A Smart System and Method for Deep Rock Property Detection and Geological Formation Inversion

Abstract

This invention discloses a system and method for intelligent detection of deep rock properties and geological formation inversion, relating to the field of mine safety assessment technology. The system includes a terahertz incoherent detection module, a multi-dimensional external field control module, a signal acquisition and control module, and an intelligent inversion and disaster prediction module. The terahertz incoherent detection module includes a terahertz incoherent wave source, a wavefront optimization unit, and a sample coupling unit. The multi-dimensional external field control module includes a magnetic field control unit and a light field control unit. The signal acquisition and control module includes an intrinsically safe synchronous controller, a high-precision current detector, and a signal preprocessing unit. The intelligent inversion and disaster prediction module includes a feature extraction unit, a correlation model library, and a disaster prediction output unit. This invention achieves high-precision, rapid in-situ detection of rock properties and geological formation conditions, as well as accurate early warning of disaster risks, effectively supporting the safe and efficient mining of deep resources.

Description

Technical Field

[0001] This invention relates to the field of mine safety assessment technology, and in particular to a system and method for intelligent detection of deep rock properties and geological formation inversion. Background Technology

[0002] In the field of geological exploration and deep-earth resource extraction, accurately grasping the rock physical parameters (such as lithology, water content, pyrite content, etc.) and geological formation conditions (including pressure field, temperature field, seepage field, chemical field distribution, etc.) of the mining face is the core prerequisite for ensuring the safe and efficient advancement of mining operations and preventing geological disasters such as water hazards, gas, and rock bursts.

[0003] Traditional geological exploration methods mainly rely on technologies such as drilling, seismic wave detection, and ground-penetrating radar. However, these methods have many insurmountable drawbacks in practical applications: drilling technology is costly and has a long operation cycle, significantly interferes with underground or field mining operations, and can only obtain discrete data at single points, failing to comprehensively reflect the overall geological conditions of the working face and making it difficult to meet the needs of large-scale, precise exploration; seismic wave detection has limited resolution, insufficient ability to identify fine geological structures such as small faults and micro-fractures, and its interpretation results are highly ambiguous, easily leading to biased geological judgments; ground-penetrating radar detection is susceptible to interference from underground electromagnetic waves and complex electromagnetic environments in the field, and its detection depth is relatively shallow, making it difficult to meet the exploration needs of deep-earth resource mining, while also lacking sufficient accuracy in identifying key parameters such as water content and mineral composition.

[0004] With the development of terahertz technology, its characteristics such as low photon energy, sensitivity to the microstructure and water content of materials, and non-radiation and non-damage have provided a brand-new technical path for geological exploration. However, existing terahertz detection technology still faces three major bottlenecks in its application in deep-earth resource mining scenarios: First, the signal extraction efficiency is low. Traditional terahertz coherent detection is easily affected by underground dust, humidity, or complex outdoor environments, resulting in interference distortion. Rock response signals are mostly weak signals in the pA-nA range, which are easily submerged by environmental noise, making it difficult to extract effective signals. Second, the information correlation is insufficient. A single terahertz spectrum is insufficient to distinguish the characteristic signals of different components in rocks, such as water, pyrite, and organic matter. It is impossible to establish an effective correlation between the detection signal and geological formation conditions and disaster risk factors, making it difficult to achieve multi-parameter synchronous inversion. Third, the environmental adaptability is poor. Existing terahertz detection systems mostly rely on expensive ultrafast lasers and coherent sampling modules, which are complex in structure, bulky, and costly. They also lack targeted designs for dustproof, waterproof, and explosion-proof features, making it difficult to adapt to complex working conditions such as confined underground spaces, high temperature and humidity, and harsh outdoor environments, and thus unable to achieve in-situ real-time detection.

[0005] Therefore, there is an urgent need to develop a rock physical property and geological formation condition detection technology that combines high precision, high efficiency, low cost and strong environmental adaptability, in order to solve many of the pain points of traditional detection methods, break through the application limitations of existing terahertz detection technology, and provide reliable technical support for the safe and efficient mining of deep earth resources. Summary of the Invention

[0006] The purpose of this invention is to propose an intelligent detection system and method for deep rock physical properties and geological formation inversion, so as to achieve rapid, accurate, and in-situ detection and prediction of rock physical parameters, geological formation conditions, and disaster risks such as water hazards, gas, and rock bursts, and provide technical support for the safe and efficient mining of deep earth resources.

[0007] To achieve the above objectives, this invention proposes an intelligent detection and geological formation inversion system for deep rocks, comprising: a terahertz incoherent detection module, a multi-dimensional external field control module, a signal acquisition and control module, and an intelligent inversion and disaster prediction module;

[0008] The terahertz incoherent detection module includes a terahertz incoherent wave source, a wavefront optimization unit, and a sample coupling unit. The wavefront optimization unit adopts a simple optical structure that is dustproof and waterproof, and optimizes the propagation direction and focusing state of the terahertz beam to ensure that the beam efficiently covers the rock detection area. The sample coupling unit adopts a near-field coupling design with a detection distance of less than 1 mm.

[0009] The multi-dimensional external field control module includes a magnetic field control unit and an optical field control unit. The magnetic field control unit uses a miniature Helmholtz coil as a magnetic field generator to produce a continuously adjustable uniform magnetic field of 0-1 mT, with gradient intensities of 0 mT, 0.2 mT, and 0.4 mT. The magnetic field direction is parallel to the coal and rock surface and perpendicular to the direction of charge carrier movement. The optical field control unit can output optical fields with wavelengths of 355 nm, 405 nm, 520 nm, and 633 nm.

[0010] The signal acquisition and control module is used to synchronously control the operating parameters of each module, acquire and preprocess response signals, and includes an intrinsically safe synchronous controller, a high-precision current detector, and a signal preprocessing unit; the high-precision current detector has a resolution of 10. -12 A is used to collect micro-current signals of rocks under the action of terahertz waves alone and under the combined action of terahertz waves and multidimensional external fields, including the bright and dark current signal ΔI(THz) after applying a terahertz field, the bright and dark current signal ΔI(B) after applying a magnetic field, and the bright and dark current signal ΔI(L) after applying a laser. The signal preprocessing unit integrates a low-pass filter and an intrinsically safe amplifier to perform noise reduction and amplification processing on the collected weak current signals. The filter cutoff frequency is 10Hz, the amplifier gain is adjustable, and downhole electromagnetic interference is suppressed to ensure that the signal meets the requirements of subsequent processing.

[0011] The intelligent inversion and disaster prediction module is used to extract features and perform correlation modeling on the preprocessed signal, invert coal and rock physical parameters and geological formation conditions, and predict the risk of water hazards, gas and rock bursts. It includes a feature extraction unit, a correlation model library and a disaster prediction output unit. The correlation model library includes a multiple linear regression model and a physical information neural network model.

[0012] The workflow of the inversion system is as follows: the terahertz incoherent detection module radiates terahertz waves that are incident on the rock to be detected; the signal acquisition and control module extends thin-film electrodes to both sides of the rock area to be tested, applies micro-pressure to acquire micro-current signals; the multi-dimensional external field control module applies one type of laser or weak magnetic field, two types of laser or weak magnetic fields, and three types of laser or weak magnetic fields in the detection area to detect micro-current signals respectively; the signal preprocessing unit calibrates and calculates the micro-current signals to obtain the terahertz response parameters of the rock layer and traverses the four terahertz frequencies of the terahertz incoherent wave source to obtain all the single-frequency terahertz response parameters of the rock.

[0013] Preferably, the terahertz incoherent detection module adopts a non-penetrating detection method to collect the microcurrent signal difference ΔI of the rock under terahertz irradiation and non-irradiation conditions; the terahertz incoherent wave source is a continuous terahertz incoherent wave source containing 0.1THz, 0.14THz, 0.192THz and 0.252THz, and the wave source power adjustment range is 50-150mW.

[0014] Preferably, the multi-dimensional external field control module can realize the coordinated detection of magnetic field-terahertz and optical field-terahertz, and sequentially collect the bright and dark current ΔI(B) after the magnetic field is applied and the bright and dark current ΔI(L) after the laser is applied, construct a three-dimensional response matrix, and sensitively reflect the information of rock water content, pyrite content, stress distribution and electronic transition characteristics, thereby reflecting the pressure, seepage, temperature and chemical conditions in the formation conditions.

[0015] Preferably, the signal acquisition and control module is based on quartz or diamond as a substrate, on which a subwavelength microstructure array adapted to four terahertz frequencies of 0.1THz, 0.14THz, 0.192THz, and 0.252THz is prepared; a 100-150nm thick Co3Sn2S2 active layer and gold, platinum, or silver electrodes are prepared using magnetron sputtering, with platinum being the preferred material; the active layer is disposed at the bottom of the substrate, and the electrodes are disposed on both sides of the subwavelength microstructure array, with a thickness of 100nm; a 100fs, 50MHz, 80mW, 2mm diameter laser is used at a speed of 1mm / s to process wires between the subwavelength microstructure array and the electrodes by inducing abrupt changes in conductivity.

[0016] Preferably, the feature extraction unit extracts equivalent noise power (NEP), modulation depth, and signal relaxation time feature parameters from the preprocessed signal to construct a high-dimensional feature vector. The magnetic field modulation depth reflects the rock stress distribution and pyrite content, while the optical field modulation depth sensitively reflects the water content and organic matter characteristics. The association model library, based on a multiple linear regression algorithm, constructs a quantitative association model between the feature parameters and formation depth, pressure, temperature, and seepage. Simultaneously, a physical information neural network (PINN) is integrated, constrained by the physical laws of charge carrier transport, and combined with a rock geological database to establish a disaster risk assessment model. The disaster prediction output unit receives in-situ detection data from the well, substitutes it into the association model library for calculation, and outputs the rock type, geological formation conditions, and disaster risk levels for water hazards, gas, and rockbursts, generating a geological condition and disaster risk distribution map of the mining face.

[0017] Preferably, the quantitative correlation model construction steps are as follows:

[0018] Define feature parameter vector X and geological parameter vector Y :

[0019] Feature parameter vector X The formula is as follows:

[0020] ;

[0021] in, The equivalent noise power without external field reflects dielectric loss and structural compactness; The magnetic modulation depth under high voltage reflects the stress-induced cracks and the content of magnetic impurities. , The laser modulation depth; The value is 405nm or 520nm, reflecting the content of higher organic matter / volatile matter; The value is 355nm or 633nm, reflecting the content of low-order organic matter / moisture; The signal relaxation time reflects the carrier recombination dynamics; among which, This is the magnetic modulation depth function. It is a high-voltage electric field;

[0022] Geological parameter vector Y The formula is as follows:

[0023] ;

[0024] in, For temperature field, To create depth, For pressure, This represents the seepage saturation level.

[0025] The quantitative correlation model is as follows:

[0026] ;

[0027] in, The regression coefficient matrix, The intercept vector, This is the error term.

[0028] Preferably, the steps for constructing a disaster risk assessment model are as follows:

[0029] Step S101, the formula for the Physical Information Neural Network (PINN) is as follows:

[0030] ;

[0031] in, For the output of the PINN neural network, As input to the PINN neural network, Here are the network parameters, and σ is the activation function. For the first l The weight matrix of the layer, For the first l Layer bias vector, For neural network functions, L This represents the total number of layers in the neural network.

[0032] Step S102, the physical constraint steps for carrier transport are as follows:

[0033] Step S1021, Drift-Diffusion Equation, the formula is as follows:

[0034] ;

[0035] ;

[0036] ;

[0037] ;

[0038] ;

[0039] in, The electron diffusion coefficient is... For electron mobility, Photogenerated carrier generation rate, The carrier recombination loss rate, n For electron concentration, t For time, For bias electric field, For terahertz electric fields, Boltzmann's constant, Absolute temperature For elementary charge, As the baseline mobility, For stress, It is a stress-temperature dependent function. The wavelength-dependent light absorption coefficient, The intensity of the laser light. Let be Planck's constant. For light frequency, For the composite lifespan, The Auger composite coefficient;

[0040] The PINN residual constraints are as follows:

[0041] ;

[0042] in, For the residual loss of partial differential equations, For the number of points, The carrier concentration predicted by the neural network. i For point indexing, For electric field strength, Carrier generation rate, The carrier recombination rate;

[0043] Step S1022, Lorentz force control, the formula is as follows:

[0044] The formula for the motion of charge carriers in a magnetic field is as follows:

[0045] ;

[0046] The effective mobility is obtained from the steady-state solution:

[0047] ;

[0048] High magnetic field resistance:

[0049] ;

[0050] The PINN magnetic field constraint is as follows:

[0051] ;

[0052] in, For the effective mass of electrons, For electron drift velocity, Magnetic flux density Let be the scattering relaxation time. For magnetic field-dependent effective mobility, This is the change in resistivity. With zero magnetic field resistivity, Resistivity is a magnetic field dependent property. For disorder degree parameter, For disorder correction function, Loss due to physical constraint of the magnetic field. The number of magnetic field-constrained sampling points. The magnetic modulation depth is predicted by the neural network. This refers to the magnetic field-dependent change in conductivity. With zero magnetic field conductivity, j Index of magnetic field-constrained sampling points;

[0053] Step S1023, Effective dielectric response, the formula is as follows:

[0054] Constraints based on Brugmann's effective medium approximation:

[0055] ;

[0056] The relationship with the terahertz response parameter NEP is as follows:

[0057] ;

[0058] The dielectric constraint formula for PINN is as follows:

[0059] ;

[0060] in, This represents the volume fraction of water. Where is the dielectric constant of water. For the effective dielectric constant, The dielectric constant of the rock matrix is ​​. Let be the real part of the dielectric constant. This is the imaginary part of the dielectric constant. DC conductivity Angular frequency, The vacuum permittivity, The imaginary unit, The imaginary part of the effective dielectric constant, For the effective medium approximate constraint loss, The number of dielectric-constrained samples. The effective dielectric constant predicted by the neural network. k For dielectric constraint sampling index;

[0061] Step S103, Physical constraints of disaster risk assessment, including:

[0062] The risk of water damage is coupled with the seepage field and pressure field, as shown in the following formula:

[0063] ;

[0064] in, The risk level is determined by the level of water damage. The saturation influence coefficient is... Critical saturation This is the pressure influence coefficient. For maximum pressure;

[0065] Gas risk is coupled with temperature and chemical fields, as shown in the following formula:

[0066] ;

[0067] in, According to the gas risk level, The content of volatile matter on a dry, ash-free basis is as follows. For activation energy, This is the universal gas constant. For reference temperature;

[0068] The risk of rock bursts is coupled with the stress field and structural field, as shown in the following formula:

[0069] ;

[0070] in, To assess the risk level of rockburst, For differential stress, The density of the fracture. The critical differential stress, Maximum fracture density;

[0071] The formula for magnetic modulation depth inversion is as follows:

[0072] ;

[0073] in, The magnetic modulation depth gradient coefficient, Let be the partial derivative of the magnetic modulation depth with respect to the voltage. It is a high magnetic field modulation coefficient;

[0074] Step S104, Total Loss Function:

[0075] ;

[0076] ;

[0077] in, For the total loss function, For data-driven loss, , , , For physical constraint weights, To constrain losses due to disaster risks, The number of training data points, For the first p The neural network predicts the output of each input sample. For the first p The true observations of each input sample.

[0078] This invention also provides a method for intelligent detection of deep rock physical properties and geological formation inversion, the specific steps of which are as follows:

[0079] Step S1: In the safe area, turn on each module to perform device self-test, adjust the terahertz wave source power, magnetic field strength accuracy and laser wavelength stability, and measure the dark current as the signal processing baseline.

[0080] Step S2: Place the detection device close to the surface of the rock to be detected, remove dust and water from the rock surface, ensure that the detection interface is tightly fitted, and control the coupling distance to 0.5-1mm;

[0081] Step S3: Perform terahertz individual detection, magnetic field-terahertz co-detection, and optical field-terahertz co-detection in sequence, with the detection time controlled within 5 seconds for each detection. Collect the bright and dark current signals after applying a single terahertz field, the bright and dark current signals after applying a magnetic field, and the bright and dark current signals after applying a laser, respectively.

[0082] Step S4: Perform noise reduction and amplification preprocessing on the acquired signal, extract the equivalent noise power NEP, modulation depth and signal relaxation time feature parameters, and construct a high-dimensional terahertz fingerprint feature vector.

[0083] Step S5: Input the feature vector into the intelligent inversion and disaster prediction module to invert rock physical parameters and geological formation conditions, predict disaster risk level and generate geological condition distribution map.

[0084] Preferably, in step S3, when performing magnetic field-terahertz coordinated detection, the magnetic field strength is switched in a gradient of 0mT, 0.2mT, 0.4mT, 0.6mT, 0.8mT, and 1.0mT to enhance charge mobility through Lorentz force and strengthen the sensitivity to rock stress distribution.

[0085] Preferably, when performing optical field-terahertz co-detection, wavelengths that match the electronic transition energy levels of rocks are selected first. The optical field modulation strategy can dynamically switch visible and near-infrared wavelengths according to rock type to enhance the specificity of photocurrent response.

[0086] Preferably, in step S5, the rock physical property parameters output by the inversion include rock type, vitrinite reflectance, formation depth, pressure, water content, and pyrite content; the geological formation conditions include pressure field, temperature field, seepage field, and chemical field characteristics.

[0087] Therefore, this invention proposes a system and method for intelligent detection of deep rock physical properties and geological formation inversion, the beneficial effects of which are as follows:

[0088] (1) This invention achieves accurate inversion of geological formation conditions such as rock pressure field and seepage field by combining terahertz incoherent detection with multidimensional synergistic control of 0-1mT adjustable magnetic field and multi-wavelength optical field, combined with multivariate linear regression and physical information neural network fusion model, effectively breaking through the bottleneck of traditional detection in identifying fine structure and weak signal.

[0089] (2) The inversion system in this invention adopts a near-field non-penetrating detection mode with a single detection time of ≤5 seconds, which can realize detection as mining and output results in real time, greatly improving the efficiency of mining operation connection; at the same time, it has dustproof, waterproof and fogproof and intrinsically safe explosion-proof design, and the working temperature covers -20°C to +60°C. It can adapt to the narrow space, high temperature and humidity and complex working conditions in the field. It can be directly embedded in the tunneling machine or mining machine to realize in-situ detection.

[0090] (3) This invention eliminates the expensive ultrafast laser and coherent sampling module of traditional terahertz systems and adopts commercially mature core components, which greatly reduces the overall cost. It also eliminates the need for consumables and is reusable, significantly reducing the investment in detection. At the same time, terahertz wave photons have low energy, pose no radiation hazard to the human body, and do not damage rock samples. It avoids the environmental damage and safety hazards of traditional drilling and other methods from the source, and has both significant economic value and safety production guarantee.

[0091] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0092] Figure 1 This is an architectural diagram of a deep rock physical property intelligent detection and geological formation inversion system according to the present invention;

[0093] Figure 2 This is a flowchart of a method for intelligent detection of deep rock physical properties and geological formation inversion according to the present invention;

[0094] Figure 3 This is a structural diagram of the signal acquisition and control module in this invention;

[0095] Figure 4 This is a schematic diagram of the inversion and prediction results of the 5307 working surface in Embodiment 2 of the present invention;

[0096] Figure 5This is a schematic diagram of the inversion and prediction results of the 7304 working surface in Embodiment 2 of the present invention;

[0097] Figure 6 This is a schematic diagram of the inversion and prediction results of the No. 1 return airway in the north in Embodiment 2 of the present invention. Attached Figure Description

[0099] 1. Electrode; 201, 0.1THz subwavelength structure; 202, 0.14THz subwavelength structure; 203, 0.192THz subwavelength structure; 204, 0.252THz subwavelength structure; 3. Quartz; 4. Active layer. Detailed Implementation

[0100] To make the technical solutions, advantages, and objectives of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below. The described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the protection scope of the present invention.

[0101] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0102] Example 1

[0103] Experimental setup.

[0104] This embodiment takes the intelligent detection of physical properties and geological formation conditions of a coal mine as an example, according to... Figure 1 The experimental setup shown is a downhole-adaptive system for building a rock property intelligent detection and geological formation condition inversion system.

[0105] Terahertz incoherent wave source: an intrinsically safe continuous terahertz source is selected, with output frequencies of 0.1THz, 0.14THz, 0.192THz, and 0.252THz, and a power adjustment range of 50-150mW;

[0106] Magnetic field control unit: adopts miniature Helmholtz coil, with a maximum output magnetic field of 1mT and magnetic field uniformity of ±2%, which meets the requirements for explosion protection in underground mines;

[0107] Optical field modulation unit: Utilizes intrinsically safe LED / laser diodes, with output wavelengths of 355nm, 405nm, 520nm, and 633nm, and an optical power density of 10mW / cm². 2 ;

[0108] like Figure 3As shown, the signal acquisition and control module is based on quartz 3 as a substrate. Subwavelength microstructure arrays adapted to four terahertz frequencies (0.1THz, 0.14THz, 0.192THz, and 0.252THz) are fabricated on the substrate. A 100-150nm thick Co3Sn2S2 active layer 4 and electrode 1 are fabricated using magnetron sputtering. Electrode 1 is made of platinum. The active layer 4 is located at the bottom of the substrate, and electrodes 1 are located on both sides of the subwavelength microstructure array, with a thickness of 100nm. A 100fs, 50MHz, 2mm diameter laser is used at a speed of 1mm / s to induce abrupt changes in conductivity between the subwavelength microstructure array and electrode 1 to fabricate a conductive wire. In this embodiment, the signal acquisition and control module uses an intrinsically safe high-precision digital source meter with a current measurement resolution of 10. -12 A, sampling rate 1kSPS;

[0109] Intelligent inversion and disaster prediction module: Based on ARM Cortex processor + NPU accelerator, it integrates multiple linear regression model and PINN model, with weights stored in on-chip Flash less than 1MB, and has the function of local processing and wireless transmission of coal mine data.

[0110] Example 2

[0111] Real-time underground detection.

[0112] like Figure 2 As shown, this invention provides a method for intelligent detection of deep rock properties and geological formation inversion. Taking the geological detection of the 5307 working face, 7304 working face, and the No. 1 return airway in the north of a coal mine as an example, the specific steps are as follows:

[0113] System initialization: In the safe area downhole, turn on all modules, calibrate the dark current to 5pA, set the terahertz source power to 100mW, and set the magnetic field gradients to 0mT, 0.2mT, 0.4mT, 0.6mT, 0.8mT, and 1.0mT, and the laser wavelengths to 355nm, 405nm, 520nm, and 633nm.

[0114] Sample coupling: Place the detection device close to the coal and rock surface of the working face, remove surface dust, and maintain a coupling distance of 0.5 mm;

[0115] Multi-field coordinated detection: Sequentially complete terahertz individual detection, magnetic field-terahertz coordinated detection, and optical field-terahertz coordinated detection, and collect micro-current signals under various conditions;

[0116] Feature extraction: The NEP values ​​and modulation depths of the coal and rock in each working face were calculated; among them, the initial value of the terahertz response parameter NEP in working face 5307 without external field was >2000pW / Hz. 1 / 2The laser modulation depth is close to 1.0, and the magnetic field modulation depth is 0.1-0.4; the peak terahertz response parameter (NEP) of the 7304 working surface is approximately 700 pW / Hz. 1 / 2 The magnetic field modulation depth under high voltage exceeds 0.9; the terahertz response parameter NEP value of the No. 1 return airway in the north is the lowest and most stable, and the 520nm laser modulation depth is about 0.9.

[0117] Intelligent Inversion and Disaster Prediction: Input the feature parameters into the intelligent inversion and disaster prediction module. The inversion and prediction results are as follows:

[0118] like Figure 4 As shown, the coal and rock type is coking coal, with a vitrinite reflectance of 1.35% and a formation depth of 0–1000 m; high ground stress, affected by four faults, with local stress concentration zones (moderate risk of rockburst); moderate limestone water pressure of 2.66 MPa, high Ordovician limestone water pressure of 11.13 MPa, and a strong runoff zone (normal inflow of 493 m³ / h). 3 / h), high risk of water damage; gas emission rate 0.51m 3 / min, high risk of spontaneous combustion;

[0119] like Figure 5 As shown, the coal and rock type is coking coal, with a vitrinite reflectance of 1.48% and a formation depth of 0–1080 m. High ground stress and complex tectonic stress are observed, with 22 faults leading to uneven stress fields (high risk of rockburst). The water pressure of the three limestone formations is 0.4 / 1.9 MPa, and that of the Ordovician limestone formation is 11.7 MPa, resulting in a complex seepage field (with a moderate inflow of 80 m³ / h). 3 / h), moderate risk of water hazard; gas emission rate 0.20m 3 / min, the risk of spontaneous combustion is the highest;

[0120] like Figure 6 As shown, the coal and rock type is lean coal, with a vitrinite reflectance of 1.62%, and a formation depth of 0–1100 m; it has high ground stress and is cut by 13 faults (moderate risk of rockburst); it has extremely high limestone water pressure of 8.31 MPa and Ordovician limestone water pressure of 11.71 MPa, and a stable seepage field (water inflow of 39 m³ / h). 3 / h), the risk of sudden water inrush to the bottom slab is the highest; gas emission rate is 0.09m³. 3 / min, with a low risk of spontaneous combustion.

[0121] The above implementation examples demonstrate that the intelligent rock property detection and geological formation condition inversion system and method proposed in this invention have the following advantages:

[0122] Sensitivity: Terahertz response parameters (NEP) and modulation depth are highly sensitive to the micro-electrical structure of coal and rock masses, and can accurately reflect macro-geological characteristics and disaster risk factors such as pressure, temperature, seepage, and chemical field.

[0123] Distinguishing features: The terahertz response data of the three working faces show significant differences in magnitude, trend and modulation characteristics, which can clearly distinguish different geological conditions and disaster risk levels;

[0124] Efficiency: Single detection time is 3 seconds, which is far superior to traditional drilling, ground-penetrating radar and other detection methods;

[0125] Environmental adaptability: The system has a volume of 12L and a weight of 8kg. It is designed to be dustproof, waterproof, and intrinsically safe. It can work stably in underground environments with temperatures ranging from -20°C to +60°C and humidity ≤95%, meeting the explosion-proof requirements of coal mines.

[0126] It is worth noting that all contents not described in detail in this invention are existing technologies and are well known to those skilled in the art.

[0127] Therefore, this invention provides a system and method for intelligent detection of deep rock properties and geological formation inversion. Through terahertz incoherent detection and multi-dimensional external field control, it achieves high-precision, rapid in-situ detection of rock properties and geological formation conditions, as well as accurate early warning of disaster risks. It has strong detection performance, strong environmental adaptability, and significant economic and safety benefits, effectively supporting the safe and efficient mining of deep earth resources.

[0128] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A deep rock physical property intelligent detection and geological formation inversion system, characterized in that, It includes a terahertz incoherent detection module, a multi-dimensional external field control module, a signal acquisition and control module, and an intelligent inversion and disaster prediction module; The terahertz incoherent detection module includes a terahertz incoherent wave source, a wavefront optimization unit, and a sample coupling unit; the sample coupling unit adopts a near-field coupling design with a detection distance of less than 1 mm. The terahertz incoherent detection module uses a non-penetrating detection method to collect the difference ΔI of microcurrent signals in rocks under terahertz irradiation and non-irradiation conditions; the terahertz incoherent wave source is a continuous terahertz incoherent wave source containing 0.1THz, 0.14THz, 0.192THz, and 0.252THz, and the wave source power adjustment range is 50-150mW; The multi-dimensional external field control module includes a magnetic field control unit and an optical field control unit. The magnetic field control unit is a miniature Helmholtz coil that can generate a continuously adjustable uniform magnetic field of 0-1mT. The optical field control unit can output optical fields with wavelengths of 355nm, 405nm, 520nm, and 633nm. The signal acquisition and control module is used to synchronously control the working parameters of each module, acquire and preprocess response signals, including an intrinsically safe synchronous controller, a high-precision current detector and a signal preprocessing unit; The intelligent inversion and disaster prediction module is used to extract features and perform correlation modeling on the preprocessed signal, invert coal and rock physical parameters and geological formation conditions, and predict the risk of water hazards, gas and rock bursts. It includes a feature extraction unit, a correlation model library and a disaster prediction output unit. The correlation model library includes a multiple linear regression model and a physical information neural network model. The feature extraction unit extracts equivalent noise power (NEP), modulation depth, and signal relaxation time feature parameters from the preprocessed signal to construct a high-dimensional feature vector. The magnetic field modulation depth reflects rock stress distribution and pyrite content, while the optical field modulation depth sensitively reflects water content and organic matter characteristics. The correlation model library, based on a multiple linear regression algorithm, constructs a quantitative correlation model between feature parameters and formation depth, pressure, temperature, and seepage. Simultaneously, a physical information neural network (PINN) is integrated, constrained by the physical laws of charge carrier transport, and combined with a rock geological database to establish a disaster risk assessment model. The disaster prediction output unit receives in-situ detection data from the well, substitutes it into the correlation model library for calculation, and outputs the rock type, geological formation conditions, and disaster risk levels for water hazards, gas, and rockbursts, generating a geological condition and disaster risk distribution map of the mining face. The steps for constructing a quantitative correlation model are as follows: Define feature parameter vector X and geological parameter vector Y : Feature parameter vector X The formula is as follows: ； in, The equivalent noise power without external field reflects dielectric loss and structural compactness; The magnetic modulation depth under high voltage reflects the stress-induced cracks and the content of magnetic impurities. , The laser modulation depth; The value is 405nm or 520nm, reflecting the content of higher organic matter / volatile matter; The value is 355nm or 633nm, reflecting the content of low-order organic matter / moisture; The signal relaxation time reflects the carrier recombination dynamics; among which, This is the magnetic modulation depth function. It is a high-voltage electric field; Geological parameter vector Y The formula is as follows: ； in, For temperature field, To create depth, For pressure, This represents the seepage saturation level. The quantitative correlation model is as follows: ； in, This is the regression coefficient matrix. The intercept vector, This is the error term; The workflow of the inversion system is as follows: the terahertz incoherent detection module radiates terahertz waves that are incident on the rock to be detected; the signal acquisition and control module extends thin-film electrodes to both sides of the rock area to be tested, applies micro-pressure to acquire micro-current signals; the multi-dimensional external field control module applies one type of laser or weak magnetic field, two types of laser or weak magnetic fields, and three types of laser or weak magnetic fields in the detection area to detect micro-current signals respectively; the signal preprocessing unit calibrates and calculates the micro-current signals to obtain the terahertz response parameters of the rock layer and traverses the four terahertz frequencies of the terahertz incoherent wave source to obtain all the single-frequency terahertz response parameters of the rock.

2. The intelligent detection and geological formation inversion system for deep rocks according to claim 1, characterized in that, The multi-dimensional external field control module can realize the coordinated detection of magnetic field-terahertz and optical field-terahertz, and sequentially collect the bright and dark current signals after the magnetic field is applied and the bright and dark current signals after the laser is applied, and construct a three-dimensional response matrix to sensitively reflect information such as rock water content, pyrite content, stress distribution and electronic transition characteristics.

3. The intelligent detection and geological formation inversion system for deep rocks according to claim 2, characterized in that, The signal acquisition and control module uses quartz or diamond as a substrate, on which a subwavelength microstructure array adapted to four terahertz frequencies of 0.1THz, 0.14THz, 0.192THz, and 0.252THz is fabricated. A 100-150nm thick Co3Sn2S2 active layer and gold, platinum, or silver electrodes are fabricated using magnetron sputtering. The active layer is located at the bottom of the substrate, and the electrodes are located on both sides of the subwavelength microstructure array, with a thickness of 100nm. A 100fs, 50MHz, 80mW, 2mm diameter laser is used at a speed of 1mm / s to fabricate wires between the subwavelength microstructure array and the electrodes by inducing abrupt changes in conductivity.

4. The intelligent detection and geological formation inversion system for deep rocks according to claim 1, characterized in that, The steps for constructing a disaster risk assessment model are as follows: Step S101, the formula for the Physical Information Neural Network (PINN) is as follows: ； in, For the output of the PINN neural network, As input to the PINN neural network, Here are the network parameters, and σ is the activation function. For the first l The weight matrix of the layer, For the first l The layer's bias vector, For neural network functions, L This represents the total number of layers in the neural network. Step S102, the physical constraint steps for carrier transport are as follows: Step S1021, Drift-Diffusion Equation, the formula is as follows: ； ； ； ； ； in, The electron diffusion coefficient is... For electron mobility, Photogenerated carrier generation rate, The carrier recombination loss rate, n For electron concentration, t For time, For bias electric field, For terahertz electric fields, Boltzmann's constant, Absolute temperature For elementary charge, As the baseline mobility, For stress, It is a stress-temperature dependent function. The wavelength-dependent light absorption coefficient, The intensity of the laser light. is Planck's constant. For light frequency, For the composite lifespan, The Auger composite coefficient; The PINN residual constraints are as follows: ； in, For the residual loss of partial differential equations, For the number of points, The carrier concentration predicted by the neural network. i For point indexing, For electric field strength, Carrier generation rate, The carrier recombination rate; Step S1022, Lorentz force control, the formula is as follows: The formula for the motion of charge carriers in a magnetic field is as follows: ； The effective mobility is obtained from the steady-state solution: ； High magnetic field resistance: ； The PINN magnetic field constraint is as follows: ； in, For the effective mass of electrons, For electron drift velocity, It represents the magnetic flux density. Let be the scattering relaxation time. For magnetic field-dependent effective mobility, This is the change in resistivity. With zero magnetic field resistivity, For magnetic field-dependent resistivity, For disorder degree parameter, For disorder correction function, Loss due to physical constraint of the magnetic field. The number of magnetic field-constrained sampling points. The magnetic modulation depth is predicted by the neural network. This refers to the magnetic field-dependent change in conductivity. With zero magnetic field conductivity, j Index of magnetic field-constrained sampling points; Step S1023, Effective dielectric response, the formula is as follows: Constraints based on Brugmann's effective medium approximation: ； The relationship with the terahertz response parameter NEP is as follows: ； The dielectric constraint formula for PINN is as follows: ； in, This represents the volume fraction of water. Where is the dielectric constant of water. For the effective dielectric constant, The dielectric constant of the rock matrix is ​​. Let be the real part of the dielectric constant. This is the imaginary part of the dielectric constant. DC conductivity Angular frequency, The vacuum permittivity, The imaginary unit, The imaginary part of the effective dielectric constant, For the effective medium approximate constraint loss, The number of dielectric-constrained samples. The effective dielectric constant predicted by the neural network. k For dielectric constraint sampling index; Step S103, Physical constraints of disaster risk assessment, including: The risk of water damage is coupled with the seepage field and pressure field, as shown in the following formula: ； in, The risk level is determined by the level of water damage. The saturation influence coefficient. Critical saturation This is the pressure influence coefficient. For maximum pressure; Gas risk is coupled with temperature and chemical fields, as shown in the following formula: ； in, According to the gas risk level, The content of volatile matter on a dry, ash-free basis is as follows. For activation energy, This is the universal gas constant. For reference temperature; The risk of rock bursts is coupled with the stress field and structural field, as shown in the following formula: ； in, To assess the risk level of rockburst, For differential stress, For crack density, The critical differential stress, Maximum fracture density; The formula for magnetic modulation depth inversion is as follows: ； in, The magnetic modulation depth gradient coefficient, Let be the partial derivative of the magnetic modulation depth with respect to the voltage. It is a high magnetic field modulation coefficient; Step S104, Total Loss Function: ； ； in, For the total loss function, For data-driven loss, , , , For physical constraint weights, To constrain losses due to disaster risks, The number of training data points, For the first p The neural network predicts the output of each input sample. For the first p The true observations of each input sample.

5. A method for implementing the intelligent detection and geological formation inversion system for deep rock properties as described in claim 1, characterized in that, The specific steps are as follows: Step S1: In the safe area, turn on each module to perform device self-test, adjust the terahertz wave source power, magnetic field strength accuracy and laser wavelength stability, and measure the dark current as the signal processing baseline. Step S2: Place the detection device close to the surface of the rock to be detected, remove dust and water from the rock surface, ensure that the detection interface is tightly fitted, and control the coupling distance to 0.5-1mm; Step S3: Perform terahertz individual detection, magnetic field-terahertz co-detection, and optical field-terahertz co-detection in sequence, with the detection time controlled within 5 seconds for each detection. Collect the bright and dark current signals after applying a single terahertz field, the bright and dark current signals after applying a magnetic field, and the bright and dark current signals after applying a laser, respectively. Step S4: Perform noise reduction and amplification preprocessing on the acquired signal, extract the equivalent noise power NEP, modulation depth and signal relaxation time feature parameters, and construct a high-dimensional terahertz fingerprint feature vector. Step S5: Input the feature vector into the intelligent inversion and disaster prediction module to invert rock physical parameters and geological formation conditions, predict disaster risk level and generate geological condition distribution map.

6. The method for intelligent detection of deep rock physical properties and geological formation inversion according to claim 5, characterized in that, In step S3, when performing magnetic field-terahertz coordinated detection, the magnetic field strength is switched in gradients of 0mT, 0.2mT, 0.4mT, 0.6mT, 0.8mT, and 1.0mT. The charge mobility is increased by the Lorentz force, thereby enhancing the sensitivity to the stress distribution of the rock.

7. The method for intelligent detection of deep rock physical properties and geological formation inversion according to claim 5, characterized in that, When conducting optical field-terahertz co-detection, wavelengths that match the electronic transition energy levels of rocks are preferentially selected. The optical field modulation strategy can dynamically switch visible and near-infrared wavelengths according to rock type to enhance the specificity of photocurrent response.

Images