Foundation pit support scheme intelligent recommendation method and system

Abstract

The application discloses a kind of foundation pit support scheme intelligent recommendation method and system, including obtaining the knowledge data in the field of foundation pit support, constructs foundation pit support knowledge graph;Based on the basic data of target foundation pit, knowledge matching is carried out from foundation pit support knowledge graph, and candidate foundation pit support scheme is obtained;The basic data of target foundation pit and the scheme parameter corresponding to candidate foundation pit support scheme are input into the pre-constructed scheme evaluation model, the performance of each candidate foundation pit support scheme is evaluated, and the corresponding scheme evaluation result is obtained;The compliance detection result of candidate foundation pit support scheme is obtained by compliance detection;According to scheme evaluation result and compliance detection result, scheme optimization objective function is constructed, and intelligent optimization algorithm is used to iteratively optimize candidate foundation pit support scheme, and the target foundation pit support scheme satisfying the preset optimization condition is obtained.The application can integrate multi-source heterogeneous knowledge, quantitatively analyze influencing factors, and intelligently recommend the optimal foundation pit support scheme.

Description

Technical Field

[0001] This invention relates to the field of geotechnical engineering technology, specifically to an intelligent recommendation method and system for foundation pit support schemes. Background Technology

[0002] Foundation pit support is an important technical link in the construction process of geotechnical engineering. The selection of its scheme needs to comprehensively consider various factors such as foundation pit depth, geological conditions, groundwater distribution, surrounding environment and construction constraints. The applicable support forms and parameter configurations vary greatly under different engineering conditions.

[0003] Currently, the formulation of foundation pit support schemes typically relies on the accumulated experience of engineering technicians, involving manual analysis based on relevant specifications, historical engineering cases, and site conditions. Due to differences in the understanding and experience of different technicians regarding engineering conditions, the existing scheme design process suffers from insufficient knowledge utilization, low scheme matching efficiency, and difficulty in effectively reusing historical engineering experience. In recent years, with the development of intelligent computing technology, some technical solutions have attempted to utilize numerical simulation, optimization algorithms, and knowledge graphs to assist engineering design; however, these existing technologies mainly target single engineering parameter analysis or general knowledge retrieval scenarios, and their classification and expression of foundation pit geological conditions are relatively coarse, making it difficult to establish refined matching relationships for different geological processes. Furthermore, the correlation between foundation pit depth and key control factors such as geological conditions is not fully utilized, affecting the accurate judgment of the support scheme's suitability. In addition, key factors affecting the implementation of support schemes, such as groundwater conditions and land use boundaries, are usually treated as independent constraints, lacking unified data association and reasoning.

[0004] Therefore, to solve the above problems, an intelligent recommendation method and system for foundation pit support schemes is needed, which can integrate multi-source heterogeneous knowledge, quantitatively analyze influencing factors, and intelligently recommend the optimal foundation pit support scheme. Summary of the Invention

[0005] In view of this, the purpose of this invention is to overcome the deficiencies in the prior art and provide an intelligent recommendation method and system for foundation pit support schemes, which can integrate multi-source heterogeneous knowledge, quantitatively analyze influencing factors, and intelligently recommend the optimal foundation pit support scheme.

[0006] The intelligent recommendation method for foundation pit support schemes of the present invention includes:

[0007] Acquire knowledge data in the field of foundation pit support, and construct a foundation pit support knowledge graph based on the knowledge data; acquire basic data of the target foundation pit;

[0008] Based on the basic data of the target foundation pit, knowledge matching is performed from the foundation pit support knowledge graph to obtain candidate foundation pit support schemes;

[0009] The basic data of the target foundation pit and the corresponding scheme parameters of the candidate foundation pit support schemes are input into the pre-built scheme evaluation model to evaluate the performance of each candidate foundation pit support scheme and obtain the corresponding scheme evaluation results.

[0010] Based on the preset technical specifications and rules, the candidate foundation pit support schemes are tested for compliance, and the compliance test results are obtained.

[0011] Based on the evaluation results and compliance test results, an objective function for optimizing the scheme is constructed, and an intelligent optimization algorithm is used to iteratively optimize the candidate foundation pit support schemes to obtain the target foundation pit support scheme that meets the preset optimization conditions.

[0012] Furthermore, the knowledge data includes technical specifications for foundation pit support, historical cases of foundation pit support, and geotechnical engineering investigation reports;

[0013] The foundation pit support knowledge graph includes support scheme nodes, influencing factor nodes, evaluation index nodes, and the relationships between each node.

[0014] The influencing factor nodes include core control element nodes, key influencing factor nodes, and boundary condition nodes. The core control element nodes include the depth of the foundation pit and geological conditions. The geological conditions are further subdivided into rock slope type, rock-soil mixed type, gravelly soil type, sandy soil type, cohesive soil type, and soft soil type. Among them, the rock slope type is further subdivided into sub-types controlled by intensity, sub-types controlled by rock strata, and sub-types controlled by fissures. The key influencing factor nodes include groundwater parameters. The boundary condition nodes include whether excavation beyond the land use boundary line is allowed and whether there are important buildings nearby.

[0015] Furthermore, a knowledge graph of foundation pit support is constructed, specifically including:

[0016] Based on the ontology framework of geotechnical engineering, natural language processing algorithms are used to perform entity recognition and relation extraction on the clauses of the technical specification for foundation pit support, resulting in specification triplet;

[0017] Extract geological conditions, support scheme types, testing and monitoring status, project cost, construction period, slope area of ​​the foundation pit, and foundation pit depth data from historical foundation pit support cases to construct case feature nodes;

[0018] The graph embedding algorithm is used to embed case feature nodes into the standard triples to construct an initial knowledge graph;

[0019] Multi-hop reasoning optimization is performed on the case feature nodes and canonical triples in the initial knowledge graph to form a structured and semantic second knowledge graph;

[0020] The second knowledge graph is divided into modules to obtain the foundation pit support knowledge graph.

[0021] Furthermore, candidate foundation pit support schemes were obtained, specifically including:

[0022] Based on the basic data of the target foundation pit, geological parameters, groundwater characteristics, depth parameters, red line constraint information, and surrounding building information are extracted from the foundation pit support knowledge graph;

[0023] First, eliminate foundation pit support schemes whose construction scope exceeds the maximum allowable distance beyond the red line or encroaches on the most unfavorable safety protection distance of surrounding buildings, thus forming a constrained candidate set;

[0024] For the constrained candidate set, a module query vector is constructed using each module of the knowledge graph as the query benchmark; the modules include a depth adaptation module, a geological adaptation module, a groundwater adaptation module, and a red line constraint adaptation module;

[0025] Based on the graph attention mechanism, the query vector is fused with the module to generate a comprehensive query vector, and the multi-dimensional semantic similarity score between the comprehensive query vector and the case feature nodes in the knowledge graph is calculated.

[0026] Based on multi-dimensional semantic similarity scores and preset thresholds, the top-ranked candidates are selected when each dimension's score exceeds its corresponding threshold and the total score is among the highest. The support schemes corresponding to the feature nodes of the case are used as candidate foundation pit support schemes.

[0027] Furthermore, the multi-dimensional semantic similarity score The formula for calculation is:

[0028] ;

[0029] Among them, k=1 corresponds to the depth adaptation dimension, k=2 corresponds to the geological adaptation dimension, k=3 corresponds to the groundwater adaptation dimension, and k=4 corresponds to the red line adaptation dimension. The semantic weight coefficient for the k-th dimension is... Let k be the engineering fit attenuation factor. To project the query vector into the feature space of the k-th dimension, For transpose operation, Let be the embedding vector of the feature node in the k-th dimension of the case. Let the feature vector be the engineering parameter of the k-th dimension case feature node. Let k be the ideal feature vector of the case feature node in the kth dimension. Let be the tolerance coefficient for the engineering parameter in the k-th dimension.

[0030] Furthermore, performance evaluations were conducted on each candidate foundation pit support scheme, yielding corresponding evaluation results, including:

[0031] Based on the target foundation pit engineering requirements and candidate foundation pit support schemes, evaluation indicators are determined, including stability adaptability, construction convenience, economic adaptability, and surrounding environmental protection.

[0032] A feasibility assessment dataset for candidate schemes is constructed based on evaluation indicators and basic data of the foundation pit. In this dataset, the indicator feature data corresponding to the candidate schemes are extracted, and feature attributes are selected based on knowledge graph theory triples. The dataset is then balanced.

[0033] The evaluation dataset is input into the random forest model to generate multidimensional evaluation parameters. Combined with the indicator weight allocation rules, the feasibility score is calculated.

[0034] Furthermore, compliance testing was conducted on the candidate foundation pit support schemes, specifically including:

[0035] Based on the geological conditions, depth, and surrounding environment of the target foundation pit, applicable national standards, industry standards, and local standards and regulations are matched from the knowledge graph to extract a set of applicable regulations.

[0036] Technical parameters are extracted from candidate foundation pit support schemes. These technical parameters are then compared item by item with the mandatory indicators specified in the set of applicable provisions. For candidate schemes with parameters exceeding the limits, the following actions are taken: if the technical parameters can be adjusted, they are corrected to the lower limit of the specification and the scheme parameters are updated; if the technical parameters do not meet the structural requirements, the candidate foundation pit support scheme is removed from the candidate set.

[0037] For candidate foundation pit support schemes that pass the compliance verification, a compliance score is calculated based on the weighted average of the margins by which each parameter meets the specifications.

[0038] Furthermore, the objective function for optimizing the proposed scheme adopts a genetic fitness function; and a genetic algorithm or similar optimization algorithm is used to iteratively optimize the candidate foundation pit support scheme.

[0039] Furthermore, it also includes: incremental updates to the knowledge graph:

[0040] After the recommended foundation pit support scheme is completed, the actual construction period, actual project cost, foundation pit support structure deformation monitoring data, and project quality evaluation data of the project are collected and compared with the original predicted data. When the deviation between the actual data and the predicted data exceeds the preset deviation threshold, multi-hop inference optimization is triggered by the updated case feature node data to incrementally update the foundation pit support knowledge graph.

[0041] An intelligent recommendation system for foundation pit support schemes includes a knowledge graph construction module, a basic data acquisition module, a candidate scheme matching module, a feasibility assessment module, and a scheme optimization and recommendation module.

[0042] The knowledge graph construction module is used to acquire knowledge data in the field of foundation pit support and construct a foundation pit support knowledge graph based on the knowledge data.

[0043] The basic data acquisition module is used to acquire basic data of the target foundation pit;

[0044] The candidate scheme matching module is used to perform knowledge matching from the foundation pit support knowledge graph based on the basic data of the target foundation pit to obtain candidate foundation pit support schemes.

[0045] The feasibility assessment module is used to input the basic data of the target foundation pit and the corresponding scheme parameters of the candidate foundation pit support schemes into a pre-built scheme evaluation model, to evaluate the performance of each candidate foundation pit support scheme, and to obtain the corresponding scheme evaluation results; based on the preset technical specification rules, to perform compliance testing on the candidate foundation pit support schemes, and to obtain compliance testing results.

[0046] The scheme optimization and recommendation module is used to construct a scheme optimization objective function based on the scheme evaluation results and compliance detection results, and to use an intelligent optimization algorithm to iteratively optimize the candidate foundation pit support schemes to obtain a target foundation pit support scheme that meets the preset optimization conditions.

[0047] The beneficial effects of this invention are as follows: The intelligent recommendation method and system for foundation pit support schemes disclosed in this invention constructs a foundation pit support knowledge graph based on multi-source engineering knowledge, combines it with the basic data of the target foundation pit to perform scheme matching, and generates candidate support schemes; it uses a scheme evaluation model to evaluate the performance of candidate schemes, and combines it with technical specification rules to perform compliance checks; and iteratively optimizes the schemes through an intelligent optimization algorithm to obtain recommended support schemes. This invention can achieve effective integration and intelligent utilization of foundation pit engineering knowledge, overcome the subjective defects of manual matching, and improve the accuracy and intelligence of support scheme selection. Attached Figure Description

[0048] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0049] Figure 1 This is a schematic diagram of the intelligent recommendation method for foundation pit support schemes of the present invention;

[0050] Figure 2 This is a structural block diagram of the intelligent recommendation system for foundation pit support schemes of the present invention. Detailed Implementation

[0051] The present invention will be further described below with reference to the accompanying drawings, as shown in the figures:

[0052] This embodiment discloses an intelligent recommendation method for foundation pit support schemes, including the following steps:

[0053] S1. Acquire knowledge data in the field of foundation pit support, and construct a foundation pit support knowledge graph based on the knowledge data; acquire basic data of the target foundation pit;

[0054] S2. Based on the basic data of the target foundation pit, knowledge matching is performed from the foundation pit support knowledge graph to obtain candidate foundation pit support schemes;

[0055] S3. Input the basic data of the target foundation pit and the corresponding scheme parameters of the candidate foundation pit support schemes into the pre-built scheme evaluation model, evaluate the performance of each candidate foundation pit support scheme, and obtain the corresponding scheme evaluation results;

[0056] S4. Based on the preset technical specifications and rules, conduct compliance testing on the candidate foundation pit support schemes and obtain the compliance testing results;

[0057] S5. Based on the scheme evaluation results and compliance test results, construct the scheme optimization objective function, and use intelligent optimization algorithm to iteratively optimize the candidate foundation pit support schemes to obtain the target foundation pit support scheme that meets the preset optimization conditions.

[0058] In this embodiment, the knowledge data includes the technical specifications for foundation pit support, historical foundation pit support cases, and geotechnical engineering investigation reports;

[0059] The foundation pit support knowledge graph includes support scheme nodes, influencing factor nodes, evaluation index nodes, and the relationships between each node.

[0060] The influencing factor nodes include core control element nodes, key influencing factor nodes, and boundary condition nodes. The core control element nodes include the depth of the foundation pit and geological conditions. The geological conditions are further subdivided into rock slope type, rock-soil mixed type (total depth ≥ 5m, and soil thickness ≥ 3m), gravelly soil type, sandy soil type, cohesive soil type, and soft soil type. Among them, the rock slope type is further subdivided into sub-types controlled by intensity, sub-types controlled by rock strata, and sub-types controlled by fissures. The key influencing factor nodes include groundwater parameters. Groundwater parameters include groundwater level depth, permeability coefficient, and confined water properties. The boundary condition nodes include whether excavation beyond the land use boundary line is permitted and whether there are important buildings nearby.

[0061] Constructing a knowledge graph for foundation pit support, specifically including:

[0062] Based on the ontology framework of geotechnical engineering, natural language processing algorithms are used to perform entity recognition and relation extraction on the clauses of the technical specification for foundation pit support, resulting in specification triplet;

[0063] The entity identification includes factors such as the depth range of the foundation pit, geological type, rock subtype, groundwater conditions, land use boundary conditions, support scheme, and geological risks. The extracted relationships include applicable, recommended scheme, key parameters, risks, and not applicable. For example, entity identification and relationship extraction are performed on the clauses of the "Technical Specification for Foundation Pit Support" JGJ 120 and other specifications, resulting in specification triplets such as single soil nailing wall, not applicable, and soft soil.

[0064] Extract geological conditions, support scheme types, testing and monitoring status, project cost, construction period, slope area of ​​the foundation pit, and foundation pit depth data from historical foundation pit support cases to construct case feature nodes;

[0065] The graph embedding algorithm is used to embed case feature nodes into the standard triples to construct an initial knowledge graph;

[0066] Multi-hop reasoning optimization is performed on case feature nodes and canonical triples in the initial knowledge graph to form a structured and semantic second knowledge graph; the multi-hop reasoning optimization includes graph entity alignment and relation completion;

[0067] For example, the Rete forward reasoning engine is used to optimize the initial knowledge graph through multi-hop reasoning, uncovering implicit relationships, such as the relationship between rock slope - controlled by fissures → anchor support - applicable; the relationship between solid underground continuous walls and the depth attribute range of the foundation pit [10m, 30m] is used to connect them to nodes with geological conditions such as soft soil, gravelly soil, and cohesive soil; the relationship between groundwater abundance and groundwater richness is used to connect them to nodes requiring larger construction sites. Based on this, a second knowledge graph is obtained.

[0068] The second knowledge graph is divided into modules to obtain a foundation pit support knowledge graph. These modules include a depth adaptation module, a geological adaptation module, a groundwater adaptation module, and a red line constraint adaptation module.

[0069] In this embodiment, the basic data includes the depth of the foundation pit, the type of geological conditions, groundwater parameters, and boundary conditions information regarding whether excavation beyond the land use boundary is permitted. For a deep foundation pit project, the basic data of the target foundation pit can be obtained through geological survey reports, design task books, and other materials. These data may include: foundation pit depth: 18m; geological conditions: the upper 0-5m is fill, 5-15m is strongly weathered granite (rock slope, controlled by fissures), and 15-25m is moderately weathered granite, classifying it as a mixed soil-rock foundation pit; groundwater parameters: groundwater level at a depth of 3m, with moderate water volume; boundary conditions: excavation beyond the land use boundary is not permitted.

[0070] In this embodiment, based on the basic data of the target foundation pit, a candidate foundation pit support scheme is obtained by performing knowledge matching from the foundation pit support knowledge graph through a semantic matching algorithm.

[0071] The candidate foundation pit support schemes were obtained, including:

[0072] Based on the basic data of the target foundation pit, geological parameters, groundwater characteristics, depth parameters, red line constraint information, and surrounding building information are extracted from the foundation pit support knowledge graph;

[0073] First, eliminate foundation pit support schemes whose construction scope exceeds the maximum allowable distance beyond the red line or encroaches on the most unfavorable safety protection distance of surrounding buildings, forming a constrained candidate set. Specifically, store the construction distance requirements for various support processes in the knowledge graph in advance, including slope excavation (horizontal distance between the basement outline and the red line, slope ratio and depth), anchor cable (rod) support (horizontal distance between the basement outline and the red line, anchor cable (rod) design length), cantilever pile and anchor pile (horizontal distance between the basement outline and the red line, pile cross-section size), and internal support system. When the horizontal distance between the basement outline and the red line of the target foundation pit does not meet the horizontal distance required for the slope excavation slope ratio, automatically filter out slope ratio excavation schemes. When the allowable distance beyond the red line is a limited value d, filter out all schemes whose construction distance requirements are greater than "d + horizontal distance between the basement outline and the red line".

[0074] For the constrained candidate set, a module query vector is constructed using each module of the knowledge graph as the query benchmark. The modules include a depth adaptation module, a geological adaptation module, a groundwater adaptation module, and a red line constraint adaptation module. The depth adaptation module corresponds to the foundation pit depth dimension, the geological adaptation module corresponds to the geological conditions (including three subclasses of rock) dimension, the groundwater adaptation module corresponds to the groundwater parameter dimension, and the red line constraint adaptation module corresponds to whether exceeding the red line is allowed dimension.

[0075] Based on the graph attention mechanism, the query vector is fused with the module to generate a comprehensive query vector, and the multi-dimensional semantic similarity score between the comprehensive query vector and the case feature nodes in the knowledge graph is calculated.

[0076] Based on multi-dimensional semantic similarity scores and preset thresholds, the top-ranked candidates are selected when each dimension's score exceeds its corresponding threshold and the total score is among the highest. The support schemes corresponding to the feature nodes of the case are used as candidate foundation pit support schemes.

[0077] The multi-dimensional semantic similarity score The formula for calculation is:

[0078] ;

[0079] Among them, k=1 corresponds to the depth adaptation dimension, k=2 corresponds to the geological adaptation dimension, k=3 corresponds to the groundwater adaptation dimension, and k=4 corresponds to the red line adaptation dimension. The semantic weight coefficient for the k-th dimension is... Let k be the engineering fit attenuation factor. To project the query vector into the feature space of the k-th dimension, For transpose operation, Let be the embedding vector of the feature node in the k-th dimension of the case. Let the feature vector be the engineering parameter of the k-th dimension case feature node. Let k be the ideal feature vector of the case feature node in the kth dimension. Let be the tolerance coefficient for the engineering parameter in the k-th dimension.

[0080] Specifically, semantic weights in four dimensions Set to [0.3, 0.3, 0.2, 0.2] (depth, geology, groundwater, red line), attenuation factor Set to [0.1, 0.1, 0.1, 0.1]. Set the score threshold for each dimension to 0.7, and select the top 5 cases with scores ≥0.7 in each dimension and the top total score as candidates, including: pile foundation + anchor cable (excluded because it requires space outside the red line), pile foundation + internal support, and diaphragm wall.

[0081] In this embodiment, the basic data of the target foundation pit and the corresponding scheme parameters of the candidate foundation pit support scheme are input into the feasibility assessment machine learning model for quantitative evaluation; the feasibility assessment machine learning model is a random forest model.

[0082] Then, the performance of each candidate foundation pit support scheme is evaluated to obtain the corresponding scheme evaluation results, which specifically include:

[0083] Based on the target foundation pit project requirements and candidate foundation pit support schemes, evaluation indicators are determined. These evaluation indicators include stability adaptability, construction convenience, economic adaptability, and surrounding environmental protection. Among them, the surrounding environmental protection measure is the degree to which the candidate foundation pit support scheme controls the settlement and horizontal displacement of urban infrastructure such as adjacent building foundations, underground pipelines (including water supply, drainage, gas, communication, and power lines), and municipal roads.

[0084] A feasibility assessment dataset for candidate schemes is constructed based on evaluation indicators and basic data of the foundation pit. The dataset is then processed by extracting the indicator feature data corresponding to the candidate schemes, filtering feature attributes based on knowledge graph theory triples, and performing equalization processing. The SMOTE algorithm is used for equalization processing. By employing this equalization method, the category distribution of the assessment dataset can be improved, enabling the model to fully learn the correlation between different foundation pit conditions, support schemes, and evaluation indicators, thereby improving the accuracy and stability of multidimensional evaluation parameter prediction.

[0085] The evaluation dataset is input into the random forest model to generate multidimensional evaluation parameters. Combined with the indicator weight allocation rules, the feasibility score is calculated.

[0086] Specifically, based on the dynamic calculation of index weights according to the comprehensive risk level of the target foundation pit (determined by the depth level of the foundation pit, the complexity of the surrounding environment, and the complexity of the geological conditions), the weight of stability adaptability is increased for Level 1 risk foundation pits, and the weight of economic adaptability is increased for Level 3 risk foundation pits. For example, the weights corresponding to stability adaptability, construction convenience, economic adaptability, and surrounding environmental protection are set to [0.35, 0.2, 0.25, 0.2]. For pile foundations with internal bracing, the multi-dimensional evaluation parameters include stability adaptability (0.9), construction convenience (0.7), economic adaptability (0.6), and surrounding environmental protection (0.9); for diaphragm walls, the multi-dimensional evaluation parameters include stability adaptability (0.95), construction convenience (0.6), economic adaptability (0.5), and surrounding environmental protection (0.95); therefore, the feasibility score for pile foundations with internal bracing is 0.785, and the feasibility score for diaphragm walls is 0.7675.

[0087] In this embodiment, the applicable national and industry standard technical specifications for the target foundation pit are extracted from the foundation pit support knowledge graph, and the technical parameters of the candidate foundation pit support schemes are verified for compliance, generating a compliance score; among them, the parameters of the candidate foundation pit support schemes that fail the compliance verification are adjusted or removed from the candidate set.

[0088] The compliance testing of candidate foundation pit support schemes includes:

[0089] Based on the geological conditions, depth, and surrounding environment of the target foundation pit, applicable national standards, industry standards, and local standards and regulations are matched from the knowledge graph to extract a set of applicable regulations.

[0090] Technical parameters are extracted from candidate foundation pit support schemes. These technical parameters are then compared item by item with the mandatory indicators specified in the applicable clauses. For candidate schemes with parameters exceeding the limits, the following actions are taken: if the technical parameters can be adjusted, they are corrected to the lower limit of the specification and the scheme parameters are updated; if the technical parameters do not meet the structural requirements, the candidate foundation pit support scheme is removed from the candidate set. The technical parameters include the dimensions of the support piles, pile spacing, embedment depth, design length and prestress value of the anchor cables (rods), cross-section and spacing of the wainscoting beams, drainage and interception measures, slope ratio, etc.

[0091] For candidate foundation pit support schemes that pass compliance verification, a compliance score is calculated based on the weighted average of the margins by which each parameter meets the code requirements. Specifically, the safety margin of each technical parameter relative to the code limits is quantified. For example, if the embedment depth of a support pile exceeds the minimum code requirement by a certain percentage, the corresponding parameter has a high compliance margin. Weights are then assigned based on the importance of different technical parameters, and the margins of each parameter are weighted and integrated to obtain a comprehensive compliance score. The compliance score reflects the safety reserve level and code compliance degree of the candidate support scheme while meeting code constraints.

[0092] In this embodiment, the genetic fitness function value is set based on the feasibility score and compliance score. The candidate foundation pit support scheme is iteratively optimized using a genetic algorithm or similar optimization algorithm. When the termination condition is met, the candidate foundation pit support scheme with the highest feasibility score is selected as the recommended foundation pit support scheme (target foundation pit support scheme).

[0093] The genetic termination conditions include the number of iterations exceeding a preset upper limit and / or the fitness fluctuation value of consecutive iterations being less than a preset fluctuation threshold; the genetic fitness function value is composed of a weighted sum of feasibility score and compliance score; the gene sequence with the highest fitness value in the population is selected, decoded, and a recommended foundation pit support scheme is generated.

[0094] For example, an iteration limit of 40 generations is set, and the fitness fluctuation threshold for continuous iteration is 0.05. Parameters such as pile diameter, pile spacing, and number of supports for the pile + internal bracing scheme are encoded. Genetic operations such as selection, crossover, and mutation are used for iterative optimization to obtain the optimized scheme and its new evaluation value. A similar process is performed on all candidate schemes, and they are ranked according to their final evaluation value. Assuming the final pile + internal bracing scheme has the highest evaluation value due to its superior stability and relatively lower cost under site constraints, it is selected as the primary recommended scheme. Simultaneously, alternative schemes such as diaphragm walls and a detailed evaluation report are output.

[0095] The intelligent recommendation method for foundation pit support schemes of the present invention also includes incremental updates of the knowledge graph:

[0096] After the recommended foundation pit support scheme is completed, the actual construction period, actual project cost, foundation pit support structure deformation monitoring data, and project quality evaluation data of the project are collected and compared with the original predicted data. When the deviation between the actual data and the predicted data exceeds the preset deviation threshold, multi-hop inference optimization is triggered by the updated case feature node data to incrementally update the foundation pit support knowledge graph, thereby forming a closed loop of continuous learning and engineering knowledge accumulation.

[0097] This invention also relates to an intelligent recommendation system for foundation pit support schemes. The recommendation system corresponds to the intelligent recommendation method for foundation pit support schemes in the above embodiments and can be understood as a system that implements the above recommendation method. The recommendation system includes a knowledge graph construction module, a basic data acquisition module, a candidate scheme matching module, a feasibility assessment module, and a scheme optimization recommendation module.

[0098] The knowledge graph construction module is used to acquire knowledge data in the field of foundation pit support and construct a foundation pit support knowledge graph based on the knowledge data.

[0099] The basic data acquisition module is used to acquire basic data of the target foundation pit;

[0100] The candidate scheme matching module is used to perform knowledge matching from the foundation pit support knowledge graph based on the basic data of the target foundation pit to obtain candidate foundation pit support schemes; wherein, the candidate scheme matching module uses a graph structure optimization algorithm to traverse the foundation pit support knowledge graph, and the graph structure optimization algorithm is either Dijkstra's algorithm or Floyd-Warshall's algorithm.

[0101] The feasibility assessment module is used to input the basic data of the target foundation pit and the corresponding scheme parameters of the candidate foundation pit support schemes into a pre-built scheme evaluation model, to evaluate the performance of each candidate foundation pit support scheme, and to obtain the corresponding scheme evaluation results; based on the preset technical specification rules, to perform compliance testing on the candidate foundation pit support schemes, and to obtain compliance testing results.

[0102] The scheme optimization and recommendation module is used to construct a scheme optimization objective function based on the scheme evaluation results and compliance detection results, and to use an intelligent optimization algorithm to iteratively optimize the candidate foundation pit support schemes to obtain a target foundation pit support scheme that meets the preset optimization conditions.

[0103] This invention provides an objective and scientific intelligent recommendation method and system for foundation pit support schemes. The constructed knowledge graph specifically integrates the core constraints of foundation pit support selection, taking foundation pit depth and rock slope geological conditions subdivided into three subcategories as core control elements, groundwater as a key influencing factor, and whether the red line is exceeded as a boundary condition, covering all dimensions of selection and making the knowledge base more comprehensive. Furthermore, the multi-dimensional semantic matching has high accuracy, and a graph attention mechanism is used to fuse query vectors from four dimensions, accurately matching conditions such as rock slope sub-types and red line constraints, overcoming the subjective defects of manual matching.

[0104] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for intelligently recommending foundation pit support schemes, characterized in that: include: Acquire knowledge data in the field of foundation pit support, and construct a foundation pit support knowledge graph based on the knowledge data; Obtain basic data of the target foundation pit; Based on the basic data of the target foundation pit, knowledge matching is performed from the foundation pit support knowledge graph to obtain candidate foundation pit support schemes; The basic data of the target foundation pit and the corresponding scheme parameters of the candidate foundation pit support schemes are input into the pre-built scheme evaluation model to evaluate the performance of each candidate foundation pit support scheme and obtain the corresponding scheme evaluation results. Based on the preset technical specifications and rules, the candidate foundation pit support schemes are tested for compliance, and the compliance test results are obtained. Based on the evaluation results and compliance test results, an objective function for optimizing the scheme is constructed, and an intelligent optimization algorithm is used to iteratively optimize the candidate foundation pit support schemes to obtain the target foundation pit support scheme that meets the preset optimization conditions.

2. The intelligent recommendation method for foundation pit support schemes according to claim 1, characterized in that: The knowledge data includes the technical specifications for foundation pit support, historical cases of foundation pit support, and geotechnical engineering investigation reports. The foundation pit support knowledge graph includes support scheme nodes, influencing factor nodes, evaluation index nodes, and the relationships between each node. The influencing factor nodes include core control element nodes, key influencing factor nodes, and boundary condition nodes. The core control element nodes include the depth of the foundation pit and geological conditions. The geological conditions are further subdivided into rock slope type, rock-soil mixed type, gravelly soil type, sandy soil type, cohesive soil type, and soft soil type. Among them, the rock slope type is further subdivided into sub-types controlled by intensity, sub-types controlled by rock strata, and sub-types controlled by fissures. The key influencing factor nodes include groundwater parameters. The boundary condition nodes include whether excavation beyond the land use boundary line is allowed and whether there are important buildings nearby.

3. The intelligent recommendation method for foundation pit support schemes according to claim 2, characterized in that: Constructing a knowledge graph for foundation pit support, specifically including: Based on the ontology framework of geotechnical engineering, natural language processing algorithms are used to perform entity recognition and relation extraction on the clauses of the technical specification for foundation pit support, resulting in specification triplet; Extract geological conditions, support scheme types, testing and monitoring status, project cost, construction period, slope area of ​​the foundation pit, and foundation pit depth data from historical foundation pit support cases to construct case feature nodes; The graph embedding algorithm is used to embed case feature nodes into the standard triples to construct an initial knowledge graph; Multi-hop reasoning optimization is performed on the case feature nodes and canonical triples in the initial knowledge graph to form a structured and semantic second knowledge graph; The second knowledge graph is divided into modules to obtain the foundation pit support knowledge graph.

4. The intelligent recommendation method for foundation pit support schemes according to claim 1, characterized in that: The candidate foundation pit support schemes were obtained, including: Based on the basic data of the target foundation pit, geological parameters, groundwater characteristics, depth parameters, red line constraint information, and surrounding building information are extracted from the foundation pit support knowledge graph; First, eliminate foundation pit support schemes whose construction scope exceeds the maximum allowable distance beyond the red line or encroaches on the most unfavorable safety protection distance of surrounding buildings, thus forming a constrained candidate set; For the constrained candidate set, a module query vector is constructed using each module of the knowledge graph as the query benchmark; the modules include a depth adaptation module, a geological adaptation module, a groundwater adaptation module, and a red line constraint adaptation module; Based on the graph attention mechanism, the query vector is fused with the module to generate a comprehensive query vector, and the multi-dimensional semantic similarity score between the comprehensive query vector and the case feature nodes in the knowledge graph is calculated. Based on multi-dimensional semantic similarity scores and preset thresholds, the top-ranked candidates are selected when each dimension's score exceeds its corresponding threshold and the total score is among the highest. The support schemes corresponding to the feature nodes of the case are used as candidate foundation pit support schemes.

5. The intelligent recommendation method for foundation pit support schemes according to claim 1, characterized in that: The performance of each candidate foundation pit support scheme was evaluated, and the corresponding scheme evaluation results were obtained, including: Based on the target foundation pit engineering requirements and candidate foundation pit support schemes, evaluation indicators are determined, including stability adaptability, construction convenience, economic adaptability, and surrounding environmental protection. A feasibility assessment dataset for candidate schemes is constructed based on evaluation indicators and basic data of the foundation pit. In this dataset, the indicator feature data corresponding to the candidate schemes are extracted, and feature attributes are selected based on knowledge graph theory triples. The dataset is then balanced. The evaluation dataset is input into the random forest model to generate multidimensional evaluation parameters. Combined with the indicator weight allocation rules, the feasibility score is calculated.

6. The intelligent recommendation method for foundation pit support schemes according to claim 1, characterized in that: The compliance testing of candidate foundation pit support schemes includes: Based on the geological conditions, depth, and surrounding environment of the target foundation pit, applicable national standards, industry standards, and local standards and regulations are matched from the knowledge graph to extract a set of applicable regulations. Technical parameters are extracted from candidate foundation pit support schemes. These technical parameters are then compared item by item with the mandatory indicators specified in the set of applicable provisions. For candidate schemes with parameters exceeding the limits, the following actions are taken: if the technical parameters can be adjusted, they are corrected to the lower limit of the specification and the scheme parameters are updated; if the technical parameters do not meet the structural requirements, the candidate foundation pit support scheme is removed from the candidate set. For candidate foundation pit support schemes that pass the compliance verification, a compliance score is calculated based on the weighted average of the margins by which each parameter meets the specifications.

7. The intelligent recommendation method for foundation pit support schemes according to claim 1, characterized in that: The objective function for optimizing the proposed scheme is a genetic fitness function; and a genetic algorithm or similar optimization algorithm is used to iteratively optimize the candidate foundation pit support scheme.

8. The intelligent recommendation method for foundation pit support schemes according to claim 1, characterized in that: Also includes: Incremental updates to knowledge graphs: After the recommended foundation pit support scheme is completed, the actual construction period, actual project cost, foundation pit support structure deformation monitoring data, and project quality evaluation data of the project are collected and compared with the original predicted data. When the deviation between the actual data and the predicted data exceeds the preset deviation threshold, multi-hop inference optimization is triggered by the updated case feature node data to incrementally update the foundation pit support knowledge graph.

9. An intelligent recommendation system for foundation pit support schemes, characterized in that: It includes a knowledge graph construction module, a basic data acquisition module, a candidate solution matching module, a feasibility assessment module, and a solution optimization and recommendation module; The knowledge graph construction module is used to acquire knowledge data in the field of foundation pit support and construct a foundation pit support knowledge graph based on the knowledge data. The basic data acquisition module is used to acquire basic data of the target foundation pit; The candidate scheme matching module is used to perform knowledge matching from the foundation pit support knowledge graph based on the basic data of the target foundation pit to obtain candidate foundation pit support schemes. The feasibility assessment module is used to input the basic data of the target foundation pit and the corresponding scheme parameters of the candidate foundation pit support schemes into a pre-built scheme evaluation model, to evaluate the performance of each candidate foundation pit support scheme, and to obtain the corresponding scheme evaluation results. Based on the preset technical specifications and rules, the candidate foundation pit support schemes are tested for compliance, and the compliance test results are obtained. The scheme optimization and recommendation module is used to construct a scheme optimization objective function based on the scheme evaluation results and compliance detection results, and to use an intelligent optimization algorithm to iteratively optimize the candidate foundation pit support schemes to obtain a target foundation pit support scheme that meets the preset optimization conditions.

Images