An integrated geotechnical engineering integrated intelligent platform

Abstract

This invention discloses a comprehensive intelligent platform for geotechnical engineering. "Comprehensive" refers to meeting the needs of geotechnical engineering throughout its entire lifecycle, multi-disciplinary collaboration, and three-dimensional digitization and intelligent operation. The intelligent platform adopts a front-end / back-end separated B / S architecture. The back-end server includes a data layer and a logic layer, while the web front-end browser includes an application layer. The data layer includes various knowledge graphs, knowledge-driven intelligent agents, and engineering databases. The logic layer consists of multiple independent algorithm engines that intelligently interact with the data layer to complete complex data processing. The application layer is divided into multiple functional modules, combined according to professional attributes to form professional subsystems such as exploration, design, monitoring, and risk control. Each subsystem is configured with different versions according to industry operation standards, possessing the ability to adapt to different scenarios. This invention achieves a generational upgrade of geotechnical engineering data processing technology, software, and operation methods.

Description

Technical Field

[0001] This invention belongs to the field of artificial intelligence, and specifically relates to an integrated intelligent platform for geotechnical engineering, in order to achieve a generational upgrade of geotechnical engineering data processing technology, application methods and software. Background Technology

[0002] Since the 1980s, geotechnical engineering investigation and design have still relied on 2D desktop software and human-computer interaction. The effectiveness of these applications is influenced by individual skill levels, and data silos naturally arise during the process. Professional digital technologies and intelligent upgrades based on the internet can effectively solve these problems and have become a hot research area worldwide. Intelligent upgrades are reflected in the productization process, replacing desktop software with a networked system that separates the front-end and back-end. A knowledge base corresponding to various digital technologies is built on the system back-end, giving these technologies learning capabilities, and intelligent sensing replaces human-computer interaction during application. However, due to technical limitations, current networked upgrade systems developed based on design technologies have limited capabilities, limited to early-stage, low-detail planning and design, and do not yet support higher-stage, intelligent detailed design. This ensures that desktop design software remains dominant, meaning that the intelligent upgrade of digital technologies in geotechnical engineering is still in an exploratory stage.

[0003] Originating in the 1950s, artificial intelligence (AI) technology has evolved from knowledge-driven and mechanism-driven stages to a more advanced data-driven stage thanks to breakthroughs in algorithms and computing power. The former two remain indispensable foundational components of AI. The expert systems and SaaS boom in geotechnical engineering during the 1980s represented early manifestations of knowledge-driven and mechanism-driven approaches, but these did not continue to the present day. Therefore, developing AI in geotechnical engineering requires first solidifying the foundation and developing both knowledge-driven and mechanism-driven AI technologies. Both the knowledge involved and the complex mechanisms governing data processing are highly specialized and cannot be achieved by directly referencing existing results such as multimodal large models. They must be developed independently based on AI principles and the specific characteristics of geotechnical engineering. The platform data layer and its contained knowledge-driven intelligent agents described in this invention represent a breakthrough achievement under this approach.

[0004] The most demanding task in geotechnical engineering production practice is daily data processing based on specifications. Building a knowledge-driven geotechnical engineering intelligent agent based on rules is an innovative and crucial foundational task, aiming to complete daily tasks compliantly and intelligently, constituting a key component of this invention. Currently, digital technologies used in geotechnical engineering, such as numerical simulation, 3D modeling, and 2D drawing, are primarily applied in desktop environments, unsuitable for the application requirements of today's internet and AI era. They must be upgraded and made intelligent. The key task is to enable these mature digital technologies to acquire learning capabilities and intelligently perceive and acquire the required input data during application, replacing the human-computer interaction of desktop software; this is another important component of this invention.

[0005] Mastering all the key digital technologies required for geotechnical engineering production processes is a fundamental condition for achieving intelligent upgrading. These key technologies mainly refer to 3D geological modeling, 3D geotechnical design, and mechanical calculation technologies. After years of effort, my country's geotechnical engineering field has mastered all key digital technologies. Among them, CnGIM_ma (Jiahua Geosciences 3D Geological Modeling and Analysis Software), launched in 2016 and achieving domestic substitution, has realized independent control over 3D geological modeling technology. CnGIM_sd (Jiahua Geosciences Slope 3D Design Software), released in 2024, enables 3D contour design and interactive mechanical calculation of excavated slopes based on 3D geological models. This signifies the integration and fusion of all 3D digital technologies in the geotechnical engineering field, surpassing international counterparts such as Dassault and Bentley, and placing China in a leading position in the intelligent upgrading process.

[0006] Currently, the network-based upgrades of BIM technology in the manufacturing and construction industries almost entirely adopt a client / server (C / S) architecture. While this solves the data silo problem caused by desktop software applications, it doesn't eliminate the need for human-computer interaction in digital technology operations and fails to achieve intelligent perception, thus hindering intelligent upgrades. In contrast, CnGIM_dg (Jiahua Geosciences Digital Geotechnical Integration Platform), launched in 2025, uses a browser / server (B / S) architecture, completely replacing human-computer interaction with intelligent perception, representing the latest development trend. This platform includes three professional subsystems—exploration, design, and monitoring—and a comprehensive system, meeting the requirements for 3D digitization and intelligent operation in the early stages of geotechnical engineering exploration and design, and achieving the upgrade of industrial software to platformization and intelligence. Building upon this, CnGIM_hw (Jiahua Geosciences Geological Risk Intelligent Early Warning System), released in 2026, also adopts a B / S architecture, developed specifically for the 3D digitization and artificial intelligence operation requirements during geotechnical engineering construction and operation. These two B / S architecture systems are different products compiled from a single codebase based on a unified architecture, essentially constructing a comprehensive integrated intelligent platform for geotechnical engineering. This invention systematically reviews the platform's innovative technologies to protect its original innovative achievements.

[0007] This invention embodies systematic innovation in the platform construction process and is a natural result of continuous technological breakthroughs and product innovation based on prior accumulation. The prior accumulation obtained solely by the inventor and directly related to this accumulation includes: ZL202210168837.1 addresses the issues of multi-disciplinary data integration and cross-disciplinary interaction in the platform's backend data layer. ZL202210459633.3 addresses data compatibility issues arising from differences in industry and regional standards within the same professional field during the construction of the platform's backend data layer. This involves the construction technologies for resource repositories and knowledge bases. ZL202111297351.X solves key technical problems in the logical layer's three-dimensional slope design based on a three-dimensional geological model, and is directly related to the logical layer three-dimensional design engine described in this invention. ZL202211393074.7, an empirical method for geological disaster early warning taking rockburst as an example, is closely related to the mechanism model engine in the platform and constructs a self-learning parameter optimization learning mechanism; ZL202510137943.7 is a geotechnical engineering intelligent design method that uses three-dimensional digital models as sensing data, replacing human-computer interaction with data perception.

[0008] CN122114124A, A method for constructing geotechnical engineering knowledge graphs for structured data.

[0009] The innovations listed above address several technical challenges in constructing an integrated intelligent platform for geotechnical engineering, but not all. From the perspective of meeting the needs of full-lifecycle 3D digital and intelligent operations, the platform construction process also requires solving digital technology issues such as 3D layering and model-data interaction; further enriching the knowledge graph composition by constructing a knowledge graph of unstructured data, integrating it with the structured data knowledge graph into the platform's data layer, and solidifying the foundation for artificial intelligence technology applications; another important task and innovation is to upgrade mature digital technologies to mechanistic models, focusing on building a configurable resource library and front-end functional design that supports parameter optimization, thereby gaining learning capabilities.

[0010] In view of this, the present invention proposes a comprehensive intelligent platform for geotechnical engineering, which focuses on solving the problems existing in the integration of multi-disciplinary data and multiple technologies and the realization of intelligent applications. Summary of the Invention

[0011] In view of the technical problems existing in the prior art, the present invention discloses a comprehensive integrated intelligent platform for geotechnical engineering.

[0012] This invention adopts a B / S architecture and, through data and technology integration, constructs a multi-purpose (comprehensive) intelligent platform, achieving a leapfrog development and upgrade of data processing technology, software, and application models in the field of geotechnical engineering. The intelligent platform employs a browser / server B / S architecture with a front-end and back-end separation; the back-end server includes a data layer and a logic layer, while the web front-end browser includes an application layer.

[0013] The data layer is constructed according to the requirements of an embedded knowledge-driven intelligent agent, which is the fundamental manifestation of the platform's "intelligence" and distinguishes it from any B / S architecture-based digital geotechnical engineering system.

[0014] The data layer includes a resource repository, a knowledge base, a knowledge graph, and an engineering database, supporting diverse and heterogeneous data; it constructs a multi-complementary search engine and provides access services through a Web API interface; and it further constructs a knowledge-driven intelligent agent based on the structured data in the data layer.

[0015] The knowledge of the knowledge-driven intelligent agent refers to the rules, basic professional knowledge, and experience requirements followed in geotechnical engineering production operations. Based on this, a resource library is constructed, and a corresponding knowledge graph is constructed according to the application conditions. For the programmable knowledge in it, a computational element library is further constructed using computer coding technology. After obtaining the input data required for production operations, the data layer constructed in this way automatically selects computational elements, completes data processing, and saves the data.

[0016] The resource and knowledge bases in the data layer are first constructed according to the specialization of the data, thus including at least several specialized resource and knowledge bases such as surveying resource and knowledge bases, design resource and knowledge bases, and monitoring resource and knowledge bases. Secondly, for each specialized resource and knowledge base, structured data is constructed using forms or relational databases, while unstructured data is constructed using file management services. Finally, general resource bases, industry resource bases, and enterprise resource bases are constructed according to their scope of application, corresponding to general standards, industry standards, and enterprise standards in production practice, respectively. The reference relationships between them are reflected in the platform through a parent-child-descendant node structure. Preferably, the resource base is a list of terms, and the knowledge base is the values ​​of the terms in the list; Preferably, the terms such as borehole diameter and geophysical methods in the exploration resource database are common to all industries and constitute the general resource database; the engineering stage division and rock mass quality grading methods vary from industry to industry and constitute the industry resource database; the report templates and drawing templates are formulated by each enterprise in accordance with industry rules and constitute the resource database. Preferably, the industrial profiles in the design resource library are common to various industries and constitute the general resource library; the design parameters and calculation methods often vary with the industry and constitute the industry resource library; the design drawing templates constitute the enterprise resource library.

[0017] The method of building a knowledge base depends on the data type and purpose. Structured data is built using relational databases; unstructured data in document form, when used as a whole template, uses relational databases to record relationships; otherwise, it is built by slicing documents and creating indexes.

[0018] The multi-dimensional complementary retrieval engine includes precise query, fuzzy matching, and semantic reasoning: Exact query: refers to a query using deterministic relationships and SQL statements. It is suitable for situations where the retrieval process must follow hard constraints, i.e., the prompts are fixed. Fuzzy matching: Based on precise query results, it uses fuzzy matching between the meaning of the prompt words and the corresponding field values ​​(multiple values) in the knowledge base to address situations where the prompt words may vary during the query. Lexical reasoning: Using vector semantics or knowledge graphs, conceptual expansion and semantic association are performed on free text prompts to capture deep similarities, targeting scenarios based on free input statements for retrieval.

[0019] Preferably, when constructing the retrieval mechanism in the platform, precise query is used first, followed by fuzzy matching, and finally semantic reasoning.

[0020] Furthermore, the logic layer in the backend server consists of multiple independent algorithm engines, including a proxy AI model engine and a mechanism model engine; each algorithm engine communicates with the knowledge graph data layer through a registration interface and dynamically selects and executes the corresponding algorithm based on the recommendation results of the knowledge graph.

[0021] AI agent models are generated through data training or fine-tuning for specific application scenarios. When applied, they return results after retrieving relevant data through knowledge graphs based on prompt words and a constructed multi-complementary retrieval mechanism. Mechanism models refer to algorithm engines developed according to different principles, often involving complex computational processes. They are used to generate text reports, 2D diagrams, 3D layering, 3D modeling, engineering design, and mechanical calculations. The core is to enable these digital technologies that follow rigorous mechanisms to have learning capabilities and upgrade to mechanism models. When applied, the model and perception data are correctly selected by the deterministic query in multivariate complementary retrieval.

[0022] Preferably, the learning ability of the mechanism model is obtained using different methods for different applications: Generate text reports using configurable templates; Two-dimensional plotting is generated using configurable templates and parameter optimization. 3D modeling, engineering design, and mechanical calculations are performed using parameter optimization.

[0023] More preferably, the report template and the drawing template are unstructured data in document form and stored in the enterprise resource library; More preferably, the configurable items of the report template include, but are not limited to, table of contents, illustrations, tables and text, and the dynamically filled data in all configurable items comes from the data of the current project recorded in the platform data layer; More preferably, the drawing templates stored in the enterprise resource library are constructed according to industry drawing requirements and two-dimensional drawing types; the configurable items in the templates include fields that may or may not be drawn, depending on the requirements. More preferably, graphic specifications (scale) and other parameters are used as optimization parameters, and their values ​​are determined by a combination of recommended default values ​​and manual modification. More preferably, in 3D modeling, 3D design, and mechanical calculations, the selected input parameters are determined by using recommended default values ​​and by allowing manual modification.

[0024] Furthermore, the three-dimensional stratigraphic layering algorithm engine is applicable to sedimentary strata, using the stratigraphic depositional regularity as a constraint and employing topological methods to construct the spatial connection relationships between strata; The sedimentary patterns include: the stratum thickness is not less than zero, the thickness is zero at the pinch-out position, and the lenticular sedimentary rhythm; the lenticular sedimentary rhythm means that the lenticular body is embedded as an independent stratigraphic unit between the upper and lower strata. The topological method includes: constructing proximity and connectivity relationships between strata with the same name to form an updatable stratigraphic topology network; When the name of the soil or rock, the stratigraphic code, the pinch-out type, or the lens connection method is modified through a web front-end browser, the engine only updates the affected stratigraphic topology, automatically maintaining the non-negative constraint of the stratigraphic thickness and the consistency of the lens connection with the upper and lower strata, without the need for global recalculation.

[0025] Furthermore, the application layer of the web front-end browser is divided into multiple application function modules. Each function module generates instructions, selects and drives the back-end algorithm engine, obtains the required input data (i.e., intelligent perception) from the data layer through the interface, completes the calculation process, and returns the results to the front end to realize intelligent application. The functional module consists of one or more commands to perform operations that meet specific requirements; The functional modules are combined with professional attributes to form at least one professional subsystem, such as a subsystem for surveying, designing, and monitoring in the early stage of a service project, and a subsystem for geological risk prevention and control in the construction and operation stage. Each specialized subsystem is configured with different versions based on industry operating standards and through user authorization to meet the application requirements of the corresponding industry.

[0026] Preferably, the exploration subsystem is further divided into versions such as geotechnical, power, and rock mass versions, which are implemented through authorization without changing the platform structure and integrity.

[0027] Preferably, the design subsystems are divided according to the type of project and the characteristics of the industry. The former includes slope design, foundation pit design and pile foundation design, while the latter includes tailings dam design and shaft design. Preferably, the application modules of the design subsystem can be invoked as a whole to meet the requirements of "composite" engineering designs that include multiple types.

[0028] Furthermore, the intelligent platform has constructed a geotechnical engineering model data format based on ASCII encoding as a unified data exchange standard, and configured a bidirectional conversion interface to achieve interoperability with various mainstream data formats.

[0029] The mainstream data formats include, but are not limited to, general formats such as dxf, obj, stl and gltf, as well as some proprietary formats such as mx and dgn.

[0030] Beneficial effects This invention has comprehensive beneficial effects, including the following aspects: Firstly, the technology and software for the entire process of geotechnical engineering investigation and design have undergone a generational upgrade and leapfrog development. From the current desktop software based on two-dimensional technology, it has been upgraded in one step to a three-dimensional and intelligent multi-professional collaborative platform, filling the gap in the nearly 50-year development history of industrial software. This has enabled my country's data processing technology and software capabilities in the field of geotechnical engineering to enter the world's forefront and take a leading position. Secondly, there is a revolutionary change in the production efficiency and quality assurance system. The knowledge-driven and mechanism-driven artificial intelligence technologies not only complete the relevant tasks efficiently and intelligently, but also strictly follow the requirements of the relevant work standards in the process of operation. They play the role of experience equality and technology equality respectively, greatly reducing the dependence on individual ability and revolutionarily improving efficiency and quality assurance capabilities. Thirdly, it has a revolutionary impact on operating models, business formats, and operations. When the platform user is an enterprise, the network-wide collaborative capability enables the enterprise to optimize human resource allocation and help to truly make the best use of everyone's talents. When the platform user is a regional (city) or industry management organization, it can integrate data to the greatest extent possible by region or industry, realize resource optimization within the industry, and create new positions and professions. The high-quality three-dimensional data accumulated with the continuous application of the platform naturally realizes asset accumulation and lays a solid foundation for carrying out data services. Attached Figure Description

[0031] Figure 1 The platform consists of a three-tier architecture and a backend data layer. Figure 2 : A multi-professional, multi-version application system configured using an authorization model; Figure 3 The integration of the platform's data layer composition and the construction of knowledge-driven intelligent agents; Figure 4 : Construction of integrated platform resource library and knowledge base; Figure 5 : Bar chart templates and configurable items in the enterprise resource library of the integration platform; Figure 6 The integrated platform provides a parameter tuning input window for 2D plotting. Figure 7 The application effect of the integrated platform's logic layer 3D layered engine; Figure 8 The integrated platform's application effect of intelligent pile length design based on a 3D geological model; Figure 9 The integrated platform provides intelligent assessment results of block instability risk based on mechanistic models. Figure 10 The application effect of the support and optimization functions of the integrated platform. Detailed Implementation

[0032] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0033] To address the shortcomings of existing technologies, this invention discloses a comprehensive intelligent platform for geotechnical engineering. It adopts a B / S architecture and integrates data, technology, and functions to achieve three-dimensional digitalization and intelligent operation throughout the entire lifecycle of geotechnical engineering.

[0034] one, Figure 1 This indicates the platform's composition, comprehensiveness, and intelligence.

[0035] Figure 1 The platform shown consists of two separate parts: a front-end and a back-end. The front-end (browser) integrates multiple application modules, organizes the required processing into instructions, and transmits them to the back-end via network communication. Based on the association, selection, and driving logic layer established by knowledge graphs and other methods, the algorithm engine further obtains the required input data from the data layer through the back-end WebAPI interface in an intelligent perception manner, completes the calculation, and returns the results.

[0036] Its comprehensiveness lies in its ability to adapt to the application requirements of multiple professions and different industries, which is also a manifestation of intelligence, thus distinguishing it from current digital PC software. Figure 1The platform lists four subsystems: integration, surveying, design, and monitoring. This means that the platform simultaneously meets the application requirements of three different professions and cross-professional collaborative management (integration). Correspondingly, the logic layer integrates multiple algorithm engines to meet the application requirements of data processing for all different professions. The engineering database, resource library, and knowledge base of the data layer are all built according to professions, storing relevant data from different professions and meeting the application requirements of different professions through data integration.

[0037] Figure 1 The algorithm engine in the logic layer shown in the lower left corner is divided into two categories: digital technology and AI technology. Digital technology includes text processing technology required for report generation, 2D CAD technology, 3D GIM modeling technology, and CAE technology for mechanical calculations. These technologies are developed according to specific principles (mathematics, mechanics, etc.) and are currently mostly integrated into PC software, used through human-computer interaction, lacking intelligent features. After integrating these technologies into the platform's backend server, the first step is to implement a SaaS service application, and then upgrade to AI technology that follows rigorous mechanisms, i.e., a mechanism-driven model. The intelligent upgrade will possess learning capabilities, thus distinguishing it from mere digitization; secondly, the method of obtaining input data during the application process is intelligent perception. The learning capability is achieved through no-code configuration and parameter tuning. The former is applied to the intelligent generation of reports and 2D diagrams, while the latter is applied to 2D diagrams and several other scenarios. The learning capability manifests as adaptability to more application scenarios and improved application effectiveness. Clearly, configuring templates, recommending default parameters, and supporting manual modification can all adapt to new application scenarios or achieve better application results without modifying the code, thus possessing the fundamental characteristics of "artificial intelligence technology."

[0038] Note that mechanism-driven approaches can also be mechanism models built based on the professional mechanisms (disaster mechanisms) followed by specific problems in geotechnical engineering (geological disasters).

[0039] Figure 1 The bottom right corner shows the composition of the platform's backend data layer and the method for constructing knowledge-driven intelligent agents, which constitutes the fundamental difference from all B / S architecture digital systems.

[0040] Figure 1The backend resource repository and knowledge base shown in the lower right corner are constructed according to specialization, with a more basic hierarchical relationship than engineering databases. The knowledge base is further structured according to data type, using different methods to create a structured knowledge base in the form of a relational database, and an unstructured knowledge base in the form of document slicing and indexing. The structured knowledge base records the value rules, methods, or results of each field in the resource repository, establishing deterministic relationships between them. When the rules and methods of the knowledge base are further converted into independently executable computational elements through computer encoding, and simultaneously integrated with the relationships with the resource repository, a knowledge-driven intelligent agent is constructed. This means that the intelligent agent is an integral part of the platform's data layer, thus distinguishing it from any other digital system.

[0041] two, Figure 2 This indicates the modular structure of the platform, with user entry layouts configured into different professional subsystems and industry versions according to authorization. The "Exploration Community" refers to the exploration subsystem, which is configured into industry versions (user entry points) such as "Geotechnical Exploration," "Municipal Exploration," "Power Exploration," and "Water Conservancy Exploration" according to the differences in exploration operations in different industries and through the authorization of application function modules. This is a manifestation of the comprehensive integrated platform meeting the customized application requirements of specific scenarios.

[0042] Figure 2 The different entry points shown do not affect the platform's architecture or code; they are the result of combining application modules in different ways. Each combination corresponds to a type of user group, and the combination is the result of authorizing application modules according to user groups. This reflects the platform's intelligent feature: adapting to the needs of different application scenarios without modifying the code.

[0043] three, Figure 3 The diagram illustrates the construction methods for resource repositories and knowledge bases for structured data. The resource repository records professional terms from different standards (abbreviated as shown in the list on the left). The values, or methods and criteria for assigning these terms, vary depending on the standard, thus constructing an industry-level resource repository. The knowledge base records the values, or methods and criteria for assigning these terms under different standards. Both the resource repository and knowledge base are constructed using a relational database, establishing deterministic relationships between them to create a knowledge graph suitable for structured data.

[0044] The method and criteria for retrieving values ​​from the structured knowledge base are further encoded into independently operable units, and after inheriting the association relationship with the resource base, the construction of a knowledge-driven intelligent agent is completed.

[0045] Four, Figure 4 This demonstrates the application effect of knowledge-driven intelligent agents.

[0046] After sampling and indoor testing are carried out in accordance with the specifications during geological exploration field work, and acoustic testing and geological logging are completed, and these raw data are entered into the exploration database within the platform, the relevant application functions for automatic data processing are clicked on the front end of the platform. The platform will then activate the exploration intelligent agent in the data layer, select the corresponding computing element according to the operation requirements, automatically read the data, complete the data processing, and store and update the results. This completes the entire intelligent operation process from perception (obtaining data through API) to reasoning (computation) to decision-making (saving results).

[0047] Figure 4 The example shown is weathering grading. Based on the specifications implemented in the current project, the rock mass weathering grading calculation element in the intelligent body is selected, the input data is read from the engineering database, the grading is completed and the results are saved, and then returned to the browser for viewing and inspection.

[0048] five, Figure 5 The image shows a drill bar chart template with configurable options. Users can meet application requirements in different scenarios by modifying only the configuration options in the template without modifying the code. This is an example of how no-code configuration can gain learning capabilities and expand application scenarios. It demonstrates how digital technology, after optimization of application methods, can be upgraded to a "mechanistic model," that is, the implementation of no-code configuration methods in specific scenarios. Obviously, the mechanism here refers to the geometric principles on which the drawing relies.

[0049] Figure 5 The configurable items are divided into two main parts: the map body and the map blocks. The former includes the position of the title, the total width of the map sheet and the total number of columns. The map blocks refer to the configuration of specific items, such as the name and width of each column, the format and position of the annotations, the font and size of the annotations, etc. In this way, templates that meet different drafting standards can be configured to adapt to the application requirements of different scenarios.

[0050] VI. Furthermore, after the selected drawing template is integrated, the data required for drawing is read (perceived) from the corresponding project in the engineering database through the API interface, and the graphic drawing is completed in the backend, other drawing parameters need to be set in production. Figure 6 The right side shows the final result of generating a borehole columnar section, which requires more than 10 parameters to be filled in. Filling in each one individually would affect efficiency and produce errors. To address this, the platform has built two intelligent operation methods: one is to provide default values ​​for these parameters based on known data such as borehole depth and configuration items, i.e., intelligent recommendation; the other is to record commonly used parameter combinations into a common configuration table for users to select and quickly complete individual modifications after setting, in order to obtain the best results. Essentially, this is a method for optimizing the parameters of the mechanistic model.

[0051] seven, Figure 7The image shows the application effect of the platform's logical layer 3D layering engine. Based on the layering results of each borehole, a 3D connection relationship is constructed between them using the built-in topology algorithm. The core lies in the automatic updating of the 3D connection relationship after arbitrary data editing based on topology updates. Arbitrary editing includes: 1) Modify the name or code of the soil and rock in any stratum of any borehole; 2) Interchange between the lens body and the pinch-out; 3) Adjustment of the extinction position.

[0052] After any of the above modifications are completed on the front end, the platform transmits the modified data to the back end, drives the 3D layering engine, completes the topology update of the affected part, returns the updated result, realizes 3D layering, and generates a standard formation applicable to all boreholes for layering of newly added boreholes.

[0053] Clearly, for this algorithm engine, the aforementioned modifications constitute parameter tuning, thereby obtaining the required layering results and improving the quality of the output. All the drilling data required for the application process is automatically obtained through the API interface, demonstrating an intelligent sensing effect.

[0054] eight, Figure 8 The image shows the application effect of the pile foundation design function module of the platform design subsystem, which completes all pile foundation bearing capacity calculations and length designs in one go. Its core is the use of corresponding computational elements in a knowledge-driven design agent, where the geological model is based on perceived data.

[0055] The knowledge-driven design agent follows the aforementioned construction method, namely, using a relational database to build a pile foundation design resource base and knowledge base, thereby uniquely clarifying the requirements of various standards for the calculation of bearing capacity of various pile types. Then, various calculation methods are converted into computational elements through computer encoding, constructing a computational element library with related relationships. Figure 8 The application shown can be linked to the computational element library and selected according to the specifications implemented in the current project and the pile type used for each pile, including the corresponding initial pile length design criteria and bearing capacity calculation method.

[0056] At this point, the 3D geological model stored in the platform's data layer and its associated geotechnical parameter values ​​become the complete input data required to complete all pile foundation bearing capacity calculations in one go. The calculation process first determines the initial length of each pile according to the selected initial pile length design criteria, combined with the geological model; then, it automatically obtains the geotechnical layers and their thicknesses traversed by each pile, the relative quantitative relationship between the pile tip and the bearing layer or soft soil layer, and associates this with the geotechnical parameter record table in the data layer to obtain all the input data required for bearing capacity calculation, thus achieving intelligent perception of all the input data required for bearing capacity calculation.

[0057] Nine, Figure 9This case study demonstrates the application of 3D discontinuous numerical simulation technology as a mechanistic model in the intelligent prediction of stochastic block instability risks during the construction of underground powerhouses in pumped storage power stations. It not only highlights the significant quality and efficiency improvements achieved by intelligently applying numerical simulation technology through a platform but also reveals the company's business innovation capabilities. At this point, mature third-party numerical simulation software is deployed on the platform's backend. Corresponding command stream files are constructed using ASCII encoding to drive the numerical simulation software and complete calculations via an instruction set approach. This process utilizes the model data interaction standard built by the platform.

[0058] Generally, completing a 3D numerical simulation of an underground powerhouse using PC software and a human-computer interaction approach takes several weeks, and the calculation results are mainly affected by the knowledge and experience of the personnel performing the calculations. When the numerical simulation software is used as the computing engine of the platform's backend, the platform directly reads the latest geological logging data from the backend API, automatically completes model updates, model settings, parameter assignments, excavation simulation, stability calculations, and result returns. Figure 9 The example shown took approximately 1.5 hours, representing a difference in work efficiency of several hundred times. Unlike human-computer interaction, the knowledge graph built by the platform integrates the experience of numerical simulation experts, intelligently recommending model settings based on specific conditions. This eliminates the uncertainty caused by the individual capabilities of the user on the calculation results, ensuring higher quality.

[0059] Numerical simulations based on PC software during the construction phase, due to their long operation cycle, can only predict potential long-term risks and optimize support systems. In contrast, platform-based intelligent computing compresses the construction period to 1.5 hours, providing a completely new method for predicting and warning of construction safety. This application case primarily serves the management of construction safety risks in engineering projects, and secondarily serves dynamic design optimization. It innovates the application scenarios of numerical simulation technology and expands into new business areas.

[0060] ten, Figure 10 Based on the S900, the support optimization is completed on the platform according to the results of intelligent geological risk analysis, thereby achieving risk prevention and control. Figure 9 and Figure 10 As shown, this demonstrates the upgrade of the collaborative work mode among the three disciplines of geology, computation, and design during the application of the integrated platform: collaboration between them is completed based on backend data interaction, replacing person-to-person handover or file copying, which not only systematically improves quality and efficiency, but also naturally completes the accumulation of high-quality data and guides business innovation.

Claims

1. A comprehensive intelligent platform for geotechnical engineering, characterized in that, The platform adopts a browser / server (B / S) architecture with a front-end and back-end separation; the back-end server includes a data layer and a logic layer, while the web front-end browser includes an application layer; wherein... The data layer includes a resource repository, a knowledge base, a knowledge graph, and an engineering database, supporting diverse and heterogeneous data; it constructs a multi-complementary search engine and provides access services through a WebAPI interface; and it further constructs a knowledge-driven intelligent agent based on the structured data in the data layer. The logic layer consists of multiple independent algorithm engines, including agent AI models and mechanism models. Each algorithm engine communicates with the data layer through a registration interface and dynamically selects and executes the corresponding algorithm based on the reasoning results of the knowledge graph to perform complex data processing tasks. The application layer is divided into multiple application function modules. Each function module drives the backend algorithm engine by generating instructions, obtains the required input data from the data layer through the interface, completes the calculation process, and returns the results to the front end to realize intelligent applications. The intelligent platform constructs a geotechnical engineering model data format based on ASCII encoding as a unified data exchange standard, and configures a bidirectional conversion interface to enable interactive operations with various mainstream data formats.

2. The intelligent platform according to claim 1, characterized in that, The data layer includes a resource library and a knowledge base. The resource library is a list of terms, and the knowledge base records the term values ​​of the terms contained in the resource library. The construction methods of the resource library and the knowledge base are as follows: First, professional resource bases and knowledge bases are constructed according to the professional attributes of the data. The professional resource bases and knowledge bases include at least survey resource bases and knowledge bases, design resource bases and knowledge bases, and monitoring resource bases and knowledge bases. Secondly, for the aforementioned professional resource database and knowledge base, the structured data is constructed using forms or relational databases according to the data type; for the unstructured data in document form, the knowledge base is constructed using a block-vectorization-storage indexing method. Finally, general resource libraries, industry resource libraries, and enterprise resource libraries are constructed according to the applicable scope of each professional resource, and implemented in the platform according to the hierarchical relationship of parent-child-grandchild nodes.

3. The intelligent platform according to claim 1, characterized in that, The multi-complementary retrieval engine includes precise query, fuzzy matching, and semantic reasoning; For cases with hard constraints, precise queries are achieved by using a relational database to establish a relationship between deterministic prompts and knowledge base field values. Fuzzy matching is built on the basis of precise query. In the case of diverse prompt words, it constructs a retrieval method that fuzzily matches the meanings of prompt words with terms in the knowledge base. Lexical reasoning uses vector semantics or knowledge graphs to expand the concepts and semantic associations of free text prompts, capturing deep similarities.

4. The intelligent platform according to claim 1, characterized in that, The knowledge-driven intelligent agent targets structured data and other programmable resources, and consists of a resource library, a knowledge graph, and a computational meta-library. The computational library consists of multiple computational elements that can be independently invoked and execute simple data processing tasks. Each computational element is programmed, compiled, and encapsulated to form a data processing method. The data processing method includes mathematical expressions, logical judgments, curves in the form of X~Y coordinate axes, and value criteria in the form of forms or text, which are managed through a resource library. The knowledge-driven intelligent agent dynamically selects and calls corresponding computational elements from the computational element library based on the characteristics of the input data and through the predefined association relationships in the knowledge graph, automatically completing data processing. Each matching and calling process can be traced back to the specific association rules and computational element identifiers.

5. The intelligent platform according to claim 1, characterized in that, The AI ​​agent model is designed for specific application scenarios and is obtained through data training or fine-tuning. When applied, it returns results after retrieving relevant data from the knowledge base based on prompt words and a constructed multi-complementary search engine. The aforementioned mechanism models are developed according to different principles and are used to generate text reports, two-dimensional diagrams, three-dimensional layering, three-dimensional modeling, engineering design, and mechanical calculations. When applied, the model and perceived data are correctly selected by a deterministic query in a multi-complementary search engine. The principles mentioned include the principles of computer text processing, the mathematical principles of two-dimensional drawing and three-dimensional modeling, and the mechanical principles followed by computational analysis.

6. The intelligent platform according to claim 5, characterized in that, For different mechanism models and their corresponding uses, different methods are used to build self-learning capabilities, so as to meet the needs of multiple application scenarios. The methods for building self-learning capabilities include: Generate text reports: using configurable templates; Two-dimensional plotting: using configurable templates and parameter optimization; 3D modeling, engineering design, and mechanical calculations: using parameter optimization.

7. The intelligent platform according to claim 6, characterized in that, The parameter optimization is implemented in the corresponding functional modules of the Web front-end browser application layer, and users can manually optimize it. The functional module consists of one or more operation commands, designed for specific application requirements, and is used to generate instruction sets, correctly select and drive the algorithm engine of the backend logic layer. The functional modules are combined into at least one professional subsystem according to professional attributes. The professional subsystem includes: a survey, design, and monitoring subsystem for the early stage of the project and a geological risk prevention and control subsystem for the construction and operation and maintenance stage. Each professional subsystem is configured into different versions according to industry operation standards and through user authorization to meet the application requirements of the corresponding industry.

8. The intelligent platform according to claim 5, characterized in that, The three-dimensional stratification is applicable to sedimentary strata; the three-dimensional stratification algorithm engine uses the sedimentary laws of strata as constraints, employs topological methods to construct the spatial connection relationships between strata, and supports topological updates. The sedimentary patterns include: the stratum thickness is not less than zero, the thickness is zero at the pinch-out position, and the lenticular sedimentary rhythm; the lenticular sedimentary rhythm means that the lenticular body is embedded as an independent stratigraphic unit between the upper and lower strata. The topological method includes: constructing proximity and connectivity relationships between strata with the same name to form an updatable stratigraphic topology data structure; When the name of the soil or rock, the stratigraphic code, the pinch-out type, or the lens connection method is modified through a web front-end browser, the three-dimensional layering algorithm engine only updates the affected stratigraphic topology, automatically maintaining the non-negative constraint of the stratigraphic thickness and the consistency of the lens connection with the upper and lower strata, without the need for global recalculation.

9. The intelligent platform according to claim 1, characterized in that, The mainstream data formats include general formats and some proprietary formats; The common formats include dxf, obj, stl, and gltf; The proprietary formats mentioned include mx and dgn.

Images