Mine operation big data security processing and analysis system

Abstract

The invention discloses a mining operation big data security processing and analysis system, and belongs to the technical field of data analysis. The method is used for solving the technical problems of poor data quality and high communication delay caused by mine operation environment influence in the existing scheme. Based on a mine sensor topological structure, space-time registration is carried out on environment data and operation data, an improved Kalman filtering algorithm is adopted to dynamically suppress electromagnetic interference and noise, and missing values are complemented through fractional order differential; constructing a roadway partition routing protocol, performing local feature extraction on the preprocessed data by an edge node, dynamically adjusting a transmission priority based on a data security level and time delay sensitivity, and transmitting high-priority data by adopting a 5G slicing technology; and fusing the calibrated data and the extracted key features by using a space-time diagram convolutional network, performing data analysis on the obtained fusion feature vector, outputting a real-time risk level, and dynamically adjusting a filtering scheme and a transmission bandwidth allocation scheme.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data analysis, and particularly relates to a mine operation big data security processing and analysis system. BACKGROUND

[0002] The mine operation safety-related big data refers to a mass, multi-source and dynamic data set directly related to safety production, which is collected and accumulated in real time through various sensors, monitoring systems, personnel positioning, equipment operation, environmental monitoring and other means in the process of mine production, operation and management; the data is used for risk early warning, accident prevention, emergency response and safety management decision-making, and is an important basis for realizing mine intelligence and intrinsic safety.

[0003] In the implementation of the existing mine operation big data security processing and analysis scheme, the complex environment of high temperature, high humidity and dust in the underground mine easily leads to damage or reading drift of the sensor; poor network coverage in remote mines easily leads to real-time data transmission interruption; limited bandwidth of the narrowband Internet of Things (NB-IoT) affects the immediate early warning of water leakage and gas outburst risks; and there are defects of uneven data quality and high underground communication delay. SUMMARY

[0004] The purpose of the present application is to provide a mine operation big data security processing and analysis system to solve the technical problems of poor data quality and high communication delay caused by the mine operation environment in the existing scheme.

[0005] The purpose of the present application can be achieved by the following technical scheme:

[0006] A mine operation big data security processing and analysis system comprises:

[0007] A multi-source heterogeneous data space-time calibration and enhancement module: based on the mine sensor topology structure, the environmental data and the operation data are space-time registered, the improved Kalman filter algorithm is used to dynamically suppress electromagnetic interference and noise, and the missing values are completed through fractional differential; the environmental data includes gas concentration and roof pressure; the operation data includes operation equipment vibration and current;

[0008] An edge cloud collaborative low-delay transmission module: a roadway partition routing protocol is constructed, the edge node performs local feature extraction on the preprocessed data, the transmission priority is dynamically adjusted based on the data security level and the time delay sensitivity, and the high-priority data is transmitted by using the 5G slicing technology;

[0009] A dynamic weighted fusion and risk assessment module: the space-time graph convolution network is used to fuse the calibrated data and the extracted key features, the obtained fusion feature vector is subjected to data analysis, and the real-time risk level is outputted;

[0010] Adaptive feedback control module: according to the output of the real-time risk level, dynamically adjust the filtering scheme and the transmission bandwidth allocation scheme, and dynamically implement the edge node local emergency truncation mechanism.

[0011] Preferably, based on the spatio-temporal aligned preprocessed data, an environment adaptive noise suppression model is constructed; for the missing segments in the de-noised data, a fractional differential model is used to complete:

[0012] wherein, is the state prediction value; t is the time index; D g is the fractional differential operator; ω k is the weight coefficient; N is the time window size; k is the time lag index in the fractional differential completion model.

[0013] Preferably, when constructing the roadway dynamic partition and routing protocol, when the sensor density of the monitoring area exceeds the threshold value, the partition algorithm is started, and the objective function is:

[0014] wherein, M is the number of partitions; C m is the sensor set of the mth partition; loc(s i ) is the three-dimensional coordinates of sensor s i ; μ m is the partition centroid coordinates.

[0015] Preferably, when performing edge node preprocessing and local feature extraction, outlier rejection is performed on the original sensor data, and three types of key features, namely time domain features, frequency domain features and space-time features, are extracted.

[0016] Preferably, according to the data security level S and the delay sensitivity D obtained by calculation, the priority score P trans is calculated:

[0017] P trans = α·S + (1-α)·D; wherein, α is the calculation weight; data analysis is performed on the priority score P trans , and according to the analysis result, dynamically select the URLLC slice, the eMBB plus URLLC mixed slice or the mMTC slice.

[0018] Preferably, when constructing the spatio-temporal graph convolution network model, a sensor-feature association graph is constructed; when designing the ST-GCN layer, the spatial convolution and the time convolution are included.

[0019] Preferably, through the spatial convolution and the time convolution of the ST-GCN layer, the data calibrated by spatio-temporal alignment and standardization processing and the key features are fused into a unified high-level feature vector, and the expression is:

[0020] wherein, F fusion is a fusion feature vector; χ is a coupling coefficient; respectively are gas, roof sub-network features.

[0021] Preferably, an improved Sigmoid function is used to output the real-time risk level R; the risk level R ranges from [0, 5], corresponding to five levels of safety, slight, medium, severe, fatal, respectively:

[0022] wherein, γ is a slope parameter; θ is an offset;

[0023] When R is greater than or equal to 3, an early warning signal is output and a 5G slice priority is linked.

[0024] Preferably, according to the output real-time risk level R, the process noise covariance Q(t) of Kalman filtering is adjusted; when R is greater than or equal to 3, a noise suppression mode is enabled:

[0025] wherein, R k is a measurement noise covariance; R0 is a basic measurement noise covariance; is a decay coefficient.

[0026] Preferably, when the edge node local emergency truncation mechanism is dynamically implemented, when the risk level R is greater than or equal to 4, the edge node local controller immediately performs motion truncation, data truncation and local alarm operation.

[0027] Compared with the prior art, the present application has the following beneficial effects:

[0028] The present application can solve the data mispositioning problem caused by the dispersed distribution of mine sensors through spatio-temporal registration, can effectively reduce the coordinate error; can effectively reduce the error caused by electromagnetic interference and improve the gas concentration measurement accuracy by using improved Kalman filtering combined with environment adaptive noise suppression; can effectively improve the missing data recovery accuracy by using fractional order differential to complete the historical trend and local mutation characteristics, and can avoid the interruption of safety monitoring caused by sensor offline.

[0029] The present application reduces the number of hops through partition routing, realizes end-to-end low latency combined with 5G URLLC slice, meets the real-time requirements of dangerous gas monitoring; dynamically adjusts the priority, which can effectively improve the transmission success rate of high safety level data, reduce the resource occupation of non-critical data, and realize efficient use of resources; through the closed-loop design of partition, preprocessing, priority and scheduling, the cooperative and efficient transmission of edge and cloud data is realized, so that the core parameters can support adaptive adjustment of the underground environment without human intervention.

[0030] The application models the ST-GCN layer through space-time dual domains, can effectively reduce the fusion error of gas and roof data, can effectively improve the contribution degree of key sensor data by using a dynamic weight mechanism, and realizes dynamic weighted fusion and real-time risk assessment of multi-source heterogeneous data through the whole process design of data alignment, space-time fusion and coupling prompt, improves the diversity and reliability of mine operation safety data analysis.

[0031] The filter parameter dynamic adjustment in the application can effectively reduce sensor data delay, and bandwidth allocation optimization can effectively improve key data transmission rate, thereby reducing the update period of risk level;When the sensor is disturbed by dust, the process noise covariance of the adaptive Kalman filter is adjusted, which can effectively reduce the data calibration error and improve the robustness of the system;Dynamic compression of non-critical data bandwidth can reduce 5G network congestion, so that the system remains stable in a multi-robot collaborative scene, and improves the resource optimization effect. BRIEF DESCRIPTION OF DRAWINGS

[0032] The application will be further described below with reference to the accompanying drawings.

[0033] Figure 1 The application is a flow chart of a mine operation big data safety processing and analysis system. DETAILED DESCRIPTION

[0034] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0035] As shown in Figure 1 The application is a mine operation big data safety processing and analysis system, which comprises:

[0036] The multi-source heterogeneous data space-time calibration and enhancement module: based on the mine sensor topology, the environmental data and operation data are space-time registered, the improved Kalman filter algorithm is used to dynamically suppress electromagnetic interference and noise, and the missing values are completed by fractional order differentiation;The environmental data includes gas concentration and roof pressure;The operation data includes operation equipment vibration and current;The specific steps include:

[0037] Based on a mine sensor deployment topology map, which includes latitude and longitude coordinates and roadway layering information, heterogeneous data is spatiotemporally aligned, corresponding to environmental and operational data. Coordinate transformation maps the three-dimensional positions of different sensors to a unified roadway coordinate system. This roadway coordinate system is constructed based on the actual roadway dimensions. Sensors include, but are not limited to, gas sensors and vibration sensors. The formula is:

[0038] P i ′=R′·P i +T; where i is the sensor index number; P i R is the original coordinate of the sensor, R′ is the rotation matrix, dynamically corrected based on the roadway direction, T is the translation vector used to compensate for installation errors, and P is the sensor's original coordinates. i ′ represents the coordinates after registration;

[0039] The sensor timestamps are calibrated using the IEEE 1588PTP protocol. Data with a delay exceeding Ams is marked as suspicious. A can be a value of 50. Time deviations are corrected through linear interpolation.

[0040] t sync =t raw +Δt·sign(t raw -t ref ); where t raw For the original timestamp, t ref The reference clock is Δt, which is a dynamic delay compensation value that is dynamically adjusted based on the wireless signal strength and ranges from [10ms to 100ms].

[0041] Based on the spatiotemporally aligned preprocessed data, an environment-adaptive noise suppression model is constructed:

[0042] in, t is the predicted state value; t is the time index; A is the state transition matrix, reflecting the temporal correlation of mine data, such as the gas concentration diffusion coefficient; B is the control input matrix, which correlates with the operating parameters of ventilation equipment; u t To control the input vector; P t Let A be the prediction error covariance matrix. T Let A be the transpose of A; Q be the process noise covariance matrix; z t H is the observation vector; H is the observation matrix; H T v is the transpose of H; t K represents the observed noise vector. t H is the Kalman gain; H is the observation matrix; R is the observation noise covariance; P t ' is the updated error covariance matrix; I is the identity matrix, with dimensions similar to P. t Consistency is used to maintain matrix dimension matching in the error covariance update formula; To update the post-state prediction value, the prediction and the observed final state value, such as the corrected real value of the gas concentration, are fused as input data for subsequent risk assessment;

[0043] For the missing segment in the denoised data, such as 2-10s data blank caused by sensor offline, a fractional differential model is used to complete it:

[0044] where D g is a fractional differential operator, the order g∈[0.3,0.7], and the high value is taken in a high dust environment to enhance the weight of historical data; ω k is a weight coefficient, which is dynamically allocated based on the mine operation cycle, and ω T' is the blasting cycle; when it is not a mining period, N is the time window size, which is 3 times the missing length, such as 5s missing corresponding to N=15; k is the time lag index in the fractional differential completion model, representing the historical data point sequence number used to predict the current missing value, and the value range is 1≤k≤N;

[0045] It needs to be explained that the unified coordinate system data after spatio-temporal registration as the filtering input can ensure the spatial consistency of noise suppression; the denoised data can provide high-quality time sequence for fractional differential missing value completion, avoiding noise amplification.

[0046] In the embodiment of the application, the spatio-temporal registration can solve the data misplacement problem caused by the scattered distribution of mine sensors, and can effectively reduce the coordinate error; the improved Kalman filter combined with environment adaptive noise suppression can effectively reduce the error caused by electromagnetic interference and improve the measurement accuracy of gas concentration; the fractional differential completion considering historical trend and local mutation characteristics can effectively improve the recovery accuracy of missing data and avoid the interruption of safety monitoring caused by sensor offline.

[0047] Edge cloud collaborative low-latency transmission module: construct a roadway partition routing protocol, the edge node extracts local features from the preprocessed data, dynamically adjusts the transmission priority based on data security level and latency sensitivity, and transmits high-priority data using 5G slicing technology; the specific steps include:

[0048] When the sensor density in the monitoring area exceeds the threshold, which can be 5 / 100m 2 , the partition algorithm is started;

[0049] Wherein, the partition algorithm uses improved K-means clustering, and the objective function is:

[0050] Wherein, M is the number of partitions, the value range is [3, 10], the upper limit is taken when the mining face complexity is high; C m is the sensor set of the mth partition; loc(s i ) is the three-dimensional coordinates of sensor s i ; x is the transverse direction of the roadway, y is the depth, and z is the height; μ m is the centroid coordinates of the partition, the initial value is taken as the blasting point position, and the iteration convergence threshold ∈ = 0.5;

[0051] When designing the routing protocol, each partition elects an edge node as the cluster head, and the remaining energy E res > 0.6xE max is screened, E max is the maximum energy, and the geographical routing plus time slot ALOHA mechanism is used between cluster heads; wherein, the time slot length T s = 20 ms, which matches the 5G frame structure;

[0052] The cluster head sends the priority Wherein, E min is the minimum energy, E min = 0.2E max , and the collision avoidance is realized;

[0053] When preprocessing the edge node and extracting the local features, the original sensor data x raw (t) is subjected to outlier rejection, and three types of key features, i.e., time domain features, frequency domain features and space-time features, are extracted, and the dimension is compressed to 1 / 10 of the original data; wherein, the expression of outlier rejection is:

[0054] Wherein, x clean (t) is the cleaned data; W is the window size, the default value is 5, corresponding to 5 sampling points, and σ is the window standard deviation;

[0055] In addition, the time domain features include: mean kurtosis Normal working condition K≈3, and when fault occurs, K> 5;

[0056] Frequency domain features: the main frequency f0 is extracted through fast Fourier transform, which corresponds to the equipment vibration frequency; for example, the ventilator f0 = 50 Hz; the coal mining machine f0 = 25 Hz;

[0057] Space-time features: the correlation coefficient ρ i,j between sensor i and j; x i , x j are the real-time sampling data of sensors i and j; cov(x i , x j ) is xi , x j covariance of x i , σ j standard deviation of x i , x j standard deviation of x

[0058] When performing data security level and latency sensitivity calculation and dynamically adjusting transmission priority, the data leakage consequences are quantified, and the security level S is calculated:

[0059] where ω k is different weight coefficients ω1, ω2, ω3; c k is different transmission data c1, c2, c3; c1 is the gas concentration data, and the corresponding weight coefficient ω1 default value is 0.5, and 5 is taken when it is over limit, and the over limit is specifically >1%; c2 is personnel positioning data, and the corresponding weight coefficient ω2 default value is 0.3, and 5 is taken when it enters the danger zone; c3 is the equipment control instruction, and the corresponding weight coefficient ω3 default value is 0.2, and 5 is taken for emergency stop instruction;

[0060] And, when the latency sensitivity D is calculated, the corresponding expression is: where T deadline is the data failure time, and the gas data is 500ms, and the video stream is 200ms; T trans is the current network transmission delay, which is obtained in real time through ping test, and the sampling period is 100ms;

[0061] According to the data security level S and the latency sensitivity D, the priority score P trans is calculated:

[0062] P trans = α·S + (1-α)·D; Wherein, α is the calculation weight, the default value is 0.7, the mine scene safety priority, α can be dynamically updated through the blockchain smart contract;

[0063] When P trans ≥0.8, it corresponds to high priority, and URLLC slice is selected; The latency of URLLC slice is <10ms;

[0064] When 0.5≤P trans <0.8, it corresponds to medium priority, and eMBB plus URLLC mixed slice is selected; The bandwidth of eMBB plus URLLC mixed slice is 100Mbps, and the latency is <50ms;

[0065] When P trans <0.5, it corresponds to low priority, and mMTC slice is selected; The latency of mMTC slice is <1s, and it supports massive connection;

[0066] Wherein, the base station side adopts proportional fair algorithm to allocate time-frequency resources:

[0067] Wherein, R alloc is the total available bandwidth of the base station; P trans,m is the average transmission priority of the mth partition.

[0068] In the embodiment of the application, the number of hops is reduced through partition routing, and low latency is realized end-to-end in combination with 5G URLLC slices to meet the real-time requirements of dangerous gas monitoring; dynamic priority scheduling can effectively improve the transmission success rate of high safety level data, while reducing the resource occupation of non-critical data, and realize efficient use of resources; through the closed-loop design of partition, preprocessing, priority, and scheduling, the cooperative and efficient transmission of edge and cloud data is realized, so that the core parameters can support adaptive adjustment of the downhole environment without manual intervention.

[0069] The dynamic weighted fusion and risk assessment module: adopt a spatio-temporal graph convolution network to fuse the calibrated data and extracted key features, and perform data analysis on the obtained fusion feature vector to output real-time risk level; the specific steps include:

[0070] The calibrated data and the extracted key features are spatio-temporally aligned and standardized, and the standardization can be Z-score normalization; spatio-temporal alignment and standardization are existing conventional technical solutions, and the specific implementation steps are not described here;

[0071] When constructing the spatio-temporal graph convolution network model, a sensor-feature association graph G=(V,E,A) is constructed;

[0072] Wherein, the node set V=V sen ∪V feat ; V sen is a sensor node; V feat is an edge feature node;

[0073] E is the edge set: calculate the Euclidean distance d ij of all sensor pairs, and retain the sensor pairs with a Euclidean distance d ij ≤5cm;

[0074] The adjacency matrix A∈R N×N ; the element is defined as:

[0075] Wherein, d ij is the Euclidean distance between nodes i and j; β is the distance attenuation coefficient, and the default value is 0.1; d0 is the maximum communication distance, and the default value is 50; e is the natural constant;

[0076] The ST-GCN layer is designed to include a spatial convolution and a temporal convolution;

[0077] The spatial convolution uses a Chebyshev polynomial to approximate a graph convolution kernel.

[0078] Y spatial is a spatial convolution output feature matrix; Θ k′ is a learnable matrix; k' is a polynomial order index; K is a polynomial order, and is 3; is a k' order Chebyshev polynomial. is a normalized Laplacian matrix; X is an input feature matrix; and the Chebyshev polynomial is a Chebyshev polynomial.

[0079] The temporal convolution uses a 1D convolution kernel with a kernel size t ker nel = 3, an input channel c in = 64, and an output channel c out = 128.

[0080] The spatial convolution and the temporal convolution of the ST-GCN layer fuse the data and the key features after the spatio-temporal alignment and the standardization processing into a unified high-level feature vector, and the expression is as follows:

[0081] F fusion is a fusion feature vector; χ is a coupling coefficient, and the dynamic update formula is as follows: gas th is a gas safety threshold. are respectively a gas and a roof sub-network feature.

[0082] An improved Sigmoid function is used to output a real-time risk level R. The risk level R ranges from 0 to 5, and corresponds to five levels of safety, slight, medium, severe, and fatal.

[0083] wherein γ is a slope parameter, and the default value is 0.8; and θ is an offset, and the default value is 0.5.

[0084] When R ≥ 3, an early warning signal A alert = 1 is output, and the 5G slice priority is parallelly moved, and the value of P trans is increased to 0.95.

[0085] In the embodiment of the present application, the ST-GCN layer is modeled through space-time dual domains, which can effectively reduce the fusion error of gas and roof data, and the dynamic weight mechanism can effectively improve the contribution of key sensor data; through the whole process design of data alignment, space-time fusion and coupling prompt, dynamic weighted fusion and real-time risk assessment of multi-source heterogeneous data are realized, and the diversity and reliability of mine operation safety data analysis are improved.

[0086] The adaptive feedback control module dynamically adjusts the filtering scheme and the transmission bandwidth allocation scheme according to the output real-time risk level, and dynamically implements the edge node local emergency truncation mechanism. The specific steps include:

[0087] According to the output real-time risk level R, the process noise covariance Q(t) of Kalman filtering is adjusted, and the formula is:

[0088] Q(t) = Q0 · (1 + k R · R); wherein Q0 is the basic process noise covariance, corresponding to the filtering stability requirement in the safe state; k R is the risk influence coefficient, the default value is 0.2, which is obtained by training the mine historical data;

[0089] When R≥3, the noise suppression mode is enabled:

[0090] Wherein, R k is the measurement noise covariance; R0 is the basic measurement noise covariance; is the attenuation coefficient, the default value is 0.3;

[0091] When dynamically adjusting the transmission bandwidth allocation scheme, the transmission data is divided into three categories according to the risk sensitivity; specifically:

[0092] High priority: gas concentration, roof pressure; the corresponding weight value is 0.6;

[0093] Medium priority: temperature, humidity; the corresponding weight value is 0.3;

[0094] Low priority: equipment status log; the corresponding weight value is 0.1;

[0095] When dynamically allocating bandwidth, the expression of dynamic allocation is:

[0096] Wherein, a R is the risk bandwidth coefficient, the default value is 0.4; B total is the total transmission bandwidth; ω i is the data priority weight; when R≥2, it is triggered to ensure that the bandwidth proportion of high priority data is increased to ≥70%;

[0097] When the edge node local emergency interruption mechanism is dynamically implemented, when the risk level R is greater than or equal to 4, the edge node local controller immediately performs the motion interruption, data interruption and local alarm operation;

[0098] When the motion interruption operation is performed, the stop command Stop = 1 is sent to the working device, and the trajectory tracking control signal is cut off;

[0099] When the data interruption operation is performed, only the transmission of gas, roof and emergency stop signals is reserved, and other non-critical data such as environmental images are closed, and the bandwidth occupation is reduced to the preset minimum threshold B min ; The specific value of the minimum threshold B min is not limited, and can be customized according to the application requirements of the actual application scene;

[0100] When the local alarm operation is performed, the audible and visual alarm is started, and the fault snapshot is recorded, and the fault snapshot includes the original data of 3s before triggering.

[0101] It should be noted that the real-time risk level R is used as the only adjusting variable, the step-by-step response from normal control, early warning and emergency interruption is realized by quantifying the risk level, and control shock caused by a single threshold can be avoided.

[0102] In the embodiment of the application, the filter parameter dynamic adjustment can effectively reduce the sensor data delay, the bandwidth allocation optimization can effectively improve the critical data transmission rate, and then the risk level update period can be reduced; when the sensor is disturbed by dust, the process noise covariance of the Kalman filter is adaptively adjusted, which can effectively reduce the data calibration error and improve the robustness of the system; the non-critical data bandwidth is dynamically compressed, which can reduce the 5G network congestion, so that the system still maintains stability in the multi-robot collaborative scene, and the resource optimization effect is improved; through the three-level adaptive adjustment of filtering, bandwidth and control driven by the risk level, the full-link closed loop of risk perception, resource allocation and safety control in the unknown curved surface tracking process is realized, and the core parameters support the dynamic adaptation of the complex mine environment.

[0103] In the several embodiments provided in the present application, it should be understood that the disclosed system can be implemented in other ways. For example, the above-described embodiments of the application are only illustrative, for example, the division of the modules is only a logical function division, and other division manners can be used in actual implementation.

[0104] The modules described as separate components can or can not be physically separated, and the components shown as modules can or can not be physical modules, which can be located in one place or distributed on multiple network modules. According to actual needs, part or all of the modules can be selected to achieve the purpose of the embodiment scheme.

[0105] In addition, each functional module in each embodiment of the present application can be integrated in one processing module, or each module can exist physically alone, or two or more modules can be integrated in one module. The integrated module can be realized in the form of hardware or in the form of hardware plus software functional module.

[0106] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be realized in other specific forms without departing from the essential characteristics of the present application.

[0107] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A mine operation big data security processing analysis system, characterized in that, Comprise: Multi-source heterogeneous data space-time calibration and enhancement module: Based on the mine sensor topology, the environmental data and operation data are space-time registered, the improved Kalman filter algorithm is used to dynamically suppress electromagnetic interference and noise, and the missing values are completed by fractional order differential; Environmental data includes gas concentration, roof pressure; Operation data includes operation equipment vibration, current; Edge cloud collaborative low delay transmission module: Constructing roadway partition routing protocol, edge nodes extract local features from preprocessed data, dynamically adjusting transmission priority based on data security level and delay sensitivity, high priority data using 5G slicing technology transmission; Dynamic weighted fusion and risk assessment module: Using spatio-temporal graph convolution network to fuse calibrated data and extracted key features, performing data analysis on the obtained fusion feature vector, and outputting real-time risk level; Adaptive feedback control module: According to the output real-time risk level, dynamically adjust the filtering scheme and transmission bandwidth allocation scheme, and dynamically implement the edge node local emergency truncation mechanism.

2. The mine operation big data security processing and analysis system according to claim 1, characterized in that, Based on the preprocessed data aligned in space and time, an environment adaptive noise suppression model is constructed; For the missing segments in the denoised data, a fractional order differential model is used to complete: where, is the state prediction value; t is the time index; D g is the fractional differential operator; ω k is the weight coefficient; N is the time window size; k is the time lag index in the fractional differential completion model.

3. The mine operation big data security processing and analysis system of claim 2, wherein, When the sensor density of the monitoring area exceeds the threshold, the partition algorithm is started, and the objective function is: where M is the number of partitions; C m is the sensor set of the mth partition; loc(s i ) is the three-dimensional coordinate of sensor s i m is the partition centroid coordinate.​ 4. The mine operation big data security processing and analysis system according to claim 3, characterized in that, When preprocessing and local feature extraction of edge nodes, outlier rejection is performed on the original sensor data, and three types of key features, time domain features, frequency domain features and space-time features, are extracted.

5. The big data security processing and analysis system of mine operations of claim 4, wherein, Based on the data security level S and the time delay sensitivity D obtained by calculation, a priority score P is calculated trans : P trans = a · S + (1 - a) · D; where a is a calculation weight; the priority score P trans perform data analysis, and dynamically select a URLLC slice, an eMBB plus URLLC mixed slice, or an mMTC slice according to the analysis result.

6. The big data security processing and analysis system of mine operations of claim 1, wherein, When constructing the spatio-temporal graph convolution network model, a sensor-feature association graph is constructed; When designing the ST-GCN layer, it contains spatial convolution and temporal convolution.

7. The big data security processing and analysis system of mine operations of claim 6, wherein, Through the spatial convolution and temporal convolution of the ST-GCN layer, the data calibrated by space-time alignment and standardization processing and the key features are fused into a unified high-level feature vector, the expression is: where F fusion is the fusion feature vector; χ is the coupling coefficient; are the gas and roof sub-network features, respectively.

8. The big data security processing and analysis system of mine operations of claim 7, wherein, The improved Sigmoid function is used to output the real-time risk level R; The risk level R ranges from 0 to 5, corresponding to five levels of safety, slight, moderate, severe and fatal: where γ is a slope parameter; and θ is an offset. When R≥3, output the warning signal and link the 5G slicing priority. 9.The mine operation big data security processing and analysis system of claim 8, wherein, According to the output real-time risk level R, adjust the process noise covariance Q(t) of Kalman filter; When R≥3, enable noise suppression mode: where R k is the measurement noise covariance; R0is the underlying measurement noise covariance; is the attenuation factor.

10. The big data security processing and analysis system of mine operations of claim 9, wherein, When implementing the edge node local emergency truncation mechanism dynamically, when the risk level R≥4, the edge node local controller immediately performs motion truncation, data truncation and local alarm operation.

Images