Geological disaster prediction method and device for underground engineering, computer device and readable storage medium

Abstract

This application relates to a method, apparatus, computer equipment, and computer-readable storage medium for predicting geological hazards in underground engineering projects. The method includes: acquiring engineering geological information of the target underground engineering project; performing feature extraction processing on the engineering geological information to obtain a geological feature vector of the target underground engineering project; obtaining similar historical cases corresponding to the target underground engineering project based on the similarity between the geological feature vector and the feature vectors of various historical cases; inferring the associated hazard paths corresponding to the target underground engineering project using a geological knowledge graph based on the associated geological entities corresponding to the engineering geological information and each similar historical case; and obtaining the target geological interpretation information and hazard prediction results corresponding to the target underground engineering project based on the associated hazard paths. This method can improve the accuracy of geological hazard prediction for underground engineering projects.

Description

Technical Field

[0001] This application relates to the field of underground engineering safety technology, and in particular to a method, apparatus, computer equipment, computer-readable storage medium and computer program product for predicting geological hazards in underground engineering. Background Technology

[0002] During the construction of underground projects, construction safety and project quality are easily affected by geological disasters. To ensure the smooth progress of underground projects, relevant technologies typically rely on geophysical exploration to obtain corresponding geological information and predict geological disasters accordingly. However, due to the complexity of underground media, the same geophysical data may correspond to multiple geological conditions, making it difficult to accurately predict the geological disaster risks faced by underground projects based on single geophysical data. Summary of the Invention

[0003] Therefore, it is necessary to provide a method, device, computer equipment, computer-readable storage medium, and computer program product for predicting geological hazards in underground engineering, in response to the above-mentioned technical problems.

[0004] Firstly, this application provides a method for predicting geological hazards in underground engineering projects, including:

[0005] Obtain engineering geological information of the target underground project;

[0006] The engineering geological information is processed by feature extraction to obtain the geological feature vector of the target underground project in one embodiment;

[0007] Based on the similarity between the geological feature vector and the feature vectors of each historical case, similar historical cases corresponding to the target underground project are obtained;

[0008] Based on the engineering geological information and the associated geological entities corresponding to each of the similar historical cases, the associated disaster paths corresponding to the target underground project are inferred using a geological knowledge graph.

[0009] Based on the associated disaster path, the target geological interpretation information and disaster prediction results corresponding to the target underground project are obtained.

[0010] In one embodiment, the step of inferring the associated disaster path corresponding to the target underground project using a geological knowledge graph based on the engineering geological information and the associated geological entities corresponding to each of the similar historical cases includes: obtaining candidate geological interpretation information corresponding to the similar historical cases based on the associated geological entities of the similar historical cases; inferring the candidate disaster path corresponding to the target underground project using a geological knowledge graph based on the engineering geological information and the candidate geological interpretation information, and calculating the inference confidence level corresponding to each candidate disaster path; and obtaining the associated disaster path corresponding to the target underground project based on the candidate disaster path with the highest inference confidence level.

[0011] In one embodiment, calculating the inference confidence level corresponding to each of the candidate disaster paths includes: obtaining the path confidence level of the candidate disaster path based on the knowledge confidence level of each geological knowledge triple in the candidate disaster path; and adjusting the path confidence level based on the engineering geological information to obtain the inference confidence level of the candidate disaster path.

[0012] In one embodiment, the geological knowledge graph is constructed according to the following steps: performing semantic analysis on geological knowledge data to identify geological entities and relationship types in the geological knowledge data; predicting the relationship type between each pair of geological entities based on the geological knowledge data to obtain the relationship prediction result between each pair of geological entities; when the relationship prediction result indicates that there is a relationship type between the pairs of geological entities with a relationship confidence greater than a confidence threshold, obtaining a geological knowledge triple based on the pair of geological entities and the relationship type; and storing each geological knowledge triple in a graph database to obtain the geological knowledge graph.

[0013] In one embodiment, the method further includes: obtaining historical case feature vectors for each historical case based on the case geological information of each historical case, and obtaining geological knowledge associated with each historical case; obtaining associated geological entities corresponding to the historical cases in the geological knowledge map based on the geological knowledge associated with the historical cases; and constructing a bidirectional index between the historical case feature vectors corresponding to the historical cases and the associated geological entities.

[0014] In one embodiment, the method further includes: obtaining the knowledge confidence of the geological knowledge triples based on the data source information, review result information, distribution information in the geological knowledge data, and the relationship confidence; performing a consistency check on each of the geological knowledge triples in the geological knowledge map; when there are conflicting geological knowledge triples, retaining the geological knowledge triple with higher knowledge confidence, and verifying the geological knowledge triple with lower knowledge confidence; when the verification passes, retaining the geological knowledge triple with lower knowledge confidence; when the verification fails, discarding the geological knowledge triple with lower knowledge confidence.

[0015] In one embodiment, the method further includes: after the target underground project is excavated, obtaining geological verification information of the target underground project; verifying the associated disaster path based on the geological verification information to obtain the prediction verification result of the associated disaster path; and updating the knowledge confidence of each geological knowledge triple in the associated disaster path based on the prediction verification result.

[0016] Secondly, this application also provides a geological hazard prediction device for underground engineering, comprising:

[0017] The information acquisition module is used to acquire engineering geological information of the target underground project;

[0018] The feature extraction module is used to perform feature extraction processing on the engineering geological information to obtain the geological feature vector of the target underground project;

[0019] The case retrieval module is used to obtain similar historical cases corresponding to the target underground project based on the similarity between the geological feature vector and the feature vectors of each historical case.

[0020] The path reasoning module is used to reason about the associated disaster paths corresponding to the target underground project based on the engineering geological information and the associated geological entities corresponding to each of the similar historical cases, using a geological knowledge graph.

[0021] The result acquisition module is used to obtain the target geological interpretation information and disaster prediction results corresponding to the target underground project based on the associated disaster path.

[0022] 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 perform the following steps:

[0023] Obtain engineering geological information of the target underground project;

[0024] The engineering geological information is processed by feature extraction to obtain the geological feature vector of the target underground project in one embodiment;

[0025] Based on the similarity between the geological feature vector and the feature vectors of each historical case, similar historical cases corresponding to the target underground project are obtained;

[0026] Based on the engineering geological information and the associated geological entities corresponding to each of the similar historical cases, the associated disaster paths corresponding to the target underground project are inferred using a geological knowledge graph.

[0027] Based on the associated disaster path, the target geological interpretation information and disaster prediction results corresponding to the target underground project are obtained.

[0028] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0029] Obtain engineering geological information of the target underground project;

[0030] The engineering geological information is processed by feature extraction to obtain the geological feature vector of the target underground project in one embodiment;

[0031] Based on the similarity between the geological feature vector and the feature vectors of each historical case, similar historical cases corresponding to the target underground project are obtained;

[0032] Based on the engineering geological information and the associated geological entities corresponding to each of the similar historical cases, the associated disaster paths corresponding to the target underground project are inferred using a geological knowledge graph.

[0033] Based on the associated disaster path, the target geological interpretation information and disaster prediction results corresponding to the target underground project are obtained.

[0034] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0035] Obtain engineering geological information of the target underground project;

[0036] The engineering geological information is processed by feature extraction to obtain the geological feature vector of the target underground project in one embodiment;

[0037] Based on the similarity between the geological feature vector and the feature vectors of each historical case, similar historical cases corresponding to the target underground project are obtained;

[0038] Based on the engineering geological information and the associated geological entities corresponding to each of the similar historical cases, the associated disaster paths corresponding to the target underground project are inferred using a geological knowledge graph.

[0039] Based on the associated disaster path, the target geological interpretation information and disaster prediction results corresponding to the target underground project are obtained.

[0040] The aforementioned method, apparatus, computer equipment, computer-readable storage medium, and computer program product for predicting geological hazards in underground engineering projects first acquire the engineering geological information of the target underground engineering project. Then, feature extraction processing is performed on the engineering geological information to obtain the geological feature vector of the target underground engineering project. Next, based on the similarity between the geological feature vector and the feature vectors of various historical cases, similar historical cases corresponding to the target underground engineering project are obtained. Then, based on the engineering geological information and the associated geological entities corresponding to each similar historical case, and using a geological knowledge graph, the associated hazard paths corresponding to the target underground engineering project are inferred. Finally, based on the associated hazard paths, the target geological interpretation information and hazard prediction results corresponding to the target underground engineering project are obtained. This scheme, by first acquiring the engineering geological information of the target underground engineering project and then extracting features, then searching for similar historical cases corresponding to the target underground engineering project based on the extracted feature vectors, and then inferring associated hazard paths based on the associated geological entities of similar historical cases and a geological knowledge graph, can fully combine historical data and geological expertise to predict geological hazards in target underground engineering projects. It provides a complete reasoning path from geological interpretation to geological conclusions, effectively reducing the interference of multiple interpretations of engineering geological information and achieving accurate prediction of geological hazard risks for target underground engineering projects. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a flowchart illustrating a geological hazard prediction method for underground engineering in one embodiment;

[0043] Figure 2 This is a schematic diagram of the process of obtaining a geological knowledge map in one embodiment;

[0044] Figure 3 This is a flowchart illustrating a geological hazard prediction method for underground engineering in another embodiment;

[0045] Figure 4 This is a schematic diagram of a dual-database architecture in one embodiment;

[0046] Figure 5 This is a schematic diagram of a causal reasoning network in one embodiment;

[0047] Figure 6 This is a schematic diagram of a partial structural segment of a geological knowledge map in one embodiment.

[0048] Figure 7 This is a structural block diagram of a geological disaster prediction device for underground engineering in one embodiment;

[0049] Figure 8 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0050] 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.

[0051] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various objects, but these objects are not limited by these terms. These terms are only used to distinguish the first object from the second object. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0052] In one embodiment, such as Figure 1 As shown, a method for predicting geological hazards in underground engineering is provided. This embodiment illustrates the method by applying it to a server. It is understood that this method can also be applied to a terminal, or to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0053] Specifically, the geological hazard prediction method for underground engineering in this embodiment can predict the geological hazard risk of the unexcavated portion of the target underground engineering project based on the engineering geological information collected before or during excavation. The target underground engineering project may include, but is not limited to, tunnel engineering, underground space development engineering, and geotechnical engineering. The engineering geological information of the target underground engineering project may include, but is not limited to, geological background information, geophysical data, on-site images, and remote sensing interpretation data. For example, taking a tunnel engineering project as an example, its engineering geological information may include, but is not limited to, Tunnel Seismic Prediction (TSP) wave velocity data, tunnel face photographs, remote sensing interpretation data, mileage markers, depth, data acquisition time, and geological background information. For example, the geological background information of a tunnel engineering project may be represented as "sandstone slate strata + well-developed bedding + groundwater infiltration".

[0054] Step S102: Perform feature extraction processing on the engineering geological information to obtain the geological feature vector of the target underground project.

[0055] In this process, one or more data points in the engineering geological information can be processed to extract features in order to obtain the geological feature vector of the target underground engineering project.

[0056] For example, taking TSP waveform data, frequency domain features, amplitude features, and phase features can be obtained from the TSP waveform data encoding, and then these features can be concatenated to obtain the corresponding feature vector. Alternatively, a CNN encoder trained on geological data can be used to convert the TSP waveform data into a feature vector of a specific dimension. Furthermore, the feature vectors obtained through the two methods described above can be concatenated to obtain the feature vector corresponding to the TSP waveform data.

[0057] For example, taking a photograph of the working face as an example, a trained geological feature recognition model can be used to perform image recognition on the photograph of the working face, extract features such as lithological features, geological structure features, and weathering degree features, and then these feature vectors are spliced ​​together to form the feature vector corresponding to the photograph of the working face.

[0058] Among them, the geological feature vector of the target underground project can be obtained by extracting feature vectors from one or more data in engineering geological information.

[0059] Step S103: Based on the similarity between the geological feature vector and the feature vectors of each historical case, obtain the similar historical cases corresponding to the target underground project.

[0060] The historical case feature vector can be obtained by extracting features from the geological information of historical cases, and each historical case feature vector can correspond to a specific historical case. Historical cases can be underground engineering projects carried out during historical periods.

[0061] For example, the feature vectors of multiple historical cases can be stored in a vector database. In this step, the feature vectors of each historical case in the vector database can be traversed, and a similarity matching algorithm (e.g., calculating cosine similarity) can be used to calculate the similarity between the geological feature vector and each historical case feature vector. Then, historical cases whose similarity meets the preset similarity conditions can be used as similar historical cases corresponding to the target underground project.

[0062] For example, historical cases can be sorted from highest to lowest similarity, and a preset number of historical cases at the top of the list can be considered as similar historical cases. Alternatively, historical cases with a similarity higher than a preset similarity threshold can also be considered as similar historical cases.

[0063] Step S104: Based on engineering geological information and the associated geological entities corresponding to each similar historical case, use the geological knowledge graph to infer the associated disaster paths corresponding to the target underground project.

[0064] In this step, after obtaining similar historical cases corresponding to the target underground project, we can further query the associated geological entities of similar historical cases in the geological knowledge graph, and combine the engineering geological information of the target underground project with the queried associated geological entities to deduce the associated disaster paths corresponding to the target underground project using the geological knowledge graph.

[0065] Specifically, a geological knowledge graph can be constructed based on geological knowledge data. It can include multiple geological knowledge triplets, each of which can include a head geological entity, a tail geological entity, and the relationship between the two. For example, the geological entities in a geological knowledge graph can include various entities such as adverse geological bodies, geological structures, geophysical anomalies, disaster types, disaster-causing mechanisms, risk levels, and mitigation measures. The relationships between geological entities can include various relationships such as causal relationships, accompanying relationships, influencing relationships, triggering relationships, aggravating relationships, and mitigating relationships. For example, the geological knowledge triplets in a geological knowledge graph can be represented as "bedding development leads to collapse," "weathering exacerbates surrounding rock damage," and "joints and fissures affect surrounding rock stability," etc.

[0066] For example, a geological knowledge graph can be stored in a graph database. Geological entities can be stored as nodes in the graph database, and the relationship between two geological entities can be stored as edges connecting the corresponding geological entity nodes. Based on the geological information corresponding to each historical case, the associated geological entities for each historical case can be determined, and a bidirectional mapping relationship can be established between the feature vector of the historical case and the nodes corresponding to the associated geological entities. For example, in a vector database, the node identifiers of the associated geological entities corresponding to the feature vector of each historical case can be recorded, while in the graph database, the vector identifiers of the feature vectors of the historical cases corresponding to each geological entity node can be recorded. For example, if a historical case is “a mountain tunnel DK235+680 sandy slate fracture zone, TSP wave velocity 2650m / s, confirmed as a moderately fractured zone after excavation, treated with I-beam support,” then its corresponding associated geological entities can include the unfavorable geological body node “sandy slate fracture zone”, the treatment measure node “I-beam support treatment,” and other geological entity nodes, thereby establishing a bidirectional index between the feature vector of the historical case and these geological entity nodes.

[0067] Among them, the index between the feature vector of historical cases and the corresponding associated geological entities can be used to query the associated geological entities in the geological knowledge graph based on similar historical cases corresponding to the target underground project. For example, assuming that similar historical cases corresponding to the target underground project include: Case 1 "a mountain tunnel DK235+680 sandy slate fracture zone, TSP wave velocity 2650m / s, after excavation confirmed as a moderately fractured zone, treated with I-beam support, similarity 0.92", Case 2 "a railway tunnel ZDK158+420 schist weathering zone, TSP wave velocity 2800m / s, after excavation confirmed as strongly weathered schist, treated with advanced small pipe grouting, similarity 0.87", Case 3 "a highway tunnel K89+150 sandstone jointed zone, TSP wave velocity 2900m / s, after excavation confirmed as a densely jointed zone, treated with densely arranged anchor bolts, similarity 0.84", then related geological entities including "sandy slate fracture zone", "schist weathering zone", and "sandstone jointed zone" can be obtained.

[0068] By combining the engineering geological information of the target underground project with the retrieved associated geological entities, the associated disaster paths corresponding to the target underground project can be derived using a geological knowledge graph. For example, a path search can be performed in the geological knowledge graph based on the associated geological entities corresponding to similar historical cases to obtain potential paths from the associated geological entities to geological entities such as disaster type, risk level, and mitigation measures. Each associated geological entity can correspond to one or more potential paths. Then, the confidence level of each path can be verified based on the engineering geological information, and the path with the highest confidence level can be taken as the associated disaster path corresponding to the target underground project. Alternatively, a hybrid solution engine integrating a local lightweight large language model and using logical symbol guidance can be used to obtain multiple potential paths based on the associated geological entities under logical symbol guidance, and then the large language model can be used to determine the associated disaster path corresponding to the target underground project with the highest confidence level. For example, the reasoning of associated disaster paths can also support multi-factor coupling analysis. For instance, based on the engineering geological information of the target underground project, the corresponding geological entity can be determined in the geological knowledge graph, and the path can be searched and reasoned by combining the geological entity and associated geological entities. For example, a composite reasoning path of "sandstone slate fracture zone + bedding development + groundwater → high collapse risk" can be obtained.

[0069] For example, based on the correlation between geological knowledge triplets in the geological knowledge graph, a multi-level causal reasoning network can be constructed. For example, it can include a first-level causal chain "geological element → geophysical anomaly", a second-level causal chain "geophysical anomaly → disaster risk", and a third-level causal chain "disaster risk → mitigation measures". Thus, the geological knowledge graph can be used to deduce a multi-level causal chain of "geological element → geophysical anomaly → disaster risk → mitigation measures".

[0070] Step S105: Obtain the target geological interpretation information and disaster prediction results corresponding to the target underground project based on the associated disaster path.

[0071] Specifically, based on the associated disaster paths of the target underground project, the target geological interpretation information corresponding to the target underground project can be obtained from the "geological elements" part of the path (such as geological entities such as adverse geological bodies and geological structures), and the disaster prediction results corresponding to the target underground project can be obtained from the "disaster risks" part of the associated disaster paths (such as geological entities such as disaster type, disaster-causing mechanism, and risk level).

[0072] Optionally, a detailed reasoning report of the target underground project can be generated based on the process of steps S101 to S105. The report may include information such as an overview of the input data, feature extraction results, similar case matching, knowledge graph reasoning path, confidence calculation process, risk assessment conclusions, and recommended disposal measures.

[0073] In the above-mentioned geological hazard prediction method for underground engineering, after obtaining the engineering geological information of the target underground engineering, feature extraction is first performed on it. Then, based on the extracted feature vectors, similar historical cases corresponding to the target underground engineering are searched. Subsequently, based on the associated geological entities of similar historical cases, and combined with the geological knowledge graph, the associated hazard path is inferred. This method can fully combine historical data and geological expertise to predict geological hazards of the target underground engineering, and provide a complete reasoning path from geological interpretation to geological conclusion. It effectively reduces the interference of multiple interpretations of engineering geological information and achieves accurate prediction of the geological hazard risk of the target underground engineering.

[0074] In an exemplary embodiment, based on engineering geological information and associated geological entities corresponding to similar historical cases, a geological knowledge graph is used to infer associated disaster paths corresponding to the target underground project. This includes: obtaining candidate geological interpretation information corresponding to similar historical cases based on associated geological entities; inferring candidate disaster paths corresponding to the target underground project based on engineering geological information and candidate geological interpretation information using a geological knowledge graph, and calculating the inference confidence level corresponding to each candidate disaster path; and obtaining the associated disaster path corresponding to the target underground project based on the candidate disaster path with the highest inference confidence level.

[0075] Specifically, by identifying geological knowledge entities such as adverse geological bodies and geological structures within the associated geological entities of similar historical cases, candidate geological interpretation information corresponding to these cases can be obtained. Simultaneously, based on the engineering geological information of the target underground project, corresponding geological knowledge entities can be queried in the geological knowledge graph. Subsequently, based on the engineering geological information and candidate geological interpretation information, paths containing the corresponding geological knowledge entities can be searched in the geological knowledge graph, thereby obtaining multiple candidate hazard paths corresponding to the target underground project. For example, multi-factor coupling analysis can be performed in path reasoning to obtain composite reasoning paths such as "sandstone slate fracture zone + bedding development + groundwater → high collapse risk".

[0076] For each candidate hazard path, the inference confidence level can be obtained by combining engineering geological information and candidate geological interpretation information. This inference confidence level is used to characterize the probability that the candidate hazard path is correct under the conditions of engineering geological information and candidate geological interpretation information. The higher the inference confidence level of a candidate hazard path, the more likely that the path is the correct geological hazard risk inference path for the target underground project. Therefore, the candidate hazard path with the highest inference confidence level can be regarded as the associated hazard path of the target underground project.

[0077] In this embodiment, by searching for corresponding candidate geological interpretation information in the geological knowledge graph based on similar historical cases, and then combining engineering geological information and candidate geological interpretation information to perform path reasoning in the geological knowledge graph, and calculating the confidence of path reasoning, it is possible to comprehensively consider historical data, geological expertise, and the actual situation of the target underground engineering to reason about and evaluate disaster paths, which helps to eliminate the problem of multiple solutions and obtain more accurate associated disaster paths.

[0078] In an exemplary embodiment, calculating the inference confidence level corresponding to each candidate disaster path may include: obtaining the path confidence level of the candidate disaster path based on the knowledge confidence level of each geological knowledge triple in the candidate disaster path; and adjusting the path confidence level based on engineering geological information to obtain the inference confidence level of the candidate disaster path.

[0079] The knowledge confidence score of a geological knowledge triple can be used to measure its credibility. For example, the knowledge confidence score of a geological knowledge triple can be quantitatively assessed based on one or more information sources, such as the data source of the triple, expert verification results, and the frequency of its appearance in geological knowledge materials. A higher knowledge confidence score indicates higher credibility of the geological knowledge triple. Specifically, for each candidate hazard path, the knowledge confidence scores of the geological knowledge triples within that path can be obtained first, then fused to obtain the path confidence score. Finally, engineering geological information and candidate geological interpretation information are used to adjust the path confidence score to obtain the inference confidence score of the candidate hazard path.

[0080] Specifically, when calculating the inference confidence of candidate disaster paths, an improved Bayesian inference framework can be used, combined with the uncertainty quantification of geological evidence and the theory of multi-source information fusion, to construct a mathematical model based on geological evidence fusion:

[0081]

[0082] In the formula, For the i-th candidate disaster path The confidence level of the inference, In order to provide engineering geological information Under the condition of the i-th candidate disaster path The posterior probability, For the i-th candidate disaster path The prior probability, For the i-th candidate disaster path Path confidence, For regularization parameters, Representative engineering geological information The probability of, where For the first Several disaster candidate paths For path When true, observed The likelihood probability, For the j-th candidate disaster path The prior probability. These are the probability parameters of the edges obtained from the graph database. Prior statistics derived from a historical case database.

[0083] In this embodiment, by comprehensively considering the actual situation of the target underground project and the reliability of the candidate disaster path itself, the inference confidence of the candidate disaster path is calculated based on the improved Bayesian framework. This enables a quantitative assessment of the probability of the candidate disaster path occurring in the target underground project, resulting in a more accurate inference confidence, which is beneficial to improving the accuracy of geological disaster prediction.

[0084] In one exemplary embodiment, such as Figure 2 As shown, a geological knowledge map can be constructed according to the following steps:

[0085] Step S201: Perform semantic analysis on the geological knowledge data to identify the geological entities and relationship types in the geological knowledge data.

[0086] Geological knowledge data can include, but is not limited to, industry standards, literature, engineering reports, and online documents in the geological field. For example, a deep learning model pre-trained on geological corpora (e.g., a deep learning model based on BERT-BiLSTM-CRF) can be used to perform semantic analysis on geological knowledge data to extract geological entities and relation types.

[0087] This can be achieved by establishing a geological entity classification system based on industry standards in the geological field, and defining the schema of each geological entity. For example, geological entities can be categorized into various types, such as adverse geological bodies, geological structures, geophysical anomalies, hazard types, hazard-causing mechanisms, risk levels, and mitigation measures. Some categories can be further subdivided into subcategories. For instance, the adverse geological body category can include subcategories such as faults, fracture zones, and weak interlayers. The fault subcategory can be further subdivided into normal faults, reverse faults, and slip faults; the fracture zone subcategory can be subdivided into fault fracture zones and densely jointed zones; and the weak interlayer subcategory can be subdivided into argillaceous interlayers, carbonaceous interlayers, and gypsum interlayers. Similarly, the geophysical anomaly category can include a TSP anomaly subcategory, which can be further subdivided into low-velocity anomalies, high-velocity anomalies, velocity abrupt changes, and reflection anomalies. For example, the disaster type category can include subcategories such as collapse, deformation, and water inrush. The collapse subcategory can be further subdivided into face collapse, arch waist collapse, and arch crown collapse. The deformation subcategory can be further subdivided into arch crown settlement, sidewall convergence, and invert arch bottom bulge. The water inrush subcategory can be further subdivided into fault water, karst cave water, underground river water, rock stratum contact zone water, and sandstone and conglomerate stratum water.

[0088] The extracted relationship types can include various relationships such as causal relationship, accompanying relationship, influence relationship, triggering relationship, aggravating relationship, and mitigating relationship.

[0089] Step S202: Based on geological knowledge, predict the relationship type between each pair of geological entities to obtain the prediction results of the relationship between each pair of geological entities.

[0090] Based on the extracted geological entities and relationship types, a double loop can be used to traverse all entity pairs, determining the relationship type between each pair of geological entities in two directions to obtain the predicted relationship results between each pair of geological entities. For example, for each pair of geological entities, an input sequence containing the original text and two geological entities can be constructed. A relationship classification model is then used to predict the relationship type between the two geological entities to obtain the predicted relationship result. This relationship classification model can be a neural network model pre-trained on a geological domain corpus.

[0091] For example, assuming a geological entity pair includes a head geological entity A and a tail geological entity B, the corresponding relationship prediction result can include the relationship confidence score for various relationship types from A to B. This relationship confidence score can be used to characterize the probability that the entity relationship between the geological entity pair is of that relationship type. It is understood that for another geological entity pair formed by the head geological entity B and the tail geological entity A, the relationship type can also be predicted, and the relationship prediction result for that geological entity pair can be obtained.

[0092] Step S203: When the relationship prediction result indicates that there is a relationship type between geological entity pairs with a relationship confidence level greater than the confidence level threshold, a geological knowledge triplet is obtained based on the geological entity pairs and the relationship type.

[0093] In this step, based on the predicted relationships between pairs of geological entities, the relationship confidence levels for various relationship types corresponding to the entity relationships between the two geological entities in the pair can be obtained. The relationship confidence levels for various relationship types between geological entity pairs can be compared with a preset confidence threshold (e.g., 0.8). When a relationship type exists with a confidence level greater than the threshold, geological knowledge triplets in the geological knowledge graph can be obtained based on the geological entity pair and the relationship type. It is understood that if no relationship type exists between geological entity pairs with a confidence level greater than the threshold, the geological knowledge graph may not contain geological knowledge triplets corresponding to that geological entity pair. It is also understood that if multiple relationship types exist between geological entity pairs with confidence levels greater than the threshold, multiple geological knowledge triplets corresponding to that geological entity pair can be constructed based on each relationship type that meets the condition.

[0094] Step S204: Store the geological knowledge triples from various regions into the graph database to obtain the geological knowledge graph.

[0095] Specifically, geological knowledge triples from various locations can be stored in a graph database to obtain a geological knowledge graph. For example, geological entities can be stored as nodes in the graph database, and the relationship between two geological entities can be stored as edges connecting the corresponding geological entity nodes.

[0096] In the graph database, nodes and edges can have multiple fields. For example, for nodes representing adverse geological bodies, attribute fields such as number, name, type, orientation, and scale can be set. For instance, for a node representing a sandy slate fracture zone, attribute fields such as number FZ001, name Sandy Slate Fracture Zone, type Fracture Zone, orientation NE 32 degrees, angle with tunnel SE 65 degrees, and scale Medium scale can be set. Similarly, for nodes representing disaster types, attribute fields such as number, name, type, risk level, and consequences can be set. For instance, for a node representing a surrounding rock collapse, attribute fields such as number H001, name Surrounding Rock Collapse, type Collapse, risk level High risk, and consequences Surrounding rock instability affecting construction safety can be set. Furthermore, for edges (i.e., relationships between geological entities) in the graph database, parameter fields such as probability and condition can be set. For example, when nodes are connected through a trigger relationship, parameter fields such as trigger probability 0.78 and trigger condition softening upon contact with water can be set for this relationship. The graph database can establish unique number constraints for each geological entity node, and can also create indexes on one or more fields of the geological entity node. For example, it can create indexes on the type field of adverse geological entity nodes and the risk level field of disaster type nodes.

[0097] In this embodiment, by extracting geological entities and relationship types based on geological knowledge data, predicting the relationship types between pairs of geological entities, and then adding geological knowledge triples with relationship confidence levels greater than a confidence threshold to the geological knowledge graph, the accuracy of the geological knowledge graph can be ensured. This is beneficial for providing an accurate domain knowledge foundation for predicting geological hazards in underground engineering. Furthermore, the graph database also enables efficient storage and management of the geological knowledge graph.

[0098] In an exemplary embodiment, the method may further include: obtaining historical case feature vectors for each historical case based on the case geological information of each historical case, and obtaining geological knowledge associated with each historical case; obtaining associated geological entities in the geological knowledge graph corresponding to the historical cases based on the associated geological knowledge of the historical cases; and constructing a bidirectional index between the historical case feature vectors corresponding to the historical cases and the associated geological entities.

[0099] Historical cases can be underground engineering projects conducted during historical periods. The geological information of these historical cases can include engineering geological information collected before or during the excavation of the underground project, as well as verification information obtained after the excavation. The same feature extraction method used for engineering geological information can be applied to the geological information of the historical cases to obtain a historical case feature vector for each historical case. For example, the historical case feature vectors of each historical case can be stored in a vector database.

[0100] Furthermore, based on the geological information of historical cases, associated geological knowledge can be obtained. For example, for the historical case "A mountain tunnel at DK235+680, sandy slate fracture zone, TSP wave velocity 2650m / s, confirmed as a moderately fractured zone after excavation, treated with I-beam support, similarity 0.92", associated geological knowledge can be obtained, including that the case corresponds to the geological body "sandy slate fracture zone" and the treatment measure "I-beam support treatment". Subsequently, the geological knowledge map can be queried based on the associated geological knowledge of historical cases to obtain the corresponding geological entities in the geological knowledge map for each associated geological knowledge of the historical case. These geological entities can then be used as associated geological entities corresponding to the historical case. Optionally, geological entities corresponding to historical cases can be created in the geological knowledge map, and relationships can be established between these entities and the entities of associated geological knowledge.

[0101] Based on the correspondence between historical cases and their associated geological entities (including entities corresponding to associated geological knowledge and entities corresponding to historical cases), a bidirectional index can be constructed between historical case feature vectors and associated geological entities. For example, the vector database can store the feature vectors of each historical case in association with the node identifiers of their associated geological entities in the graph database. This allows retrieval of the corresponding associated geological entities and historical cases from the graph database using the index relationship based on the historical case feature vectors. For example, the graph database can record the vector identifiers of the historical case feature vectors associated with each geological entity node in the vector database. This allows retrieval of the historical case feature vectors associated with a specific geological entity from the vector database using the index relationship.

[0102] In this embodiment, by extracting the corresponding historical case feature vectors and obtaining the associated geological knowledge based on the case geological information of historical cases, and then retrieving the geological entities associated with the geological knowledge in the geological knowledge graph, and establishing a bidirectional index between the geological entities and the historical case feature vectors, cross-source queries between historical cases and geological professional knowledge can be realized. This is beneficial for efficiently querying associated geological entities based on similar historical cases of the target underground project when predicting geological disasters, thereby improving the efficiency of geological disaster risk prediction.

[0103] In an exemplary embodiment, the method may further include: obtaining the knowledge confidence of the geological knowledge triples based on the data source information, review result information, distribution information in the geological knowledge data, and relation confidence; performing consistency checks on the geological knowledge triples in the geological knowledge map; retaining the geological knowledge triples with higher knowledge confidence when there are conflicting geological knowledge triples, and verifying the geological knowledge triples with lower knowledge confidence; retaining the geological knowledge triples with lower knowledge confidence when the verification passes; and discarding the geological knowledge triples with lower knowledge confidence when the verification fails.

[0104] Specifically, for each geological knowledge triple in the geological knowledge graph, the corresponding knowledge confidence can be calculated, and consistency checks and conflict resolution can be performed on each geological knowledge triple.

[0105] Specifically, the knowledge confidence of a geological knowledge triple can be used to measure the credibility of the geological knowledge triple. This knowledge confidence can be obtained based on the data source information, review results information, distribution information in geological knowledge data, and the relationship confidence of the geological knowledge triple.

[0106] For example, the source information of a geological knowledge triple can be used to indicate the type of geological knowledge data from which the triple originates. Based on the source information of the geological knowledge triple, the reliability of the data source of the geological knowledge triple can be obtained. For example, when the geological knowledge triple comes from authoritative literature, It can be set to 0.9 when the geological knowledge triple comes from the engineering report. It can be set to 0.8 when the geological knowledge triples come from online documents. It can be set to 0.6.

[0107] For example, the audit results of the geological knowledge triplet can be used to indicate whether the triplet has been verified and confirmed by experts. Based on the audit results of the geological knowledge triplet, the audit credibility of the geological knowledge triplet can be obtained. For example, when the audit result information indicates that the geological knowledge ternary set is a relationship confirmed by experts, It can be set to 0.2 when the audit result information indicates that the geological knowledge ternary is a relationship that has not been confirmed by experts. It can be set to 0.

[0108] For example, the distribution information of geological knowledge triples in geological knowledge data can be used to indicate the frequency or occurrence of the triple in the geological knowledge data. Based on the distribution information of geological knowledge triples, the statistical support confidence of the geological knowledge triples can be obtained. For example, when the distribution information indicates that the frequency of the geological knowledge triplet in the geological knowledge data is not less than a preset number (e.g., 5 times), It can be set to 0.1, otherwise It can be set to 0.

[0109] For example, the knowledge confidence of a geological knowledge triple can be calculated using the following formula:

[0110]

[0111] In the formula, The knowledge confidence level of the geological knowledge triplet. To assess the reliability of the data sources for the geological knowledge triplet, To assess the credibility of the geological knowledge ternary set, The statistical reliability of geological knowledge triples, The confidence level of the relationship between the geological knowledge triples.

[0112] Specifically, for each geological knowledge triple in the geological knowledge graph, a consistency check mechanism based on ontology reasoning can be used to check the consistency of each geological knowledge triple to detect logical conflicts in the knowledge graph. When a conflicting geological knowledge triple is detected, such as a circular causal relationship where "A causes B" and "B causes A" coexist, the conflicting geological knowledge triple can be marked as logically conflicted and conflict resolution processing can be performed. During the conflict resolution process, the knowledge confidence levels of the two conflicting geological knowledge triples can be compared. The geological knowledge triple with the higher knowledge confidence level is retained, while the geological knowledge triple with the lower knowledge confidence level is marked as pending verification.

[0113] For geological knowledge triplets in a state to be verified, they can be further verified based on the knowledge of geological experts to obtain corresponding verification results. For example, the verification can be performed based on a pre-built geological expert knowledge base, or the geological knowledge triplets in a state to be verified can be sent to a terminal used by geological experts, and the verification results input by the experts can be obtained. When the verification result indicates that the geological knowledge triplet in a state to be verified has passed verification, the geological knowledge triplet can be retained in the geological knowledge graph; when the verification result indicates that the geological knowledge triplet in a state to be verified has failed verification, the geological knowledge triplet can be discarded, that is, the triplet is not added to the geological knowledge graph.

[0114] In this embodiment, the knowledge confidence of geological knowledge triples is calculated by comprehensively considering multiple dimensions such as data sources, review results, distribution, and relationship confidence. This allows for a more accurate quantification of the reliability of geological knowledge triples. Therefore, based on the knowledge confidence of geological knowledge triples, the path confidence of each reasoning path in the geological knowledge graph can be accurately quantified, and a data foundation can be provided for conflict resolution within geological knowledge triples.

[0115] In an exemplary embodiment, the method may further include: after the excavation of the target underground project, obtaining geological verification information of the target underground project; verifying the associated disaster path based on the geological verification information to obtain the prediction verification result of the associated disaster path; and updating the knowledge confidence of each geological knowledge triple in the associated disaster path based on the prediction verification result.

[0116] After the target underground project is excavated, its corresponding geological verification information can be obtained. For example, the geological verification information may include, but is not limited to, one or more of the following: the predicted number of the excavated portion of the target underground project, the actual geological conditions, the prediction accuracy assessment results, the analysis of deviation causes, and the summary of lessons learned. The prediction accuracy assessment results can be obtained based on the differences between the actual geological conditions and their corresponding target geological interpretation information and disaster prediction results. The analysis of deviation causes and the summary of lessons learned can be obtained based on the prediction accuracy assessment results and the reasoning process of related disaster paths.

[0117] For example, for a target underground project, the actual exposure after excavation is as follows: a sandy slate fracture zone was found, approximately 2.8 meters wide, with relatively fractured rock mass and groundwater seepage. The corresponding geological verification information could include fields such as prediction number, actual geological conditions, prediction accuracy assessment result, and verification time. The prediction number is the number used in the previous geological hazard prediction for this excavated portion of the target underground project; the actual geological conditions field can record the geological type of the target underground project as a sandy slate fracture zone, the size of the fracture zone (including width, height, and length), the degree of fracture as moderate, the hydrological condition as groundwater seepage, the hazard occurrence status as true, and the hazard type as small-scale surrounding rock collapse; the prediction accuracy assessment result field can record that the hazard type prediction accuracy is true, the scale prediction accuracy is basically accurate, and the risk prediction accuracy is accurately assessed; the verification time field can be the time when this geological verification information was generated.

[0118] Based on geological verification information, the associated disaster paths corresponding to the target underground project obtained in previous geological disaster predictions can be verified to obtain the prediction verification results of the associated disaster paths. For example, when verifying associated disaster paths, it can be determined whether the entity relationships of each geological knowledge triplet in the path correspond to the actual geological conditions of the target underground project. Based on the accuracy judgment results of each geological knowledge triplet, the prediction verification results of the associated disaster paths are obtained. Subsequently, the knowledge confidence of each geological knowledge triplet in the associated disaster paths can be updated based on these prediction verification results.

[0119] For example, the knowledge confidence of geological knowledge triples can be updated based on the triple's number, accuracy assessment result, and preset learning rate. Specifically, the current knowledge confidence of the geological knowledge triple can be obtained first, and then adjusted based on the accuracy assessment result and learning rate. If the prediction is accurate, the knowledge confidence is increased; if the prediction is inaccurate, the knowledge confidence is decreased. For example, when increasing knowledge confidence, the new knowledge confidence can be the current knowledge confidence plus the learning rate, with an upper limit of 1.0; when decreasing knowledge confidence, the new knowledge confidence can be the current knowledge confidence minus the learning rate, with a lower limit of 0.1.

[0120] In this embodiment, by establishing an update system based on construction feedback, the knowledge confidence of each geological knowledge triple in the geological knowledge graph can be updated according to the geological verification information after the excavation of the target underground project. This is conducive to the continuous learning and optimization of the geological knowledge graph and improves the accuracy of subsequent geological disaster prediction.

[0121] In one exemplary embodiment, such as Figure 3 As shown, a method for predicting geological hazards in underground engineering is provided, which may include the following steps:

[0122] Step S1: Construct a unified acquisition and standardized processing architecture for multi-source geological data.

[0123] Specifically, before conducting geological hazard prediction for underground engineering projects, a unified processing architecture covering four major data sources—geophysical data, image data, document data, and monitoring data—can be established.

[0124] This architecture may include a data acquisition interface that supports the automated access of various data sources, such as TSP wave velocity data, tunnel face photographs, and remote sensing interpretation data. For example, the data acquisition interface can be encapsulated in a standardized JSON format, containing key fields such as data type, mileage marker, timestamp, depth, data sampling rate, numerical value, and geological background. For example, different acquisition frequencies can be set for different geological data; for instance, TSP data can be acquired once every 100m, and tunnel face photographs can be taken every 3 meters of excavation.

[0125] For the collected geological data, a three-dimensional coordinate system of mileage-depth-time can be established. For example, taking tunnel engineering as an example, a three-dimensional coordinate system can be established with the tunnel centerline mileage as the X-axis, the stratum depth as the Y-axis, and the acquisition time as the Z-axis. All data is precisely located according to the format "D3K mileage + offset:depth:timestamp". Furthermore, a coordinate registration algorithm can be established to achieve mileage coordinate standardization, depth coordinate transformation, and timestamp standardization. For example, mileage coordinate standardization unifies the format of the input mileage data; depth coordinate transformation converts relative elevation to absolute elevation, achieving precise conversion by combining surface elevation data; and timestamp standardization converts all time data to the UTC standard time format. Finally, a coordinate system object containing standard mileage, absolute depth, and UTC timestamp is returned.

[0126] The unified processing architecture may also include a geological data quality control and preprocessing module. For example, this module can denoise TSP waveform data by using wavelet transform to filter out high-frequency noise; perform brightness equalization and contrast enhancement on tunnel face photographs; and perform geometric and radiometric correction on remote sensing images. Simultaneously, this module can also establish a data quality scoring mechanism. For instance, for TSP data, records with a signal-to-noise ratio below 20dB can be marked as low-quality data, and for tunnel face photographs, images with a blur score below 0.7 can be reacquired.

[0127] Furthermore, a spatiotemporal indexing system for geological data can be established. Specifically, a hybrid index structure, including spatial indexes based on R-trees and temporal indexes based on B+ trees, can be used to index geological data. This index structure can support complex spatiotemporal queries. For example, the spatial index can be divided into 15×15 meter grids, and the temporal index can be layered by day. The index structure can support queries with complex conditions such as "all TSP anomaly data collected in month Y of year X within the interval from D3K mileage 1+offset 1 to D3K mileage 2+offset 2".

[0128] This can also include establishing a data version management mechanism for geological data. For example, a distributed version control approach can be adopted, generating a version identifier and version information for each data update. For instance, when it is found that the TSP interpretation results need to be corrected, the original version can be retained and a new version created, ensuring the traceability of data changes. For example, the version information of geological data can include: version number, modification time, modifier, reason for modification, and scope of impact.

[0129] Step S2: Implement a dual-database architecture based on vector and graph databases.

[0130] Specifically, the dual-database architecture in this embodiment can be as follows: Figure 4 As shown, it can include a vector database and a graph database. Based on this dual-database architecture, it is possible to store and index historical cases and geological knowledge data, and provide a foundation for subsequent geological hazard prediction for target underground engineering projects.

[0131] For multi-source geological data, appropriate data preprocessing can be performed first according to the corresponding data type.

[0132] For example, for the geological information of historical cases, feature extraction processing can be performed on one or more data such as TSP waveform data and tunnel face photographs to obtain the historical case feature vectors corresponding to the historical cases, as well as the geological knowledge associated with each historical case. A geological data vectorization storage mechanism can be established to store the historical case feature vectors corresponding to the historical cases in a vector database.

[0133] For example, for geological knowledge data, a geological knowledge graph and entity relationship extraction system can be constructed, and this system can be used to extract textual knowledge from the geological knowledge data and construct a geological knowledge graph. First, semantic analysis can be performed on the geological knowledge data to identify geological entities and relationship types. Then, based on the geological knowledge data, the relationship type between each pair of geological entities can be predicted, obtaining the prediction results for the relationships between each pair of geological entities. When the relationship prediction result indicates that there is a relationship type between geological entity pairs with a confidence level greater than a confidence threshold, the geological knowledge triples to be added to the geological knowledge graph can be obtained based on the geological entity pairs and relationship types. Finally, each geological knowledge triple can be stored in a graph database to obtain the geological knowledge graph. For example, geological entities can be stored as nodes in the graph database, and the relationship between two geological entities can be stored as an edge connecting the corresponding two geological entity nodes. For example, based on the association relationships between geological knowledge triples in the geological knowledge graph, a system can be constructed as follows: Figure 5The diagram illustrates a multi-level causal reasoning network. For example, this network may include a first-level causal chain "geological elements → geophysical anomalies," a second-level causal chain "geological anomalies → disaster risk," and a third-level causal chain "disaster risk → mitigation measures." Thus, a multi-level causal chain of "geological elements → geophysical anomalies → disaster risk → mitigation measures" can be derived using a geological knowledge graph.

[0134] For example, after constructing a geological knowledge graph, the knowledge confidence of each geological knowledge triplet can be calculated based on its source information, review results, distribution information in geological knowledge data, and relationship confidence. Then, a consistency check can be performed on each geological knowledge triplet in the geological knowledge graph. When conflicting geological knowledge triplets exist, the triplet with the higher knowledge confidence is retained, and the triplet with the lower knowledge confidence is verified. If the verification passes, the triplet with the lower knowledge confidence can be retained; if the verification fails, the triplet with the lower knowledge confidence can be discarded.

[0135] After constructing the vector database and the graph database, a bidirectional indexing mechanism can be established between them. For example, based on the geological knowledge associated with historical cases, the corresponding geological entities in the geological knowledge graph can be obtained, and then a bidirectional index can be constructed between the historical case feature vector and the associated geological entities. The vector database can record the graph node identifier corresponding to each vector, and the graph database can record the vector identifier corresponding to each graph node. Furthermore, the historical case feature vector of each historical case can be queried using the corresponding spatiotemporal index.

[0136] Step S3: Implement the intelligent disambiguation inference engine for multiple solutions of geophysical data.

[0137] Among them, the geophysical data multi-solution intelligent disambiguation inference engine can be used to predict the geological hazard risks faced by a target underground project based on its engineering geological information. For example, when it is necessary to predict geological hazards for a target underground project, its engineering geological information can be obtained first, and feature extraction processing can be performed on the engineering geological information to obtain the geological feature vector of the target underground project. Then, based on the similarity between the geological feature vector and the feature vectors of various historical cases, similar historical cases corresponding to the target underground project can be obtained. Subsequently, based on the engineering geological information and the associated geological entities corresponding to each similar historical case, the associated hazard paths corresponding to the target underground project can be inferred using a geological knowledge graph. Finally, based on the associated hazard paths, the target geological interpretation information and hazard prediction results corresponding to the target underground project can be obtained.

[0138] Specifically, after extracting the geological feature vector of the target underground project, cosine similarity can be used to calculate the similarity between the geological feature vector and the feature vectors of each historical case in the vector database, and the historical cases with the highest similarity can be selected as similar historical cases of the target underground project.

[0139] For example, taking the low-velocity TSP anomaly at D3K670+074.1 of the target underground project as an example, similar historical cases retrieved by the system may include: 1. A mountain tunnel at DK235+680, a sandy slate fracture zone, with a TSP wave velocity of 2650m / s. After excavation, it was confirmed to be a moderately fractured zone and was treated with I-beam support, with a similarity of 0.92; 2. A railway tunnel at ZDK158+420, a schist weathering zone, with a TSP wave velocity of 2800m / s. After excavation, it was confirmed to be strongly weathered schist and was treated with advanced small-diameter pipe grouting, with a similarity of 0.87; 3. A highway tunnel at K89+150, a sandstone jointed zone, with a TSP wave velocity of 2900m / s. After excavation, it was confirmed to be a densely jointed zone and was treated with densely arranged anchor bolts, with a similarity of 0.84. The search results can include fields such as case number, similarity value, geological interpretation, location information, and verification results for each historical case, and store them in a similarity list. Finally, the results are sorted from high to low similarity and the top k most similar cases are returned.

[0140] Specifically, after retrieving similar historical cases, disambiguation reasoning based on a geological knowledge graph can be performed. This combines the geological interpretations of historical cases with the professional knowledge in the knowledge graph to provide the most probable geological explanation for the target underground project. Specifically, a bidirectional index between the vector database and the graph database can be used to query associated geological entities of similar historical cases, and candidate geological interpretation information corresponding to these cases can be obtained. Then, based on engineering geological information and candidate geological interpretation information, candidate hazard paths corresponding to the target underground project can be inferred using the geological knowledge graph. Next, the path confidence of the candidate hazard path can be obtained based on the knowledge confidence of each geological knowledge triple in the candidate hazard path. This path confidence is then adjusted based on engineering geological information to obtain the inference confidence of the candidate hazard path. Finally, the associated hazard path corresponding to the target underground project can be obtained based on the candidate hazard path with the highest inference confidence. When calculating the inference confidence of the candidate hazard path, an improved Bayesian inference framework can be used, combined with the uncertainty quantification of geological evidence and multi-source information fusion theory, to adjust the path confidence.

[0141] After obtaining the target geological interpretation information and disaster prediction results corresponding to the target underground project, a detailed reasoning report can be generated for the prediction results. This report may include multiple parts such as: an overview of the input data, feature extraction results, similar case matching, knowledge graph reasoning path, confidence calculation process, risk assessment conclusions, and recommended disposal measures.

[0142] For example, taking the aforementioned underground project D3K670+074.1 as an example, based on its engineering geological information and similar historical cases, some structural fragments obtained by searching in the geological knowledge graph can be as follows: Figure 6 As shown. The reasoning report for this underground project can include basic information such as the predicted location D3K670+074.1, the predicted distance of 50 meters, and the predicted time. The input data section records TSP wave velocity data of 2650-3200 m / s, showing obvious low-velocity anomalies. The geological background is sandy slate with well-developed bedding and groundwater infiltration. Feature analysis shows that the TSP characteristics match the characteristics of the fracture zone, the wave velocity reduction is 32% lower than that of the intact surrounding rock, which is consistent with the characteristics of the fracture zone, and the waveform features show significant attenuation of reflected wave energy and significant scattering phenomena. Similar historical case matching results include a mountain tunnel DK235+680 sandy slate fracture zone with a similarity of 0.92 (confirmed after excavation), a railway tunnel ZDK158+420 schist weathering zone with a similarity of 0.87 (confirmed after excavation), and a highway tunnel K89+150 sandstone jointed area with a similarity of 0.84 (confirmed after excavation). Geological knowledge graph analysis shows that the sandy slate strata, combined with the low-velocity TSP anomaly and well-developed stratification, indicate a probability of a fractured zone of 0.88. The fractured zone, combined with groundwater infiltration, indicates a collapse risk probability of 0.85. The overall conclusion is that the predicted sandy slate fractured zone has a confidence level of 87%, a high risk level, and the main risk is large deformation and collapse of the surrounding rock. Recommended measures include strengthening advanced geological exploration to determine the scale and degree of fracture, developing a specific construction plan using advanced small-diameter pipe grouting reinforcement, strengthening initial support and appropriately increasing steel frame density, strictly controlling the excavation advance, and promptly closing the ring during excavation.

[0143] Step S5: Establish a closed-loop system for continuous learning and verification of knowledge graphs based on construction feedback.

[0144] This involves establishing a construction verification data collection mechanism to obtain corresponding geological verification information after the excavation of the target underground project. For example, after excavation at point D3K670+074.1 of the target underground project, the actual exposed conditions were: a fractured zone of sandy slate was discovered, approximately 2.8 meters wide, with relatively fractured rock mass and groundwater seepage. Verification data can be automatically entered into the system, including: actual geological conditions, prediction accuracy assessment results, deviation cause analysis, and lessons learned.

[0145] This module can be used to compare and analyze the predicted results with the actual situation. It can be used to assess the accuracy of geological hazard predictions by combining the results of the target underground project and historical data, and automatically calculate the prediction accuracy rate. For example, based on the most recent 100 predictions and verification records, the accuracy rate for geological type prediction is 87.3%, the accuracy rate for scale prediction is 74.8%, and the accuracy rate for risk assessment is 83.6%.

[0146] Simultaneously, a dynamic weight adjustment algorithm for the knowledge graph can be used to automatically adjust the knowledge confidence of geological knowledge triples in the geological knowledge graph based on geological verification information. For example, taking the target underground project D3K670+074.1 as an example, its geological verification information shows that the correlation strength of "TSP low-velocity anomaly → sandy slate fracture zone" is verified, and the knowledge confidence of this geological knowledge triple increases from 0.82 to 0.85; at the same time, the relation weight of "sandy slate fracture zone → high collapse risk" remains unchanged from 0.85 because the actual risk is consistent with expectations. For example, the knowledge confidence of geological knowledge triples can also be updated according to a preset learning rate. For instance, the current knowledge confidence of the geological knowledge triple can be obtained first, and then adjusted according to the geological verification information and the learning rate; if the prediction is accurate, the knowledge confidence is increased; if the prediction is inaccurate, the knowledge confidence is decreased. For example, when increasing knowledge confidence, the new knowledge confidence can be the current knowledge confidence plus the learning rate, with an upper limit of 1.0; when decreasing knowledge confidence, the new knowledge confidence can be the current knowledge confidence minus the learning rate, with a lower limit of 0.1.

[0147] This can be achieved by establishing an incremental update mechanism for model parameters, using geological verification information to incrementally train the machine learning model applied to the system. Online learning algorithms can be employed to avoid the computational overhead of large-scale retraining. For example, incremental learning can be implemented through a learner model update method that receives new verification data as parameters. Taking updating the feature extraction model of geophysical data as an example, new feature vectors can first be extracted from the TSP data of geological verification information to obtain the actual geological type as the label. Then, incremental training is performed to update the model parameters. Next, the similarity matching weight parameters are updated based on the new verification data. Finally, model checkpoints are saved to ensure the persistence of model updates.

[0148] This can be further enhanced by constructing an expert feedback integration module. Specifically, expert review meetings can be organized periodically, inviting geological experts to evaluate the prediction results. The expert feedback integration module can provide standardized forms for inputting expert feedback data into the system. For example, expert feedback data for each geological hazard prediction result can include: the reasonableness of the prediction result (1-5 points), the clarity of the reasoning logic (1-5 points), the accuracy of the risk assessment (1-5 points), and improvement suggestions. After obtaining the expert feedback data, it can be integrated into the optimization process of the geological knowledge graph. When a reasoning path receives more than three consecutive negative expert evaluations, the system automatically marks the path as pending review and arranges for in-depth expert review. Optionally, expert feedback can be implemented through an integration method. This method receives feedback data as parameters, including prediction number, expert scores, expert comments, etc. First, the prediction quality score is updated based on the expert scores. Then, expert comments are analyzed to extract improvement suggestions, which are then transformed into specific system optimization tasks. Finally, these system optimization tasks are executed iteratively.

[0149] Optionally, based on the dual-database architecture in this embodiment, a web-based visual interface can also be used to provide intelligent conversational query functionality, supporting professional question answering in natural language and real-time risk assessment of multiple tunnel monitoring points. Specifically, a geological-specific prompt word project can be constructed, leveraging the knowledge representation capabilities of the DIKW hierarchical structure to improve the model's inference accuracy and response speed in the tunnel geology field. Furthermore, intelligent cross-source query technology can be employed, utilizing entity links and semantic alignment mechanisms to link field monitoring data with expert knowledge, establishing a multi-scale, multi-dimensional geological information association analysis mechanism. Through dynamic metadata management, semantic bridging between different data sources can be achieved, enabling rapid fusion and retrieval of geological monitoring data, historical cases, and expert experience.

[0150] The geological hazard prediction method for underground engineering provided in this embodiment has at least the following beneficial effects:

[0151] 1. Significantly solves the problem of multiple interpretations in geophysical data and reduces the false alarm rate. Addressing the technical challenge of a false alarm rate as high as 40-50% in complex geological environments, this embodiment utilizes an intelligent disambiguation inference engine for geophysical data, integrating historical case similarity matching and confidence ranking algorithms, combined with geological professional knowledge map inference. This effectively eliminates the interference of multiple interpretations in geophysical signals, reducing the false alarm rate to below 5%.

[0152] 2. By establishing a three-dimensional correlation mechanism of mileage-depth-time and intelligent cross-source query technology, a unified data format and coordinate system are established to achieve standardized processing and spatiotemporal registration of multi-source geological data.

[0153] 3. Through the incremental learning function of the continuous learning closed-loop system based on construction feedback and the checkpoint mechanism, the prediction accuracy and model parameters can be automatically adjusted according to the actual construction feedback, realizing the continuous learning closed loop of prediction-verification-learning-optimization, and continuously improving the intelligence level and prediction reliability of the system.

[0154] 4. Significantly improves system response speed, reducing system response time to less than 5 minutes, improving efficiency by more than 60% compared to traditional methods, and effectively meeting the real-time decision support needs in underground engineering construction such as tunnels.

[0155] 5. Achieve interpretable output of the geological prediction process, establish an interpretable output module for the reasoning process, provide a complete reasoning path and supporting explanations from geophysical data to geological conclusions, and provide reliable decision support for engineering and technical personnel.

[0156] 6. The establishment of a human-machine collaborative intelligent prediction system ensured the system's level of intelligence and made full use of the rich experience of geological experts, achieving an organic combination of technological innovation and engineering practice.

[0157] 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 in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0158] Based on the same inventive concept, this application also provides a geological hazard prediction device for underground engineering projects, which implements the geological hazard prediction method for underground engineering projects described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the geological hazard prediction device for underground engineering projects provided below can be found in the limitations of the geological hazard prediction method for underground engineering projects described above, and will not be repeated here.

[0159] In one exemplary embodiment, such as Figure 7 As shown, a geological hazard prediction device for underground engineering is provided, comprising:

[0160] The information acquisition module 701 is used to acquire the engineering geological information of the target underground project;

[0161] The feature extraction module 702 is used to perform feature extraction processing on the engineering geological information to obtain the geological feature vector of the target underground project;

[0162] The case retrieval module 703 is used to obtain similar historical cases corresponding to the target underground project based on the similarity between the geological feature vector and the feature vectors of each historical case.

[0163] The path reasoning module 704 is used to reason about the associated disaster path corresponding to the target underground project based on the engineering geological information and the associated geological entities corresponding to each of the similar historical cases, using a geological knowledge graph.

[0164] The result acquisition module 705 is used to obtain the target geological interpretation information and disaster prediction results corresponding to the target underground project based on the associated disaster path.

[0165] In an exemplary embodiment, the path reasoning module 704 is configured to: obtain candidate geological interpretation information corresponding to the similar historical cases based on the associated geological entities of the similar historical cases; use a geological knowledge graph to reason about candidate disaster paths corresponding to the target underground project based on the engineering geological information and the candidate geological interpretation information, and calculate the reasoning confidence of each candidate disaster path; and obtain the associated disaster path corresponding to the target underground project based on the candidate disaster path with the highest reasoning confidence.

[0166] In an exemplary embodiment, the path reasoning module 704 is configured to: obtain the path confidence of the candidate disaster path based on the knowledge confidence of the geological knowledge triples in the candidate disaster path; and adjust the path confidence based on the engineering geological information to obtain the reasoning confidence of the candidate disaster path.

[0167] In an exemplary embodiment, the geological knowledge graph is constructed according to the following steps: performing semantic analysis on geological knowledge data to identify geological entities and relationship types in the geological knowledge data; predicting the relationship type between each pair of geological entities based on the geological knowledge data to obtain the relationship prediction result between each pair of geological entities; when the relationship prediction result indicates that there is a relationship type between the pairs of geological entities with a relationship confidence greater than a confidence threshold, obtaining a geological knowledge triple based on the pair of geological entities and the relationship type; and storing each geological knowledge triple in a graph database to obtain the geological knowledge graph.

[0168] In an exemplary embodiment, the apparatus further includes: a case processing module, configured to obtain historical case feature vectors for each historical case and geological knowledge associated with each historical case based on the case geological information of each historical case; an entity association module, configured to obtain associated geological entities corresponding to the historical cases in the geological knowledge graph based on the geological knowledge associated with the historical cases; and an index construction module, configured to construct a bidirectional index between the historical case feature vectors corresponding to the historical cases and the associated geological entities.

[0169] In an exemplary embodiment, the apparatus further includes: a confidence level acquisition module, configured to obtain the knowledge confidence level of the geological knowledge triplet based on the data source information, review result information, distribution information in the geological knowledge data, and the relationship confidence level of the geological knowledge triplet; a conflict resolution module, configured to perform consistency checks on each of the geological knowledge triplets in the geological knowledge graph, retain the geological knowledge triplet with higher knowledge confidence level when there are conflicting geological knowledge triplets, and verify the geological knowledge triplet with lower knowledge confidence level; and a verification processing module, configured to retain the geological knowledge triplet with lower knowledge confidence level when the verification passes, and discard the geological knowledge triplet with lower knowledge confidence level when the verification fails.

[0170] In an exemplary embodiment, the apparatus further includes: a verification information acquisition module, configured to acquire geological verification information of the target underground project after excavation; a path verification module, configured to verify the associated disaster path based on the geological verification information to obtain a prediction verification result of the associated disaster path; and a confidence update module, configured to update the knowledge confidence of each geological knowledge triple in the associated disaster path based on the prediction verification result.

[0171] Each module in the geological hazard prediction device for the aforementioned underground engineering can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0172] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 8As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores engineering geological information, historical case feature vectors, geological knowledge maps, and other data related to the target underground engineering project. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for predicting geological hazards in underground engineering projects.

[0173] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0174] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0175] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0176] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0177] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0178] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0179] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0180] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for predicting geological hazards in underground engineering projects, characterized in that, The method includes: Obtain engineering geological information of the target underground project; The engineering geological information is processed by feature extraction to obtain the geological feature vector of the target underground project; Based on the similarity between the geological feature vector and the feature vectors of each historical case, similar historical cases corresponding to the target underground project are obtained; Based on the engineering geological information and the associated geological entities corresponding to each of the similar historical cases, the associated disaster paths corresponding to the target underground project are inferred using a geological knowledge graph. Based on the associated disaster path, the target geological interpretation information and disaster prediction results corresponding to the target underground project are obtained.

2. The method according to claim 1, characterized in that, The step of reasoning about the associated disaster paths corresponding to the target underground project based on the engineering geological information and the associated geological entities corresponding to each of the similar historical cases using a geological knowledge graph includes: Based on the associated geological entities of the similar historical cases, candidate geological interpretation information corresponding to the similar historical cases is obtained; Based on the engineering geological information and the candidate geological interpretation information, the candidate disaster paths corresponding to the target underground project are inferred using a geological knowledge graph, and the inference confidence level corresponding to each candidate disaster path is calculated. Based on the candidate disaster path with the highest inference confidence, the associated disaster path corresponding to the target underground project is obtained.

3. The method according to claim 2, characterized in that, Calculate the inference confidence score for each of the candidate disaster paths, including: The path confidence of the candidate disaster path is obtained based on the knowledge confidence of the geological knowledge triples in each candidate disaster path. The path confidence is adjusted based on the engineering geological information to obtain the inference confidence of the candidate disaster path.

4. The method according to any one of claims 1 to 3, characterized in that, The geological knowledge graph is constructed according to the following steps: Semantic analysis is performed on geological knowledge data to identify the geological entities and relationship types within the data. Based on the geological knowledge data, the relationship type between each pair of geological entities is predicted, and the prediction results of the relationship between each pair of geological entities are obtained. When the relationship prediction result indicates that there is a relationship type between the geological entity pairs with a relationship confidence level greater than a confidence threshold, a geological knowledge triple is obtained based on the geological entity pairs and the relationship type; The geological knowledge triplets are stored in the graph database to obtain the geological knowledge graph.

5. The method according to claim 4, characterized in that, The method further includes: Based on the geological information of each historical case, the historical case feature vector of each historical case is obtained, as well as the geological knowledge associated with each historical case; Based on the geological knowledge associated with the historical cases, the associated geological entities corresponding to the historical cases in the geological knowledge map are obtained; Construct a bidirectional index between the historical case feature vector corresponding to the historical case and the associated geological entity.

6. The method according to claim 4, characterized in that, The method further includes: The knowledge confidence of the geological knowledge triple is obtained based on the data source information, review result information, distribution information in the geological knowledge data, and the relationship confidence. A consistency check is performed on each of the geological knowledge triples in the geological knowledge graph. When there are conflicting geological knowledge triples, the geological knowledge triple with higher knowledge confidence is retained, and the geological knowledge triple with lower knowledge confidence is verified. If the verification passes, the geological knowledge triple with the lower knowledge confidence level is retained; if the verification fails, the geological knowledge triple with the lower knowledge confidence level is discarded.

7. The method according to claim 6, characterized in that, The method further includes: After the target underground project is excavated, obtain the geological verification information of the target underground project; Based on the geological verification information, the associated disaster paths are verified to obtain the prediction verification results of the associated disaster paths; Based on the prediction verification results, update the knowledge confidence of each geological knowledge triple in the associated disaster path.

8. A geological hazard prediction device for underground engineering, characterized in that, The device includes: The information acquisition module is used to acquire engineering geological information of the target underground project; The feature extraction module is used to perform feature extraction processing on the engineering geological information to obtain the geological feature vector of the target underground project; The case retrieval module is used to obtain similar historical cases corresponding to the target underground project based on the similarity between the geological feature vector and the feature vectors of each historical case. The path reasoning module is used to reason about the associated disaster paths corresponding to the target underground project based on the engineering geological information and the associated geological entities corresponding to each of the similar historical cases, using a geological knowledge graph. The result acquisition module is used to obtain the target geological interpretation information and disaster prediction results corresponding to the target underground project based on the associated disaster path.

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