Safe recovery process based on weak rock mass condition

Abstract

The invention relates to a safe stoping process based on a weak rock mass condition, which comprises the following steps: S1, through three-dimensional geological modeling and crustal stress field measurement, dividing geological units and determining differentiated stope structure parameters; s2, layered stoping, wherein the stope advancing speed is dynamically adjusted according to the ground stress real-time monitoring data; s3, layered filling operation is implemented, a shape memory alloy wire mesh is preset in a filling body, and a grouting channel is reserved in the contact face of the filling body and the surrounding rock; according to the method, through a dynamic correlation model of crustal stress real-time monitoring and propelling speed, the crustal stress change rate is directly converted into a mining parameter adjustment instruction, and dynamic balance of mining strength and crustal stress release is achieved; multi-dimensional data such as ground stress, micro-seismic energy and displacement are collected in real time, data are fused through an algorithm, rock mass mechanical parameters are inverted, a numerical model is optimized, and regulation and control schemes such as stope advancing speed, supporting parameters and filling ratio are automatically generated based on a model prediction result.

Description

Technical Field

[0001] The present invention relates to the field of mining technology, in particular to a safe mining process based on soft rock conditions. Background Art

[0002] In the soft rock mining scenario, geostress is the core driving force that causes rock deformation and destruction. Its dynamic changes play a direct and decisive role in the stability of the mining site. However, current technologies generally face prominent problems such as delayed dynamic response of geostress and lack of control means. It is impossible to perceive the geostress field in real time and it is difficult to achieve adaptive control. To effectively solve the above problems, a safe mining process based on soft rock conditions is proposed. Summary of the Invention

[0003] The present invention aims to solve the problems existing in the prior art and provides the following technical solutions:

[0004] The safe mining process based on soft rock conditions includes the following steps:

[0005] S1: Divide geological units and determine differentiated stope structural parameters through 3D geological modeling and geostress field measurement;

[0006] S2: Slice mining, where the stope advancement speed is dynamically adjusted based on real-time monitoring data of ground stress;

[0007] S3: Implement layered filling operations, pre-place shape memory alloy wire mesh in the filling body, and reserve grouting channels on the contact surface with the surrounding rock;

[0008] S4: Dynamically optimize support parameters and filling ratios based on microseismic monitoring, fiber optic sensing, and iterative updates of numerical simulation models;

[0009] S5: By integrating the physical information neural network (PINN) and genetic algorithm to invert the rock mechanical parameters, the numerical model is iteratively updated to ensure that the simulation accuracy residual is ≤1mm.

[0010] As an improvement to the above technical solution, the differentiated stope structure parameters are determined by the following formula:

[0011]

[0012] Wherein, k is the structural correction factor, ranging from 0.8 to 1.2; RMR is the rock mass quality score; E is the rock mass deformation modulus (GPa); γ is the overburden density (kN / m 3 ); H is the mining depth (m); σ max , σ min are the maximum and minimum principal stresses (MPa), respectively; P0 is atmospheric pressure (0.1 MPa).

[0013] As an improvement to the above technical solution, the layered filling in step S2 includes:

[0014] The lower layer is a cemented filling body with a thickness of ≥2m, a 28-day compressive strength of ≥4MPa, and an elastic modulus E1 of 0.2-0.5 times the elastic modulus of the surrounding rock;

[0015] The upper layer is a non-cemented filling body, with a thickness of 60%-80%. Ni-Ti alloy wire mesh is evenly distributed inside, with a wire diameter of 0.5-1mm and a mesh density of ≥4 wires / m 2 .

[0016] As an improvement to the above technical solution, in step S5: the rock mass mechanical parameters are inverted by fusing the physical information neural network (PINN) with the genetic algorithm, and the numerical model is iteratively updated. The specific steps are as follows:

[0017] S51: Multi-source data acquisition and preprocessing

[0018] Use displacement / stress sensors and microseismic monitoring systems to collect rock mass dynamic data in real time, and combine it with static parameters such as rock strength and elastic modulus obtained from geological exploration to construct a data set;

[0019] S52: Discretization of physical equations

[0020] The finite element method is used to discretize the equilibrium equations and constitutive equations into mesh node algebraic equations;

[0021] S53: PINN model construction and constraint setting

[0022] Design a neural network structure with displacement as the output variable, derive strain and stress through automatic differentiation, construct a loss function, integrate the data-driven term (the error between model prediction and measured data) and the physical constraint term (the residual of the equilibrium / constitutive equation), and adjust the proportion of the two through the weight coefficient α;

[0023] S54: Genetic algorithm to optimize initial parameters

[0024] The value range of the parameters to be inverted (such as elastic modulus and Poisson's ratio) is set, and the parameter population is randomly generated. Through iterative optimization of genetic operations such as selection, crossover, and mutation, the optimal initial parameter combination is screened to provide a high-quality starting point for PINN.

[0025] S55: Parameter Inversion and Model Iteration

[0026] The optimized initial parameters are input into the PINN model, and the model is trained with the monitoring data. The inverted parameters are substituted into the numerical model to calculate stress, displacement and other responses, and compared with the measured data. If the simulation residual is greater than 1mm, the data is updated and iterated again until the residual is ≤1mm.

[0027] As an improvement to the above technical solution, in step S53,

[0028] Data-driven items

[0029] Where N is the number of data samples, and are the predicted value and true value of the i-th sample respectively) and are used to measure the error between the model prediction value and the actual monitoring data;

[0030] Physical constraints M is the number of discrete nodes, r j is the residual of the physical equation at the jth node;

[0031] Final loss function α is the weight coefficient, ranging from 0 to 1.

[0032] As an improvement to the above technical solution, in step S2, the correlation model between the propulsion speed and the stress change rate is:

[0033]

[0034] v0: initial velocity;

[0035] λ: attenuation coefficient (set according to the rock mass brittleness index, the greater the brittleness, the larger the λ value);

[0036] When Δσ / Δt→0, v→v0; when Δσ / Δt increases, v decreases exponentially, forcing the mining rhythm to slow down.

[0037] As an improvement to the above technical solution, the dynamic adjustment in step S2 includes:

[0038] When the energy of microseismic events is greater than 1×10 3 J and when the frequency is greater than 5 times / h, the stope advancement speed is reduced to 50%-70% of the original value;

[0039] When the top plate displacement rate is greater than 3 mm / h, the anchor cables are activated for secondary tensioning to 150%-200% of the initial prestress.

[0040] As an improvement to the above technical solution, the numerical simulation model in step S2 adopts PFC3D discrete element and FLAC3D finite difference coupling modeling, and the rock mass parameters are updated every 5 m of advancement. When the displacement residual is greater than 1 mm for three consecutive iterations, the high-pressure splitting grouting reinforcement program is started, and the grouting pressure is ≥5 MPa.

[0041] Beneficial effects of the present invention:

[0042] Through the dynamic correlation model of real-time monitoring of ground stress and advancement speed, the ground stress change rate is directly converted into mining parameter adjustment instructions, achieving a dynamic balance between mining intensity and ground stress release;

[0043] Real-time collection of multi-dimensional data such as ground stress, microseismic energy, displacement, etc., data fusion through algorithms, inversion of rock mechanical parameters and optimization of numerical models, and automatic generation of control plans such as mining field advancement speed, support parameters, and filling ratio based on model prediction results. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0045] The safe mining process based on soft rock conditions includes the following steps:

[0046] S1: Divide geological units and determine differentiated stope structural parameters through 3D geological modeling and geostress field measurement;

[0047] S2: Slice mining, where the stope advancement speed is dynamically adjusted based on real-time monitoring data of ground stress;

[0048] S3: Implement layered filling operations, pre-place shape memory alloy wire mesh in the filling body, and reserve grouting channels on the contact surface with the surrounding rock;

[0049] S4: Dynamically optimize support parameters and filling ratios based on microseismic monitoring, fiber optic sensing, and iterative updates of numerical simulation models;

[0050] S5: By integrating the physical information neural network (PINN) and genetic algorithm to invert the rock mechanical parameters, the numerical model is iteratively updated to ensure that the simulation accuracy residual is ≤1mm.

[0051] The differentiated stope structure parameters are determined by the following formula:

[0052]

[0053] Wherein, k is the structural correction factor, ranging from 0.8 to 1.2; RMR is the rock mass quality score; E is the rock mass deformation modulus (GPa); γ is the overburden density (kN / m 3 ); H is the mining depth (m); σ max , σ min are the maximum and minimum principal stresses (MPa), respectively; P0 is atmospheric pressure (0.1 MPa);

[0054] The layered filling in step S2 includes:

[0055] The lower layer is a cemented filling body with a thickness of ≥2m, a 28-day compressive strength of ≥4MPa, and an elastic modulus E1 of 0.2-0.5 times the elastic modulus of the surrounding rock;

[0056] The upper layer is a non-cemented filling body, with a thickness of 60%-80%. Ni-Ti alloy wire mesh is evenly distributed inside, with a wire diameter of 0.5-1mm and a mesh density of ≥4 wires / m 2 ;

[0057] Specifically, S1 3D geological modeling: constructing a 3D geological model of the mining area through drilling data, geophysical exploration (such as seismic waves, electromagnetic waves), and other means to intuitively present the rock mass lithology and structure (faults, joints) distribution, and divide geological units with similar engineering characteristics; geostress field measurement: using hydraulic fracturing and stress relief methods to obtain the magnitude and direction of geostress, clarify the regional stress concentration area and principal stress direction; differential parameter calculation: using formulas (combining structural correction coefficient k, rock mass quality score RMR, geostress and other parameters) to customize stope structure parameters (such as span, height, and pillar size) for different geological units;

[0058] S2 stratified mining: The ore body is mined in layers (3-5m per layer) to reduce the disturbance range of a single excavation to the surrounding rock, reduce the exposed area of the roof, and facilitate timely support or filling. Dynamic monitoring of ground stress: Stress changes around the stope (such as maximum principal stress σ1 and stress change rate Δσ / Δt) are monitored in real time through stress sensors. Based on the stress change rate, the advancement speed is dynamically adjusted to achieve a dynamic balance of "stress release-mining-support" and prevent surrounding rock instability caused by excessive advancement. The correlation model between advancement speed and stress change rate is as follows:

[0059]

[0060] v0: initial velocity (e.g. 2m / d);

[0061] λ: attenuation coefficient (set according to the rock mass brittleness index, the greater the brittleness, the larger the λ value);

[0062] When Δσ / Δt→0, v→v0; when Δσ / Δt increases, v decreases exponentially, forcing the mining rhythm to slow down;

[0063] Layered filling design: Lower layer cemented filling: provides high-strength support (strength ≥ 4MPa), elastic modulus matches that of the surrounding rock (0.2-0.5 times), and suppresses bottom heave and side wall deformation;

[0064] The upper non-cemented filling body fills the space and reduces costs. A Ni-Ti shape memory alloy wire mesh is pre-installed inside. Utilizing its superelasticity and shape memory properties, the wire mesh is stretched and dissipated to dissipate energy when the surrounding rock deforms, thus preventing the filling body from cracking as a whole. A grouting channel is reserved: Grouting is later used to enhance the interfacial adhesion between the filling body and the surrounding rock, forming a "surrounding rock-filling body" collaborative bearing system, compensating for the insufficient self-stabilization capacity of the weak rock mass, and controlling the surrounding rock deformation in the long term.

[0065] S3 microseismic monitoring: Captures rock fracture signals and locates damaged areas; Fiber optic sensing: Real-time monitoring of strain distribution and identification of stress concentration areas; Numerical simulation: Based on current geological and mining parameters, predicts rock mass response (such as stress, displacement, and plastic zone) under different support / filling schemes; Enables real-time adaptive adjustment of support and filling parameters to reduce costs and improve safety;

[0066] Multi-source data input: Integrate displacement / stress monitoring data, microseismic signals, and geological exploration parameters to construct a training dataset.

[0067] Physical equation constraints: Discretize the rock mechanics equations (equilibrium equations, constitutive equations) and embed them into the PINN model to ensure that the inversion parameters conform to physical laws;

[0068] S5 genetic algorithm optimization: randomly generates a population of rock mass mechanical parameters (elastic modulus, Poisson's ratio, etc.), and selects the optimal initial parameters through selection, crossover, and mutation operations with the loss function as the target;

[0069] Iterative training: Optimized parameters are input into the PINN model for training. The inverse parameters are substituted into the numerical model to calculate stress and displacement, which are then compared with the measured data. If the simulation residual is greater than 1mm, the data is updated and iterated again until the accuracy meets the standard. This improves the numerical model's accuracy in fitting actual working conditions, providing a reliable prediction basis for optimizing mining plans.

[0070] Each step forms a closed loop through "monitoring data drive-intelligent algorithm analysis-engineering measures response".

[0071] In one embodiment, in step S5, the rock mass mechanical parameters are inverted by fusing the physical information neural network (PINN) with the genetic algorithm, and the numerical model is iteratively updated. The specific steps are as follows:

[0072] S51: Multi-source data acquisition and preprocessing

[0073] Use displacement / stress sensors and microseismic monitoring systems to collect rock mass dynamic data in real time, and combine it with static parameters such as rock strength and elastic modulus obtained from geological exploration to construct a data set;

[0074] S52: Discretization of physical equations

[0075] The finite element method is used to discretize the equilibrium equations and constitutive equations into mesh node algebraic equations;

[0076] S53: PINN model construction and constraint setting

[0077] Design a neural network structure with displacement as the output variable, derive strain and stress through automatic differentiation, construct a loss function, integrate the data-driven term (the error between model prediction and measured data) and the physical constraint term (the residual of the equilibrium / constitutive equation), and adjust the proportion of the two through the weight coefficient α;

[0078] S54: Genetic algorithm to optimize initial parameters

[0079] The value range of the parameters to be inverted (such as elastic modulus and Poisson's ratio) is set, and the parameter population is randomly generated. Through iterative optimization of genetic operations such as selection, crossover, and mutation, the optimal initial parameter combination is screened to provide a high-quality starting point for PINN.

[0080] S55: Parameter Inversion and Model Iteration

[0081] The optimized initial parameters are input into the PINN model, and the model is trained with the monitoring data. The inverted parameters are substituted into the numerical model to calculate stress, displacement and other responses, and compared with the measured data. If the simulation residual is greater than 1mm, the data is updated and iterated again until the residual is ≤1mm.

[0082] Specifically, S51: displacement / stress sensor: acquires deformation and stress changes on the tunnel surface or inside the borehole, reflecting the real-time response of the rock mass;

[0083] Microseismic monitoring system: Infer the evolution process of internal rock fracture through event location and energy release patterns.

[0084] Static data (geological exploration): rock strength (uniaxial compressive strength UCS, tensile strength), elastic modulus (laboratory core test), geological structural characteristics (fault distribution, joint density);

[0085] Provide high-quality, multi-dimensional input data for subsequent model training to ensure the authenticity of parameter inversion;

[0086] S52: Physics equations:

[0087] Balanced equation: The equilibrium relationship between the stress tensor σ and the body force f;

[0088] Constitutive equation: σ = D·ε, stress σ and strain ε are related through the elastic matrix D.

[0089] Finite element discretization: Divide the computational domain into grid elements, discretize the continuous physical field (displacement, stress) into nodal variables through shape functions, and transform them into a set of algebraic equations: K u = F, where K is the stiffness matrix, u is the nodal displacement vector, and F is the nodal force vector. This transforms the basic laws of rock mechanics into a computable mathematical form as the physical constraints of the PINN model.

[0090] S53: PINN model construction and constraint setting;

[0091] Neural network structure:

[0092] Input layer: coordinate points (x, y, z), time t, known boundary conditions;

[0093] Hidden layer: a multi-layer fully connected network that captures nonlinear relationships through activation functions;

[0094] Output layer: displacement components (ux,uy,uz).

[0095] Automatic differentiation to derive stress and strain: Based on the displacement field u, the strain is calculated by automatic differentiation Then the stress σ is calculated according to the constitutive equation,

[0096] Loss Function

[0097] Data-driven items Where N is the number of data samples, and are the predicted value and true value of the i-th sample respectively) and are used to measure the error between the model prediction value and the actual monitoring data;

[0098] Physical constraints Where M is the number of discrete nodes, r j is the residual of the physical equation at the jth node;

[0099] Through physical constraints, the neural network not only fits the data but also follows the basic laws of rock mechanics, improving the extrapolation ability and the rationality of parameter inversion;

[0100] S54: Encode the parameters to be inverted (such as elastic modulus E, Poisson's ratio ν, internal friction angle φ) as chromosomes, and the fitness function is: in is the loss function of PINN, the higher the fitness, the better the parameter combination;

[0101] Genetic Operations:

[0102] Selection: Select individuals with high fitness;

[0103] Crossover: exchange chromosome segments to generate new parameter combinations;

[0104] Mutation: Randomly perturb some genes to avoid falling into local optimality.

[0105] Iterative optimization: Repeat the selection-crossover-mutation process until the population converges, and output the optimal parameter combination as the initial value of PINN.

[0106] S55: Input the parameters optimized by the genetic algorithm into PINN and minimize the loss function by gradient descent Update the neural network weights and parameters to be inverted;

[0107] Substitute the inversion parameters (e.g., E = 2.5 GPa, ν = 0.35) into the numerical model (e.g., FLAC3D) to calculate the stress and displacement fields;

[0108] Residual verification: Compare the numerical simulation results with the measured data (such as key point displacements) and calculate the residuals: Where K is the total number of monitoring points or data sample size, u sim (x k ) is the numerical simulation displacement value of the kth monitoring point, u obs (x k ) is the actual observed displacement value of the kth monitoring point. If the residual is greater than 1 mm, update the monitoring data and repeat S53-S55 until the residual is ≤ 1 mm.

[0109] In one embodiment, the dynamic adjustment in step S2 includes:

[0110] When the energy of microseismic events is greater than 1×10 3 J and when the frequency is greater than 5 times / h, the stope advancement speed is reduced to 50%-70% of the original value;

[0111] When the top plate displacement rate is greater than 3 mm / h, the anchor cables are activated for secondary tensioning to 150%-200% of the initial prestress;

[0112] Microseismic events usually occur before significant displacement. By reducing the advancement speed, the further development of the rupture can be suppressed and prevented from evolving into large deformation. When the displacement rate exceeds the limit, indicating that significant deformation has occurred, the secondary tensioning anchor cable can quickly control the displacement and prevent the accident from escalating. Microseismic positioning can identify high-risk areas. Combined with displacement monitoring to determine the specific location, precise control of "regional early warning + local enhanced support" is achieved.

[0113] In one embodiment, the numerical simulation model in step S2 is modeled by coupling PFC3D discrete element and FLAC3D finite difference, and the rock mass parameters are updated every 5 m of advancement. When the displacement residual is greater than 1 mm for three consecutive iterations, the high-pressure splitting grouting reinforcement program is started, and the grouting pressure is ≥5 MPa.

[0114] Mining in fault fracture zones: When the stope approaches a fault, PFC3D simulates block sliding within the fault zone, while FLAC3D calculates the activation effect of mining stress on the fault. The coupled model predicts slip risk and triggers grouting to reinforce the fault zone.

[0115] High-stress soft rock tunnel: During excavation, FLAC3D monitored the expansion of the plastic zone in the sidewall, and PFC3D simultaneously simulated the development of cracks into the depths. After the residual exceeded the limit, grouting was used to form a "stone body-rock mass" composite bearing structure. PFC3D and FLAC3D coupled modeling combined with a dynamic grouting mechanism achieved the unification of continuous deformation analysis and discrete fracture prediction in soft rock mining. The rock failure mode was evaluated in real time through displacement residuals, accurately triggering grouting reinforcement.

[0116] The above are all preferred embodiments of the present application, and are not intended to limit the scope of protection of the present application. Therefore, any equivalent changes made based on the structure, shape, and principle of the present application should be included in the scope of protection of the present application.

Claims

1. A safe mining process based on soft rock conditions, characterized by: The following steps are involved: S1: Divide geological units and determine differentiated stope structural parameters through 3D geological modeling and geostress field measurement; S2: Slice mining, where the stope advancement speed is dynamically adjusted based on real-time monitoring data of ground stress; S3: Implement layered filling operations, pre-place shape memory alloy wire mesh in the filling body, and reserve grouting channels on the contact surface with the surrounding rock; S4: Dynamically optimize support parameters and filling ratios based on microseismic monitoring, fiber optic sensing, and iterative updates of numerical simulation models; S5: By integrating the physical information neural network (PINN) and genetic algorithm to invert the rock mechanical parameters, the numerical simulation model is iteratively updated to ensure that the simulation accuracy residual is ≤1mm.

2. The safe mining process under weak rock conditions according to claim 1 is characterized in that: The differentiated stope structure parameters are determined by the following formula: Wherein, k is the structural correction factor, ranging from 0.8 to 1.2; RMR is the rock mass quality score; E is the rock mass deformation modulus (GPa); γ is the overburden density (kN / m 3 ); H is the mining depth (m); σ max , σ min are the maximum and minimum principal stresses (MPa), respectively; P0 is atmospheric pressure (0.1MPa).

3. The safe mining process based on weak rock conditions according to claim 1 is characterized in that: The layered filling in step S2 includes: The lower layer is a cemented filling body with a thickness of ≥2m, a 28-day compressive strength of ≥4MPa, and an elastic modulus E1 of 0.2-0.5 times the elastic modulus of the surrounding rock; The upper layer is a non-cemented filling body, with a thickness of 60%-80%. Ni-Ti alloy wire mesh is evenly distributed inside, with a wire diameter of 0.5-1mm and a mesh density of ≥4 wires / m 2 .

4. The safe mining process under weak rock conditions according to claim 1 is characterized in that: In step S5, the rock mass mechanical parameters are inverted by fusing the physical information neural network (PINN) with the genetic algorithm, and the numerical simulation model is iteratively updated. The specific steps are as follows: S51: Multi-source data acquisition and preprocessing Use displacement / stress sensors and microseismic monitoring systems to collect real-time rock mass dynamic data, and combine it with static parameters such as rock strength and elastic modulus obtained from geological exploration to construct a data set; S52: Discretization of physical equations The finite element method is used to discretize the equilibrium equations and constitutive equations into mesh node algebraic equations; S53: PINN model construction and constraint setting Design a neural network structure, use displacement as the output variable, derive strain and stress through automatic differentiation, construct a loss function, integrate data-driven terms and physical constraints, and adjust the proportion of the two through the weight coefficient α; S54: Genetic algorithm to optimize initial parameters Set the value range of the parameters to be inverted, randomly generate parameter populations, and iteratively optimize through genetic operations such as selection, crossover, and mutation to screen out the optimal initial parameter combination, providing a high-quality starting point for PINN; S55: Parameter Inversion and Model Iteration The optimized initial parameters are input into the PINN model, and the model is trained with the monitoring data. The inverted parameters are substituted into the numerical model to calculate stress, displacement and other responses, and compared with the measured data. If the simulation residual is greater than 1mm, the data is updated and iterated again until the residual is ≤1mm.

5. The safe mining process based on weak rock conditions according to claim 4 is characterized in that: In step S53, the data driven item Where N is the number of data samples, and are the predicted value and true value of the i-th sample respectively) and are used to measure the error between the model prediction value and the actual monitoring data; Physical constraints M is the number of discrete nodes, r j is the residual of the physical equation at the jth node; Final loss function α is the weight coefficient, ranging from 0 to 1.

6. The safe mining process under weak rock conditions according to claim 1 is characterized in that: In step S2, the correlation model between the advancing speed and the stress change rate is: v0: initial velocity (e.g. 2m / d); λ: attenuation coefficient (set according to the rock mass brittleness index, the greater the brittleness, the larger the λ value); When Δσ / Δt→0, v→v0; when Δσ / Δt increases, v decreases exponentially, forcing the mining rhythm to slow down.

7. The safe mining process under weak rock mass conditions according to claim 1 is characterized in that: The dynamic adjustment in step S2 includes: When the energy of microseismic events is greater than 1×10 3 J and when the frequency is greater than 5 times / h, the stope advancement speed is reduced to 50%-70% of the original value; When the top plate displacement rate is greater than 3 mm / h, the anchor cables are activated for secondary tensioning to 150%-200% of the initial prestress.

8. The safe mining process under weak rock conditions according to claim 1 is characterized in that: The numerical simulation model in step S2 is modeled by coupling PFC3D discrete element and FLAC3D finite difference. The rock mass parameters are updated every 5 m of advancement. When the displacement residual is greater than 1 mm for three consecutive iterations, the high-pressure splitting grouting reinforcement program is started, and the grouting pressure is ≥5 MPa.

Images