Method and system for blending filling materials for controlling open slope of high and cold mining area
By measuring the performance of backfill materials in open-pit coal mines in high-altitude and cold regions and optimizing them through data-driven deep learning, combined with digital twin technology for dynamic adjustment, the instability problem of backfill material ratio design in high-altitude and cold environments has been solved, and the stability control and safety assurance of slopes have been achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-21
- Publication Date
- 2026-04-10
AI Technical Summary
In open-pit coal mining in cold and high-altitude areas, the hydration reaction rate of backfill materials decreases and the early strength development is slow, making it difficult to meet the requirements for slope stability control. The existing mix design lacks a systematic adaptation method, resulting in unstable backfilling effect and difficulty in ensuring slope safety.
By conducting basic performance tests on the filling materials and establishing a set of performance parameters, and combining deep hybrid neural networks and multi-objective Bayesian optimization frameworks, we can screen out the mix proportions that meet the requirements of slope bearing capacity and freeze-thaw resistance. We can then use a digital twin-driven adaptive model to make dynamic adjustments and achieve precise allocation.
It significantly improves the strength development capacity and structural stability of the backfill under low temperature conditions, reduces the risk of slope instability, improves the scientificity and reliability of the backfill material ratio, and enhances the stability and adaptability of backfill operations in open-pit mines.
Smart Images
Figure CN121836036A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of open-pit mine slope engineering and backfilling mining technology, and in particular to a method and system for preparing backfilling materials for open-pit slope control in high-altitude and cold mining areas. Background Technology
[0002] In open-pit coal mining in high-altitude and cold regions, large-scale goaf spaces are easily formed after the side coal is mined, weakening the integrity of the original rock mass and easily leading to slope deformation and instability risks. Backfilling these goaf areas using backfilling technology is an important means to improve the overall stability of open-pit mine slopes, reduce stripping, and achieve green mining. However, in high-altitude and cold mining areas, the environment is often at low or even sub-zero temperatures, significantly reducing the hydration rate of the backfill material, resulting in slow early strength development, and making it difficult to effectively densify the internal pore structure of the backfill, thus failing to meet the requirements for load-bearing and deformation constraints for open-pit slope stability control.
[0003] Current open-pit mine backfill material mix design is mostly based on empirical parameters under normal or moderately cold conditions, lacking a systematic adaptation method for high-altitude and cold environments. This is particularly problematic under freeze-thaw cycles, leading to issues such as strength attenuation and permeability degradation in the backfill. Furthermore, traditional mix design methods often rely on static tests and manual adjustments, failing to comprehensively consider the coupling relationship between material property differences, environmental temperature variations, and slope stability control requirements. They also lack on-site dynamic feedback and real-time optimization capabilities, resulting in unstable backfill effects and jeopardizing long-term slope safety. Therefore, a technical method is urgently needed that addresses the control objectives of open-pit slopes in high-altitude and cold mining areas, enabling precise mixing and dynamic control of backfill material proportions. Summary of the Invention
[0004] This solution addresses the problems and needs raised above by proposing a method and system for the allocation of backfill materials for controlling open-pit slopes in high-altitude and cold mining areas. The above-mentioned technical objectives can be achieved by adopting the following technical features, and other technical effects are also brought about.
[0005] One objective of this invention is to provide a method for preparing backfill materials for controlling open-pit slopes in high-altitude and cold mining areas, comprising the following steps: S10: Conduct basic performance tests on backfill materials used for backfilling coal at the side of open-pit mines, obtain their physical, chemical and mechanical property parameters, and establish a set of performance parameters for backfill materials; S20: Based on different material ratio schemes, conduct indoor mechanical and durability tests to obtain the performance indicators of the filling body under low temperature and freeze-thaw conditions, and form a performance test database of different material ratios; S30: Based on the requirements for stability control of open-pit slopes in high-altitude and cold mining areas, the test results were analyzed to screen the range of fill material proportions that meet the requirements for slope bearing capacity, impermeability and freeze-thaw resistance; S40: Based on the screening of the filling material ratio range, a deep hybrid neural network model based on the attention mechanism is constructed, and the optimal ratio parameters are obtained by using a multi-objective Bayesian optimization framework; S50: The optimized mix proportion parameters are applied to the on-site filling operation control system. Through the digital twin-driven adaptive model predictive control system, the mix proportion is dynamically adjusted to achieve precise allocation under the control target of the open slope.
[0006] Furthermore, the filling material preparation method and system for controlling open-pit slopes in high-altitude and cold mining areas according to the present invention may also have the following technical features: In one example of the present invention, in step S10, the physical, chemical and mechanical properties parameters specifically include: aggregate particle size distribution, density, specific surface area of cementitious material, activity index, initial setting time of filling material, final setting time of filling material, and compressive strength at room temperature.
[0007] In one example of the present invention, step S30 further includes introducing open slope control correlation constraints, linking the filling body parameters with the slope stability target for screening, and the linkage screening includes at least: S31: Using the slope safety factor Fs≥Fs0 as a constraint, the target range of the cohesion c of the filling body and the internal friction angle φ is used as the screening condition. S32: Using the slope displacement Δ≤Δ0 as a constraint, the target range of the deformation modulus E or compression modulus M of the filling body is used as the screening condition. Among them, Fs0 is set to 1.20 to 1.50, and Δ0 is set to 10 to 50 mm.
[0008] In one example of the present invention, in step S40, the filling material proportioning parameters include at least the water-cement ratio, the amount of cementitious material, the amount of admixture, and the aggregate gradation parameters.
[0009] In one example of the present invention, in step S40, the deep hybrid neural network model based on the attention mechanism includes an input layer, a one-dimensional convolutional layer, a long short-term memory network layer, an attention mechanism layer, and a fully connected output layer connected in sequence. The input layer is configured to receive parameter vectors that affect the performance of the filling material to form input features; The one-dimensional convolutional layer is configured to perform local perception and fusion of input features, and extract local interaction patterns between different material parameters; The Long Short-Term Memory (LSTM) network layer is configured to capture the dynamic patterns of infill performance over time, addressing long-term dependency issues. Internally, it utilizes a forgetting gate... Input gate Output gate and cell state and hidden state Collaborative updates; The attention mechanism layer is configured to adaptively handle the hidden states of the Long Short-Term Memory (LSTM) network layer at different times. Assign appropriate weights to focus on the historical information segments that are most critical to the current prediction target; The fully connected output layer is configured to aggregate a context vector containing key information. Mapped to the final predicted performance metric.
[0010] In one example of the present invention, step S40, obtaining the optimal ratio parameters using a multi-objective Bayesian optimization framework, specifically includes the following steps: S41: Proxy Model Construction: For each performance objective that needs optimization, such as strength, penetration coefficient, and cost. A Gaussian process regression model is established as its surrogate model, and its expression is as follows: In the formula, It is a mean function. It is the covariance function; S42: Acquisition Functions and Search: Through Measure and evaluate a new candidate point The goal is to achieve hypervolume gain for the current Pareto front, and the next evaluation point will be selected by iteratively optimizing and maximizing the EHVI. To efficiently approximate the real Pareto frontier; among them, The expression is: In the formula, the hypervolume HV is the spatial volume enclosed by the Pareto front and a defined reference point; S43: Termination and Output: Repeat the iterative process until the preset number of evaluations or convergence conditions are reached, and finally output a set of Pareto optimal solutions; select an appropriate ratio from the solution set as the final solution according to the actual needs of emphasizing strength or cost.
[0011] In one example of the present invention, the detailed steps of the iterative optimization in step S42 are as follows: S421: Initialization: Using a small amount of experimental data from the initial proportioning scheme, initialize the Gaussian process proxy model for each objective; S422: Model Update: Update each agent model based on all currently evaluated datasets; S423: Candidate point selection: through optimization The acquisition function selects the point most likely to improve the current Pareto front from the feasible matching parameter space. ; S424: Performance Evaluation: [Regarding...] Conduct real-world physics experiments to obtain its multi-objective performance value f( ); S425: Data Expansion: Adding new data pairs ( ,f( Add it to the evaluation dataset.
[0012] In one example of the present invention, in step S50, the dynamic adjustment of the digital twin-driven adaptive model prediction control system specifically includes the following steps: S51: Constructing a Digital Twin: A high-fidelity model that fully maps to the physical filling site is built in virtual space. This model integrates multi-physics processes including slope rock mechanics, hydration and hardening of the filling material, heat conduction, and freeze-thaw damage. Its governing equations are expressed as follows: In the formula, These are the mass, damping, and stiffness matrices of the system, respectively. It is a displacement vector. For temperature ,stress and time Related load vectors; S52: Constructing the core of adaptive model predictive control, specifically including: Online system identification: Quickly predict the system state. The expression for predicting the system state is: In the formula, For system status, To control the input, For observation output, and Process noise and measurement noise; model parameters It is not fixed, but rather uses recursive least squares with a forgetting factor for online real-time identification and updating, and its expression is: In the formula, For the parameter estimation vector, For data vectors, forgetting factor A value of 0.95 to 0.99 is used to enable the model to track dynamic changes in the system; Rolling time-domain optimization: in each control cycle The controller estimates based on the current state. Using the initial conditions and the updated prediction model, we solve a finite-time optimal control problem, expressed as: In the formula, To predict the time domain, To control the time domain, and This is the weight matrix. Set values for performance targets; Feedback closed-loop execution: The first control increment in the optimization sequence - The water-cement ratio fine-tuning amount and the instantaneous addition amount of admixtures are sent to the automated batching and mixing equipment on site for execution; when entering the next control cycle, the new system output measurement values are collected. The value is compared with the predicted value, the error is used to correct the state estimation, and the above "identification-optimization-execution" process is repeated to form an adaptive closed-loop control system.
[0013] Another objective of this invention is to provide a filling material preparation system for open-pit slope control in high-altitude and cold mining areas, comprising: The performance parameter set establishment module is configured to perform basic performance tests on backfill materials used for backfilling coal at the side of open-pit mines, obtain their physical, chemical and mechanical property parameters, and establish a performance parameter set for backfill materials. The performance test database module is configured to conduct indoor mechanical and durability tests based on different material ratio schemes, obtain the performance indicators of the filling body under low temperature and freeze-thaw conditions, and form a performance test database with different material ratios. The backfill material screening module is configured to analyze test results in conjunction with the stability control requirements of open-pit slopes in high-altitude and cold mining areas, and screen the backfill material mix ratio range that meets the requirements of slope bearing capacity, impermeability and freeze-thaw resistance. The optimal ratio parameter acquisition module is configured to construct a deep hybrid neural network model based on the attention mechanism based on the ratio range of the screened filling materials, and obtain the optimal ratio parameters using a multi-objective Bayesian optimization framework. The dynamic adjustment and precise allocation module is configured to apply the optimized mix ratio parameters to the on-site filling operation control system. Through the digital twin-driven adaptive model predictive control system, the mix ratio is dynamically adjusted to achieve precise allocation under the control target of the open slope.
[0014] In one example of the present invention, the filling material screening module includes: The first constraint unit is configured with the slope safety factor Fs≥Fs0 as the constraint, and the target range of the cohesion c of the filling body and the internal friction angle φ as the screening condition; where Fs0 is taken as 1.20~1.50; The second constraint unit is configured to use the slope displacement Δ≤Δ0 as a constraint and the target range of the deformation modulus E or compression modulus M of the filling body as a screening condition; where Δ0 is a set value of 10 to 50 mm.
[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention addresses the characteristics of low temperature and freeze-thaw environment in high-altitude mining areas by constructing a precise method for the preparation of backfill materials with the goal of controlling the stability of open-pit slopes. This method can significantly improve the strength development capacity and structural stability of the backfill body under low temperature conditions, and effectively reduce the risk of slope instability.
[0016] 2. By systematically measuring the physical, chemical, and mechanical properties of the filling material and combining the results of indoor mechanical and durability tests for proportioning, this invention realizes the transformation of filling material proportioning from experience-based design to data-driven design, thereby improving the scientificity and reliability of proportioning determination.
[0017] 3. This invention introduces a mathematical model or machine learning model and combines iterative optimization algorithms to accurately optimize the proportion of filling materials. This can meet the requirements of slope bearing capacity, impermeability and freeze-thaw resistance, while taking into account material utilization efficiency and engineering applicability.
[0018] 4. Through an automatic monitoring and feedback control mechanism, this invention can dynamically adjust the mix ratio according to the on-site temperature, slurry state and slope response, thereby improving the stability and adaptability of open-pit mine backfilling operations in cold and high-altitude mining areas and possessing good engineering promotion value.
[0019] The preferred embodiments of the invention will be described in more detail below with reference to the accompanying drawings, so as to facilitate an understanding of the features and advantages of the invention. Attached Figure Description
[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. The drawings are merely illustrative of some embodiments of the present invention and are not intended to limit the scope of the present invention to all embodiments.
[0021] Figure 1 This is a flowchart of a method for precisely preparing filling materials according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating the precise control and proportioning process according to an embodiment of the present invention. Detailed Implementation
[0022] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. The same reference numerals in the drawings represent the same components. It should be noted that 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 scope of protection of the present invention.
[0023] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms “first,” “second,” and similar terms used in this patent application specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, “an” or “a” and similar terms do not necessarily indicate a quantity limitation. Terms such as “comprising” or “including” mean that the element or object preceding the word encompasses the element or object listed following the word and its equivalents, without excluding other elements or objects. Terms such as “connected” or “linked” are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as “upper,” “lower,” “left,” and “right” are used only to indicate relative positional relationships; these relative positional relationships may change accordingly when the absolute position of the described object changes.
[0024] According to a first aspect of the present invention, a method for preparing backfill material for controlling open-pit slopes in high-altitude and cold mining areas, such as... Figure 1 As shown, it includes the following steps: S10: Conduct basic performance tests on backfill materials used for backfilling coal at the side of open-pit mines, obtain their physical, chemical and mechanical property parameters, and establish a set of performance parameters for backfill materials; S20: Based on different material ratio schemes, conduct indoor mechanical and durability tests to obtain the performance indicators of the filling body under low temperature and freeze-thaw conditions, and form a performance test database of different material ratios; S30: Based on the requirements for stability control of open-pit slopes in high-altitude and cold mining areas, the test results were analyzed to screen the range of fill material proportions that meet the requirements for slope bearing capacity, impermeability and freeze-thaw resistance; S40: Based on the screening of the filling material ratio range, a deep hybrid neural network model based on the attention mechanism is constructed, and the optimal ratio parameters are obtained by using a multi-objective Bayesian optimization framework; S50: The optimized mix proportion parameters are applied to the on-site filling operation control system. Through the digital twin-driven adaptive model predictive control system, the mix proportion is dynamically adjusted to achieve precise allocation under the control target of the open slope.
[0025] This formulation method is designed for the low temperature and freeze-thaw environment characteristics of high-altitude mining areas. It is a precise formulation method for backfill materials with the goal of controlling the stability of open-pit slopes. It can significantly improve the strength development capacity and structural stability of the backfill body under low temperature conditions and effectively reduce the risk of slope instability.
[0026] This formulation method, through systematic determination of the physical, chemical and mechanical properties of the filling material, and combined with the results of indoor mechanical and durability tests for proportion screening, realizes the transformation of filling material proportion design from experience-based design to data-driven design, thereby improving the scientificity and reliability of proportion determination.
[0027] This formulation method introduces mathematical or machine learning models and combines them with iterative optimization algorithms to precisely optimize the proportion of filling materials. It can meet the requirements of slope bearing capacity, impermeability and freeze-thaw resistance while taking into account material utilization efficiency and engineering applicability.
[0028] This mixing method, through an automatic monitoring and feedback control mechanism, enables dynamic adjustment of the mix ratio based on on-site temperature, slurry condition, and slope response, thereby improving the stability and adaptability of open-pit mine backfilling operations in cold and high-altitude mining areas and possessing significant engineering promotion value.
[0029] In one example of the present invention, in step S10, the filling material includes: aggregate, cementing material, water and functional additives.
[0030] In one example of the present invention, in step S10, the physical, chemical and mechanical properties parameters specifically include: aggregate particle size distribution, density, specific surface area of cementitious material, activity index, initial setting time of filling material, final setting time of filling material, and compressive strength at room temperature.
[0031] In one example of the present invention, step S20, conducting indoor mechanical and durability testing includes the following steps: S21: Compressive strength test, with test results for 7 days, 28 days and low-temperature curing conditions; S22: Permeability coefficient test, used to evaluate the impermeability of the filling material; S23: Freeze-thaw cycle test, with a freeze-thaw temperature range of −20℃ to +20℃ and a cycle count of 10 to 50 times.
[0032] In one example of the present invention, in step S30, the slope stability control requirements include: the 28-day compressive strength of the backfill body is not less than 3-5 MPa, and the permeability coefficient is less than 1×10⁻. 6cm / s and the strength loss rate after freeze-thaw cycles is no more than 20%.
[0033] In one example of the present invention, step S30 further includes introducing open slope control correlation constraints, linking the filling body parameters with the slope stability target for screening, and the linkage screening includes at least: S31: Using the slope safety factor Fs≥Fs0 as a constraint, the target range of the cohesion c of the filling body and the internal friction angle φ is used as the screening condition. S32: Using the slope displacement Δ≤Δ0 as a constraint, the target range of the deformation modulus E or compression modulus M of the filling body is used as the screening condition. Among them, Fs0 is set to 1.20 to 1.50, and Δ0 is set to 10 to 50 mm.
[0034] In one example of the present invention, in step S40, the filling material proportioning parameters include at least the water-cement ratio, the amount of cementitious material, the amount of admixture, and the aggregate gradation parameters.
[0035] In one example of the present invention, in step S40, the deep hybrid neural network model based on the attention mechanism includes an input layer, a one-dimensional convolutional layer, a long short-term memory network layer, an attention mechanism layer, and a fully connected output layer connected in sequence. The input layer is configured to receive parameter vectors that affect the performance of the filling material to form input features; The one-dimensional convolutional layer is configured to perform local perception and fusion of input features, and extract local interaction patterns between different material parameters; The Long Short-Term Memory (LSTM) network layer is configured to capture the dynamic patterns of infill performance over time, addressing long-term dependency issues. Internally, it utilizes a forgetting gate... Input gate Output gate and cell state and hidden state Collaborative updates; The attention mechanism layer is configured to adaptively handle the hidden states of the Long Short-Term Memory (LSTM) network layer at different times. Assign appropriate weights to focus on the historical information segments that are most critical to the current prediction target; The fully connected output layer is configured to aggregate a context vector containing key information. Mapped to the final predicted performance metric.
[0036] Specifically, this model is used to establish a high-precision, interpretable nonlinear mapping relationship from the proportioning parameters of filling materials to key performance indicators. Its purpose is to achieve accurate prediction of the performance of complex material systems under extreme environments through deep learning technology. Its model structure and working principle are as follows: The model adopts a hierarchical fusion architecture, which includes an input layer, a one-dimensional convolutional layer, a long short-term memory network layer, an attention mechanism layer, and a fully connected output layer.
[0037] The input layer receives the matching parameter vector X=[x1, x2, ..., x... m The components represent key variables affecting the performance of the filling material, such as water-cement ratio, cementitious material content, aggregate gradation parameters, admixture content, curing temperature, and curing age.
[0038] The purpose of a series of convolutional layers is to perform local perception and fusion of input features, extracting local interaction patterns between different material parameters. Its computational process can be represented as follows: in, and These are the kernel weights and biases, respectively. This is the activation function.
[0039] Long Short-Term Memory (LSTM) network layers are used to capture the dynamic patterns of infiller performance over time, addressing the problem of long-term dependency. Internally, they utilize forgetting gates... Input gate Output gate and cell state and hidden state Collaborative updates. The specific update mechanism is as follows: in, for Activation function It represents the Hadamah accumulation. , , , and , , , These are trainable parameters.
[0040] The core function of the attention mechanism layer is to adaptively assign hidden states of the LSTM at different time steps. Assign appropriate weights to focus on historical information segments that are most critical to the current prediction target. Attention weights. Calculation and context vector The generation process is as follows: in, , and For learnable parameters, context vectors It integrates the key parts of all time-series information.
[0041] The fully connected output layer will aggregate the context vector containing key information. Mapped to the final predicted performance metric: in, These represent the model-predicted compressive strength, permeability coefficient, and strength loss rate after freeze-thaw cycles, respectively.
[0042] Model Application: The model is trained using the "performance test database" established in step S20. Model parameters are optimized by minimizing the error between predicted and actual test values. After training, the model can quickly and accurately predict the performance of filling materials with any given mix proportions, and the attention weights can be visualized. To analyze the impact of different input characteristics and maintenance stages on the final performance.
[0043] In one example of the present invention, step S40, obtaining the optimal ratio parameters using a multi-objective Bayesian optimization framework, specifically includes the following steps: S41: Proxy Model Construction: For each performance objective that needs optimization, such as strength, penetration coefficient, and cost. A Gaussian process regression model is established as its surrogate model, and its expression is as follows: In the formula, It is a mean function. The covariance function is used; the Matrn kernel is typically chosen to accommodate the non-smoothness that may exist in engineering data. The Gaussian process model provides the predicted mean of the objective function value and gives the variance of the prediction uncertainty.
[0044] S42: Acquisition Functions and Search: Through Measure and evaluate a new candidate point The goal is to achieve hypervolume gain for the current Pareto front, and the next evaluation point will be selected by iteratively optimizing and maximizing the EHVI. To efficiently approximate the real Pareto frontier; among them, The expression is: In the formula, the hypervolume HV is the spatial volume enclosed by the Pareto front and a set reference point, which is an important indicator for measuring the quality of multi-objective solution sets.
[0045] S43: Termination and Output: Repeat the iterative process until the preset number of evaluations or convergence conditions are reached, and finally output a set of Pareto optimal solutions; select an appropriate ratio from the solution set as the final solution according to actual needs such as emphasis or cost.
[0046] In one example of the present invention, the detailed steps of the iterative optimization in step S42 are as follows: S421: Initialization: Using a small amount of experimental data from the initial proportioning scheme, initialize the Gaussian process proxy model for each objective; S422: Model Update: Update each agent model based on all currently evaluated datasets; S423: Candidate point selection: through optimization The acquisition function selects the point most likely to improve the current Pareto front from the feasible matching parameter space. ; S424: Performance Evaluation: [Regarding...] Conduct real-world physics experiments to obtain its multi-objective performance value f( ); S425: Data Expansion: Adding new data pairs ( ,f( Add it to the evaluation dataset.
[0047] In one example of the present invention, in step S50, the dynamic adjustment of the digital twin-driven adaptive model prediction control system specifically includes the following steps: S51: Constructing a Digital Twin: A high-fidelity model is built in virtual space that fully maps to the physical filling site. This model integrates multiple physical processes, including slope rock mechanics, hydration and hardening of the filling material, heat conduction, and freeze-thaw damage. Its governing equations can be abstractly represented as: In the formula, These are the mass, damping, and stiffness matrices of the system, respectively. It is a displacement vector. For temperature ,stress and time The relevant load vectors; through a sensor network deployed on the physical site, the slope deformation, the internal state of the filling body and environmental parameters are collected in real time, and the digital twin is driven to update synchronously to ensure that the virtual model and the physical entity are consistent.
[0048] S52: Constructing the core of adaptive model predictive control, specifically including: Online system identification: Quickly predict the system state. The expression for predicting the system state is: In the formula, For system status, To control the input, For observation output, and Process noise and measurement noise; model parameters It is not fixed, but rather uses recursive least squares with a forgetting factor for online real-time identification and updating, and its expression is: In the formula, For the parameter estimation vector, For data vectors, forgetting factor A value of 0.95 to 0.99 is used to enable the model to track dynamic changes in the system; Rolling time-domain optimization: in each control cycle The controller estimates based on the current state. Using the initial conditions and the updated prediction model, we solve a finite-time optimal control problem, expressed as: In the formula, To predict the time domain, To control the time domain, and This is the weight matrix. Set a value for the performance target; the optimization problem aims to find a sequence of control actions that make the future output of the system as close as possible to the desired target, while the control changes are gradual.
[0049] Feedback closed-loop execution: The first control increment in the optimization sequence - The water-cement ratio fine-tuning amount and the instantaneous addition amount of admixtures are sent to the automated batching and mixing equipment on site for execution; when entering the next control cycle, the new system output measurement values are collected. The value is compared with the predicted value, the error is used to correct the state estimation, and the above "identification-optimization-execution" process is repeated to form an adaptive closed-loop control system.
[0050] In other words, the digital twin-driven adaptive model predictive control system establishes an intelligent control system that integrates virtual and real worlds, enables real-time data interaction, and achieves closed-loop optimization throughout the entire process. Its core objective is to achieve real-time monitoring, early prediction, and precise control of the filling process for open-pit slopes in high-altitude mining areas, ensuring that the performance of the filling material consistently meets the requirements for slope stability.
[0051] In one example of the present invention, in step S50, the system may optionally adopt a distributed intelligent decision-making architecture based on federated learning, wherein local nodes in multiple mining areas use local data to train models, and a central server securely aggregates model parameters to update the global model, thereby realizing cross-mining area collaborative optimization and knowledge sharing.
[0052] According to a second aspect of the present invention, a filling material preparation system for open-pit slope control in high-altitude and cold mining areas includes: The performance parameter set establishment module is configured to perform basic performance tests on backfill materials used for backfilling coal at the side of open-pit mines, obtain their physical, chemical and mechanical property parameters, and establish a performance parameter set for backfill materials. The performance test database module is configured to conduct indoor mechanical and durability tests based on different material ratio schemes, obtain the performance indicators of the filling body under low temperature and freeze-thaw conditions, and form a performance test database with different material ratios. The backfill material screening module is configured to analyze test results in conjunction with the stability control requirements of open-pit slopes in high-altitude and cold mining areas, and screen the backfill material mix ratio range that meets the requirements of slope bearing capacity, impermeability and freeze-thaw resistance. The optimal ratio parameter acquisition module is configured to construct a deep hybrid neural network model based on the attention mechanism based on the ratio range of the screened filling materials, and obtain the optimal ratio parameters using a multi-objective Bayesian optimization framework. The dynamic adjustment and precise allocation module is configured to apply the optimized mix ratio parameters to the on-site filling operation control system. Through the digital twin-driven adaptive model predictive control system, the mix ratio is dynamically adjusted to achieve precise allocation under the control target of the open slope.
[0053] This material preparation system is designed for the low temperature and freeze-thaw environment of high-altitude mining areas. It is a precise material preparation method for open-pit slope stability control. It can significantly improve the strength development and structural stability of the backfill under low temperature conditions and effectively reduce the risk of slope instability.
[0054] This formulation system, through systematic measurement of the physical, chemical and mechanical properties of filling materials and combined with indoor mechanical and durability test results for proportion screening, realizes the transformation of filling material proportion design from experience-based design to data-driven design, improving the scientificity and reliability of proportion determination.
[0055] This allocation system incorporates mathematical or machine learning models and iterative optimization algorithms to precisely optimize the proportions of filling materials. It can meet the requirements of slope bearing capacity, impermeability, and freeze-thaw resistance while taking into account material utilization efficiency and engineering applicability.
[0056] Through an automatic monitoring and feedback control mechanism, this mixing system can dynamically adjust the mix ratio according to the on-site temperature, slurry state and slope response, thereby improving the stability and adaptability of open-pit mine backfilling operations in cold and high-altitude mining areas and possessing good engineering promotion value.
[0057] In one example of the present invention, the filling material screening module includes: The first constraint unit is configured with the slope safety factor Fs≥Fs0 as the constraint, and the target range of the cohesion c of the filling body and the internal friction angle φ as the screening condition; where Fs0 is taken as 1.20~1.50; The second constraint unit is configured to use the slope displacement Δ≤Δ0 as a constraint and the target range of the deformation modulus E or compression modulus M of the filling body as a screening condition; where Δ0 is a set value of 10 to 50 mm.
[0058] It should be noted that the filling material distribution system for open-pit slope control in high-altitude and cold mining areas of the present invention can also perform any of the processing steps described in the previously described method for filling material distribution for open-pit slope control in high-altitude and cold mining areas, and the specific details will not be repeated here.
[0059] Specific examples: Taking a high-altitude, cold-weather open-pit coal mine as an example, the specific implementation steps are as follows: S10: The lowest winter temperature in this mining area can reach -30℃. To address the backfilling needs of the goaf formed after side coal mining, coal gangue was selected as the main aggregate, ordinary Portland cement and fly ash as cementing materials, with appropriate amounts of antifreeze and water-reducing agents added. Basic performance tests were conducted on all raw materials, with aggregates focusing on particle size distribution, density, and water absorption; cementing materials focusing on specific surface area, setting time, and activity index; and admixtures testing their effective content and low-temperature suitability. A set of performance parameters for the backfill material was established.
[0060] S20: Based on the completion of material performance determination, multiple candidate mix proportions were designed according to different water-cement ratios, cementitious material dosages, and admixture dosages, and corresponding filling specimens were prepared. The specimens were cured under both ambient and low-temperature conditions, with the low-temperature curing temperature controlled at approximately -20℃. Compressive strength tests were conducted on each group of specimens, obtaining strength data at 3 days, 7 days, and 28 days. Simultaneously, permeability tests and freeze-thaw cycle tests were performed, with the freeze-thaw cycle temperature range from -20℃ to +20℃, and the number of cycles set at 25. Through these tests, the mechanical and durability performance indicators of different mix proportions under low-temperature and freeze-thaw conditions were systematically obtained, forming a mix proportion-performance test database.
[0061] S30: Based on the mix design-performance test database obtained from indoor tests, and combined with the engineering requirements for stability control of open-pit slopes in high-altitude and cold mining areas, a comprehensive analysis of various mix design schemes was conducted. The compressive strength at 28 days of low temperature, permeability coefficient, and strength retention rate after freeze-thaw cycles were used as the main screening indicators, eliminating mix design schemes with insufficient strength, excessively high permeability, or significant freeze-thaw degradation. Simultaneously, considering the safety factor and deformation control requirements of open-pit slopes, the mix design range of filling materials that can meet the requirements of slope bearing capacity, seepage suppression, and resistance to freeze-thaw damage was selected to determine the feasible mix design range suitable for the engineering conditions of this high-altitude and cold mining area.
[0062] S40: The proportioning parameters must meet at least the following ranges: water-cement ratio of 0.35 to 0.70; cementitious material content of 6% to 25%; admixture content of 0.1% to 5.0%; maximum aggregate size of 10 to 40 mm, and the gradation must be continuous or optimized by segmented gradation to meet pumping and compaction requirements.
[0063] Within a feasible mix design range, a deep hybrid neural network model based on an attention mechanism was constructed and trained. The model inputs included the water-cement ratio, cementitious material proportions, admixture dosage, and aggregate gradation parameters. It successfully achieved performance evaluation on low-temperature strength, permeability, and freeze-thaw durability of the test set. 2 A high-precision prediction of >0.92 was achieved. Model analysis revealed that the performance after freeze-thaw cycles was most strongly correlated with the strength development trajectory over 7 days. Subsequently, a multi-objective Bayesian optimization framework was applied, with three optimization objectives: maximizing the predicted 28-day strength, minimizing the predicted permeability coefficient, and minimizing material costs. Starting from 30 sets of initial experimental data, the optimization process, after 15 rounds of iterative search guided by the EHVI function, efficiently converged and obtained a clear Pareto optimal front after only 15 additional ratio points were evaluated. Based on current cost constraints, the engineering team selected the final optimal ratio from the Pareto solution set as follows: water-cement ratio 0.41, total cementitious material content 21% (of which cement accounts for 70% and fly ash accounts for 30%), and composite admixture content 2.2%.
[0064] S50: Using the above optimal mix ratio as the baseline parameter, input it into the fully automated intelligent mixing plant for filling operations. Simultaneously, the digital twin-driven adaptive model predictive control system is activated. A sensor network distributed throughout the slope and filling pipelines synchronizes displacement, temperature, pressure, and slurry rheological data to the digital twin virtual model in real time. During construction on a certain day, the system, through twin model simulation, predicted that a cold front would cause the core temperature of the newly filled material to drop below -6℃ within the next 40 minutes, potentially affecting early strength development. The controller immediately initiated rolling optimization, calculating, based on the online-updated system model, a control instruction to gradually increase the cementitious material content by 1.8% over the next three control cycles. This instruction was automatically issued and precisely executed. Subsequent monitoring data showed that the internal temperature of the backfill remained consistently above -2℃, and the slope displacement remained stable. The system continuously compared the deviations between predicted and measured values, automatically correcting the parameters of the internal prediction model, successfully completing this dynamic control, ensuring the stable performance of the backfill under high-altitude and cold conditions, and guaranteeing the long-term safety of the slope.
[0065] The foregoing description, with reference to preferred embodiments, details an exemplary implementation of the method and system for preparing backfill material for controlling open-pit slopes in cold mining areas proposed by the present invention. However, those skilled in the art will understand that various modifications and alterations can be made to the above specific embodiments without departing from the concept of the present invention, and various combinations can be made to the various technical features and structures proposed by the present invention without exceeding the protection scope of the present invention, which is determined by the appended claims.
Claims
1. A method for preparing filling materials for controlling open-pit slopes in high-altitude and cold mining areas, characterized in that, Includes the following steps: S10: Conduct basic performance tests on backfill materials used for backfilling coal at the side of open-pit mines, obtain their physical, chemical and mechanical property parameters, and establish a set of performance parameters for backfill materials; S20: Based on different material ratio schemes, conduct indoor mechanical and durability tests to obtain the performance indicators of the filling body under low temperature and freeze-thaw conditions, and form a performance test database of different material ratios; S30: Based on the requirements for stability control of open-pit slopes in high-altitude and cold mining areas, the test results were analyzed to screen the range of fill material proportions that meet the requirements for slope bearing capacity, impermeability and freeze-thaw resistance; S40: Based on the screening of the filling material ratio range, a deep hybrid neural network model based on the attention mechanism is constructed, and the optimal ratio parameters are obtained by using a multi-objective Bayesian optimization framework; S50: The optimized mix proportion parameters are applied to the on-site filling operation control system. Through the digital twin-driven adaptive model predictive control system, the mix proportion is dynamically adjusted to achieve precise allocation under the control target of the open slope.
2. The method for preparing filling materials for open-pit slope control in high-altitude and cold mining areas according to claim 1, characterized in that, In step S10, the physical, chemical and mechanical properties parameters specifically include: aggregate particle size distribution, density, specific surface area of cementitious materials, activity index, initial setting time of the filling material, final setting time and compressive strength at room temperature.
3. The method for preparing filling materials for controlling open-pit slopes in high-altitude and cold mining areas according to claim 1, characterized in that, Step S30 further includes introducing open slope control correlation constraints, linking the filling body parameters with the slope stability target for screening. The linkage screening includes at least: S31: Using the slope safety factor Fs≥Fs0 as a constraint, the target range of the cohesion c of the filling body and the internal friction angle φ is used as the screening condition. S32: Using the slope displacement Δ≤Δ0 as a constraint, the target range of the deformation modulus E or compression modulus M of the filling body is used as the screening condition. Among them, Fs0 is set to 1.20 to 1.50, and Δ0 is set to 10 to 50 mm.
4. The method for preparing filling materials for open-pit slope control in high-altitude and cold mining areas according to claim 1, characterized in that, In step S40, the filling material proportioning parameters include at least the water-cement ratio, cementitious material dosage, admixture dosage, and aggregate gradation parameters.
5. The method for preparing filling materials for open-pit slope control in high-altitude and cold mining areas according to claim 1, characterized in that, In step S40, the deep hybrid neural network model based on the attention mechanism includes, in sequence, an input layer, a one-dimensional convolutional layer, a long short-term memory network layer, an attention mechanism layer, and a fully connected output layer. The input layer is configured to receive parameter vectors that affect the performance of the filling material to form input features; The one-dimensional convolutional layer is configured to perform local perception and fusion of input features, and extract local interaction patterns between different material parameters; The Long Short-Term Memory (LSTM) network layer is configured to capture the dynamic patterns of infill performance over time, addressing long-term dependency issues. Internally, it utilizes a forgetting gate... Input gate Output gate and cell state and hidden state Collaborative updates; The attention mechanism layer is configured to adaptively handle the hidden states of the Long Short-Term Memory (LSTM) network layer at different times. Assign appropriate weights to focus on the historical information segments that are most critical to the current prediction target; The fully connected output layer is configured to aggregate a context vector containing key information. Mapped to the final predicted performance metric.
6. The method for preparing filling materials for open-pit slope control in high-altitude and cold mining areas according to claim 1, characterized in that, In step S40, obtaining the optimal ratio parameters using a multi-objective Bayesian optimization framework specifically includes the following steps: S41: Proxy Model Construction: For each performance objective that needs optimization, such as strength, penetration coefficient, and cost. A Gaussian process regression model is established as its surrogate model, and its expression is as follows: In the formula, It is a mean function. It is the covariance function; S42: Acquisition Functions and Search: Through Measure and evaluate a new candidate point The goal is to achieve hypervolume gain for the current Pareto front, and the next evaluation point will be selected by iteratively optimizing and maximizing the EHVI. To efficiently approximate the real Pareto frontier; among them, The expression is: In the formula, the hypervolume HV is the spatial volume enclosed by the Pareto front and a defined reference point; S43: Termination and Output: Repeat the iterative process until the preset number of evaluations or convergence conditions are reached, and finally output a set of Pareto optimal solutions; select an appropriate ratio from the solution set as the final solution according to the actual needs of emphasizing strength or cost.
7. The method for preparing filling materials for open-pit slope control in high-altitude and cold mining areas according to claim 6, characterized in that, In step S42, the detailed steps of the iterative optimization are as follows: S421: Initialization: Using a small amount of experimental data from the initial proportioning scheme, initialize the Gaussian process proxy model for each objective; S422: Model Update: Update each agent model based on all currently evaluated datasets; S423: Candidate point selection: through optimization The acquisition function selects the point most likely to improve the current Pareto front from the feasible matching parameter space. ; S424: Performance Evaluation: [Regarding...] Conduct real-world physics experiments to obtain its multi-objective performance value f( ); S425: Data Expansion: Adding new data pairs ( ,f( Add it to the evaluation dataset.
8. The method for preparing filling materials for open-pit slope control in high-altitude and cold mining areas according to claim 1, characterized in that, In step S50, the dynamic adjustment of the digital twin-driven adaptive model predictive control system specifically includes the following steps: S51: Constructing a Digital Twin: A high-fidelity model that fully maps to the physical filling site is built in virtual space. This model integrates multi-physics processes including slope rock mechanics, hydration and hardening of the filling material, heat conduction, and freeze-thaw damage. Its governing equations are expressed as follows: In the formula, These are the mass, damping, and stiffness matrices of the system, respectively. It is a displacement vector. For temperature ,stress and time Related load vectors; S52: Constructing the core of adaptive model predictive control, specifically including: Online system identification: Quickly predict the system state. The expression for predicting the system state is: In the formula, For system status, To control the input, For observation output, and Process noise and measurement noise; model parameters It is not fixed, but rather uses recursive least squares with a forgetting factor for online real-time identification and updating, and its expression is: In the formula, For the parameter estimation vector, For data vectors, forgetting factor A value of 0.95 to 0.99 is used to enable the model to track dynamic changes in the system; Rolling time-domain optimization: in each control cycle The controller estimates based on the current state. Using the initial conditions and the updated prediction model, we solve a finite-time optimal control problem, expressed as: In the formula, To predict the time domain, To control the time domain, and This is the weight matrix. Set values for performance targets; Feedback closed-loop execution: The first control increment in the optimization sequence - The water-cement ratio fine-tuning amount and the instantaneous addition amount of admixtures are sent to the automated batching and mixing equipment on site for execution; when entering the next control cycle, the new system output measurement values are collected. The state is estimated by comparing the value with the predicted value, correcting the error, and repeating the identification-optimization-execution process to form an adaptive closed-loop control system.
9. A filling material preparation system for open-pit slope control in high-altitude and cold mining areas, characterized in that, include: The performance parameter set establishment module is configured to perform basic performance tests on backfill materials used for backfilling coal at the side of open-pit mines, obtain their physical, chemical and mechanical property parameters, and establish a performance parameter set for backfill materials. The performance test database module is configured to conduct indoor mechanical and durability tests based on different material ratio schemes, obtain the performance indicators of the filling body under low temperature and freeze-thaw conditions, and form a performance test database with different material ratios. The backfill material screening module is configured to analyze test results in conjunction with the stability control requirements of open-pit slopes in high-altitude and cold mining areas, and screen the backfill material mix ratio range that meets the requirements of slope bearing capacity, impermeability and freeze-thaw resistance. The optimal ratio parameter acquisition module is configured to construct a deep hybrid neural network model based on the attention mechanism based on the ratio range of the screened filling materials, and obtain the optimal ratio parameters using a multi-objective Bayesian optimization framework. The dynamic adjustment and precise allocation module is configured to apply the optimized mix ratio parameters to the on-site filling operation control system. Through the digital twin-driven adaptive model predictive control system, the mix ratio is dynamically adjusted to achieve precise allocation under the control target of the open slope.
10. The backfill material distribution system for open-pit slope control in high-altitude and cold mining areas according to claim 9, characterized in that, The filling material screening module includes: The first constraint unit is configured with the slope safety factor Fs≥Fs0 as the constraint, and the target range of the cohesion c of the filling body and the internal friction angle φ as the screening condition; where Fs0 is taken as 1.20~1.50; The second constraint unit is configured to use the slope displacement Δ≤Δ0 as a constraint and the target range of the deformation modulus E or compression modulus M of the filling body as a screening condition; where Δ0 is a set value of 10 to 50 mm.