Pile foundation buckling behavior intelligent analysis method and system based on transparent soil model test
By extracting displacement field data of the soil around the pile from transparent soil model tests and using machine learning models to generate the spatial constraint stiffness distribution of the soil around the pile, weak and reinforced areas are identified, solving the problem of insufficient accuracy in the analysis of pile buckling behavior in existing technologies, and realizing intelligent and refined design of pile foundation reinforcement.
Patent Information
- Application Number
- CN202610021132.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-08
- Publication Date
- 2026-05-12
AI Technical Summary
Existing methods for analyzing the buckling behavior of pile foundations based on transparent soil tests cannot effectively convert high-dimensional visualized constraint information into parameters that can be used for precise mechanical analysis. This leads to reinforcement design relying on empirical judgment and rough estimation, making it difficult to achieve refined design.
By acquiring sequential image data from transparent soil model tests, displacement field data of the soil around the pile is extracted. A machine learning model is used to map and generate the spatial constraint stiffness distribution of the soil around the pile, identify weak and reinforced constraint areas, and evaluate reinforcement schemes through virtual parameter perturbation.
It enables refined characterization of the discontinuous and non-uniform constraint morphology of the soil around the pile, generates accurate reinforcement scheme parameters, improves the pertinence and economy of reinforcement measures, and realizes intelligent design of pile foundation buckling stability reinforcement.
Smart Images

Figure CN122021264A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent analysis technology for engineering parameters, and more specifically, to an intelligent analysis method and system for pile foundation buckling behavior based on transparent soil model tests. Background Technology
[0002] In pile foundation engineering, especially in the design and reinforcement of pile foundations involving deep and long piles or complex soil conditions, accurate prediction and assessment of pile buckling stability are crucial. Physical model testing is an important tool for studying pile-soil interaction and buckling mechanisms. Transparent soil model testing technology, due to its ability to achieve non-invasive, visual observation of the deformation field within the soil surrounding the pile, has been applied to the study of pile buckling behavior. This technology, through laser slicing and image acquisition, can obtain sequential image data of pile displacement and the displacement field of the surrounding soil during buckling. Based on this image data, existing analytical methods mainly rely on qualitative descriptions of the displacement field morphology or by fitting and inverting the macroscopic load-displacement curves obtained from experiments with classical mechanical models, such as the Winkler foundation model, to assess the overall stability of the pile foundation and guide reinforcement design.
[0003] However, existing analytical methods based on transparent soil tests have limitations: classical mechanical models simplify the lateral constraint of soil on the pile as a continuous, uniform, or linearly distributed spring system. This highly simplified representation is fundamentally different from the complex constraint morphology and failure modes exhibited by the soil around the pile in actual transparent soil tests, especially after local reinforcement, which are discontinuous, non-uniform, and highly nonlinear. This results in the inability to effectively transform the high-dimensional, spatialized, and visualized constraint information obtained from transparent soil tests into parameters that can be used for precise mechanical analysis and quantitative design. Consequently, the design and effectiveness evaluation of reinforcement schemes still largely rely on empirical judgment and rough estimation of macroscopic parameters, making it difficult to achieve refined design and optimization based on the actual constraint mechanism. Summary of the Invention
[0004] To overcome the aforementioned deficiencies of the prior art, this invention provides an intelligent analysis method and system for pile foundation buckling behavior based on transparent soil model tests to solve the problems mentioned in the background art.
[0005] To achieve the above objectives, the present invention provides the following technical solution:
[0006] The intelligent analysis method for pile foundation buckling behavior based on transparent soil model tests includes the following steps:
[0007] S1. Obtain sequential image data of the pile buckling process in the transparent soil model test, and extract displacement field data of the soil around the pile from the sequential image data;
[0008] S2. Based on displacement field data, by analyzing the spatiotemporal evolution characteristics of displacement with the loading process, identify the spatial regions where displacement occurs first, continues to develop, or accelerates during the loading process, so as to determine potential instability paths.
[0009] S3. Based on the spatiotemporal evolution characteristics of potential instability paths, quantitative data representing the spatial constraint stiffness distribution of the soil around the pile is mapped and generated through a trained machine learning model.
[0010] S4. Identify the weakly constrained areas and strongly constrained areas in the soil around the pile that play a dominant role in the buckling of the pile body based on quantitative data.
[0011] S5. Calculate the constraint contribution of each strengthened constraint region by virtually changing the constraint parameters of the strengthened constraint region and evaluating the changes in system stability; and calculate the system redundancy improvement potential of each weak constraint region by virtually increasing the constraint parameters of the weak constraint region and evaluating the changes in system stability.
[0012] S6. Based on the potential for improvement in constraint contribution and system redundancy, generate parameters for pile foundation reinforcement schemes in areas with weak constraints.
[0013] Furthermore, sequential image data of the pile buckling process in the transparent soil model test were obtained, and displacement field data of the soil around the pile was extracted from the sequential image data, including:
[0014] Spatial coordinate correction is performed on the sequence image data using preset calibration points;
[0015] Select the analysis area containing the soil around the pile in the sequence image data, and set the calculation grid within the analysis area;
[0016] Based on digital image correlation methods, the displacement vector of each grid point in the computational grid is calculated frame by frame, where the displacement vector includes both magnitude and direction;
[0017] Based on the displacement vectors of all grid points, the displacement field data of the soil around the pile is generated.
[0018] Furthermore, based on displacement field data, by analyzing the spatiotemporal evolution characteristics of displacement during the loading process, the spatial regions where displacement first occurs, continues to develop, or accelerates during loading are identified to determine potential instability paths, including:
[0019] Based on displacement field data, the moment when the displacement of each spatial location first exceeds a preset threshold during the entire loading process is calculated to form the first time distribution;
[0020] Calculate the average rate of change of displacement at each spatial location throughout the entire loading process to form a rate of change distribution;
[0021] Extract the earliest spatial location corresponding to the first time in the first time distribution, and combine it with the spatial location corresponding to the highest rate of change in the rate of change distribution to perform spatial clustering;
[0022] The connected regions formed after clustering are used as the spatial regions where displacement occurs first, continues to develop, or accelerates, thereby determining potential instability paths.
[0023] Furthermore, based on the spatiotemporal evolution characteristics of potential instability paths, a trained machine learning model is used to map and generate quantitative data characterizing the spatial constraint stiffness distribution of the soil around the pile, including:
[0024] Extract the temporal and spatial relationship features of displacement development at each point along the potential instability path to form a spatiotemporal feature set;
[0025] Input the spatiotemporal feature set into the trained machine learning model;
[0026] The spatiotemporal feature set is processed by a trained machine learning model, and the soil constraint stiffness value corresponding to each point on the potential instability path is output.
[0027] Based on the spatial location of each point on the potential instability path and its corresponding soil constraint stiffness value, combined with the benchmark constraint stiffness value assigned to the remaining areas of the soil around the pile that are not on the potential instability path, quantitative data characterizing the spatial constraint stiffness distribution of the soil around the pile is generated.
[0028] Furthermore, the trained machine learning model processes the spatiotemporal feature set and outputs the soil constraint stiffness values corresponding to each point on the potential instability path. This is achieved by performing nonlinear transformation and feature fusion on the input spatiotemporal feature set through the trained machine learning model, directly mapping it into continuous stiffness values.
[0029] Furthermore, based on quantitative data, weakly constrained areas and strongly constrained areas in the soil surrounding the pile that play a dominant role in pile buckling are identified, including:
[0030] Based on quantitative data characterizing the spatial constraint stiffness distribution of the soil around the pile, the statistical characteristic value of the spatial constraint stiffness within the soil area around the pile is calculated.
[0031] The constraint stiffness values at each spatial location in the quantified data are compared with the statistical characteristic values of spatial constraint stiffness.
[0032] Spatial locations with constraint stiffness values lower than the statistical characteristic value of spatial constraint stiffness are classified, and the classification results of adjacent spatial locations are aggregated to form weak constraint regions.
[0033] Spatial locations with constraint stiffness values higher than the statistical characteristic value of spatial constraint stiffness are classified, and the classification results of adjacent spatial locations are aggregated to form a reinforced constraint region.
[0034] Furthermore, based on the quantitative data characterizing the spatial constraint stiffness distribution of the soil around the pile, the statistical characteristic value of the spatial constraint stiffness within the soil area around the pile is calculated by traversing the constraint stiffness values of all spatial locations covered by the quantitative data and calculating their arithmetic mean or median value.
[0035] Furthermore, by virtually changing the constraint parameters of the strengthened constraint regions and evaluating changes in system stability, the constraint contribution of each strengthened constraint region is calculated. Additionally, by virtually increasing the constraint parameters of the weak constraint regions and evaluating changes in system stability, the system redundancy enhancement potential of each weak constraint region is calculated. This includes:
[0036] For each reinforced constraint region, while keeping the constraint parameters of other regions unchanged, the constraint parameters of the reinforced constraint region are virtually reduced, and the pile foundation buckling stability index is recalculated based on the adjusted overall constraint parameter distribution. The constraint contribution of the reinforced constraint region is calculated according to the magnitude of the index change.
[0037] For each weak constraint region, while keeping the constraint parameters of other regions unchanged, the constraint parameters of the weak constraint region are virtually increased, and the buckling stability index of the pile foundation is recalculated based on the adjusted overall constraint parameter distribution. The system redundancy enhancement potential of the weak constraint region is calculated according to the magnitude of the index change.
[0038] Furthermore, based on the potential for improvement in constraint contribution and system redundancy, parameters for pile foundation reinforcement schemes targeting weakly constrained areas are generated, including:
[0039] The priority order for reinforcement implementation is determined by ranking the system redundancy enhancement potential of each weak constraint area.
[0040] For each weak constraint region to be reinforced, the target constraint parameter improvement range that the weak constraint region needs to achieve is determined based on the potential for improving system redundancy.
[0041] By combining the constraint contribution of the reinforced constraint area, the location and scale of the reinforcement measures are adjusted so that the target constraint parameters of the weak constraint area after reinforcement can be improved, while maintaining or optimizing the buckling stability of the overall pile foundation system, thereby generating pile foundation reinforcement scheme parameters.
[0042] On the other hand, the present invention provides an intelligent analysis system for pile foundation buckling behavior based on transparent soil model tests, comprising the following modules:
[0043] The data extraction module is used to acquire sequential image data of the pile buckling process in the transparent soil model test, and extract displacement field data of the soil around the pile from the sequential image data;
[0044] The path determination module is used to identify the spatial regions where displacement first occurs, continues to develop, or accelerates during the loading process by analyzing the spatiotemporal evolution characteristics of displacement as the loading process based on displacement field data, so as to determine potential instability paths.
[0045] The data generation module is used to map and generate quantitative data representing the spatial constraint stiffness distribution of the soil around the pile based on the spatiotemporal evolution characteristics of potential instability paths through a trained machine learning model.
[0046] The region identification module is used to identify weakly constrained regions and strongly constrained regions in the soil around the pile that play a dominant role in the buckling of the pile body based on quantitative data.
[0047] The virtual analysis module is used to calculate the constraint contribution of each strengthened constraint region by virtually changing the constraint parameters of the strengthened constraint region and evaluating the changes in system stability, and to calculate the system redundancy improvement potential of each weak constraint region by virtually improving the constraint parameters of the weak constraint region and evaluating the changes in system stability.
[0048] The parameter generation module is used to generate parameters for pile foundation reinforcement schemes in areas with weak constraints, based on the potential for improvement in constraint contribution and system redundancy.
[0049] Compared with the prior art, the present invention has the following beneficial effects:
[0050] 1. By extracting displacement fields and identifying potential instability paths from sequence images of transparent soil tests, and then using machine learning models to directly generate spatial constraint stiffness distribution data based on the spatiotemporal evolution characteristics of the paths, this method effectively overcomes the gap between the simplified representation of traditional mechanical models and the real complex constraint morphology. It achieves a refined and data-driven representation of the discontinuous and non-uniform constraint morphology of the soil around the pile, and directly transforms the high-dimensional visualization information observed in the transparent soil tests into a stiffness parameter field that truly reflects spatial heterogeneity and can be used for subsequent quantitative mechanical analysis, thus laying a data foundation for analysis based on real constraint mechanisms.
[0051] 2. Based on the stiffness distribution obtained by quantification, weak and reinforced areas are identified. The constraint contribution and system redundancy enhancement potential of each area are evaluated by quantification through virtual parameter perturbation. Finally, reinforcement scheme parameters are generated accordingly, forming a technical closed loop from intelligent identification, quantitative evaluation to scheme generation. This transforms reinforcement design from relying on macro-experience judgment to precise decision-making based on local real constraint characteristics and global stability contribution. The location, scale, and strength parameters of reinforcement measures directly correspond to the engineering requirements obtained through intelligent analysis, thereby significantly improving the pertinence and economy of reinforcement measures and realizing the refined and intelligent design of pile foundation buckling stability reinforcement. Attached Figure Description
[0052] Figure 1 This is a flowchart of the intelligent analysis method for pile foundation buckling behavior based on transparent soil model tests according to the present invention;
[0053] Figure 2 This is a schematic diagram of the intelligent analysis system for pile foundation buckling behavior based on transparent soil model tests according to the present invention. Detailed Implementation
[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.
[0055] Example 1: Figure 1 The present invention provides an intelligent analysis method for pile foundation buckling behavior based on transparent soil model tests, which includes the following steps:
[0056] S1. Obtain sequential image data of the pile buckling process in the transparent soil model test, and extract displacement field data of the soil around the pile from the sequential image data;
[0057] S2. Based on displacement field data, by analyzing the spatiotemporal evolution characteristics of displacement with the loading process, identify the spatial regions where displacement occurs first, continues to develop, or accelerates during the loading process, so as to determine potential instability paths.
[0058] S3. Based on the spatiotemporal evolution characteristics of potential instability paths, quantitative data representing the spatial constraint stiffness distribution of the soil around the pile is mapped and generated through a trained machine learning model.
[0059] S4. Identify the weakly constrained areas and strongly constrained areas in the soil around the pile that play a dominant role in the buckling of the pile body based on quantitative data.
[0060] S5. Calculate the constraint contribution of each strengthened constraint region by virtually changing the constraint parameters of the strengthened constraint region and evaluating the changes in system stability; and calculate the system redundancy improvement potential of each weak constraint region by virtually increasing the constraint parameters of the weak constraint region and evaluating the changes in system stability.
[0061] S6. Based on the potential for improvement in constraint contribution and system redundancy, generate parameters for pile foundation reinforcement schemes in areas with weak constraints.
[0062] The specific implementation details of step S1 are as follows. Before the transparent soil model test begins, multiple physical markers with fixed and known spatial coordinates are pre-arranged on the inner wall of the model box and inside the transparent soil material as preset calibration points. These preset calibration points are clearly visible in all subsequent sequence image data. After acquiring the sequence image data of the pile foundation buckling process, the positions of these preset calibration points in the image pixel coordinate system are identified for each frame of the image. By establishing the correspondence between at least three non-collinear preset calibration points in the real-world coordinate system and the image pixel coordinate system, the spatial coordinate correction parameters for each frame of the image are calculated using a direct linear transformation algorithm. The spatial coordinate correction parameters are used to perform geometric correction on all sequence image data to ensure that the coordinates of any pixel in the corrected image can be mapped to the real three-dimensional spatial coordinate system with the bottom surface of the model box as the reference plane through a unified transformation relationship. This eliminates image spatial distortion caused by camera lens distortion or shooting angle, ensuring that all image data are under a unified and accurate spatial measurement system. This spatial coordinate correction process is the basis for subsequent quantitative displacement analysis; uncorrected image data cannot guarantee the geometric accuracy of displacement calculations.
[0063] After spatial coordinate correction, a fixed rectangular region is selected from all sequential image data as the analysis region encompassing the soil around the pile. The boundary of this analysis region is determined according to the experimental design, for example, extending outwards from the pile center to a range where soil deformation is negligible. Within the selected analysis region, a computational grid consisting of uniformly distributed nodes is established, discretizing the continuous image region. The size of the computational grid is determined based on the required analytical precision and image resolution, specifically by setting a target physical resolution, for example, expecting each grid cell to correspond to an actual physical size of 2 mm × 2 mm. Based on the known imaging system resolution, for example, 10 pixels per millimeter, the pixel spacing of the grid in the image can be calculated. Once this computational grid is set, it is applied to all spatially corrected sequential image data frames to ensure a consistent spatial reference frame for displacement calculations. The determination of the analysis region and the computational grid provides clear boundaries and calculation points for subsequent displacement quantization calculations of local areas.
[0064] The core displacement field data is obtained frame-by-frame through a digital image correlation method. For each grid point in the computational grid, a sub-region of odd-numbered pixels is defined around it, such as a 21-pixel x 21-pixel square region. This size must ensure that the sub-region contains sufficient texture information for reliable matching. In the reference image frame before loading begins, the grayscale distribution characteristics of the corresponding sub-region for each grid point are recorded. In each subsequent image frame during the loading process, for the same grid point, within a certain search area near its expected position, a new position that best matches the grayscale distribution of the sub-region of that point in the reference image is found using a normalized cross-correlation coefficient algorithm. The effectiveness of the matching is judged by a preset normalized cross-correlation coefficient threshold. The specific value of this threshold is determined as follows: before the experiment begins or using an image sequence in an unloaded state, multiple pairs of image regions with known small relative displacements are selected for cross-correlation coefficient calculation. The typical distribution of the correlation coefficient is statistically obtained, and the critical value that can effectively distinguish between correct and incorrect matches is set as this threshold, for example, 0.85. This threshold is used to determine the confidence level of the matching result. A match is considered valid when the cross-correlation coefficient of the found matching positions is higher than this normalized cross-correlation coefficient threshold. The displacement vector of the grid point in the current image frame relative to the reference image frame is calculated using the difference in the center positions of the matched pixels. The magnitude of the displacement vector is recorded in pixels and then converted to the actual physical length unit of millimeters based on the known image spatial resolution, for example, 0.1 millimeters per pixel. The direction of the displacement vector is determined by the direction of the position difference. This calculation process is repeated for all grid points frame by frame to obtain the instantaneous displacement vector field covering the entire analysis area at each image frame.
[0065] Finally, the instantaneous displacement vector fields from all time frames are integrated to generate complete pile-surround soil displacement field data. This displacement field data is a spatiotemporal dataset, defined spatially by the node positions of the computational grid and temporally by the time series of image acquisition. For each grid point, its displacement history is recorded as a sequence of displacement vector changes over time, with the magnitude of the displacement vector in millimeters and the direction expressed as angle or direction cosine. The final generated displacement field data will be used for subsequent analysis steps. This data is stored in a structured array or database, explicitly linking the spatial coordinates, timestamp, displacement vector magnitude, and direction of each data point. Through the complete process described above—from image acquisition, geometric correction, region gridding, and sub-pixel precision displacement matching to data integration—accurate, reliable, and directly usable pile-surround soil displacement field data is extracted from the original sequence image data, providing an accurate data foundation for the entire method.
[0066] The specific implementation of step S2 is explained below. This step is based on the analysis of the pile perimeter soil displacement field data generated in step S1. The displacement field data is a structured data set containing spatial location, time, and corresponding displacement vectors. The analysis first focuses on the magnitude component of the displacement field data. For each spatial location defined in the displacement field data, i.e., each node of the computational grid, the magnitude of the displacement over the entire loading time series is extracted. This data is a sequence of displacement values changing over time. Based on this time series, the loading time or image frame number corresponding to the first time the displacement magnitude at that location exceeds a preset displacement threshold is calculated. The specific value of this preset displacement threshold is determined based on the observation accuracy of the experiment and the initial nonlinear stage characteristics of soil deformation. Its purpose is to effectively identify the moment when significant deformation begins to occur and eliminate the influence of minor measurement noise or initial compaction deformation. The specific method for setting the preset displacement threshold includes: in the initial stage of loading, selecting a period in which soil deformation is considered to be linear elastic or not significantly developed, calculating the average and standard deviation of the displacement magnitude at all spatial locations within this period, and setting the average plus two or three times the standard deviation as the preset displacement threshold; or, based on the scale and pile diameter of the model test, setting the threshold as an empirical value related to the model geometry, such as the physical displacement corresponding to 0.5% of the model pile diameter. By traversing all spatial locations and performing this judgment, a first threshold exceeding moment is assigned to each location, thereby forming a first time distribution in space corresponding to the computational grid, which characterizes the chronological order in which significant deformation begins at different locations.
[0067] While calculating the initial time distribution, the average rate of change of displacement magnitude at each spatial location throughout the entire loading process is calculated in parallel to quantify the speed of deformation development. For each spatial location, a sequence of displacement magnitude changes over time is used as input. The calculation of the average rate of change is achieved by determining the overall growth trend of this displacement time series. Specifically, the ratio of the total displacement increment from the start to the end of loading to the total loading time is calculated as the average rate of change for that location. The total loading time refers to the actual time or number of loading steps elapsed from the first reference frame to the last frame. By performing this calculation on all spatial locations, a rate of change distribution corresponding to the computational grid space is obtained, which visually reflects the average rate of soil deformation development at different locations.
[0068] After obtaining the first-time distribution and rate of change distribution, key spatial location information needs to be extracted for identifying potential regions. From the first-time distribution, the earliest spatial locations corresponding to each time point are extracted, for example, the top 10% of spatial locations at each time point. From the rate of change distribution, the spatial locations corresponding to the highest average rate of change are extracted, for example, the top 10% of spatial locations with the highest average rate of change. These two sets of spatial locations together constitute a key point set characterizing the earliest occurrence and continuous acceleration of displacement. Subsequently, spatial clustering analysis is performed on these key spatial points. Spatial clustering analysis is based on the spatial coordinates of each point within the point set, grouping points that are spatially adjacent or close together. Specifically, a distance-based clustering method is used, setting a spatial neighborhood radius and aggregating points whose distance to each other is less than this radius. The setting of this spatial neighborhood radius is related to the size of the computational grid and the physical width of the instability path to be identified; for example, it can be set to the size of 2 to 3 grid cells to ensure that neighboring points representing the same potential deformation zone can be merged. Through spatial clustering, the discrete key point set is divided into several spatially connected point groups.
[0069] Finally, the spatial extent covered by each group of connected points formed after spatial clustering is defined as the spatial region where displacement first occurs, continues to develop, or accelerates. The spatial extension of these regions constitutes potential soil shear zones or deformation concentration zones, which are the potential instability paths that need to be identified. Each potential instability path is marked by a series of spatially continuous grid nodes or soil elements with significant spatiotemporal evolution characteristics. These potential instability paths are weak spatial channels most likely to experience gradual failure, quantitatively identified from the continuous displacement field, providing a clear target and spatial framework for in-depth analysis of the spatial heterogeneity and evolution of soil constraint stiffness in subsequent steps. Through the above systematic analysis process, objective and quantitative identification of key instability paths from full-field displacement data is achieved.
[0070] The specific implementation of step S3 is explained below. This step is based on the potential instability path and its spatiotemporal evolution characteristics determined in step S2. For each potential instability path, feature data of all discrete points on the path are extracted to construct a spatiotemporal feature set. For each point on the path, the feature extraction mainly focuses on two aspects: displacement development time sequence and spatial positional relationship. The displacement development time sequence feature refers to the complete sequence data of the displacement magnitude evolving over time during the entire loading process, which is obtained through the displacement field data generated in step S1. The spatial positional relationship feature includes the absolute spatial coordinates of the point itself, as well as its relative positional relationship with adjacent points on the path, such as relative distance and azimuth. Combining these time and spatial features, a multi-dimensional feature vector is constructed for each point on the path. The set of feature vectors of all points constitutes the spatiotemporal feature set used to describe the spatiotemporal evolution mode of the path. The construction of the spatiotemporal feature set transforms the continuous physical evolution process into a structured and digital feature expression, providing standardized input for subsequent machine learning models.
[0071] The constructed spatiotemporal feature set is input into a pre-trained machine learning model. This pre-trained machine learning model refers to a neural network model whose parameters have been optimized through a large amount of training data. Its training process is usually completed in other independent training stages, with training data derived from historical transparent soil model experiments or high-fidelity numerical simulations. The training objective is to enable the model to accurately predict the soil constraint stiffness value at the corresponding location from the input spatiotemporal features. In this step, this pre-trained machine learning model is directly called for inference and prediction. The model receives the spatiotemporal feature set as input, and the dimension of its features must match the design dimension of the model's input layer.
[0072] The trained machine learning model processes the input spatiotemporal feature set and outputs the soil constraint stiffness value corresponding to each point on the potential instability path. This processing essentially involves a series of nonlinear transformations and feature fusion calculations within the model. Specifically, the model contains multiple network layers, each with a large number of neurons and connection weights. The input feature vector is first linearly combined with the weight matrix of the first layer, plus a bias term, and then passed through a nonlinear activation function, such as the ReLU function, to produce the output of the first layer. This output serves as the input to the next layer, repeating a similar linear combination and nonlinear transformation process. Through successive propagation and abstraction across multiple hidden layers, the model finally maps a high-dimensional feature representation incorporating complex spatiotemporal patterns to a single, continuous real value at the output layer. This value is the soil constraint stiffness value predicted by the model at that point, typically measured in megapascals per meter (MPa). All weights and bias parameters within the model are determined and fixed during the training phase and remain unchanged in the current inference application. By performing independent forward propagation calculations on the feature vectors of each point in the spatiotemporal feature set, the soil constraint stiffness values corresponding to all points on the path can be obtained.
[0073] After obtaining the points and soil constraint stiffness values along the potential instability path, it is necessary to generate quantitative data on the spatial constraint stiffness distribution covering the entire analysis area of the soil around the pile. Since the machine learning model only predicts the points along the path, a baseline constraint stiffness value needs to be assigned to the vast soil areas not along the potential instability path. This baseline constraint stiffness value can be determined based on the known stiffness empirical range of similar soils under undisturbed conditions. For example, the small strain stiffness value of the transparent soil material can be measured through laboratory geotechnical tests and assigned as a constant to all non-path areas. Another method is to perform sparse sampling within the analysis area, use a trained machine learning model to predict the stiffness values of these sampled points, and then take their statistical average as the baseline constraint stiffness value. Finally, the spatial coordinates of each point along the path and the predicted stiffness values, as well as the coordinates of each grid point in the non-path areas and the baseline constraint stiffness value, are integrated to form a complete data structure. In this data structure, each spatial location corresponds to a specific constraint stiffness value, which together form a quantitative data map characterizing the spatial constraint stiffness distribution of the soil around the pile. This data map is stored and expressed in the form of a two-dimensional matrix or geographic information system raster data, providing direct quantitative input for the area identification in step S4.
[0074] The specific implementation of step S4 is explained below. This step is based on the quantitative data representing the spatial constraint stiffness distribution of the soil around the pile generated in step S3. The quantitative data is a two-dimensional matrix or raster dataset, where each data cell corresponds to a specific spatial location and stores a value representing the magnitude of the soil constraint stiffness at that location, typically in megapascals per meter. The primary task of the analysis is to calculate the statistical characteristic value of the spatial constraint stiffness within the entire soil area around the pile. This value will serve as a quantitative benchmark for distinguishing regions with different constraint strengths. The method for calculating this statistical characteristic value of the spatial constraint stiffness is to traverse all spatial locations covered by the quantitative data, read the constraint stiffness values at all locations, and form a sample set containing all stiffness values. Then, statistical analysis is performed on this sample set to calculate a representative value of its central tendency. Specifically, the arithmetic mean of all stiffness values can be calculated as the statistical characteristic value of the spatial constraint stiffness. The calculation process for the arithmetic mean involves adding all stiffness values in the sample set and then dividing by the total number of samples. Another feasible approach is to calculate the median value of the sample set as the statistical characteristic value of the spatial constraint stiffness. This involves first sorting all stiffness values by size, and then taking the value that is exactly in the middle of the sorted values. The rationale for choosing between the arithmetic mean and the median is that the mean reflects the average overall stiffness level, while the median is less sensitive to extreme values in the data and is more robust. In practice, one can choose based on the distribution characteristics of the quantified data; for example, the mean is used when the stiffness value distribution is relatively symmetrical, while the median is used when the distribution is significantly skewed. This calculation process is completed automatically by a computer program.
[0075] After determining the statistical characteristic value of spatial constraint stiffness, the constraint stiffness value corresponding to each spatial location in the quantified data is compared one by one with the statistical characteristic value. The purpose of the comparison is to determine whether the soil constraint stiffness at each location is lower than, equal to, or higher than the average or median level of the entire area. This comparison process is a simple scalar numerical magnitude judgment. For each spatial location, if its constraint stiffness value is strictly less than the statistical characteristic value, it is marked as a potential low-constraint point; if its constraint stiffness value is strictly greater than the statistical characteristic value, it is marked as a potential high-constraint point. In practice, for the sake of rigor, the case where the constraint stiffness value is equal to the statistical characteristic value can be specially handled, for example, by classifying it into the adjacent mainstream category, or assigning it according to the category of its surrounding points.
[0076] Next, the spatial locations marked as potential low-constraint points need to be spatially aggregated to form continuous weakly constrained regions. The principle of aggregation is to merge spatially adjacent or connected low-constraint points into the same region. Spatial adjacency is usually defined based on pixel or grid adjacency relationships, such as using four-neighbor connectivity or eight-neighbor connectivity rules. The aggregation process can be implemented using connected component labeling algorithms or region growing algorithms in image processing. In practice, starting from any unvisited low-constraint point, all its adjacent low-constraint points that satisfy the adjacency rules are found, and these points are grouped into the same region. This process is then recursively continued to find adjacent low-constraint points for newly added points until no new adjacent low-constraint points can be added, thus forming a complete connected region. After traversing all low-constraint points, several unconnected connected regions composed of low-constraint points are obtained. These regions are the identified weakly constrained regions, representing the spatial range in the soil around the pile where the constraint capacity is significantly lower than the overall average level, and are areas that adversely affect pile buckling.
[0077] Parallel to or similar to the process of forming weakly constrained regions, spatial aggregation is performed on the spatial locations marked as potentially high-constraint points to form continuous reinforced-constraint regions. Using the same spatial adjacency definition and aggregation algorithm as described above, all spatially connected high-constraint points are merged together. Each connected region thus formed is a reinforced-constraint region. A reinforced-constraint region represents a spatial range in the soil surrounding the pile where the constraint capacity is significantly higher than the overall average level, providing favorable support against pile buckling. Finally, the output of step S4 is two distinct sets of spatial regions: a set of weakly constrained regions and a set of reinforced-constraint regions. Each region in these sets is defined by its boundary coordinates or the index of the grid points it contains. This regional information provides a clear spatial analysis object for calculating the constraint contribution and system redundancy enhancement potential in step S5. Through the above method based on statistical benchmark comparison and spatial connectivity analysis, a discretized and regionalized interpretation of engineering significance from continuous stiffness distribution data is achieved.
[0078] The specific implementation of step S5 is explained below. This step is based on the set of reinforced and weakly constrained regions identified in step S4, and the quantitative data characterizing the spatial constraint stiffness distribution of the soil around the pile generated in step S3. The goal of this step is to quantitatively assess the importance of each specific region to the overall buckling stability of the pile foundation through virtual perturbation analysis. First, the constraint contribution is calculated sequentially for each independent region in the set of reinforced constrained regions. At the start of the calculation, the complete spatial constraint stiffness distribution quantitative data generated in step S3 is used as the baseline condition. For a particular reinforced constrained region currently being analyzed, in the computer numerical analysis environment, the constraint stiffness parameter values at all spatial locations covered by this region are virtually reduced, while keeping the constraint stiffness parameter values of all other regions of the soil around the pile completely consistent with the baseline condition, thereby constructing a new, specifically modified overall constraint parameter distribution. The virtual reduction needs to be a quantifiable value. For example, the constraint stiffness values of all points in the region can be uniformly multiplied by a reduction factor less than 1, such as 0.5, meaning the stiffness is reduced to 50% of the original value. Alternatively, a specific reduction percentage can be determined based on the degree to which the stiffness value of the region exceeds the overall average level. This process involves manually setting parameters in the analysis model, not performing the operation in an actual physical experiment.
[0079] Based on the new overall constraint parameter distribution obtained after the above adjustments, the buckling stability analysis of the pile foundation system is recalculated. This recalculation process is usually achieved through the finite element method or other numerical methods. Its core is to solve for the critical load or minimum eigenvalue at which the pile foundation structure buckles and becomes unstable under the updated soil constraints. This critical load or eigenvalue is the recalculated pile foundation buckling stability index. This index is compared with the original pile foundation buckling stability index under the benchmark condition, and its relative change is calculated. The change can be calculated, for example, by subtracting the new index value from the original index value and then dividing by the original index value. The percentage obtained is the rate of decrease in system stability caused by the weakening of the stiffness of the reinforced constraint region. This rate of decrease is defined as the constraint contribution of the reinforced constraint region. The higher the constraint contribution, the more critical the role of this region in maintaining the overall stability of the system, and the greater the negative impact on system safety if its stiffness is lost. For each region in the set of reinforced constraint regions, the above virtual reduction, recalculation, and comparison process is repeated independently to obtain the constraint contribution value of each region.
[0080] After completing the analysis of all reinforced constraint regions, the system redundancy enhancement potential is calculated in parallel for each independent region in the set of weakly constrained regions. The calculation also starts with the stiffness distribution of the baseline load case. For a particular weakly constrained region currently being analyzed, the constraint stiffness parameter values at all spatial locations covered by that region are virtually increased in the computer numerical analysis environment, while keeping the parameters of all other regions unchanged, thus constructing a new overall constraint parameter distribution. The magnitude of this virtual increase also needs to be quantified. For example, the constraint stiffness values at all points within the region can be uniformly multiplied by an enhancement coefficient greater than 1, such as 2.0, meaning the stiffness is increased to twice the original value; or a specific increment value can be set. Similarly, based on this adjusted constraint parameter distribution, the pile foundation buckling stability analysis is performed again to obtain a new pile foundation buckling stability index.
[0081] The new index is compared with the original index under the baseline condition, and their relative change is calculated. The change can be calculated, for example, by subtracting the original index value from the new index value and then dividing by the original index value. The resulting percentage represents the system stability improvement rate due to the increased stiffness of the weak constraint region. This improvement rate is defined as the system redundancy improvement potential of the weak constraint region. The higher the system redundancy improvement potential, the more significant the gain effect on the overall system stability that engineering measures can bring to strengthen the region, i.e., the greater the cost-effectiveness or improvement potential of strengthening the region. For each region in the set of weak constraint regions, the above virtual improvement, recalculation, and comparison process is repeated independently to obtain the system redundancy improvement potential value for each region. Finally, the output of step S5 is the constraint contribution value assigned to each strengthened constraint region and the system redundancy improvement potential value assigned to each weak constraint region. These two sets of quantitative indicators provide a direct decision-making basis for formulating targeted and prioritized strengthening schemes in step S6. Through the above virtual parameter perturbation and stability reassessment process, a deeper analysis from static region identification to dynamic quantitative ranking of region importance is achieved.
[0082] The specific implementation of step S6 is explained below. This step generates pile foundation reinforcement scheme parameters with clear engineering guidance based on the system redundancy enhancement potential values of each weak constraint region and the constraint contribution values of each reinforced constraint region calculated in step S5. First, all identified weak constraint regions are sorted from largest to smallest according to the specific system redundancy enhancement potential values of each region calculated in step S5. This sorting is based on direct numerical comparison; regions with larger system redundancy enhancement potential values are ranked higher in the sorting sequence. This sorting result directly defines the priority order of reinforcement project implementation. The weak constraint region ranked highest is determined as the target region for highest priority reinforcement because reinforcing this region is expected to bring the greatest improvement benefit to the buckling stability of the entire pile foundation system. This sorting process is completed automatically, and its purpose is to provide a scientific decision-making basis for step-by-step or selective reinforcement when engineering resources are limited.
[0083] After determining the reinforcement priority order, for each weakly constrained area to be reinforced in the sequence, it is necessary to specifically determine the target to be achieved by the reinforcement project, that is, the extent to which the soil constraint parameters in that area need to be improved. The determination of this target constraint parameter improvement directly depends on the system redundancy improvement potential value of that area. One specific method is to establish a proportional relationship between the target improvement magnitude and the system redundancy improvement potential value. For example, a baseline improvement coefficient can be set, and the system redundancy improvement potential value can be multiplied by this baseline improvement coefficient; the product is then used as the percentage increase in the target constraint parameters for that area. The setting of the baseline improvement coefficient needs to comprehensively consider the feasibility of the reinforcement technology, economic costs, and the urgency of the stability improvement requirement; for example, it can be set to 0.8 or 1.0. Another method is to assign different target improvement magnitude levels based on the different ranges in which the system redundancy improvement potential value falls. For example, regions with a redundancy enhancement potential greater than 15% are defined as Level 1 enhancement regions, with a target of increasing the constraint stiffness of these regions by 30%; regions with a redundancy enhancement potential between 5% and 15% are defined as Level 2 enhancement regions, with a target of increasing it by 20%. Through this method, a specific and quantifiable reinforcement performance index is derived for each weak constraint region.
[0084] Generating the final pile foundation reinforcement scheme parameters requires considering not only the improvement target of the weak area itself, but also the overall impact of the reinforced constraint area, especially its criticality represented by its constraint contribution. In this step, the constraint contribution of the reinforced constraint area must be taken into account as a constraint condition. Specifically, when planning reinforcement measures for a weak constraint area, such as grouting reinforcement, soil replacement, or the installation of reinforced bodies, the precise location and scale of the reinforcement measures should be optimized based on the constraint contribution of neighboring reinforced constraint areas. A reinforced constraint area with a higher constraint contribution indicates a more critical supporting role for the overall stability of the system. Therefore, when implementing reinforcement of adjacent weak areas, detailed numerical simulations should be used to ensure that the proposed reinforcement technology, reinforcement range, and reinforcement intensity, while achieving the target improvement in the weak area's constraint parameters, will not adversely disturb or weaken adjacent reinforced constraint areas with high constraint contributions, such as avoiding grouting pressure leading to damage to the soil structure in critical areas. The final pile foundation reinforcement scheme parameters are a collection of specific engineering instructions, including but not limited to the number of each weakly constrained area requiring reinforcement, its geographical boundary, target stiffness enhancement value, recommended reinforcement methods, suggested dimensions and material parameters of the reinforcement body, and the technological constraints to be followed during construction to protect the critical reinforcement areas. These parameters together constitute an executable reinforcement design blueprint directly derived from the intelligent analysis results, thus completing a full technical loop from experimental observation, intelligent identification, quantitative evaluation to the generation of engineering measures.
[0085] Example 2: Figure 2 A schematic diagram of the intelligent analysis system for pile foundation buckling behavior based on transparent soil model tests is provided. The intelligent analysis system for pile foundation buckling behavior based on transparent soil model tests includes the following modules:
[0086] The data extraction module is used to acquire sequential image data of the pile buckling process in the transparent soil model test, and extract displacement field data of the soil around the pile from the sequential image data;
[0087] The path determination module is used to identify the spatial regions where displacement first occurs, continues to develop, or accelerates during the loading process by analyzing the spatiotemporal evolution characteristics of displacement as the loading process based on displacement field data, so as to determine potential instability paths.
[0088] The data generation module is used to map and generate quantitative data representing the spatial constraint stiffness distribution of the soil around the pile based on the spatiotemporal evolution characteristics of potential instability paths through a trained machine learning model.
[0089] The region identification module is used to identify weakly constrained regions and strongly constrained regions in the soil around the pile that play a dominant role in the buckling of the pile body based on quantitative data.
[0090] The virtual analysis module is used to calculate the constraint contribution of each strengthened constraint region by virtually changing the constraint parameters of the strengthened constraint region and evaluating the changes in system stability, and to calculate the system redundancy improvement potential of each weak constraint region by virtually improving the constraint parameters of the weak constraint region and evaluating the changes in system stability.
[0091] The parameter generation module is used to generate parameters for pile foundation reinforcement schemes in areas with weak constraints, based on the potential for improvement in constraint contribution and system redundancy.
[0092] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.
[0093] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.
[0094] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and inventive constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0095] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.
[0096] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.
[0097] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
[0098] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. An intelligent analysis method for pile foundation buckling behavior based on transparent soil model tests, characterized in that, Includes the following steps: S1. Obtain sequential image data of the pile buckling process in the transparent soil model test, and extract displacement field data of the soil around the pile from the sequential image data; S2. Based on displacement field data, by analyzing the spatiotemporal evolution characteristics of displacement with the loading process, identify the spatial regions where displacement occurs first, continues to develop, or accelerates during the loading process, so as to determine potential instability paths. S3. Based on the spatiotemporal evolution characteristics of potential instability paths, quantitative data representing the spatial constraint stiffness distribution of the soil around the pile is mapped and generated through a trained machine learning model. S4. Identify the weakly constrained areas and strongly constrained areas in the soil around the pile that play a dominant role in the buckling of the pile body based on quantitative data. S5. Calculate the constraint contribution of each strengthened constraint region by virtually changing the constraint parameters of the strengthened constraint region and evaluating the changes in system stability; and calculate the system redundancy improvement potential of each weak constraint region by virtually increasing the constraint parameters of the weak constraint region and evaluating the changes in system stability. S6. Based on the potential for improvement in constraint contribution and system redundancy, generate parameters for pile foundation reinforcement schemes in areas with weak constraints.
2. The intelligent analysis method for pile foundation buckling behavior based on transparent soil model tests according to claim 1, characterized in that, Acquire sequential image data of the pile buckling process in the transparent soil model test, and extract displacement field data of the soil around the pile from the sequential image data, including: Spatial coordinate correction is performed on the sequence image data using preset calibration points; Select the analysis area containing the soil around the pile in the sequence image data, and set the calculation grid within the analysis area; Based on digital image correlation methods, the displacement vector of each grid point in the computational grid is calculated frame by frame, where the displacement vector includes both magnitude and direction; Based on the displacement vectors of all grid points, the displacement field data of the soil around the pile is generated.
3. The intelligent analysis method for pile foundation buckling behavior based on transparent soil model tests according to claim 1, characterized in that, Based on displacement field data, by analyzing the spatiotemporal evolution characteristics of displacement during the loading process, we can identify the spatial regions where displacement first occurs, continues to develop, or accelerates during loading, in order to determine potential instability paths, including: Based on displacement field data, the moment when the displacement of each spatial location first exceeds a preset threshold during the entire loading process is calculated to form the first time distribution; Calculate the average rate of change of displacement at each spatial location throughout the entire loading process to form a rate of change distribution; Extract the earliest spatial location corresponding to the first time in the first time distribution, and combine it with the spatial location corresponding to the highest rate of change in the rate of change distribution to perform spatial clustering; The connected regions formed after clustering are used as the spatial regions where displacement occurs first, continues to develop, or accelerates, thereby determining potential instability paths.
4. The intelligent analysis method for pile foundation buckling behavior based on transparent soil model tests according to claim 1, characterized in that, Based on the spatiotemporal evolution characteristics of potential instability paths, a trained machine learning model is used to map and generate quantitative data characterizing the spatial constraint stiffness distribution of the soil around the pile, including: Extract the temporal and spatial relationship features of displacement development at each point along the potential instability path to form a spatiotemporal feature set; Input the spatiotemporal feature set into the trained machine learning model; The spatiotemporal feature set is processed by a trained machine learning model, and the soil constraint stiffness value corresponding to each point on the potential instability path is output. Based on the spatial location of each point on the potential instability path and its corresponding soil constraint stiffness value, combined with the benchmark constraint stiffness value assigned to the remaining areas of the soil around the pile that are not on the potential instability path, quantitative data characterizing the spatial constraint stiffness distribution of the soil around the pile is generated.
5. The intelligent analysis method for pile foundation buckling behavior based on transparent soil model tests according to claim 4, characterized in that, The trained machine learning model processes the spatiotemporal feature set and outputs the soil constraint stiffness values corresponding to each point on the potential instability path. This is achieved by performing nonlinear transformation and feature fusion on the input spatiotemporal feature set through the trained machine learning model, directly mapping it into continuous stiffness values.
6. The intelligent analysis method for pile foundation buckling behavior based on transparent soil model tests according to claim 1, characterized in that, Based on quantitative data, weakly constrained areas and strongly constrained areas in the soil surrounding the pile that play a dominant role in pile buckling are identified, including: Based on quantitative data characterizing the spatial constraint stiffness distribution of the soil around the pile, the statistical characteristic value of the spatial constraint stiffness within the soil area around the pile is calculated. The constraint stiffness values at each spatial location in the quantified data are compared with the statistical characteristic values of spatial constraint stiffness. Spatial locations with constraint stiffness values lower than the statistical characteristic value of spatial constraint stiffness are classified, and the classification results of adjacent spatial locations are aggregated to form weak constraint regions. Spatial locations with constraint stiffness values higher than the statistical characteristic value of spatial constraint stiffness are classified, and the classification results of adjacent spatial locations are aggregated to form a reinforced constraint region.
7. The intelligent analysis method for pile foundation buckling behavior based on transparent soil model tests according to claim 6, characterized in that, Based on the quantitative data characterizing the spatial constraint stiffness distribution of the soil around the pile, the statistical characteristic value of the spatial constraint stiffness within the soil area around the pile is calculated by traversing the constraint stiffness values of all spatial locations covered by the quantitative data and calculating their arithmetic mean or median value.
8. The intelligent analysis method for pile foundation buckling behavior based on transparent soil model tests according to claim 1, characterized in that, The constraint contribution of each strengthened constraint region is calculated by virtually changing the constraint parameters of the strengthened constraint region and evaluating changes in system stability. The system redundancy enhancement potential of each weak constraint region is calculated by virtually increasing the constraint parameters of the weak constraint region and evaluating changes in system stability. This includes: For each reinforced constraint region, while keeping the constraint parameters of other regions unchanged, the constraint parameters of the reinforced constraint region are virtually reduced, and the pile foundation buckling stability index is recalculated based on the adjusted overall constraint parameter distribution. The constraint contribution of the reinforced constraint region is calculated according to the magnitude of the index change. For each weak constraint region, while keeping the constraint parameters of other regions unchanged, the constraint parameters of the weak constraint region are virtually increased, and the buckling stability index of the pile foundation is recalculated based on the adjusted overall constraint parameter distribution. The system redundancy enhancement potential of the weak constraint region is calculated according to the magnitude of the index change.
9. The intelligent analysis method for pile foundation buckling behavior based on transparent soil model tests according to claim 1, characterized in that, Based on the potential for improving constraint contribution and system redundancy, parameters for pile foundation reinforcement schemes in weakly constrained areas are generated, including: The priority order for reinforcement implementation is determined by ranking the system redundancy enhancement potential of each weak constraint area. For each weak constraint region to be reinforced, the target constraint parameter improvement range that the weak constraint region needs to achieve is determined based on the potential for improving system redundancy. By combining the constraint contribution of the reinforced constraint area, the location and scale of the reinforcement measures are adjusted so that the target constraint parameters of the weak constraint area after reinforcement can be improved, while maintaining or optimizing the buckling stability of the overall pile foundation system, thereby generating pile foundation reinforcement scheme parameters.
10. An intelligent analysis system for pile foundation buckling behavior based on transparent soil model tests, used to implement the intelligent analysis method for pile foundation buckling behavior based on transparent soil model tests as described in any one of claims 1-9, characterized in that, Includes the following modules: The data extraction module is used to acquire sequential image data of the pile buckling process in the transparent soil model test, and extract displacement field data of the soil around the pile from the sequential image data; The path determination module is used to identify the spatial regions where displacement first occurs, continues to develop, or accelerates during the loading process by analyzing the spatiotemporal evolution characteristics of displacement as the loading process based on displacement field data, so as to determine potential instability paths. The data generation module is used to map and generate quantitative data representing the spatial constraint stiffness distribution of the soil around the pile based on the spatiotemporal evolution characteristics of potential instability paths through a trained machine learning model. The region identification module is used to identify weakly constrained regions and strongly constrained regions in the soil around the pile that play a dominant role in the buckling of the pile body based on quantitative data. The virtual analysis module is used to calculate the constraint contribution of each strengthened constraint region by virtually changing the constraint parameters of the strengthened constraint region and evaluating the changes in system stability, and to calculate the system redundancy improvement potential of each weak constraint region by virtually improving the constraint parameters of the weak constraint region and evaluating the changes in system stability. The parameter generation module is used to generate parameters for pile foundation reinforcement schemes in areas with weak constraints, based on the potential for improvement in constraint contribution and system redundancy.
