component offset; The anisotropic deflection angle of foundation jacking is derived from the foundation excavation structure data according to the vector field divergence-curl decomposition data, and the deflection angle of foundation jacking direction is obtained.

7. The construction land approval risk early warning method according to claim 3 is characterized in that: Step S25 includes the following steps: Step S251: Analyze the displacement jump values between the interfaces of each soil layer on the quantitative data of the soil expansion building displacement to obtain the displacement jump values between the interfaces of the soil layers; Step S252: performing an offset direction component analysis on the quantitative data of building displacement caused by soil expansion based on the displacement jump value between soil layer interfaces to obtain the offset direction component of the building displacement; Step S253: performing displacement acceleration geometric increment identification based on the displacement jump value between the soil layer interfaces and the building displacement direction component to obtain displacement acceleration geometric increment data; Step S254: performing a progressive periodic change integration process on the offset acceleration geometric increment data to obtain offset acceleration progressive integral data; Step S255: Performing slight-change expansion displacement accumulation processing according to the offset acceleration progressive integral data to obtain slight-change expansion displacement accumulation data.

8. The construction land approval risk early warning method according to claim 1 is characterized in that: Step S3 includes the following steps: Step S31: normalizing the cumulative data of the slight change in expansion displacement to obtain the normalized cumulative data of the slight change in expansion displacement; Step S32: performing convergence constraint training on the expansion displacement slight change accumulated normalized data based on the deep Q network to obtain expansion displacement slight change convergence constraint data; Step S33: Predicting the risk of building expansion cracking based on the expansion displacement micro-change convergence constraint data to obtain building expansion cracking risk prediction data.

9. The construction land approval risk early warning method according to claim 8 is characterized in that: Step S32 includes the following steps: Step S321: extract random samples from the cumulative normalized data of the slight change in expansion displacement to obtain a random sample of the slight change in expansion displacement; Step S322: performing multi-step learning processing on the cumulative random samples of the slight change in the expansion displacement to obtain sample multi-step learning convergence data; Step S333: Based on the deep Q network, convergence constraint training is performed on the sample multi-step learning convergence data to obtain expansion displacement slight change convergence constraint data.

10. A construction land approval risk warning system, characterized by: For executing the construction land approval risk warning method according to claim 1, the construction land approval risk warning system comprises: The geological profile layered analysis module is used to obtain the geological structure data of the construction land uploaded by the client; perform geological profile layered analysis on the geological structure data of the construction land to obtain geological profile layered data of the construction land; The expansion displacement accumulation module is used to analyze the soil layer viscosity state of the construction land geological profile layer data to obtain soil layer viscosity state data; simulate and quantify the soil layer expansion building displacement based on the soil layer viscosity state data to obtain soil layer expansion building displacement quantitative data; and perform micro-variation expansion displacement accumulation processing on the soil layer expansion building displacement quantitative data to obtain micro-variation expansion displacement accumulation data; A cracking risk prediction module is used to predict the risk of building expansion cracking based on the cumulative data of micro-changes in expansion displacement, and obtain building expansion cracking risk prediction data; The risk pre-factor identification module is used to identify the risk pre-factors of construction land based on the building expansion and cracking risk prediction data, and obtain the risk pre-factor data of construction land; the risk pre-factor data of construction land is fed back to the terminal to implement the risk warning of construction land approval.

Images