A method for generating a stratigraphic description based on a caliper log

Abstract

The present application relates to the technical field of geological exploration production operation, especially to a method for generating stratum description based on intelligent merging of return-to-shank records, which adopts a three-layer architecture of data collection layer, intelligent processing layer and interactive verification layer, collects multi-source information such as return-to-shank records and logging data through an upgraded APP and configures scene tags, generates description through CNN-LSTM model fusion, adaptive rule matching, complex geological body intelligent identification and closed-loop self-optimization in a cloud platform, and provides visual verification and multi-format output in an interactive layer. The system has the advantages of improving stratum description accuracy and scene adaptability, significantly reducing labor cost, supporting multi-industry standard and BIM docking, and being applicable to various engineering geological exploration scenes.

Description

Technical Field

[0001] This invention relates to the field of geological exploration and production operation technology, and in particular to a method for generating stratigraphic descriptions based on intelligent merging of backtrack records. Background Technology

[0002] Currently, engineering geological survey data acquisition has been digitized, enabling systematic management of massive amounts of engineering geological survey data. However, many problems still exist on-site, including the following shortcomings: (1) In the traditional data collection process, due to technical limitations and human error, a large amount of data is often collected repeatedly, resulting in an ineffective increase in data volume. This not only wastes valuable time and resources, but may also lead to inaccurate and chaotic data.

[0003] (2) In the traditional data collection process, due to human factors and operational differences, the collected data often has problems such as inconsistent format and inconsistent standards, which brings great trouble to subsequent data processing and analysis.

[0004] (3) In the traditional data processing process, the office staff need to spend a lot of time and energy to organize, classify and label the raw data. This is not only inefficient, but also prone to human error, mainly reflected in the extraction of stratigraphic description information based on the retrieval record. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of the existing technology by providing a method for generating stratigraphic descriptions based on intelligent merging of backtrack records. Through four major technological breakthroughs—multi-source data fusion modeling, adaptive rule engine, intelligent identification of complex geological bodies, and closed-loop self-optimization verification—the method achieves more accurate, scenario-based, and intelligent stratigraphic descriptions.

[0006] The objective of this invention is achieved through the following technical solutions: A method for generating stratigraphic descriptions based on intelligent merging of film reel records includes a mobile terminal program and a cloud platform. The mobile terminal program contains film reel records and stratigraphic descriptions. Data interaction occurs between the mobile terminal program and the cloud platform. The cloud platform receives the film reel records and stratigraphic descriptions sent by the mobile terminal program. The method is characterized by: The mobile terminal program serves as a data acquisition layer, used for the acquisition of multi-source geological data, scene configuration, and data uploading. The cloud platform includes an intelligent processing layer and an interactive verification layer. The intelligent processing layer receives data uploaded by the data acquisition layer, generates standardized stratigraphic description data through multi-source data fusion, adaptive rule matching, complex geological body identification, and closed-loop self-optimization processing, and pushes the processing results to the interactive verification layer. The interactive verification layer receives the processing results from the intelligent processing layer, provides a visual verification interface and correction function, feeds back user correction traces to the intelligent processing layer, and supports multi-format output and export of stratigraphic description results. The intelligent processing layer automatically merges soil and rock descriptions with consistent soil and rock names, states, and densities from the back-tracking records. The starting depth of the merged strata layer description is the starting depth of the first soil and rock description in the back-tracking records before the merger. The ending depth of the merged strata layer description is the ending depth of the last soil and rock description in the back-tracking records before the merger. Soil and rock categories, classifications, colors, moisture, inclusions, and soil layer descriptions are all based on the corresponding content in the first soil and rock description in the back-tracking records before the merger.

[0007] The mobile terminal program includes a basic data acquisition module, a multi-source data interface module, a scene configuration module, and a multimedia enhanced acquisition module. The basic data acquisition module is equipped with a return-to-the-base record input interface. The multi-source data interface module has a built-in data docking protocol, supporting the import or real-time docking of well logging data (including quantitative data such as resistivity and acoustic velocity), ground-penetrating radar data, and adjacent borehole data. The scene configuration module is used for users to configure project scene tags. The multimedia enhanced acquisition module integrates an image acquisition unit, a video acquisition unit, and a sensor data receiving unit, supporting the capture and acquisition of core microscopic images and on-site videos, as well as the reception of environmental sensor data such as temperature, humidity, and groundwater level dynamics.

[0008] The intelligent processing layer includes a hardware support unit, a multi-source data fusion module, an adaptive rule engine, a complex geological body identification module, a closed-loop self-optimization module, and a data traceability module.

[0009] The adaptive rule engine's parameter threshold dynamic adjustment unit is configured with a gradient formation determination subunit, and the determination logic of the subunit is as follows: (1) Preset parameter similarity thresholds: the humidity similarity threshold for the railway industry is set to 0.8, and the humidity similarity threshold for the construction industry is set to 0.75; (2) When the soil and rock names of the continuous back-track records are consistent, the similarity of the core parameters is higher than the corresponding threshold, and the variance of the stratigraphic probability distribution output by the CNN-LSTM model is less than 0.1, it is determined to be a gradually changing stratum. (3) The merging description of the gradient strata adopts the "parametric gradient description". The starting depth of the merged strata is associated with the starting depth of the first backtrack record, and the ending depth is associated with the ending depth of the last backtrack record.

[0010] The interactive verification layer includes an APP-side interactive unit and a Web-side interactive unit, both of which communicate with the intelligent processing layer via a network. It includes an intelligent auxiliary verification module, a special geological body description template library module, and a multi-format output module.

[0011] The mobile terminal programs are all equipped with a timestamp binding unit and a GPS positioning unit, and the data upload module supports the synchronous upload of multimedia materials such as photos and videos as well as structured data.

[0012] The return footage record includes: drilling method, termination time, starting depth, total drill string length, drill string above-ground length, termination depth, wall protection method, footage per return, total core length, core recovery rate, soil type, soil classification, soil name, color, state, humidity, density, inclusions, RQD, and description.

[0013] The stratigraphic description includes: starting depth, ending depth, soil type, soil classification, soil name, color, state, moisture content, density, inclusions, and soil layer description.

[0014] A single tracking record may contain one or more stratigraphic layers, and a stratigraphic layer may contain one or more tracking records.

[0015] The data in the tracking record and stratigraphic description share the same information, including soil type, soil classification, soil name, color, state, humidity, density, inclusions, and soil layer description. The tracking information is retained as the original exploration record.

[0016] The advantages of this invention are: 1) It can significantly reduce the repeated collection of the same data, thereby greatly reducing the workload of on-site work; 2) By establishing unified data collection standards and specifications, the standardization of data collection results has been improved, which greatly facilitates subsequent data processing and analysis; 3) It can automatically extract and generate stratigraphic layer description information, thereby greatly reducing the workload of office staff and improving the efficiency and accuracy of data processing; 4) It can significantly improve the accuracy of stratigraphic description; 5) It can accurately handle complex geological scenarios such as gradually changing strata, faults, and karst caves. The generated strata description can be directly connected to the BIM model to support subsequent engineering design and construction decisions. 6) The full-chain traceability chain enables the source of parameters to be traceable and the correction traces to be tracked, meeting the compliance requirements of engineering geological exploration. Attached Figure Description

[0017] Figure 1 This is a basic functional flowchart of the present invention. Detailed Implementation

[0018] The features and other related features of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments, so as to facilitate understanding by those skilled in the art: Example: A method for generating stratigraphic descriptions based on intelligent merging of backtrack records in this invention includes an APP + cloud platform, the overall architecture of which includes: Data Acquisition Layer (Upgraded App): The original film tracking records and stratigraphic description acquisition functions are retained. Added multi-source data interface: supports importing / interconnecting well logging data (resistivity, sonic velocity), ground-penetrating radar data, and adjacent borehole data; Added scene tag configuration: Supports users to select the industry (railway / highway / construction), region (such as East China Hills / Northwest Gobi), and survey specification version of the project; New multimedia enhancement acquisition: Supports the acquisition of core microscopic images (magnified 20-50 times) and on-site environmental sensor data (temperature, humidity, and dynamic groundwater level data).

[0019] Intelligent processing layer (upgraded cloud platform): Multi-source data fusion module: Based on a deep learning model (CNN-LSTM hybrid model), it fuses retrieval records (qualitative data), well logging data (quantitative data), and adjacent borehole data (spatial correlation data) to output the formation stratification probability distribution; Adaptive rule engine: Built-in geological knowledge base (including 100+ industry standards, 50+ regional geological features, and 30+ lithological gradation models), automatically matches basic merging rules based on scene tags, and dynamically adjusts parameter thresholds in combination with stratigraphic probability distribution; Complex geological body identification module: Based on feature engineering, 12 types of features are extracted, such as "abrupt core recovery rate, sharp drop in RQD value, abnormal fluctuation of logging curve, and abnormal processing mark of film record". The module automatically identifies special geological bodies such as faults, interlayers, and karst caves through gradient boosting tree model. Closed-loop self-optimization module: records manual verification and correction traces, and iteratively optimizes merging rules and identification model parameters through reinforcement learning algorithms; Data traceability module: Establishes a full-link traceability chain of "raw data - fusion results - hierarchical description - verification and correction", and supports reverse parameter query.

[0020] Interactive verification layer (upgraded APP + Web): Intelligent auxiliary verification interface: Visually displays the comparison of three columns: "original footage data + multi-source verification data + stratified description results", automatically marks contradictory data (such as mismatch between stratification boundaries and logging anomalies) and pushes correction suggestions; Special geological body description template library: Provides standardized description templates for different special geological bodies (e.g., faults need to include attitude, filling material, and influence zone range, and karst caves need to include size, filling rate, and stability evaluation). Multi-format output module: Supports automatic generation of Word / PDF format stratigraphic description reports according to target industry standards, and supports export of BIM model data.

[0021] (II) Core Technology Process: Multi-source data acquisition and preprocessing: The APP collects logging records (including all existing fields), multimedia data (core microscopic images, on-site videos), and environmental sensor data, and imports well logging data and adjacent borehole data through the interface. Users configure project scenario tags (industry, region, standard version), and the APP automatically binds timestamps and GPS location information and uploads them to the cloud platform; The cloud platform preprocesses multi-source data: standardizes and cleans the field of the logging record, reduces noise in the logging data, and aligns the spatial coordinates of adjacent borehole data.

[0022] Adaptive stratigraphic merging and multi-source validation: The adaptive rule engine calls basic merging rules from the geological knowledge base based on scene tags (e.g., the railway industry prioritizes setting the rock and soil naming matching threshold according to the "Railway Engineering Geological Investigation Specification" GB50021-2001). The multi-source data fusion module inputs the preprocessed retrieval records (qualitative features) and well logging data (quantitative features) into the CNN-LSTM model and outputs the probability distribution of formation categories at each depth. The engine dynamically adjusts the merging parameters by combining "basic rules + probability distribution": if the soil and rock parameters of a certain section of the traverse record change gradually (such as the humidity changing from "slightly wet" to "wet", and the probability distribution shows that it is the same stratum), the gradual description rule is automatically triggered to generate a layered description of "silty clay, slightly wet to wet, medium density"; if the probability distribution shows that there are multiple stratum probability peaks in a certain merging section, it is automatically split and marked for verification.

[0023] Intelligent identification and description of complex geological bodies: The complex geological body identification module extracts anomalous features from the back-tracking records (such as a sudden drop in core recovery rate from 85% to 30%) and anomalous fluctuations in well logging data (such as abrupt changes in resistivity), and identifies special geological body types (faults / interlayers / caves, etc.) through a gradient boosting tree model. The system calls the corresponding special geological body description template, and automatically generates a standard description (such as "fault fracture zone, burial depth 12.3-13.1m, main filling material is silty clay mixed with gravel, core recovery rate 32%, RQD=15%, poor stability") by combining the collected multimedia data and parameter data.

[0024] Intelligent Assisted Verification and Closed-Loop Optimization: The app / web platform displays a three-column comparison interface, automatically marks contradictory data (such as inconsistencies between the end depth of stratification and the abnormal depth of well logging), and pushes correction suggestions based on the geological knowledge base; After user verification and adjustment, the system will feed back the correction traces (such as merging threshold adjustment and geological body type correction) to the closed-loop self-optimization module, and improve the accuracy of subsequent generation by updating model parameters and merging rules through reinforcement learning; Once the verification is successful, the cloud platform generates a stratigraphic description report that conforms to the target industry standards, supporting export, printing, and BIM data integration.

[0025] (III) Key Innovative Algorithms and Rules: Multi-source data fusion CNN-LSTM model: Input layer: 12 types of qualitative features (encoded as vectors) such as soil and rock naming and state from the logging record; 5 types of quantitative features (time series) such as resistivity and sonic velocity from the well logging data; and formation boundaries of adjacent boreholes (spatial features). Convolutional layer (CNN): Extracts local anomalies in well logging data (such as resistivity abrupt change intervals); Long Short-Term Memory (LSTM): Captures the temporal correlation features of the retracement records (such as the gradual trend of parameter changes); Output layer: Outputs the probability values ​​of different strata categories corresponding to each depth segment (e.g., "silty clay" probability 0.92, "clay" probability 0.07).

[0026] Unified stratigraphic merging rules: Automatically merge soil and rock descriptions that have the same soil and rock name, condition, and density in the back-tracking records; The starting depth of the merged stratigraphic description is the starting depth of the first soil and rock description recorded in the previous traverse. The end depth of the stratigraphic description after merging is the end depth of the last soil and rock description in the previous back-tracking record before merging. The information on soil type, soil classification, color, humidity, inclusions, and soil layer description is based on the corresponding information in the first soil description recorded in the previous measurement.

[0027] Gradual stratigraphic merging rules: Define parameter similarity thresholds (dynamically adjusted, such as a humidity similarity threshold of 0.8 for the railway industry and 0.75 for the construction industry). If the soil and rock names recorded in consecutive traverses are consistent, and the similarity of the core parameters (state, moisture, density) is higher than the threshold, and the variance of the stratigraphic probability distribution output by the model is less than 0.1, then it is determined to be a gradually changing stratum. The merged description uses a "parameter gradient expression" (e.g., "humidity: slightly humid → humid"), with the starting depth taken from the first record and the ending depth taken from the last record.

[0028] Special geological body identification feature set: Key features: abrupt change in core recovery rate (Δ≥50%), abnormal range of RQD value (<30%), abrupt change in logging curve slope (|k|≥2), abnormal processing markers in the film footage (such as "core breakage"), and microscopic image texture features of the core (fragmentation, porosity). Model training: A gradient boosting tree model was trained based on 100,000+ borehole data (including annotations of special geological bodies), with an accuracy of ≥92%.

[0029] This embodiment overcomes the limitations of single-source data by fusing multi-source data, achieving dual verification of "qualitative + quantitative": it introduces multi-source quantitative / spatial data such as well logging, nearby boreholes, and ground-penetrating radar, and uses CNN-LSTM model to fuse and analyze the data, thus solving the problem of "the accuracy of description depends on a single data and the reliability is insufficient under complex geological conditions" in the original technology.

[0030] This embodiment uses an adaptive rule engine to adapt to multiple scenarios and overcomes the bottleneck of "rigid rules": by building a geological knowledge base and a dynamic threshold adjustment mechanism, it realizes an adaptive process of "scenario labeling → rule matching → parameter optimization", and especially designed exclusive merging rules for gradually changing strata.

[0031] This embodiment fills a technological gap by intelligently identifying complex geological bodies and breaks through the limitations of "conventional strata": it constructs a dedicated identification model and description template library through feature engineering to achieve automatic identification and standardized description of special geological bodies.

[0032] This embodiment achieves a virtuous cycle of "generation-verification-iteration" through a closed-loop self-optimization mechanism, breaking through the limitations of "static rules": by using reinforcement learning, manual correction traces are transformed into the driving force for model iteration, realizing the dynamic optimization of rules and models.

[0033] Specific application scenario 1: A railway survey project, located in the hilly area of ​​East China, based on the "Code for Geological Survey of Railway Engineering" GB50021-2001; Data acquisition: The APP collects the footage records (15 in total, depth 0-30m), resistivity logging data, layer data of two adjacent boreholes, and core micro images; Adaptive merging: The system calls the basic rules of the railway industry. The CNN-LSTM model analysis found that the "humidity" of the 8-12m depth section of the retrieval record gradually changed from "slightly wet" to "wet". The parameter similarity is ≥0.85 and the probability distribution variance is 0.08. It is determined to be a gradually changing stratum and is automatically merged to generate "silty clay, 8.0-12.0m, slightly wet → wet, medium density, containing a small amount of quartz sand". Verification optimization: The system has no contradictory data markings, and users can directly confirm the upload. The closed-loop module records the parameter thresholds for this gradient scene, which can be automatically reused in subsequent similar projects.

[0034] Specific Application Scenario 2: Identification of Karst Caves in Building Engineering Project Scenario: A building survey project located in the karst region of South China, based on the standard GB50007-2011 "Code for Investigation of Building Foundation". Data acquisition: The APP acquired the core recovery record (the core recovery rate in the 18-19.5m depth section dropped sharply from 78% to 25%, RQD=10%), sonic logging data (the sonic velocity in this section decreased abruptly), and core micro images (showing a large number of pores). Special geological body identification: The system extracts the above-mentioned abnormal features, determines them as "karst caves" through the gradient boosting tree model, calls the karst cave description template, and generates "karst cave, burial depth 18.0-19.5m, cave diameter 1.5m, filling rate 30%, filling material is silty clay, core recovery rate 25%, poor stability" by combining the collected data; Verification optimization: After the user verifies the core image and logging data, the closed-loop module updates the characteristic parameters of the cave (core recovery rate mutation rate, sonic velocity threshold) to the model, improving the accuracy of subsequent cave identification.

[0035] Although the above embodiments have described the concept and embodiments of the present invention in detail with reference to the accompanying drawings, those skilled in the art will recognize that various improvements and modifications can still be made to the present invention without departing from the scope of the claims, and therefore will not be elaborated here.

Claims

1. A method for generating stratigraphic descriptions based on intelligent merging of film reel records, comprising a mobile terminal program and a cloud platform, wherein the mobile terminal program contains film reel records and stratigraphic descriptions, the mobile terminal program and the cloud platform engage in data interaction, and the cloud platform receives the film reel records and stratigraphic descriptions sent by the mobile terminal program, characterized in that: The mobile terminal program serves as the data acquisition layer, which is used for the acquisition of multi-source geological data, scene configuration, and data uploading. The cloud platform includes an intelligent processing layer and an interactive verification layer. The intelligent processing layer receives data uploaded by the data acquisition layer, generates standardized stratigraphic description data through multi-source data fusion, adaptive rule matching, complex geological body identification, and closed-loop self-optimization processing, and pushes the processing results to the interactive verification layer. The interactive verification layer receives the processing results from the intelligent processing layer, provides a visual verification interface and correction function, feeds back user correction traces to the intelligent processing layer, and supports multi-format output and export of stratigraphic description results. The intelligent processing layer automatically merges soil and rock descriptions with consistent soil and rock names, states, and densities from the back-tracking records. The starting depth of the merged strata layer description is the starting depth of the first soil and rock description in the back-tracking records before the merger. The ending depth of the merged strata layer description is the ending depth of the last soil and rock description in the back-tracking records before the merger. Soil and rock categories, classifications, colors, moisture, inclusions, and soil layer descriptions are all based on the corresponding content in the first soil and rock description in the back-tracking records before the merger.

2. The method for generating stratigraphic description based on intelligent merging of back-scale records according to claim 1, characterized in that: The mobile terminal program includes a basic data acquisition module, a multi-source data interface module, a scene configuration module, and a multimedia enhanced acquisition module. The basic data acquisition module is equipped with a return-to-the-base record input interface. The multi-source data interface module has a built-in data docking protocol, supporting the import or real-time docking of well logging data (including quantitative data such as resistivity and acoustic velocity), ground-penetrating radar data, and adjacent borehole data. The scene configuration module is used for users to configure project scene tags. The multimedia enhanced acquisition module integrates an image acquisition unit, a video acquisition unit, and a sensor data receiving unit, supporting the capture and acquisition of core microscopic images and on-site videos, as well as the reception of environmental sensor data such as temperature, humidity, and groundwater level dynamics.

3. The method for generating stratigraphic description based on intelligent merging of back-scale records according to claim 1, characterized in that: The intelligent processing layer includes a hardware support unit, a multi-source data fusion module, an adaptive rule engine, a complex geological body identification module, a closed-loop self-optimization module, and a data traceability module.

4. The method for generating stratigraphic description based on intelligent merging of back-scale records according to claim 3, characterized in that: The adaptive rule engine's parameter threshold dynamic adjustment unit is configured with a gradient formation determination subunit, and the determination logic of the subunit is as follows: (1) Preset parameter similarity thresholds: the humidity similarity threshold for the railway industry is set to 0.8, and the humidity similarity threshold for the construction industry is set to 0.75; (2) When the soil and rock names of the continuous back-track records are consistent, the similarity of the core parameters is higher than the corresponding threshold, and the variance of the stratigraphic probability distribution output by the CNN-LSTM model is less than 0.1, it is determined to be a gradually changing stratum. (3) The merging description of the gradient strata adopts the "parametric gradient description". The starting depth of the merged strata is associated with the starting depth of the first backtrack record, and the ending depth is associated with the ending depth of the last backtrack record.

5. The method for generating stratigraphic description based on intelligent merging of back-scale records according to claim 1, characterized in that: The interactive verification layer includes an APP-side interactive unit and a Web-side interactive unit, both of which communicate with the intelligent processing layer via a network. It includes an intelligent auxiliary verification module, a special geological body description template library module, and a multi-format output module.

6. The method for generating stratigraphic description based on intelligent merging of back-scale records according to claim 1, characterized in that: The mobile terminal programs are all equipped with a timestamp binding unit and a GPS positioning unit, and the data upload module supports the synchronous upload of multimedia materials such as photos and videos as well as structured data.

7. The method for generating stratigraphic description based on intelligent merging of back-scale records according to claim 1, characterized in that: The return footage record includes: drilling method, termination time, starting depth, total drill string length, drill string above-ground length, termination depth, wall protection method, footage per return, total core length, core recovery rate, soil type, soil classification, soil name, color, state, humidity, density, inclusions, RQD, and description.

8. The method for generating stratigraphic description based on intelligent merging of back-scale records according to claim 1, characterized in that: The stratigraphic description includes: starting depth, ending depth, soil type, soil classification, soil name, color, state, moisture content, density, inclusions, and soil layer description.

9. The method for generating stratigraphic description based on intelligent merging of back-scale records according to claim 1, characterized in that: A single tracking record may contain one or more stratigraphic layers, and a stratigraphic layer may contain one or more tracking records.

10. The method for generating stratigraphic description based on intelligent merging of back-scale records according to claim 1, characterized in that: The data in the tracking record and stratigraphic description share the same information, including soil type, soil classification, soil name, color, state, humidity, density, inclusions, and soil layer description. The tracking information is retained as the original exploration record.

Images