Karst region pile foundation intelligent positioning and accurate construction method

Abstract

This invention provides a method for intelligent positioning and precise construction of pile foundations in karst areas. The method includes acquiring a dataset of karst development characteristics of the target engineering area, comprising a 3D geological exploration data set and a karst attribute data set. The dataset is preprocessed to obtain a karst preprocessed dataset. Based on a feature extraction model, features are extracted from the preprocessed 3D geological exploration data set to generate a 3D geological feature map. A karst cavity identification model is used to identify the map, generating a karst cavity distribution heat map. The karst preprocessed dataset undergoes karst cavity feature transformation processing to obtain a pile foundation construction feature dataset. Based on the karst cavity distribution heat map, the pile foundation construction feature dataset is constrained to generate a pile foundation constraint boundary map. Based on the pile foundation positioning optimization algorithm and the pile foundation constraint boundary map, the final borehole positioning coordinate set is output. This invention can improve the safety and accuracy of pile foundation construction in karst areas, and reduce construction risks and costs.

Description

Technical Field

[0001] This invention relates to the field of karst geological engineering and intelligent construction technology, and more specifically, to a method for intelligent positioning and precise construction of pile foundations in karst areas. Background Technology

[0002] Traditional methods face two major technical bottlenecks when constructing pile foundations in karst areas: First, due to insufficient spatial positioning accuracy of karst caves, traditional geological exploration methods such as advanced drilling can only obtain discrete point data, making it difficult to construct a complete three-dimensional model of the karst cave. This limitation leads to significant deviations between the pile foundation design coordinates and the actual bearing layer location, resulting in construction quality problems such as insufficient pile length and unsupported pile ends. Especially when the thickness of the karst cave roof slab is less than 3 meters or the filling material is in a fluid plastic state, and the pile foundation needs to penetrate the karst cave or undergo reinforcement treatment, the error rate of manual judgment remains high. Second, during dynamic construction, changes in geological conditions such as karst fissures and groundwater seepage can lead to sudden engineering disasters, such as rapid mud loss, negative pressure collapse of the overburden layer, and abnormal drill bit slippage.

[0003] Existing technologies rely excessively on human experience and judgment, making it difficult to respond promptly to dynamic changes, resulting in delays in emergency response and significantly increasing the incidence of construction accidents. For example, when constructing steel casings in multi-layered karst cave areas, it is necessary to penetrate the cave roof layer by layer. However, in the face of sudden grout leakage, the casing sinking often lags behind the speed of disaster development, easily leading to serious drill burial accidents. These technical deficiencies severely restrict the safety and reliability of pile foundation construction in karst areas. Summary of the Invention

[0004] This invention provides a method for intelligent positioning and precise construction of pile foundations in karst areas, comprising:

[0005] Obtain a karst development feature dataset for the target engineering area, wherein the karst development feature dataset includes a three-dimensional geological exploration data set and a karst attribute data set;

[0006] The karst development characteristic dataset is preprocessed to obtain a karst preprocessed dataset, wherein the karst preprocessed dataset includes a three-dimensional geological exploration preprocessed data set and a karst attribute preprocessed data set;

[0007] Based on the feature extraction model, features are extracted from the three-dimensional geological exploration preprocessing data set to obtain a three-dimensional geological feature map;

[0008] The karst cavity identification model is used to identify the three-dimensional geological feature map and generate a karst cavity distribution heat map.

[0009] The karst preprocessing dataset is subjected to karst cave feature transformation processing to obtain a pile foundation construction feature dataset, wherein the pile foundation construction feature dataset includes a pile foundation coordinate dataset and a bearing layer feature dataset;

[0010] Based on the heat map of karst cavity distribution, the pile foundation construction feature dataset is constrained to generate a pile foundation constraint boundary map;

[0011] Based on the pile foundation positioning optimization algorithm and the pile foundation constraint boundary diagram, the final hole positioning coordinate set of the pile foundation is output.

[0012] Furthermore, the acquisition of the karst development characteristic dataset of the target engineering area includes:

[0013] The original geological exploration dataset is obtained as the initial exploration dataset, which includes a three-dimensional seismic wave data set and a ground-penetrating radar data set.

[0014] Based on the pre-defined pile foundation construction area, the three-dimensional seismic wave data set in the initial exploration set is spatially filtered to obtain the target area exploration set;

[0015] Based on the karst development intensity threshold, the geological radar data set of the target area is subjected to feature screening to obtain the karst feature data set.

[0016] Based on the coordinate range of the pile foundation construction, the karst feature data set is filtered to obtain the target location feature set;

[0017] Based on the karst attribute data set of the target location features, the data is divided into a three-dimensional geological exploration data set and a karst attribute data set.

[0018] Further, the preprocessing of the karst development characteristic dataset to obtain a karst preprocessed dataset includes:

[0019] For each three-dimensional geological exploration data point in the karst development feature dataset, perform the following steps:

[0020] Obtain the attribute fields of the three-dimensional geological exploration data;

[0021] Based on the attribute fields of the three-dimensional geological exploration data, a karst anomaly zone identifier is constructed;

[0022] Based on the karst anomaly zone identifiers, preprocessed geological exploration data is generated;

[0023] The preprocessed geological exploration data is converted into a preset data format to obtain three-dimensional geological exploration preprocessed data.

[0024] The various 3D geological exploration preprocessing data were integrated into a karst preprocessing dataset.

[0025] Furthermore, the feature extraction based on the feature extraction model is used to extract features from the three-dimensional geological exploration preprocessing data set to obtain a three-dimensional geological feature map, including:

[0026] Based on the karst identification feature dimensions, feature reconstruction is performed on each of the three-dimensional geological exploration preprocessing data in the three-dimensional geological exploration preprocessing data group to obtain exploration data that meet the feature dimensions.

[0027] Spatial interpolation is performed on the exploration data that meet the characteristic dimensions to obtain a three-dimensional geological feature map.

[0028] Further, the step of performing karst cave feature transformation processing on the karst preprocessing dataset to obtain a pile foundation construction feature dataset includes:

[0029] According to the pile foundation construction parameter standard, the karst attribute preprocessing data group in the karst preprocessing dataset is transformed to obtain attribute data that meet the construction parameter standard;

[0030] Coordinate encoding is performed on the pile foundation coordinate data and the bearing stratum feature data that meet the construction parameter standards to obtain the pile foundation coordinate dataset and the bearing stratum feature dataset.

[0031] Based on the pile foundation design specifications, the pile foundation coordinate dataset is calibrated to obtain the pile foundation construction coordinate dataset;

[0032] The pile foundation construction coordinate dataset and the bearing layer feature dataset are integrated into a pile foundation construction feature dataset.

[0033] Further, the step of outputting the final borehole positioning coordinate set of the pile foundation based on the pile foundation positioning optimization algorithm and the pile foundation constraint boundary diagram includes:

[0034] Based on the pile foundation diameter specifications, the pile foundation constraint boundary diagram is processed to obtain a pile foundation obstacle avoidance buffer zone diagram;

[0035] Based on the pile foundation obstacle avoidance buffer map, the pile foundation construction coordinate dataset is positionally optimized to generate an optimized pile foundation coordinate set.

[0036] By using the pile foundation positioning verification model, collision detection is performed on the optimized pile foundation coordinate set to obtain the verified pile foundation coordinate set.

[0037] The verification is converted from the pile foundation coordinate set to the final hole positioning coordinate output format to obtain the pile foundation final hole positioning coordinate set.

[0038] Furthermore, the step of identifying the three-dimensional geological feature map using a karst cavity identification model to generate a karst cavity distribution heat map includes:

[0039] Based on the karst development grading standard, a color mapping relationship for different karst rates is set; the three-dimensional geological feature map is continuously sliced ​​to obtain karst feature layers at each elevation.

[0040] Based on the parameters of the cavity identification model, the cavity probability of each karst characteristic layer at each elevation is calculated to generate a karst probability matrix.

[0041] The karst probability matrix is ​​converted into a three-dimensional thermal distribution map using the color mapping relationship.

[0042] Furthermore, the step of constraining the pile foundation construction feature dataset based on the karst cavity distribution heat map to generate a pile foundation constraint boundary map includes:

[0043] Extract the vertex coordinate set of high-risk karst areas from the karst cavity distribution heat map;

[0044] Based on a preset safety distance threshold, construct a constrained polygon corresponding to the coordinate set of each vertex.

[0045] Perform a spatial Boolean operation between the constrained polygon and the pile foundation design coordinates to obtain the constructible area of ​​the pile foundation.

[0046] Generate the boundary topology data of the constructable area of ​​the pile foundation to form the pile foundation constraint boundary diagram.

[0047] Furthermore, the step of performing collision detection on the optimized pile foundation coordinate set through the pile foundation positioning verification model includes:

[0048] Establish a cylindrical spatial model for pile foundation construction, wherein the diameter of the cylindrical spatial model is equal to 1.2 times the design diameter of the pile foundation;

[0049] Calculate the minimum net distance between adjacent pile foundation spatial models;

[0050] When the minimum clearance value is less than the preset safe clearance threshold, it is marked as a collision pile foundation;

[0051] Adjust the spatial coordinates of the colliding pile foundations until all minimum clearance values ​​are greater than the preset safe clearance threshold.

[0052] Furthermore, it also includes:

[0053] The drilling rig positioning system is controlled based on the final hole positioning coordinate set of the pile foundation to carry out precise drilling construction;

[0054] Real-time acquisition of borehole offset data to generate construction error distribution maps;

[0055] When the offset value in the construction error distribution map is greater than a preset threshold, an automatic correction command is triggered.

[0056] The matching degree between the corrected borehole coordinates and the final borehole positioning coordinates of the pile foundation is verified, and a pile foundation borehole quality report is output.

[0057] The embodiments of the present invention have at least the following beneficial effects:

[0058] 1. By acquiring and preprocessing the karst development feature dataset, and combining the feature extraction model and the karst cavity identification model, a heat map of karst cavity distribution can be accurately generated, thereby providing a detailed geological risk reference for pile foundation construction, effectively avoiding engineering accidents caused by karst cavities during pile foundation construction, and ensuring construction safety.

[0059] 2. Based on the heat map of karst cavity distribution, the pile foundation construction feature dataset is constrained to generate a pile foundation constraint boundary map. Combined with the pile foundation positioning optimization algorithm, the final hole positioning coordinate set of the pile foundation is output, realizing the intelligent and precise positioning of the pile foundation. This reduces the pile foundation positioning deviation caused by human factors or inaccurate traditional methods, improves the pile foundation construction quality, and ensures that the pile foundation bearing capacity meets the engineering requirements.

[0060] 3. This method covers the entire process from data acquisition and processing to construction positioning, forming a complete solution for pile foundation construction in karst areas. Compared with traditional construction methods, it can more systematically address the complex geological conditions in karst areas, reduce construction difficulty, shorten the construction cycle, and improve construction efficiency, which is of great significance for promoting engineering construction in karst areas. Attached Figure Description

[0061] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent from the following detailed description taken in conjunction with the accompanying drawings. Several embodiments of the invention are illustrated in the drawings by way of example and not limitation, wherein:

[0062] Figure 1 This is a flowchart illustrating a method for intelligent positioning and precise construction of pile foundations in karst areas, provided in an embodiment of the present invention. Detailed Implementation

[0063] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application. It should be noted that similar reference numerals and letters in the following drawings indicate similar items; therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0064] like Figure 1 As shown, this application proposes a method for intelligent positioning and precise construction of pile foundations in karst areas, comprising:

[0065] S1. Obtain the karst development feature dataset of the target engineering area, which includes a three-dimensional geological exploration data set and a karst attribute data set.

[0066] S2. Preprocess the karst development characteristic dataset to obtain the karst preprocessed dataset, which includes the three-dimensional geological exploration preprocessed data set and the karst attribute preprocessed data set.

[0067] S3. Based on the feature extraction model, feature extraction is performed on the three-dimensional geological exploration preprocessing data set to obtain a three-dimensional geological feature map.

[0068] S4. Using the karst cavity identification model, identify the three-dimensional geological feature map and generate a karst cavity distribution heat map;

[0069] S5. Perform karst cave feature transformation on the karst preprocessing dataset to obtain the pile foundation construction feature dataset, which includes the pile foundation coordinate dataset and the bearing layer feature dataset.

[0070] S6. Based on the heat map of karst cavity distribution, constrain the pile foundation construction feature dataset to generate a pile foundation constraint boundary map.

[0071] S7. Based on the pile foundation positioning optimization algorithm and the pile foundation constraint boundary diagram, output the pile foundation final hole positioning coordinate set.

[0072] It should be noted that the karst development characteristic dataset refers to a multi-source data set integrating 3D geological exploration and karst attribute parameters. Specifically, it can be obtained by combining 3D seismic wave detection and ground-penetrating radar scanning, and is used to construct a full-space geological model covering the construction area. This dataset provides high-resolution input for subsequent intelligent analysis, solving the problem of incomplete 3D modeling caused by traditional discrete data.

[0073] Preprocessing refers to noise filtering and format standardization of the raw exploration data. This can be achieved using outlier removal algorithms and a unified coordinate transformation protocol to ensure consistency in spatial coordinate systems and accuracy levels across data collected by different detection devices. Preprocessing eliminates the interference of data heterogeneity on model training and improves the reliability of subsequent feature extraction.

[0074] Specifically, the feature extraction model uses a 3D convolutional neural network to process the 3D geological exploration preprocessing data set, with the input being standardized grid data (resolution 0.2m × 0.2m × 0.5m). The model extracts geological features through progressive convolutional layers:

[0075] The first two layers use a 3×3×3 convolution kernel (stride 1) combined with the ReLU activation function to capture the rock strata boundary and fracture distribution;

[0076] The third layer uses a 5×5×5 convolution kernel (stride 2) to analyze macroscopic structures such as cave connectivity; the fourth layer performs global average pooling to compress data dimensions.

[0077] The fifth fully connected layer outputs a 256-dimensional feature vector, which forms a three-dimensional geological feature map.

[0078] The model was trained on 10,000 sets of labeled data. Its robustness was enhanced by random rotation (±15°) and Gaussian noise (standard deviation 0.05). The model was pre-trained using the public dataset OpenKIM and then fine-tuned for 50 rounds with target region data.

[0079] The optimizer used is Adam (learning rate 0.001, decay rate...). The loss function is the mean squared error (MSE). In practical applications, after preprocessing the data and inputting it into the model, geological feature maps are automatically output for use by the karst cavity identification model.

[0080] The karst cavity identification model is based on a probability density estimation algorithm for predicting karst cavity distribution. Specifically, it uses a Gaussian mixture model to perform probabilistic clustering of three-dimensional geological features, generating a thermal distribution map that quantifies the intensity of karst cavity development. This model transforms complex geological conditions into visible risk levels, providing a dynamic decision-making basis for pile foundation obstacle avoidance.

[0081] The karst cavity identification model is based on a 3D U-Net architecture, with the input being a 3D geological feature map generated by the feature extraction model. The encoder extracts deep features through four levels of downsampling (each level contains two 3×3×3 convolutional layers and 2×2×2 max pooling); the decoder symmetrically performs four levels of upsampling (transposed convolution and encoder features are concatenated to fuse boundary information); the output layer generates a karst probability matrix (value range [0,1]) through 1×1×1 convolution and the Sigmoid function.

[0082] During implementation, the probability matrix is ​​converted into a three-dimensional thermogram according to a preset karst development grading standard:

[0083] A probability value ≤ 0.3 is mapped to blue (low risk).

[0084] A score of 0.3–0.7 indicates a yellow to orange level (medium risk).

[0085] A value ≥0.7 is considered red (high risk).

[0086] Karst cave feature conversion processing refers to the mathematical process of mapping geological parameters to engineering parameters. Specifically, the finite element method can be used to calculate the correspondence between the bearing capacity of the karst cave roof and the settlement of the pile foundation, realizing the standardized output of geological data to construction parameters such as pile diameter and pile length. This processing ensures that the pile foundation design parameters are accurately matched with the underground karst conditions.

[0087] A pile foundation constraint boundary map refers to a spatial restricted area topology generated based on a safety distance threshold. Specifically, the Voronoi diagram algorithm can be used to delineate the geometric boundaries between high-risk karst cave areas and constructible areas, forming rigid constraints on pile placement. This boundary map quantifies geological risks as construction restricted areas, preventing pile tips from falling into unreinforced karst cave areas.

[0088] The pile foundation positioning optimization algorithm refers to a coordinate iterative adjustment mechanism that considers multiple constraints. Specifically, a genetic algorithm can be used to globally optimize the pile position coordinates, maximizing the pile group layout density while meeting the net distance requirements and bearing capacity standards. This algorithm dynamically balances geological risks and engineering economics, solving the suboptimal problem of traditional manual pile placement.

[0089] The pile foundation positioning optimization algorithm is based on a genetic algorithm framework, using the pile foundation constraint boundary diagram as spatial constraints to optimize the pile foundation construction coordinate dataset. The algorithm initializes 100-200 sets of pile position coordinate solutions as a population, and the fitness function is defined as:

[0090]

[0091] in This is the minimum distance between the center of the pile foundation and the boundary of the high-risk karst cave. The effective bearing area of ​​the bearing stratum covered by the pile group. It is the sum of the Euclidean distances from the original design to the pile positions. The iterative process involves three steps:

[0092] Option: Retain the top 20% of elite solutions in terms of fitness;

[0093] Crossover: Swap 50% of the random pile coordinates with a probability of 0.7;

[0094] Mutation: A random offset of ±0.5 meters is applied to the coordinates of a single pile with a probability of 0.02. After each iteration, a cylindrical spatial model of the pile foundation (diameter = design pile diameter × 1.2) is established, and the minimum net distance between adjacent piles is detected. If the net distance is less than 0.8 meters, the colliding pile is moved 0.3-0.5 meters away using the gradient descent method. The optimization loop terminates when the fitness improvement rate is <1% for 10 consecutive generations or after 100 iterations, and the final borehole positioning coordinate set is output. In practice, the algorithm automatically reads the pile foundation constraint boundary diagram and construction coordinate data, and after optimization, generates executable instructions for the drilling rig, such as a CSV coordinate file.

[0095] Specifically, the first step is to acquire a dataset of karst development characteristics for the target engineering area, including a 3D geological exploration data set and a karst attribute data set. These datasets cover the geological structure and characteristics of the karst region.

[0096] Next, the karst development characteristic dataset is preprocessed to obtain a preprocessed karst dataset. The preprocessing process may include data cleaning, standardization, and other operations to improve the accuracy of subsequent analyses.

[0097] Then, based on the feature extraction model, features are extracted from the 3D geological exploration preprocessing data set to generate a 3D geological feature map. This step aims to extract key geological features from the preprocessed data to form a visualized 3D map.

[0098] A karst cavity identification model is used to identify three-dimensional geological features and generate a heat map showing the distribution of karst cavities. This heat map visually displays the distribution and risk level of karst cavities.

[0099] The karst preprocessing dataset undergoes karst cave feature transformation to obtain a pile foundation construction feature dataset. This step converts geological features into engineering parameters, including a pile foundation coordinate dataset and a bearing stratum feature dataset.

[0100] Based on the heat map of karst cavity distribution, the characteristic dataset of pile foundation construction is constrained to generate a pile foundation constraint boundary map. This step determines the safety boundary of pile foundation construction.

[0101] Finally, based on the pile foundation positioning optimization algorithm and the pile foundation constraint boundary diagram, the final borehole positioning coordinate set of the pile foundation is output. This step comprehensively considers geological conditions and engineering requirements to obtain the optimal pile foundation positioning scheme.

[0102] The entire process, driven by data and intelligent algorithms, achieves precise and automated positioning of pile foundations in karst areas, effectively solving the problems of insufficient spatial positioning accuracy of karst caves and delayed disaster response during construction in traditional methods.

[0103] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0104] When acquiring a dataset of karst development characteristics for the target engineering area, a combination of 3D seismic exploration and ground-penetrating radar (GPR) scanning can be used. 3D seismic exploration provides information on a wide range of underground structures, while GPR scanning provides higher-resolution near-surface karst details.

[0105] When preprocessing datasets of karst development characteristics, filtering algorithms can be used to remove noise, and interpolation methods can be used to fill in data gaps. The preprocessed datasets should have a uniform format and resolution.

[0106] The feature extraction model can employ a convolutional neural network, which, trained on karst geological data, can effectively identify karst features. Extracted features include the shape, size, and connectivity of karst caves.

[0107] Karst cavity identification models can be based on machine learning algorithms, such as support vector machines or random forests, to identify new cavities by learning the characteristics of known karst cavities. The identification results are presented in the form of a heatmap, where the intensity of the color indicates the probability of the cavity's existence.

[0108] The transformation of karst features can be achieved by establishing a mathematical model to convert karst properties such as cavity size, depth, and filling material properties into pile foundation design parameters, such as pile length, diameter, and bearing capacity.

[0109] During the constraint process, safety thresholds can be set. For example, when the heat map shows that the degree of karst development in a certain area exceeds the preset threshold, the area is marked as a prohibited construction area.

[0110] Pile foundation positioning optimization algorithms can employ genetic algorithms or particle swarm optimization algorithms to find the optimal pile foundation layout scheme while satisfying constraints. The objective function of the algorithm can include factors such as minimizing the total pile length and balancing the pile stress.

[0111] This application further proposes obtaining the original geological exploration dataset as the initial exploration set, which includes a 3D seismic wave data set and a ground-penetrating radar data set; spatially filtering the 3D seismic wave data set in the initial exploration set according to the pre-defined pile foundation construction area to obtain the target area exploration set; feature filtering of the ground-penetrating radar data set in the target area exploration set according to the karst development intensity threshold to obtain the karst feature data set; location filtering of the karst feature data set according to the pile foundation construction coordinate range to obtain the target location feature set; and dividing the karst attribute data set in the target location feature set into a 3D geological exploration data set and a karst attribute data set.

[0112] The pile foundation construction area is defined as a polygonal coordinate range, for example, the set of boundary coordinate points extracted from the pile foundation layout diagram in the construction drawings. When spatially filtering the 3D seismic wave data set, only seismic wave data within a 5-10 meter range extending beyond the boundary of the pile foundation construction area are retained; data exceeding this range is discarded. The karst development intensity threshold is quantified into a numerical range of 0.5-0.8; signals with reflected wave amplitudes below this threshold in the ground-penetrating radar data set are marked as invalid data. The pile foundation construction coordinate range is determined by the coordinate extreme values ​​of the pile foundation design plan; data from the karst characteristic data set whose cave vertex coordinates exceed this range are excluded. The division between the 3D geological exploration data set and the karst attribute data set is achieved through data field classification; for example, stratigraphic density and wave velocity parameters are classified into the 3D geological exploration data set, while cave size and filling type are classified into the karst attribute data set.

[0113] Specifically, after the original geological exploration dataset is acquired, the 3D seismic data set is first spatially filtered based on the pile foundation construction area. For example, if the pile foundation construction area is defined as a square area with a side length of 200 meters, only data within an 8-meter radius outside the square boundary is retained in the 3D seismic data set, and the rest is filtered out. Subsequently, the reflected wave amplitude in the ground-penetrating radar data set is calculated and compared with a karst development intensity threshold. For example, data with a reflected wave amplitude below 0.6 is identified as low-intensity karst and removed from the target area exploration set. The karst feature data set, after feature filtering, is further matched with the pile foundation construction coordinate range. For example, if the pile foundation construction coordinate range is 1000-1200 meters on the X-axis and 500-700 meters on the Y-axis, only data with the coordinates of the karst cave apex within this range is retained. The final 3D geological exploration data set includes the filtered seismic velocity field data, and the karst attribute data set includes structured parameters such as karst cave height and filling density.

[0114] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0115] The raw geological exploration dataset is obtained as the initial exploration set. This dataset includes 3D seismic wave data and ground-penetrating radar (GPR) data. The 3D seismic wave data set may contain parameters such as P-wave velocity, S-wave velocity, and wave impedance. The GPR data set may contain parameters such as electromagnetic wave reflection time and amplitude.

[0116] Based on a pre-defined pile foundation construction area, spatial filtering is performed on the 3D seismic wave data set in the initial exploration set to obtain the target area exploration set. For example, the pile foundation construction area can be set as a 100m × 100m rectangular area, and only the 3D seismic wave data within this area can be retained.

[0117] Based on the karst development intensity threshold, feature filtering is performed on the concentrated ground-penetrating radar data set of the target area to obtain a karst feature data set. The karst development intensity threshold can be set to 0.5, and karst feature data with a development intensity greater than 0.5 are selected.

[0118] Based on the coordinate range of the pile foundation construction, the karst feature data set is filtered to obtain the target location feature set. The coordinate range of the pile foundation construction can be set as a circular area with a radius of 5m centered on the designed pile location, and only the karst feature data within this range is retained.

[0119] Based on the karst attribute data set concentrated on the target location characteristics, the data is divided into a 3D geological exploration data set and a karst attribute data set. The 3D geological exploration data set may contain information such as stratigraphic structure and lithological distribution. The karst attribute data set may contain information such as cave size and infill material properties.

[0120] This application further proposes to perform the following steps on each three-dimensional geological exploration data in the karst development feature dataset: obtaining the attribute fields of the three-dimensional geological exploration data; constructing karst anomaly zone identifiers based on the attribute fields of the three-dimensional geological exploration data; generating preprocessed geological exploration data based on the karst anomaly zone identifiers; converting the preprocessed geological exploration data into a preset data format to obtain three-dimensional geological exploration preprocessed data; and integrating each three-dimensional geological exploration preprocessed data into a karst preprocessed dataset.

[0121] The acquisition of attribute fields can include key geological parameters such as rock layer density, karst cave distribution, or groundwater permeability. Structured features are extracted from multidimensional data through a data parsing module. The construction of karst anomaly zone identifiers can employ logical rules or threshold judgment methods. For example, identifier generation is triggered when the karst cave distribution density exceeds 0.5 per cubic meter or the rock layer density is below 2.5 g / cm³. The boundary coordinates of the anomaly area are extended and marked using a spatial interpolation algorithm. During the generation of preprocessed geological exploration data, anomaly zone identifiers are encoded as independent data layers, forming a mapping relationship with the original exploration data. Missing attribute values ​​are supplemented using a weighted average of neighboring data. Preset data format conversion includes unifying heterogeneous data into standardized grid data, with a grid resolution set to 0.1 to 0.5 meters. Coordinate system conversion uses Gauss-Kruger projection parameters to align multi-source data. In the data integration stage, data associations are established using spatial indexing technology, storage efficiency is optimized using an octree data structure, and data integrity verification is achieved through hash value comparison.

[0122] Specifically, when acquiring attribute fields, the data parsing module separates lithological parameters, geophysical response values, and geometric coordinate information from the original exploration data. The geophysical response values ​​may include seismic wave reflection intensity or electromagnetic wave attenuation coefficient. During the construction of karst anomaly zone identifiers, areas with abrupt changes in reflection intensity are marked based on a cave boundary detection algorithm, and fractured zones are identified through spatial clustering of continuous low-density value areas. When generating preprocessed data, anomaly zone identifiers are converted into binary mask layers, noisy data is eliminated using a median filtering algorithm, and redundant data outside the anomaly zones is pruned to reduce computational load. In the data format conversion stage, discrete point cloud data collected by different devices are resampled into a voxel grid with a uniform resolution, coordinate system deviations are corrected using a feature point matching algorithm, and data conversion errors are controlled within the millimeter range.

[0123] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0124] The following steps are performed on each 3D geological exploration data point in the karst development feature dataset: First, obtain the attribute fields of the 3D geological exploration data, including key geological parameters such as lithology, density, and porosity. Then, construct karst anomaly zone identifiers based on these attribute fields, for example, by setting density thresholds or porosity ranges to identify potential karst cave areas. Next, generate preprocessed geological exploration data based on the karst anomaly zone identifiers, removing noisy data and filling in missing values. Convert the preprocessed geological exploration data to a preset data format, such as a unified grid data format or coordinate system. Finally, integrate the various 3D geological exploration preprocessed data points into a karst preprocessed dataset.

[0125] This application further proposes a feature extraction model for extracting features from a three-dimensional geological exploration preprocessing data set to obtain a three-dimensional geological feature map. This includes reconstructing the features of each data in the three-dimensional geological exploration preprocessing data set according to the karst identification feature dimension to obtain exploration data that meets the feature dimension, and performing spatial interpolation on these exploration data to obtain a three-dimensional geological feature map.

[0126] Feature reconstruction transforms heterogeneous parameters such as seismic wave reflection intensity and ground-penetrating radar dielectric constant into standardized feature vectors. For example, seismic wave amplitude values ​​are normalized to the 0-1 range, and the ground-penetrating radar dielectric constant is converted into relative dielectric constant differences. Spatial interpolation employs the Kriging algorithm, with the variogram set to a spherical model, nugget values ​​controlled between 0.05 and 0.1, and the maximum hysteresis distance set to three times the distance between exploration points, thus integrating regional geological trends and local variation characteristics. The reconstructed data must meet preset dimensional alignment conditions, such as mapping the 128-dimensional features of 3D seismic wave data and the 64-dimensional features of ground-penetrating radar data to a unified 256-dimensional feature space through a fully connected layer. During spatial interpolation, the weight allocation of adjacent exploration points uses an inverse distance squared weighting method, and weight calculation is automatically cut off when the distance between two points exceeds a critical value of 50 meters.

[0127] Specifically, the feature reconstruction stage converts physical quantities output by different exploration equipment into dimensionless feature values ​​by establishing a feature mapping table. For example, the seismic wave reflection time difference Δt and the rock mass wave velocity v are converted into velocity disturbance values ​​using the formula Δv=Δt / (2d), where d is the exploration depth reference value. The spatial interpolation stage employs adaptive grid partitioning. When a known fault exists in the interpolation area, the grid density is automatically increased to twice the conventional density, and independent interpolation domains are set on both sides of the fault. In the three-dimensional geological feature map formed through dual processing, the gradient change rate of the karst cave boundary is enhanced to the order of 0.3-0.5 / meter, enabling accurate capture of karst cave outlines with diameters greater than 1.5 meters during subsequent heat map generation.

[0128] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0129] Feature extraction is performed on a 3D geological exploration preprocessing data set based on a feature extraction model to obtain a 3D geological feature map. Specifically, firstly, based on the karst identification feature dimensions, features of each 3D geological exploration preprocessing data set are reconstructed to obtain exploration data that meet each feature dimension. For example, heterogeneous parameters such as seismic wave reflection intensity and ground-penetrating radar dielectric constant can be converted into standardized feature vectors. Further, spatial interpolation is performed on each exploration data that meets the feature dimensions to obtain the 3D geological feature map. The spatial interpolation can employ Kriging interpolation, integrating regional geological trends and local variation characteristics, so that the karst cave boundaries exhibit clear gradient changes in the 3D map.

[0130] This application further proposes to perform karst cave feature transformation processing on the karst pretreatment dataset to obtain a pile foundation construction feature dataset, including: performing parameter transformation on the karst attribute pretreatment data group in the karst pretreatment dataset according to the pile foundation construction parameter standard to obtain attribute data that meets the construction parameter standard; performing coordinate encoding on the pile foundation coordinate data and the bearing layer feature data that meet the construction parameter standard to obtain a pile foundation coordinate dataset and a bearing layer feature dataset; performing coordinate calibration on the pile foundation coordinate dataset according to the pile foundation design specifications to obtain a pile foundation construction coordinate dataset; and integrating the pile foundation construction coordinate dataset and the bearing layer feature dataset into a pile foundation construction feature dataset.

[0131] The parameter conversion process includes converting the diameter of karst caves and the elastic modulus of infill material in the karst attribute data into pressure units required for pile foundation bearing capacity calculations. For example, the cave diameter is converted from meters to a multiple of the pile diameter, and the infill material velocity is converted from centimeters per second to a matching value for the mud pump discharge rate. Coordinate encoding employs a dual mapping between the geographic coordinate system and the local coordinate system of the pile foundation. For instance, the WGS84 coordinates of the cave vertices are converted to Cartesian coordinates of the construction area, while preserving millimeter-level accuracy of the elevation data. Coordinate calibration eliminates the deviation between the construction coordinate system and the design coordinate system through a coordinate translation matrix. For example, when the pile foundation design specifications require a pile spacing error of less than 5 centimeters, sub-pixel-level interpolation calibration is performed on the coordinate dataset. During the integration process, the bearing layer feature dataset and the pile foundation construction coordinate dataset are associated through spatial indexing, such as using an R-tree index to achieve fast query matching.

[0132] Specifically, the geometric parameters of karst caves in the karst attribute preprocessing data set are first converted into equivalent diameters under the pile foundation bearing capacity calculation standard. For example, the karst cave diameter is divided by a safety factor of 1.2 to obtain the pile foundation design diameter. The compressive strength of the rock strata in the bearing layer characteristic data is normalized to megapascals and mapped to the pile foundation concrete strength grade. During coordinate encoding, the elevation data of the karst cave top is converted into the bearing layer elevation of the pile foundation. For example, when the karst cave top is buried at a depth of 15 meters and the designed pile length is 20 meters, the bearing layer elevation is set to -35 meters. During coordinate calibration, the planar position of the pile foundation coordinate dataset is fitted to the construction control network using the least squares method. For example, piles with coordinate deviations greater than 3 cm are translated and corrected. The integrated pile foundation construction characteristic dataset is stored using a key-value pair structure, where the pile foundation construction coordinates serve as the primary key to associate with the bearing layer characteristic data, enabling real-time retrieval and verification during construction. Therefore, through multi-level data conversion and spatial calibration, the karst geological features and pile foundation construction parameters are accurately mapped, solving the problem of pile position deviation exceeding the allowable error caused by incompatibility of parameter systems in traditional methods. This improves the accuracy of pile foundation coordinate positioning to the centimeter level and increases the accuracy of bearing layer feature matching to over 95%.

[0133] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0134] A karst preprocessing dataset was processed to transform karst cavity features, resulting in a pile foundation construction feature dataset. First, according to pile foundation construction parameter standards, the karst attribute preprocessing data groups in the karst preprocessing dataset underwent parameter transformation. For example, karst cavity volume data was converted into bearing capacity parameters required for pile foundation design, and karst development degree was converted into foundation treatment coefficients. Second, coordinate encoding was performed on the transformed pile foundation coordinate data and bearing stratum feature data. Specifically, the UTM projection coordinate system was used to convert geographic coordinates into plane rectangular coordinates, and a unique code was assigned to each pile location. Further, the pile foundation coordinate dataset was calibrated according to the pile foundation design specifications. For example, the pile foundation coordinates were aligned with the construction machinery operating system through coordinate translation and rotation. Finally, the calibrated pile foundation construction coordinate dataset and bearing stratum feature dataset were integrated into a single pile foundation construction feature dataset.

[0135] This application further proposes to perform buffer boundary processing on the pile foundation constraint boundary diagram according to the pile foundation diameter specification to obtain a pile foundation obstacle avoidance buffer diagram; to perform position optimization on the pile foundation construction coordinate dataset based on the pile foundation obstacle avoidance buffer diagram to generate an optimized pile foundation coordinate set; to perform collision detection on the optimized pile foundation coordinate set through the pile foundation positioning verification model to obtain a verified pile foundation coordinate set; and to convert the verified pile foundation coordinate set into the final hole positioning coordinate output format to obtain the pile foundation final hole positioning coordinate set.

[0136] Buffer boundary processing creates a safe operating space by extending the pile foundation constraint boundary outwards. The extension amount is dynamically adjusted according to the pile diameter; for example, when the pile diameter is 1.2 meters, the buffer zone extension is set to 0.6 meters. Position optimization uses a gradient descent algorithm to iteratively adjust the pile foundation coordinates within the obstacle avoidance buffer map, ensuring that the distance between the pile center point and the boundary of the high-risk karst cave area is always greater than the safety threshold. Collision detection calculates the minimum clearance between adjacent pile foundations by establishing an enlarged diameter cylindrical spatial model. When the pile diameter is 1.5 meters, the cylindrical model diameter is expanded to 1.8 meters, and the safe clearance threshold is set to 0.8 meters. Format conversion uses the mapping rules between the WGS84 coordinate system and the local coordinate system of the construction drilling rig to achieve lossless coordinate data transmission.

[0137] Specifically, based on the pile foundation constraint boundary diagram, the buffer expansion amount is first determined according to the pile diameter specifications, for example, 0.5 times the pile diameter is used as the buffer radius, and the original boundary is expanded outward to form an obstacle avoidance buffer diagram. Subsequently, the initial coordinate points in the pile foundation construction coordinate dataset are imported into the obstacle avoidance buffer diagram. Through coordinate optimization algorithms, the pile position layout is optimized while avoiding high-risk areas of karst caves, so that the spacing between adjacent pile foundations meets the drilling rig operation requirements. The optimized coordinate set is input into the pile foundation positioning verification model. This model simulates the actual construction site by constructing a cylindrical spatial model with an enlarged diameter, calculates the minimum clear distance between the outer walls of adjacent models, and automatically triggers a secondary coordinate adjustment when insufficient clear distance is detected. Finally, the verified coordinate data is converted into a binary encoding format recognizable by the drilling rig control system through the coordinate transformation module, such as using the IEEE 754 floating-point standard for coordinate value storage, to achieve direct interface between the final hole positioning data and the construction equipment.

[0138] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0139] Based on the pile foundation diameter specifications, the pile foundation constraint boundary diagram is processed to obtain a pile foundation obstacle avoidance buffer diagram. Specifically, the buffer analysis method in GIS spatial analysis can be used, taking the boundary line in the pile foundation constraint boundary diagram as a reference and extending it outward by a certain distance to form a buffer zone. The buffer distance can be set to 1.5 times the pile foundation diameter to ensure sufficient construction space.

[0140] Based on the pile foundation obstacle avoidance buffer map, the location of the pile foundation construction coordinate dataset is optimized to generate an optimized pile foundation coordinate set. Furthermore, a genetic algorithm can be used to further optimize the pile foundation coordinates. First, the initial pile foundation coordinates are used as a population, and the fitness function is defined as the minimum distance between the pile foundation and the buffer boundary. New coordinate schemes are generated through crossover and mutation operations, and iterative optimization continues until preset conditions are met.

[0141] By verifying the pile foundation positioning model, collision detection is performed on the optimized pile foundation coordinate set to obtain the verified pile foundation coordinate set. Based on this, a cylindrical spatial model for pile foundation construction can be established, with the cylinder diameter set to 1.2 times the designed pile foundation diameter. The minimum clear distance between adjacent pile foundation spatial models is calculated. When the minimum clear distance is less than a preset safe clear distance threshold, it is marked as a colliding pile foundation and adjustments are made.

[0142] The verified pile foundation coordinate set is converted into the final borehole positioning coordinate output format to obtain the pile foundation final borehole positioning coordinate set. For example, the coordinate data can be converted into a CSV format that construction equipment can recognize, containing information such as pile number, X coordinate, Y coordinate, and Z coordinate.

[0143] This application further proposes to set color mapping relationships for different karst rates according to karst development grading standards, to continuously slice the three-dimensional geological feature map to obtain karst feature layers at each elevation, to calculate the cavity probability of each karst feature layer based on cavity identification model parameters to generate a karst probability matrix, and to convert the karst probability matrix into a three-dimensional thermal distribution map through color mapping relationships.

[0144] The color mapping relationship uses a karst rate threshold to define color gradients; for example, a 10% increase in karst rate corresponds to a change in color gradation, creating a transition from cool to warm colors. Continuous slicing processes layer the 3D geological feature map at preset elevation intervals (0.5 meters or 1 meter), ensuring each karst feature layer corresponds to an independent elevation range. The cavity probability calculation uses convolutional neural network model parameters to identify features in each slice and output a probability value; a probability threshold of 0.7 indicates a high-risk area. The 3D thermal distribution map uses voxel rendering technology to fuse the probability matrix with the color gradient, where the color depth of each voxel is positively correlated with the karst probability.

[0145] Specifically, after the 3D geological feature map is segmented into multiple horizontal karst feature layers, the karst development characteristics within each layer are analyzed independently. The parameters of the cavity identification model are optimized through training data; for example, karst sample data is used to train the convolutional kernel weights, improving the model's accuracy in identifying cavity boundaries to over 92%. Each element in the karst probability matrix represents the probability of a cavity existing at the corresponding coordinate point. When the probability value is higher than 0.7, it is marked as a high-risk area using a red color gradient. The 3D thermal distribution map, by overlaying the color mapping results of each elevation layer, forms a spatially continuous risk distribution visualization model. Thus, construction personnel can quickly identify high-risk karst areas at different depths through the color block distribution; for example, red areas correspond to densely populated karst areas that pile foundations need to avoid, while blue areas represent safe construction zones.

[0146] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0147] Based on the karst development grading standard, a color mapping relationship is set for different karst ratios. For example, 0-20% karst ratio can be mapped to green, 20%-40% to yellow, 40%-60% to orange, 60%-80% to red, and 80%-100% to dark red.

[0148] The three-dimensional geological feature map is continuously sliced ​​to obtain karst feature layers at each elevation. Specifically, slices can be taken every 0.5 meters along the vertical direction to generate a series of two-dimensional planar images.

[0149] Based on the cavity identification model parameters, the cavity probability is calculated for each karst feature layer at each elevation, generating a karst probability matrix. The cavity identification model can employ a convolutional neural network, with model parameters optimized using a training dataset. Convolution operations are performed on the image of each elevation layer to extract features and calculate the cavity probability value for each pixel, forming a probability matrix.

[0150] By using color mapping, the karst probability matrix is ​​converted into a three-dimensional thermal distribution map. Furthermore, 3D visualization software can be used to overlay the probability matrices of each elevation layer and apply color mapping to generate a three-dimensional thermal distribution map. This allows engineers to visually observe the degree of karst development at different depths and locations.

[0151] This application further proposes to extract the vertex coordinate set of high-risk karst areas from the karst cavity distribution heat map; construct the constraint polygon corresponding to each vertex coordinate set according to the preset safety distance threshold; perform spatial Boolean operation on the constraint polygon and the pile foundation design coordinates to obtain the pile foundation construction area; and generate the boundary topology data of the pile foundation construction area to form the pile foundation constraint boundary map.

[0152] The extraction of vertex coordinate sets can be achieved using the Delaunay triangulation algorithm or the α-shape algorithm. The contour capture accuracy is controlled by setting a vertex density threshold; for example, sampling points are automatically increased when the radius of curvature of the cave edge is less than 0.5 meters. The construction of the constrained polygon is achieved using a spatial buffering algorithm. The safety distance threshold can be set to 1.5 to 2 times the pile diameter, and the buffering direction is dynamically adjusted according to the inclination angle of the cave roof. Spatial Boolean operations are performed using a BSP tree structure for 3D spatial cutting, with computational accuracy controlled within ±5 cm. Boundary topology data generation employs NURBS surface fitting technology, outputting 3D vector data in a format containing closed loops and hole markers.

[0153] Specifically, in the karst cavity distribution heatmap, the outlines of high-risk areas are identified using an edge detection algorithm, and a discrete vertex set is extracted using a moving cube algorithm. After planar projection, the vertex coordinates are used to generate initial polygons using a convex hull algorithm, and then the vertex density is adjusted based on the karst cavity roof thickness distribution data. A constraint polygon is then extended outwards using a 3D buffer algorithm to form a safety barrier layer, where the vertical buffer distance is set to 0.8 times the karst cavity roof thickness, and the horizontal buffer distance is dynamically calculated based on the pile diameter. During spatial Boolean operations, the pile design coordinates are converted into a cylindrical spatial model, and after a difference operation with the constraint polygon, the coordinates of piles completely outside the safe zone are retained. The final generated boundary topology data contains multi-level LOD details and can be adapted to drilling rig positioning systems of varying precision.

[0154] As a preferred embodiment, the specific implementation of this application is as follows: Identify the boundary vertices of high-risk areas from the karst cavity distribution heat map, and obtain the vertex coordinate set through a spatial coordinate extraction algorithm; set a safety distance threshold based on engineering safety specifications, and generate a constrained polygon by expanding along the normal vector direction of each vertex in a three-dimensional coordinate system; input the pile foundation design coordinates into the spatial calculation engine, perform Boolean difference set operation between the constrained polygon and the pile foundation coordinates, and automatically eliminate pile positions that spatially overlap with the high-risk karst cavity area; extract the closed polygon boundary of the constructable area through a boundary topology generation algorithm, and output vector graphic data containing vertex sequences and topological relationships.

[0155] This application further proposes to perform collision detection on the optimized pile foundation coordinate set through a pile foundation positioning verification model, including establishing a cylindrical spatial model of the pile foundation construction, wherein the diameter of the cylindrical spatial model is equal to 1.2 times the design diameter of the pile foundation; calculating the minimum net distance between adjacent pile foundation spatial models; marking a colliding pile foundation when the minimum net distance is less than a preset safe net distance threshold; and adjusting the spatial coordinates of the colliding pile foundation until all minimum net distance values ​​are greater than the preset safe net distance threshold.

[0156] The diameter expansion ratio of the cylindrical spatial model can be dynamically adjusted based on the accuracy of the construction equipment or geological conditions. For example, a 1.25-fold expansion coefficient is used in karst fissure areas, and a 1.15-fold expansion coefficient is used in stable strata. The minimum net distance between adjacent models is obtained through three-dimensional spatial vector calculation, specifically by subtracting the sum of their radii from the shortest distance between the axes of the two cylinders. The safe net distance threshold is set according to the pile foundation construction specifications; for example, a safe net distance of 0.8m corresponds to a pile foundation with a diameter of 1.5m. The coordinate adjustment process uses a gradient descent algorithm, which iteratively calculates to move colliding pile foundations along the optimal path, avoiding new collisions.

[0157] Specifically, the construction of the cylindrical spatial model for pile foundation construction involves expanding the design diameter by 20% to cover the maximum offset range caused by drill bit oscillation. For example, for a pile foundation with a design diameter of 1.0m, the spatial model diameter is set to 1.2m, which can absorb drill rig guidance error of ±0.1m and karst strata compression deformation of 0.05m. When calculating the minimum clearance value, a spatial geometric algorithm is used to solve for the shortest distance between the axes of the two cylinders, and the expanded radius value is deducted. For example, when the distance between the two axes is 3.0m, the actual clearance is 3.0m - (0.6m + 0.6m) = 1.8m. When the clearance is detected to be lower than the safety threshold, a coordinate translation algorithm is used to move the colliding pile foundation to a low-density area, with a movement step size set to 10%-15% of the pile foundation diameter, for example, adjusting 0.12m-0.18m each time. During the adjustment process, the topological relationship of the spatial model is updated in real time until the clearance between all adjacent pile foundations reaches the safety standard, eliminating the risk of equipment interference.

[0158] As a preferred embodiment, the specific implementation of this application is as follows: During the pile foundation positioning verification process, a cylindrical spatial model is first constructed based on the pile foundation design diameter. The diameter of this model is set to 1.2 times the pile foundation design diameter to simulate the dynamic offset range that may occur during actual construction. Subsequently, the minimum net distance between the cylindrical spatial models corresponding to adjacent pile foundations is obtained through three-dimensional spatial geometric calculation, using a bounding box collision detection algorithm for rapid distance estimation. When the minimum net distance is detected to be lower than a preset safe net distance threshold, the corresponding pile foundation coordinates are automatically marked as colliding pile foundations. The coordinate adjustment of colliding pile foundations is achieved through an iterative optimization algorithm, specifically using the gradient descent method to fine-tune the pile foundation coordinates. After each adjustment, the minimum net distance between adjacent pile foundations is recalculated until the spacing between all pile foundations meets the safety requirements. During the adjustment process, priority is given to maintaining the overall layout stability of the pile foundation group, and coordinate offset correction is only performed on local conflict areas.

[0159] This application further proposes a technical solution to control the drilling rig positioning system to carry out precise drilling construction based on the pile foundation final hole positioning coordinate set, collect drilling offset data in real time to generate a construction error distribution map, trigger an automatic correction command when the offset value is greater than a preset threshold, and verify the matching degree between the corrected drilling coordinates and the final hole positioning coordinates to output a pile foundation hole formation quality report.

[0160] The control commands for the drilling rig positioning system are generated by converting the final hole positioning coordinate set into mechanical drive parameters based on a coordinate transformation algorithm. For example, the drill rod travel angle and depth are calculated using a coordinate translation matrix. The construction error distribution map is generated by real-time data acquisition from tilt and displacement sensors installed on the drill rod. Specifically, the tilt sensor monitors the drill bit deflection angle with an accuracy of 0.1°, and the displacement sensor records the drill rod pressure with millimeter-level resolution. The data is then processed by Kalman filtering to generate a three-dimensional error vector field. The preset thresholds can be dynamically adjusted based on the characteristics of karst formations. For example, in areas with dense karst caves, the horizontal offset threshold is set to 5cm and the vertical offset threshold to 3cm, while in stable bedrock areas, these thresholds are relaxed to 8cm and 5cm, respectively. The automatic correction command is triggered by generating a reverse compensation amount through a PID control algorithm. This is specifically manifested in adjusting the pressure distribution ratio of the hydraulic propulsion system. For example, when the horizontal offset reaches the threshold, the oil supply pressure ratio of the hydraulic cylinder on the opposite side of the offset direction is increased to 65%-70%. The matching degree verification uses the least squares method to calculate the sum of squared residuals between the corrected coordinates and the design coordinates. When the residual is less than 0.5cm, it is judged as qualified, and a quality report containing a coordinate deviation heatmap and compliance rating is automatically generated.

[0161] Specifically, during the drilling rig startup phase, the final hole positioning coordinate set is transmitted to the drilling rig control system via an industrial bus, driving the servo motor to adjust the initial angle of the drill rod to the designed axis. During drilling, the multi-axis inertial measurement unit captures the drill bit's movement trajectory at a 50Hz sampling rate. After data preprocessing, a real-time updated error distribution map is generated, displaying the cumulative offset at different depths in contour lines. When the horizontal offset of a certain section of the borehole exceeds a 5cm threshold, the control system immediately interrupts drilling and initiates a correction program. This is achieved by adjusting the hydraulic extension of the drill rod guide to generate a counterforce; for example, if the offset is 6cm in the positive X-axis direction, the guide applies a 2MPa correction pressure in the negative X-axis direction for 30 seconds. After correction is completed, the laser ranging module built into the drill rod re-acquires coordinate data and compares it point-by-point with the final hole positioning coordinates, generating an evaluation report containing the maximum deviation value, average deviation value, and pass rate indicators.

[0162] Through the above technical solution, this application effectively solves the problem of borehole location deviation caused by dynamic changes in karst strata. By employing a closed-loop control mechanism of real-time monitoring and automatic correction, the borehole trajectory is dynamically corrected to within the design allowable deviation range, avoiding the accumulation of errors caused by delayed manual intervention. Simultaneously, a coordinate matching verification method is used to replace traditional sampling inspection, ensuring that the contact area of ​​each pile foundation borehole meets the requirements of the bearing stratum, fundamentally improving the bearing stability and geological disaster resistance of the pile foundation structure.

[0163] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for intelligent positioning and precise construction of pile foundations in karst areas, characterized in that, include: Obtain a karst development feature dataset for the target engineering area, wherein the karst development feature dataset includes a three-dimensional geological exploration data set and a karst attribute data set; The karst development characteristic dataset is preprocessed to obtain a karst preprocessed dataset, wherein the karst preprocessed dataset includes a three-dimensional geological exploration preprocessed data set and a karst attribute preprocessed data set; Based on the feature extraction model, features are extracted from the three-dimensional geological exploration preprocessing data set to obtain a three-dimensional geological feature map; The karst cavity identification model is used to identify the three-dimensional geological feature map and generate a karst cavity distribution heat map. The karst preprocessing dataset is subjected to karst cave feature transformation processing to obtain a pile foundation construction feature dataset, wherein the pile foundation construction feature dataset includes a pile foundation coordinate dataset and a bearing layer feature dataset; Based on the heat map of karst cavity distribution, the pile foundation construction feature dataset is constrained to generate a pile foundation constraint boundary map; Based on the pile foundation positioning optimization algorithm and the pile foundation constraint boundary diagram, output the pile foundation final hole positioning coordinate set; The step of performing karst cave feature transformation processing on the karst preprocessing dataset to obtain a pile foundation construction feature dataset includes: Based on the pile foundation construction parameter standards, the karst properties in the karst pretreatment data are pre-defined. The data sets are processed and transformed to obtain attribute data that meet the construction parameter standards. Coordinate encoding is performed on the pile foundation coordinate data and the bearing stratum feature data that meet the construction parameter standards to obtain the pile foundation coordinate dataset and the bearing stratum feature dataset. Based on the pile foundation design specifications, the pile foundation coordinate dataset is calibrated to obtain the pile foundation construction coordinate dataset; The pile foundation construction coordinate dataset and the bearing layer feature dataset are integrated into a pile foundation construction feature dataset.

2. The method according to claim 1, characterized in that, The acquisition of the karst development characteristic dataset of the target engineering area includes: The original geological exploration dataset is obtained as the initial exploration dataset, which includes a three-dimensional seismic wave data set and a ground-penetrating radar data set. Based on the pre-defined pile foundation construction area, the three-dimensional seismic wave data set in the initial exploration set is spatially filtered to obtain the target area exploration set; Based on the karst development intensity threshold, the geological radar data set of the target area is subjected to feature screening to obtain the karst feature data set. Based on the coordinate range of the pile foundation construction, the karst feature data set is filtered to obtain the target location feature set; Based on the karst attribute data set of the target location features, the data is divided into a three-dimensional geological exploration data set and a karst attribute data set.

3. The method according to claim 1, characterized in that, The preprocessing of the karst development characteristic dataset to obtain a karst preprocessed dataset includes: For each three-dimensional geological exploration data point in the karst development feature dataset, perform the following steps: Obtain the attribute fields of the three-dimensional geological exploration data; Based on the attribute fields of the three-dimensional geological exploration data, a karst anomaly zone identifier is constructed; Based on the karst anomaly zone identifiers, preprocessed geological exploration data is generated; The preprocessed geological exploration data is converted into a preset data format to obtain three-dimensional geological exploration preprocessed data. The various 3D geological exploration preprocessing data were integrated into a karst preprocessing dataset.

4. The method according to claim 1, characterized in that, The feature extraction model is used to extract features from the three-dimensional geological exploration preprocessing data set to obtain a three-dimensional geological feature map, including: Based on the karst identification feature dimensions, feature reconstruction is performed on each of the three-dimensional geological exploration preprocessing data in the three-dimensional geological exploration preprocessing data group to obtain exploration data that meet the feature dimensions. Spatial interpolation is performed on the exploration data that meet the characteristic dimensions to obtain a three-dimensional geological feature map.

5. The method according to claim 1, characterized in that, The step of outputting the final borehole positioning coordinate set based on the pile foundation positioning optimization algorithm and the pile foundation constraint boundary diagram includes: Based on the pile foundation diameter specifications, the pile foundation constraint boundary diagram is processed to obtain a pile foundation obstacle avoidance buffer zone diagram; Based on the pile foundation obstacle avoidance buffer map, the pile foundation construction coordinate dataset is positionally optimized to generate an optimized pile foundation coordinate set. By using the pile foundation positioning verification model, collision detection is performed on the optimized pile foundation coordinate set to obtain the verified pile foundation coordinate set. The verification is converted from the pile foundation coordinate set to the final hole positioning coordinate output format to obtain the pile foundation final hole positioning coordinate set.

6. The method according to claim 1, characterized in that, The process of identifying the three-dimensional geological feature map using a karst cavity identification model and generating a karst cavity distribution heat map includes: Based on the karst development grading standard, a color mapping relationship is set for different karst rates; The three-dimensional geological feature map is continuously sliced ​​to obtain karst feature layers at each elevation; Based on the parameters of the cavity identification model, the cavity probability of each karst characteristic layer at each elevation is calculated to generate a karst probability matrix. The karst probability matrix is ​​converted into a three-dimensional thermal distribution map using the color mapping relationship.

7. The method according to claim 1, characterized in that, The process of constraining the pile foundation construction feature dataset based on the karst cavity distribution heat map to generate a pile foundation constraint boundary map includes: Extract the vertex coordinate set of high-risk karst areas from the karst cavity distribution heat map; Based on a preset safety distance threshold, construct a constrained polygon corresponding to the coordinate set of each vertex. Perform a spatial Boolean operation between the constrained polygon and the pile foundation design coordinates to obtain the constructible area of ​​the pile foundation. Generate the boundary topology data of the constructable area of ​​the pile foundation to form the pile foundation constraint boundary diagram.

8. The method according to claim 5, characterized in that, The step of verifying the pile foundation positioning model and performing collision detection on the optimized pile foundation coordinate set includes: Establish a cylindrical spatial model for pile foundation construction, wherein the diameter of the cylindrical spatial model is... It is equal to 1.2 times the design diameter of the pile foundation; Calculate the minimum net distance between adjacent pile foundation spatial models; When the minimum clearance value is less than the preset safe clearance threshold, it is marked as a collision pile foundation; Adjust the spatial coordinates of the colliding pile foundations until all minimum clearance values ​​are greater than the preset safe clearance threshold.

9. The method according to claim 1, characterized in that, Also includes: The drilling rig positioning system is controlled based on the final hole positioning coordinate set of the pile foundation to carry out precise drilling construction; Real-time acquisition of borehole offset data to generate construction error distribution maps; When the offset value in the construction error distribution map is greater than a preset threshold, an automatic correction command is triggered. The matching degree between the corrected borehole coordinates and the final borehole positioning coordinates of the pile foundation is verified, and a pile foundation borehole quality report is output.

Images