A method and system for staged filling of a goaf after endwall mining

Abstract

This invention discloses a method and system for graded backfilling of goaf areas after end-face mining, relating to the field of goaf filling technology. The method includes: acquiring the structure of the target goaf, dividing it into backfilling units to obtain multiple backfilling units; acquiring multiple state-aware data sequences; acquiring multiple state-aware features; identifying the risk levels of the multiple backfilling units to obtain multiple identified backfilling units; performing independent analysis of graded backfilling schemes to obtain multiple independent backfilling schemes, correcting for adjacent influences, and obtaining a target graded backfilling scheme; remotely controlling an unmanned forklift to place prefabricated lightweight inflatable blocks according to the target graded backfilling scheme; using a directional drilling rig to construct grouting boreholes; and performing secondary grouting control on the initial backfilling area of ​​the prefabricated lightweight inflatable blocks. This solves the technical problem in existing technologies where uneven distribution of backfilling strength leads to insufficient overall stability and safety, achieving the technical effect of improving the stability and safety of goaf backfilling.

Description

Technical Field

[0001] This invention relates to the field of goaf filling technology, specifically to a graded filling method and system for goaf areas after end-face mining. Background Technology

[0002] If the goaf formed after end-face mining is not backfilled in a timely and effective manner, it can easily lead to engineering disasters such as overburden movement and instability, surface subsidence, and sidewall collapse. Currently, there are two main traditional technical approaches for treating goafs after end-face mining. One is to dump gangue into the goaf. This method is low-cost and simple to operate, but the gangue is randomly piled up, making it difficult to guarantee the density and mechanical properties of the filling, which can easily affect slope stability. The other is to directly inject gangue slurry into the goaf. This method can form a relatively dense and uniform filling body with good stability, but the construction efficiency is low and the filling speed is slow, making it difficult to meet the rapid treatment needs under complex geological conditions. Furthermore, both methods generally lack the ability to monitor the spatiotemporal mechanical state and risk zoning within the goaf, making it difficult to achieve refined backfilling control based on regional differences. Summary of the Invention

[0003] This application provides a graded filling method and system for goaf areas after end-face mining, which solves the technical problem of insufficient overall stability and safety caused by uneven distribution of filling intensity in goaf areas in the prior art.

[0004] A first aspect of this application provides a method for graded backfilling of goaf areas after endwall mining, the method comprising:

[0005] The structure of the target goaf is obtained, and the filling units are divided based on the strip length, mining height, overlying stratum thickness and lithology, and geological structure, resulting in multiple filling units. A sensor network is deployed in the goaf to perceive the state of the multiple filling units, obtaining multiple state perception data sequences. Based on the multiple state perception data sequences, state time sequence perception is performed to obtain multiple state perception features. The multiple filling units are identified by risk level according to the multiple state perception features, resulting in multiple identified filling units. The multiple identified filling units are traversed to perform independent analysis of graded filling schemes, resulting in multiple independent filling schemes. The adjacency influence is corrected for the multiple independent filling schemes based on the position of the multiple identified filling units, resulting in the target graded filling scheme. A remotely controlled unmanned forklift is used to place prefabricated lightweight inflatable blocks according to the target graded filling scheme. A directional drilling rig is used to construct grouting boreholes to perform secondary grouting control on the initial filling area of ​​the prefabricated lightweight inflatable blocks.

[0006] Furthermore, the multiple state-aware data sequences are traversed to perform multi-scale feature extraction, resulting in multiple multi-scale temporal feature sets; the multiple multi-scale temporal feature sets are then enhanced within the sets to obtain multiple state-aware features.

[0007] Furthermore, the multiple multi-scale temporal feature sets are sorted in descending order of recognition scale to obtain multiple multi-scale temporal feature sequences; the multi-scale temporal feature in the preceding position of the multiple multi-scale temporal feature sequences is used to enhance the subsequent multi-scale temporal feature to obtain multiple multi-scale enhanced temporal feature sequences; state perception is performed based on the multiple multi-scale enhanced temporal feature sequences to obtain multiple state-aware features.

[0008] Furthermore, a preset graded filling rule is obtained, and a graded filling scheme is identified based on the risk level identifiers of the multiple identified filling units to obtain multiple independent filling schemes. The preset graded filling rule includes: for low-risk filling units, standard lightweight inflatable blocks are used for filling; for medium-risk filling units, lightweight inflatable blocks are used for initial filling, followed by secondary grouting using micro-expansion inorganic grouting material; for high-risk filling units, lightweight inflatable blocks are used for initial filling, followed by double grouting using micro-expansion inorganic grouting material and high-pressure-resistant filling material.

[0009] Furthermore, multiple high-risk identifier filling units are extracted from the multiple identifier filling units, and neighborhoods of multiple high-risk identifier filling units are constructed according to a preset adjacency influence bandwidth. The complexity of the neighborhoods of the multiple high-risk identifier filling units is identified to obtain multiple neighborhood complexities. The multiple neighborhood complexities are filtered according to a preset threshold to map and extract multiple leading high-risk identifier filling units. Based on the multiple independent filling schemes corresponding to the multiple leading high-risk identifier filling units, the independent filling schemes in the neighborhoods of the multiple leading high-risk identifier filling units are adjusted to achieve optimal collaborative stability and adjacency influence correction, resulting in a set of multiple corrected hierarchical filling schemes. Based on the set of multiple corrected hierarchical filling schemes, the multiple independent filling schemes are updated to obtain the target hierarchical filling scheme.

[0010] Furthermore, according to the risk level type and the number of units of the same risk level type, data is extracted from the neighborhood of the multiple high-risk identifier filling units to obtain multiple neighborhood risk level type sets and multiple sets of the same type of quantity; weighted complexity identification is performed on the multiple neighborhood risk level type sets and multiple sets of the same type of quantity to obtain multiple neighborhood complexities.

[0011] Furthermore, from multiple high-risk indicator filling units and multiple independent filling schemes, the first high-risk indicator filling unit and the first independent filling scheme corresponding to the first high-risk indicator filling unit are extracted; combined with the positions between neighboring indicator filling units in the neighborhood of the first high-risk indicator filling unit and the set of neighboring independent filling schemes, the first independent filling scheme is subjected to cooperative stability identification to obtain a first cooperative stability; the set of neighboring independent filling schemes is randomly adjusted multiple times according to a preset adjustment range to obtain multiple adjusted neighboring independent filling scheme sets; again, based on the positions between neighboring indicator filling units and the multiple adjusted neighboring independent filling scheme sets, the first independent filling scheme is subjected to cooperative stability identification to obtain multiple adjusted cooperative stability; it is determined whether there is an adjusted cooperative stability greater than or equal to the first cooperative stability among the multiple adjusted cooperative stability; if so, the adjusted neighboring independent filling scheme set corresponding to the maximum value among the multiple adjusted cooperative stability is taken as the first corrected hierarchical filling scheme set, and the first corrected hierarchical filling scheme set is added to the multiple corrected hierarchical filling scheme sets.

[0012] Furthermore, if there is no adjustment collaborative stability greater than or equal to the first collaborative stability among the multiple adjustment collaborative stabilityes, the number of times the adjustment methods of the multiple adjustment neighborhood independent filling scheme sets are disabled will be marked, and the neighborhood independent filling scheme sets will be randomly adjusted multiple times according to the preset adjustment range. The multiple adjustment neighborhood independent filling scheme sets will be updated according to the adjustment results.

[0013] Furthermore, by employing a sensorless network to perform stability identification on the graded filling process of the multiple filling units, stability identification results are obtained; based on the stability identification results, early warning and maintenance of the target goaf area are carried out.

[0014] A second aspect of this application provides a graded backfilling system for goaf areas after end-face mining, the system comprising:

[0015] The backfilling division module acquires the structure of the target goaf, and divides it into backfilling units based on strip length, mining height, overlying stratum thickness and lithology, and geological structure, obtaining multiple backfilling units. The status perception module deploys a goaf sensor network to perceive the status of multiple backfilling units, obtaining multiple status perception data sequences. The feature extraction module performs status time-series perception based on the multiple status perception data sequences, obtaining multiple status perception features. The risk identification module identifies the risk level of the multiple backfilling units based on the multiple status perception features, obtaining multiple identified backfilling units. The scheme formulation module traverses the multiple identified backfilling units to perform independent analysis of graded backfilling schemes, obtaining multiple independent backfilling schemes. It then performs adjacency influence correction on the multiple independent backfilling schemes based on the location of the multiple identified backfilling units, obtaining the target graded backfilling scheme. The backfilling control module remotely controls unmanned forklifts to place prefabricated lightweight inflatable blocks according to the target graded backfilling scheme, uses a directional drilling rig to construct grouting boreholes, and performs secondary grouting control on the initial backfilling area of ​​the prefabricated lightweight inflatable blocks.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0017] First, the structure of the target goaf is acquired. Combining strip length, mining height, overlying stratum thickness and lithology, and geological structure, filling units are divided, resulting in multiple filling units. Next, a sensor network is deployed in the goaf to monitor the status of these multiple filling units, obtaining multiple status monitoring data sequences. Further, status time-series monitoring is performed based on these data sequences to obtain multiple status monitoring features. Risk levels are then assigned to each filling unit based on these features, resulting in multiple identified filling units. Next, independent analysis of graded filling schemes is performed across these identified filling units, resulting in multiple independent filling schemes. Adjacency influence corrections are applied to these independent filling schemes based on the location of the identified filling units, yielding the target graded filling scheme. Finally, a remotely controlled unmanned forklift places prefabricated lightweight inflatable blocks according to the target graded filling scheme. A directional drilling rig is used to drill grouting holes to control secondary grouting in the initial filling area of ​​the prefabricated lightweight inflatable blocks. This application utilizes precast gas-concrete blocks for primary filling, rapidly forming a basic support structure in the initial stages of the filling operation. This significantly improves construction speed and filling efficiency, and effectively prevents subsequent filling quality from being affected by rockfall before full filling. Subsequent secondary grouting reinforces the initially filled area, achieving higher density within the fill material and further enhancing the stability of the slope and overburden. Therefore, this application solves the technical problem of uneven distribution of filling strength in goaf areas leading to insufficient overall stability and safety in existing technologies, achieving the technical effect of improving the stability and safety of goaf filling. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A schematic flowchart of a graded backfilling method for goaf after end-face mining provided in this application embodiment;

[0020] Figure 2 This is a schematic diagram of a graded backfilling system for goaf areas after end-face mining, provided as an embodiment of this application.

[0021] Figure labeling: 11 Filling division module, 12 State perception module, 13 Feature extraction module, 14 Risk identification module, 15 Scheme formulation module, 16 Filling control module. Detailed Implementation

[0022] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0023] In Example 1, after the end-face mining is completed and a goaf is formed, the graded backfilling method provided in this application can be directly applied to the actual backfilling operation of the goaf. The specific engineering application process is as follows:

[0024] First, after the goaf is ready for entry, the spatial structure of the target goaf is acquired using geological data, mining design parameters, and on-site measurement data. This clarifies the goaf's strip length, mining height, overlying stratum thickness, and lithological distribution. Combined with geological structural information such as faults and densely jointed zones, the goaf is spatially divided into multiple filling units with clearly defined boundaries. These filling units serve as the basic objects for subsequent risk identification, filling decisions, and construction control. Subsequently, a sensor network is deployed within the goaf according to the spatial distribution of the filling units to monitor the condition of the roof, sidewalls, and key stress areas. Real-time data on roof pressure, surrounding rock deformation, microseismic activity, and pore pressure are collected. The collected data is then aggregated according to the filling unit number, forming a state-sensing data sequence corresponding to each filling unit. Based on this state-sensing data sequence, the system performs state-time sensing and extracts state-sensing features reflecting the stability differences of the filling units. After obtaining the status perception characteristics of each filling unit, the system identifies the risk level of each filling unit according to preset risk level determination rules, dividing the goaf into low-risk, medium-risk, and high-risk filling units. The risk level identification guides differentiated filling methods for different areas, rather than applying a uniform filling intensity to the entire goaf. Furthermore, the system generates corresponding graded filling schemes based on the risk level identification of each filling unit. For low-risk filling units, conventional filling is performed using prefabricated lightweight inflatable blocks; for medium-risk filling units, secondary grouting reinforcement is implemented after the initial filling with prefabricated lightweight inflatable blocks, combined with micro-expansion inorganic grouting material; for high-risk filling units, double grouting is performed after the initial block placement, combining micro-expansion inorganic grouting material and high-pressure-resistant filling material to improve the bearing capacity and stability of local areas.

[0025] During the actual construction process, based on the generated target-level filling plan, unmanned forklifts are remotely controlled to enter the goaf area. Following the preset layout path and stacking density, prefabricated lightweight inflatable blocks are transported and placed into the corresponding filling units, forming the initial filling area. This allows the goaf to establish a basic spatial support structure in a relatively short time. After the initial filling is completed, directional drilling rigs are used to drill grouting holes within the area covered by the prefabricated lightweight inflatable blocks, extending the grouting channels to the bottom or lateral spaces requiring reinforcement. After the grouting holes are completed, secondary grouting is controlled according to the graded filling plan for the corresponding filling unit. By controlling the grouting pressure, grouting volume, and grouting time, the grouting material diffuses directionally along the borehole, fully filling the gaps between the blocks and localized weak load-bearing areas. This constructs a continuous and dense filling body based on the skeleton structure formed by the initial block arrangement.

[0026] During and after the graded filling construction, the system continuously monitors the stability of each filling unit using a sensor network and identifies the stability of the filling effect based on the monitoring results. When stress concentration, abnormal deformation, or abnormal microseismic activity is detected in a local area, an early warning can be triggered and supplementary grouting or local reinforcement treatment can be carried out in the corresponding area, thereby realizing the whole process control and subsequent maintenance of goaf filling.

[0027] Through the above-described embodiments, the graded filling method of this application can be directly applied to the engineering filling operation of the goaf after end-side mining, realizing graded and differentiated filling of different areas of the goaf, and improving the overall stability and safety of the filling structure.

[0028] Example 2 further illustrates the application of the graded backfilling method of this application in the actual backfilling process of a goaf formed after end-side mining:

[0029] In this embodiment, the target goaf is a strip-shaped goaf formed after the end-side mining is completed. Its length along the strike direction is about 180m, the average mining height is about 12m, and the maximum mining height in some areas is about 15m. The thickness of the overlying strata in the goaf is between 18 and 26m. The overlying lithology is mainly composed of interbedded moderately weathered sandstone and mudstone, and there are structural zones with relatively well-developed joints and fissures in some areas.

[0030] Based on the aforementioned goaf structural parameters, the system divides the goaf into grids along the strike and vertical directions, dividing it into multiple filling units. In this embodiment, a total of 24 filling units are obtained, with each filling unit having a planar dimension of approximately 10m × 15m. Each filling unit has a clearly defined spatial boundary, serving as the basic unit for subsequent state perception, risk identification, and graded filling control.

[0031] Sensor networks are deployed at corresponding locations in each filling unit to continuously monitor status information such as roof pressure, surrounding rock convergence deformation, and microseismic activity in the goaf. The monitoring data are then collected according to the filling unit number to form a status perception data sequence for each filling unit.

[0032] The system extracts state-aware features based on the state-aware data sequence and identifies the risk level of the filling unit according to a preset risk level determination rule. Taking some filling units in this embodiment as examples, their state-aware features and risk identification results are as follows:

[0033] Filling unit U-03: The roof pressure change rate is about 0.02 MPa / d, the surrounding rock convergence rate is about 0.3 mm / d, and the frequency of microseismic events is about 1 time / d. It is identified as a low-risk filling unit.

[0034] Filling unit U-11: The roof pressure change rate is about 0.08 MPa / d, the surrounding rock convergence rate is about 1.2 mm / d, and the frequency of microseismic events is about 6 times / d. It is identified as a medium-risk filling unit.

[0035] Filling unit U-19: The roof pressure change rate is about 0.15 MPa / d, the surrounding rock convergence rate is about 2.6 mm / d, and the frequency of microseismic events is about 15 times / d. It is identified as a high-risk filling unit.

[0036] The system presets the roof pressure change rate threshold to 0.1 MPa / d, the surrounding rock convergence rate threshold to 2.0 mm / d, and the microseismic event frequency threshold to 10 times / d. When at least two of the state perception features exceed the corresponding thresholds, the corresponding filling unit will be marked as a high-risk filling unit.

[0037] Based on the risk level identification results of each filling unit, the system generates corresponding graded filling schemes.

[0038] For low-risk filling units (such as U-03), prefabricated lightweight inflatable blocks are used for filling. The density of the lightweight inflatable blocks is about 0.85 blocks / m³, and no secondary grouting operation is performed.

[0039] For medium-risk filling units (such as U-11), after the initial filling with prefabricated lightweight inflatable blocks, a secondary grouting reinforcement is carried out using micro-expansion inorganic grouting material. The grouting pressure is controlled at 0.6~0.8MPa, and the grouting volume per unit volume is approximately 0.12m³ / m³.

[0040] For high-risk filling units (such as U-19), after the initial filling of precast lightweight inflatable blocks is completed, dual grouting control is implemented. The first grouting uses micro-expansion inorganic grouting material, and the grouting pressure is controlled at 0.8 to 1.0 MPa. The second grouting uses high-pressure resistant filling material, and the grouting pressure is controlled at 1.2 to 1.5 MPa. The grouting coverage area is extended to the adjacent filling unit by about 3m.

[0041] During the actual construction process, based on the generated target tiered filling plan, unmanned forklifts are remotely controlled to enter the goaf area. Following the preset layout path and stacking density requirements, prefabricated lightweight inflatable blocks are transported and placed at the corresponding filling unit locations, forming the initial filling area. After the initial filling is completed, directional drilling rigs are used to drill grouting holes within the area covered by the prefabricated lightweight inflatable blocks. These holes penetrate the blocks and extend to the bottom or lateral areas of the goaf requiring reinforcement. After drilling, secondary grouting is controlled according to the tiered filling plan for the corresponding filling unit. By controlling the grouting pressure, grouting volume, and grouting time, the grouting material diffuses directionally along the boreholes, filling the gaps between the blocks and areas with weak load-bearing capacity, thus forming a continuous and dense filling body. During and after the tiered filling construction, the system continuously monitors the stability of each filling unit using a sensor network, analyzing changes in surrounding rock convergence, roof pressure fluctuations, and microseismic activity. When monitoring results indicate that the deformation of the surrounding rock of the filling unit tends to stabilize, the fluctuation amplitude of the roof pressure decreases, and the microseismic activity is significantly attenuated, the filling unit is determined to be in a stable state; when abnormal changes occur, supplementary grouting or local reinforcement treatment can be further triggered.

[0042] Example 3, as Figure 1 As shown, this application provides a method for graded backfilling of goaf areas after end-face mining, wherein the method includes:

[0043] The structure of the target goaf is obtained, and the filling units are divided by combining the strip length, mining height, thickness and lithology of the overlying strata, and geological structure, resulting in multiple filling units.

[0044] First, the spatial extent of the target goaf is reconstructed using geological data, 3D laser scanning, and measured data from the tunnels to obtain its geometric parameters, including strip length, strip width, mining height, and the top and bottom morphology of the goaf. Then, the thickness and lithological parameters of the overlying strata are read, including strata hardness, degree of interlayer joint development, and layer stability coefficient. Simultaneously, the geological structural information of the goaf area is analyzed, including the location and distribution of faults, folds, collapse columns, and densely jointed zones. Based on these structural, lithological, and structural parameters, the goaf is subdivided into grids along the strip direction and vertically according to preset unit division rules: when the strip length variation rate, mining height variation rate, or overlying strata stability criterion exceeds a preset threshold, the corresponding area is independently divided into separate filling units; when a geological structure crosses a region, that region is further subdivided according to the width of the structural influence zone. Finally, the geometric boundaries, strata properties, and structural influence range of each subdivided region are merged to generate multiple filling units.

[0045] A sensor network was deployed in the goaf area to monitor the status of multiple filling units and obtain multiple status monitoring data sequences.

[0046] Based on the spatial location and stability distribution of the filling units, sensor nodes are deployed in the roof, sidewalls, and key stress locations of the goaf to construct a sensor network covering all filling units. The sensor network includes roof pressure sensors, surrounding rock convergence displacement sensors, microseismic monitoring nodes, pore water pressure gauges, and environmental vibration accelerometers. Sensor density is pre-allocated according to the risk level of the filling units, with high-density sensor clusters deployed in high-risk units.

[0047] All sensors are networked via wired or wireless mining communication links, enabling each sensor node to collect real-time data on roof loading changes, surrounding rock deformation, microseismic energy events, pore pressure changes, and dynamic vibration response according to a preset sampling period. The collected data is aggregated according to the filling unit number, and the various types of monitoring data from each filling unit within a continuous time window are arranged along a time axis to generate multiple state-sensing data sequences. Each state-sensing data sequence consists of a timestamp, sensor type identifier, and corresponding mechanical response value, used to characterize the dynamic state changes of the filling unit before, during, and after filling, as well as during the stabilization process.

[0048] Based on the multiple state-aware data sequences, state timing perception is performed to obtain multiple state-aware features.

[0049] Furthermore, based on the multiple state-aware data sequences, state temporal perception is performed to obtain multiple state-aware features, including:

[0050] Multi-scale feature extraction is performed by traversing the multiple state-aware data sequences to obtain multiple sets of multi-scale temporal features; in-set enhancement is performed on the multiple sets of multi-scale temporal features to obtain multiple state-aware features.

[0051] First, multiple state-aware data sequences are traversed, and features are extracted from each sequence according to different time scales. The sequences are divided into short-period windows, medium-period windows, and long-period windows. Multiple features reflecting instantaneous fluctuations, trend changes, and stability evolution are extracted using methods such as time-domain statistics, frequency-domain analysis, and microseismic event density calculation, forming multiple multi-scale time-series feature sets. Each set includes multi-dimensional time-series indices such as pressure gradient change rate, surrounding rock convergence velocity, microseismic energy accumulation, pore pressure attenuation coefficient, and vibration spectrum characteristics. Next, intra-set enhancement processing is performed on the multiple multi-scale time-series feature sets. By analyzing the temporal correlation between features at different scales within the same set, features with higher sensitivity or higher stability are selected as dominant features, and other scale features are weighted and enhanced. This allows the enhanced feature sets to more fully express the mechanical behavior response of the infill unit at different time scales. Finally, the enhanced multi-scale time-series feature sets are aggregated to obtain multiple state-aware features that comprehensively characterize the dynamic state of the infill unit.

[0052] Furthermore, intra-set enhancement is performed on the multiple multi-scale temporal feature sets to obtain multiple state-aware features, including:

[0053] The multiple multi-scale temporal feature sets are sorted in descending order of recognition scale to obtain multiple multi-scale temporal feature sequences; the multi-scale temporal feature of the next position in the multiple multi-scale temporal feature sequences is enhanced to obtain multiple multi-scale enhanced temporal feature sequences; state perception is performed based on the multiple multi-scale enhanced temporal feature sequences to obtain multiple state-aware features.

[0054] First, the multiple multi-scale temporal feature sets are sorted in descending order of recognition scale, placing high-scale features (such as long-period stability features) at the beginning of the sequence and low-scale features (such as short-period instantaneous fluctuation features) at the end, thus obtaining multiple multi-scale temporal feature sequences. Then, for each multi-scale temporal feature sequence, the preceding multi-scale temporal feature is used as the dominant feature, and enhancement processing is performed on the following multi-scale temporal feature: the temporal similarity between the preceding and following features is calculated, including trend similarity, fluctuation consistency, and peak synchronization. Trend similarity is calculated based on the Pearson correlation coefficient, fluctuation consistency is calculated based on the differential sequence correlation, and peak synchronization is calculated based on the difference between peak position and peak amplitude. The three types of similarity are weighted and fused to obtain a comprehensive similarity result. The similarity result is normalized to construct a feature enhancement matrix. The feature enhancement matrix is ​​then used to perform a convolution operation on the following multi-scale temporal feature. Based on the convolution output, the feature values ​​are either enhanced or suppressed to enhance their expressive power in the direction of the dominant feature, resulting in multiple multi-scale enhanced temporal feature sequences. Finally, state perception is performed based on multiple multi-scale enhanced temporal feature sequences. By evaluating the enhanced features in terms of trend consistency, abnormal response sensitivity, and energy release patterns in the time dimension, multiple state perception features that can more accurately reflect the dynamic stability of the filling unit are generated.

[0055] Based on the multiple state-aware features, the multiple filling units are identified by risk level to obtain multiple identified filling units.

[0056] First, the state-aware characteristics corresponding to each filling unit are input into the risk assessment model. These characteristics include comprehensive indicators reflecting structural stability, such as the roof pressure change rate, surrounding rock convergence velocity, microseismic energy accumulation trend, pore pressure attenuation law, vibration spectrum stability, and multi-scale enhancement time-series characteristics. Then, based on preset risk level determination rules, these characteristics are quantitatively scored. The rules calculate the risk index for each filling unit by comparing each characteristic with risk thresholds, inflection points, and abnormal fluctuation amplitudes. A low-risk unit is identified when the risk index is in the low range, a medium-risk unit when it is in the middle range, and a high-risk unit when it exceeds the high-range threshold. Further, a multi-feature cross-validation mechanism is used to verify the stability of the initial judgment results to prevent misjudgments caused by fluctuations in a single feature. Finally, the risk level identifiers corresponding to each filling unit are numbered and registered, forming multiple identified filling units, providing a basis for subsequent identification of graded filling schemes and correction of adjacent influences.

[0057] The risk index for each filling unit is calculated as follows: First, corresponding risk threshold ranges are set for various state-aware features, including the roof pressure safety threshold, the critical rate of surrounding rock convergence, the upper limit of microseismic energy event frequency, the threshold for abnormal increase in pore pressure, and the vibration spectrum offset limit. The feature deviation is obtained by calculating the difference between the actual feature value of the filling unit and its corresponding threshold. Then, based on the relationship between the deviation and the inflection point of feature change, the rising, stabilizing, or abrupt change phases of the feature in the time dimension are identified. For features in the rapid rising or abrupt change phase, additional weighting is applied according to a preset risk gain coefficient to improve sensitivity to dangerous abrupt changes. Further, for features with significant abnormal fluctuation amplitudes, the change gradient within a short-time window is calculated, and the change amplitude is mapped to a penalty term and added to the score value. Finally, the quantitative scores of each feature are weighted and summed according to preset weights, and combined with the abnormal fluctuation penalty term to obtain the risk index. This risk index is used as the basis for identifying the risk level of the filling unit's stability.

[0058] The hierarchical filling scheme is analyzed independently by traversing the multiple marker filling units to obtain multiple independent filling schemes. The adjacency effect correction of the multiple independent filling schemes is performed by combining the positions of the multiple marker filling units to obtain the target hierarchical filling scheme.

[0059] Furthermore, by traversing the multiple identifier filling units and performing independent hierarchical filling scheme analysis, multiple independent filling schemes are obtained, including:

[0060] A preset graded filling rule is obtained, and a graded filling scheme is identified based on the risk level identifiers of the multiple identified filling units to obtain multiple independent filling schemes. The preset graded filling rule includes: for low-risk filling units, standard lightweight inflatable blocks are used for filling; for medium-risk filling units, lightweight inflatable blocks are used for initial filling, followed by secondary grouting using micro-expansion inorganic grouting material; for high-risk filling units, lightweight inflatable blocks are used for initial filling, followed by double grouting using micro-expansion inorganic grouting material and high-pressure-resistant filling material.

[0061] The system invokes a pre-defined tiered filling rule library and identifies tiered filling schemes based on the risk level of each identified filling unit. Specifically, for filling units identified as low-risk, an independent filling scheme using only lightweight inflatable blocks for conventional filling is generated by retrieving standard lightweight inflatable block layout parameters. For filling units identified as medium-risk, a composite reinforcement scheme is formed by configuring micro-expansion inorganic grouting material according to pre-defined grouting strength and mix proportion parameters, based on the initial filling scheme using lightweight inflatable blocks. For filling units identified as high-risk, high-pressure-resistant filling material is introduced in addition to lightweight inflatable block layout and micro-expansion inorganic grouting, implementing dual grouting reinforcement control to improve local bearing capacity and deformation suppression. The multiple independent filling schemes generated based on the pre-defined tiered filling rules correspond to the filling requirements of different risk levels, providing basic scheme inputs for subsequent adjacent impact correction.

[0062] Furthermore, by combining the positions of multiple marker filling units, adjacency influence correction is applied to the multiple independent filling schemes to obtain a target hierarchical filling scheme, including:

[0063] Multiple high-risk identifier filling units are extracted from the multiple identifier filling units. Neighborhoods of these high-risk identifier filling units are constructed according to a preset adjacency influence bandwidth. The complexity of these neighborhoods is identified to obtain multiple neighborhood complexities. These neighborhood complexities are then filtered according to a preset threshold to map and extract multiple leading high-risk identifier filling units. Based on multiple independent filling schemes corresponding to these leading high-risk identifier filling units, the independent filling schemes within the neighborhoods of these leading high-risk identifier filling units are adjusted to achieve optimal collaborative stability and adjacency influence correction, resulting in a set of multiple corrected hierarchical filling schemes. Finally, the multiple independent filling schemes are updated based on this set of corrected hierarchical filling schemes to obtain the target hierarchical filling scheme.

[0064] First, all high-risk marked filling units are extracted from multiple marked filling units. Based on the three-dimensional spatial structure of the goaf and the spatial distance between units, and according to a preset adjacency influence bandwidth, a corresponding high-risk filling neighborhood is constructed for each high-risk marked filling unit to characterize the impact range of the high-risk area on the surrounding area. Then, the complexity of the neighborhoods of multiple high-risk marked filling units is identified. By statistically analyzing the distribution types of units of different risk levels and the number of units of the same risk level within the neighborhood, the risk superposition degree and spatial non-uniformity of the neighborhood are calculated to obtain multiple neighborhood complexities. Further, the multiple neighborhood complexities are compared with a preset complexity screening threshold, and neighborhoods with a complexity greater than or equal to the threshold are selected. The corresponding high-risk marked filling units are mapped and extracted as leading high-risk marked filling units, which serve as the dominant units for adjacency influence correction. Next, based on multiple independent filling schemes corresponding to multiple leading high-risk marked filling units, the independent filling schemes within their respective neighborhoods are adjusted to achieve optimal collaborative stability and perform adjacency influence correction. Specifically, by analyzing the spatial relationships of filling units within a neighborhood, and comparing the initial filling strength, grouting material usage, and block density of each unit, strategic conflicts that may lead to uneven local load distribution or stress concentration risks are identified. Based on the independent scheme leading the high-risk unit, adjustments are made to the corresponding independent schemes within the neighborhood, including increasing the grouting pressure of adjacent units, expanding the secondary grouting range, increasing the density of lightweight aerated blocks, or introducing higher-grade reinforcement materials, to achieve an overall improvement in the coordinated stability of the neighborhood. The results of these adjustments are aggregated to generate multiple sets of modified graded filling schemes. Finally, based on these sets of modified graded filling schemes, multiple independent filling schemes are updated, replacing or merging the corresponding initial independent schemes with modified schemes to form a target graded filling scheme that satisfies both spatial coordination consistency and risk distribution balance.

[0065] Furthermore, the complexity of the neighborhood of the multiple high-risk marker filling units is identified to obtain multiple neighborhood complexities, including:

[0066] Based on the risk level type and the number of units of the same risk level type, data is extracted from the neighborhood of the multiple high-risk identifier filling units to obtain multiple neighborhood risk level type sets and multiple sets of quantities of the same type; weighted complexity identification is performed on the multiple neighborhood risk level type sets and multiple sets of quantities of the same type to obtain multiple neighborhood complexities.

[0067] First, for each high-risk marker filling unit's neighborhood, data is extracted from all filling units within the neighborhood according to risk level type. The risk level (low risk, medium risk, high risk) of the units is recorded according to their occurrence order and spatial distribution relationship, forming multiple neighborhood risk level type sets. Simultaneously, the number of different risk level types within the neighborhood is independently counted, forming multiple sets of the same type, which reflect the distribution density and proportion of each risk level in the neighborhood. Subsequently, weighted complexity identification is performed on the multiple neighborhood risk level type sets and the multiple sets of the same type, respectively. By calculating indicators such as the mixing degree between different risk level types, the influence coefficient of level differences on neighborhood stability, the clustering coefficient of the same type, and the spatial non-uniformity of risk level combinations, the identification results of the two sets are weighted and fused according to a preset weight factor to obtain multiple neighborhood complexities, which are used to quantify the risk accumulation degree and spatial complexity of each high-risk neighborhood.

[0068] The process of obtaining multiple neighborhood complexities involves the following steps: For each set of neighborhood risk level types, firstly, the spatial mixing degree of different risk level types is calculated. By identifying the alternating distribution of low-risk, medium-risk, and high-risk units in the neighborhood, a risk level mixing degree index is obtained. Simultaneously, the influence coefficient of the level difference between different risk levels on neighborhood stability is calculated to measure the potential local load unevenness or stress concentration effect caused by the adjacency of high-risk and low-risk units. For multiple sets of similar quantities, the clustering degree of infill units of the same risk level within the neighborhood is statistically analyzed to obtain the similar quantity clustering coefficient. Furthermore, based on the degree of quantity skewness of risk levels, a spatial non-uniformity index of risk level combinations is derived to reflect the balance of risk level distribution within the neighborhood. Subsequently, the above multiple indicators are weighted and fused according to preset weighting factors: the mixing degree and the level difference influence coefficient are used as the main stability factors, and the clustering coefficient and spatial non-uniformity are used as auxiliary distribution factors. Through linear weighting or non-linear fusion methods, multiple neighborhood complexities characterizing the risk complexity of the neighborhood are calculated and used for subsequent guiding unit selection and adjacency effect correction.

[0069] Furthermore, based on multiple independent filling schemes corresponding to multiple high-risk marker filling units, the independent filling schemes within the neighborhood of multiple high-risk marker filling units are adjusted to optimize the adjacency impact with the goal of achieving optimal collaborative stability, resulting in a set of multiple corrected hierarchical filling schemes, including:

[0070] From multiple high-risk indicator filling units and multiple independent filling schemes, the first high-risk indicator filling unit and the first independent filling scheme corresponding to the first high-risk indicator filling unit are extracted. Combining the positions of neighboring indicator filling units within the neighborhood of the first high-risk indicator filling unit and the set of neighboring independent filling schemes, the first independent filling scheme is subjected to cooperative stability identification to obtain a first cooperative stability. The set of neighboring independent filling schemes is randomly adjusted multiple times according to a preset adjustment range to obtain multiple adjusted neighboring independent filling scheme sets. Again, based on the positions of neighboring indicator filling units and the multiple adjusted neighboring independent filling scheme sets, the first independent filling scheme is subjected to cooperative stability identification to obtain multiple adjusted cooperative stability. It is determined whether there exists an adjusted cooperative stability greater than or equal to the first cooperative stability among the multiple adjusted cooperative stability sets. If so, the adjusted neighboring independent filling scheme set corresponding to the maximum value among the multiple adjusted cooperative stability sets is taken as the first corrected hierarchical filling scheme set, and the first corrected hierarchical filling scheme set is added to the multiple corrected hierarchical filling scheme sets.

[0071] From multiple high-risk marker filling units, a leading unit is randomly selected, and its corresponding independent filling scheme is extracted. This leading unit is designated as the first high-risk marker filling unit, and its corresponding independent filling scheme is designated as the first independent filling scheme. Subsequently, all neighboring marker filling units and their independent filling schemes within the neighborhood of the first high-risk marker filling unit are aggregated. Combining the spatial relationship between the neighboring marker filling units, the first independent filling scheme is subjected to cooperative stability identification. By calculating indicators such as the overall bearing consistency, stress gradient coordination, grout diffusion matching degree, and block density adaptability of the neighborhood, the first cooperative stability value is obtained to characterize the stability performance of the scheme within the current neighborhood. Next, the set of independent filling schemes in the neighborhood is randomly adjusted multiple times according to a preset adjustment range. The adjustments include fine-tuning of grouting pressure, increasing or decreasing the density of lightweight aerated blocks, slight changes in the proportion of micro-expansion inorganic grouting material, and local expansion of the grouting coverage area, thereby generating multiple sets of adjusted independent filling schemes in the neighborhood. Subsequently, based on the spatial relationship of the neighborhood identifier filling units, the cooperative stability of the multiple sets of independent neighborhood filling schemes is identified with the first independent filling scheme, resulting in multiple adjustment cooperative stability values. These values ​​are used to evaluate the improvement effect of different adjustment strategies on the overall stability. Finally, the multiple adjustment cooperative stability values ​​are compared with the first cooperative stability. If there is an adjustment cooperative stability value greater than or equal to the first cooperative stability, the set of independent neighborhood filling schemes corresponding to the maximum value is selected as the first corrected hierarchical filling scheme set. This first corrected hierarchical filling scheme set is then added to the multiple corrected hierarchical filling scheme sets and used as the input basis for the final hierarchical filling scheme update.

[0072] Furthermore, if there is no adjustment collaborative stability greater than or equal to the first collaborative stability among the multiple adjustment collaborative stabilityes, the number of times the adjustment methods of the multiple adjustment neighborhood independent filling scheme sets are disabled will be marked, and the neighborhood independent filling scheme sets will be randomly adjusted multiple times according to a preset adjustment range. The multiple adjustment neighborhood independent filling scheme sets will be updated according to the adjustment results.

[0073] After identifying the collaborative stability of multiple sets of independent filling schemes in the neighborhood, if the collaborative stability of all adjustments is lower than the first collaborative stability, it indicates that the current adjustment strategy cannot improve the overall collaborative stability of the neighborhood. In this case, the number of times each adjustment method was disabled is marked, and the number of failures of each method is recorded for subsequent adjustment strategy selection and avoidance. Subsequently, based on the adjustment methods that were not disabled or had fewer disable times, multiple random adjustments are performed on the set of independent filling schemes in the neighborhood according to a preset adjustment range. This is achieved by randomly perturbing the combination of parameters such as grouting pressure, grouting coverage, lightweight inflatable block density, and material ratio to generate new sets of independent filling schemes in the neighborhood. The adjusted scheme sets are updated based on the latest random adjustment results, allowing them to continue participating in optimization iterations in subsequent collaborative stability identification, thereby improving the overall collaborative stability of the filling schemes and the regional bearing capacity coordination.

[0074] The remotely controlled unmanned forklift places the precast lightweight inflatable blocks according to the target tiered filling plan. A directional drilling rig is used to drill grouting holes and perform secondary grouting control on the initial filling area of ​​the precast lightweight inflatable blocks.

[0075] First, the generated target-graded filling plan is uploaded to the filling construction control platform. Control commands are sent to the unmanned forklift via wireless communication link, enabling the forklift to transport prefabricated lightweight inflatable blocks one by one to the corresponding filling unit positions according to the spatial layout path and stacking density requirements in the plan. The forklift completes the initial filling area placement operation in a specified direction and stacking method. During placement, the unmanned forklift, combined with its onboard positioning module, vision assistance module, and collision avoidance sensors, achieves automatic obstacle avoidance and precise positioning in the complex terrain of the goaf area, ensuring the stability and structural integrity of the lightweight inflatable block placement. Subsequently, based on the completion of the initial block placement, a directional drilling rig is dispatched to construct grouting boreholes at the outer edge of the lightweight inflatable block stack and within its coverage area. By adjusting the angle and controlling the depth along the preset drilling trajectory, the boreholes penetrate the blocks and extend to the bottom and lateral spatial areas requiring reinforcement, forming grouting flow channels. After drilling is completed, secondary grouting is performed on the corresponding areas according to the filling level of the target graded filling scheme. The grouting materials include micro-expansion inorganic grouting materials or high-pressure resistant filling materials. During the grouting process, the grouting pressure, grouting volume, and diffusion trend are monitored in real time, and the grouting rate and grouting time are dynamically adjusted to ensure that the grouting material fully penetrates the gaps between the blocks and the adjacent weak load-bearing areas, forming a continuous and dense reinforced body. Finally, by completing the secondary grouting reinforcement, the area where the blocks were initially laid out is structurally sealed, the mechanical bearing capacity is improved, and the overall stability is enhanced, achieving differentiated filling control for areas with different risk levels in the graded filling scheme.

[0076] Furthermore, this includes:

[0077] By employing a sensorless network to perform stability identification on the graded filling process of the multiple filling units, stability identification results are obtained; based on the stability identification results, early warning and maintenance of the target goaf area are carried out.

[0078] The stability of the graded filling process of multiple filling units is identified by a sensor network deployed in the goaf, and the stability identification results are obtained. Specifically, during the graded filling process, the mechanical response data of the filling units in the block placement, drilling and secondary grouting stages are collected in real time by roof pressure sensors, surrounding rock convergence displacement sensors, microseismic monitoring nodes and grouting pressure monitoring devices deployed in the sensor network. The process data is compared and analyzed with the initial state perception characteristics of the corresponding units to calculate the deformation recovery degree of the surrounding rock after filling, the uniformity of stress redistribution, the attenuation rate of microseismic activity and the integrity of grout diffusion, etc. Based on this, the stability level of each filling unit is identified, and the stability identification results are formed.

[0079] During the filling process, real-time monitoring data such as surrounding rock displacement, roof pressure, frequency of microseismic events, and grouting pressure changes were continuously collected for each filling unit. These process data were compared item by item with the initial state perception characteristics obtained before filling, arranged chronologically. Key mechanical indicators such as the degree of recovery of surrounding rock deformation, uniformity of stress redistribution, and attenuation rate of microseismic events were calculated by quantifying the rebound amplitude of surrounding rock displacement, the temporal stability of roof stress, and the attenuation trend of microseismic energy events. Furthermore, based on the pressure diffusion trajectory and the matching degree between grouting volume and diffusion range during the grouting process, a grouting diffusion integrity index was calculated to assess whether the filling body formed a stable and continuous support structure. Subsequently, these indicators were input into a stability assessment model, and by comparing them with preset stability level thresholds, a comprehensive judgment was made as to whether the filling unit was stable, metastable, or unstable, thus forming a stability identification result.

[0080] By quantifying the rebound amplitude of surrounding rock displacement, the temporal stability of roof stress, and the attenuation trend of microseismic energy events, key mechanical indicators such as the degree of recovery of surrounding rock deformation, the uniformity of stress redistribution, and the attenuation rate of microseismic activity are calculated. These include: First, based on the continuous displacement sequence collected by surrounding rock convergence displacement sensors, the difference in surrounding rock displacement before and after filling and its rebound amount are calculated. The rebound value is compared with the initial deformation to obtain the degree of recovery of surrounding rock deformation, which characterizes the mechanical stability of the surrounding rock structure after filling. Second, using the temporal changes in pressure monitored by roof pressure sensors, the pressure sequence is subjected to fluctuation analysis and stability assessment. By calculating parameters such as the peak pressure variation amplitude, the standard deviation of pressure fluctuation, and the time for pressure to recover to the stable range, a stress redistribution uniformity index is formed, which reflects whether the stress field of the roof tends to be uniform and stable after filling. Next, a time-series decay analysis was performed on the energy sequence of microseismic events collected by the microseismic monitoring nodes. By statistically analyzing the frequency of occurrence of microseismic events, changes in energy level, and the slope of the decay curve, the microseismic activity decay rate was calculated to determine whether the energy release between the filling body and the surrounding rock is stable and whether there is a new trend of rupture extension.

[0081] Based on the stability identification results, the system provides early warning and maintenance for the target goaf. When the identification results indicate that a filling unit has signs of instability such as stress concentration, continuous deformation, or abnormal microseismic activity, the system automatically determines its risk level and generates early warning information. Remote maintenance measures are triggered through the control center, including additional local grouting, adding lightweight inflatable blocks, adjusting support parameters, or restricting personnel and equipment from entering the risk area. At the same time, the early warning status is recorded in real time and synchronized to the goaf maintenance system to continuously monitor areas with potential instability, ensuring the overall stability and operational safety of the goaf after filling.

[0082] In summary, the embodiments of this application have at least the following technical effects:

[0083] First, the structure of the target goaf is acquired. Combining strip length, mining height, overlying stratum thickness and lithology, and geological structure, filling units are divided, resulting in multiple filling units. Next, a sensor network is deployed in the goaf to monitor the status of these multiple filling units, obtaining multiple status monitoring data sequences. Further, status time-series monitoring is performed based on these data sequences to obtain multiple status monitoring features. Risk levels are then assigned to each filling unit based on these features, resulting in multiple identified filling units. Next, independent analysis of graded filling schemes is performed across these identified filling units, resulting in multiple independent filling schemes. Adjacency influence corrections are applied to these independent filling schemes based on the location of the identified filling units, yielding the target graded filling scheme. Finally, a remotely controlled unmanned forklift places prefabricated lightweight inflatable blocks according to the target graded filling scheme. A directional drilling rig is used to drill grouting holes to control secondary grouting in the initial filling area of ​​the prefabricated lightweight inflatable blocks. This invention solves the technical problem of uneven distribution of goaf filling strength in existing technologies, which leads to insufficient overall stability and safety, and achieves the technical effect of improving the stability and safety of goaf filling.

[0084] Example 4, based on the same inventive concept as the graded backfilling method for goaf after end-face mining in the aforementioned examples, such as... Figure 2 As shown, this application provides a graded backfilling system for goaf areas after endwall mining, wherein the system includes:

[0085] Backfilling Division Module 11: Acquires the structure of the target goaf, and divides it into backfilling units based on strip length, mining height, overlying stratum thickness and lithology, and geological structure, obtaining multiple backfilling units; State Sensing Module 12: Deploys a goaf sensor network to perform state sensing on multiple backfilling units, obtaining multiple state sensing data sequences; Feature Extraction Module 13: Performs state temporal sensing based on the multiple state sensing data sequences, obtaining multiple state sensing features; Risk Identification Module 14: Identifies the risk of the multiple backfilling units based on the multiple state sensing features. Risk level identification, obtaining multiple identification filling units; Scheme formulation module 15: Traversing the multiple identification filling units to perform independent analysis of graded filling schemes, obtaining multiple independent filling schemes, and combining the positions of the multiple identification filling units to perform adjacency influence correction on the multiple independent filling schemes to obtain the target graded filling scheme; Filling control module 16: Remotely controlling unmanned forklifts to place prefabricated lightweight inflatable blocks according to the target graded filling scheme, using directional drilling rigs to construct grouting holes, and performing secondary grouting filling control on the initial filling area of ​​the prefabricated lightweight inflatable blocks.

[0086] Furthermore, the feature extraction module 13 is used to perform the following method:

[0087] Multi-scale feature extraction is performed by traversing the multiple state-aware data sequences to obtain multiple sets of multi-scale temporal features; in-set enhancement is performed on the multiple sets of multi-scale temporal features to obtain multiple state-aware features.

[0088] Furthermore, the feature extraction module 13 is used to perform the following method:

[0089] The multiple multi-scale temporal feature sets are sorted in descending order of recognition scale to obtain multiple multi-scale temporal feature sequences; the multi-scale temporal feature of the next position in the multiple multi-scale temporal feature sequences is enhanced to obtain multiple multi-scale enhanced temporal feature sequences; state perception is performed based on the multiple multi-scale enhanced temporal feature sequences to obtain multiple state-aware features.

[0090] Furthermore, the scheme formulation module 15 is used to perform the following methods:

[0091] A preset graded filling rule is obtained, and a graded filling scheme is identified based on the risk level identifiers of the multiple identified filling units to obtain multiple independent filling schemes. The preset graded filling rule includes: for low-risk filling units, standard lightweight inflatable blocks are used for filling; for medium-risk filling units, lightweight inflatable blocks are used for initial filling, followed by secondary grouting using micro-expansion inorganic grouting material; for high-risk filling units, lightweight inflatable blocks are used for initial filling, followed by double grouting using micro-expansion inorganic grouting material and high-pressure-resistant filling material.

[0092] Furthermore, the scheme formulation module 15 is used to perform the following methods:

[0093] Multiple high-risk identifier filling units are extracted from the multiple identifier filling units. Neighborhoods of these high-risk identifier filling units are constructed according to a preset adjacency influence bandwidth. The complexity of these neighborhoods is identified to obtain multiple neighborhood complexities. These neighborhood complexities are then filtered according to a preset threshold to map and extract multiple leading high-risk identifier filling units. Based on multiple independent filling schemes corresponding to these leading high-risk identifier filling units, the independent filling schemes within the neighborhoods of these leading high-risk identifier filling units are adjusted to achieve optimal collaborative stability and adjacency influence correction, resulting in a set of multiple corrected hierarchical filling schemes. Finally, the multiple independent filling schemes are updated based on this set of corrected hierarchical filling schemes to obtain the target hierarchical filling scheme.

[0094] Furthermore, the scheme formulation module 15 is used to perform the following methods:

[0095] Based on the risk level type and the number of units of the same risk level type, data is extracted from the neighborhood of the multiple high-risk identifier filling units to obtain multiple neighborhood risk level type sets and multiple sets of quantities of the same type; weighted complexity identification is performed on the multiple neighborhood risk level type sets and multiple sets of quantities of the same type to obtain multiple neighborhood complexities.

[0096] Furthermore, the scheme formulation module 15 is used to perform the following methods:

[0097] From multiple high-risk indicator filling units and multiple independent filling schemes, the first high-risk indicator filling unit and the first independent filling scheme corresponding to the first high-risk indicator filling unit are extracted. Combining the positions of neighboring indicator filling units within the neighborhood of the first high-risk indicator filling unit and the set of neighboring independent filling schemes, the first independent filling scheme is subjected to cooperative stability identification to obtain a first cooperative stability. The set of neighboring independent filling schemes is randomly adjusted multiple times according to a preset adjustment range to obtain multiple adjusted neighboring independent filling scheme sets. Again, based on the positions of neighboring indicator filling units and the multiple adjusted neighboring independent filling scheme sets, the first independent filling scheme is subjected to cooperative stability identification to obtain multiple adjusted cooperative stability. It is determined whether there exists an adjusted cooperative stability greater than or equal to the first cooperative stability among the multiple adjusted cooperative stability sets. If so, the adjusted neighboring independent filling scheme set corresponding to the maximum value among the multiple adjusted cooperative stability sets is taken as the first corrected hierarchical filling scheme set, and the first corrected hierarchical filling scheme set is added to the multiple corrected hierarchical filling scheme sets.

[0098] Furthermore, the scheme formulation module 15 is used to perform the following methods:

[0099] If none of the multiple adjustment coordination stabilityes is greater than or equal to the first coordination stability, then the number of times the adjustment method of the multiple adjustment neighborhood independent filling scheme sets is disabled will be marked, and the neighborhood independent filling scheme sets will be randomly adjusted multiple times according to the preset adjustment range. The multiple adjustment neighborhood independent filling scheme sets will be updated according to the adjustment results.

[0100] Furthermore, the filling control module 16 is used to perform the following methods:

[0101] By employing a sensorless network to perform stability identification on the graded filling process of the multiple filling units, stability identification results are obtained; based on the stability identification results, early warning and maintenance of the target goaf area are carried out.

[0102] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method of staged filling of a stope behind a sill pillar, characterised in that, The method includes: The structure of the target goaf is obtained, and the filling units are divided by combining the strip length, mining height, thickness and lithology of the overlying strata, and geological structure to obtain multiple filling units; A sensor network was deployed in the goaf area to monitor the status of multiple filling units and obtain multiple status monitoring data sequences. Based on the multiple state-aware data sequences, state time sequence perception is performed to obtain multiple state-aware features; Based on the multiple state-aware features, the multiple filling units are identified by risk level to obtain multiple identified filling units; The hierarchical filling scheme is analyzed independently by traversing the multiple marker filling units to obtain multiple independent filling schemes. The adjacency influence correction of the multiple independent filling schemes is performed by combining the positions of the multiple marker filling units to obtain the target hierarchical filling scheme. The remotely controlled unmanned forklift places the precast lightweight inflatable blocks according to the target graded filling plan. The directional drilling rig is used to construct grouting holes and perform secondary grouting control on the initial filling area of ​​the precast lightweight inflatable blocks. Based on the multiple state-aware data sequences, state timing perception is performed to obtain multiple state-aware features, including: By traversing the multiple state-aware data sequences, multi-scale feature extraction is performed to obtain multiple multi-scale temporal feature sets; Intra-set enhancement is performed on the multiple multi-scale temporal feature sets to obtain multiple state-aware features; By combining the positions of multiple marker filling units, adjacency correction is applied to the multiple independent filling schemes to obtain a target hierarchical filling scheme, including: Extract multiple high-risk identifier filling units from the multiple identifier filling units, and construct multiple high-risk identifier filling unit neighborhoods according to the preset adjacency influence bandwidth; Complexity is identified in the neighborhood of the multiple high-risk marker filling units to obtain multiple neighborhood complexities; Multiple neighborhood complexity are filtered according to a preset threshold, and multiple high-risk indicator filling units are mapped and extracted. Based on multiple independent filling schemes corresponding to multiple high-risk marker filling units, the independent filling schemes in the neighborhood of multiple high-risk marker filling units are adjusted in a guiding manner. Adjacency influence is corrected with the goal of optimal collaborative stability, resulting in a set of multiple corrected hierarchical filling schemes. The multiple independent filling schemes are updated based on the multiple sets of modified hierarchical filling schemes to obtain the target hierarchical filling scheme.

2. A method of staged filling of a stope behind a breast after mining according to claim 1, characterised in that, In-set enhancement is performed on the multiple multi-scale temporal feature sets to obtain multiple state-aware features, including: The multiple multi-scale temporal feature sets are sorted in descending order of recognition scale to obtain multiple multi-scale temporal feature sequences; Based on the multi-scale time-series feature of the preceding multi-scale time-series feature, the multi-scale time-series feature of the following multi-scale time-series feature is enhanced to obtain multiple multi-scale enhanced time-series feature sequences. State perception is performed based on the multiple multi-scale enhanced temporal feature sequences to obtain multiple state perception features.

3. A method of staged filling of a stope behind a breast after mining according to claim 1, characterised in that, By traversing the multiple identifier filling units and performing independent hierarchical filling scheme analysis, multiple independent filling schemes are obtained, including: Obtain preset graded filling rules, identify graded filling schemes based on the risk level identifiers of the multiple identified filling units, and obtain multiple independent filling schemes; The preset graded filling rules include: For low-risk filling units, standard lightweight inflatable blocks are used for filling; For medium-risk filling units, lightweight inflatable blocks are used for initial filling, followed by secondary grouting using micro-expansion inorganic grouting materials. For high-risk filling units, lightweight inflatable blocks are used for initial filling, followed by double grouting using micro-expansion inorganic grouting material and high-pressure resistant filling material.

4. The method for graded backfilling of goaf after end-face mining as described in claim 1, characterized in that, Complexity identification is performed on the neighborhoods of the multiple high-risk marker filling units to obtain multiple neighborhood complexities, including: Based on the risk level type and the quantity of the same risk level type, data is extracted from the neighborhood of the multiple high-risk identifier filling units to obtain multiple neighborhood risk level type sets and multiple sets of the same type of quantity. Weighted complexity identification is performed on the multiple neighborhood risk level type sets and the multiple sets of quantities of the same type to obtain multiple neighborhood complexities.

5. A method of staged filling of a stope behind a breast after mining according to claim 4, characterised in that, Based on multiple independent filling schemes corresponding to multiple high-risk marker filling units, the independent filling schemes within the neighborhood of multiple high-risk marker filling units are adjusted to optimize collaborative stability and correct adjacency effects, resulting in a set of multiple corrected hierarchical filling schemes, including: From multiple high-risk sign filling units and multiple independent filling schemes, extract the first high-risk sign filling unit and the first independent filling scheme corresponding to the first high-risk sign filling unit; By combining the positions of neighboring sign filling units within the neighborhood of the first leading high-risk sign filling unit, and the set of independent filling schemes in the neighborhood, the first independent filling scheme is identified for cooperative stability to obtain the first cooperative stability. The set of independent neighborhood filling schemes is randomly adjusted multiple times according to a preset adjustment range to obtain multiple sets of adjusted independent neighborhood filling schemes. The first independent filling scheme is then identified again based on the position between the neighboring identifier filling units and the multiple sets of adjusted neighboring independent filling schemes to obtain multiple adjusted cooperative stability. Determine whether there exists an adjustment co-stability greater than or equal to the first co-stability among the multiple adjustment co-stability. If so, take the set of independent filling schemes in the neighborhood corresponding to the maximum value among the multiple adjustment co-stability as the first modified hierarchical filling scheme set, and add the first modified hierarchical filling scheme set into the multiple modified hierarchical filling scheme set.

6. A method of staged filling of a stope behind a breast after mining according to claim 5 wherein, If none of the multiple adjustment coordination stabilityes is greater than or equal to the first coordination stability, then the number of times the adjustment method of the multiple adjustment neighborhood independent filling scheme sets is disabled will be marked, and the neighborhood independent filling scheme sets will be randomly adjusted multiple times according to the preset adjustment range. The multiple adjustment neighborhood independent filling scheme sets will be updated according to the adjustment results.

7. A method of staged filling of a stope behind a breast after mining according to claim 1, characterised in that, include: By employing a sensorless network to perform stability identification on the hierarchical filling process of the multiple filling units, stability identification results are obtained. Based on the stability identification results, early warning and maintenance are carried out for the target goaf area.

8. A staged filling system for a stope behind a face after the face has been mined, characterised in that, For implementing the graded backfilling method for goaf after endwall mining as described in any one of claims 1-7, the system comprises: Filling division module: Obtain the structure of the target goaf, and divide it into filling units based on strip length, mining height, thickness and lithology of overlying strata, and geological structure to obtain multiple filling units; Status perception module: Deploy a sensor network in the goaf area to perceive the status of multiple filling units and obtain multiple status perception data sequences; Feature extraction module: performs state temporal perception based on the multiple state-aware data sequences to obtain multiple state-aware features; Risk identification module: Identifies the risk level of the multiple filling units based on the multiple state perception features to obtain multiple identified filling units; Scheme formulation module: Iterates through the multiple marker filling units to perform independent analysis of hierarchical filling schemes, obtains multiple independent filling schemes, and performs adjacency influence correction on the multiple independent filling schemes based on the positions of the multiple marker filling units to obtain the target hierarchical filling scheme; Filling control module: Remotely control unmanned forklifts to place prefabricated lightweight inflatable blocks according to the target graded filling plan, and use directional drilling rigs to construct grouting holes to perform secondary grouting control on the initial filling area of ​​the prefabricated lightweight inflatable blocks.

Images