Geotechnical data structured conversion method

Abstract

The invention belongs to the technical field of data processing, and particularly relates to a rock-soil data structured conversion method. According to the method, unstructured data of rock and soil logs is split into minimum information units of entity, attribute and relation clues through information particle deconstruction, standardized entities and relations are generated through clue association network clustering and double-track reasoning model processing, and a knowledge graph is constructed and mapped into structured data; and constructing an initial geologic model by using structured data, extracting core data, converting the core data into an extended IFC format intermediate file containing a geologic special entity, and finally associating a BIM model to realize data fusion. According to the method, structured conversion of unstructured data is realized, a data format barrier is broken through, and the engineering collaboration efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and in particular relates to a method for structured conversion of geotechnical data. Background Art

[0002] In the current field of geological and mining engineering, a large amount of geotechnical data exists in unstructured forms such as log files. The data is scattered and the format is chaotic, making it difficult to use directly for modeling and analysis. At the same time, there are barriers between the data formats of geological models and BIM models. The existing IFC format lacks dedicated entities that adapt to the unique properties of strata. It is difficult to effectively map geological core data (geometry, attributes, spatial relationships) to BIM models, resulting in the separation of geological information and building information, seriously affecting the efficiency of data sharing and collaborative decision-making in engineering design and construction, and increasing modeling errors and engineering risks. Summary of the Invention

[0003] In view of the technical problems existing in the above background technology, the present invention proposes a structured conversion method for geotechnical data.

[0004] In order to achieve the above object, the technical solution adopted by the present invention includes the following steps:

[0005] Obtaining an unstructured data set of geotechnical log files, performing information granular deconstruction on the data set, breaking the data into minimum information units containing entity clues, attribute clues, and relationship clues, and marking each information unit with a clue type and confidence level;

[0006] Construct a clue association network, calculate the association weights between information units based on clue co-occurrence frequency and semantic consistency, and form a weighted network topology;

[0007] Clustering the weighted network topology using an adaptive community discovery algorithm to obtain a number of cluster communities;

[0008] Construct a dual-track reasoning model, which includes a clue aggregation layer and a knowledge completion layer:

[0009] The clue aggregation layer performs feature fusion on the information units in each cluster community to generate entity candidate sets and relationship candidate sets;

[0010] The knowledge completion layer is used to perform logical verification and missing information completion on the entity candidate set and the relationship candidate set, and output standardized entity results and relationship results;

[0011] Constructing a geotechnical knowledge graph based on the entity results and the relationship results, and mapping the knowledge graph into a structured data format;

[0012] Constructing an initial geological model using structured data, extracting core data from the initial geological model, including stratum geometry data, attribute data, and spatial relationship data, and converting the data into an intermediate data file that conforms to an extended IFC format. The extended IFC format adds geological-specific entities to accommodate stratum-specific attributes.

[0013] The intermediate data file is associated with the BIM model, and the geological-specific entities are matched with the component attributes in the BIM model through data mapping rules to obtain data that can generate a BIM model containing geological information.

[0014] Preferably, the data set is decomposed into information granules, breaking the data into the smallest information units containing entity clues, attribute clues, and relationship clues, including:

[0015] Natural language processing technology is used to perform word segmentation and part-of-speech tagging on the unstructured data of geotechnical log files, identifying noun phrases in the text as entity clue candidates, quantity phrases and adjective phrases as attribute clue candidates, and prepositional phrases and verb phrases as relationship clue candidates;

[0016] Classify candidate clues based on the pre-trained BERT model in the geotechnical engineering field and determine the clue type;

[0017] The confidence of clue classification is calculated using a confusion matrix, which is obtained by the ratio of the number of correctly classified clues to the total number of clues, where the total number of clues is the sum of the number of entity clues, attribute clues, and relationship clues.

[0018] Preferably, the construction of the clue association network, calculating the association weights between information units based on clue co-occurrence frequency, semantic consistency and domain logic association, and forming a weighted network topology is specifically implemented as follows:

[0019] First, determine the network nodes: the smallest information unit obtained by deconstructing information particles is the network node, and each node carries its clue type and confidence label;

[0020] Calculate the co-occurrence frequency of clues, count the number of times any two information units co-occur in the same text paragraph, table row or adjacent sentences of the unstructured data of the geotechnical log file, and take the ratio of this number to the total number of paragraphs in the data set as the co-occurrence frequency;

[0021] Calculate semantic consistency. Based on the pre-trained BERT model, extract the context vector corresponding to the text content of each information unit, and calculate the semantic consistency of two information units through cosine similarity;

[0022] The co-occurrence frequency of clues and semantic consistency are weighted and fused to obtain the association weights between information units;

[0023] A clue association network is constructed with information units as nodes and association weights as edges. If the association weight of two information units is greater than or equal to the set weight threshold, a connection is established in the network, and finally a weighted network topology containing association relationships is obtained.

[0024] Preferably, the method of clustering the weighted network topology using an adaptive community discovery algorithm to obtain a number of cluster communities is specifically implemented as follows:

[0025] First, the initial seed node is generated based on the weight density. The weight density of the node is the sum of the association weights of the node and its adjacent nodes, and is calculated as follows: ,in, is the i-th information unit node in the network, Representation node The set of adjacent nodes of Representation node and adjacent nodes The association weight between Represented as a node The weight density of

[0026] Calculate the mean weight density of all nodes in the weighted network and standard deviation , set dynamic threshold , the weight density is greater than or equal to The nodes are selected as candidate seed nodes;

[0027] De-redundancy processing is performed on candidate seed nodes. If two candidate seed nodes The association weight between If the value is greater than or equal to the set strong correlation threshold, the nodes with higher weight density are retained as the initial clustering centers, and finally a set of seed nodes is formed. , where m is the number of initial cluster centers;

[0028] Compute nodes and cluster centers The weighted distance is calculated as ,in, is the center node of the kth cluster, For nodes and cluster centers The association weight between is the average value of the association weights between all nodes in the kth cluster. Based on the above weighted distance, the nodes are Assign to the cluster with the smallest distance;

[0029] The kth cluster is defined as a community, and the community cluster cohesion is calculated as follows: ,in, Expressed as the cohesion of the k-th community cluster, For nodes The association weight of is the sum of the total network weight, For nodes The sum of the association weights with all adjacent nodes, is the penalty coefficient, is the number of nodes in the kth community;

[0030] like If the value is less than zero, the community will be split according to the maximum weight difference, and the difference in the associated weights between all nodes in the community will be calculated. The pair of nodes with the largest difference will be selected as the new center, and the original community will be split into two sub-communities. If the cohesion of the community cluster after the two adjacent communities are merged is When the percentage reaches the set threshold compared to before the merge, the merge operation is performed; the above split-merge process is repeated until the set number of iterations is reached and the iteration is stopped.

[0031] Preferably, the clue aggregation layer performs feature fusion on the information units in each cluster community to generate entity candidate sets and relationship candidate sets, including:

[0032] Extract core features from the information units within each cluster community. These core features include the textual semantic features of entity clues, the numerical and descriptive features of attribute clues, and the semantic association features of relationship clues. The clue type label and confidence level of each information unit are also introduced as auxiliary features.

[0033] Based on the association weights between information units within the clustering community, a weighted feature fusion strategy is adopted to combine entity clues with closely associated attribute clues to form preliminary entity candidates, and match relationship clues with associated entity clues to form relationship candidates, generating entity candidate sets and relationship candidate sets corresponding to each clustering community.

[0034] Preferably, the knowledge completion layer is used to perform logic verification and missing information completion on the entity candidate set and the relationship candidate set, and output standardized entity results and relationship results, including:

[0035] Perform logical verification on the entity candidate set and the relationship candidate set, checking the rationality of the entity attributes of the entity candidate set and verifying the logic of the relationship between entities for the relationship candidate set;

[0036] For missing information found after verification, reasoning and completion are performed based on high-confidence related clues in the clue association network and existing knowledge;

[0037] Output standardized entity results and relationship results.

[0038] Preferably, the method of constructing an initial geological model using structured data, extracting core data from the initial geological model, including stratum geometry data, attribute data, and spatial relationship data, and converting the data into an intermediate data file conforming to the extended IFC format is specifically implemented as follows:

[0039] Design and extend the IFC format, add geological entities, including stratigraphic unit entities, geological attribute set entities, and spatial topology entities;

[0040] Establish a mapping relationship between core data and extended IFC format, map stratigraphic geometry data to geometric parameters of stratigraphic unit entities, map attribute data to characteristic parameters of geological attribute set entities, and map spatial relationship data to association parameters of spatial topology entities;

[0041] According to the mapping relationship, the extracted core data is converted into an intermediate data file that complies with the extended IFC format.

[0042] Preferably, associating the intermediate data file with the BIM model includes: importing the intermediate data file into the BIM platform using the IFC standard interface, and aligning the coordinate system of the geological model with the coordinate system of the BIM model through a coordinate conversion algorithm to achieve spatial position alignment.

[0043] Compared with the existing technology, the advantages and positive effects of the present invention are that, through information granule deconstruction and dual-track reasoning model, unstructured geotechnical log data is converted into standardized entities and relationships, which solves the problem of scattered and chaotic data and difficulty in direct modeling, and provides a structured basis for geological analysis; innovative design expands the IFC format, adds geological-specific entities, breaks through the data barriers between geological models and BIM models, and realizes the effective mapping of stratigraphic geometry, attributes and spatial relationship data; through coordinate unification and attribute matching, it promotes the deep integration of geological information and building information, improves data sharing and collaborative decision-making efficiency in engineering design and construction, reduces modeling errors and engineering risks, and provides a new solution for the efficient application of geological and mining engineering data. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following is a brief introduction to the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0045] Figure 1 This is a structural flow chart of a geotechnical data structured conversion method. DETAILED DESCRIPTION

[0046] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described below in conjunction with the accompanying drawings and embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features therein can be combined with each other.

[0047] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways than those described herein. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0048] In the current field of geological and mining engineering, data management and application face two core issues. On the one hand, a large amount of data exists in unstructured forms such as log files and survey reports. These data contain key information such as stratum distribution, geotechnical properties, and geological structure. However, due to chaotic formats and inconsistent representations, they are difficult to use directly for modeling and analysis. On the other hand, there is a natural barrier between the data formats of geological models and building information models (BIM). The existing Industrial Foundation Class (IFC) standard lacks dedicated entities adapted to the unique properties of strata (such as permeability and compression modulus). As a result, core geological data (stratum geometry, physical and mechanical properties, and spatial topological relationships) cannot be effectively mapped to BIM models. Based on this, a structured conversion method for geotechnical data is proposed.

[0049] First, an unstructured data set of geotechnical log files is acquired and deconstructed into information granules, breaking the data into minimal information units consisting of entity clues, attribute clues, and relationship clues. Each information unit is labeled with a clue type and confidence score. Natural language processing techniques are used to segment and tag the unstructured geotechnical log file data. Noun phrases in the text are identified as entity clue candidates, quantity phrases and adjective phrases as attribute clue candidates, and prepositional phrases and verb phrases as relationship clue candidates. Candidate clues are classified based on a pretrained BERT model for geotechnical engineering to determine their type. A confusion matrix is ​​used to calculate the confidence score of the clue classification, which is calculated as the ratio of the number of correctly classified clues to the total number of clues, which is the sum of the number of entity clues, attribute clues, and relationship clues. Specifically, the unstructured text of the geotechnical log files is preprocessed to remove redundant symbols (such as extra punctuation and repeated spaces) and standardize unit representations. The word segmentation tool is combined with a professional dictionary in the field of geotechnical engineering for word segmentation, and the HanLP tool is used for part-of-speech tagging. The tagging results include part-of-speech tags such as noun (n), quantifier (m), adjective (a), preposition (p), and verb (v). Based on the tagging results, clue candidates are identified: segments with part-of-speech tags of nouns or noun phrases (such as "n+n" and "n+ns") are selected as entity clue candidates, such as "silty clay" and "drilling ZK3"; quantitative phrases ("m+unit", such as "28%" and "5×10- ³cm / s") and adjective phrases ("a" or "a+a", such as "loose" and "hard plastic") are selected as attribute clue candidates; prepositional phrases ("p+noun phrase", such as "located under the plain fill") and verb phrases ("v+noun phrase", such as "contacting the pebble layer") are selected as relation clue candidates. A pre-trained BERT model is selected and fine-tuned using 5,000 geotechnical log samples labeled with "entity clues", "attribute clues", and "relation clues": the candidate clue text is converted into a word vector and input into the model through The fully connected layer outputs classification probabilities, with the category with the highest probability being the clue type (e.g., "silty clay" is classified as an entity clue). The classification results are summarized using a confusion matrix: rows represent the actual categories (entity, attribute, relationship), columns represent the predicted categories, and the diagonal elements represent the number of correctly classified clues. The confidence score is calculated as "total number of correctly classified clues divided by (number of entity clues + number of attribute clues + number of relationship clues)." For example, if there are 150 total clues and 138 are correctly classified, the confidence score is 138 ÷ 150 = 0.92.

[0050] Construct a clue association network, calculate the association weights between information units based on clue co-occurrence frequency and semantic consistency, and form a weighted network topology. The specific implementation of constructing a clue association network, calculating the association weights between information units based on clue co-occurrence frequency, semantic consistency and domain logical association, and forming a weighted network topology is as follows: first determine the network nodes: the smallest information unit obtained by deconstructing information particles is the network node, and each node carries its clue type and confidence label; calculate the clue co-occurrence frequency, count the number of times any two information units co-occur in the same text paragraph, table row or adjacent sentence of the unstructured data of the geotechnical log file, and take the ratio of this number to the total number of paragraphs in the data set as the co-occurrence frequency; calculate semantic consistency, based on the pre-trained BERT model, extract the context vector corresponding to the text content of each information unit, and calculate the semantic consistency of the two information units through cosine similarity; perform weighted fusion on the clue co-occurrence frequency and semantic consistency to obtain the association weights between the information units; construct a clue association network with information units as nodes and association weights as edges, and establish a connection in the network if the association weight of the two information units is greater than or equal to the set weight threshold, and finally obtain a weighted network topology containing the association relationship. Specifically, the smallest information unit obtained by deconstructing information particles serves as a network node. This smallest information unit is an independent semantic fragment obtained through word segmentation, part-of-speech tagging, and clue classification. Each node must carry two types of labels: a clue type label, clearly marked as "entity clue," "attribute clue," or "relationship clue." For example, the node "plain fill" is labeled "entity clue," and the node "water content 28%" is labeled "attribute clue." The second label is a confidence value, which represents the confidence value of the classified information unit. Nodes are stored in a structured data format (such as JSON). For example, the following example is: {"node ID":"N1","text content":"silty clay","clue type":"entity clue","confidence":0.93}. By traversing all deconstruction results, a node set containing all information units is generated, ensuring that there are no duplicate or redundant nodes. The co-occurrence frequency and semantic consistency of the clues are then calculated. A weighted summation is used to combine the co-occurrence frequency and semantic consistency. The weight coefficients are set based on domain experience, with a weight of 0.6 for co-occurrence frequency and a weight of 0.4 for semantic consistency. The formula for calculating association weight is: Association weight = Co-occurrence frequency weight × Co-occurrence frequency + Semantic consistency weight × Semantic consistency. Finally, a weight threshold is set and the association weight matrix is ​​traversed. For any two nodes whose association weight is greater than or equal to the threshold, a connection is established between them, and the edge weight is the association weight. The network topology is stored in an adjacency table format. Each node entry contains the ID of the connected node and the corresponding weight. The resulting weighted network topology needs to be visually verified to ensure that strongly associated nodes are correctly connected and that weakly associated nodes have no redundant connections.

[0051] Then, an adaptive community discovery algorithm is used to cluster the weighted network topology to obtain a number of cluster communities. The specific implementation of clustering the weighted network topology to obtain a number of cluster communities by the adaptive community discovery algorithm is as follows: first, an initial seed node is generated based on the weight density. The weight density of the node is the sum of the association weights of the node and its adjacent nodes, and is calculated as follows: ,in, is the i-th information unit node in the network, Representation node The set of adjacent nodes of Representation node and adjacent nodes The association weight between Represented as a node Calculate the weight density of all nodes in the weighted network. and standard deviation , set dynamic threshold , the weight density is greater than or equal to The nodes are used as candidate seed nodes; the candidate seed nodes are deduplicated. If two candidate seed nodes The association weight between If the value is greater than or equal to the set strong correlation threshold, the nodes with higher weight density are retained as the initial clustering centers, and finally a set of seed nodes is formed. , where m is the number of initial cluster centers; computing nodes and cluster centers The weighted distance is calculated as ,in, is the center node of the kth cluster, For nodes and cluster centers The association weight between is the average value of the association weights between all nodes in the kth cluster. Based on the above weighted distance, the nodes are Assign to the cluster with the smallest distance; define the kth cluster as a community, and calculate the community cluster cohesion, which is calculated as follows: ,in, Expressed as the cohesion of the k-th community cluster, For nodes The association weight of is the sum of the total network weight, For nodes The sum of the association weights with all adjacent nodes, is the penalty coefficient, is the number of nodes in the kth community; if If the value is less than zero, the community will be split according to the maximum weight difference, and the difference in the associated weights between all nodes in the community will be calculated. The pair of nodes with the largest difference will be selected as the new center, and the original community will be split into two sub-communities. If the cohesion of the community cluster after the two adjacent communities are merged is When the percentage reaches the set threshold compared to before the merge, the merge operation is performed; the above split-merge process is repeated until the set number of iterations is reached and the iteration is stopped.

[0052] A dual-track reasoning model is constructed, comprising a clue aggregation layer and a knowledge completion layer. The clue aggregation layer performs feature fusion on information units within each cluster community to generate entity candidate sets and relationship candidate sets. The knowledge completion layer performs logical verification and missing information completion on the entity candidate sets and relationship candidate sets, outputting standardized entity and relationship results. The clue aggregation layer performs feature fusion on information units within each cluster community to generate entity candidate sets and relationship candidate sets. The implementation includes extracting core features from information units within each cluster community, including textual semantic features of entity clues, numerical and descriptive features of attribute clues, and semantic association features of relationship clues. The clue type label and confidence level of each information unit are also introduced as auxiliary features. Based on the association weights between information units within a cluster community, a weighted feature fusion strategy is employed to combine entity clues with closely associated attribute clues to form preliminary entity candidates, and to match relationship clues with associated entity clues to form relationship candidates. This generates entity candidate sets and relationship candidate sets corresponding to each cluster community. Specifically, for information units within a cluster community, core features are extracted by clue type and fused with auxiliary features. For entity clues, a pre-trained BERT model is used. After inputting the entity text, a 768-dimensional vector at the [CLS] position is extracted as a textual semantic feature, accurately capturing the entity's unique semantics within the domain. For attribute clues, numerical values ​​are first normalized to floating-point numbers to generate a 1-dimensional numerical feature. For descriptive attributes, a 768-dimensional semantic vector is extracted using the BERT model as a descriptive feature. For relationship clues, the context containing the entity pair is input, and the 768-dimensional vector output by BERT is extracted as a semantic association feature, reflecting the directionality of the relationship. Furthermore, a clue type label (one-hot encoding of entity [1,0,0], attribute [0,1,0], and relationship [0,0,1]) and confidence are introduced and concatenated with the core features to form a 772-dimensional fused feature vector, achieving comprehensive feature integration. A weighted strategy is used to generate candidate sets based on the association weights of information units within cluster communities. For entity clues, an association weight threshold of 0.4 was set. This was determined by calculating a weight distribution across a large number of known valid and unrelated information units. A threshold value was selected that balanced association recognition accuracy and recall. Attribute clues with weights ≥ 0.4 were selected (for example, in historical data, "particle size 20-100mm" had a weight of 0.53, and "slightly dense" had a weight of 0.47). Entity and attribute features were weighted and fused using the normalized weight value (0.53 / (0.53+0.47)=0.53) to form a preliminary entity candidate for "pebble layer (particle size 20-100mm, slightly dense)" consisting of an entity name and attribute key-value pair. For relationship clues, the association network was used to locate associated entity pairs (for example, "silty clay" and "pebble layer" with weights of 0.49 and 0.51). The cosine similarity between the relationship feature and the two entity features was calculated (a value ≥ 0.6 was considered a match), resulting in a relationship candidate for "(silty clay, contact with..., pebble layer)" (see Figure 1).Finally, duplicate candidates are removed to generate entity candidate sets and relationship candidate sets for each community.

[0053] The knowledge completion layer is used to perform logical verification and missing information completion on the entity candidate set and the relationship candidate set, and the implementation of outputting standardized entity results and relationship results includes: performing logical verification on the entity candidate set and the relationship candidate set, wherein the rationality of the entity attributes of the entity candidate set is checked, and the logic of the relationship between entities is verified for the relationship candidate set; for the missing information found after verification, reasoning and completion are performed based on high-confidence association clues in the clue association network and existing knowledge; and outputting standardized entity results and relationship results.

[0054] Specifically, during the logical verification phase, the entity candidate set's attributes are checked for plausibility from three perspectives. The first step is a value range check based on a reasonable data range to verify that attribute values ​​fall within a reasonable range. The second step is an attribute relevance check, verifying that the attributes of the same entity are logically consistent. For example, the "clay content" of a "sand" entity must be ≤3%. If a "clay content of 15%" is detected, it is considered a contradiction. When an entity is labeled "slightly dense" (density), its "porosity" must be between 0.7 and 0.9. Values ​​outside this range are considered attribute conflicts. Finally, a type matching check is performed to verify that the reported attribute matches the entity type. For example, "weathering degree" is a rock-specific attribute; if a "silt" entity carries this attribute, it is considered a mismatch. To verify the logicality of the relationships between the entity candidate sets, the stratigraphic chronology is verified based on geochronological principles to ensure that the entity relationships conform to the sedimentary pattern of older strata below and newer strata above. At the same time, spatial topological consistency is verified to ensure that the relationships between the same entity pairs are unique and conflict-free. For example, if entities A and B are both labeled as being above B and horizontally adjacent to each other, and there is no geological structural explanation, this is considered a topological conflict, and it is necessary to confirm whether it conforms to the physical association rules of the entity space. In addition, it is necessary to match the preset geotechnical domain relationship library, such as typical rules such as "aquifers must be adjacent to aquicludes" and "strongly weathered rock layers are often located above moderately weathered rock layers." If the relationship conflicts with the records in the library, it is marked as pending verification. All abnormal relationships must be further reviewed in light of the geological context to ultimately ensure that the relationships between entities conform to objective geological laws and logical rules. Then, for the missing information discovered after verification, high-confidence clues with association weights greater than or equal to a set threshold are first screened from the clue association network. The confidence is obtained by the ratio of the number of correctly classified clues to the total number of clues, which is the sum of the number of entity clues, attribute clues, and relationship clues. The existing knowledge mentioned above is historical case data and industry standard data. Combining high-confidence association clues and existing knowledge for reasoning completion, first, high-confidence clues are screened from the clue association network, and then structured historical case data and industry specification data are called based on the high-confidence clues. The industry specifications provide mandatory constraints to define the boundaries of the completion range; historical data forms a statistical law library, and the specific completion value of the completion result is determined through probability analysis.

[0055] A geotechnical knowledge graph is constructed based on the entity results and relationship results, and the knowledge graph is mapped into a structured data format. Specifically, standardized entity results are used as nodes, each of which contains the entity's unique ID, name, and attribute set (e.g., the "plain fill" node contains the ID "E1," the attributes "burial depth 0-3m," and "dry density 1.6g / cm³"); relationship results are used as directed edges, each of which contains the relationship type, source entity ID, and target entity ID (e.g., the "below" relationship corresponds to an edge with source "E2 (silty clay)" and target "E1 (plain fill)"). Nodes and edges are stored in a graph database to form a geotechnical knowledge graph containing entities, attributes, and relationships. When mapping to structured data, entity tables and relationship tables are generated: the entity table contains the "entity ID", "name", and "attribute JSON" fields (for example, the row corresponding to E1 is: "E1", "plain fill", "{"burial depth":"0-3m","dry density":"1.6g / cm³"}"); the relationship table contains the "relationship ID", "source entity ID", "target entity ID", and "relationship type" fields (for example, the row corresponding to R1 is: "R1", "E2", "E1", "below"), achieving accurate conversion of the graph into a structured format.

[0056] An initial geological model is constructed using structured data, and the core data in the initial geological model, including stratum geometry data, attribute data, and spatial relationship data, are extracted and converted into an intermediate data file that conforms to the extended IFC format. The extended IFC format adds geological-specific entities to adapt to the unique attributes of the strata. The specific implementation of the method of constructing an initial geological model using structured data, extracting the core data in the initial geological model, including stratum geometry data, attribute data, and spatial relationship data, and converting into an intermediate data file that conforms to the extended IFC format is to design an extended IFC format, add geological-specific entities, and the geological-specific entities include stratum unit entities, geological attribute set entities, and spatial topology entities; establish a mapping relationship between the core data and the extended IFC format, map the stratum geometry data to the geometric parameters of the stratum unit entity, map the attribute data to the characteristic parameters of the geological attribute set entity, and map the spatial relationship data to the associated parameters of the spatial topology entity; and based on the mapping relationship, convert the extracted core data into an intermediate data file that conforms to the extended IFC format. Specifically, when designing the extended IFC format, three types of special entities are added focusing on the characteristics of geological data. Stratigraphic unit entities represent the physical form of a stratum. They include basic identifiers such as the stratigraphic unit number, stratigraphic name, and formation age, as well as geometric descriptors such as layer boundaries, layer thickness, and dip angles, defined by a set of three-dimensional coordinate points. These entities precisely define the spatial extent of a single stratigraphic layer. Geological attribute set entities specifically carry geotechnical and physical properties, including quantitative attributes such as water content, porosity, compression modulus, and internal friction angle, as well as qualitative attributes such as soil properties. These entities are bound to stratigraphic unit entities through an attribute set association mechanism. Spatial topological entities characterize the spatial relationships between strata, including the contact type, stacking order, and spatial associations with specific geological structures. The direction and strength of these relationships are captured through the structure of topological nodes and edges. Mapping relationships require precise correspondence between core data and dedicated entity parameters. Within stratigraphic geometry data, the three-dimensional coordinates (X, Y, Z) of a layer are mapped to the "Boundary Coordinate Set" parameter of the stratigraphic unit entity, layer thickness is mapped to the "Thickness" parameter, and layer elevation is mapped to the "Elevation Datum" parameter, ensuring accurate digital restoration of the geometry. In attribute data, the physical properties of rock and soil (such as natural density of 1.8g / cm³) are mapped to the "Density" field of the geological attribute set entity, mechanical properties are mapped to the "Compression Modulus" field, and soil type names are mapped to the "Soil Property Type" field, forming a complete chain of attribute parameter correspondence. In spatial relationship data, the relationship between strata is mapped to the "Vertical Order" parameter of the spatial topology entity, the contact type (such as "Gradual Contact") is mapped to the "Contact Characteristics" parameter, and the spatial distance to the fault is mapped to the "Tectonic Association Distance" parameter, achieving a structured expression of spatial logic.The conversion process is performed in three steps: first, core data is extracted from the initial geological model, classified according to "stratum geometry-attribute-spatial relationship", the original data is standardized, and the coordinate data is converted to the WGS84 coordinate system; second, based on the preset mapping relationship, the classified data is filled one by one into the corresponding fields of the extended IFC format. For example, the "beam boundary coordinates" of a silty clay layer are filled into the "boundary coordinate set" of the stratigraphic unit entity, and the "water content of 25%" is filled into the "water content" field of the geological attribute set entity; finally, the data is organized according to the grammatical rules of the IFC format (such as entity instantiation and attribute assignment statements) to generate an intermediate data file containing three types of entities: stratigraphic unit, attribute set, and topological relationship. The file structure integrity is verified using a format verification tool to ensure that there are no missing fields or format errors.

[0057] The intermediate data file is associated with the BIM model and imported into the BIM platform using the IFC standard interface. A coordinate conversion algorithm is used to align the coordinate system of the geological model with the coordinate system of the BIM model, achieving spatial alignment. Data mapping rules are used to match geological entities with component attributes in the BIM model, generating data that can be used to generate a BIM model containing geological information. Specifically, associating the intermediate data file with the BIM model requires three steps. First, data is imported using the IFC standard interface: Using the IFC universal interface supported by BIM platforms (such as Revit and Bentley), the intermediate data file in the extended IFC format is directly imported. The interface automatically parses the structural information of the stratigraphic unit entities, geological attribute set entities, and spatial topological entities in the file, ensuring that the hierarchical relationships of the geological data (such as the binding of strata and attributes and the orientation of topological relationships) are fully preserved in the BIM environment, avoiding data loss or structural distortion. Second, spatial alignment is achieved using a coordinate conversion algorithm. Geological models typically use a geodetic coordinate system, while BIM models often use a construction coordinate system (such as a site-relative coordinate system). This requires a seven-parameter coordinate conversion method to unify the coordinate systems. First, three or more common points (such as borehole locations and site boundary corners) are selected, their coordinate values ​​in both coordinate systems are obtained, and the translation, rotation angle, and scaling factor are calculated. Based on these parameters, the 3D coordinates of the geological model are then batch-converted, ensuring that the spatial positions of the strata (such as burial depth and strike) precisely correspond to the building components in the BIM model in the same coordinate system, ensuring no offset when the two are spatially superimposed. Finally, data mapping rules are used to match the attributes of entities and components. Attribute mapping tables are developed between geological entities and BIM components. For example, geometric parameters such as "layer thickness" and "inclination" of the stratigraphic unit entity are mapped to the "geometric characteristics" attribute group of the BIM geological component; parameters such as "compression modulus" and "water content" of the geological attribute set entity are mapped to the "physical and mechanical properties" field of the BIM component; and relationships such as "overburden" and "contact" of the spatial topological entity are mapped to the "spatial association" attribute of the BIM. Using scripting tools (such as Dynamo), these mapping rules are implemented to embed geological information into the attribute structure of the BIM model, ultimately generating a BIM model containing complete geological data.

[0058] The above description is merely a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any person skilled in the art may utilize the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes for application in other fields. However, any simple modification, equivalent change, and modification of the above embodiments made in accordance with the technical essence of the present invention without departing from the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A method for structured conversion of geotechnical data, characterized in that: The following steps are involved: Obtaining an unstructured data set of geotechnical log files, performing information granular deconstruction on the data set, breaking the data into minimum information units containing entity clues, attribute clues, and relationship clues, and marking each information unit with a clue type and confidence level; Construct a clue association network, calculate the association weights between information units based on clue co-occurrence frequency and semantic consistency, and form a weighted network topology; Clustering the weighted network topology using an adaptive community discovery algorithm to obtain a number of cluster communities; Construct a dual-track reasoning model, which includes a clue aggregation layer and a knowledge completion layer: The clue aggregation layer performs feature fusion on the information units in each cluster community to generate entity candidate sets and relationship candidate sets; The knowledge completion layer is used to perform logical verification and missing information completion on the entity candidate set and the relationship candidate set, and output standardized entity results and relationship results; Constructing a geotechnical knowledge graph based on the entity results and the relationship results, and mapping the knowledge graph into a structured data format; Constructing an initial geological model using structured data, extracting core data from the initial geological model, including stratum geometry data, attribute data, and spatial relationship data, and converting the data into an intermediate data file that conforms to an extended IFC format. The extended IFC format adds geological-specific entities to accommodate stratum-specific attributes. The intermediate data file is associated with the BIM model, and the geological-specific entities are matched with the component attributes in the BIM model through data mapping rules to obtain data that can generate a BIM model containing geological information.

2. A method for structural conversion of geotechnical data according to claim 1, characterized in that: Deconstruct the data set into information granules, breaking the data into the smallest information units containing entity clues, attribute clues, and relationship clues, including: Natural language processing technology is used to perform word segmentation and part-of-speech tagging on the unstructured data of geotechnical log files, identifying noun phrases in the text as entity clue candidates, quantity phrases and adjective phrases as attribute clue candidates, and prepositional phrases and verb phrases as relationship clue candidates; Classify candidate clues based on the pre-trained BERT model in the geotechnical engineering field and determine the clue type; The confidence of clue classification is calculated using a confusion matrix, which is obtained by the ratio of the number of correctly classified clues to the total number of clues, where the total number of clues is the sum of the number of entity clues, attribute clues, and relationship clues.

3. The method for structural conversion of geotechnical data according to claim 1, characterized in that: The specific implementation of constructing the clue association network and calculating the association weights between information units based on clue co-occurrence frequency, semantic consistency and domain logic association to form a weighted network topology is as follows: First, determine the network nodes: the smallest information unit obtained by deconstructing information particles is the network node, and each node carries its clue type and confidence label; Calculate the co-occurrence frequency of clues, count the number of times any two information units co-occur in the same text paragraph, table row or adjacent sentences of the unstructured data of the geotechnical log file, and take the ratio of this number to the total number of paragraphs in the data set as the co-occurrence frequency; Calculate semantic consistency. Based on the pre-trained BERT model, extract the context vector corresponding to the text content of each information unit, and calculate the semantic consistency of two information units through cosine similarity; The co-occurrence frequency of clues and semantic consistency are weighted and fused to obtain the association weights between information units; A clue association network is constructed with information units as nodes and association weights as edges. If the association weight of two information units is greater than or equal to the set weight threshold, a connection is established in the network, and finally a weighted network topology containing association relationships is obtained.

4. The method for structural conversion of geotechnical data according to claim 1, characterized in that: The specific implementation of clustering the weighted network topology using the adaptive community discovery algorithm to obtain several cluster communities is as follows: First, the initial seed node is generated based on the weight density. The weight density of the node is the sum of the association weights of the node and its adjacent nodes, and is calculated as follows: ,in, is the i-th information unit node in the network, Representation node The set of adjacent nodes of Representation node and adjacent nodes The association weight between Represented as a node The weight density of Calculate the mean weight density of all nodes in the weighted network and standard deviation , set dynamic threshold , the weight density is greater than or equal to The nodes are selected as candidate seed nodes; De-redundancy processing is performed on candidate seed nodes. If two candidate seed nodes The association weight between If the value is greater than or equal to the set strong correlation threshold, the nodes with higher weight density are retained as the initial clustering centers, and finally a set of seed nodes is formed. , where m is the number of initial cluster centers; Compute nodes and cluster centers The weighted distance is calculated as ,in, is the center node of the kth cluster, For nodes and cluster centers The association weight between is the average value of the association weights between all nodes in the kth cluster. Based on the above weighted distance, the nodes are Assign to the cluster with the smallest distance; The kth cluster is defined as a community, and the community cluster cohesion is calculated as follows: ,in, Expressed as the cohesion of the k-th community cluster, For nodes The association weight of is the sum of the total network weight, For nodes The sum of the association weights with all adjacent nodes, is the penalty coefficient, is the number of nodes in the kth community; like If the value is less than zero, the community will be split according to the maximum weight difference, and the difference in the associated weights between all nodes in the community will be calculated. The pair of nodes with the largest difference will be selected as the new center, and the original community will be split into two sub-communities. If the cohesion of the community cluster after the two adjacent communities are merged is When the percentage reaches the set threshold compared to before the merge, the merge operation is performed; the above split-merge process is repeated until the set number of iterations is reached and the iteration is stopped.

5. The method for structural conversion of geotechnical data according to claim 1, characterized in that: The clue aggregation layer performs feature fusion on the information units in each cluster community to generate entity candidate sets and relationship candidate sets, including: Extract core features from the information units within each cluster community. These core features include the textual semantic features of entity clues, the numerical and descriptive features of attribute clues, and the semantic association features of relationship clues. The clue type label and confidence level of each information unit are also introduced as auxiliary features. Based on the association weights between information units within the clustering community, a weighted feature fusion strategy is adopted to combine entity clues with closely associated attribute clues to form preliminary entity candidates, and match relationship clues with associated entity clues to form relationship candidates, generating entity candidate sets and relationship candidate sets corresponding to each clustering community.

6. The method for structural conversion of geotechnical data according to claim 1, characterized in that: The knowledge completion layer is used to perform logical verification and missing information completion on the entity candidate set and the relationship candidate set, and output standardized entity results and relationship results. The implementation includes: Perform logical verification on the entity candidate set and the relationship candidate set, checking the rationality of the entity attributes of the entity candidate set and verifying the logic of the relationship between entities for the relationship candidate set; For missing information found after verification, reasoning and completion are performed based on high-confidence related clues in the clue association network and existing knowledge; Output standardized entity results and relationship results.

7. The method for structural conversion of geotechnical data according to claim 1, characterized in that: The specific implementation of constructing an initial geological model using structured data, extracting core data from the initial geological model, including stratum geometry data, attribute data, and spatial relationship data, and converting the data into an intermediate data file conforming to the extended IFC format is as follows: Design and extend the IFC format, add geological entities, including stratigraphic unit entities, geological attribute set entities, and spatial topology entities; Establish a mapping relationship between core data and extended IFC format, map stratigraphic geometry data to geometric parameters of stratigraphic unit entities, map attribute data to characteristic parameters of geological attribute set entities, and map spatial relationship data to association parameters of spatial topology entities; According to the mapping relationship, the extracted core data is converted into an intermediate data file that complies with the extended IFC format.

8. The method for structural conversion of geotechnical data according to claim 1, characterized in that: Establishing an association between the intermediate data file and the BIM model includes: using the IFC standard interface to import the intermediate data file into the BIM platform, and using the coordinate conversion algorithm to align the coordinate system of the geological model with the coordinate system of the BIM model to achieve spatial position alignment.

Images