Lateral deformation prediction system and method based on deep foundation pit wall

Abstract

The invention relates to the technical field of deformation prediction, in particular to a lateral deformation prediction system and method based on a deep foundation pit wall body, and the system comprises a deformation data collection module which collects a ground surface settlement monitoring value and a lateral displacement monitoring value of a corresponding measuring point of the deep foundation pit wall body, carries out the time sequence alignment processing of data points, removes an abnormal monitoring value, and obtains a deformation data collection module; and establishing calibration deformation sequence data. According to the invention, through real-time acquisition and strict execution of data time sequence alignment processing, the consistency of ground surface settlement data and wall lateral displacement data is ensured, abnormal monitoring data is effectively eliminated, and the data accuracy and consistency are improved; in the spatial interpolation process, width measurement of a settlement area is introduced, the width characteristic of a settlement groove is determined, and the earth surface deformation trend is captured according to a deformation index; and extracting the maximum lateral displacement of each time section of the monitoring section, establishing a dynamic corresponding relation between the maximum displacement and the width of the settling tank, and identifying the regular influence of the change of the construction process on the lateral deformation.

Description

Technical Field

[0001] The present invention relates to the technical field of deformation prediction, and particularly to a lateral deformation prediction system and method based on a deep foundation pit wall. Background Art

[0002] The technical field of deformation prediction is an important part of civil engineering safety monitoring, and its main research objects are civil structures such as bridges, high-rise buildings, slopes, tunnels, and deep foundation pits. In this field, deformation data such as settlement, inclination, and displacement of structures are collected in real time through sensors, and potential risks are identified in advance and early warnings are given by means of data analysis, trend prediction, and risk assessment to ensure the structural safety during the construction and use stages of the project.

[0003] In the prior art, the data averaging interpolation of the overall area is mainly used, and the key settlement areas are not accurately measured and analyzed, resulting in the monitoring results being unable to reflect the important change trends of local deformations; at the same time, the research on the relationship between the extraction of the maximum displacement of the wall and the process evolution is insufficient, making the analysis of the wall deformation trend relatively rough and difficult to capture the key risk nodes during the construction process. Therefore, improvements are needed. Summary of the Invention

[0004] The purpose of the present invention is to solve the deficiencies existing in the prior art, and to propose a lateral deformation prediction system and method based on a deep foundation pit wall.

[0005] In order to achieve the above purpose, the present invention adopts the following technical solution: A lateral deformation prediction system based on a deep foundation pit wall includes: A deformation data acquisition module, which collects the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points on the deep foundation pit wall, performs time series alignment processing and abnormal monitoring value elimination on the data points, and establishes calibrated deformation sequence data; A surface pattern recognition module, based on the surface settlement monitoring values in the calibrated deformation sequence data, generates surface settlement isograms through spatial interpolation to obtain a surface settlement isogram map, and based on the surface settlement isogram map, measures the width of the settlement trough formed by the key settlement areas to establish the settlement trough width characteristics; A wall response correlation module, based on the lateral displacement monitoring values of the wall in the calibrated deformation sequence data, extracts the maximum lateral displacement values of each monitoring section at the same time section to obtain the maximum wall displacement value, and based on the maximum wall displacement value, analyzes the corresponding relationship between the evolution with the excavation process and the change of the settlement trough width characteristics to establish the surface-wall evolution correlation rule; The deformation trend prediction module, based on the width characteristics of the settlement tank, queries the wall behavior pattern corresponding to the width characteristics in the evolution association rules of the surface wall, obtains the expected wall response pattern, and based on the expected wall response pattern, combines the deformation state of the current deep foundation pit wall to predict the lateral deformation value of the deep foundation pit wall within the next time window, and obtains the prediction result of the lateral deformation of the wall.

[0006] Preferably, the steps for obtaining the calibrated deformation sequence data are as follows: Collect the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points of the deep foundation pit wall, match the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points of the deep foundation pit wall one by one according to the collection time, and mark and eliminate the unmatched monitoring values to generate the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points of the deep foundation pit wall after time series alignment and outlier elimination; Based on the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points of the deep foundation pit wall after time series alignment and outlier elimination, group the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points of the deep foundation pit wall according to the measuring point numbers, and check the data integrity within each group to eliminate the incomplete data groups, and generate data pairs grouped according to the measuring point numbers and ensuring integrity; Based on the data pairs grouped according to the measuring point numbers and ensuring integrity, sort the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points of each group of data pairs to form calibrated deformation sequence data.

[0007] Preferably, the steps for obtaining the surface settlement isogram are as follows: Based on the surface settlement monitoring values in the calibrated deformation sequence data, sort the surface settlement monitoring values according to the spatial coordinates of the measuring points, and calculate the difference between the surface settlement monitoring values between adjacent measuring points and the distance measurement of each measuring point to generate the difference between adjacent measuring point settlements and distance measurement data; According to the difference between adjacent measuring point settlements and distance measurement data, calculate the surface settlement isogram interpolation factor; Based on the surface settlement isogram interpolation factor, perform grid distribution fitting on the surface settlement monitoring values through spatial interpolation, and adjust the interpolation density of each grid unit according to the surface settlement isogram interpolation factor during the grid process to form a surface settlement isogram.

[0008] Preferably, the steps for obtaining the width characteristics of the settlement tank are as follows: Based on the surface settlement isogram, identify the area where the settlement isograms are the densest in the surface settlement isogram, and mark the outer edge coordinate points of the area to obtain the outer edge coordinate points of the key settlement area; Based on the outer edge coordinate points of the key settlement area, the horizontal distance between the outer edge coordinate points is measured, and the representative distance is selected according to the principle of maximum horizontal spacing to obtain the maximum horizontal width of the key settlement area; Based on the maximum horizontal width of the key sedimentation area, the maximum horizontal width is used as a characteristic indicator of the sedimentation tank, and the width value is recorded as the sedimentation tank width feature to form the sedimentation tank width feature.

[0009] Preferably, the steps for obtaining the maximum displacement value of the wall are: Based on the wall lateral displacement monitoring values ​​in the calibration deformation sequence data, the wall lateral displacement monitoring values ​​are grouped according to the monitoring section numbers, and all observation values ​​of the same time section are extracted for the wall lateral displacement monitoring values ​​in each group of monitoring sections, and the valid observation values ​​in the time section are selected by comparing the timestamps to obtain the valid lateral displacement monitoring values ​​of each monitoring section at the same time section; Based on the effective lateral displacement monitoring values ​​of each monitoring section at the same time section, traverse each group of data in turn and compare the size of each observation value in the group, select the lateral displacement monitoring value with the largest value in each group, and record the corresponding monitoring section number and observation time at the same time to obtain the maximum lateral displacement value set of each monitoring section; Based on the maximum lateral displacement value set of each monitoring section, the sizes of all the maximum lateral displacement values ​​in the set are compared, and the highest group is selected as the representative value. At the same time, the monitoring section number and observation time of the representative value are marked to form the maximum displacement value of the wall.

[0010] Preferably, the steps for obtaining the surface wall evolution association rules are: Based on the maximum displacement value of the wall, the maximum displacement values ​​of the wall are aggregated according to the time sequence of the excavation process, and a one-to-one correspondence is established between the width characteristics of the settlement trough corresponding to each process and the maximum displacement value of the wall, so as to obtain combined data of the excavation process, the maximum displacement value of the wall and the width characteristics of the settlement trough; Calculating a comprehensive factor of wall settlement evolution based on the combined data of the excavation process, the maximum displacement value of the wall and the width characteristics of the settlement trough; Based on the comprehensive factor of wall settlement evolution, the comprehensive factor of wall settlement evolution is correlated with the excavation process number, the maximum displacement value of the wall and the width characteristics of the settlement trough, and the combination representing the relationship between the wall deformation and the settlement trough change is extracted to form the surface wall evolution association rules.

[0011] Preferably, the steps of obtaining the expected wall response mode are: Based on the width feature of the settling tank, match the width feature of the settling tank with the historical width features recorded in the correlation rules of the surface wall evolution. Check one by one whether the width range of the settling tank in each historical record covers the current width feature of the settling tank, screen the historical records that meet the matching conditions, and extract the corresponding historical excavation process numbers, settling tank width ranges, and associated wall behavior descriptions to generate a set of historical correlation records; Based on the set of historical correlation records, parse the wall behavior descriptions in each historical correlation record one by one, extract the associated wall displacement change trend, displacement change rate, and key deformation nodes in the description. Group the records of the same type of wall displacement change trend, displacement change rate, and key deformation nodes into the same group, and summarize the wall behavior descriptions with the highest frequency of occurrence in each group to obtain the main wall behavior pattern groups; Based on the main wall behavior pattern groups, match and verify the main wall behavior pattern groups corresponding to the current settling tank width feature with the real-time monitoring data. By comparing whether the displacement change trend, displacement change rate, and key deformation nodes are consistent, screen the behavior descriptions that meet the current working conditions to generate the expected wall response pattern.

[0012] Preferably, the steps for obtaining the prediction result of the lateral deformation of the wall are as follows: Based on the expected wall response pattern, perform data matching between the expected wall response pattern and the deformation state of the current deep foundation pit wall, and extract the displacement increment, acceleration, and time change rate of the current deep foundation pit wall to generate a set of parameters required for predicting the current deep foundation pit wall; According to the set of parameters required for predicting the current deep foundation pit wall, calculate the predicted value of the lateral deformation of the deep foundation pit wall within the next time window; Based on the predicted value of the lateral deformation of the deep foundation pit wall within the next time window, compare the predicted value item by item with the historical deformation threshold, real-time monitoring data, and expected wall response pattern to obtain the prediction result of the lateral deformation of the wall.

[0013] The present invention provides a method for predicting the lateral deformation of a deep foundation pit wall, including the following steps: Collect the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points of the deep foundation pit wall, perform time series alignment processing and abnormal monitoring value elimination on the data points, and establish calibrated deformation sequence data; Based on the surface settlement monitoring values in the calibrated deformation sequence data, generate surface settlement isograms by spatial interpolation to obtain a surface settlement isogram map. Based on the surface settlement isogram map, measure the width of the settling tank formed in the key settlement area to establish a width feature of the settling tank; Based on the lateral displacement monitoring values of the wall in the calibrated deformation sequence data, extract the maximum lateral displacement values of each monitoring section at the same time section to obtain the maximum displacement value of the wall. Based on the maximum displacement value of the wall, analyze the corresponding relationship between the evolution with the excavation process and the change of the settlement trough width characteristics, and establish the surface-wall evolution association rules; Based on the settlement trough width characteristics, query the wall behavior pattern corresponding to the width characteristics in the surface-wall evolution association rules to obtain the expected wall response pattern. Based on the expected wall response pattern, combine the current deformation state of the deep foundation pit wall to predict the lateral deformation value of the deep foundation pit wall within the next time window, and obtain the prediction result of the lateral deformation of the wall.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In the present invention, by collecting in real time and strictly performing data time series alignment processing, the consistency of the surface settlement data and the lateral displacement data of the wall is ensured, and abnormal monitoring data is effectively eliminated, improving the data accuracy and consistency; during the spatial interpolation process, the width measurement of the settlement area is introduced to determine the settlement trough width characteristics, and the surface deformation trend is captured by deformation indicators; the maximum lateral displacement of each time section of the monitoring section is extracted, and the dynamic correspondence between the maximum displacement and the settlement trough width is established to identify the regular influence of the construction process change on the lateral deformation; through the matching of the settlement trough width characteristics and the wall response pattern, the accuracy and adaptability of the predicted lateral deformation value are dynamically improved; by combining the current deformation state and the prediction mode for comparative analysis, the accurate prediction result is corrected in real time, enhancing the ability to perceive the deformation trend in advance and pre-control risks, and reducing the risk hidden dangers during the construction stage. Description of the Drawings

[0015] Figure 1 It is the system flow chart of the present invention. Detailed Embodiment

[0016] In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention and are not used to limit the present invention.

[0017] Please refer to Figure 1 , the present invention provides a technical solution: a lateral deformation prediction system based on a deep foundation pit wall includes: A deformation data acquisition module, which acquires the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points of the deep foundation pit wall, performs data point time series alignment processing and abnormal monitoring value elimination, and establishes calibrated deformation sequence data; The ground surface pattern recognition module, based on the ground settlement monitoring values in the calibrated deformation sequence data, generates ground settlement isograms through spatial interpolation to obtain a ground settlement isogram map. Based on the ground settlement isogram map, it measures the width of the settlement trough formed by the key settlement areas and establishes the settlement trough width feature; The wall response correlation module, based on the lateral displacement monitoring values of the wall in the calibrated deformation sequence data, extracts the maximum lateral displacement values of each monitoring section at the same time section to obtain the maximum wall displacement value. Based on the maximum wall displacement value, it analyzes the corresponding relationship between the evolution with the excavation process and the change of the settlement trough width feature, and establishes the ground-wall evolution correlation rule; The deformation trend prediction module, based on the settlement trough width feature, queries the wall behavior pattern corresponding to the width feature in the ground-wall evolution correlation rule to obtain the expected wall response pattern. Based on the expected wall response pattern, combined with the current deformation state of the deep foundation pit wall, it predicts the lateral deformation value of the deep foundation pit wall in the next time window and obtains the wall lateral deformation prediction result.

[0018] The steps for obtaining the calibrated deformation sequence data are as follows: Collect the ground settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points of the deep foundation pit wall, match the ground settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points of the deep foundation pit wall one by one according to the collection time, and mark and eliminate the unmatched monitoring values to generate the ground settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points of the deep foundation pit wall after time series alignment and outlier elimination; Based on the ground settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points of the deep foundation pit wall after time series alignment and outlier elimination, group the ground settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points of the deep foundation pit wall according to the measuring point numbers, and check the data integrity within each group to eliminate the incomplete data groups, and generate the data pairs grouped according to the measuring point numbers and ensuring integrity; Based on the data pairs grouped according to the measuring point numbers and ensuring integrity, sort the ground settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points of each group of data pairs to form the calibrated deformation sequence data.

[0019] Specifically, when collecting the ground settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points of the deep foundation pit wall, first, match these two types of monitoring data one by one according to the collection time. The specific operation is to traverse each ground settlement monitoring value, record its collection timestamp, and then search for the record with the same timestamp in the lateral displacement monitoring values of the corresponding measuring points of the deep foundation pit wall. Considering the possible slight time asynchrony in actual monitoring, a time matching tolerance is set, for example Seconds. That is, if the acquisition time difference between two monitoring values is within this tolerance range, it is considered a successful match. Mark the monitoring values for which no corresponding timestamp match can be found within this tolerance. For example, add a "not matched" label. Subsequently, remove all the monitoring data points marked as "not matched" from their respective data sets. Next, perform the removal of abnormal monitoring values for the time-aligned data pairs that have been successfully matched. Here, the abnormal values mainly refer to the abnormalities in the data values themselves. The method based on the interquartile range (IQR) is used for judgment. For the surface settlement monitoring value sequence, first calculate its first quartile (Q1) and third quartile (Q3) to obtain the interquartile range , and set the boundary parameter for abnormal value judgment to 1.5. This value is a commonly used standard in statistics to identify moderately abnormal points. For example, if Q1 of a batch of surface settlement monitoring values is 5 mm and Q3 is 12 mm, then the IQR is 7 mm. Therefore, the effective data range is mm to mm. Any surface settlement monitoring value (and its paired wall lateral displacement value) that exceeds this range will be considered abnormal and removed. Similarly, the same abnormal value removal process is performed on the lateral displacement monitoring value sequence of the corresponding measuring points of the deep foundation pit wall. Calculate its interquartile range , and use the same boundary parameter of 1.5 (this parameter can be adjusted according to specific engineering experience and data characteristics. For example, for data with large fluctuations, it can be appropriately relaxed to 2.0). Set the normal range. For example, if Q1 of a batch of wall lateral displacement monitoring values is 3 mm and Q3 is 10 mm, then the IQR is 7 mm. The effective data range is mm to mm. The wall lateral displacement monitoring values (and their paired surface settlement values) that exceed this range are also removed. After the above time matching and two-way numerical abnormal point removal, the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points of the deep foundation pit wall after time series alignment and abnormal value removal are generated.

[0020] Based on the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points of the deep foundation pit wall after time series alignment and abnormal value removal, group these data according to the measuring point numbers. Specifically, when implementing, a data structure can be constructed. For example, use the measuring point number as the key, and its value is a dictionary containing a list of all valid (i.e., paired) surface settlement monitoring values and wall lateral displacement monitoring values of that measuring point. Traverse all the cleaned data points, and according to the measuring point number recorded in each data point, assign it to the list of the corresponding key. After grouping, check the data integrity within each group to remove incomplete data groups. The standard for checking data integrity is based on a preset minimum number threshold of valid data point pairs , the setting of this threshold generally considers the total monitoring duration, monitoring frequency, and the basic requirements of the analysis model for the amount of data. For example, if the monitoring period of a deep foundation pit project is 60 days and the monitoring equipment automatically collects data 4 times a day (once every 6 hours), then theoretically each measuring point should have data pairs. Set a data integrity percentage , such as 70%. This percentage is determined based on engineering experience and aims to ensure that the amount of data used for analysis is sufficient to reflect the statistical characteristics of the deformation trend. Data sequences below this standard may not provide a reliable analysis basis due to too few data points. Then the minimum number of valid data pair thresholds data pairs. During the inspection process, calculate the actual number of data pairs included in each group of measuring point numbers . If , then the data group corresponding to this measuring point number is determined to be an incomplete data group and removed from the dataset for subsequent analysis. For example, if measuring point T001 has 180 data pairs ( ), it is retained. If measuring point T002 has 150 data pairs ( ), it is marked as incomplete and removed. After this step is completed, data pairs grouped by measuring point number and ensuring integrity are generated.

[0021] Based on the data pairs grouped by measuring point number and ensuring integrity, next, sort the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points on the deep foundation pit wall within each group of data pairs. The core basis for this sorting is the acquisition timestamp of the monitoring data pairs, with the aim of ensuring that the data for each measuring point forms a time series arranged strictly in chronological order, laying a foundation for subsequent trend analysis and pattern recognition. Specifically, for each group of measuring point numbers that have passed the integrity check, extract all the data pairs it contains (each data pair consists of a surface settlement monitoring value, a wall lateral displacement monitoring value, and the corresponding acquisition timestamp). Then, within each group, apply a standard sorting algorithm, such as quicksort or mergesort, using the acquisition timestamp as the sorting key for ascending order, that is, from the earliest monitoring time point to the latest monitoring time point. If there are two data pairs with exactly the same timestamp (although this situation is extremely rare after the previous processing, it may still occur in some low-precision time recording or specific concurrent acquisition scenarios), a secondary sorting rule can be set, such as sorting according to the magnitude of the surface settlement monitoring value or the magnitude of the wall lateral displacement monitoring value, or simply maintaining their relative order in the original data. However, the main goal is to ensure strict chronological order. After completing this sorting process, each measuring point number corresponds to one or more time series of surface settlement monitoring values and wall lateral displacement monitoring values, and the data points in these series are strictly arranged in chronological order, thus forming calibrated deformation sequence data.

[0022] The steps for obtaining the ground settlement contour map are as follows: Based on the ground settlement monitoring values in the calibrated deformation sequence data, sort the ground settlement monitoring values according to the spatial coordinates of the measuring points, calculate the difference in ground settlement monitoring values between adjacent measuring points and the distance measurement of each measuring point, and generate the adjacent measuring point settlement difference and distance measurement data; According to the adjacent measuring point settlement difference and distance measurement data, calculate the ground settlement contour interpolation factor. The calculation formula is: ; wherein, is the ground settlement contour interpolation factor of the measuring point, is the ground settlement monitoring value of the measuring point, is the ground settlement monitoring value of the measuring point, is the horizontal distance between the measuring point and the measuring point, is the historical settlement average value of the measuring point, is the historical settlement average value of the measuring point; Based on the ground settlement contour interpolation factor, fit the ground settlement monitoring values through spatial interpolation into a grid distribution. During the grid process, adjust the interpolation density of each grid cell according to the ground settlement contour interpolation factor to form the ground settlement contour map.

[0023] Specifically, based on the ground settlement monitoring values in the calibrated deformation sequence data, first record and organize each ground settlement monitoring value according to its corresponding three-dimensional spatial coordinates of the measuring point (for example, X, Y, Z coordinates, where X and Y are plane coordinates and Z is the elevation) for spatial proximity analysis. Next, determine the "adjacent" relationship between each measuring point. A clear method is to set a spatial search radius for each measuring point. This search radius is determined according to the actual measuring point layout density of the construction site. For example, if the average distance between measuring points is 10 meters, can be set to 15 meters to ensure that the most directly affected surrounding measuring points are included. Traverse all measuring points. For the currently processed measuring point , calculate the spatial straight-line distance between it and all other measuring points . If and , then the measuring point is regarded as the measuring point For an adjacent measurement point, after identifying the adjacent measurement points, for each pair of determined adjacent measurement points ( ), calculate the difference in the surface settlement monitoring values between them, that is , where and are the surface settlement monitoring values obtained from the calibrated deformation sequence data at the same time section for measurement points and measurement point respectively. At the same time, calculate the horizontal distance between these two adjacent measurement points, that is, the distance measurement . Store these calculated differences and the corresponding horizontal distances as well as the identification of the measurement point pairs participating in the calculation. For example, if the coordinates of measurement point A are (10, 20, 5) and the settlement value is 15 mm, and the coordinates of measurement point B are (10, 30, 4.8) and the settlement value is 12 mm, and the set is 15 m, then the spatial distance between them is approximately 10.02 m, less than 15 m, considered adjacent, and the settlement difference between them is mm, and the horizontal distance measurement is m. After processing all measurement point pairs in this way, generate a set containing the settlement differences and corresponding distance measurements of each adjacent measurement point pair, that is, the settlement differences and distance measurement data of adjacent measurement points.

[0024] Formula: , the advantage of the formula is that the interpolation factor of the surface settlement contour line takes into account the instantaneous settlement differences between the current measurement point and its adjacent measurement points, their spatial distances, and the similarity of their historical settlement behaviors. By introducing the square term of the difference in historical settlement means and placing it inside the square root in the denominator, the contribution weight of adjacent point pairs with large differences in historical settlement behaviors to the current interpolation factor is reduced. Conversely, the weight of point pairs with similar behaviors is relatively increased. This design can reflect the complexity and differences of local surface deformation, avoid the smoothing effect that may be brought by simple distance or settlement difference weighting, make the identified settlement change area more accurate, and the use of absolute values ensures that all difference contributions are positive. Finally, dividing by the total number of adjacent points achieves normalized averaging, making the factors between measurement points with different neighborhood numbers comparable, which helps to more accurately adjust the density of subsequent grid interpolation, thereby improving the accuracy and expression ability of the surface settlement contour map; is the surface settlement monitoring value of the th measurement point. This value is derived from the "calibrated deformation sequence data" generated in the previous steps of this application, specifically referring to the measurement point The surface settlement readings after data cleaning and calibration. For example, at 3:00 PM on May 8, 2025, by referring to the "Calibrated Deformation Sequence Data", the surface settlement monitoring value of the measuring point numbered CP001 (i.e., the th measuring point) is 22.5 mm; For the surface settlement monitoring value of the th measuring point, this value is also obtained from the "Calibrated Deformation Sequence Data" and represents the surface settlement reading of the measuring point adjacent to the measuring point at the same analysis time. The measuring point is one of the adjacent measuring points of the determined measuring point . For example, for the above-mentioned measuring point CP001, the surface settlement monitoring value of one of its adjacent measuring points CP002 (i.e., the th measuring point) at the same time is 18.0 mm; For the horizontal distance between the th measuring point and the th measuring point, this data is obtained when generating the "Adjacent Measuring Point Settlement Difference and Distance Measurement Data" by calculation. It is calculated according to the plane coordinates (such as X and Y coordinates) of the measuring point and the measuring point in the engineering site coordinate system using the distance formula between two points. For example, the plane coordinates of the measuring point CP001 are (50.5 m, 100.2 m), and the plane coordinates of the adjacent measuring point CP002 are (50.5 m, 110.2 m), then the horizontal distance between them m; For the historical settlement mean value of the th measuring point, this value is the result of arithmetic averaging of all the surface settlement monitoring values in the "Calibrated Deformation Sequence Data" of the measuring point since the start of monitoring until before the current analysis time. It reflects the average settlement level of this measuring point over a historical period. For example, the measuring point CP001 has 180 valid settlement records in the past 90 days, and the sum of these records is 2700 mm, then its historical settlement mean value mm; For the historical settlement mean value of the th measuring point, its acquisition method is exactly the same as , except that it is for the adjacent measuring point , and the historical settlement data of the measuring point is used for calculation. For example, the adjacent measuring point CP002 has 175 valid settlement records in the past 90 days, and the sum of its records is 2100 mm, then its historical settlement mean value mm; is the total number of adjacent measuring points, which is obtained by counting when determining adjacent measuring points, representing the number of measuring points that meet the proximity condition (for example, within a preset range). For example, for measuring point CP001, after proximity judgment, a total of 3 adjacent measuring points (CP002, CP003, CP004) are found, then ; ; Calculation process: Taking measuring point CP001 as an example, assume there are three adjacent measuring points CP002, CP003, and CP004, that is . The specific values of each parameter are as follows: For CP001 ( ): mm, mm.

[0025] Substituting into the formula for calculation gives 17.78. This result indicates that the interpolation factor of the ground settlement contour line for measuring point CP001 is approximately 17.78. This value comprehensively reflects an average measure of the difference in current settlement, spatial distance, and historical settlement behavior of measuring point CP001 relative to its surrounding points.

[0026] Based on the interpolation factor of the ground settlement contour line, in specific operations, first define a regular geographical grid covering the entire monitoring area. The initial size of the grid cells is initially set according to the scope of the monitoring area and the required accuracy of the contour map. For example, for a 100 m × 100 m area, a 1 m × 1 m grid cell can be initially set. Next, for the center point of each grid cell, use spatial interpolation techniques (such as inverse distance weighting method IDW or Kriging method) to estimate its ground settlement monitoring value. This interpolation process utilizes the ground settlement monitoring values in the original calibrated deformation sequence data and their corresponding measuring point spatial coordinates. The key point is that during the gridification process, it is necessary to dynamically adjust the interpolation density of each grid cell according to the interpolation factors of the ground settlement contour lines calculated in the previous step . Specifically, first interpolate the values of each measuring point onto the grid to obtain the values at the center point of each grid cell. Then set one or more threshold intervals. For example, define a low - impact area ( ), a medium - impact area ( ), and a high - impact area ( ). These thresholds are set based on the analysis of the statistical characteristics (such as mean, standard deviation) of the value distribution in the study area and combined with engineering experience. For grid cells in different impact areas, their interpolation density adjustment strategies are different. In the high - impact area, that is Regions with larger values indicate complex settlement changes or significant differences. In this case, a higher-density interpolation strategy can be adopted. For example, when performing the inverse distance weighting method, reduce the radius for searching neighboring monitoring points (e.g., from the default 50 meters to 25 meters) or increase the number of nearest neighbor points forcibly included in the calculation (e.g., from the default 5 points to 8 points), or locally densify the grid in this region (e.g., subdivide 1 m × 1 m cells into 0.5 m × 0.5 m cells for interpolation). Conversely, in low-impact areas, a relatively sparse interpolation strategy can be adopted or the initial settings can be maintained. After estimating the settlement values of all grid cells through this adaptive interpolation density adjustment based on the interpolation factor of the ground settlement isoline, the standard isoline generation algorithm can be used to draw the ground settlement isoline map based on these gridded settlement data.

[0027] The steps for obtaining the characteristics of the settlement trough width are as follows: Based on the ground settlement isoline map, identify the region where the settlement isolines are the densest in the ground settlement isoline map, and mark the outer edge coordinate points of the region to obtain the outer edge coordinate points of the key settlement region; Based on the outer edge coordinate points of the key settlement region, measure the horizontal distances between the outer edge coordinate points, and screen the representative distances according to the principle of the maximum lateral spacing to obtain the maximum horizontal width of the key settlement region; Based on the maximum horizontal width of the key settlement region, take the maximum horizontal width as the characteristic index of the settlement trough, and record the width value as the settlement trough width characteristic to form the settlement trough width characteristic.

[0028] Specifically, based on the ground settlement isoline map, it is first necessary to perform digital processing on the map and convert it into raster data or vector data that can be analyzed by a computer. If it is raster data, each pixel point corresponds to a settlement value. To identify the region where the settlement isolines are the densest, a specific method is to calculate the numerical magnitude of the ground settlement gradient, that is, calculate the settlement change rates in the X and Y directions at each grid point, and then synthesize the total gradient value at this point. For example, for the settlement value at the grid point , its X-direction gradient is approximately , and its Y-direction gradient is approximately , where and are the grid spacings, and the total gradient . The region with a larger gradient value corresponds to denser isolines. Set a gradient threshold to define the "densest" region. This threshold The setting can be based on the statistical analysis of the gradient values of all grid points in the entire monitoring area. For example, the 90th percentile of all gradient values can be taken, or it can be set as the average gradient value plus twice the standard deviation. If the calculated regional average gradient is 0.004 m / m and the standard deviation is 0.001 m / m, then can be set as m / m, and all grid points with gradient values greater than or equal to are initially identified as part of the dense area. Subsequently, clustering is performed on these point sets, and the upper-connected dense point groups are identified as independent dense areas. If there are multiple such independent dense areas, the area with the largest area or the area containing the point with the largest settlement value is selected as the "key settlement area". Once the key settlement area is determined, a boundary extraction algorithm, such as the Moore neighborhood tracking algorithm or the radial scanning algorithm, is used to track and record the planar coordinates (X, Y coordinates) of the grid points that form the outermost contour of this area. The set of these coordinate points is the outer edge coordinate points of the key settlement area.

[0029] Based on the outer edge coordinate points of the key settlement area, it is first necessary to clarify the definition of "lateral", which is usually relative to the deep foundation pit excavation direction or the main deformation direction. For example, if the foundation pit is excavated along the north-south direction, then "lateral" generally refers to the east-west direction. For example, if it is determined that the lateral is the X-axis direction of the site coordinate system, the process of measuring the horizontal distance between the outer edge coordinate points and screening the representative distance is as follows: Traverse each point in the set of outer edge coordinate points of the key settlement area , for each point , find its corresponding point in the "lateral" direction. A scanning line method can be used, that is, along multiple parallel scanning lines perpendicular to the "lateral" (for example, along the Y-axis direction), the key settlement area is cut at a certain interval (for example, set to 0.5 m according to the measurement point density or the desired accuracy). Each scanning line will generate two or more intersection points with the outer edge of the key settlement area. Record a pair of intersection points located on the outermost side of the area boundary on each scanning line and , and the horizontal distance between them is the lateral width at the position of this scanning line . Record the lateral widths calculated on all scanning lines, and screen them according to the "principle of maximum lateral spacing", that is, among all the calculated , select the largest value as the representative distance. For example, through the above scanning line method, widths of 8.5 m are obtained at m, widths of 9.2 m are obtained at m, and widths of 9.0 m are obtained at m. Then 9.2 m is the currently selected maximum lateral spacing, and the maximum horizontal width of the key settlement area is obtained.

[0030] Based on the maximum horizontal width of the critical settlement area, this specific value is directly used as a key geometric parameter characterizing the current ground settlement trough shape, that is, the characteristic index of the settlement trough. The previously calculated maximum horizontal width value is 9.2 meters. This width value is saved together with the corresponding monitoring date, time, and the identification information of the relevant ground settlement contour map. For example, it is recorded as: "Monitoring date: May 8, 2025, Ground settlement contour map number: SCM - 20250508 - 01, Maximum horizontal width of the critical settlement area: 9.2 meters". This recording method ensures the traceability and relevance of the data, enabling the accurate connection of the settlement trough width characteristics at a specific time point with other working condition parameters (such as excavation depth, wall displacement, etc.) during subsequent deformation trend analysis and prediction. The recorded width value is officially adopted as the settlement trough width characteristic at that moment, forming a data point that can be used for time series analysis or pattern matching, thus forming the settlement trough width characteristic.

[0031] The steps to obtain the maximum wall displacement value are as follows: Based on the lateral wall displacement monitoring values in the calibrated deformation sequence data, the lateral wall displacement monitoring values are grouped according to the monitoring profile number, and all the observed values at the same time section are extracted for the lateral wall displacement monitoring values within each group of monitoring profiles. By comparing the timestamps, the valid observed values within the time section are selected to obtain the valid lateral displacement monitoring values of each monitoring profile at the same time section. Based on the valid lateral displacement monitoring values of each monitoring profile at the same time section, each group of data is traversed in sequence and the magnitudes of the observed values within the group are compared to select the largest lateral displacement monitoring value in each group. At the same time, the corresponding monitoring profile number and observation time are recorded to obtain the set of maximum lateral displacement values of each monitoring profile. Based on the set of maximum lateral displacement values of each monitoring profile, the magnitudes of all the maximum lateral displacement values in the set are compared, and the highest group is selected as the representative value. At the same time, the monitoring profile number and observation time of the representative value are marked to form the maximum wall displacement value.

[0032] Specifically, based on the lateral wall displacement monitoring values in the calibration deformation sequence data, first, these monitoring data are grouped according to the monitoring profile numbers they belong to. For example, all monitoring data with a profile number of PM01 are grouped into one group, PM02 into another group, and so on. After grouping, for the data within each monitoring profile number group, all observed values representing a specific "same time section" need to be extracted. Here, the "same time section" refers to a target analysis time point, such as a fixed time of each day, like 10:00 am. Considering the possible small time deviations in data acquisition of each sensor in actual monitoring, an effective time window is set around this target time. The size of this window is determined according to the synchronization accuracy of the monitoring system and the data acquisition frequency. For example, if the data is collected once an hour, a 5-minute time window before and after the target time can be set. That is, if the target time is 10:00:00, the time window is from 09:55:00 to 10:05:00. Traverse all the lateral wall displacement monitoring values recorded within this time window for the specified monitoring profile (such as PM01). These values usually correspond to the sensor readings at different depths within the profile. If there are multiple readings for the same-depth sensor within the effective time window, select the reading with the time stamp closest to the target time as the effective observed value for that depth. Or, if the readings are very dense and the values are similar, the average value can also be taken, but usually, the single value closest to the target time is used as the standard to ensure that there is only one unique effective lateral displacement monitoring value for each monitoring depth at this time section. By operating in this way, the effective lateral displacement monitoring values of each monitoring profile at the target same time section are obtained.

[0033] Based on the effective lateral displacement monitoring values of each monitoring profile at the same time section, the data within each monitoring profile will be processed next. The specific operation is to sequentially access a set of effective lateral displacement monitoring values included in each monitoring profile (such as PM01, PM02, etc.) at this specified time section. This set of values represents the lateral displacement of the wall at different depths of the profile. For example, for the monitoring profile PM01, at the time section of 10:00 am on May 8, 2025, its effective lateral displacement monitoring values are: 5.3 mm at a depth of -2.0 m, 8.1 mm at a depth of -4.0 m, 11.5 mm at a depth of -6.0 m, and 9.2 mm at a depth of -8.0 m. Compare these values (5.3, 8.1, 11.5, 9.2) to find the maximum value. In this case, the maximum lateral displacement monitoring value of the PM01 profile at this moment is 11.5 mm, which occurs at a depth of -6.0 m. Record this maximum value, and at the same time record the monitoring profile number (PM01) to which it belongs and the observation time of this time section (10:00 am on May 8, 2025). Perform the same operation on all other monitoring profiles (PM02, PM03, etc.), that is, find the maximum value among all the effective lateral displacement monitoring values within their respective profiles at the same time section, and record the corresponding profile number, maximum displacement value, and observation time. After completing the traversal and maximum value extraction of all monitoring profiles, a set containing the maximum lateral displacement values of each monitoring profile at this moment and their relevant information is obtained, that is, the set of maximum lateral displacement values of each monitoring profile.

[0034] Based on the maximum lateral displacement value set of each monitoring section, the set contains the maximum wall lateral displacement readings of each monitoring section at the same specific time section. For example, the set may contain the following data: (monitoring section PM01, maximum displacement 11.5 mm, observation time 2025-05-08 10:00), (monitoring section PM02, maximum displacement 14.2 mm, observation time 2025-05-08 10:00), (monitoring section PM03, maximum displacement 10.8 mm, observation time 2025-05-08 10:00). The processing process is to traverse each record in this set and extract the maximum lateral displacement value of each monitoring section. (i.e. 11.5 mm, 14.2 mm, 10.8 mm), directly compare these values ​​to screen out the one with the highest value. In this example, 14.2 mm is the highest value among the maximum displacement values ​​of all sections. This 14.2 mm is selected as the representative maximum displacement value of the entire deep foundation pit wall at this time section. At the same time, the information associated with this highest value is fully recorded, including the monitoring section number of its source (PM02 here) and the exact observation time (2025-05-0810:00). The screened out highest value, its corresponding monitoring section number and observation time are combined together and recorded as a whole to form the maximum displacement value of the wall.

[0035] The steps for obtaining the association rules of surface wall evolution are as follows: Based on the maximum displacement value of the wall, the maximum displacement value of the wall is aggregated according to the time sequence of the excavation process, and a one-to-one correspondence is established between the width characteristics of the settlement trough corresponding to each process and the maximum displacement value of the wall, so as to obtain the combined data of the excavation process, the maximum displacement value of the wall and the width characteristics of the settlement trough; According to the combined data of excavation process, maximum displacement value of the wall and width characteristics of the settlement trough, the comprehensive factor of wall settlement evolution is calculated. The calculation formula is: ; in, For the Comprehensive factors of wall settlement evolution in each process, For the The maximum displacement value of the wall in each process, For the The maximum displacement value of the wall in each process, For the The width characteristics of the sedimentation tank in each process, For the The width characteristics of the sedimentation tank in each process, For the The duration of each process, For the The duration of each process, is the amount of monitoring data for the th process, is the total number of processes; Based on the comprehensive factor of wall settlement evolution, the comprehensive factor of wall settlement evolution is correlated with the excavation process number, the maximum wall displacement value, and the characteristics of the settlement trough width for correlation analysis, extracting the combination representing the relationship between wall deformation and settlement trough change to form the surface wall evolution association rule.

[0036] Specifically, based on the maximum wall displacement value, first obtain the detailed excavation process plan from the project construction management records. This plan should include the unique numbers of all excavation processes planned in the project, the estimated start time and end time of each process, and the main content of the process. For example, process S01 (excavation of the first layer of soil) is planned to start at 08:00 on May 1, 2025, and end at 17:00 on May 7, 2025; process S02 (excavation of the second layer of soil and application of the first support) is planned to start at 08:00 on May 8, 2025, and end at 17:00 on May 14, 2025, until the last planned process S , and then, the data sequence of "maximum wall displacement value" with timestamps obtained through the previous steps up to the current time point, and the data sequence of "settlement trough width characteristics" also with timestamps are accurately allocated to the time intervals of each completed or ongoing excavation process according to their respective observation times. For each completed excavation process , select the "maximum wall displacement value" observed at the end time point of this process (or the time closest to this end time point) as the representative maximum wall displacement value of this process, denoted as , and similarly, select the "settlement trough width characteristics" at the corresponding time as the representative settlement trough width characteristics of this process, denoted as , for example, for the completed process S01, if the finally determined maximum wall displacement value at its end (May 7, 2025) is 15.0 mm and the corresponding settlement trough width characteristic is 8.0 m, then the combined data entry for process S01 is (S01, 15.0 mm, 8.0 m). This process continues until all occurred processes have corresponding representative data, forming a combined data of excavation processes, maximum wall displacement values, and settlement trough width characteristics arranged in the order of process time.

[0037] Formula: , The benefit of this formula is that this comprehensive factor of wall settlement evolution is intended to quantitatively evaluate a specific The relative evolution intensity and data support degree of each process in the context of the entire project. In the numerator part of the formula, by accumulating the square root of the product of the maximum wall displacement change and the settlement trough width change of all (completed and planned) processes in the project (while considering the adjustment effect of the process duration ratio), an index representing the "total potential energy" or "total cumulative change" of the overall project deformation is constructed. When evaluating the th process, if the parameters of future processes are estimated values, then this "total potential energy" also contains an estimated component. The denominator part is composed of the logarithm of the total number of project processes and the amount of monitoring data for the current th process . This makes the value vary due to the difference in a specific process : The larger the amount of monitoring data , the larger the denominator, which will cause the to decrease, and vice versa. If the monitoring data for a certain process is less, then the risk or uncertainty corresponding to this process (manifested as a higher value) is relatively higher, or in other words, the "weight" or "sensitivity" of this process in the overall project evolution is amplified; is the maximum wall displacement value of the th process. This data is mainly extracted from the generated "combined data of excavation process, maximum wall displacement value and settlement trough width characteristics", representing the maximum lateral displacement value of the wall observed or determined at the end of the th excavation process (or at the representative time point determined for this process), with the unit of millimeters. For example, the actual maximum wall displacement value of process S01 ( ) is 15.0 millimeters, the actual maximum wall displacement value of process S02 ( ) is 25.0 millimeters, and the estimated maximum wall displacement value of the planned process S03 ( ) is 32.0 millimeters; is the maximum wall displacement value of the th process, and the acquisition method is the same as . For the first process (i.e., when ), represents the initial state before the foundation pit excavation, usually set to 0 millimeters. For example, when calculating the term of , millimeters, and when calculating the term of , millimeters; is the The width characteristics of the settling tank for each process are also mainly extracted from the generated combined data, representing the width of the settling tank at the end of the th process (or representative time point). For completed processes, it is the actual observed value, and for future processes, it is the estimated value. For example, the actual width of the settling tank for process S01 ( ) is 8.0 meters, and the actual width of the settling tank for process S02 ( ) is 10.5 meters. The estimated width of the settling tank for the planned process S03 ( ) is 12.5 meters; is the width characteristic of the settling tank for the th process, and the acquisition method is the same as . For the first process ( ), represents the width of the settling tank in the initial state, usually set to 0 meters. For example, when calculating the term of , meters, and when calculating the term of , meters; is the duration of the th process. This data is obtained from the project construction management records or the excavation process plan and is calculated by subtracting the start time from the planned or actual end time of the th process. For example, the actual duration of process S01 is 7 days, the actual duration of process S02 is 7 days, and the estimated duration of the planned process S03 is 6 days; is the duration of the th process, and the acquisition method is the same as . For the calculation term of the first process ( ), to ensure that , that is, is meaningful, it is agreed that when , the value of this ratio is 1, that is, the calculation result of the first process is not adjusted through this item. For example, when calculating the term of , days, and when calculating the term of , it is set that ; is the amount of monitoring data for the th process, referring to the corresponding to the calculation factor at the During the actual duration of each specific process, the total number of observations of the effective "maximum wall displacement value" and "settlement trough width characteristics" collected and used for analysis. For example, during process S02 (here, for example ), effective records of the maximum wall displacement value and settlement trough width characteristics are made daily for a total of 7 days. Then (for example, one representative observation value per day); is the total number of processes, which refers to the number of all independent processes planned for the entire deep foundation pit excavation project. For example, if a project is divided into 10 main excavation and support processes, then , in this example, the total number of project processes is set ; is the specific process number for which the comprehensive factor of wall settlement evolution is currently being calculated. For example, to calculate the factor for process S02, then ; Comprehensive factor of change . Known and estimated parameters: millimeters, meters for the actual process S01( ): millimeters, meters, days for the actual process S02( ): millimeters, meters, days, and the amount of monitoring data for this process Planned process S03( ): millimeters, meters, days; Substitute the parameters into the formula to calculate 3.823. This result indicates that in the context of a project with a total of 3 planned processes, when evaluating the 2nd process (and considering the estimated deformation of the 3rd process), the comprehensive factor of wall settlement evolution is approximately 3.823. This value reflects a relative relationship between the cumulative change intensity of the entire project (partially based on estimation) and the amount of monitoring data for the 2nd process, and is used for subsequent establishment of association rules.

[0038] Based on the comprehensive factor of wall settlement evolution, for each completed process The calculated comprehensive factor of wall settlement evolution value, together with the number of this process, the maximum wall displacement value actually observed at the end of this process and the actual settlement trough width characteristics (these and (from the actual observation part of the generated "Combined Data of Excavation Process, Maximum Wall Displacement Value and Settlement Trough Width Characteristics") are organized together to form a dataset containing (process number, , , ). Association analysis is performed on this dataset with the aim of mining potential regular combinations among these parameters that can characterize specific evolution behaviors. The analysis methods can include first classifying or discretizing continuous data (such as , , , and their change amounts compared to the previous process and ). For example, the value is divided into three levels of "low", "medium", and "high" according to its statistical distribution (such as quartiles). For example, if the lower quartile of all historical values is 1.5 and the upper quartile is 4.0, then is low, is medium, is high. Similarly, the wall displacement change amount can be divided into "small increase" (such as 0 - 2 mm), "medium increase" (such as 2 - 5 mm), and "large increase" (such as greater than 5 mm). These thresholds need to be set comprehensively based on engineering experience, design control standards, and the deformation characteristics of specific projects. For example, the threshold for "large increase" can refer to 30% to 50% of the design deformation warning value. After discretization, apply association rule mining algorithms such as Apriori or FP - Growth to find frequently occurring parameter level combinations (frequent itemsets) in the dataset, and generate association rules with high support and confidence from these frequent itemsets. For example, it may be found that the rule: "If the excavation process type is 'deep earth excavation' and its corresponding level is 'high', then the level of this process is 'large increase' and the probability that the level is'significantly expanded' is 80%". These verified and screened rule combinations constitute the surface wall evolution association rules.

[0039] The steps to obtain the expected wall response pattern are as follows: Based on the settlement trough width characteristics, match the settlement trough width characteristics with the historical width characteristics recorded in the surface wall evolution association rules. Check one by one whether the settlement trough width range of each historical record covers the current settlement trough width characteristics. Screen the historical records that meet the matching conditions, and extract the corresponding historical excavation process numbers, settlement trough width ranges, and associated wall behavior descriptions to generate a set of historical association records; Based on the set of historical association records, each description of the wall behavior in the historical association records is parsed one by one to extract the associated wall displacement change trend, displacement change rate, and key deformation nodes in the description. The records of the same type of wall displacement change trend, displacement change rate, and key deformation nodes are grouped into the same set, and the wall behavior descriptions with the highest frequency of occurrence in each set are summarized to obtain the main wall behavior pattern groups; Based on the main wall behavior pattern groups, the main wall behavior pattern groups corresponding to the current settlement trough width feature are matched and verified with the real-time monitoring data. By comparing whether the displacement change trend, displacement change rate, and key deformation nodes are consistent, the behavior descriptions that meet the current working conditions are screened to generate the expected wall response pattern.

[0040] Specifically, based on the settlement trough width feature, which is a specific value calculated for the current monitoring period, such as 9.5 meters, this current settlement trough width feature is matched with each historical record stored in the "surface-wall evolution association rule" library formed in the previous step. These historical records include historical excavation process numbers, historical settlement trough width ranges, and associated wall historical behavior descriptions. The core operation of the matching is to check one by one whether the "settlement trough width range" in the historical record "covers" the current settlement trough width feature. Specifically, if the settlement trough width range defined in a certain historical record A is [8.0 meters, 10.0 meters], then the current 9.5 meters is considered to be covered because it satisfies , so this historical record A is the record that meets the matching condition. On the contrary, if the width range in historical record B is [10.5 meters, 12.0 meters], then the current 9.5 meters is not within this range, and record B does not meet the matching condition. The system will screen out all historical records that meet this coverage condition and extract the structured information they contain, mainly including the "historical excavation process number" (for example, process S03) when the record was formed, the "settlement trough width range" corresponding to this historical record (for example, [8.0 meters, 10.0 meters]), and the most critical "associated wall behavior description" (for example, "The wall deforms uniformly at a rate of 0.3 mm / day in the middle, and the maximum displacement points are concentrated near the -8-meter depth of the CX-02 section"). Combining all the extracted information forms a temporary data set containing multiple matching historical situations, that is, the historical association record set.

[0041] Based on the historical association record set, perform structured parsing on the "wall behavior description" text contained in each historical record in the set. This parsing process aims to extract three key quantitative or classification indicators: the wall displacement change trend, the displacement change rate, and the key deformation nodes. The wall displacement change trend is classified into predefined trend types by analyzing keywords in the description, such as "accelerating", "constant speed", "decelerating", "tending to be stable", "increasing", "decreasing", etc. For example, "accelerating increase" (such as the average rate in the current few days exceeds 20% compared to the average rate in the previous cycle and the current rate is greater than 0.1 mm / day), "constant speed increase" (such as the rate is between 0.1 and 0.5 mm / day and the rate change in the recent 3 days does not exceed ±10%), "decelerating increase" (such as the rate is still increasing but the increase amount decreases by more than 20% compared to the previous cycle), "stable" (such as the rate remains within ±0.1 mm / day for more than 3 days), or "decreasing". The displacement change rate directly extracts the specific numerical value and unit from the text (for example, 0.3 mm / day), and can be classified into "low speed" (such as 0 to 0.2 mm / day), "medium speed" (such as 0.2 to 0.5 mm / day), "high speed" (such as greater than 0.5 mm / day) according to a preset threshold. These rate classification thresholds are set based on engineering experience and warning levels. For example, "high speed" may correspond to the early warning level. The key deformation nodes are determined based on the position information in the description, such as "wall top", "middle of the wall", "specific section number (CX - 02)", or "specific depth (-8 m)", and are standardized. For example, they are uniformly classified into "upper part", "middle part", "lower part", or specific section / depth intervals. After completing the parsing and feature extraction of all historical records, group the records with the same or similar (according to the predefined classification criteria) wall displacement change trends, the same level of displacement change rates, and the same type of key deformation node positions. For example, all records with the trend of "constant speed increase", the rate of "medium speed", and the node in the "middle of the wall" are grouped together. Within each formed group, count the frequency of occurrence of each original "wall behavior description" text, and select the "wall behavior description" with the highest frequency of occurrence in the group as the representative summary of the group, obtaining the main wall behavior pattern groups.

[0042] Based on the main group of wall behavior patterns, first, according to the current "settlement trough width feature", select the "main group of wall behavior patterns" that match it, that is, select those behavior pattern groups whose historical settlement trough width ranges cover the current width feature. Then, match and verify these selected behavior pattern groups with the "real-time monitoring data" of the deep foundation pit wall obtained through the sensor network. This real-time monitoring data needs to be processed through parsing similar to historical data to extract the current actual wall displacement change trend, the current actual displacement change rate, and the position of the current key deformation node. For example, if the real-time data shows that the wall is undergoing uniform displacement in the middle at a rate of 0.25 mm / day, then the current trend is "uniform increase", the rate is "medium speed" (for example, 0.25 mm / day falls within the medium speed range), and the node is "middle". The verification process is to compare one by one whether the trend, rate level, and node type defined by each "main wall behavior pattern" selected are "consistent" with these three features parsed from the current real-time data. The judgment criterion for "consistent" is: the current trend classification is exactly the same as the trend classification in the pattern, the level to which the current rate value belongs is the same as the rate level in the pattern (for example, the current 0.25 mm / day belongs to "medium speed", and the pattern also defines "medium speed"), and the position classification of the current key deformation node is the same as the node position classification in the pattern. Only when all three aspects of features match, this "main wall behavior pattern" is considered consistent with the current working condition. From all the behavior patterns determined to be consistent, select the corresponding original "wall behavior description" text. If multiple behavior patterns are determined to be consistent, they can be sorted according to the total frequency of their appearance in historical data or the confidence level of their association rules, and select the "wall behavior description" text corresponding to the behavior pattern with the highest ranking as the current most likely wall response state to generate the expected wall response pattern.

[0043] The steps to obtain the prediction result of the lateral deformation of the wall are as follows: Based on the expected wall response pattern, match the expected wall response pattern with the deformation state of the current deep foundation pit wall, extract the displacement increment, acceleration, and time change rate of the current deep foundation pit wall, and generate the parameter set required for predicting the current deep foundation pit wall; According to the parameter set required for predicting the current deep foundation pit wall, calculate the predicted value of the lateral deformation of the deep foundation pit wall within the next time window. The calculation formula is: ; Among them, is the duration of the next time window, is the predicted value of the lateral deformation of the deep foundation pit wall within the next time window, is the displacement increment of the th monitoring point, The acceleration of each monitoring point, is the rate of change of time for the th monitoring point, and is the reference speed, which is the average lateral speed in the most recent month, is the expected response complexity parameter for the current process, is the total number of monitoring points;

[0044] Specifically, based on the expected wall response pattern, which is a text description of the most likely behavior of the wall under the current working conditions determined in the previous steps, such as "Accelerated displacement will occur in the middle of the wall, with a rate of about 0.8 mm / day, mainly affecting Section CX02", first conduct a macroscopic compliance check of this expected wall response pattern with the overall deformation state of the current deep foundation pit wall obtained from the real-time monitoring system to confirm that the current actual deformation trend is generally consistent with the description of the expected pattern. If there are significant deviations, such as acceleration expected but deceleration actually occurring, it may be necessary to re-evaluate the applicability of the expected pattern or mark potential unknown influencing factors. After confirming basic compliance, for each key monitoring point (e.g., the measuring points in all active inclinometers) that constitutes the current wall deformation state, extract the dynamic parameters required for the next-stage prediction from its latest "calibrated deformation sequence data". The specific extraction content includes: the displacement increment of each monitoring point within the most recent observation period (e.g., the past 24 hours, this period is set according to the monitoring frequency and prediction requirements, denoted as ), obtained by subtracting the previous displacement value from the current latest displacement value. For example, if the current displacement of a measuring point is 20.5 mm and it was 19.5 mm 24 hours ago, then its is 1.0 mm. Secondly, it is the current acceleration of each monitoring point , obtained by analyzing the displacement data of this point in at least the most recent three consecutive observation periods. For example, if the displacements of a measuring point in the past three 6-hour intervals are 19.0 mm, 19.5 mm, and 20.5 mm respectively, then its acceleration can be calculated accordingly. Furthermore, it is the current rate of change of time (i.e., the instantaneous speed) of each monitoring point, usually obtained by dividing the displacement increment in the most recent observation period by the duration of this observation period, that is , for example, if mm and days, then mm / day. These parameters of all monitoring points ( ​)Aggregate them to generate the parameter set required for predicting the current deep foundation pit wall.

[0045] Formula: , the benefit of the formula is that the prediction model combines the instantaneous dynamic parameters (displacement increment, acceleration, velocity) of multiple monitoring points to predict the overall lateral deformation in the next time window. By accumulating and averaging the influence factors ( ), it can reflect the overall trend of wall deformation rather than a single most dangerous point. The use of absolute values ensures that the contributions of both accelerating and decelerating deformations are regarded as the magnitudes of positive influence factors. The introduction of the reference velocity normalizes the product of the dynamic parameters of each monitoring point, reducing the influence caused by differences in velocity magnitudes and making it more focused on the relative intensity of dynamic behavior. The expected response complexity parameter allows adjusting the prediction sensitivity according to the current process and the predicted deformation mode, making the model have a certain engineering adaptability. Finally, multiplying by the prediction duration converts the rate or acceleration effect into a displacement amount, providing a quantitative prediction basis for engineering decisions; is the duration of the next time window, which refers to the time length for predicting future deformations and is jointly determined by engineering management requirements and the monitoring feedback cycle. For example, during the rapid excavation stage, daily prediction may be required, then is set to 1 day. During the maintenance stage when the deformation tends to be stable, it may be adjusted to weekly prediction, then is set to 7 days. This parameter directly affects the magnitude of the predicted displacement. For example, here it is set days; is the displacement increment of the th monitoring point. This data is obtained from the generated "parameter set required for predicting the current deep foundation pit wall" and represents the actual lateral displacement change amount of the th monitoring point in the most recent observation cycle (such as 24 hours). For example, for monitoring point k = 1, its displacement increment in the past 24 hours is 1.2 mm; for monitoring point k = 2, is 0.8 mm; is the acceleration of the th monitoring point. This data is also obtained from the "parameter set required for predicting the current deep foundation pit wall" and is the instantaneous lateral displacement acceleration calculated by analyzing the recent series (at least three) of displacement readings of the th monitoring point. For example, the acceleration of monitoring point k = 1 is 0.15 mm / day²; the acceleration is -0.05 mm / day² (indicating deceleration); is the time change rate of the th monitoring point, that is, its current lateral displacement speed, which is obtained from the "set of parameters required for predicting the current deep foundation pit wall" and is usually calculated by dividing the displacement increment in the most recent observation period by the duration of this observation period to get, that is , for example, if is 1 day, then the time change rate of monitoring point k = 1; the time change rate of monitoring point k = 2; is the reference speed, representing the average deformation rate performance of the foundation pit wall over a relatively long period of time. Specifically, it is obtained by retrieving the lateral displacement data of all relevant monitoring points in the most recent month (30 days) from the "calibrated deformation sequence data", calculating the average daily displacement rate of each monitoring point within these 30 days, and then averaging these daily rates of all monitoring points again. For example, after calculation, the average lateral speed in the most recent month is 1.0 mm / day; is the expected response complexity parameter of the current process. This parameter is an adjustment coefficient set according to the current construction process type and the deformation complexity interpreted from the "expected wall response mode", and its value is obtained by referring to the predefined "process-mode-complexity parameter mapping table", which is established based on historical engineering data and expert experience. For example, the corresponding to "shallow excavation-uniform settlement mode" is 0.9, and the corresponding to "deep excavation-differential settlement acceleration mode" is 1.3. For example, currently in the "main structure construction" stage, the expected response mode is "the wall deformation slows down and there are local fine-tunings", and by looking up the table, ; is the total number of monitoring points participating in this prediction calculation, referring to the total number of effective wall lateral displacement monitoring points arranged in the current deep foundation pit project. For example, if there are 50 effective wall lateral displacement monitoring points participating in the calculation in the project, then ; Calculation process: Set days, , mm / day, . Monitoring point 1: mm, mm / day², mm / day. Monitoring point 2: mm, mm / day², mm / day

[0046] Calculate the contribution terms of each monitoring point : For monitoring point 1 ( ): ; For monitoring point 2 ( ): ; Calculate the summation term : ; Calculate the predicted value of the lateral deformation of the deep foundation pit wall within the next time window : ; The result shows that, based on the current monitoring data and the selected parameters, it is predicted that within the next 1 day, the lateral deformation increment of the deep foundation pit wall (on average) will be approximately 0.1054 mm. This predicted value is a comprehensive average increment used to evaluate the overall deformation trend of the wall within the next time window.

[0047] Based on the predicted value of the lateral deformation of the deep foundation pit wall within the next time window (for example, the calculated mm), add this predicted displacement increment to the cumulative displacement value of the current wall to obtain the estimated total cumulative displacement , and then compare this estimated total cumulative displacement as well as the average deformation rate within the next time window deduced from item by item with the preset "historical deformation thresholds". These historical deformation thresholds include multiple levels, setting the monitoring alarm values, orange warning values, and red warning values for displacement and rate. For example, if the monitoring alarm value is 25 mm for cumulative displacement, the orange warning is 30 mm, and the current is 24.9 mm, then mm, which means the prediction result has triggered the monitoring alarm value. At the same time, check the consistency of the predicted deformation trend (for example, judging whether it is continuous deformation, accelerating deformation, or decelerating deformation based on the sign and magnitude of ) with the current actual deformation trend reflected by the "real-time monitoring data", and check the compliance with the deformation behavior described in the "expected wall response mode" obtained in the previous step. For example, if the expected response mode is "deformation slows down", but the predicted value still shows an obvious increment, then this difference needs to be concerned. Combining these comparison results, including whether the predicted value exceeds the limit, the coincidence degree of the predicted trend with the actual and expected modes, finally form a comprehensive evaluation and description of the safety state of the wall in the next stage, that is, the prediction result of the lateral deformation of the wall.

Claims

1. A lateral deformation prediction system based on the deep foundation pit wall, characterized in that The system includes: A deformation data acquisition module that acquires the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points on the deep foundation pit wall, performs time-series alignment processing of data points and elimination of abnormal monitoring values, and establishes calibrated deformation sequence data; A surface pattern recognition module that, based on the surface settlement monitoring values in the calibrated deformation sequence data, generates surface settlement isograms through spatial interpolation to obtain a surface settlement isogram map, and based on the surface settlement isogram map, measures the width of the settlement trough formed by the key settlement areas to establish settlement trough width characteristics; A wall response correlation module that, based on the lateral displacement monitoring values of the wall in the calibrated deformation sequence data, extracts the maximum lateral displacement values of each monitoring section at the same time section to obtain the maximum wall displacement values, and based on the maximum wall displacement values, analyzes the corresponding relationship between the evolution with the excavation process and the change in the settlement trough width characteristics to establish surface-wall evolution correlation rules; A deformation trend prediction module that, based on the settlement trough width characteristics, queries the wall behavior pattern corresponding to the width characteristics in the surface-wall evolution correlation rules to obtain the expected wall response pattern, and based on the expected wall response pattern, combines the current deformation state of the deep foundation pit wall to predict the lateral deformation value of the deep foundation pit wall within the next time window to obtain the wall lateral deformation prediction result.

2. The lateral deformation prediction system based on the deep foundation pit wall according to claim 1, wherein The steps for obtaining the calibrated deformation sequence data are as follows: Acquire the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points on the deep foundation pit wall, match the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points on the deep foundation pit wall one by one according to the acquisition time, and mark and eliminate the unmatched monitoring values to generate the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points on the deep foundation pit wall after time-series alignment and outlier elimination; Based on the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points on the deep foundation pit wall after time-series alignment and outlier elimination, group the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points on the deep foundation pit wall according to the measuring point numbers, and check the data integrity within each group to eliminate incomplete data groups to generate data pairs grouped according to the measuring point numbers and ensuring integrity; Based on the data pairs grouped according to the measuring point numbers and ensuring integrity, sort the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points on the deep foundation pit wall in each group to form calibrated deformation sequence data.

3. The lateral deformation prediction system based on the deep foundation pit wall according to claim 1, wherein The steps for obtaining the surface settlement isogram map are as follows: Based on the surface settlement monitoring values in the calibrated deformation sequence data, sort the surface settlement monitoring values according to the spatial coordinates of the measuring points, and calculate the difference in the surface settlement monitoring values between adjacent measuring points and the distance measurement of each measuring point to generate adjacent measuring point settlement difference and distance measurement data; Calculate the surface settlement isogram interpolation factor according to the adjacent measuring point settlement difference and distance measurement data; Based on the surface settlement isogram interpolation factor, perform grid distribution fitting on the surface settlement monitoring values through spatial interpolation, and adjust the interpolation density of each grid unit according to the surface settlement isogram interpolation factor during the gridification process to form a surface settlement isogram map.

4. The lateral deformation prediction system based on the deep foundation pit wall according to claim 1, wherein, The steps for obtaining the settlement trough width characteristics are as follows: Based on the surface settlement contour map, identifying the area with the densest settlement contour lines in the surface settlement contour map, and marking the outer edge coordinate points of the area to obtain the outer edge coordinate points of the key settlement area; Based on the outer edge coordinate points of the key settlement area, the horizontal distance between the outer edge coordinate points is measured, and the representative distance is selected according to the principle of maximum horizontal spacing to obtain the maximum horizontal width of the key settlement area; Based on the maximum horizontal width of the key sedimentation area, the maximum horizontal width is used as a characteristic indicator of the sedimentation tank, and the width value is recorded as the sedimentation tank width feature to form the sedimentation tank width feature.

5. The lateral deformation prediction system based on the deep foundation pit wall according to claim 1, characterized in that The steps for obtaining the maximum displacement value of the wall are: Based on the wall lateral displacement monitoring values ​​in the calibration deformation sequence data, the wall lateral displacement monitoring values ​​are grouped according to the monitoring section numbers, and all observation values ​​of the same time section are extracted for the wall lateral displacement monitoring values ​​in each group of monitoring sections, and the valid observation values ​​in the time section are selected by comparing the timestamps to obtain the valid lateral displacement monitoring values ​​of each monitoring section at the same time section; Based on the effective lateral displacement monitoring values ​​of each monitoring section at the same time section, traverse each group of data in turn and compare the size of each observation value in the group, select the lateral displacement monitoring value with the largest value in each group, and record the corresponding monitoring section number and observation time at the same time to obtain the maximum lateral displacement value set of each monitoring section; Based on the maximum lateral displacement value set of each monitoring section, the sizes of all the maximum lateral displacement values ​​in the set are compared, and the highest group is selected as the representative value. At the same time, the monitoring section number and observation time of the representative value are marked to form the maximum displacement value of the wall.

6. The lateral deformation prediction system based on the deep foundation pit wall according to claim 1, wherein, The steps for obtaining the surface wall evolution association rules are as follows: Based on the maximum displacement value of the wall, the maximum displacement values ​​of the wall are aggregated according to the time sequence of the excavation process, and a one-to-one correspondence is established between the width characteristics of the settlement trough corresponding to each process and the maximum displacement value of the wall, so as to obtain combined data of the excavation process, the maximum displacement value of the wall and the width characteristics of the settlement trough; Calculating a comprehensive factor of wall settlement evolution based on the combined data of the excavation process, the maximum displacement value of the wall and the width characteristics of the settlement trough; Based on the comprehensive factor of wall settlement evolution, the comprehensive factor of wall settlement evolution is correlated with the excavation process number, the maximum displacement value of the wall and the width characteristics of the settlement trough, and the combination representing the relationship between the wall deformation and the settlement trough change is extracted to form the surface wall evolution association rules.

7. The lateral deformation prediction system based on the deep foundation pit wall according to claim 1, characterized in that, The steps for obtaining the expected wall response mode are: Based on the settlement trough width feature, the settlement trough width feature is matched with the historical width feature recorded in the surface wall evolution association rule, and each historical record is checked one by one to see whether the settlement trough width range covers the current settlement trough width feature, and the historical records that meet the matching conditions are screened, and the corresponding historical excavation process number, settlement trough width range and associated wall behavior description are extracted to generate a historical association record set; Based on the set of historical association records, each description of the wall behavior in the historical association records is parsed one by one, extracting the trend of wall displacement change, displacement change rate, and key deformation nodes associated in the description. The records of the same type of wall displacement change trend, displacement change rate, and key deformation nodes are grouped into the same set, and the wall behavior descriptions with the highest frequency of occurrence in each set are summarized to obtain the main wall behavior pattern groups; Based on the main wall behavior pattern groups, the main wall behavior pattern groups corresponding to the current settlement trough width characteristics are matched and verified with the real-time monitoring data. By comparing whether the trends of displacement change, displacement change rates, and key deformation nodes are consistent, the behavior descriptions that conform to the current working conditions are screened to generate the expected wall response pattern.

8. The lateral deformation prediction system based on the deep foundation pit wall according to claim 1, wherein, The steps for obtaining the prediction result of the lateral deformation of the wall are as follows: Based on the expected wall response pattern, the expected wall response pattern is matched with the deformation state of the current deep foundation pit wall in terms of data, extracting the displacement increment, acceleration, and time change rate of the current deep foundation pit wall to generate the parameter set required for predicting the current deep foundation pit wall; According to the parameter set required for predicting the current deep foundation pit wall, calculate the predicted value of the lateral deformation of the deep foundation pit wall within the next time window; Based on the predicted value of the lateral deformation of the deep foundation pit wall within the next time window, compare the predicted value item by item with the historical deformation threshold, real-time monitoring data, and expected wall response pattern to obtain the prediction result of the lateral deformation of the wall.

9. The lateral deformation prediction method for the deep foundation pit wall-based lateral deformation prediction system according to any one of claims 1-8, characterized in that, Including the following steps: Collect the surface settlement monitoring values and the lateral displacement monitoring values of the corresponding measuring points of the deep foundation pit wall, perform time series alignment processing of the data points and removal of abnormal monitoring values, and establish calibrated deformation sequence data; Based on the surface settlement monitoring values in the calibrated deformation sequence data, generate surface settlement isograms by spatial interpolation to obtain the surface settlement isogram map. Based on the surface settlement isogram map, measure the width of the settlement trough formed in the key settlement area to establish the settlement trough width characteristics; Based on the lateral displacement monitoring values of the wall in the calibrated deformation sequence data, extract the maximum lateral displacement values of each monitoring section at the same time section to obtain the maximum wall displacement value. Based on the maximum wall displacement value, analyze the corresponding relationship between the evolution with the excavation process and the change of the settlement trough width characteristics to establish the surface-wall evolution association rule; Based on the settlement trough width characteristics, query the wall behavior pattern corresponding to the width characteristics in the surface-wall evolution association rule to obtain the expected wall response pattern. Based on the expected wall response pattern and combined with the deformation state of the current deep foundation pit wall, predict the lateral deformation value of the deep foundation pit wall within the next time window to obtain the prediction result of the lateral deformation of the wall.

Images