Intelligent design system and method for coal-pillar-free self-forming roadway mining

Abstract

The invention relates to the field of coal mining design, and discloses an intelligent design system and method for coal-pillar-free self-forming roadway mining, and the system comprises a geological analysis module, a parameter design module, an analogy check module and an output module. The geological analysis module is used for screening and processing mine engineering geological data to obtain geological parameters required by design; the parameter design module is used for designing key mining parameters based on the geological parameters; the analogy checking module is used for checking, optimizing and adjusting key mining parameters by using historical engineering case big data; and the output module is used for completing mining scheme design and outputting a design report and a drawing based on the optimized and adjusted key mining parameters. Compared with a traditional design method depending on artificial experience, the method has the advantages that design efficiency is greatly improved, personal errors are reduced, meanwhile, complex and changeable geological conditions can be accurately adapted, a safer, more economical and higher-operability mining design scheme is generated, and the standardization level of design is further improved.

Description

Technical Field

[0001] This invention relates to the field of coal mine design, specifically to an intelligent design system and method for pillarless self-forming roadway mining. Background Technology

[0002] For a long time, coal mining has mainly relied on the longwall mining method 121. This method requires the retention of coal pillars to maintain roadway safety, which has problems such as low resource recovery rate, large roadway excavation volume, and stress concentration in coal pillars, which can easily lead to hidden dangers such as large deformation of surrounding rock and uneven surface subsidence.

[0003] In recent years, the pillarless self-forming roadway mining technology (110 / N00 method), which is based on the "top-cutting short-arm beam" theory, has been successfully applied in many mining areas. It achieves automatic roadway formation and avoids leaving coal pillars by actively cutting off the stress transmission of the roof through directional top cutting. It has advantages such as high resource recovery rate, less roadway excavation, and less environmental disturbance.

[0004] However, the current mining design of this method still relies on manual experience and field tests by geological technicians, resulting in problems such as low design efficiency, poor repeatability, and insufficient promotion effectiveness, making it difficult to meet the requirements of efficient, intelligent, and green mining in modern coal mines. Therefore, there is an urgent need for an intelligent design system and method for pillarless self-forming roadway mining, which can realize automatic analysis of geological data, intelligent design and analog optimization of key parameters, and ultimately automatically output the optimal design scheme and result documents. Summary of the Invention

[0005] This invention addresses the problems of heavy reliance on manual experience, low design efficiency, and insufficient promotion effectiveness in the design process of existing pillarless self-forming roadways (110 / N00 method). It proposes an intelligent design method and system that combines AI intelligent analysis of geological data, automatic parameter design, and big data comparison and optimization. This method supports the efficient and repeatable generation of optimal pillarless self-forming roadway mining schemes under different geological conditions in different mining areas, and promotes the intelligent, green, and safe development of coal mines.

[0006] To achieve the above objectives, the present invention provides an intelligent design system for pillarless self-forming roadway mining, comprising: a geological analysis module, a parameter design module, an analogy verification module, and an output module;

[0007] The geological analysis module is used to screen and process mine engineering geological data to obtain the geological parameters required for the design.

[0008] The parameter design module is used to design key mining parameters based on the geological parameters;

[0009] The analogy verification module is used to verify and optimize the key mining parameters using big data from historical engineering cases;

[0010] The output module is used to complete the design of the mining scheme and output the design report and drawings based on the optimized and adjusted key mining parameters.

[0011] Preferably, the geological analysis module cleans and standardizes the imported mining area engineering geological data, including removing redundant information, unifying measurement units, converting coordinate systems, and extracting key data.

[0012] The geological analysis module utilizes a geological feature recognition model constructed using deep learning algorithms to automatically extract key geological parameters; the structure of the geological feature recognition model includes: an input layer, a hidden layer, and an output layer;

[0013] The input layer is used to input engineering geological data of the mining area, which can be in the form of text and drawings, and includes geological exploration reports, borehole columnar sections, preliminary mine design, operating procedures, and geological specifications.

[0014] Hidden layers use LSTM or Transformer structures to extract geological features;

[0015] The output layer outputs geological parameters, including tunnel parameters and strata parameters.

[0016] Preferably, the key mining parameters include: roof cutting, roof support, temporary support, and rock retaining support;

[0017] The parameters of the top plate cut include the cut angle and the cut height. The cut angle is determined as follows: when the hardness coefficient of the top plate is less than 2, the cut angle is 10°; when the hardness coefficient of the top plate is 2-4, the cut angle is 15°; when the hardness coefficient of the top plate is greater than 4, the cut angle is 20°. The cut height H0 is determined as follows:

[0018] H0 = f1 × k × h;

[0019] Where f1 represents the mining height influence coefficient; k represents the roof influence coefficient; and h represents the mining height.

[0020] Preferably, the parameters of the top plate support include anchor cable length and anchor cable spacing;

[0021] The method for determining the anchor cable length L is as follows: when the cut height H0 < 15m, L = H0 + 2m; when the cut height H0 ≥ 15m, the anchor cable length is taken as 17m.

[0022] The method for determining the spacing between anchor cables is as follows: when the hardness coefficient of the top plate is >4, the spacing between anchor cables is 1.6m-2.0m, and the number of anchor cables in each row is greater than or equal to 2; when the hardness coefficient of the top plate is 2-4, the spacing between anchor cables is 1.0m-1.6m, and the number of anchor cables in each row is greater than or equal to 3; when the hardness coefficient of the top plate is <2, the spacing between anchor cables is less than 1.0m, and the number of anchor cables in each row is greater than or equal to 3.

[0023] Preferably, the key parameters of the rock retaining support include the rock retaining support type and the support spacing;

[0024] All rock retaining walls are supported by a combination of metal mesh and U-shaped steel beams. The size of the metal mesh is determined according to the roadway height, and the spacing between the U-shaped steel beams is 400mm-500mm. For mining heights less than 2m, 25U beams are used; for mining heights between 2m and 4m, 29U beams are used; and for mining heights greater than 4m, 36U beams are used. Each set of U-shaped steel beams consists of two overlapping beams, with the overlap length being 1 / 3 of the roadway height.

[0025] Preferably, the workflow of the analogy verification module includes:

[0026] Construct a large engineering database, which is a distributed database architecture;

[0027] Based on the aforementioned key mining parameters, similar geological condition cases are identified in the engineering database, and a multi-attribute similarity evaluation model is established.

[0028] Based on the aforementioned multi-attribute similarity evaluation model, a key indicator analogy evaluation system is established;

[0029] The key mining parameters are optimized and adjusted using the key indicator analogy evaluation system.

[0030] Preferably, the workflow of the output module includes:

[0031] Output several alternative design schemes, establish an evaluation system that includes indicators of economy, safety and operability, and use the analytic hierarchy process (AHP) to determine the optimal scheme.

[0032] Automatically generate design reports that conform to industry standards and output parametric CAD design drawings;

[0033] Develop a detailed monitoring plan, including monitoring equipment, monitoring point layout, monitoring frequency, etc.

[0034] This invention also provides an intelligent design method for pillarless self-forming roadway mining, the method being applied to the above-mentioned system, and the steps include:

[0035] S1. Screen and process the engineering geological data of the mine to obtain the geological parameters required for the design;

[0036] S2. Based on the geological parameters, design key mining parameters;

[0037] S3. Use big data from historical engineering cases to verify and optimize the key mining parameters;

[0038] S4. Based on the optimized and adjusted key mining parameters, complete the mining scheme design and output the design report and drawings.

[0039] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0040] Compared to traditional design methods that rely on manual experience, this system significantly improves design efficiency, reduces human error, and can accurately adapt to complex and changing geological conditions, generating safer, more economical, and more operable mining solutions. Its automatically generated design reports and CAD drawings conform to industry standards, further enhancing the standardization of design and providing strong support for the intelligent, green, and safe development of the coal mining industry. Attached Figure Description

[0041] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a schematic diagram of the process framework of an embodiment of the present invention. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0045] Example 1:

[0046] This embodiment provides an intelligent design system for pillarless self-forming roadway mining, including: a geological analysis module, a parameter design module, an analogy verification module, and an output module. The following will explain in detail how this embodiment solves practical technical problems.

[0047] First, the geological analysis module is used to process the mine engineering geological data to obtain the geological parameters required for the design.

[0048] (1) Data Import: The geological analysis module supports the import of geological data in various formats, including geological exploration reports, borehole columnar sections, mine design reports, coal mine safety operation specifications, and other technical documents in PDF, CAD, Word, and Excel formats. This module has the function of automatically recognizing document structure (such as recognizing descriptions of roadway cross-sections, shapes, dimensions, and engineering geological parameters in Word documents, corresponding roadway support parameters in CAD documents, and rock strata mechanical parameters in Excel spreadsheets), and can accurately distinguish between different content formats such as text, tables, and drawings.

[0049] (2) Data Preprocessing: The imported raw data is cleaned and standardized, including removing redundant information, unifying units of measurement, coordinate system conversion, and extracting key data. This module uses natural language processing technology to segment the text content, identify entities, and extract relationships (using BERT or SpaCy models for entity identification, mainly including the extraction and identification of keyword content, such as "coal seam thickness = 4.0m"). Relationship extraction: establishing parameter associations (such as "roof lithology = sandstone → compressive strength ≥ 50MPa").

[0050] (3) Intelligent Analysis: A geological feature recognition model built using deep learning algorithms automatically extracts key geological parameters. The model structure includes: an input layer, a hidden layer, and an output layer;

[0051] Input layer: consists of text feature input units and image feature input units.

[0052] The text feature input unit is used to receive text information from engineering geological data of the mining area, such as geological exploration reports, borehole columnar section descriptions, preliminary mine design documents, operating procedures, geological specifications, etc.; the text is converted into vector representation through word segmentation and word embedding (Word2Vec / BERT embedding).

[0053] The image feature input unit is used to receive image data such as CAD drawings and scanned borehole columnar diagrams, and extracts spatial structure and symbol features through a convolutional neural network (CNN).

[0054] Hidden layer: Employs a dual-channel fusion structure, wherein:

[0055] The text channel uses a bidirectional long short-term memory network (BiLSTM) to extract time series and contextual features.

[0056] The image channels use a convolutional neural network (CNN) to extract spatial structural features and local patterns.

[0057] In the fusion stage, a multi-head attention mechanism is introduced to achieve deep fusion of text and image features, and feature mapping and dimensionality reduction are performed through a fully connected layer.

[0058] Output layer: Consists of multi-task regression and classification units, which output structured geological parameters, including:

[0059] Tunnel parameters: cross-sectional shape, dimensions, slope, etc.; support parameters: existing support type, material specifications, support density, etc.

[0060] Rock strata parameters: lithological composition, thickness distribution, mechanical parameters, mining height, breccia coefficient, joint development, etc.

[0061] (4) Output: The output results are provided in standardized JSON or table format for direct use by the subsequent parameter design module.

[0062] Then, the parameter design module is used to design key mining parameters based on geological parameters.

[0063] (1) Design of roof slab cuts. The key parameters of roof slab cuts include cut angle and cut height.

[0064] The method for determining the cutting angle is as follows: when the hardness coefficient of the top plate is less than 2, the cutting angle is determined to be 10°; when the hardness coefficient of the top plate is 2-4, the cutting angle is determined to be 15°; when the hardness coefficient of the top plate is greater than 4, the cutting angle is determined to be 20°.

[0065] Based on parameters such as rock mechanics, mining height, and fragmentation coefficient, the optimal cut height H0 is determined:

[0066] H0 = f1 × k × h;

[0067] Where f1 represents the influence coefficient of mining height (f1 = 2 when mining height < 2m, f1 = 2.5 when mining height 2m-4m, f1 = 3 when mining height > 4m); k represents the influence coefficient of roof (k = 0.8 when roof hardness coefficient < 2; k = 1 when roof hardness coefficient is 2-4; k = 1.2 when roof hardness coefficient > 4); h represents the mining height.

[0068] In addition, the drilling angle for directional pre-splitting cuts in the roof should be determined based on the working face height and comply with the following regulations:

[0069] When H c When ≤1m, β=15°~20°;

[0070] When H c When the value is greater than 1m, β = 10° to 15°.

[0071] Where β is the angle between the pre-splitting borehole and the vertical line; HC To raise the working face.

[0072] The calculation method for the height of the directional pre-splitting cut in the top slab is as follows:

[0073] H F =(H C -ΔH1-ΔH2) / (K-1);

[0074] Among them, H F ΔH1 represents the height of the pre-splitting cut; ΔH2 represents the amount of subsidence of the stope roof; and K represents the coefficient of breakage and expansion of the stope roof.

[0075] The influence of different cutting angles on rock strata collapse was analyzed to determine the optimal cutting angle. Considering the tensile strength and fracture development of the top strata, parameters such as charge density, aperture, and hole spacing were calculated to determine the optimal charge parameters.

[0076] (2) Roof support design: Based on the roof cutting design, a short-arm beam mechanical model is established. The forces acting on the model in the vertical direction are: rock mass gravity G, which is vertically downward and acts equivalently at the centroid; constant resistance anchor cable support load f; and the supporting force of the actual coal mass on the structure can be simplified to a linearly varying load σ. x The residual stress σ in the coal seam at the roadway c The reaction force σ0 generated at the fixed end of the coal body. Based on the above model and boundary forces, the roof support length and support strength can be obtained through static equilibrium calculations, thereby confirming the support form to be adopted and recommending the optimal support combination scheme according to geological conditions and mining technology.

[0077] The design of roof support parameters revolves around the stability of the "pressure relief structure," requiring calculation of the constant-resistance anchor cable support strength based on a mechanical model. Key input parameters include roadway width b, average unit weight γ of the overlying strata, burial depth H, coal body cohesion c, and internal friction angle. Lateral pressure coefficient λ and coal seam support strength P.

[0078] The width x0 of the ultimate stress equilibrium zone of the coal seam is calculated using the following formula:

[0079]

[0080] Then, combine the following two formulas to calculate the required anchor cable support strength f per unit length of roadway:

[0081]

[0082] Where k represents the stress concentration factor; σ c σ represents the residual stress at the edge of the coal seam; G represents the self-weight of the "pressure relief structure"; M represents the geometric parameters; θ represents the fracture angle of the roof strata; σ0 represents the peak value of the support force of the coal body on the "pressure relief structure".

[0083] Key parameters for roof support include anchor cable length and anchor cable spacing. The anchor cable length is determined as follows: when the cut height H0 < 15m, L = H0 + 2m; when the cut height H0 ≥ 15m, the anchor cable length is 17m. The anchor cable spacing is determined as follows: when the roof hardness coefficient > 4, the anchor cable spacing is 1.6m-2.0m, with at least 2 anchor cables per row; when the roof hardness coefficient is 2-4, the anchor cable spacing is 1.0m-1.6m, with at least 3 anchor cables per row; when the roof hardness coefficient < 2, the anchor cable spacing is less than 1.0m, with at least 3 anchor cables per row.

[0084] (3) Temporary support design: Considering the different collapse stages of the roof after the working face is mined, temporary support calculations are carried out before the overlying rock behind the working face stabilizes. The strength of the temporary support for the overlying rock is calculated, and the distance of the temporary support is determined according to the distance of the overlying rock stabilization.

[0085] Temporary reinforced support design needs to match the three-stage characteristics of roadway surrounding rock deformation: the pressure relief influence period (0–40m), the dynamic pressure deformation period (40m–160m), and the compaction stabilization period (after 160m). During the dynamic pressure deformation period, the surrounding rock undergoes significant deformation due to the compression of the roof fractured rock blocks, requiring the installation of temporary supports (such as roof cutting and side support supports) that also have the functions of "pressure control" and "pressure relief".

[0086] Key parameters for temporary support include support strength and support spacing. The methods for determining temporary support strength and support spacing are as follows: if a single column + π-beam support method is used, it is arranged in a "one beam, five columns" pattern with a row spacing of 0.8m-1.0m; if a top-cutting and side-supporting support + π-beam support method is used, it is arranged in a "paired" pattern with a row spacing of 2.4m-3.0m.

[0087] (4) Design of rock retaining wall support: To prevent rock from the goaf from flowing into the roadway, rock retaining wall support is required on the goaf side. In this embodiment, the discrete element method is used for simulation analysis to calculate the stress of the rock retaining structure, optimize the support density and form, and determine the rock retaining wall support strength, support spacing, and support form. The specific steps are as follows:

[0088] Geometric model establishment: Based on the measured cross-sectional dimensions of the goaf and roadway, construct the geometric model of the goaf, roadway and surrounding rock in 3D modeling software, and determine the boundary conditions and initial state.

[0089] Particle parameter assignment: Based on geological exploration data and indoor test results, parameters such as particle size distribution, density, elastic modulus, internal friction angle and cohesion of gangue particles are set; the particle shape adopts the multi-sphere approximation method to ensure the authenticity of particle contact and movement in the simulation.

[0090] Collapse process simulation: Apply gravity field and surrounding rock constraints to the goaf in numerical simulation software (such as PFC, UDEC), gradually remove roof support, and trigger gangue collapse; use time history analysis to record the movement trajectory, velocity and contact force changes of gangue.

[0091] Stress and stability analysis: After the collapse and accumulation stabilizes, the distribution of normal and tangential forces in the contact area between the gangue and the retaining structure is extracted, and the maximum bearing capacity requirement of the retaining structure is calculated; combined with the surrounding rock pressure analysis, the support strength coefficient is determined.

[0092] Support scheme optimization: Based on the stress analysis results, the finite element method was used to verify the impact of different support spacing and support types (steel frame, single column, infill wall, etc.) on the rock retaining effect, and the combination scheme with a safety factor ≥1.5 and the best economic efficiency was selected.

[0093] Outputs include design tables for support strength, spacing, and type, as well as stress distribution diagrams and gangue accumulation diagrams for reference during construction and subsequent monitoring.

[0094] Key parameters for rock retaining support include the support type and spacing. Rock retaining support uses a "metal mesh + U-shaped steel" method. The size of the metal mesh is determined according to the roadway height, and the spacing of the U-shaped steel is 400mm-500mm (the same as the spacing of the slotted boreholes). For mining heights less than 2m, 25U type is used; for mining heights of 2m-4m, 29U type is used; and for mining heights greater than 4m, 36U type is used. Each set of U-shaped steel consists of two overlapping sections, with an overlap length of 1 / 3 of the roadway height.

[0095] Subsequently, the analogy verification module optimizes and adjusts key mining parameters using historical engineering cases.

[0096] (1) Construction of the engineering big data database and design of a distributed database architecture. The distributed database architecture designed in this embodiment adopts a multi-level technology combination to support the efficient operation of the intelligent design system for pillarless self-forming roadway mining.

[0097] The core architecture includes a data access layer (processing multi-source data such as PDF / CAD / Word), a distributed storage layer (PostgreSQL sharding and table storage of structured parameters, MongoDB management of unstructured cases, and Redis caching of high-frequency data), a computing layer (Spark / Flink implementation of KNN similarity analysis and real-time monitoring and processing), and accelerated retrieval through Elasticsearch.

[0098] This architecture features horizontal scalability, employs the Raft protocol to ensure consistency, and combines with Kafka to implement a real-time monitoring data pipeline, providing reliable data support for intelligent design, supporting the storage of multiple engineering data, and establishing a standardized database management system.

[0099] (2) Intelligent screening algorithm automatically identifies historical cases with similar geological conditions to the target mining area in the engineering database, and performs accurate screening and sorting based on the multi-attribute similarity evaluation model.

[0100] ① Feature system construction:

[0101] The feature system consists of geological parameter features (such as coal seam thickness T). c The characteristics of coal seam dip angle α, burial depth H, roof compressive strength σ1, roof joint development index J, etc., and engineering parameters (such as cut height H0, support strength S) are also important. s Support density D s S-shaped rock retaining strength g It consists of two parts (etc.).

[0102] All features are normalized:

[0103]

[0104] Where, X′ i Represents the normalized features; X i This indicates the selected feature.

[0105] ② Feature dimensionality reduction:

[0106] An improved t-SNE-PCA hybrid dimensionality reduction algorithm is adopted, firstly extracting principal components Z using PCA. k (Retaining a cumulative variance contribution rate of ≥95%), the principal components are then input into the t-SNE model to obtain a low-dimensional manifold embedding representation Y, in order to preserve the global structure and local neighborhood relationships.

[0107] After dimensionality reduction, the dimension of the feature vector is reduced from the original n to m (m << n).

[0108] ③ Similarity calculation model:

[0109] A weighted composite distance metric model is adopted:

[0110]

[0111] Where: w j A represents the weight of the j-th feature; j B j Let represent the j-th feature value of two samples A and B after dimensionality reduction.

[0112] ④ Search Architecture:

[0113] First layer (coarse screening): Elasticsearch quickly matches candidate cases within a specified feature range based on the inverted index.

[0114] The second layer (refinement): Spark distributed computing performs weighted composite distance calculations and sorts the results by similarity.

[0115] The third layer (dynamic threshold adjustment): The similarity threshold θ1 is dynamically adjusted based on the historical matching accuracy. When Sim≥θ1, it is judged as a highly similar case.

[0116] ⑤ Online learning mechanism:

[0117] The system receives subsequent new engineering cases and actual mining feedback results, and assigns weights w to them. j The threshold θ1 is iteratively updated to achieve adaptive optimization, thereby improving matching accuracy and generalization ability.

[0118] ⑥ Output results:

[0119] Output the top K historical cases with the highest similarity, along with their corresponding geological parameters, engineering parameters, and implementation effects, and provide input data for subsequent analogy optimization.

[0120] (3) Parameter analogy analysis: Establish a key indicator analogy evaluation system. Geological condition analogy parameters include coal seam thickness, coal seam dip angle, coal seam burial depth and roof conditions, etc. Design analogy parameters include roof cutting, roof support, temporary support and rock retaining support, etc.

[0121] The key indicator analogy evaluation system constructed in this embodiment adopts a combination of multi-dimensional feature fusion and dynamic weight allocation to comprehensively evaluate different geological conditions and engineering design parameters.

[0122] The key to multi-attribute similarity evaluation lies in determining the similarity coefficient M; the similarity coefficient M = mining height coefficient A × burial depth coefficient B × roof hardness coefficient C × coal seam dip angle coefficient D. The closer M is to 1, the higher the similarity. The top 3 samples with the highest similarity are used to compare and verify the design parameters. The difference between each design parameter value and the sample parameters should not exceed 20%; otherwise, manual confirmation and correction are required.

[0123] The method for determining the sampling height coefficient is as follows: if the design target sampling height is greater than or equal to the sample sampling height, then A = design target sampling height / sample sampling height; if the design target sampling height is less than the sample sampling height, then A = sample sampling height / design target sampling height.

[0124] The method for determining the burial depth coefficient B is as follows: if the design target burial depth is greater than or equal to the sample burial depth, then B = design target burial depth / sample burial depth; if the design target burial depth is less than the sample burial depth, then B = sample burial depth / design target burial depth.

[0125] The method for determining the roof hardness coefficient C is as follows: if the design target roof compressive strength is greater than or equal to the sample roof compressive strength, then C = design target roof compressive strength / sample roof compressive strength; if the design target roof compressive strength is less than the sample roof compressive strength, then C = sample roof compressive strength / design target roof compressive strength.

[0126] The method for determining the coal seam dip angle coefficient D is as follows: if the design target coal seam dip angle is greater than or equal to the sample coal seam dip angle, then D = design target coal seam dip angle / sample coal seam dip angle; if the design target coal seam dip angle is less than the sample coal seam dip angle, then D = sample coal seam dip angle / design target coal seam dip angle.

[0127] The final output module, based on the optimized and adjusted key mining parameters, completes the mining scheme design and outputs a design report and drawings.

[0128] (1) Multidimensional feature fusion:

[0129] The system first standardizes the geological conditions (such as coal seam thickness, coal seam dip angle, roof strength, surrounding rock integrity, burial depth, etc.) and engineering parameters (such as cut height, support density, support strength, cost coefficient, construction period, etc.) into a unified dimension.

[0130] Standardization not only ensures that indicators with different dimensions can be compared within the same calculation framework, but also facilitates subsequent weight allocation and similarity calculation.

[0131] (2) Weighting and Evaluation Model:

[0132] In terms of weight allocation, the Analytic Hierarchy Process (AHP) is used to construct a three-level indicator system of "objective-criteria-indicators". The relative importance of each core indicator is determined through expert scoring, pairwise comparison and other methods.

[0133] In terms of similarity assessment, a fuzzy comprehensive evaluation model is introduced to integrate the membership degrees of different indicators into a unified evaluation matrix, thereby obtaining the overall similarity between the target case and historical cases.

[0134] (3) Dynamic feedback and adaptive optimization:

[0135] To improve the adaptability of the evaluation system under different mining conditions, a dynamic feedback mechanism is introduced:

[0136] When there are discrepancies between historical cases and actual mining results, the system will automatically adjust the weight allocation ratio based on on-site monitoring data (such as roof subsidence, roadway deformation rate, and changes in support stress).

[0137] Regarding the early warning mechanism, users can set risk thresholds. For example, when the roof subsidence exceeds 15mm, the system will automatically issue a safety alarm and suggest adjustments to the design scheme.

[0138] Based on the above evaluation results, the preliminary design parameters are optimized and adjusted, and multiple optional design schemes (usually 3-5 schemes) are generated. A comprehensive comparison is made from multiple dimensions such as economy, safety, and operability. The optimal scheme is selected again through AHP, and a design report that conforms to industry standards is automatically generated.

[0139] It also outputs parametric CAD design drawings and detailed monitoring plans (including monitoring point layout, monitoring frequency, required equipment type and deployment method).

[0140] In this way, the entire evaluation and optimization process relies on both big data-driven similarity analysis and human experience-based weighted judgment, achieving accurate analogy, adaptive optimization, and traceable verification of design solutions.

[0141] The overall system workflow in this embodiment is as follows: Figure 1 As shown.

[0142] Example 2:

[0143] This embodiment also provides an intelligent design method for pillarless self-forming roadway mining, the steps of which include:

[0144] S1. Screen and process the geological data of the mine engineering to obtain the geological parameters required for the design.

[0145] (1) Data import: Use AI analysis to import geological data in various formats, including technical documents such as geological exploration reports, mine design reports, and coal mine safety operation specifications in PDF, CAD, Word, and Excel formats.

[0146] (2) Data preprocessing: The imported raw data is cleaned and standardized, including removing redundant information, unifying units of measurement, and coordinate system conversion. Natural language processing technology is used to segment the text content, identify entities, and extract relations (using BERT or SpaCy models for entity identification, mainly including the extraction and identification of keyword content, such as "coal seam thickness = 4.0m"). Relation extraction: establishing parameter associations (such as "roof lithology = sandstone → compressive strength ≥ 50MPa").

[0147] (3) Intelligent Analysis: A geological feature recognition model built using deep learning algorithms automatically extracts key geological parameters. The model structure includes: an input layer, a hidden layer, and an output layer;

[0148] Input layer: consists of text feature input units and image feature input units.

[0149] The text feature input unit is used to receive text information from engineering geological data of the mining area, such as geological exploration reports, borehole columnar section descriptions, preliminary mine design documents, operating procedures, geological specifications, etc.; the text is converted into vector representation through word segmentation and word embedding (Word2Vec / BERT embedding).

[0150] The image feature input unit is used to receive image data such as CAD drawings and scanned borehole columnar diagrams, and extracts spatial structure and symbol features through a convolutional neural network (CNN).

[0151] Hidden layer: Employs a dual-channel fusion structure, wherein:

[0152] The text channel uses a bidirectional long short-term memory network (BiLSTM) to extract time series and contextual features.

[0153] The image channels use a convolutional neural network (CNN) to extract spatial structural features and local patterns.

[0154] In the fusion stage, a multi-head attention mechanism is introduced to achieve deep fusion of text and image features, and feature mapping and dimensionality reduction are performed through a fully connected layer.

[0155] Output layer: Consists of multi-task regression and classification units, which output structured geological parameters, including:

[0156] Tunnel parameters: cross-sectional shape, dimensions, slope, etc.; support parameters: existing support type, material specifications, support density, etc.

[0157] Rock strata parameters: lithological composition, thickness distribution, mechanical parameters, mining height, breccia coefficient, joint development, etc.

[0158] (4) Output: The system automatically generates standardized geological parameters and outputs roadway basic parameters and drawings.

[0159] S2. Design key mining parameters based on geological parameters.

[0160] (1) Design of roof slab cuts. The key parameters of roof slab cuts include cut angle and cut height.

[0161] The method for determining the cutting angle is as follows: when the hardness coefficient of the top plate is less than 2, the cutting angle is determined to be 10°; when the hardness coefficient of the top plate is 2-4, the cutting angle is determined to be 15°; when the hardness coefficient of the top plate is greater than 4, the cutting angle is determined to be 20°.

[0162] Based on parameters such as rock mechanics parameters, mining height, and fragmentation coefficient, the optimal cut height H0 is determined:

[0163] H0 = f1 × k × h;

[0164] Where f1 represents the influence coefficient of mining height (f1 = 2 when mining height < 2m, f1 = 2.5 when mining height 2m-4m, f1 = 3 when mining height > 4m); k represents the influence coefficient of roof (k = 0.8 when roof hardness coefficient < 2; k = 1 when roof hardness coefficient is 2-4; k = 1.2 when roof hardness coefficient > 4); h represents the mining height.

[0165] In addition, the drilling angle for directional pre-splitting cuts in the roof should be determined based on the working face height and comply with the following regulations:

[0166] When H c When ≤1m, β=15°~20°;

[0167] When H c When the value is greater than 1m, β = 10° to 15°.

[0168] Where β is the angle between the pre-splitting borehole and the vertical line; H C To raise the working face.

[0169] The calculation method for the height of the directional pre-splitting cut in the top slab is as follows:

[0170] H F =(H C -ΔH1-ΔH2) / (K-1);

[0171] Among them, H F ΔH1 represents the height of the pre-splitting cut; ΔH2 represents the amount of subsidence of the stope roof; and K represents the coefficient of breakage and expansion of the stope roof.

[0172] The influence of different cutting angles on rock strata collapse was analyzed to determine the optimal cutting angle. Considering the tensile strength and fracture development of the top strata, parameters such as charge density, aperture, and hole spacing were calculated to determine the optimal charge parameters.

[0173] (2) Roof support design: Based on the roof cutting design, a short-arm beam mechanical model is established. The forces acting on the model in the vertical direction are: rock mass gravity G, which is vertically downward and acts equivalently at the centroid; constant resistance anchor cable support load f; and the supporting force of the actual coal mass on the structure can be simplified to a linearly varying load σ. x The residual stress σ in the coal seam at the roadway c The reaction force σ0 generated at the fixed end of the coal body. Based on the above model and boundary forces, the roof support length and support strength can be obtained through static equilibrium calculations, thereby confirming the support form to be adopted and recommending the optimal support combination scheme according to geological conditions and mining technology.

[0174] The design of roof support parameters revolves around the stability of the "pressure relief structure," requiring calculation of the constant-resistance anchor cable support strength based on a mechanical model. Key input parameters include roadway width b, average unit weight γ of the overlying strata, burial depth H, coal body cohesion c, and internal friction angle. Lateral pressure coefficient λ and coal seam support strength P.

[0175] The width x0 of the ultimate stress equilibrium zone of the coal seam is calculated using the following formula:

[0176]

[0177] Then, combine the following two formulas to calculate the required anchor cable support strength f per unit length of roadway:

[0178]

[0179] Where k represents the stress concentration factor; σ c σ represents the residual stress at the edge of the coal seam; G represents the self-weight of the "pressure relief structure"; M represents the geometric parameters; θ represents the fracture angle of the roof strata; σ0 represents the peak value of the support force of the coal body on the "pressure relief structure".

[0180] Key parameters for roof support include anchor cable length and anchor cable spacing. The anchor cable length is determined as follows: when the cut height H0 < 15m, L = H0 + 2m; when the cut height H0 ≥ 15m, the anchor cable length is 17m. The anchor cable spacing is determined as follows: when the roof hardness coefficient > 4, the anchor cable spacing is 1.6m-2.0m, with at least 2 anchor cables per row; when the roof hardness coefficient is 2-4, the anchor cable spacing is 1.0m-1.6m, with at least 3 anchor cables per row; when the roof hardness coefficient < 2, the anchor cable spacing is less than 1.0m, with at least 3 anchor cables per row.

[0181] (3) Temporary support design: Considering the different collapse stages of the roof after the working face is mined, temporary support calculations are carried out before the overlying rock behind the working face stabilizes. The strength of the temporary support for the overlying rock is calculated, and the distance of the temporary support is determined according to the distance of the overlying rock stabilization.

[0182] Temporary reinforced support design needs to match the three-stage characteristics of roadway surrounding rock deformation: the pressure relief influence period (0–40m), the dynamic pressure deformation period (40m–160m), and the compaction stabilization period (after 160m). During the dynamic pressure deformation period, the surrounding rock undergoes significant deformation due to the compression of the roof fractured rock blocks, requiring the installation of temporary supports (such as roof cutting and side support supports) that also have the functions of "pressure control" and "pressure relief".

[0183] Key parameters for temporary support include support strength and support spacing. The methods for determining temporary support strength and support spacing are as follows: if a single column + π-beam support method is used, it is arranged in a "one beam, five columns" pattern with a row spacing of 0.8m-1.0m; if a top-cutting and side-supporting support + π-beam support method is used, it is arranged in a "paired" pattern with a row spacing of 2.4m-3.0m.

[0184] (4) Design of rock retaining wall support: To prevent rock from the goaf from flowing into the roadway, rock retaining wall support is required on the goaf side. In this embodiment, the discrete element method is used for simulation analysis to calculate the stress of the rock retaining structure, optimize the support density and form, and determine the rock retaining wall support strength, support spacing, and support form. The specific steps are as follows:

[0185] Geometric model establishment: Based on the measured cross-sectional dimensions of the goaf and roadway, construct the geometric model of the goaf, roadway and surrounding rock in 3D modeling software, and determine the boundary conditions and initial state.

[0186] Particle parameter assignment: Based on geological exploration data and indoor test results, parameters such as particle size distribution, density, elastic modulus, internal friction angle and cohesion of gangue particles are set; the particle shape adopts the multi-sphere approximation method to ensure the authenticity of particle contact and movement in the simulation.

[0187] Collapse process simulation: Apply gravity field and surrounding rock constraints to the goaf in numerical simulation software (such as PFC, UDEC), gradually remove roof support, and trigger gangue collapse; use time history analysis to record the movement trajectory, velocity and contact force changes of gangue.

[0188] Stress and stability analysis: After the collapse and accumulation stabilizes, the distribution of normal and tangential forces in the contact area between the gangue and the retaining structure is extracted, and the maximum bearing capacity requirement of the retaining structure is calculated; combined with the surrounding rock pressure analysis, the support strength coefficient is determined.

[0189] Support scheme optimization: Based on the stress analysis results, the finite element method was used to verify the impact of different support spacing and support types (steel frame, single column, infill wall, etc.) on the rock retaining effect, and the combination scheme with a safety factor ≥1.5 and the best economic efficiency was selected.

[0190] Outputs include design tables for support strength, spacing, and type, as well as stress distribution diagrams and gangue accumulation diagrams for reference during construction and subsequent monitoring.

[0191] Key parameters for rock retaining support include the support type and spacing. Rock retaining support uses a "metal mesh + U-shaped steel" method. The size of the metal mesh is determined according to the roadway height, and the spacing of the U-shaped steel is 400mm-500mm (the same as the spacing of the slotted boreholes). For mining heights less than 2m, 25U type is used; for mining heights of 2m-4m, 29U type is used; and for mining heights greater than 4m, 36U type is used. Each set of U-shaped steel consists of two overlapping sections, with an overlap length of 1 / 3 of the roadway height.

[0192] S3. Use big data from historical engineering cases to verify and optimize key mining parameters.

[0193] (1) Construction of the engineering big data database and design of a distributed database architecture. The distributed database architecture designed in this embodiment adopts a multi-level technology combination to support the efficient operation of the intelligent design system for pillarless self-forming roadway mining.

[0194] The core architecture includes a data access layer (processing multi-source data such as PDF / CAD / Word), a distributed storage layer (PostgreSQL sharding and table storage of structured parameters, MongoDB management of unstructured cases, and Redis caching of high-frequency data), a computing layer (Spark / Flink implementation of KNN similarity analysis and real-time monitoring and processing), and accelerated retrieval through Elasticsearch.

[0195] This architecture features horizontal scalability, employs the Raft protocol to ensure consistency, and combines with Kafka to implement a real-time monitoring data pipeline, providing reliable data support for intelligent design, supporting the storage of multiple engineering data, and establishing a standardized database management system.

[0196] (2) Intelligent screening algorithm automatically identifies historical cases with similar geological conditions to the target mining area in the engineering database, and performs accurate screening and sorting based on the multi-attribute similarity evaluation model.

[0197] ① Feature system construction:

[0198] The feature system consists of geological parameter features (such as coal seam thickness T). c The characteristics of coal seam dip angle α, burial depth H, roof compressive strength σ1, roof joint development index J, etc., and engineering parameters (such as cut height H0, support strength S) are also important. s Support density D s S-shaped rock retaining strength g It consists of two parts (etc.).

[0199] All features are normalized:

[0200]

[0201] Where, X′ i Represents the normalized features; X i This indicates the selected feature.

[0202] ② Feature dimensionality reduction:

[0203] An improved t-SNE-PCA hybrid dimensionality reduction algorithm is adopted, firstly extracting principal components Z using PCA. k (Retaining a cumulative variance contribution rate of ≥95%), the principal components are then input into the t-SNE model to obtain a low-dimensional manifold embedding representation Y, in order to preserve the global structure and local neighborhood relationships.

[0204] After dimensionality reduction, the dimension of the feature vector is reduced from the original n to m (m << n).

[0205] ③ Similarity calculation model:

[0206] A weighted composite distance metric model is adopted:

[0207]

[0208] Among them, w j A represents the weight of the j-th feature; j B j Let represent the j-th feature value of two samples A and B after dimensionality reduction.

[0209] ④ Search Architecture:

[0210] First layer (coarse screening): Elasticsearch quickly matches candidate cases within a specified feature range based on the inverted index.

[0211] The second layer (refinement): Spark distributed computing performs weighted composite distance calculations and sorts the results by similarity.

[0212] The third layer (dynamic threshold adjustment): The similarity threshold θ1 is dynamically adjusted based on the historical matching accuracy. When Sim≥θ1, it is judged as a highly similar case.

[0213] ⑤ Online learning mechanism:

[0214] The system receives subsequent new engineering cases and actual mining feedback results, and assigns weights w to them. j The threshold θ1 is iteratively updated to achieve adaptive optimization, thereby improving matching accuracy and generalization ability.

[0215] ⑥ Output results:

[0216] Output the top K historical cases with the highest similarity, along with their corresponding geological parameters, engineering parameters, and implementation effects, and provide input data for subsequent analogy optimization.

[0217] (3) Parameter analogy analysis: Establish a key indicator analogy evaluation system. Geological condition analogy parameters include coal seam thickness, coal seam dip angle, coal seam burial depth and roof conditions, etc. Design analogy parameters include roof cutting, roof support, temporary support and rock retaining support, etc.

[0218] The key indicator analogy evaluation system constructed in this embodiment adopts a combination of multi-dimensional feature fusion and dynamic weight allocation to comprehensively evaluate different geological conditions and engineering design parameters. The key to multi-attribute similarity evaluation lies in determining the similarity coefficient M; the similarity coefficient M = mining height coefficient A × burial depth coefficient B × roof hardness coefficient C × coal seam dip angle coefficient D. The closer M is to 1, the higher the similarity. The top 3 samples with the highest similarity are used to compare and verify the design parameters. The difference between each design parameter value and the sample parameters should not exceed 20%; otherwise, manual confirmation and correction are required.

[0219] The method for determining the sampling height coefficient is as follows: if the design target sampling height is greater than or equal to the sample sampling height, then A = design target sampling height / sample sampling height; if the design target sampling height is less than the sample sampling height, then A = sample sampling height / design target sampling height.

[0220] The method for determining the burial depth coefficient B is as follows: if the design target burial depth is greater than or equal to the sample burial depth, then B = design target burial depth / sample burial depth; if the design target burial depth is less than the sample burial depth, then B = sample burial depth / design target burial depth.

[0221] The method for determining the roof hardness coefficient C is as follows: if the design target roof compressive strength is greater than or equal to the sample roof compressive strength, then C = design target roof compressive strength / sample roof compressive strength; if the design target roof compressive strength is less than the sample roof compressive strength, then C = sample roof compressive strength / design target roof compressive strength.

[0222] The method for determining the coal seam dip angle coefficient D is as follows: if the design target coal seam dip angle is greater than or equal to the sample coal seam dip angle, then D = design target coal seam dip angle / sample coal seam dip angle; if the design target coal seam dip angle is less than the sample coal seam dip angle, then D = sample coal seam dip angle / design target coal seam dip angle.

[0223] S4. Based on the optimized and adjusted key mining parameters, complete the mining scheme design and output the design report and drawings.

[0224] (1) Multidimensional feature fusion:

[0225] First, the geological conditions (such as coal seam thickness, coal seam dip angle, roof strength, surrounding rock integrity, burial depth, etc.) and engineering parameters (such as cut height, support density, support strength, cost coefficient, construction period, etc.) are standardized in terms of unified dimensions.

[0226] Standardization not only ensures that indicators with different dimensions can be compared within the same calculation framework, but also facilitates subsequent weight allocation and similarity calculation.

[0227] (2) Weighting and Evaluation Model:

[0228] In terms of weight allocation, the Analytic Hierarchy Process (AHP) is used to construct a three-level indicator system of "objective-criteria-indicators". The relative importance of each core indicator is determined through expert scoring, pairwise comparison and other methods.

[0229] In terms of similarity assessment, a fuzzy comprehensive evaluation model is introduced to integrate the membership degrees of different indicators into a unified evaluation matrix, thereby obtaining the overall similarity between the target case and historical cases.

[0230] (3) Dynamic feedback and adaptive optimization:

[0231] To improve the adaptability of the evaluation system under different mining conditions, a dynamic feedback mechanism is introduced:

[0232] When there are discrepancies between historical cases and actual mining results, the system will automatically adjust the weight allocation ratio based on on-site monitoring data (such as roof subsidence, roadway deformation rate, and changes in support stress).

[0233] Regarding the early warning mechanism, users can set risk thresholds. For example, when the roof subsidence exceeds 15mm, the system will automatically issue a safety alarm and suggest adjustments to the design scheme.

[0234] (4) Solution optimization and output:

[0235] Based on the above evaluation results, the system optimizes and adjusts the preliminary design parameters and generates multiple optional design schemes (usually 3-5 schemes). It then conducts a comprehensive comparison from multiple dimensions such as economy, safety, and operability. The optimal scheme is selected again through AHP and a design report that conforms to industry standards is automatically generated.

[0236] It also outputs parametric CAD design drawings and detailed monitoring plans (including monitoring point layout, monitoring frequency, required equipment type and deployment method).

[0237] In this way, the entire evaluation and optimization process relies on both big data-driven similarity analysis and human experience-based weighted judgment, achieving accurate analogy, adaptive optimization, and traceable verification of design solutions.

[0238] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A pillarless self-forming roadway intelligent design system, characterized in that, include: Geological analysis module, parameter design module, analogy verification module, and output module; The geological analysis module is used to screen and process mine engineering geological data to obtain the geological parameters required for the design. The parameter design module is used to design key mining parameters based on the geological parameters; The analogy verification module is used to verify and optimize the key mining parameters using big data from historical engineering cases; The output module is used to complete the design of the mining scheme and output the design report and drawings based on the optimized and adjusted key mining parameters.

2. The intelligent design system for pillarless self-forming roadway mining according to claim 1, characterized in that, The geological analysis module cleans and standardizes the imported mining area engineering geological data, including removing redundant information, unifying measurement units, converting coordinate systems, and extracting key data. The geological analysis module utilizes a geological feature recognition model constructed using deep learning algorithms to automatically extract key geological parameters; the structure of the geological feature recognition model includes: an input layer, a hidden layer, and an output layer; The input layer is used to input engineering geological data of the mining area, which can be in the form of text and drawings, and includes geological exploration reports, borehole columnar sections, preliminary mine design, operating procedures, and geological specifications. Hidden layers use LSTM or Transformer structures to extract geological features; The output layer outputs geological parameters, including tunnel parameters and strata parameters.

3. The intelligent design system for pillarless self-forming roadway mining according to claim 1, characterized in that, The key mining parameters include: roof cutting, roof support, temporary support, and rock retaining support; The parameters of the top plate cut include the cut angle and the cut height. The cut angle is determined as follows: when the hardness coefficient of the top plate is less than 2, the cut angle is 10°; when the hardness coefficient of the top plate is 2-4, the cut angle is 15°; when the hardness coefficient of the top plate is greater than 4, the cut angle is 20°. The cut height H0 is determined as follows: H0 = f1 × k × h; Where f1 represents the mining height influence coefficient; k represents the roof influence coefficient; and h represents the mining height.

4. The intelligent design system for pillarless self-forming roadway mining according to claim 3, characterized in that, The parameters of the top plate support include anchor cable length and anchor cable spacing; The method for determining the anchor cable length L is as follows: when the cut height H0 < 15m, L = H0 + 2m; when the cut height H0 ≥ 15m, the anchor cable length is taken as 17m. The method for determining the spacing between anchor cables is as follows: when the hardness coefficient of the top plate is >4, the spacing between anchor cables is 1.6m-2.0m, and the number of anchor cables in each row is greater than or equal to 2; when the hardness coefficient of the top plate is 2-4, the spacing between anchor cables is 1.0m-1.6m, and the number of anchor cables in each row is greater than or equal to 3; when the hardness coefficient of the top plate is <2, the spacing between anchor cables is less than 1.0m, and the number of anchor cables in each row is greater than or equal to 3.

5. The intelligent design system for pillarless self-forming roadway mining according to claim 3, characterized in that, The key parameters of the rock-blocking support include the rock-blocking support type and the support spacing; All rock retaining walls are supported by a combination of metal mesh and U-shaped steel. The size of the metal mesh is determined according to the height of the roadway, and the spacing of the U-shaped steel is 400mm-500mm. When the mining height is less than 2m, the 25U type is used; when the mining height is 2m-4m, the 29U type is used; and when the mining height is greater than 4m, the 36U type is used. Each set of U-shaped steel consists of two overlapping sections, with the overlap length being 1 / 3 of the roadway height.

6. The intelligent design system for pillarless self-forming roadway mining according to claim 1, characterized in that, The workflow of the analogy verification module includes: Construct a large engineering database, which is a distributed database architecture; Based on the aforementioned key mining parameters, similar geological condition cases are identified in the engineering database, and a multi-attribute similarity evaluation model is established. Based on the aforementioned multi-attribute similarity evaluation model, a key indicator analogy evaluation system is established; The key mining parameters are optimized and adjusted using the key indicator analogy evaluation system.

7. The intelligent design system for pillarless self-forming roadway mining according to claim 1, characterized in that, The workflow of the output module includes: Output several alternative design schemes, establish an evaluation system that includes indicators of economy, safety and operability, and use the analytic hierarchy process (AHP) to determine the optimal scheme. Automatically generate design reports that conform to industry standards and output parametric CAD design drawings; Develop a detailed monitoring plan, including monitoring equipment, monitoring point layout, and monitoring frequency.

8. A smart design method for pillarless self-forming roadway mining, wherein the method is applied to the system described in any one of claims 1-7, characterized in that the steps include... include: S1. Screen and process the engineering geological data of the mine to obtain the geological parameters required for the design; S2. Based on the geological parameters, design key mining parameters; S3. Use big data from historical engineering cases to verify and optimize the key mining parameters; S4. Based on the optimized and adjusted key mining parameters, complete the mining scheme design and output the design report and drawings.

Images