Rock slope ecological protection scheme intelligent recommendation method, system, equipment and medium

Abstract

The invention relates to a rock slope ecological protection scheme intelligent recommendation method and system, equipment and a medium. The method comprises the following steps: acquiring a rock slope ecological protection theory and a historical rock slope ecological protection case, and constructing an ecological protection knowledge graph according to the rock slope ecological protection theory and the historical rock slope ecological protection case; obtaining slope basic data of the target rock slope area, and matching from the knowledge graph through a semantic matching algorithm to obtain a candidate protection scheme; inputting the slope basic data and the parameters of the candidate protection schemes into a protection scheme feasibility evaluation machine learning model, performing feasibility quantitative evaluation and generating an evaluation value; and setting a genetic fitness function value based on the evaluation value, optimizing the candidate protection schemes by adopting a genetic algorithm, and selecting the candidate protection scheme with the highest feasibility quantitative evaluation value as a recommended ecological protection scheme when a genetic termination condition is identified to be met, thereby realizing intelligent recommendation of the rock slope ecological protection scheme. And the scientificity and efficiency of scheme recommendation are improved.

Description

Technical Field

[0001] This invention belongs to the field of geotechnical engineering and ecological protection, and in particular relates to an intelligent recommendation method, system, equipment and medium for ecological protection schemes for rock slopes. Background Technology

[0002] With the development of geotechnical engineering and ecological protection technologies, ecological protection technology for rock slopes has emerged. This technology combines the dual characteristics of slope structural stability reinforcement and surrounding ecological environment restoration, representing an important development direction in geotechnical engineering. Currently, the industry largely relies on the experience of engineering technicians, combined with basic survey data, to formulate protection plans. In traditional techniques, technicians first conduct geological surveys of the target rock slope area to obtain basic data such as slope geology and topography. Combining their own engineering experience with reference to some similar historical cases, they select a protection plan through manual analysis. In some scenarios, the plan is determined simply by comparing the applicability of several conventional plans. This current approach has the following problems: Limited by the personal experience and knowledge of technicians, the protection plan is not well-matched to the actual slope conditions, making it difficult to consider slope geological characteristics, ecological restoration needs, and engineering implementation conditions; there is a lack of systematic exploration and utilization of historical engineering cases, resulting in low efficiency in plan formulation; and the absence of a standardized quantitative evaluation system leads to strong subjectivity in the feasibility judgment, easily resulting in poor protection effects or wasted engineering costs, failing to meet the current scientific and intelligent development needs of rock slope ecological protection engineering. Summary of the Invention

[0003] Therefore, it is necessary to provide an intelligent recommendation method, system, equipment, and medium for ecological protection solutions for rock slopes that can solve the above problems.

[0004] Firstly, this application provides an intelligent recommendation method for ecological protection schemes for rock slopes, including:

[0005] Acquire theories and historical cases of ecological protection for rock slopes, and construct an ecological protection knowledge graph based on these theories and cases.

[0006] Acquire basic slope data for the target rock slope area, and based on the basic slope data, use a semantic matching algorithm to match candidate protection schemes from the ecological protection knowledge graph;

[0007] The basic slope data and the protection scheme parameters of the candidate protection schemes are input into the machine learning model for feasibility assessment of the protection schemes to conduct a quantitative feasibility assessment and generate a quantitative feasibility assessment value.

[0008] The genetic fitness function value of candidate protection schemes is set based on the feasibility quantitative evaluation value. The candidate protection schemes are genetically optimized according to the genetic algorithm. When the genetic termination condition is identified, the candidate protection scheme with the highest feasibility quantitative evaluation value is selected as the recommended ecological protection scheme.

[0009] In one embodiment, an ecological protection knowledge graph is constructed based on the theory of ecological protection of rock slopes and historical cases of ecological protection of rock slopes, including:

[0010] Based on the geotechnical engineering ontology framework, a natural language algorithm is used to identify entities in the theory of ecological protection of rock slopes. The entities in the theory of ecological protection of rock slopes are identified, and the relationships of the entities in the theory of ecological protection of rock slopes are extracted to obtain theoretical triples.

[0011] Engineering parameters, protection schemes, and effect data were extracted from historical cases of ecological protection of rock slopes. Based on these data, case feature nodes were constructed.

[0012] A graph embedding algorithm is used to embed case feature nodes into theoretical triples to construct the first knowledge graph;

[0013] Multi-hop reasoning optimization is performed on the case feature nodes and theoretical triples in the first knowledge graph to obtain the second knowledge graph;

[0014] The second knowledge graph is divided into knowledge modules to obtain an ecological protection knowledge graph.

[0015] In one embodiment, based on basic slope data, candidate protection schemes are obtained from an ecological protection knowledge graph using a semantic matching algorithm, including:

[0016] Based on basic slope data, geological parameters, environmental characteristics, and engineering requirements are extracted from the ecological protection knowledge graph.

[0017] Based on geological parameters, environmental characteristics, and engineering requirements, a module query vector is constructed using each knowledge module in the ecological protection knowledge graph as the query benchmark. Among them, the knowledge modules include stability adaptation modules, ecological adaptation modules, and engineering implementation adaptation modules.

[0018] Based on the graph attention mechanism, the query vector is fused with the module to generate a comprehensive query vector. Based on the comprehensive query vector and the case feature nodes in the ecological protection knowledge graph, the multi-dimensional semantic similarity score of the comprehensive query vector is calculated.

[0019] Based on multi-dimensional semantic similarity scores and multi-dimensional score thresholds, the protection schemes corresponding to the case feature nodes whose semantic similarity scores in each dimension exceed the score thresholds in each dimension, and whose total semantic similarity scores rank in the top N, are selected to obtain candidate protection schemes.

[0020] In one embodiment, the multi-dimensional semantic similarity score is calculated as follows:

[0021]

[0022] in, Let be the semantic weight coefficient of the k-th knowledge module. Let be the engineering fit attenuation factor for the k-th knowledge module. To project the query vector into the feature dimension space corresponding to the k-th knowledge module, For transpose operation, Let be the embedding vector of the case feature node corresponding to the candidate protection scheme in the dimension corresponding to the k-th knowledge module. This represents the feature vector of engineering parameters in the case feature node corresponding to the k-th knowledge module. Let be the ideal feature vector corresponding to the engineering parameter feature vector in the case feature node of the k-th knowledge module. This is the tolerance coefficient for the engineering parameters of the dimension corresponding to the k-th knowledge module.

[0023] In one embodiment, the machine learning model for assessing the feasibility of the protection scheme is a random forest model;

[0024] The basic slope data and the parameters of candidate protection schemes are input into a machine learning model for feasibility assessment of protection schemes to perform quantitative feasibility evaluation and generate quantitative feasibility evaluation values, including:

[0025] Based on the engineering requirements and candidate protection schemes in the basic slope data, the evaluation indicators are determined.

[0026] Based on the evaluation indicators and basic slope data, a feasibility evaluation dataset for candidate solutions is constructed.

[0027] The feasibility assessment dataset of candidate solutions is input into the random forest model to generate multidimensional assessment index parameters. Based on the multidimensional assessment index parameters and the weight allocation rules of the assessment index, the quantitative feasibility assessment value is calculated.

[0028] In one embodiment, the evaluation metrics include stability compatibility, ecological compatibility, engineering implementation difficulty, and economic compatibility.

[0029] Based on evaluation indicators and basic slope data, a feasibility assessment dataset for candidate solutions is constructed, including:

[0030] Based on the evaluation indicators, the indicator feature data corresponding to each evaluation indicator is extracted from the candidate protection schemes;

[0031] Based on the theoretical triples in each knowledge module of the ecological protection knowledge graph, the correlation between basic slope data and evaluation indicators is determined;

[0032] Calculate the correlation strength of the associations, and based on the correlation strength, filter out key feature attributes according to the correlation threshold;

[0033] Based on key feature attributes and relationships, an initial feasibility assessment dataset for the proposed solution is constructed.

[0034] The initial feasibility assessment dataset is balanced to obtain the candidate feasibility assessment dataset.

[0035] In one embodiment, the genetic fitness function value of candidate protection schemes is set based on the feasibility quantification evaluation value. The candidate protection schemes are then genetically optimized using a genetic algorithm. When a genetic termination condition is identified, the candidate protection scheme with the highest feasibility quantification evaluation value is selected as the recommended ecological protection scheme, including:

[0036] Based on the feasibility quantitative evaluation value, construct the fitness function matrix of the genetic algorithm;

[0037] Based on the fitness function matrix and the iteration termination condition, a genetic algorithm is used for iterative calculation to obtain the iterative calculation results;

[0038] Iteration termination conditions include the number of iterations exceeding the upper limit and / or the fitness fluctuation value of consecutive iterations being less than the fitness fluctuation threshold;

[0039] Based on the iterative calculation results, the gene sequence with the highest fitness value in the population is selected, and the gene sequence is decoded to obtain the recommended ecological protection scheme.

[0040] Secondly, this application also provides an intelligent recommendation system for ecological protection schemes for rock slopes, including:

[0041] The knowledge graph construction module is used to acquire theories and historical cases of ecological protection of rock slopes, and to construct an ecological protection knowledge graph based on these theories and cases.

[0042] The candidate scheme matching module is used to acquire the basic slope data of the target rock slope area, and based on the basic slope data, to match candidate protection schemes from the ecological protection knowledge graph through a semantic matching algorithm.

[0043] The feasibility assessment module is used to input the basic slope data and the protection parameters of the candidate protection schemes into the machine learning model for feasibility assessment of the protection schemes, to conduct a quantitative assessment of feasibility and generate a quantitative feasibility assessment value.

[0044] The scheme optimization and recommendation module is used to set the genetic fitness function value of candidate protection schemes based on the feasibility quantitative evaluation value, perform genetic optimization on the candidate protection schemes according to the genetic algorithm, and select the candidate protection scheme with the highest feasibility quantitative evaluation value as the recommended ecological protection scheme when the genetic termination condition is identified.

[0045] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-mentioned intelligent recommendation method for ecological protection schemes of rock slopes.

[0046] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-mentioned intelligent recommendation method for ecological protection schemes of rock slopes.

[0047] The aforementioned intelligent recommendation method, system, equipment, and medium for ecological protection schemes of rock slopes acquire ecological protection theories and historical cases to construct an ecological protection knowledge graph. This systematically integrates multi-source knowledge to form a structured database, thereby reducing reliance on human experience and improving the objectivity and consistency of recommendations. Based on the target slope's basic data, a semantic matching algorithm is used to match candidate protection schemes from the knowledge graph, achieving multi-dimensional and accurate adaptation of geological parameters, environmental characteristics, and engineering requirements, overcoming the subjective and low-precision defects of traditional methods. Slope data and scheme parameters are input into a machine learning model for feasibility quantification assessment, generating quantitative evaluation values ​​that provide scientific and comprehensive judgment criteria, enhancing the accuracy and quantifiability of the assessment. Based on the evaluation values, a genetic fitness function is set, and candidate schemes are optimized through a genetic algorithm. The optimal recommendation is selected when the termination condition is met, efficiently exploring the solution space and achieving the optimal balance between stability, ecology, and economy. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1This is a flowchart of an intelligent recommendation method for ecological protection schemes of rock slopes according to the present invention;

[0050] Figure 2 This is a structural diagram of an intelligent recommendation system for ecological protection schemes of rock slopes according to the present invention. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0052] In one embodiment, such as Figure 1 As shown, an intelligent recommendation method for ecological protection schemes of rock slopes is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, and can also be applied to an architecture that includes both a terminal and a server, and can be implemented through the interaction between the terminal and the server. In the implementation environment, the hardware architecture of this application includes a terminal device (such as a mobile data acquisition device or engineering tablet) and a server. The terminal is responsible for acquiring basic slope data (such as geological parameters and environmental characteristics) of the target rock slope area. The server stores the ecological protection knowledge graph and runs semantic matching and machine learning models. Data synchronization and computation distribution are achieved through a cloud platform or local area network. The application scenario includes: when there is an engineering need for ecological protection of rock slopes, the terminal device collects on-site data and uploads it to the server. The server performs candidate scheme matching and feasibility assessment based on the knowledge graph, generates quantitative values, optimizes them through a genetic algorithm, and returns the recommended scheme to the terminal for display.

[0053] In this embodiment, the method includes the following steps:

[0054] S01. Obtain the theory and historical cases of ecological protection of rock slopes, and construct an ecological protection knowledge graph based on the theory and historical cases of ecological protection of rock slopes.

[0055] Among them, the theory of ecological protection of rock slopes is a professional theoretical system for the structural stability reinforcement and ecological environment restoration of rock slopes; historical cases of ecological protection of rock slopes are a complete record of engineering parameters, protection schemes, implementation effects, and other engineering data corresponding to the implemented ecological protection projects of rock slopes; in practice, the theoretical texts and historical case data of ecological protection of rock slopes can be collected and acquired through multiple channels such as professional literature databases and engineering case databases. When constructing the ecological protection knowledge graph, knowledge graph construction technology can be used according to the relevant ontological framework of geotechnical engineering to extract and associate knowledge elements of theoretical content, and to perform feature mining and structuring of historical cases. By integrating, optimizing, and structuring theoretical knowledge and case features, an ecological protection knowledge graph that can represent the knowledge associations in the field of ecological protection of rock slopes can be formed.

[0056] S02: Obtain basic slope data of the target rock slope area, and based on the basic slope data, obtain candidate protection schemes from the ecological protection knowledge graph through semantic matching algorithm.

[0057] Among them, the basic slope data consists of full-volume basic survey data of geological parameters, topographic features, surrounding environmental conditions, and engineering construction requirements obtained through geological surveys, on-site monitoring, and instrument data collection for the target rock slope area; the semantic matching algorithm refers to various semantic similarity analysis and feature matching algorithms that combine geotechnical engineering with natural language processing and graph computing; in practice, the basic slope data of the target area can be collected and preprocessed using engineering survey equipment. Based on the preprocessed basic slope data, corresponding feature information is extracted from the ecological protection knowledge graph and a query feature vector is constructed. Various semantic matching algorithms are used to perform multi-dimensional matching calculations between the protection scheme-related features stored in the graph and the target slope features. Based on the preset adaptation and screening rules, protection schemes that meet the basic characteristics of the target slope are selected from the graph as candidate protection schemes.

[0058] S03. Input the basic slope data and the protection scheme parameters of the candidate protection schemes into the machine learning model for feasibility assessment of the protection schemes, perform a quantitative feasibility assessment, and generate a quantitative feasibility assessment value.

[0059] The protection scheme parameters include the relevant technical parameters of the candidate protection schemes, such as engineering implementation, structural reinforcement, and ecological restoration. The protection scheme feasibility assessment machine learning model is a variety of machine learning models capable of quantitative analysis of scheme feasibility. The feasibility quantitative assessment value is a numerical result that quantitatively represents the comprehensive feasibility of the candidate protection scheme for the target rock slope. In implementation, the mechanical energy characteristics of the slope's basic data and the protection scheme parameters of the candidate protection schemes are extracted and integrated into an input dataset. This dataset is then input into the protection scheme feasibility assessment machine learning model. The model, relying on its own algorithmic logic, performs multi-dimensional quantitative analysis and calculation of the feasibility of the candidate schemes, generating a feasibility quantitative assessment value that characterizes the level of feasibility of the scheme.

[0060] S04. Based on the feasibility quantitative evaluation value, set the genetic fitness function value of the candidate protection scheme, perform genetic optimization on the candidate protection scheme according to the genetic algorithm, and when the genetic termination condition is identified, select the candidate protection scheme with the highest feasibility quantitative evaluation value as the recommended ecological protection scheme.

[0061] The genetic fitness function value is a quantitative value converted from the feasibility quantitative evaluation value of candidate protection schemes and used as the basis for the genetic algorithm's iterative optimization judgment. The genetic algorithm is a global optimization algorithm based on the principles of natural selection and genetic inheritance. The genetic termination condition consists of various preset criteria for determining the termination of the genetic algorithm's iterative optimization process (set according to computing power and actual engineering needs). The recommended ecological protection scheme is the optimal ecological protection scheme for the target rock slope. In implementation, the corresponding genetic fitness function value can be set for each candidate protection scheme based on the feasibility quantitative evaluation value through numerical transformation, function mapping, etc., and used as the basis for the genetic algorithm to perform global genetic optimization iterative calculation on the candidate protection schemes to generate the recommended ecological protection scheme.

[0062] In one embodiment, an ecological protection knowledge graph is constructed based on the theory of ecological protection of rock slopes and historical cases of ecological protection of rock slopes, including:

[0063] S11. Based on the geotechnical engineering ontology framework, natural language algorithms are used to identify entities in the theory of ecological protection of rock slopes, identify entities in the theory of ecological protection of rock slopes, and extract relations from the entities in the theory of ecological protection of rock slopes to obtain theoretical triples.

[0064] S12 extracts engineering parameters, protection schemes, and effect data from historical ecological protection cases of rock slopes, and constructs case feature nodes based on these parameters, schemes, and effect data.

[0065] S13, using a graph embedding algorithm, embeds case feature nodes into theoretical triples to construct the first knowledge graph;

[0066] S14, perform multi-hop reasoning optimization on the case feature nodes and theoretical triples in the first knowledge graph to obtain the second knowledge graph;

[0067] S15, divide the second knowledge graph into knowledge modules to obtain the ecological protection knowledge graph.

[0068] For example, based on the geotechnical engineering ontology framework, natural language processing algorithms can be used to perform entity recognition on theoretical texts of ecological protection for rock slopes. This identifies theoretical entities such as slope type and protection methods, and uses relation extraction techniques to form theoretical triples, such as structured representations like "anchor support - applicable to - rock slopes," laying the foundation for a knowledge graph. Engineering parameters (such as geological conditions and slope), protection schemes (such as vegetation concrete technology), and effect data (such as stability indicators) can be extracted from historical rock slope ecological protection cases to construct case feature nodes. Each node encapsulates the key attributes of the case, enabling data traceability and standardization. Graph embedding algorithms (such as Node2Vec or graph neural networks) can be used to embed the case feature nodes into the theoretical triples. Vectorization is used to achieve semantic alignment between cases and theories, constructing a first knowledge graph that initially integrates domain knowledge. The first knowledge graph is optimized through multi-hop reasoning, utilizing path reasoning or a rule engine to uncover implicit relationships between entities, such as transitioning from "support methods" to "ecological benefits," enhancing the graph's depth and logical coherence to obtain a more comprehensive second knowledge graph. Based on engineering requirements, the second knowledge graph is divided into knowledge modules, forming specialized subgraphs such as stability adaptation modules, ecological adaptation modules, and engineering implementation adaptation modules. This modularizes the graph and makes it easy to query, constructing an ecological protection knowledge graph that provides structured support for subsequent solution matching.

[0069] In one embodiment, based on basic slope data, candidate protection schemes are obtained from an ecological protection knowledge graph using a semantic matching algorithm, including:

[0070] S21, based on slope basic data, extracts geological parameters, environmental characteristics and engineering requirements from the ecological protection knowledge graph;

[0071] S22, based on geological parameters, environmental characteristics and engineering requirements, uses each knowledge module in the ecological protection knowledge graph as the query benchmark to construct a module query vector; among which, the knowledge modules include stability adaptation module, ecological adaptation module and engineering implementation adaptation module;

[0072] S23. Based on the graph attention mechanism, the query vector is fused with the query vector to generate a comprehensive query vector. Based on the comprehensive query vector and the case feature nodes in the ecological protection knowledge graph, the multi-dimensional semantic similarity score of the comprehensive query vector is calculated.

[0073] S24. Based on the multi-dimensional semantic similarity score and the multi-dimensional score threshold, select the protection schemes corresponding to the case feature nodes whose semantic similarity scores in each dimension exceed the score thresholds in each dimension and whose total semantic similarity scores rank in the top N, and obtain the candidate protection schemes.

[0074] Specifically, based on the basic slope data of the target rock slope area, geological parameters (such as rock mass strength and fracture development), environmental characteristics (such as rainfall and vegetation cover), and engineering requirements (such as safety level and budget constraints) can be extracted from the ecological protection knowledge graph. Using the pre-defined stability adaptation module, ecological adaptation module, and engineering implementation adaptation module in the ecological protection knowledge graph as query benchmarks, module query vectors are constructed. Each module corresponds to a specific feature dimension; for example, the stability module focuses on slope mechanical parameters, and through vectorization, geological parameters are mapped to a high-dimensional space to form the module query vector. A graph attention mechanism (similar to multi-head attention in a graph neural network) is used to fuse these module query vectors to generate a comprehensive query vector. This mechanism dynamically aggregates the importance of each module by calculating attention weights; for example, the stability module may have a higher weight than other modules, generating a more representative comprehensive query vector. Based on the comprehensive query vector and case feature nodes in the knowledge graph, multi-dimensional semantic similarity scores are calculated to quantify semantic and engineering matching. Based on multi-dimensional semantic similarity scores and multi-dimensional score thresholds (such as the minimum score requirement for each module), the protection schemes corresponding to the case feature nodes whose scores in each dimension exceed the threshold and whose total scores rank in the top N are selected as candidate schemes, so as to achieve a balanced optimization of the recommended scheme in terms of stability, ecologicality and implementability.

[0075] In one embodiment, S31, the multi-dimensional semantic similarity score is calculated as follows:

[0076]

[0077] in, Let be the semantic weight coefficient of the k-th knowledge module. Let be the engineering fit attenuation factor for the k-th knowledge module. To project the query vector into the feature dimension space corresponding to the k-th knowledge module, For transpose operation, Let be the embedding vector of the case feature node corresponding to the candidate protection scheme in the dimension corresponding to the k-th knowledge module. This represents the feature vector of engineering parameters in the case feature node corresponding to the k-th knowledge module. Let be the ideal feature vector corresponding to the engineering parameter feature vector in the case feature node of the k-th knowledge module. This is the tolerance coefficient for the engineering parameters of the dimension corresponding to the k-th knowledge module.

[0078] For example, this formula is used to achieve precise matching between candidate protection schemes and target rock slopes. By quantifying the semantic and engineering adaptability between the comprehensive query vector and the case feature nodes in the knowledge graph, it provides a scientific basis for subsequent scheme selection. The overall structure of the formula covers three knowledge modules (stability, ecology, and engineering feasibility), and the calculation of each module includes two parts: the first part is... The weighted cosine similarity term is used to measure the overall query vector. Case node embedding vector The degree of semantic alignment between them, where It is the projection of the query vector onto a specific module. It is an embedded representation of the case nodes, and the cosine value of the included angle is calculated through dot product and norm normalization to reflect the similarity of vector directions. As a semantic weight coefficient, it adjusts the importance of different modules in semantic matching; the second part is... Weighted exponential fit terms, calculating the eigenvectors of engineering parameters based on Gaussian function form. With ideal eigenvectors The Euclidean distance between them is the fit, where Engineering parameters representing real-world examples. It is a preset ideal value. As a tolerance factor, it controls the degree of leniency in the adaptation. The contribution weight of engineering matching is adjusted as a decay factor. By linearly combining these two parts, the fusion of semantic similarity and engineering parameter fit is achieved, enabling the score S to comprehensively reflect the matching quality of the solution across multiple dimensions. The relationship between the variables is as follows: and These are preset hyperparameters, set through training or experience. and Derived from graph attention mechanisms and knowledge graph embedding, and Defined based on historical case data Adjustments are made based on project tolerance, including the comprehensive query vector. The case feature nodes are generated by fusing the module query vectors through a graph attention mechanism. The scoring results derived from the construction of the ecological protection knowledge graph are directly used for subsequent threshold screening (such as multi-dimensional scoring thresholds) and ranking selection, supporting the optimization and recommendation of candidate solutions.

[0079] In one embodiment, the machine learning model for assessing the feasibility of the protection scheme is a random forest model;

[0080] The basic slope data and the parameters of candidate protection schemes are input into a machine learning model for feasibility assessment of protection schemes to perform quantitative feasibility evaluation and generate quantitative feasibility evaluation values, including:

[0081] S41. Based on the engineering requirements and candidate protection schemes in the slope basic data, determine the evaluation indicators;

[0082] S42, Based on the evaluation indicators and basic slope data, construct a feasibility evaluation dataset for candidate schemes;

[0083] S43. Input the candidate solution feasibility assessment dataset into the random forest model to generate multidimensional assessment index parameters. Based on the multidimensional assessment index parameters and the assessment index weight allocation rules, calculate the feasibility quantitative assessment value.

[0084] Specifically, the machine learning model can employ a random forest model to achieve quantitative feasibility assessment: Based on the engineering requirements (such as safety level and budget constraints) in the basic slope data of the target rock slope and the parameters of candidate protection schemes (such as support structure type and ecological materials), assessment indicators are determined, and a candidate scheme feasibility assessment dataset is constructed based on the assessment indicators: The indicator feature data corresponding to each assessment indicator can be extracted from the candidate protection schemes, such as rock mass strength and slope parameters for stability adaptability, and vegetation coverage for ecological adaptability; Based on the theoretical triples in the ecological protection knowledge graph, the correlation between the basic slope data and assessment indicators is analyzed, and key feature attributes (such as the impact of fracture development degree on stability) are screened by calculating the correlation strength and based on a preset correlation threshold; An initial dataset is constructed based on the key attributes, and equalization processing (such as oversampling or weight adjustment) is performed to eliminate data bias, forming a candidate scheme feasibility assessment dataset. The dataset is input into a random forest model, which processes the feature data in parallel through multiple decision trees to generate multidimensional evaluation index parameters (such as stability score and ecological score). Combined with the evaluation index weight allocation rules (such as the weights set by the analytic hierarchy process), a comprehensive feasibility quantitative evaluation value is obtained by weighted calculation. This value can serve as an objective basis for scheme optimization.

[0085] In one embodiment, the evaluation metrics include stability compatibility, ecological compatibility, engineering implementation difficulty, and economic compatibility.

[0086] Based on evaluation indicators and basic slope data, a feasibility assessment dataset for candidate solutions is constructed, including:

[0087] S51, based on the evaluation indicators, extract the indicator feature data corresponding to each evaluation indicator from the candidate protection schemes;

[0088] S52, based on the theoretical triples in each knowledge module of the ecological protection knowledge graph, determines the correlation between basic slope data and evaluation indicators;

[0089] S53, calculate the correlation strength of the association, and based on the correlation strength, filter out key feature attributes according to the correlation threshold;

[0090] S54. Construct an initial feasibility assessment dataset based on key feature attributes and relationships;

[0091] S55. The initial feasibility assessment dataset is balanced to obtain the candidate feasibility assessment dataset.

[0092] For example, the evaluation indicators include stability adaptability, ecological adaptability, engineering implementation difficulty, and economic adaptability. These indicators quantify the degree of adaptability of the scheme in terms of slope mechanical stability, ecological restoration effect, construction complexity, and cost-effectiveness. In implementation, based on the evaluation indicators, the characteristic data corresponding to each indicator can be extracted from the candidate protection schemes. For example, features such as rock mass strength and slope angle can be obtained from the support scheme parameters for stability adaptability; vegetation coverage and biodiversity index can be extracted from the ecological material types for ecological adaptability; machinery utilization rate and construction period length can be derived from the construction plan for engineering implementation difficulty; and material costs and maintenance expenditures can be collected from cost reports for economic adaptability. Based on the theoretical triples in each knowledge module of the ecological protection knowledge graph, the correlation between slope basic data (such as the degree of fracture development in geological parameters) and evaluation indicators can be analyzed through graph query technology. For example, the direct impact of slope data on stability adaptability or the indirect correlation between environmental characteristics and ecological adaptability can be derived using triple logic. The correlation strength of these relationships is calculated using statistical methods such as Pearson correlation coefficient or mutual information metric to quantify the dependence between slope characteristics and indicators. Key feature attributes are selected based on a preset correlation threshold (e.g., a correlation coefficient greater than 0.7 is considered a strong correlation), such as identifying rock mass strength and rainfall as core stability factors, thus avoiding redundant data interference. Based on the key feature attributes and the established correlations, an initial feasibility assessment dataset is constructed. Slope basic data, scheme parameters, and key attributes are organized into tables using data integration techniques. The initial dataset is then balanced using oversampling methods (such as the SMOTE algorithm) or weight adjustment strategies to balance the distribution of different categories of data, preventing model bias, and resulting in a candidate feasibility assessment dataset.

[0093] In one embodiment, the genetic fitness function value of candidate protection schemes is set based on the feasibility quantification evaluation value. The candidate protection schemes are then genetically optimized using a genetic algorithm. When a genetic termination condition is identified, the candidate protection scheme with the highest feasibility quantification evaluation value is selected as the recommended ecological protection scheme, including:

[0094] S61, Based on the feasibility quantification evaluation value, construct the fitness function matrix of the genetic algorithm;

[0095] S62, based on the fitness function matrix and the iteration termination condition, uses a genetic algorithm to perform iterative calculations and obtain the iterative calculation results;

[0096] S63, the iteration termination conditions include the number of iterations exceeding the upper limit of the number of iterations and / or the fitness fluctuation value of consecutive iterations being less than the fitness fluctuation threshold;

[0097] S64. Based on the iterative calculation results, the gene sequence with the highest fitness value in the population is selected, and the gene sequence is decoded to obtain the recommended ecological protection scheme.

[0098] For example, a fitness function matrix for the genetic algorithm can be constructed based on the feasibility quantification evaluation value: the feasibility quantification evaluation value (the comprehensive score output by the random forest model) of each candidate protection scheme is converted into a fitness value through a linear mapping or normalization function, such as normalizing the evaluation value to the [0,1] interval using min-max scaling, forming a fitness function matrix, where the rows of the matrix correspond to the individual individuals in the scheme population, and the columns represent the fitness scores, providing a quantitative evaluation basis for the genetic algorithm. Based on this fitness function matrix and the preset iteration termination condition, the genetic algorithm is used for iterative calculation: during initialization, the candidate protection scheme is encoded as a gene sequence, such as using binary or real number encoding to represent the scheme parameters (such as support type, material usage), generating an initial population; in each generation iteration, parent individuals are selected based on fitness values ​​using roulette wheel selection or tournament selection, single-point crossover or uniform crossover operations are performed to generate offspring, random mutations (such as bit flips or Gaussian perturbations) are introduced to maintain population diversity, and the scheme parameters are gradually optimized. The iterative calculation process continues until the termination condition is met, and the iterative calculation result is obtained. The iteration termination conditions include the number of iterations exceeding a preset upper limit (e.g., 40 generations) or the fluctuation value (e.g., standard deviation) of the population fitness during continuous iterations being less than the fitness fluctuation threshold (e.g., 0.05). Based on the iterative calculation results, the gene sequence with the highest fitness value is selected from the population, and the gene sequence is restored to specific protection scheme parameters through decoding operations. For example, the binary code is mapped back to engineering indicators such as anchor length or vegetation density to obtain a recommended ecological protection scheme. This scheme achieves an optimal balance in terms of stability, ecology, and economy.

[0099] The aforementioned intelligent recommendation method for ecological protection schemes of rock slopes constructs an ecological protection knowledge graph, integrating rock slope ecological protection theories and historical cases into a structured knowledge base. It utilizes natural language processing and graph embedding algorithms to extract theoretical triples and case feature nodes, achieving standardized fusion of multi-source knowledge, reducing reliance on the subjective experience of technical personnel, and improving the objectivity and consistency of scheme recommendations. Based on the target slope's basic data, it matches candidate protection schemes using a semantic matching algorithm. This algorithm employs a graph attention mechanism to generate a comprehensive query vector and calculate multi-dimensional semantic similarity scores, achieving adaptation to geological parameters, environmental characteristics, and engineering requirements. This overcomes the shortcomings of traditional methods, such as subjective matching and low accuracy, and enhances the overall effectiveness of the recommendation. The system enhances the targeting and reliability of scheme selection; it inputs slope data and scheme parameters into a random forest model for quantitative feasibility assessment, constructs a feasibility assessment dataset for candidate schemes and generates multi-dimensional assessment index parameters, and calculates a comprehensive assessment value based on weight allocation rules, providing a scientific and quantifiable basis for judgment and improving the accuracy and comprehensiveness of the assessment; it sets a genetic fitness function based on the assessment value and uses a genetic algorithm for iterative optimization, exploring the solution space through encoding, selection, crossover, and mutation operations, and decoding the optimal gene sequence when the iteration termination condition is met, efficiently achieving the optimal balance of schemes in terms of stability, ecology, and economy, and comprehensively promoting the development of rock slope ecological protection engineering towards intelligence and efficiency.

[0100] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0101] Based on the same inventive concept, this application also provides an intelligent recommendation system for ecological protection schemes of rock slopes, used to implement the intelligent recommendation method for ecological protection schemes of rock slopes described above. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the intelligent recommendation system for ecological protection schemes of rock slopes provided below can be found in the limitations of the intelligent recommendation method for ecological protection schemes of rock slopes described above, and will not be repeated here.

[0102] In one exemplary embodiment, such as Figure 2As shown, an intelligent recommendation system for ecological protection schemes of rock slopes is provided, including:

[0103] The knowledge graph construction module 101 is used to acquire the theory of ecological protection of rock slopes and historical cases of ecological protection of rock slopes, and to construct an ecological protection knowledge graph based on the theory of ecological protection of rock slopes and historical cases of ecological protection of rock slopes.

[0104] The candidate scheme matching module 102 is used to acquire the basic slope data of the target rock slope area, and based on the basic slope data, to match candidate protection schemes from the ecological protection knowledge graph through a semantic matching algorithm.

[0105] The feasibility assessment module 103 is used to input the basic data of the slope and the protection parameters of the candidate protection schemes into the machine learning model for feasibility assessment of the protection scheme, to conduct a quantitative feasibility assessment and generate a quantitative feasibility assessment value.

[0106] The scheme optimization and recommendation module 104 is used to set the genetic fitness function value of candidate protection schemes based on the feasibility quantitative evaluation value, perform genetic optimization on the candidate protection schemes according to the genetic algorithm, and select the candidate protection scheme with the highest feasibility quantitative evaluation value as the recommended ecological protection scheme when the genetic termination condition is identified.

[0107] In one embodiment, the knowledge graph construction module 101 is further configured to:

[0108] Based on the geotechnical engineering ontology framework, a natural language algorithm is used to identify entities in the theory of ecological protection of rock slopes. The entities in the theory of ecological protection of rock slopes are identified, and the relationships of the entities in the theory of ecological protection of rock slopes are extracted to obtain theoretical triples.

[0109] Engineering parameters, protection schemes, and effect data were extracted from historical cases of ecological protection of rock slopes. Based on these data, case feature nodes were constructed.

[0110] A graph embedding algorithm is used to embed case feature nodes into theoretical triples to construct the first knowledge graph;

[0111] Multi-hop reasoning optimization is performed on the case feature nodes and theoretical triples in the first knowledge graph to obtain the second knowledge graph;

[0112] The second knowledge graph is divided into knowledge modules to obtain an ecological protection knowledge graph.

[0113] In one embodiment, the candidate solution matching module 102 is further configured to:

[0114] Based on basic slope data, geological parameters, environmental characteristics, and engineering requirements are extracted from the ecological protection knowledge graph.

[0115] Based on geological parameters, environmental characteristics, and engineering requirements, a module query vector is constructed using each knowledge module in the ecological protection knowledge graph as the query benchmark. Among them, the knowledge modules include stability adaptation modules, ecological adaptation modules, and engineering implementation adaptation modules.

[0116] Based on the graph attention mechanism, the query vector is fused with the module to generate a comprehensive query vector. Based on the comprehensive query vector and the case feature nodes in the ecological protection knowledge graph, the multi-dimensional semantic similarity score of the comprehensive query vector is calculated.

[0117] Based on multi-dimensional semantic similarity scores and multi-dimensional score thresholds, the protection schemes corresponding to the case feature nodes whose semantic similarity scores in each dimension exceed the score thresholds in each dimension, and whose total semantic similarity scores rank in the top N, are selected to obtain candidate protection schemes.

[0118] In one embodiment, the multi-dimensional semantic similarity score in the candidate solution matching module 102 is calculated as follows:

[0119]

[0120] in, Let be the semantic weight coefficient of the k-th knowledge module. Let be the engineering fit attenuation factor for the k-th knowledge module. To project the query vector into the feature dimension space corresponding to the k-th knowledge module, For transpose operation, Let be the embedding vector of the case feature node corresponding to the candidate protection scheme in the dimension corresponding to the k-th knowledge module. This represents the feature vector of engineering parameters in the case feature node corresponding to the k-th knowledge module. Let be the ideal feature vector corresponding to the engineering parameter feature vector in the case feature node of the k-th knowledge module. This is the tolerance coefficient for the engineering parameters of the dimension corresponding to the k-th knowledge module.

[0121] In one embodiment, the machine learning model for assessing the feasibility of the protection scheme in the scheme feasibility assessment module 103 is a random forest model;

[0122] The feasibility assessment module 103 is also used for:

[0123] Based on the engineering requirements and candidate protection schemes in the basic slope data, the evaluation indicators are determined.

[0124] Based on the evaluation indicators and basic slope data, a feasibility evaluation dataset for candidate solutions is constructed.

[0125] The feasibility assessment dataset of candidate solutions is input into the random forest model to generate multidimensional assessment index parameters. Based on the multidimensional assessment index parameters and the weight allocation rules of the assessment index, the quantitative feasibility assessment value is calculated.

[0126] In one embodiment, the evaluation indicators in the feasibility assessment module 103 include stability adaptability, ecological adaptability, engineering implementation difficulty, and economic adaptability.

[0127] The feasibility assessment module 103 is also used for:

[0128] Based on the evaluation indicators, the indicator feature data corresponding to each evaluation indicator is extracted from the candidate protection schemes;

[0129] Based on the theoretical triples in each knowledge module of the ecological protection knowledge graph, the correlation between basic slope data and evaluation indicators is determined;

[0130] Calculate the correlation strength of the associations, and based on the correlation strength, filter out key feature attributes according to the correlation threshold;

[0131] Based on key feature attributes and relationships, an initial feasibility assessment dataset for the proposed solution is constructed.

[0132] The initial feasibility assessment dataset is balanced to obtain the candidate feasibility assessment dataset.

[0133] In one embodiment, the scheme optimization and recommendation module 104 is further configured to:

[0134] Based on the feasibility quantitative evaluation value, construct the fitness function matrix of the genetic algorithm;

[0135] Based on the fitness function matrix and the iteration termination condition, a genetic algorithm is used for iterative calculation to obtain the iterative calculation results;

[0136] Iteration termination conditions include the number of iterations exceeding the upper limit and / or the fitness fluctuation value of consecutive iterations being less than the fitness fluctuation threshold;

[0137] Based on the iterative calculation results, the gene sequence with the highest fitness value in the population is selected, and the gene sequence is decoded to obtain the recommended ecological protection scheme.

[0138] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the intelligent recommendation method for ecological protection schemes of rock slopes as described above.

[0139] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0140] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0141] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for intelligently recommending ecological protection schemes for rock slopes, characterized in that, The method includes: Acquire theories and historical cases of ecological protection for rock slopes, and construct an ecological protection knowledge graph based on these theories and cases. Acquire basic slope data of the target rock slope area, and based on the basic slope data, obtain candidate protection schemes from the ecological protection knowledge graph through a semantic matching algorithm; The basic slope data and the protection scheme parameters of the candidate protection schemes are input into the machine learning model for feasibility assessment of the protection schemes to perform a quantitative feasibility assessment and generate a quantitative feasibility assessment value. Based on the feasibility quantification evaluation value, the genetic fitness function value of the candidate protection scheme is set, and the candidate protection scheme is genetically optimized according to the genetic algorithm. When the genetic termination condition is identified, the candidate protection scheme with the highest feasibility quantification evaluation value is selected as the recommended ecological protection scheme.

2. The method according to claim 1, characterized in that, The construction of an ecological protection knowledge graph based on the aforementioned ecological protection theory for rock slopes and historical cases of ecological protection for rock slopes includes: Based on the geotechnical engineering ontology framework, a natural language algorithm is used to identify entities of the ecological protection theory of rock slopes, identify entities of the ecological protection theory of rock slopes, and extract relations from the entities of the ecological protection theory of rock slopes to obtain theoretical triples. Engineering parameters, protection schemes, and effect data are extracted from the historical ecological protection cases of rock slopes. Based on the engineering parameters, protection schemes, and effect data, case feature nodes are constructed. A graph embedding algorithm is used to embed the case feature nodes into the theoretical triples to construct the first knowledge graph; Multi-hop reasoning optimization is performed on the case feature nodes and theoretical triples in the first knowledge graph to obtain the second knowledge graph; The second knowledge graph is divided into knowledge modules to obtain the ecological protection knowledge graph.

3. The method according to claim 2, characterized in that, The process of obtaining candidate protection schemes from the ecological protection knowledge graph based on the slope baseline data using a semantic matching algorithm includes: Based on the slope basic data, geological parameters, environmental characteristics and engineering requirements are extracted from the ecological protection knowledge graph; Based on the geological parameters, environmental characteristics, and engineering requirements, a module query vector is constructed using each knowledge module in the ecological protection knowledge graph as a query benchmark; wherein, the knowledge module includes a stability adaptation module, an ecological adaptation module, and an engineering implementation adaptation module; Based on the graph attention mechanism, the module query vectors are fused to generate a comprehensive query vector. Based on the comprehensive query vector and the case feature nodes in the ecological protection knowledge graph, the multi-dimensional semantic similarity score of the comprehensive query vector is calculated. Based on the multi-dimensional semantic similarity score and the multi-dimensional score threshold, the protection schemes corresponding to the case feature nodes whose semantic similarity scores in each dimension exceed the score thresholds in each dimension of the multi-dimensional score, and whose total semantic similarity scores in each dimension rank in the top N, are selected to obtain the candidate protection schemes.

4. The method according to claim 3, characterized in that, The formula for calculating the multi-dimensional semantic similarity score is: in, The semantic weight coefficient of the k-th knowledge module is given. Let be the engineering adaptability attenuation factor for the k-th knowledge module. The projection of the comprehensive query vector into the feature dimension space corresponding to the k-th knowledge module is... For transpose operation, The embedding vector of the case feature node corresponding to the candidate protection scheme in the dimension corresponding to the k-th knowledge module. The engineering parameter feature vector is the feature vector of the case feature node corresponding to the kth knowledge module. This refers to the ideal feature vector corresponding to the engineering parameter feature vector in the case feature node of the k-th knowledge module. The tolerance coefficient is the engineering parameter of the dimension corresponding to the k-th knowledge module.

5. The method according to claim 3, characterized in that, The machine learning model for assessing the feasibility of the protection scheme is a random forest model. The step of inputting the basic slope data and the protection scheme parameters of the candidate protection schemes into the machine learning model for feasibility assessment of the protection schemes, performing a quantitative feasibility assessment, and generating a quantitative feasibility assessment value includes: Based on the engineering requirements in the slope basic data and the candidate protection schemes, the evaluation indicators are determined; Based on the evaluation indicators and the basic slope data, a feasibility evaluation dataset for candidate solutions is constructed. The candidate solution feasibility assessment dataset is input into the random forest model to generate multidimensional assessment index parameters. Based on the multidimensional assessment index parameters and the assessment index weight allocation rules, the feasibility quantitative assessment value is calculated.

6. The method according to claim 5, characterized in that, The evaluation indicators include stability adaptability, ecological adaptability, engineering implementation difficulty, and economic adaptability. The process of constructing a feasibility assessment dataset for candidate solutions based on the evaluation indicators and the basic slope data includes: Based on the evaluation indicators, extract the indicator feature data corresponding to each evaluation indicator from the candidate protection schemes; Based on the theoretical triples in each knowledge module of the ecological protection knowledge graph, the correlation between the basic slope data and the evaluation indicators is determined; Calculate the correlation strength of the association, and based on the correlation strength, filter out key feature attributes according to the correlation threshold; Based on the key feature attributes and the correlation relationships, an initial scheme feasibility assessment dataset is constructed; The initial feasibility assessment dataset is subjected to equalization processing to obtain the candidate feasibility assessment dataset.

7. The method according to claim 5, characterized in that, The process involves setting the genetic fitness function value of the candidate protection scheme based on the feasibility quantification evaluation value, performing genetic optimization on the candidate protection scheme using a genetic algorithm, and selecting the candidate protection scheme with the highest feasibility quantification evaluation value as the recommended ecological protection scheme when a genetic termination condition is identified. This includes: Based on the aforementioned feasibility quantification evaluation value, construct the fitness function matrix of the genetic algorithm; Based on the fitness function matrix and the iteration termination condition, the genetic algorithm is used to perform iterative calculations to obtain the iterative calculation results; The iteration termination conditions include the number of iterations exceeding the upper limit of the number of iterations and / or the fitness fluctuation value of consecutive iterations being less than the fitness fluctuation threshold; Based on the iterative calculation results, the gene sequence with the highest fitness value in the population is selected, and the gene sequence is decoded to obtain the recommended ecological protection scheme.

8. An intelligent recommendation system for ecological protection schemes for rock slopes, characterized in that, The system includes: The knowledge graph construction module is used to acquire ecological protection theories and historical cases of ecological protection of rock slopes, and to construct an ecological protection knowledge graph based on the ecological protection theories and historical cases of ecological protection of rock slopes. The candidate scheme matching module is used to acquire basic slope data of the target rock slope area, and based on the basic slope data, to match candidate protection schemes from the ecological protection knowledge graph through a semantic matching algorithm. The feasibility assessment module is used to input the basic slope data and the protection parameters of the candidate protection schemes into the protection scheme feasibility assessment machine learning model to perform a quantitative feasibility assessment and generate a quantitative feasibility assessment value. The scheme optimization and recommendation module is used to set the genetic fitness function value of the candidate protection scheme based on the feasibility quantitative evaluation value, perform genetic optimization on the candidate protection scheme according to the genetic algorithm, and select the candidate protection scheme with the highest feasibility quantitative evaluation value as the recommended ecological protection scheme when the genetic termination condition is identified.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images