Geological feature database construction method and system for tunnel unfavorable geological identification

Abstract

The present application belongs to the technical field of underground engineering and engineering geological informatization. A geological feature database construction method and system for tunnel unfavorable geological identification are proposed, multi-source exploration information is obtained, a mileage window working face sequence is established, and heterogeneous data is uniformly mapped and aligned; a working face data package containing quality marks is constructed, merging and conflict resolution are performed, and the output is calculated fields and evidence index; then, the abnormal section is located along the mileage standardization processing and combined with the weight, and the mutation feature vector is extracted; according to the supervision label, the index adaptive binning and discretization are realized by iteratively optimizing the split point under constraints; the whole process data is layered into the database, the original record is only increased but not changed, the derived data is versioned, and based on the write-back mechanism of the discrimination result and the exposure conclusion, the label and rule version are updated, realizing the traceable closed loop iteration of unfavorable geological identification.

Description

Technical Field

[0001] This invention relates to the field of underground engineering and engineering geology information technology, specifically to a method and system for constructing a geological feature database for identifying adverse geological conditions in tunnels. Background Technology

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Currently, tunnel construction is developing towards longer distances, deeper burial sites, and more complex geological conditions, placing extremely high demands on the advanced identification of adverse geological conditions. With advancements in exploration technology, geological data sources are becoming increasingly diversified, encompassing information from multiple stages, including geological mapping, geophysical exploration, drilling, and construction exposure. These data vary significantly in spatial representation, accuracy, and update frequency. Constructing a geological feature database capable of integrating multi-source heterogeneous data, supporting dynamic updates, and possessing traceability has become a key infrastructure for achieving tunnel construction safety early warning and intelligent decision-making, and a core component in promoting the digital transformation of geotechnical engineering.

[0004] However, existing technologies struggle to effectively address the alignment challenges of spatial mapping between multi-source exploration data and information conflicts within the same working face when constructing geological databases. The lack of unified working face sequence units and standardized conflict resolution mechanisms prevents the transformation of raw data into computable feature fields, and breaks in the data source evidence chain hinder subsequent precise anomaly segment location and feature vector extraction. Furthermore, traditional methods lack adaptive discretization processing of continuous indicators and version control of discrimination rules, preventing the iterative optimization of adverse geological judgment results based on dynamic feedback from construction and exposure findings. This severely restricts the accuracy and timeliness of the identification model. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for constructing a geological feature database for identifying adverse geological conditions in tunnels. This system achieves unified alignment of geological information during the exploration period, structured data entry, automatic extraction of abrupt change features, data-driven parameterization, discriminative field output, and write-back update closed-loop iteration, thereby improving the stability, interpretability, and traceability of the identification results.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a method for constructing a geological feature database for identifying adverse geological conditions in tunnels.

[0007] A method for constructing a geological feature database for identifying adverse geological conditions in tunnels includes the following steps: Acquire multi-source exploration data and perform field standardization processing to form an original field set; Based on a unified mileage benchmark, a mileage window sequence is constructed, mapping the point data, segment data, and text description data in the original field set to the corresponding mileage windows to generate working face alignment results. Based on the workface alignment results, a workface data package containing geological field values, source index and quality markers is constructed for each mileage window. Multiple records within the same mileage window are merged and conflict resolution is performed, and the computable field values ​​and evidence index are output. Based on the weights generated by the computable field values ​​and quality markers, mutation detection is performed along the mileage sequence to locate the start and end positions of abnormal segments, and the mutation intensity, persistence, rate of change and co-triggered features of abnormal segments are extracted to form a feature vector; Based on the supervision labels, a set of indicators to be parameterized is constructed from the computable field values ​​and feature vectors. The binning boundary is determined through iterative optimization, and continuous indicators are mapped to discrete level indicators to generate binning rule versions. The working face alignment results, working face data packages, feature vectors, discrete level indicators, and binning rule versions are stored in the database in layers. Based on the discrete level indicators, adverse geological judgment results containing five elements, namely type, cause, range, development, and hazard, are output. Subsequent disclosure conclusions are written back to update the monitoring labels and binning rule versions.

[0008] In one implementation of the first aspect of the present invention, the point data includes mileage points, coordinate points, or profile measurement point data; the segment data includes mileage intervals, stratigraphic segments, or structural segments; the text description data includes geological phenomenon descriptions or catalog descriptions; when mapping the segment data to the corresponding mileage window, the coverage ratio of the segment data to each mileage window is recorded.

[0009] In one implementation of the first aspect of the present invention, the quality markers include coverage, missing rate and conflict markers; during the merging and conflict resolution process, records with conflict markers are processed according to preset priority rules or weighted fusion rules, and the source index of each record is retained as an evidence index.

[0010] In one implementation of the first aspect of the present invention, mutation detection employs a cumulative sum detection method; when extracting collaborative triggering features, the number of indicators exceeding a preset threshold within the same abnormal segment is counted, and the degree of synchronous mutation of multiple indicators is calculated as a component of the feature vector.

[0011] In one implementation of the first aspect of the present invention, when determining the binning boundary, mutual information is used as the optimization objective. Under the premise of satisfying the minimum number of bin samples and the maximum number of bins, a complexity penalty term is introduced, and the splitting point is iteratively selected until the mutual information gain is lower than a preset threshold, thereby obtaining the binning boundary.

[0012] In one implementation of the first aspect of the present invention, mapping continuous indicators to discrete level indicators further includes: mapping the computable field values ​​and missing values ​​in the feature vector to independent missing levels using a unified encoding strategy, and incorporating the missing levels into the level system of discrete level indicators.

[0013] In one implementation of the first aspect of the present invention, the layered data entry includes an original layer, a process layer, a feature layer, a parameter layer, and a conclusion layer; wherein, the original layer only stores the original field set and source index and adopts an add-only strategy, the parameter layer stores the binning rule version, the conclusion layer stores the adverse geological judgment results and the associated evidence index, and records the differences before and after the update and the timestamp during the write-back to form a version chain.

[0014] Secondly, the present invention provides a geological feature database construction system for identifying adverse geological conditions in tunnels.

[0015] A geological feature database construction system for identifying adverse geological conditions in tunnels includes: The data standardization unit is configured to: acquire multi-source exploration data and perform field standardization processing to form an original field set; The mileage alignment unit is configured to: construct a mileage window sequence based on a unified mileage benchmark, map point data, segment data and text description data in the original field set to the corresponding mileage windows, and generate working face alignment results; The data packet construction unit is configured to: construct a working face data packet containing geological field values, source index and quality marker for each mileage window based on the working face alignment results, and perform merging and conflict resolution on multiple records within the same mileage window, outputting computable field values ​​and evidence index; The anomaly detection unit is configured to: perform mutation detection along the mileage sequence based on the weights generated by the computable field values ​​and quality labels to locate the start and end positions of the anomaly segments, and extract the mutation intensity, persistence, rate of change and co-triggered features of the anomaly segments to form a feature vector; The indicator parameterization unit is configured to: construct a set of indicators to be parameterized based on the supervision labels, computable field values ​​and feature vectors, determine binning boundaries through iterative optimization, map continuous indicators to discrete level indicators, and generate binning rule versions. The database entry discrimination unit is configured to: hierarchically enter the working face alignment results, working face data package, feature vector, discrete level index and binning rule version into the database, and output adverse geological discrimination results containing five elements including type, cause, range, development and hazard based on the discrete level index, and write back the subsequent disclosure conclusions to update the supervision label and binning rule version.

[0016] Thirdly, the present invention provides a computer-readable storage medium storing a computer program adapted to be loaded by a processor and executed by a method for constructing a geological feature database for identifying adverse geological conditions in tunnels, as described in the first aspect of the present invention.

[0017] Fourthly, the present invention provides a computer device, comprising: a processor and a computer-readable storage medium; A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method for constructing a geological feature database for identifying adverse geological conditions in tunnels, as per the first aspect of this invention.

[0018] Compared with the prior art, the beneficial effects of the present invention are: This invention establishes a working face sequence based on mileage windows, achieving efficient mapping and alignment of multi-source heterogeneous geological data under a unified spatial benchmark, effectively solving the alignment problem caused by inconsistent data spatial representation in traditional methods. By constructing working face data packages containing source indexes and quality markers, and performing merging and conflict resolution, it not only outputs reliable computable field values ​​but also retains complete evidence indexes, ensuring the traceability of data sources, completely eliminating information conflicts, and maintaining the evidence chain without breaking. Using quality markers to generate weights to locate anomalous segments and extract mutation feature vectors, combined with iterative binning technology under constraints, it achieves adaptive mapping from continuous indicators to discrete level indicators, significantly improving the accuracy of feature representation. Adopting a hierarchical database strategy of adding only original records and versioning derived data, coupled with a mechanism for writing back disclosure conclusions, it achieves dynamic iterative updates of supervision labels, binning rules, and discrimination rules. This scheme constructs a geological feature database with self-evolving capabilities, significantly improving the accuracy and timeliness of adverse geological condition identification, providing strong data support and decision-making basis for tunnel construction safety.

[0019] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0020] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0021] Figure 1 A flowchart illustrating a method for constructing a geological feature database for identifying adverse geological conditions in tunnels, provided as an exemplary embodiment of the present invention; Figure 2A schematic diagram of a geological feature database construction system for identifying adverse geological conditions in tunnels, provided as an exemplary embodiment of the present invention; Figure 3 A schematic diagram of a computer device provided for an exemplary embodiment of the present invention. Detailed Implementation

[0022] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0023] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0024] In the entire lifecycle of tunnel engineering, the exploration phase is the most critical stage for acquiring underground geological information. This stage typically generates multi-source geological information, including borehole logging, stratigraphic and lithological segmentation, surrounding rock quality and weathering grade, fault and fracture zone investigation, groundwater and karst development, as well as corresponding maps and experimental results. This information is crucial for identifying adverse geological conditions, conducting risk analysis, and recommending parameters. However, in actual engineering projects, the above data vary significantly in terms of representation, spatial positioning, and data quality. For example, borehole data may be represented by depth and coordinates, while geological sketches describe mileage and relative location, and geophysical data may be presented as a sequence of measurement points. This multi-source heterogeneity means that engineering practice still relies heavily on manual compilation and empirical interpretation to achieve data fusion and identification conclusions, making it difficult to meet the needs of large-scale, reproducible, and traceable engineering applications.

[0025] Existing technologies suffer from at least the following significant problems: First, spatial benchmarks and multi-source alignment / merging are difficult, leading to unstable fusion. Inconsistent expressions such as mileage points / intervals, coordinate points, profile measurement points, and textual descriptions make it difficult to stably correlate information from multiple sources at the same location. When multiple records coexist within the same mileage window, a unified and reproducible merging strategy is lacking, making the results prone to human error. Second, the data is heterogeneous and exhibits significant quality differences, lacking quantitative quality control and conflict resolution mechanisms. Multi-source data shows significant differences in coverage, accuracy, and reliability, with missing, duplicate, and conflicting data being common. Existing practices rely heavily on manual screening or simple rules, lacking unified quality markers (missing rate, coverage, conflict markers, etc.) and consistent conflict resolution processes, making reliability evaluation and verification difficult. Finally, identification and parameterization heavily depend on manual methods / fixed thresholds, lacking an iterative and reusable closed loop. The boundaries of unfavorable geological sections are mainly interpreted manually, lacking automatic extraction of abnormal sections; the parameterization of indicators often uses fixed thresholds / expert classification, which is unstable in terms of adaptability and discrimination ability; at the same time, binning / rules lack version management, and the identification results are often not structured enough, making it difficult to trace, compare, update in a closed loop, and reuse across projects after supplementing data.

[0026] To address the aforementioned issues encountered when using tunnel geological information from the exploration period for adverse geological identification, such as inconsistent spatial benchmarks of multi-source data, difficulty in stable alignment and fusion; coexistence of multiple records at the same mileage location with widespread missing and conflicting data, lack of reproducible merging and quality control; reliance on manual interpretation of adverse geological section boundaries, making it difficult to automatically extract anomalous sections and stable features; reliance on fixed thresholds or expert grading for parameterization of indicators, lack of data-driven adaptive binning and version management; and difficulty in establishing an evidentiary correspondence between identification conclusions and original records, making verification and traceability challenging, and hindering closed-loop updates after supplementary exploration, this invention proposes a geological feature database construction and indicator parameterization method for tunnel adverse geological identification. This invention aims to achieve unified alignment, structured database entry, automatic extraction of abrupt change features, data-driven parameterization, field-based output of discrimination data, and closed-loop iteration of write-back updates for geological information from the exploration period, thereby improving the stability, interpretability, and traceability of identification results.

[0027] The following will be combined with the appendix Figure 1 The process shown illustrates the specific implementation steps of the present invention in detail. Figure 1 This is a flowchart illustrating a method for constructing a geological feature database and parameterizing indicators for identifying adverse geological conditions in tunnels. It showcases the complete technical path from multi-source data access to final closed-loop updates. First, multi-source data information is aligned across working faces. This stage includes three parallel sub-steps: unifying mileage coordinates, dividing working face windows, and projecting various data types to their corresponding working faces. After alignment, the data package construction stage begins, followed by CUSUM mutation feature extraction to identify geological anomalies. Next, the core mutual information-driven adaptive binning module is implemented. This module determines the optimal binning boundary through four steps: generating a candidate splitting point set, implementing a scoring + penalty mechanism, performing iterative splitting optimization, and checking monotonicity. The processed data is finally stored in the database, and judgment output is performed based on this data. Finally, a write-back update mechanism feeds newly generated disclosure or verification conclusions back to the database, forming a complete technical path from data access, feature extraction, parameterization to judgment output and closed-loop updates.

[0028] S101: Geological information access during the exploration period is aligned with the working face.

[0029] This step forms the basis for building a geological feature database. Its core purpose is to solve the problem of inconsistent spatial benchmarks for multi-source data and to achieve standardized data access and spatial alignment.

[0030] S101-1: Acquisition and standardized preprocessing of multi-source geological information.

[0031] First, the system needs to acquire at least one or more data sources during the exploration period. These data sources cover the results of the entire tunnel exploration process, specifically including: Borehole logging data includes borehole number, borehole coordinates, opening elevation, final borehole depth, footage per run, core recovery rate, rock quality index (RQD), lithological description, joint and fracture statistics, infill material properties, groundwater outlet points and water volume and temperature, etc. This type of data is usually plotted with depth (from the borehole opening downwards) as the vertical coordinate.

[0032] Stratigraphic and lithological segmentation data: Stratigraphic boundaries and lithological distribution segments formed by comprehensive interpretation of geological mapping and drilling. This type of data is usually expressed in mileage intervals (such as K10+200~K10+350) or station ranges, with clear start and end positions.

[0033] Surrounding rock quality and weathering grade data: This includes evaluation results such as surrounding rock classification (e.g., Grade I, Grade II, Grade III, Grade IV, Grade V) and weathering degree (fully weathered, strongly weathered, weakly weathered, slightly weathered, fresh). These data may be derived from standard table lookups, field tests, or expert scoring, and are often expressed as paragraph attributes or discrete point evaluations.

[0034] Fault and breccia zone survey data includes fault name, attitude (strike, dip, angle), breccia zone width, affected zone extent, tectonic rock characteristics, and water-bearing properties. This type of data is often presented in the form of the mileage of the tectonic zone center and the affected area (e.g., K12+100±15m).

[0035] Groundwater and karst development information: including the location of water inflow points, water volume, water pressure, water quality analysis results, location, size, and filling status of karst caves, etc. The data may be in point form (water inflow points) or segment form (water-rich segments).

[0036] Test results data include indoor rock mechanics tests (uniaxial compressive strength, elastic modulus, Poisson's ratio, etc.) and in-situ tests (sonic testing, in-situ stress testing, water pressure testing, etc.). These data are usually associated with specific sampling locations or test borehole sections.

[0037] Map index data includes index information for electronic or paper maps such as geological plans, cross-sections, columnar sections, and geophysical interpretation maps, used to establish the connection between data and graphic evidence.

[0038] After acquiring the aforementioned multi-source data, it is essential to perform field standardization to eliminate semantic and format differences between the data sources. Standardization includes the following aspects: Unified field naming: Establish a standard data dictionary to map similar fields from different sources to a unified standard field name. For example, map "Rock Quality Designation", "RQD value", and "rock mass quality index" to the standard field "RQD"; map "inflow", "outflow", and "flow rate" to "Water_Inflow".

[0039] Dimensional Definition and Unit Conversion: Clearly define the physical dimensions of each field and convert all data to a unified International System of Units (SI) or commonly used engineering units. For example, unify pressure units to MPa, length units to m, and flow rate units to L / min or m³ / d. For non-numerical data such as lithological descriptions, establish a standard coding system (e.g., using national standard lithology codes).

[0040] Value range verification: Based on geological knowledge and standard requirements, set reasonable value ranges for each field. For outliers outside the range (such as negative depth, sampling rate exceeding 100%, etc.), mark them and trigger a manual review process to prevent erroneous data from entering subsequent calculation stages.

[0041] Risk Direction Marking: Different indicators are marked to indicate the direction of their influence on adverse geological conditions. For example, a lower RQD value indicates more fractured rock mass and higher risk, marked as a "negative indicator"; a larger water inflow indicates higher risk, marked as a "positive indicator"; and a higher surrounding rock grade number (e.g., grade V is worse than grade III) indicates higher risk, marked as a "positive indicator." This marking will be used in subsequent abrupt change detection to determine the directionality of indicator changes.

[0042] After the above preprocessing, a set of raw fields that can be stored in the database is formed, which prepares for subsequent spatial alignment.

[0043] S101-1: Establishment of working face sequence and construction of mileage window.

[0044] To achieve stable alignment of multi-source data, this invention abandons the traditional approach of using a single data source (such as boreholes) as a benchmark, and instead establishes an independent working face sequence with "mileage windows" as the smallest unit. Mileage is the only continuous spatial benchmark in tunnel engineering, which can be used throughout the entire process of design, construction, and operation and maintenance.

[0045] In practical implementation, the tunnel's design centerline mileage is first used as a unified spatial benchmark to establish a continuous coordinate axis from mileage zero to mileage end. Then, based on a balance between engineering accuracy requirements and computational efficiency, the mileage window length ΔL is set. The value of ΔL can be flexibly set according to specific engineering characteristics; for example, in sections with complex geological conditions requiring precise identification, it can be set to 1m or 2m; in sections with relatively uniform geological conditions, it can be set to 5m or 10m. The system divides the entire tunnel into a series of continuous, non-overlapping (or slidingly overlapping according to a set step size) mileage windows, each mileage window defined as a "working face." For example, if ΔL = 2m, the working face sequence is as follows: .

[0046] S101-3: Spatial mapping of multi-source data to the working surface.

[0047] After establishing the working surface sequence, the standardized data of various types need to be uniformly mapped to the corresponding mileage window working surfaces. Since the spatial representation of the original data varies, different mapping strategies are required: (1) Mapping of point data: For point data with clear spatial coordinates or mileage locations (such as borehole sampling points, geophysical detection points, water inrush points, and ground stress test points), the system directly determines the mileage window to which it belongs based on its mileage value (if it is a coordinate, it needs to be projected to the centerline mileage first).

[0048] If the mileage of the point data satisfy (in and For the first If the starting and ending mileage of the first working face is specified, then this data is directly assigned to the first... One work surface.

[0049] For borehole data, since its original coordinates are borehole depth, it is necessary to use borehole trajectory measurement data (inclinometer data) to convert the borehole depth into three-dimensional spatial coordinates, and then project it onto the tunnel centerline to obtain the corresponding mileage. If inclinometer data is lacking, the mileage can be estimated by approximately assuming the borehole is vertical or by using simple geometric relationships, but a "mileage estimation" label must be recorded in the quality marking to reduce its weight.

[0050] (2) Mapping of segmented data: For segmented data that exists in the form of intervals (such as stratigraphic segments, surrounding rock classification segments, fault influence zones, etc.), multiple working faces need to be allocated according to their coverage relationship.

[0051] If a certain geological section Complete coverage of the work surface Then, the corresponding field of the working face directly inherits the attribute value of the geological section, where, This is the starting mileage for geological / tectonic segments. This is the mileage at the end of a geological / tectonic segment. Let be the starting mileage of the i-th mileage window (working surface). This is the termination mileage of the i-th mileage window (working surface).

[0052] If a certain geological section only partially covers the first One working face (i.e.) and If there is overlap but no overlap, then calculate the coverage ratio. ,in, This is used to determine the mileage length of the interval. This attribute value can be assigned to the work surface, and the coverage ratio can be recorded. As part of the quality criteria; or, for indicators with high continuity requirements, a weighted average method can be used to handle working faces spanning multiple geological sections.

[0053] For example, a Class IV surrounding rock section is defined as K10+001~K10+003, with a mileage window of 2m (K10+000~K10+002, K10+002~K10+004). Then, in the first working face (K10+000~K10+002), 1m belongs to Class IV and 1m belongs to another class (let's assume Class III); in the second working face (K10+002~K10+004), 1m belongs to Class IV and 1m belongs to another class. The system needs to record this mixed state, or determine the surrounding rock class of the working face during merging based on the dominant principle (such as the safety principle of "choosing the lower class"), while retaining the original composition information for traceability.

[0054] (3) Mapping of text description data: For unstructured textual descriptive data (such as phenomenon descriptions and disaster sign records in geological logging), the data is assigned by parsing the associated mileage information, structural segment names, or borehole segment numbers.

[0055] If the description explicitly relates to a mileage range, it will be processed according to the segmented data mapping logic; If the description is only related to the borehole depth, it is first converted to mileage and then mapped. If the description is vague (e.g., "around K10+500"), it will be mapped to all affected working surfaces according to the preset fuzzy radius (e.g., ±5m).

[0056] Crucially, source indexes (such as "description at borehole ZK01-25m" and "geological sketch section-05 annotation") must be preserved to facilitate the establishment of a chain of evidence later.

[0057] Through the mapping process described above, all multi-source heterogeneous geological information is uniformly reorganized into a structured sequence indexed by mileage window working faces. Each working face now possesses a set of potential information from different data sources, resolving the instability issue caused by inconsistent spatial benchmarks and achieving stable alignment of multi-source data.

[0058] S102: Workface data package construction and quality marking.

[0059] After spatial alignment is completed, the system needs to construct a complete "working face data package" for each working face. This data package not only contains geological field values, but also crucial quality markers and evidence indexes, aiming to address the problems of data heterogeneity, large quality differences, and the lack of conflict resolution mechanisms.

[0060] S102-1: The composition structure of the working face data packet.

[0061] Each working face data package contains at least the following three core sets: (1) Geological Field Set: This set contains all mapped original fields and preliminary derived fields within the working face. For numerical fields (such as RQD, compressive strength, and water inflow), the specific numerical values ​​are stored; for categorical fields (such as lithology and surrounding rock grade), the coded values ​​are stored; and for textual fields, the descriptive content is stored. If the same field has multiple source values ​​within the working face (for example, both borehole data and geophysical interpretation data exist at this mileage), all candidate values ​​are temporarily retained for subsequent allocation and use.

[0062] (2) Source Index Set: Records the source information of each field value, forming a traceable chain of evidence. The index content includes, but is not limited to: data source type (drilling, surveying, geophysical exploration, experiment), original document number (such as borehole logging table number, map number), specific location identifier (such as borehole number and depth, measuring point number, description segment number), entry time and responsible person. This set ensures that any calculation result can be traced back to the most original record, meeting the requirement of "verifiable traceability".

[0063] (3) Quality Marker Set: This is the key innovation that distinguishes this invention from existing technologies. The quality marker set is used to quantitatively evaluate the reliability and completeness of the working face data, and includes at least the following indicators: Coverage: refers to the proportion of valid data for the target geological field within a working face. For point data interpolation or segmental data coverage, coverage reflects the representativeness of the data. For example, if within a 2m working face, only 1m is covered by borehole cores, and the remaining 1m is inferred, then the coverage is 0.5.

[0064] Missing Rate: This refers to the proportion of key fields (such as surrounding rock grade, lithology, and groundwater) for which there is no valid data in the working face. A high missing rate means that the identification results of the working face have a high degree of uncertainty.

[0065] Conflict Flag: A conflict flag is triggered when multiple sources exist for the same field within the same working face and the numerical differences exceed a preset threshold. For example, a borehole might show "slightly weathered granite," while surface mapping infers "strongly weathered," creating a clear conflict. The system must record the specific details of the conflict, the data sources involved, and the magnitude of the difference.

[0066] Consistency Score: A quantitative indicator calculated based on the degree of agreement between multiple data sources. A high score is awarded if there are no conflicts and the data from multiple sources corroborate each other; a low score is awarded if there are conflicts or the data is sparse. This score will be used as a weighting factor in subsequent mutation detection.

[0067] S102-2: Multi-record merging and conflict resolution strategy.

[0068] For situations where multiple records coexist within the same working surface, this invention proposes a reproducible merging and conflict resolution process to avoid the arbitrariness caused by manual intervention.

[0069] The merging rules are pre-defined, with a set of rules based on data credibility priority. Credibility priority typically follows the principles of: "measured data is superior to inference," "nearer data is superior to farther data," "higher precision is superior to lower precision," and "consistency across multiple sources is superior to single-source data." For example, lithological descriptions from borehole core samples have higher priority than surface geological inferences; wave velocities obtained from acoustic testing have higher priority than empirical estimates.

[0070] For numerical fields (such as RQD and intensity), if multiple valid values ​​exist, and there are no conflicts (differences are within the allowable error range), a weighted average is used as the final field value. The weights are determined by the reliability level of the data source and the consistency score in the quality label. If conflicts exist (differences exceed the threshold), the average is not directly applied; instead, all candidate values ​​and their sources are retained, and a "conflict label" is generated. Simultaneously, the system can generate a "conservative estimate" for preliminary identification based on safety principles (such as taking the worst-case scenario, i.e., minimum intensity or maximum flow rate). However, this value must be explicitly marked as a "conservative estimate" rather than a "measured average" in the data packet, and conflict evidence must be associated with it.

[0071] Categorical field merging: For categorical fields (such as lithology and surrounding rock grade): if multiple sources agree, confirm directly; if inconsistent, use a "voting mechanism" or the "lowest priority principle" (for safety indicators). For example, if there is a dispute between Grade III and Grade IV of the surrounding rock grade, it will be temporarily designated as Grade IV and marked as conflicting. The system will list the evidence sources supporting each grade for expert review.

[0072] The output is generated, and after merging and conflict resolution, it outputs the "computable field values" and "traceable evidence index" of the work surface. Even if conflicts exist, the system ensures the transparency and reproducibility of the processing. Any subsequent personnel who run the process again will get the same result as long as the input data and rules remain unchanged, completely solving the problem of "results changing with people".

[0073] By constructing a working surface data package containing rich quality information, this invention achieves quantitative control over data quality, providing a reliable weight basis for subsequent mutation feature extraction and avoiding interference from low-quality data on the identification results.

[0074] S103: Geological mutation feature extraction.

[0075] Unfavorable geological bodies (such as fault fracture zones, water-rich areas, and weak interlayers) typically exhibit significant abrupt changes in geological parameters along the mileage curve. Traditional methods rely on manual visual identification of these abrupt changes, which is inefficient and highly subjective. This invention uses algorithms to automatically locate anomalous segments and extract structured features, reducing reliance on human experience.

[0076] S103-1: Directional unification and standardization of indicator sequences.

[0077] Before performing mutation detection, the selected key indicator sequences for identification must first be preprocessed.

[0078] Unified Direction: Based on the "risk direction marker," all indicators are converted into "positive risk indicators." That is, the higher the value, the higher the geological risk. For example, for RQD (the lower the value, the higher the risk), the transformation is performed... or For the surrounding rock grade (the higher the number, the higher the risk), it remains unchanged. In this way, all anomalies are represented as "peak values" or "high value areas," simplifying the detection logic.

[0079] Standardization: Due to the huge differences in the dimensions and orders of magnitude of different indicators (e.g., the flow rate may be several hundred, while the friction coefficient is only a few tenths), it is necessary to standardize the series (e.g., Z-Score standardization or Min-Max normalization) to eliminate the influence of dimensions, make different indicators comparable, and facilitate subsequent collaborative analysis.

[0080] Quality weighting is introduced: weights are assigned to the indicator values ​​of each milestone using the generated quality tags (specifically, consistency scores and coverage). Points with high-quality data have higher weights, while points with low-quality data (such as those with missing data or severe conflicts) have lower weights. In subsequent calculations, the contribution of low-weight data to mutation detection will be suppressed, thereby improving noise resistance.

[0081] S103-2: Abnormal segment location based on cumulative sum (CUSUM).

[0082] This invention employs the Cumulative Sum Control Chart (CUSUM) algorithm to automatically locate the start and end points of anomaly segments. This method is highly sensitive to small shifts in the mean and is suitable for capturing gradual to abrupt changes in geological parameters.

[0083] Alternatively, a specific implementation is provided as follows: Let the standardized index sequence be The corresponding quality weight is ,in, Representing the Standardized index values ​​corresponding to each mileage window Representing the The overall weight of each working face.

[0084] Calculate background values: Select a section with relatively stable geological conditions (or calculate the mean background value using a sliding window statistical method). and standard deviation .

[0085] Construct a cumulative sum sequence and define the forward cumulative sum. With negative cumulative sum (For risk indicators, focus primarily on positive deviations): (1); (2); in, This represents the positive cumulative sum of the previous mileage window. This represents the negative cumulative sum of the previous mileage window.

[0086] Determine the start and end of an anomaly: Set a threshold. (like or ),when Start from 0 and rise above When, it is marked as the start of the abnormal segment, when When the price falls from a high level to 0, it is marked as the end of the abnormal segment.

[0087] In this way, the system can automatically identify mileage intervals where parameters significantly deviate from background values, i.e., "abnormal segments." For segments with high noise or many missing values, due to weighting... The level is low, the accumulation and growth are slow, and it is not easy to trigger the exception by mistake, or it will be accompanied by a high uncertainty flag after being triggered.

[0088] S103-3: Extraction of structural mutation features.

[0089] After locating the anomalous segment, further extract the structured features describing the nature of the anomalous segment to form a feature vector for identifying adverse geological conditions. These features include at least: Anomaly Intensity: The difference between the maximum value of the index within the anomaly segment and the background mean, or the difference between the weighted average value within the anomaly segment and the background mean, reflects the severity of adverse geological conditions.

[0090] Duration: The length of the anomalous segment (end point - start point). Reflects the size of the adverse geological body.

[0091] Rate of Change: The absolute value of the slope of the index curve at the initial boundary of the anomaly segment. It reflects the degree of drastic change in geological conditions (such as the fault contact zone, which usually changes at an extremely rapid rate).

[0092] Extent: The percentage of mileages where the metric exceeds a certain high-risk threshold.

[0093] Synergy: This counts how many different types of indicators (such as simultaneous low wave velocity, high water inflow, and low RQD) abruptly occur within the same anomaly segment. The more synergistic indicators present, the higher the reliability of the adverse geological assessment.

[0094] Morphological characteristics: such as the symmetry of anomalous segments and the number of peaks, are used to help distinguish the types of geological bodies (e.g., a single sharp peak may correspond to a fault, while a wide and gentle valley may correspond to a weak rock layer).

[0095] Through the above steps, the geological parameters that change continuously along the mileage are transformed into discrete, structured "anomaly segment feature vectors". This not only reduces the reliance on manual segmentation interpretation, but also provides high-quality input features for subsequent data-driven parameterization.

[0096] S104: Adaptive binning index parameterization.

[0097] After obtaining the structured feature vectors and original exploration fields, continuous numerical indicators need to be converted into discrete grade indicators (such as "high, medium, low" or "level I, II, III") to facilitate rule identification and engineering applications. Existing technologies mostly use fixed thresholds or expert experience for grading, which has poor adaptability. This invention proposes an adaptive binning method based on mutual information, which automatically determines the boundary of indicator levels by the discriminative power of the data itself.

[0098] S104-1: Construction of supervised labels.

[0099] Adaptive binning requires monitoring signals. The system uses the identification results of historical projects, construction disclosure conclusions, or expert review opinions as "monitoring tags".

[0100] Label definition: Mark each working face or anomalous section as "unfavorable geology" or "normal geology", or more granular type labels (such as "fault fracture zone", "karst water-rich area", "high geostress zone", etc.).

[0101] Tag sources can be measured data from completed tunnels or data from pilot tunnels in the early stages of the project. As the project progresses, newly revealed data can be continuously added to the tag set.

[0102] S104-2: Mutual information-driven iterative binning.

[0103] The continuous indicators to be parameterized (such as "abnormal intensity", "wave velocity", "inflow rate", etc.) are used as input variables. The supervisory label is used as the output variable. The goal is to find a set of splitting points. ,Will Divided into There are intervals (boxes) such that the discrete variables after partitioning are... and Maximizing mutual information (MI) between individuals; mutual information measures the knowledge... Later The degree of reduction in uncertainty, that is, the ability of the binned indicators to distinguish adverse geological types.

[0104] Optionally, a specific algorithm flow is given as follows: Initialization: Set the minimum number of bins. (To prevent overfitting, such as ensuring each bin contains at least 10 samples) and the maximum number of bins. (Control complexity, such as a maximum of 5 levels). The initial set of splitting points is empty.

[0105] Candidate split point evaluation: Iterate through all possible split points (usually the quantile points of the sample) and calculate the weighted mutual information gain of the two subsets after splitting the data in half at that point.

[0106] Optimal split selection: Select the split point with the largest mutual information gain and add it to the split point set, provided that the number of samples in both subsets after the split is greater than [a certain value]. And the current number of boxes does not exceed .

[0107] Complexity Penalty and Iteration: A complexity penalty term (such as based on the Bayesian Information Criterion (BIC)) is introduced to prevent excessive bin counts from reducing the model's generalization ability. The objective function is optimized as follows: ,in, This is the penalty coefficient. Repeat steps S102-S103 until no significant split point that improves mutual information can be found or the upper limit of the constraint is reached.

[0108] Uniform coding for missing values: Instead of simply filling in the mean or 0 for missing values ​​in the indicators, they are coded as a separate "missing" category. This is because in geological engineering, "missing data" itself may be important information (e.g., the inability to obtain a core sample may indicate that the rock mass is extremely fragmented), and binning it separately helps to preserve this characteristic.

[0109] S104-3: Versioned storage of binning rules.

[0110] After each adaptive binning operation, the system not only outputs the binning boundaries (e.g., inflow rate less than 10 L / min is low, 10-50 is medium, and greater than 50 is high), but also records the metadata that generated the rule. Sample version: Which projects and time periods' data were used as the training set.

[0111] Tag version: Monitor the source and update time of the tags.

[0112] Constraints: at that time , And the penalty coefficient setting.

[0113] Performance metrics: The discriminative power of the binning scheme on the validation set (e.g., AUC value, accuracy).

[0114] This metadata, along with the binning boundaries, is stored in a versioned manner. This means that when new reveal data (supplementary exploration or construction reveal) is added later, the system can rerun the binning algorithm to generate a new rule version (such as V2.0) and compare it with the old version (V1.0). This mechanism avoids the insufficient adaptability problem caused by fixed thresholds and supports continuous iterative optimization of the model.

[0115] S105: Structured data entry and version management.

[0116] To achieve data traceability and closed-loop updates, this invention designs a layered data entry strategy and a strict version management mechanism. The database architecture consists of five layers: raw data layer, process layer, feature layer, parameter layer, and conclusion layer.

[0117] S105-1: Layered data entry strategy.

[0118] (1) Raw Layer: Contents: Original records such as borehole logs, map indexes, and test reports stored and accessed.

[0119] Strategy: Adopt an "Append-Only" strategy. Once data is entered into the database, direct modification or deletion is strictly prohibited. If errors are found in the original data, the only recourse is to add a correction record and link it to the original record, ensuring the authenticity of the original data and the completeness of audit trails.

[0120] (2) Process Layer: Contents: Stores the generated working face alignment results, generated merged intermediate state data, conflict details, and quality tags.

[0121] Function: To record the process of data transformation from its original form to a standardized form.

[0122] (3) Feature Layer: Content: Stores structured feature vectors such as the extracted anomalous segment boundaries, mutation intensity, persistence, and rate of change.

[0123] Function: It preserves high-order geological features refined by algorithms and serves as direct input to the identification model.

[0124] (4) Parameter Layer: Contents: Stores the generated bin boundaries, discretization index values, and binning rule version information.

[0125] Function: To manage the standards and rules for parameterization of indicators, and to support the coexistence of multiple versions.

[0126] (5) Conclusion Layer: Content: Stores the final adverse geological assessment results, risk recommendations, and evidence explanations.

[0127] Function: The final delivery layer for users.

[0128] S105-2: Versioning Management and Evidence Chain Binding.

[0129] Versioning of derived features and rules: For data in the process layer, feature layer, and parameter layer, old data is not directly overwritten, but versioned management is adopted. Each record is bound to a version number (Version ID) and generation conditions (such as the algorithm version used, parameter configuration, and input data version).

[0130] Evidence chain binding: Each judgment result at the conclusion level must be bound to a complete evidence chain. The evidence chain is a set of pointers pointing to feature records at the feature layer, merge records at the process layer, and finally to the original catalog and maps at the original layer.

[0131] For example, if a working face is identified as a "fault fracture zone", its chain of evidence includes: the "low wave velocity anomaly feature" on which the identification is based (pointing to the feature layer) → the feature originates from the "merging of acoustic test data and borehole RQD data" (pointing to the process layer) → the merged data originates from the "25-30m section of borehole ZK03" and "geophysical profile Line-02" (pointing to the original layer).

[0132] Dependency Recording: The system automatically records the generation dependency graph between data at each layer. When the underlying raw data changes (such as adding a borehole) or the rules change, the system can automatically identify the affected upper-level features and conclusions and prompt that recalculation is required.

[0133] This hierarchical and version management mechanism ensures that historical results can be recalculated and compared under the condition of new data without overwriting the original records, thus truly achieving "traceability".

[0134] S106: Field-based output and write-back update of adverse geological conditions.

[0135] This step is the system's output stage and a key part of the closed-loop iteration. The system combines parameterized indicators with a rule base to output structured discrimination results and achieves self-evolution through a write-back mechanism.

[0136] S106-1: Discriminant Field-Based Output Using working faces or candidate anomaly segments as units, and based on the discretized indicators in the parameter layer and the preset / self-learned discrimination rule base, adverse geological conditions are identified. To facilitate retrieval, statistics, comparison, and report delivery, the discrimination results are strictly structured into five key fields: Type: The specific category of unfavorable geological conditions, such as "fault fracture zone", "karst conduit", "weak interlayer", "high geostress zone", "mud and water inrush zone", etc.

[0137] Cause: The main geological reasons that lead to this unfavorable geological condition, such as "tectonic compression", "karst erosion", "sedimentary facies change", "unloading relaxation", etc.

[0138] Distribution range: The spatial distribution of adverse geological bodies, including geometric information such as starting and ending mileage, length, width, and burial depth.

[0139] Development status: describes the activity level or development trend of the adverse geological body, such as "stable", "developing", "intensely active", "well filled", "poorly filled", etc.

[0140] Hazard level: The level of potential threat to engineering safety, such as "minor", "moderate", "serious" or "catastrophic", and can be associated with specific risk probability values.

[0141] In addition to the five elements, the system also outputs simultaneously: Risk recommendations: Engineering treatment suggestions based on the assessment results, such as "advanced grouting", "strengthening support", "changing construction methods", and "increasing monitoring".

[0142] Key Parameter Summary: List the key indicators that trigger this judgment and their values ​​(e.g., "Predicted inflow > 1000m³"). 3 / d”).

[0143] Evidence description: Automatically generate a natural language description or structured list that briefly outlines the chain of evidence upon which the judgment is based.

[0144] This field-based output transforms the identification results from vague text reports into structured data that can be directly queried in databases, statistically analyzed, and visualized, greatly improving the efficiency of engineering applications.

[0145] S106-2: Writeback Update and Closed-Loop Iteration The core advantage of this invention lies in its closed-loop iterative capability. During tunnel construction, as excavation reveals more details, the actual geological conditions are confirmed or corrected.

[0146] Obtaining the results of disclosure / verification: Collecting actual disclosure information such as construction face logging, advanced forecast verification, and monitoring measurement feedback. This information constitutes a new "truth value" label.

[0147] Write back to the database: Write the findings back to the database as new monitoring labels.

[0148] The update trigger mechanism is as follows: (1) Label update: Replace or supplement the original prediction labels with actual disclosure labels.

[0149] (2) Bin version update: Using the newly added ground truth samples, the adaptive binning algorithm of S104 is rerun. If the new data indicates that the original binning boundary (such as the inflow threshold) is no longer optimal, the system generates a new binning rule version (such as V2.0) and marks the applicable scope.

[0150] (3) Rule version update: Based on the new version of parameters and labels, retrain or adjust the discrimination rule base to generate a new rule version.

[0151] (4) Difference Recording: The system automatically records the differences before and after the update (e.g., "originally identified as Level IV, now revealed as Level V; original water inrush threshold was 50, new threshold is 30") and adds a timestamp.

[0152] The updated model was immediately applied to the prediction of unexcavated sections, making the identification results more and more accurate as the project progressed. At the same time, because all historical versions were retained, engineers could backtrack and compare at any time to analyze why previous identifications were biased, thereby continuously optimizing geological understanding.

[0153] Through a closed loop of "discrimination output - exposure verification - write-back update", this invention realizes the self-evolution and knowledge accumulation of the geological identification system, and solves the pain points of traditional methods being difficult to reuse and adapt to new situations.

[0154] To more clearly illustrate the effects of this invention, let's assume that this method is applied in a long railway tunnel project.

[0155] Assume the tunnel traverses complex carbonate rock strata with intense karst development and multiple regional faults. Extensive drilling and geophysical exploration were completed during the exploration phase, but the data is fragmented, leading to inconsistent assessments of the risk of karst water inrush ahead.

[0156] The implementation process is as follows: Data Access: The system accessed logging data from 50 boreholes, interpretation data from three geophysical profiles, and surface geological survey data. By aligning the working faces, the data, which were originally scattered across depth, coordinate, and mileage sections, were unified into a sequence of working faces, each 2 meters long.

[0157] Quality control: In the K15+200~K15+300 section, the system detected sparse borehole data (coverage <30%), but geophysical data showed obvious low resistivity anomalies. The system automatically marked this section as "high missing rate, high conflict risk" and gave geophysical data a higher weight during merging, while also indicating that supplementary exploration is needed.

[0158] Abrupt change detection: Abrupt change detection is performed on indicators such as resistivity, predicted flow rate, and RQD. The algorithm automatically identifies a strong "low resistivity-low RQD" co-anomaly segment in the range of K15+240~K15+280, and extracts features with high anomaly intensity and rapid change rate.

[0159] Adaptive binning: Using exposed data from adjacent sections as labels, the system automatically learns that, in this geological context, the combination of resistivity less than 150 Ω·m and RQD less than 20% has the highest distinguishability (maximum mutual information) for "large karst caves". Based on this, dynamic binning rules are generated instead of applying fixed values ​​from the specifications.

[0160] Identification Output: The system outputs the identification result: the type is "karst conduit", the cause is "dissolution collapse", the range is "K15+240~K15+280", the development is "poor filling, high water conductivity", and the hazard is "severe (high risk of water inrush)". Recommended measures: "Advanced horizontal drilling verification, pre-grouting reinforcement". Attached evidence includes: a low-resistivity anomaly zone pointing to geophysical profile P2 and a drill bit loss record in borehole ZK12.

[0161] Closed-loop update: When construction excavation reached K15+250, a large mud-filled karst cave was actually exposed, confirming the identification results. The system wrote back this discovery, strengthening the correlation between this type of feature and the "karst conduit" tag, and optimizing the identification model for subsequent sections.

[0162] This invention achieves unified alignment of multi-source geological information by constructing a working face sequence based on mileage windows; it achieves quantitative quality control of data fusion by introducing quality indicators such as coverage, missing rate, and conflict markers; it achieves automatic extraction and parameterization of adverse geological features through adaptive binning driven by accumulation, detection, and mutual information; and it achieves traceability and closed-loop iteration of identification conclusions through hierarchical storage, version management, and write-back update mechanisms.

[0163] This invention is not only applicable to tunnel engineering, but can also be extended to geological identification and risk assessment of linear underground projects such as subways, water conservancy tunnels, and underground utility tunnels. It transforms traditional qualitative and empirical geological analysis into quantitative, data-driven intelligent identification, significantly improving the safety, economy, and intelligence level of engineering construction.

[0164] Figure 2 This paper presents a geological feature database construction system for identifying adverse geological conditions in tunnels, including: Data standardization unit 201 is configured to: acquire multi-source exploration data and perform field standardization processing to form an original field set; Mileage alignment unit 202 is configured to: construct a mileage window sequence based on a unified mileage benchmark, map point data, segment data and text description data in the original field set to the corresponding mileage windows, and generate working face alignment results; The data packet construction unit 203 is configured to: construct a working face data packet containing geological field values, source index and quality mark for each mileage window based on the working face alignment result, and perform merging and conflict resolution on multiple records in the same mileage window, and output computable field values ​​and evidence index; Anomaly detection unit 204 is configured to: perform mutation detection along the mileage sequence based on the weights generated by the computable field values ​​and quality markers to locate the start and end positions of the anomaly segments, and extract the mutation intensity, persistence, rate of change and co-triggered features of the anomaly segments to form a feature vector; The indicator parameterization unit 205 is configured to: construct a set of indicators to be parameterized based on the supervision label, computable field values ​​and feature vectors, determine binning boundaries through iterative optimization, map continuous indicators to discrete level indicators, and generate binning rule versions. The database entry discrimination unit 206 is configured to: enter the working face alignment results, working face data package, feature vector, discrete level index and binning rule version into the database in layers, and output the adverse geological discrimination results containing five elements including type, cause, range, development and hazard based on the discrete level index, and write back the subsequent disclosure conclusions to update the supervision label and binning rule version.

[0165] It is understood that the aforementioned units can be individually or entirely merged into one or more other units, or some of the units can be further divided into multiple functionally smaller units. This achieves the same operation without affecting the technical effects of the embodiments of the present invention. The aforementioned units are based on logical functional division. In practical applications, the function of one unit can be implemented by multiple units, or the function of multiple units can be implemented by one unit. In other embodiments of the present invention, the system may also include other units. In practical applications, these functions can also be implemented with the assistance of other units, and can be implemented collaboratively by multiple units.

[0166] According to another embodiment of the present invention, the system of this embodiment can be constructed by running a computer program (including program code) capable of performing the steps involved in the corresponding method of the present invention on a general-purpose computing device, such as a computer, which includes processing elements and storage elements such as a central processing unit (CPU), random access memory (RAM), and read-only memory (ROM). The computer program can be recorded on, for example, a computer-readable recording medium, loaded into the aforementioned computing device through the computer-readable recording medium, and run therein.

[0167] Figure 3 A computer device is shown, which includes a processor 301, a communication interface 302, and a computer-readable storage medium 303. The processor 301, communication interface 302, and computer-readable storage medium 303 can be connected via a bus or other means.

[0168] The communication interface 302 is used to receive and send data. The computer-readable storage medium 303 can be stored in the memory of the electronic device. The computer-readable storage medium 303 is used to store computer programs, which include program instructions. The processor 301 is used to execute the program instructions stored in the computer-readable storage medium 303.

[0169] The processor 301 is the computing and control core of the electronic device. It is suitable for implementing one or more instructions, specifically for loading and executing one or more instructions to achieve the corresponding method flow or corresponding function.

[0170] Processor 301 is configured to perform the following procedure: Acquire multi-source exploration data and perform field standardization processing to form an original field set; Based on a unified mileage benchmark, a mileage window sequence is constructed, mapping the point data, segment data, and text description data in the original field set to the corresponding mileage windows to generate working face alignment results. Based on the workface alignment results, a workface data package containing geological field values, source index and quality markers is constructed for each mileage window. Multiple records within the same mileage window are merged and conflict resolution is performed, and computable field values ​​and evidence index are output. Based on the weights generated by the computable field values ​​and quality markers, mutation detection is performed along the mileage sequence to locate the start and end positions of abnormal segments, and the mutation intensity, persistence, rate of change and co-triggered features of abnormal segments are extracted to form a feature vector; Based on the supervision labels, a set of indicators to be parameterized is constructed from the computable field values ​​and feature vectors. The binning boundary is determined through iterative optimization, and continuous indicators are mapped to discrete level indicators to generate binning rule versions. The working face alignment results, working face data packages, feature vectors, discrete level indicators, and binning rule versions are stored in the database in layers. Based on the discrete level indicators, adverse geological judgment results containing five elements, namely type, cause, range, development, and hazard, are output. Subsequent disclosure conclusions are written back to update the monitoring labels and binning rule versions.

[0171] This invention also provides a computer-readable storage medium, which is a memory device in an electronic device for storing programs and data. It is understood that the computer-readable storage medium here may include both built-in storage media in the electronic device and extended storage media supported by the electronic device. The computer-readable storage medium provides storage space for storing the processing system of the electronic device.

[0172] Furthermore, this storage space also contains one or more instructions suitable for loading and execution by the processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory; alternatively, it can also be at least one computer-readable storage medium located remotely from the aforementioned processor.

[0173] In one embodiment, the computer-readable storage medium stores one or more instructions; the processor loads and executes the one or more instructions stored in the computer-readable storage medium to perform the following process: Acquire multi-source exploration data and perform field standardization processing to form an original field set; Based on a unified mileage benchmark, a mileage window sequence is constructed, mapping the point data, segment data, and text description data in the original field set to the corresponding mileage windows to generate working face alignment results. Based on the workface alignment results, a workface data package containing geological field values, source index and quality markers is constructed for each mileage window. Multiple records within the same mileage window are merged and conflict resolution is performed, and computable field values ​​and evidence index are output. Based on the weights generated by the computable field values ​​and quality markers, mutation detection is performed along the mileage sequence to locate the start and end positions of abnormal segments, and the mutation intensity, persistence, rate of change and co-triggered features of abnormal segments are extracted to form a feature vector; Based on the supervision labels, a set of indicators to be parameterized is constructed from the computable field values ​​and feature vectors. The binning boundary is determined through iterative optimization, and continuous indicators are mapped to discrete level indicators to generate binning rule versions. The working face alignment results, working face data packages, feature vectors, discrete level indicators, and binning rule versions are stored in the database in layers. Based on the discrete level indicators, adverse geological judgment results containing five elements, namely type, cause, range, development, and hazard, are output. Subsequent disclosure conclusions are written back to update the monitoring labels and binning rule versions.

[0174] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can implement the described functions using different methods for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0175] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of the present invention is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic cable, digital cable) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data processing device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc.

[0176] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for constructing a geological feature database for identifying adverse geological conditions in tunnels, characterized in that, The process includes the following: Acquire multi-source exploration data and perform field standardization processing to form an original field set; Based on a unified mileage benchmark, a mileage window sequence is constructed, mapping the point data, segment data, and text description data in the original field set to the corresponding mileage windows to generate working face alignment results. Based on the workface alignment results, a workface data package containing geological field values, source index and quality markers is constructed for each mileage window. Multiple records within the same mileage window are merged and conflict resolution is performed, and computable field values ​​and evidence index are output. Based on the weights generated by the computable field values ​​and quality markers, mutation detection is performed along the mileage sequence to locate the start and end positions of abnormal segments, and the mutation intensity, persistence, rate of change and co-triggered features of abnormal segments are extracted to form a feature vector; Based on the supervision labels, a set of indicators to be parameterized is constructed from the computable field values ​​and feature vectors. The binning boundary is determined through iterative optimization, and continuous indicators are mapped to discrete level indicators to generate binning rule versions. The working face alignment results, working face data packages, feature vectors, discrete level indicators and binning rule versions are stored in the database in layers. Based on the discrete level indicators, adverse geological judgment results containing five elements including type, cause, range, development and hazard are output. Subsequent disclosure conclusions are written back to update the supervision label and binning rule version. When determining the binning boundary, mutual information is used as the optimization objective. Under the premise of satisfying the minimum number of samples in the minimum bin and the maximum number of bins, a complexity penalty term is introduced. The splitting point is iteratively selected until the mutual information gain is lower than the preset threshold, thereby obtaining the binning boundary.

2. The method for constructing a geological feature database for identifying adverse geological conditions in tunnels as described in claim 1, characterized in that, Point data includes mileage points, coordinate points, or profile measurement points; segment data includes mileage intervals, stratigraphic segments, or structural segments; textual description data includes descriptions of geological phenomena or logging descriptions; when mapping segment data to the corresponding mileage windows, the coverage ratio of the segment data to each mileage window is recorded.

3. The method for constructing a geological feature database for identifying adverse geological conditions in tunnels as described in claim 1, characterized in that, Quality markers include coverage, missing rate, and conflict markers. During the merging and conflict resolution process, records with conflict markers are processed according to preset priority rules or weighted fusion rules, and the source index of each record is retained as an evidence index.

4. The method for constructing a geological feature database for identifying adverse geological conditions in tunnels as described in claim 1, characterized in that, The mutation detection adopts the cumulative sum detection method; when extracting co-triggered features, the number of indicators exceeding the preset threshold within the same abnormal segment is counted, and the degree of synchronous mutation of multiple indicators is calculated as a component of the feature vector.

5. The method for constructing a geological feature database for identifying adverse geological conditions in tunnels as described in claim 1, characterized in that, Mapping continuous indicators to discrete level indicators also includes: using a unified encoding strategy to map the computable field values ​​and missing values ​​in the feature vector to independent missing levels, and incorporating the missing levels into the level system of discrete level indicators.

6. The method for constructing a geological feature database for identifying adverse geological conditions in tunnels as described in claim 1, characterized in that, The layered data entry includes the original layer, process layer, feature layer, parameter layer, and conclusion layer. The original layer only stores the original field set and source index and adopts an add-only strategy. The parameter layer stores the binning rule version. The conclusion layer stores the adverse geological judgment results and related evidence indexes, and records the differences before and after the update and timestamps to form a version chain during write-back.

7. A geological feature database construction system for identifying adverse geological conditions in tunnels, characterized in that, include: The data standardization unit is configured to: acquire multi-source exploration data and perform field standardization processing to form an original field set; The mileage alignment unit is configured to: construct a mileage window sequence based on a unified mileage benchmark, map point data, segment data and text description data in the original field set to the corresponding mileage windows, and generate working face alignment results; The data packet construction unit is configured to: construct a working face data packet containing geological field values, source index and quality marker for each mileage window based on the working face alignment results, and perform merging and conflict resolution on multiple records within the same mileage window, outputting computable field values ​​and evidence index; The anomaly detection unit is configured to: perform mutation detection along the mileage sequence based on the weights generated by the computable field values ​​and quality labels to locate the start and end positions of the anomaly segments, and extract the mutation intensity, persistence, rate of change and co-triggered features of the anomaly segments to form a feature vector; The indicator parameterization unit is configured to: construct a set of indicators to be parameterized based on the supervision labels, computable field values ​​and feature vectors, determine binning boundaries through iterative optimization, map continuous indicators to discrete level indicators, and generate binning rule versions. The database entry discrimination unit is configured to: enter the working face alignment results, working face data package, feature vector, discrete level index and binning rule version into the database in layers, and output the adverse geological discrimination results containing five elements: type, cause, range, development and hazard, based on the discrete level index, and write back the subsequent disclosure conclusions to update the supervision label and binning rule version. When determining the binning boundary, mutual information is used as the optimization objective. Under the premise of satisfying the minimum number of samples in the minimum bin and the maximum number of bins, a complexity penalty term is introduced. The splitting point is iteratively selected until the mutual information gain is lower than the preset threshold, thereby obtaining the binning boundary.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed as described in any one of claims 1 to 6, for constructing a geological feature database for identifying adverse geological conditions in tunnels.

9. A computer device, characterized in that, include: Processor and computer-readable storage media; A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the method for constructing a geological feature database for identifying adverse geological conditions in tunnels as described in any one of claims 1 to 6.

Images