Geotechnical engineering intelligent reconnaissance system and method based on big data

Abstract

The invention discloses a geotechnical engineering intelligent investigation system and method based on big data. The system comprises a field sensing module, a multi-source fusion module, an intelligent analysis module, a modeling early warning module and a report generation module. The field sensing module collects drilling parameters, in-situ test data and geological images in real time; the multi-source fusion module receives an original investigation signal and accesses a historical engineering investigation database and a regional geology database; the intelligent analysis module receives the fused data signal, constructs a knowledge graph containing a rock-soil entity relationship, predicts rock-soil parameters and stratum distribution of an unexplored area based on a machine learning model, and generates a parameter prediction signal; the modeling early warning module receives the parameter prediction signal; and the report generation module receives the parameter prediction signal and the risk early warning signal. The geotechnical engineering intelligent investigation system and method based on big data can solve the problems of discrete investigation data, strong experience dependence and low intelligent degree.

Description

Technical Field

[0001] The present invention relates to the technical field of geotechnical engineering investigation and big data artificial intelligence, and in particular to a geotechnical engineering intelligent investigation system and method based on big data. Background Art

[0002] Traditional geotechnical engineering surveys rely on on-site drilling sampling, geotechnical tests and manual interpretation, and have three inherent defects: first, the survey data is distributed discretely in a point-like manner, and the inference of strata between boreholes relies on the experience of engineers, which makes misjudgment prone to complex geological conditions; second, historical project data, regional geological information and real-time survey data are separated from each other, forming data islands that cannot support the mining of macro-laws; third, report preparation requires manual data collation, drawing and text description, which is time-consuming and has a low degree of standardization.

[0003] Existing technological improvements are limited: survey data management systems enable digital storage of borehole data but lack intelligent analysis capabilities; geographic information systems manage spatial data but struggle to process deep geotechnical parameters; and a few 3D modeling software packages rely on dense borehole data and lack automatic updating. While rudimentary expert systems based on simple rules have emerged, their knowledge bases are limited and unable to adapt to changing field conditions. In recent years, machine learning techniques have attempted to predict geotechnical parameters, but these models are often tailored to a single data type and lack deep integration with engineering knowledge systems.

[0004] The industry faces a core dilemma: On the one hand, urban underground space development faces increasingly complex geological environments, and traditional methods lack efficiency and precision. On the other hand, massive amounts of historical data are not being effectively utilized, and artificial intelligence technology has failed to systematically address issues throughout the entire survey process. There is an urgent need to build intelligent systems that integrate multi-source, heterogeneous data, embed domain knowledge, and cover the entire survey chain, achieving a fundamental shift from "experience-driven" to "data-driven." Summary of the Invention

[0005] In view of the shortcomings of the above-mentioned prior art, the purpose of the present invention is to provide a geotechnical engineering intelligent survey system and method based on big data, which is used to solve the problems of discrete survey data, strong reliance on experience, and low intelligence. The present invention uses an on-site perception module to collect drilling parameters, in-situ test data, and geological images in real time to form raw signals; a multi-source fusion module integrates historical databases and regional geological data to generate fused signals; an intelligent analysis module constructs a geotechnical knowledge map and uses machine learning to predict parameters in unexplored areas; a modeling and early warning module automatically constructs a three-dimensional geological model and identifies risks; and a report generation module outputs structured reports, enabling data-driven intelligent decision-making, replacing manual reliance on experience.

[0006] The present invention provides a geotechnical engineering intelligent survey system based on big data, comprising: On-site perception module: The on-site perception module collects drilling parameters, in-situ test data and geological images in real time to form original exploration signals; Multi-source fusion module: The multi-source fusion module receives the original survey signal and accesses the historical engineering survey database and regional geological database, and generates fused data signals through data cleaning and semantic association; Intelligent analysis module: The intelligent analysis module receives the fused data signal, constructs a knowledge graph containing geotechnical entity relationships, predicts geotechnical parameters and stratum distribution in unexplored areas based on machine learning models, and generates parameter prediction signals; Modeling and early warning module: The modeling and early warning module receives parameter prediction signals, automatically constructs a three-dimensional geological model, and simultaneously identifies geological risk patterns to generate risk early warning signals; Report generation module, the report generation module receives parameter prediction signals and risk warning signals, and outputs a structured exploration report containing geological descriptions, parameter charts and risk warnings.

[0007] In one embodiment of the present invention, the on-site perception module includes a drilling pressure sensor, an in-hole camera device and an in-situ tester. The drilling pressure sensor obtains pressure change data during the drilling process of the drilling rig in real time. The in-hole camera device collects images of the stratum structure of the borehole wall through an optical lens. The in-situ tester performs static penetration testing or standard penetration testing during drilling intervals. The above-mentioned devices send the collected data to the edge computing node through the Internet of Things communication protocol. The node performs timestamp synchronization and noise filtering on the multi-source asynchronous data to form a standardized original exploration signal.

[0008] In one embodiment of the present invention, the multi-source fusion module uses anomaly detection rules based on the geotechnical engineering ontology when performing data cleaning. The ontology defines the stratigraphic lithology classification system, the threshold range of physical and mechanical parameters, and the spatial relationship constraints of geological structures. At the same time, the natural language processing engine parses the text of historical survey reports, automatically extracts key attribute terms in the stratigraphic description paragraphs and associates them with the ontology entities, and geo-aligns the remote sensing image data in the regional geological database with the spatial coordinates in the original survey signal, ultimately generating a multi-dimensional fused data signal that integrates spatiotemporal attributes, semantic attributes, and engineering parameters.

[0009] In one embodiment of the present invention, the intelligent analysis module constructs a knowledge graph using stratigraphic units as nodes and lithologic gradient relationships, parameter statistical dependencies, and geological structural influence relationships as edges. The machine learning model adopts a hybrid architecture that integrates graph neural networks and spatial interpolation algorithms. The topological structural features of the knowledge graph are used to enhance the stratigraphic continuity modeling capability. When predicting unexplored areas, a probability distribution map of geotechnical parameters and a stratigraphic interface confidence surface are generated based on spatial location similarity and geological environment similarity, and a parameter prediction signal with an uncertainty quantification indicator is output.

[0010] In one embodiment of the present invention, the modeling and warning module adopts implicit surface reconstruction technology when constructing a three-dimensional geological model, converts the discrete prediction point set in the parameter prediction signal into a continuous geological interface, and constructs a fault zone spatial model by introducing fault strike and dip data. The geological risk pattern recognition includes real-time analysis of the drilling parameter mutation characteristics and abnormal texture of the borehole image, combined with the disaster trigger rule chain predefined in the knowledge graph, and generates a landslide risk warning signal when a sudden drop in drilling pressure is detected accompanied by the appearance of a broken zone feature in the borehole wall image. If the in-situ test data mutates and meets the liquefaction judgment criteria, a sand liquefaction warning signal is generated.

[0011] In one embodiment of the present invention, the three-dimensional geological model supports a dynamic update mechanism. When new exploration point data is input, local model reconstruction is automatically triggered. The machine learning model weights are adaptively adjusted by comparing the deviation values ​​between the predicted parameters and the actual measured parameters. After the geological risk warning signal is generated, it is pushed to the on-site mobile terminal in real time. The terminal displays a three-dimensional perspective rendering of the risk location and avoidance route suggestions.

[0012] In one embodiment of the present invention, the report generation module has a built-in structured template engine, which converts the stratigraphic distribution data in the parameter prediction signal into geological profile description text, automatically matches the standardized expressions in the standard terminology library, calls the preset risk response plan text fragment when the risk warning signal is triggered, and uses dynamic contour rendering technology to generate parameter charts. Vectorized geotechnical parameter spatial distribution maps are output according to the display range and accuracy requirements set by the user, and finally all elements are integrated to generate a structured survey report that conforms to the industry standard format.

[0013] In one embodiment of the present invention, a structured template engine integrates an expert experience rule base, automatically inserts design considerations when the predicted parameters are at a critical value, associates engineering disposal plan recommendations from historical cases if a special geological structure is identified, verifies data logic contradictions in real time during report generation, and triggers a data review prompt when the predicted value of the soft soil layer thickness is greater than the total depth of the stratum.

[0014] In one embodiment of the present invention, the edge computing node connected to the on-site perception module performs raw data preprocessing and real-time risk assessment, the cloud server cluster runs the multi-source fusion module and the intelligent analysis module, the modeling warning module and the report generation module are deployed on the cloud, and the edge node and the cloud synchronize data through the incremental data compression transmission protocol. When the network is interrupted, the edge node temporarily stores the data and automatically resumes the transmission after recovery. The edge computing node embeds a lightweight risk identification model, which is derived from the cloud machine learning model through knowledge distillation technology. In a network-free environment, a primary warning signal is generated based on the drilling parameter mutation characteristics and the locally cached geological risk rule library. On-site personnel can view the warning details and temporary disposal guidelines through mobile terminals.

[0015] The present invention also provides a geotechnical engineering intelligent survey method based on big data, including: S1: Real-time acquisition of drilling parameters, in-situ test data and geological images to form original survey signals; S2: Receives the original survey signal and accesses the historical engineering survey database and regional geological database, generating a fused data signal through data cleaning and semantic association; S3: Receives fused data signals, constructs a knowledge graph containing geotechnical entity relationships, predicts geotechnical parameters and stratum distribution in unexplored areas based on machine learning models, and generates parameter prediction signals; S4: Receive parameter prediction signals, automatically construct a 3D geological model, and simultaneously identify geological risk patterns to generate risk warning signals; S5: Receive parameter prediction signals and risk warning signals, and automatically output a structured survey report containing geological descriptions, parameter charts, and risk warnings.

[0016] The big data-based intelligent geotechnical engineering survey system and method provided by the present invention uses an on-site perception module to collect drilling parameters, in-situ test data and geological images in real time to form original signals; a multi-source fusion module integrates historical databases and regional geological data to generate fused signals; an intelligent analysis module constructs a geotechnical knowledge map and uses machine learning to predict parameters of unexplored areas; a modeling and early warning module automatically constructs a three-dimensional geological model and identifies risks; and a report generation module outputs structured reports, realizing data-driven intelligent decision-making and replacing reliance on manual experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 This is the system architecture diagram of the geotechnical engineering intelligent investigation system based on big data; Figure 2 A schematic diagram showing the workflow of the geotechnical engineering intelligent investigation system based on big data; Figure 3 The figure shows a flow chart of the intelligent geotechnical engineering investigation method based on big data. DETAILED DESCRIPTION

[0019] The following describes the embodiments of the present invention through specific examples. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. The details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other unless they conflict.

[0020] It should be noted that the illustrations provided in the following embodiments are merely schematic illustrations of the basic concept of the present invention. Therefore, the illustrations only show components related to the present invention and are not drawn according to the number, shape, and size of components in actual implementation. In actual implementation, the type, quantity, and proportion of each component may be changed arbitrarily, and the component layout may also be more complex.

[0021] In the following description, numerous details are discussed to provide a more thorough explanation of the embodiments of the present invention. However, it will be apparent to those skilled in the art that the embodiments of the present invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring the embodiments of the present invention.

[0022] See Figure 1-3 , shown is the intelligent geotechnical engineering survey system and method based on big data of the present invention. The intelligent geotechnical engineering survey system based on big data of the present invention includes a field perception module, a multi-source fusion module, an intelligent analysis module, a modeling and early warning module and a report generation module. The field perception module collects drilling parameters, in-situ test data and geological images in real time to form original survey signals; the multi-source fusion module receives the original survey signals and accesses the historical engineering survey database and the regional geological database, and generates fused data signals through data cleaning and semantic association; the intelligent analysis module receives the fused data signals, constructs a knowledge graph containing geotechnical entity relationships, and predicts geotechnical parameters and stratum distribution in unexplored areas based on machine learning models to generate parameter prediction signals; the modeling and early warning module receives the parameter prediction signals, automatically constructs a three-dimensional geological model and simultaneously identifies geological risk patterns to generate risk warning signals; the report generation module receives the parameter prediction signals and the risk warning signals, and outputs a structured survey report containing geological descriptions, parameter charts and risk prompts.

[0023] like Figure 1As shown in the figure, the core system of this patent consists of five modules forming a closed-loop signal processing loop. The field perception module, as the data source, uses the pressure sensors, torque sensors, and displacement encoders deployed on the drilling equipment to capture real-time timing parameters such as drilling pressure, rotational torque, and footage speed. Combined with high-definition lithologic images captured by the in-hole panoramic camera, as well as the cone resistance and side friction resistance data of the static penetration probe or the hammer count record of the standard penetration test, it forms an original exploration signal stream containing timestamps, spatial coordinates, and equipment identification. The signal stream is transmitted to the multi-source fusion module through the industrial Internet of Things gateway. The module first accesses structured data such as drill hole histograms, geotechnical test reports, groundwater records in the historical engineering survey database, and associates spatial data such as stratigraphic contour maps, fault distribution maps, and remote sensing images in the regional geological database. The data cleaning engine removes outliers and fills in missing items. The semantic parser based on geotechnical engineering ontology is then used to convert descriptive terms such as "sandy clay" and "strongly weathered rock layer" in the historical report text into standardized entity codes. Finally, the real-time data, historical data and regional data are aligned according to spatial location and time dimensions to generate a fused data signal. After this signal is input into the intelligent analysis module, the knowledge graph construction process is initiated: starting with the formation number as the primary node, sub-nodes such as lithology classification, physical parameters, and structural characteristics are extended to establish attribute association edges such as "silt layer-permeability coefficient-seismic liquefaction risk." Simultaneously, a machine learning prediction engine is loaded. Using spatial interpolation algorithms and probabilistic graphical models, based on the distribution patterns of parameters such as cohesion and internal friction angle at known exploration points, the module derives parameter confidence intervals and a probabilistic model of formation interfaces in the undrilled area, outputting a parameter prediction signal with an error assessment indicator. After receiving this signal, the modeling and warning module uses an implicit surface generation algorithm to convert the discrete prediction point set into a continuous three-dimensional geological model. The module then compares the drilling parameter mutation characteristics with the disaster rules in the knowledge graph in real time (for example, the correlation between a sudden pressure drop exceeding a threshold and the risk of cavitation collapse) to generate a graded warning signal. The final report generation module integrates the stratigraphic contour map, lithologic distribution thermodynamic map and risk location markers in the parameter prediction signal, and automatically outputs a structured report containing engineering geological evaluation, foundation selection recommendations and risk prevention and control measures through a natural language generation template, thus achieving full process automation from raw data collection to intelligent decision-making output.

[0024] Furthermore, the hardware composition and data processing flow of the field perception module have been further refined. The drill bit pressure sensor uses a resistance strain gauge integrated into the drill pipe hydraulic system, with a sampling frequency of no less than 100 Hz, to record dynamic resistance changes during the drill bit's rock-breaking process in real time. The in-hole camera is equipped with a 360-degree rotating optical lens and a fill-light system. It uses an image stitching algorithm to generate a continuous expansion of the borehole wall, automatically identifying rock formation fracture rate and inclination information. The in-situ tester includes a static penetration double-bridge probe and an automatic drop hammer device for standard penetration tests. During drilling breaks, it tests and records cone tip resistance, side friction resistance, or penetration blows. These devices transmit data via an industrial-grade wireless transmission protocol to an edge computing node. This node uses a time synchronization protocol to align multi-source data timestamps, employs Kalman filtering to reduce noise for sensor drift, performs super-resolution reconstruction of the camera's shaken and blurred images, and performs data normalization: converting drill bit pressure values ​​to megapascals, correcting penetration blows by depth, and quantizing image texture features into grayscale statistical histograms. A key innovation lies in the edge node's built-in abnormal data interception mechanism. When it detects a persistent zero WOB reading (possibly indicating a drill pipe drop) or a completely black in-hole image (possibly due to a lens obstruction), it automatically triggers a device self-check and discards invalid data segments, ensuring that the output raw survey signal meets integrity verification rules. Furthermore, the edge node implements data compression optimization, using lossy floating-point encoding for drilling parameters and block-based discrete cosine transform compression for high-definition imagery. This reduces transmission bandwidth requirements by over 70%, providing stable support for field operations.

[0025] Specifically, the core technical means of the multi-source fusion module are deeply expanded in this claim. The data cleaning stage implements rule verification based on the geotechnical engineering ontology: this ontology defines a stratigraphic lithology classification tree (e.g., sedimentary rock → clastic rock → sandstone → coarse sandstone), physical and mechanical parameter threshold constraints (e.g., a porosity greater than 1.0 is considered highly compressible soil), and geological structural spatial rules (e.g., sudden changes in the rock formations on both sides of a fault zone). When the cleaning engine detects that the cohesion of a borehole in the historical database reaches 500 kPa (far exceeding the reasonable range for geotechnical engineering), it automatically marks it as an outlier and traces back to the original test record for correction; missing bedrock depth contours in the regional geological database are reconstructed using the Kriging interpolation method. The core of semantic association is a natural language processing engine, which includes a geological term extraction model and a contextual relationship parser. First, descriptions such as "gray-yellow silty clay intercalated with thin layers of fine sand" from historical survey reports are segmented to identify entity terms such as "silty clay" and "fine sand." Dependency parsing then identifies the interlayer relationships expressed by "intercalated thin layers." Finally, these entities and relationships are mapped to an ontology model, generating a standardized representation such as "3-5 meters from borehole B: silty clay (main layer) - containing - fine sand (thin layer)." Spatial registration employs an affine transformation algorithm to align the local coordinate system of the real-time drilling trajectory with the geodetic coordinate system of the regional geological map, maintaining a registration error within 0.1 meters. The generation of the fused data signal embodies multi-dimensional integration: temporally integrating nearly ten years of historical project data to form a temporal evolution baseline; spatially linking kilometer-scale regional structures with meter-scale borehole details; and semantically connecting natural language descriptions with machine-readable encodings. In particular, this module establishes a data lineage tracking mechanism to mark the source of each fused data unit (such as a certain porosity ratio value comes from Table 2 of the geotechnical test of the XX project in 2020) to ensure the traceability of the results.

[0026] In one embodiment of the present invention, the predictive capabilities of the intelligent analysis module stem from the collaborative architecture of a knowledge graph and a machine learning model. The knowledge graph is constructed using a graph neural network framework, with the stratigraphic unit of the project site as the root node (e.g., Quaternary cover, Cretaceous sandstone), extending downward to lithologic subnodes (gravel, silt, mudstone, etc.). Three types of associated edges are established: lithologic transition edges describe stratigraphic transitions (e.g., the probability of silt to clay transitions), parameter statistics edges record the correlation between shear strength and permeability, and tectonic influence edges indicate the impact of faults on rock fragmentation. The machine learning prediction engine is designed as a hybrid model: the front-end uses a graph convolutional network to extract topological features from the knowledge graph and generate stratigraphic continuity embedding vectors. The back-end, coupled with a spatial interpolation algorithm, performs Delaunay triangulation on the drillhole point set in the fused data signal and uses radial basis function interpolation within the triangular units to calculate geotechnical parameters at unexplored locations. The prediction process incorporates Monte Carlo simulation, generating thousands of sampling results by perturbing the input data. The resulting output parameter prediction signal includes three core components: a spatial probability distribution map of geotechnical parameters (e.g., a 95% confidence interval for the internal friction angle of 28-32 degrees), a confidence surface for stratigraphic interfaces (with an error range of ±1.5 meters for interface depth), and an uncertainty quantification heat map (using a color gradient to indicate prediction reliability). Specifically, when a paleochannel is detected at a site, the knowledge graph automatically loads the "lenticular sand liquefaction" rule from historical case studies, adding an excess pore water pressure warning indicator to the parameter prediction signal.

[0027] like Figure 2 As shown, the modeling and early warning module enables simultaneous geological model construction and risk identification. Three-dimensional geological modeling utilizes an improved implicit surface reconstruction technique: first, the predicted stratigraphic interface points in the parameter prediction signal are converted into a weighted set of spatial sampling points, with the weights inversely proportional to the prediction uncertainty. A continuous surface function is then fitted using the moving least squares method, with directional derivative constraints introduced for the fault region. A triangular mesh model of the fault zone is generated based on the strike and dip data of the fault in the regional geological database. Geological risk identification utilizes a dual-channel analysis mechanism: a real-time channel monitors the drilling parameter stream and uses a dynamic time warping algorithm to compare the current WOB curve with the "cavity collapse" feature template in the knowledge graph. A red alert is triggered when the similarity exceeds 90% for five consecutive seconds and the torque fluctuation exceeds a threshold. An imaging channel uses a convolutional neural network to analyze the real-time video stream transmitted by the in-hole camera to identify areas of dense fractures (where the number of fractures per unit area exceeds a critical value) or abnormal water seepage (where pixel chromaticity values ​​suddenly change). This is then combined with in-situ test data (e.g., a sudden drop in friction ratio during static penetration testing) to generate a risk level signal. The early warning logic implements a layered response: the first-level warning (yellow) indicates potential risks and recommends intensified exploration; the second-level warning (orange) forces drilling to be suspended and initiates emergency scanning; the third-level warning (red) links on-site sound and light alarms and automatically pushes disaster avoidance plans to mobile terminals.

[0028] Furthermore, the 3D geological model features dynamic evolution and on-site interaction. The dynamic update mechanism is implemented using an octree spatial index. When new exploration borehole data is input, the system automatically locates the affected spatial block (with a minimum granularity of 1 cubic meter) and performs a local reconstruction of the implicit surface function within that block. This reconstruction process preserves the original model's boundary constraints to ensure geometric continuity. Model self-optimization is achieved through residual learning: the residuals between the actual measured values ​​and the original predicted values ​​at newly added points are calculated. When the regional average residual exceeds a tolerance threshold, the intelligent analysis module triggers a model retraining process, using an incremental learning algorithm to adjust the graph neural network weights. The on-site response system includes a mobile terminal application. Upon receiving a risk warning signal, the terminal uses a lightweight 3D engine to render the geological risk volume (e.g., a subsidence area is displayed as a red translucent cube), which is then overlaid onto the real-time camera feed using augmented reality technology. The avoidance route is generated using the A* pathfinding algorithm, starting from the current drilling rig position and avoiding a 3-meter safety buffer zone extending beyond the boundaries of all warning areas. The optimal path is dynamically displayed on the terminal screen. New features include a real-time data annotation tool: engineers can circle the abnormal area of ​​the model on a mobile terminal and manually enter correction parameters (such as "the bedrock surface here is raised by 2 meters"). The system automatically verifies the rationality of the correction value and triggers a local model update.

[0029] like Figure 3 The figure shows the intelligent geotechnical engineering survey method based on big data of the present invention. S1: Real-time acquisition of drilling parameters, in-situ test data and geological images to form original survey signals; S2: Receive the original survey signals and access the historical engineering survey database and the regional geological database, and generate a fused data signal through data cleaning and semantic association; S3: Receive the fused data signals, construct a knowledge graph containing geotechnical entity relationships, and predict the geotechnical parameters and stratum distribution of unexplored areas based on the machine learning model to generate a parameter prediction signal; S4: Receive the parameter prediction signal, automatically construct a three-dimensional geological model and simultaneously identify geological risk patterns to generate a risk warning signal; S5: Receive the parameter prediction signal and the risk warning signal, and automatically output a structured survey report containing geological descriptions, parameter charts and risk warnings.

[0030] Furthermore, the report generation module enables fully automated intelligent document production. The structured template engine utilizes a recursive neural network architecture: the geological description component inputs stratigraphic distribution data from parameter prediction signals into a sequence generation model, outputting paragraph text that complies with the "Code for Geotechnical Engineering Investigation" (e.g., "The silty clay layer in the southeast of the site is 8-12 meters thick, with localized intercalation of silt lenses"). The risk warning component uses a semantic matching engine to link warning signals with contingency plan templates in the knowledge base. When liquefaction risk is identified, it automatically inserts treatment recommendations such as "It is recommended to adopt gravel pile infill treatment." Parameter chart generation utilizes adaptive visualization technology: based on the user-defined horizontal range and vertical depth, contour line generation algorithms of varying precision are dynamically invoked (Kriging for 100-meter ranges and Inverse Distance Weighted Method for 10-meter ranges). Geotechnical parameter spatial distribution maps undergo vector optimization processing, using the Douglas-Peucker algorithm to thin out stratigraphic interface lines, retaining curvature feature points to reduce file size. The quality control system incorporates three layers of validation: a logic check layer detects parameter inconsistencies (e.g., a compression modulus value less than 30% of the historical data mean for the same layer triggers a warning); a code check layer verifies the completeness of inputs for liquefaction discrimination formulas in seismic codes; and a version management layer records report modification traces and supports backtracking to historical versions. The final output supports multiple formats, with the Word version retaining style tags for manual revisions, and the PDF version integrating the ability to scan and view 3D models using QR codes.

[0031] The big data-based intelligent geotechnical engineering survey system and method of the present invention uses an on-site perception module to collect drilling parameters, in-situ test data, and geological images in real time to form original signals; a multi-source fusion module integrates historical databases and regional geological data to generate fused signals; an intelligent analysis module constructs a geotechnical knowledge graph and uses machine learning to predict parameters of unexplored areas; a modeling and early warning module automatically constructs a three-dimensional geological model and identifies risks; and a report generation module outputs structured reports, enabling data-driven intelligent decision-making and replacing reliance on manual experience.

[0032] Therefore, the big data-based intelligent geotechnical engineering survey system and method of the present invention solves the problems of discrete survey data, strong dependence on experience, and low intelligence.

[0033] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the present invention. Anyone skilled in the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by one of ordinary skill in the art without departing from the spirit and technical principles disclosed herein are intended to be covered by the claims of the present invention.

Claims

1. The geotechnical engineering intelligent investigation system based on big data is characterized by: include: An on-site sensing module, which collects drilling parameters, in-situ test data, and geological images in real time to form original survey signals; A multi-source fusion module receives the original survey signal and accesses the historical engineering survey database and the regional geological database, and generates a fused data signal through data cleaning and semantic association; an intelligent analysis module, which receives the fused data signal, constructs a knowledge graph containing geotechnical entity relationships, predicts geotechnical parameters and stratum distribution in unexplored areas based on a machine learning model, and generates a parameter prediction signal; A modeling and warning module receives the parameter prediction signal, automatically constructs a three-dimensional geological model, and simultaneously identifies geological risk patterns to generate a risk warning signal; A report generation module receives the parameter prediction signal and the risk warning signal, and outputs a structured exploration report including a geological description, parameter charts and risk warnings.

2. The geotechnical engineering intelligent investigation system based on big data according to claim 1 is characterized in that: The on-site perception module includes a drilling pressure sensor, an in-hole camera device and an in-situ tester. The drilling pressure sensor obtains pressure change data in real time during the drilling process of the drilling rig. The in-hole camera device collects images of the stratum structure of the borehole wall through an optical lens. The in-situ tester performs static penetration tests or standard penetration tests during drilling intervals. The above-mentioned devices send the collected data to the edge computing node through the Internet of Things communication protocol. The node performs timestamp synchronization and noise filtering on the multi-source asynchronous data to form a standardized original exploration signal.

3. The geotechnical engineering intelligent investigation system based on big data according to claim 1 is characterized in that: The multi-source fusion module performs data cleaning using outlier detection rules based on a geotechnical engineering ontology. This ontology defines a stratum lithology classification system, physical and mechanical parameter threshold ranges, and geological structural spatial relationship constraints. It also uses a natural language processing engine to parse historical survey report texts, automatically extract key attribute terms from stratum description paragraphs, and associate them with ontology entities. The remote sensing image data in the regional geological database is geo-referenced with the spatial coordinates in the original survey signal, ultimately generating a multi-dimensional fused data signal that integrates spatiotemporal attributes, semantic attributes, and engineering parameters.

4. The geotechnical engineering intelligent investigation system based on big data according to claim 1 is characterized in that: When constructing the knowledge graph, the intelligent analysis module uses stratigraphic units as nodes and lithologic gradient relationships, parameter statistical dependencies, and geological structural influence relationships as edges. The machine learning model adopts a hybrid architecture that integrates graph neural networks and spatial interpolation algorithms. It enhances the stratigraphic continuity modeling capability through the topological structural features of the knowledge graph. When predicting unexplored areas, it generates geotechnical parameter probability distribution maps and stratigraphic interface confidence surfaces based on spatial location similarity and geological environment similarity, and outputs parameter prediction signals with quantitative uncertainty indicators.

5. The geotechnical engineering intelligent investigation system based on big data according to claim 1 is characterized in that: The modeling and warning module uses implicit surface reconstruction technology to construct a three-dimensional geological model, converting the discrete prediction point set in the parameter prediction signal into a continuous geological interface, and constructing a fault zone spatial model by introducing fault strike and dip data. Geological risk pattern recognition includes real-time analysis of drilling parameter mutation characteristics and abnormal textures in the borehole image, combined with the disaster trigger rule chain predefined in the knowledge graph. When a sudden drop in drilling pressure is detected accompanied by the appearance of a broken zone feature in the borehole wall image, a landslide risk warning signal is generated. If the in-situ test data mutates and meets the liquefaction discrimination criteria, a sand liquefaction warning signal is generated.

6. The geotechnical engineering intelligent investigation system based on big data according to claim 1 is characterized in that: The three-dimensional geological model supports a dynamic update mechanism. When new exploration point data is input, local model reconstruction is automatically triggered. The machine learning model weights are adaptively adjusted by comparing the deviation values ​​between the predicted parameters and the actual measured parameters. After the geological risk warning signal is generated, it is pushed to the on-site mobile terminal in real time. The terminal displays a three-dimensional perspective rendering of the risk location and avoidance route suggestions.

7. The geotechnical engineering intelligent investigation system based on big data according to claim 1 is characterized in that: The report generation module has a built-in structured template engine, which converts the stratigraphic distribution data in the parameter prediction signal into geological profile description text, automatically matches the standardized expressions in the standard terminology library, calls the preset risk response plan text fragment when the risk warning signal is triggered, and uses dynamic contour rendering technology to generate parameter charts. It outputs vectorized geotechnical parameter spatial distribution maps based on the display range and accuracy requirements set by the user, and finally integrates all elements to generate a structured survey report that conforms to the industry standard format.

8. The geotechnical engineering intelligent investigation system based on big data according to claim 1 is characterized in that: The structured template engine integrates an expert experience rule library, automatically inserts design considerations when the predicted parameters are at critical values, associates engineering disposal plan suggestions from historical cases if special geological structures are identified, verifies data logic contradictions in real time during report generation, and triggers a data review prompt when the predicted value of the soft soil layer thickness is greater than the total depth of the stratum.

9. The geotechnical engineering intelligent investigation system based on big data according to claim 1 is characterized in that: The edge computing node connected to the on-site perception module performs raw data preprocessing and real-time risk assessment. The cloud server cluster runs the multi-source fusion module and the intelligent analysis module. The modeling warning module and the report generation module are deployed on the cloud. The edge node and the cloud synchronize data through the incremental data compression transmission protocol. When the network is interrupted, the edge node temporarily stores the data and automatically resumes the transmission after recovery. The edge computing node is embedded with a lightweight risk identification model, which is derived from the cloud machine learning model through knowledge distillation technology. In a network-free environment, a primary warning signal is generated based on the drilling parameter mutation characteristics and the locally cached geological risk rule library. On-site personnel can view the warning details and temporary disposal guidelines through mobile terminals.

10. A geotechnical engineering intelligent investigation method based on big data according to any one of claims 1 to 9, characterized in that: include: S1: Real-time acquisition of drilling parameters, in-situ test data and geological images to form original survey signals; S2: Receive the original survey signal and access the historical engineering survey database and the regional geological database, and generate a fused data signal through data cleaning and semantic association; S3: Receive the fused data signal, construct a knowledge graph containing geotechnical entity relationships, predict geotechnical parameters and stratum distribution in unexplored areas based on a machine learning model, and generate a parameter prediction signal; S4: receiving the parameter prediction signal, automatically constructing a three-dimensional geological model and simultaneously identifying geological risk patterns, and generating a risk warning signal; S5: receiving the parameter prediction signal and the risk warning signal, and automatically outputting a structured survey report including a geological description, parameter charts, and risk warnings.

Images