Filling mining safety management method and system based on intelligent monitoring

Abstract

The invention belongs to the field of intelligent safety management, and discloses a filling mining safety management method and system based on intelligent monitoring, and the method comprises the steps: collecting and preprocessing safety management data of a filling mining area; the safety management data comprises environment data, operation condition data and filling material conveying data; performing time sequence analysis and anomaly detection on the preprocessed security management data to obtain a data feature vector; establishing a safety risk assessment model by adopting an artificial bee colony algorithm, and assessing the safety state of the filling mining area according to the data feature vector and the historical accident data to obtain a safety level assessment result; and based on the safety level evaluation result, an emergency plan and an operation adjustment scheme are established, and filling mining safety management based on intelligent monitoring is completed.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent safety management, and in particular to a backfill mining safety management method and system based on intelligent monitoring. Background Art

[0002] During the backfill mining process, there are many safety hazards and risk factors due to the complex and changeable mining environment. Traditional manual inspection and supervision methods are difficult to fully and real-time grasp the safety status of the mining site, and there are blind spots and lags in supervision. At the same time, the massive amount of monitoring data also brings challenges to safety management. How to quickly identify and warn of safety risks from huge amounts of data has become an urgent problem to be solved. In addition, safety assessments often rely on the experience and subjective judgment of experts, lack of quantitative indicators and scientific decision-making basis, resulting in difficulty in improving the accuracy and efficiency of safety management. Therefore, there is an urgent need for a backfill mining safety management method and system based on intelligent monitoring. By building an integrated platform for data collection, transmission, storage, and analysis, combined with machine learning algorithms, the safety status of each link of backfill mining can be comprehensively monitored and risk assessed. Summary of the invention

[0003] In order to solve the above technical problems, the present invention provides a backfill mining safety management method and system based on intelligent monitoring.

[0004] A backfill mining safety management method based on intelligent monitoring, the method comprising:

[0005] Collect and pre-process the safety management data of the filling mining area; the safety management data includes environmental data, operating condition data and filling material transportation data;

[0006] Performing time series analysis and anomaly detection on the preprocessed security management data to obtain a data feature vector;

[0007] An artificial bee colony algorithm is used to establish a safety risk assessment model, and the safety status of the filling mining area is evaluated based on the data feature vector and historical accident data to obtain a safety level assessment result;

[0008] Based on the safety level assessment results, an emergency plan and operation adjustment plan are established to complete the safety management of filling mining based on intelligent monitoring.

[0009] Preferably, the environmental data include stress and strain, displacement, temperature, humidity, oxygen concentration, harmful gas concentration, dust concentration and geological pressure of the filling and mining area;

[0010] The operation condition data includes personnel activity data, equipment operation status and filling operation progress; wherein the personnel activity data includes personnel location, operation time and operation type;

[0011] The filling material conveying data includes the filling material type, quality, quantity, conveying speed, conveying pressure and material humidity.

[0012] Preferably, the method for obtaining the data feature vector includes:

[0013] Using an autoregressive integrated moving average model to perform time series modeling on the security management data to obtain data trends and data periodicity patterns;

[0014] Obtaining time series characteristics of security management data based on the data trend and the data periodicity pattern;

[0015] Using a long short-term memory network, predicting the multidimensional time series of the security management data, obtaining a data anomaly pattern, and using the data anomaly pattern as an anomaly pattern feature;

[0016] The statistical features of the security management data are extracted, and combined with the time series features and the abnormal pattern features to obtain the data feature vector.

[0017] Preferably, the method for obtaining data trends and data periodicity patterns includes:

[0018] The order of the autoregressive integrated moving average model is obtained by analyzing the autocorrelation function and the partial autocorrelation function; wherein the order includes the order of the autoregressive term, the number of differences, and the order of the moving average term;

[0019] Based on the least square method, the parameters of the autoregressive integrated moving average model are estimated to obtain the initial autoregressive integrated moving average model;

[0020] Performing residual analysis on the initial autoregressive integrated moving average model, and adjusting the initial autoregressive integrated moving average model according to the residual analysis result to obtain an autoregressive integrated moving average optimization model;

[0021] Long-term trends and periodic patterns in the security management data are extracted based on the autoregressive integrated moving average optimization model.

[0022] Preferably, the method for obtaining the data anomaly pattern includes:

[0023] Extracting time series of different types of data in the security management data as sub-time series, and obtaining a multidimensional time series based on the sub-time series;

[0024] Calculating a low-dimensional compressed representation of the sub-time series using an LSTM encoder;

[0025] Processing the low-dimensional compressed representation based on an LSTM decoder to obtain a reconstructed time series;

[0026] Calculating a reconstruction error between the sub-time series and the corresponding reconstructed time series;

[0027] All reconstruction error series slices are concatenated to obtain a multi-dimensional vector, and the reconstruction error of the multi-dimensional vector is calculated;

[0028] sorting the reconstruction errors of the multidimensional vectors to obtain anomaly scores of the security management data;

[0029] Based on the anomaly score, the data anomaly pattern is obtained.

[0030] Preferably, the method of establishing a security risk assessment model using an artificial bee colony algorithm includes:

[0031] Randomly generate an initial nectar source and initialize model parameters; wherein the model parameters include the maximum number of nectar sources to be mined, the number of mining times, and the maximum number of outer cycles;

[0032] Calculating a fitness function based on the model parameters and the initial nectar source;

[0033] Randomly generate a new nectar source, and calculate the fitness function value of the new nectar source based on the fitness function;

[0034] When the fitness function value of the new nectar source satisfies the preset fitness function value, the nectar source is updated;

[0035] Based on the updated honey source, the model parameters are updated, and when the model parameters meet the preset threshold, the optimal model parameters are obtained;

[0036] Based on the optimal model parameters, the security risk assessment model is constructed.

[0037] The present invention also provides a backfill mining safety management system based on intelligent monitoring, the system is used to implement any one of the methods described, including:

[0038] A data acquisition module, used to collect and pre-process the safety management data of the filling mining area; the safety management data includes environmental data, operating condition data and filling material transportation data;

[0039] A feature extraction module, used to perform time series analysis and anomaly detection on the pre-processed security management data to obtain a data feature vector;

[0040] A risk assessment module is used to establish a safety risk assessment model using an artificial bee colony algorithm, and to assess the safety status of the backfill mining area based on the data feature vector and historical accident data to obtain a safety level assessment result;

[0041] The emergency dispatch module is used to establish emergency plans and operation adjustment plans based on the safety level assessment results, and complete the backfill mining safety management based on intelligent monitoring.

[0042] Preferably, the feature extraction module includes:

[0043] A time series analysis unit, configured to use an autoregressive integrated moving average model to perform time series modeling on the security management data to obtain data trends and data periodicity patterns; and obtain time series characteristics of the security management data according to the data trends and the data periodicity patterns;

[0044] an anomaly detection unit, configured to use a long short-term memory network to predict the multidimensional time series of the security management data, obtain a data anomaly pattern, and use the data anomaly pattern as an anomaly pattern feature;

[0045] The feature vector acquisition unit is used to extract the statistical features of the security management data, and obtain the data feature vector in combination with the time series features and the abnormal pattern features.

[0046] Compared with the prior art, the beneficial effects of the present invention are: through intelligent monitoring and data analysis technology, the safety status of the backfill mining area can be grasped in real time, and potential safety hazards can be discovered and dealt with in a timely manner. This helps to improve the safety management level of mining enterprises and reduce the occurrence of safety accidents. The operation adjustment plan formulated based on the safety risk assessment results can solve the problems and bottlenecks in production in a targeted manner. By optimizing the production process, the efficiency and quality of backfill mining can be improved and the production cost can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0048] Figure 1 A flow chart of a backfill mining safety management method based on intelligent monitoring according to an embodiment of the present invention; DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0051] Example

[0052] like Figure 1 As shown, an embodiment of the present invention provides a backfill mining safety management method based on intelligent monitoring, the method comprising:

[0053] S1: Collect and pre-process safety management data of the filling and mining area; the safety management data includes environmental data, operating condition data and filling material transportation data; a further implementation method is that the environmental data includes stress and strain, displacement, temperature, humidity, oxygen concentration, harmful gas concentration, dust concentration and geological pressure of the filling and mining area; in this embodiment, the environmental data is collected in real time using a sensor network deployed in the mining area.

[0054] The operation condition data includes personnel activity data, equipment operation status and filling operation progress; among them, personnel activity data includes personnel location, operation time and operation type; in this embodiment, high-definition network cameras are installed at key locations in the filling and mining area to transmit the on-site images to the monitoring center in real time. These cameras have night vision, waterproof, explosion-proof and other functions to adapt to the complex mining environment. Through real-time analysis of the collected video stream, the operator's face recognition and behavior recognition are performed to obtain the operator's location, operation time and operation type. The built-in sensors of the equipment are used to collect the operation status of the operation equipment and the progress of the filling operation.

[0055] Specifically, regarding the identification of operator behavior, this embodiment adopts the following method:

[0056] Collect video data of the workers' working process and extract all human posture sequences in each frame of video. Specifically, based on OpenPose, the human key point information extracted from the video is used to construct a human skeleton sequence. The human key point information includes the position and direction parameters of the joint points, and their changes over time form the trajectory of human movement.

[0057] Using the keyframe localization network, the posture sequence of each person is scored, and then the abnormal posture sequence is screened out using a preset threshold; specifically, the abnormal curve score is obtained based on the transformation function of the human skeleton sequence and N layers of normalizing flows. Normalizing flows is a powerful generative model that can map simple probability distributions (such as standard normal distribution) to complex data distributions through a series of reversible transformation functions. Here, N layers of normalizing flows are used to transform the human skeleton sequence. Each layer of transformation will change the distribution of the data. By stacking multiple layers of transformation, a deeper representation of the data is obtained. After N layers of normalizing flow transformation, the data distribution of the human skeleton sequence will change. This change is used to calculate the abnormal curve score. Specifically, the data distribution difference before and after the transformation is compared, and the abnormal posture sequence is screened out using a preset threshold as the basis for the abnormal score. The abnormal curve score reflects the degree of abnormality of the human posture in time series. By monitoring the changes in this score, abnormal behavior of the human posture can be discovered and responded to in a timely manner.

[0058] The behavior recognition network that integrates attention and enhanced residual is used to classify the abnormal posture sequences and locate and identify the abnormal behaviors of operators in the video.

[0059] Specifically, the behavior recognition network structure adopted in this embodiment is composed of a spatial convolutional network EA-GCN (enhanced attention-GCN) fused with residual attention enhancement, a channel attention mechanism SE, a multi-branch temporal convolutional network EA-TCN (enhanced attention-TCN) fused with attention, and a spatial attention mechanism SAM.

[0060] EA-GCN is used to capture the spatial relationship between human joints in video frames and highlight the importance of key joints by enhancing the attention mechanism. The dynamic graph convolutional neural network is introduced in the spatial convolution, so that the adjacency matrix can automatically learn the change pattern of nodes and edges, and dynamically adjust the network structure according to the changes, so as to better adapt to and capture the dynamic features in the data.

[0061] The channel attention mechanism weights the channels of the feature map to enhance the model's ability to select important features and improve the robustness of feature expression.

[0062] EA-TCN is used to process the temporal dependencies in video sequences and improve the model's ability to recognize key frames in time series by enhancing the attention mechanism.

[0063] SAM further highlights key areas in the spatial dimension, helping the model to more accurately locate where abnormal behaviors occur.

[0064] The specific network workflow is as follows:

[0065] 1. Input processing: First, extract the human joint point coordinate sequence from the video stream as the input of the network.

[0066] 2. Spatial feature extraction: EA-GCN is used to perform spatial graph convolution operations on the input human joint point coordinate sequence to capture the spatial relationship between the joint points and highlight the key joint points through the enhanced attention mechanism.

[0067] 3. Channel attention enhancement: The output of EA-GCN is processed through the SE channel attention mechanism, each channel of the feature map is weighted, and the feature expression of important channels is enhanced.

[0068] 4. Temporal feature extraction: The feature map enhanced by channel attention is input into EA-TCN, and a temporal convolution operation is performed to capture the temporal dependencies in the video sequence and identify key frames through the enhanced attention mechanism.

[0069] 5. Spatial Attention Mechanism: The SAM spatial attention mechanism is applied on the output of EA-TCN to further highlight the key areas where abnormal behaviors occur in the spatial dimension.

[0070] 6. Behavior classification and abnormal behavior identification: Finally, the features processed by the spatial attention mechanism are input into the fully connected layer for classification, the specific behavior category of the workers in the video is identified, and abnormal behavior is defined according to the preset filling mining operation specifications to determine whether there is abnormal behavior and its type.

[0071] The filling material delivery data includes the filling material type, quality, quantity, delivery speed, delivery pressure and material humidity. The filling material delivery data is collected by reading the filling material record sheet and video monitoring.

[0072] In this embodiment, the preprocessing of the security management data includes:

[0073] Data cleaning: remove duplicate data, process missing values ​​(such as filling them with interpolation, mean value or model-based predicted values), correct erroneous data, etc.

[0074] Data formatting: Convert data into a unified format to facilitate subsequent processing and analysis.

[0075] Data normalization / standardization: Convert data of different magnitudes to the same scale to improve the accuracy of analysis results.

[0076] S2: Performing time series analysis and anomaly detection on the pre-processed security management data to obtain a data feature vector; a further implementation method is that the method for obtaining the data feature vector includes:

[0077] S21: Use the autoregressive integrated moving average model to perform time series modeling on the security management data to obtain data trends and data periodicity patterns; based on the data trends and data periodicity patterns, obtain the time series characteristics of the security management data;

[0078] In a further embodiment, the method of obtaining data trends and data periodicity patterns includes:

[0079] S211: Obtaining the order of the autoregressive integrated moving average model by analyzing the autocorrelation function and the partial autocorrelation function; wherein the order includes the order of the autoregressive term, the number of differences, and the order of the moving average term;

[0080] Specifically, the autocorrelation function of the safety management data is calculated, the correlation of the data at different lag periods shown in the autocorrelation graph is observed, and the order (p) of the autoregressive term is determined by finding the autocorrelation coefficient that is significantly not zero.

[0081] Calculate the partial autocorrelation function of the safety management data, observe the correlation of the data shown in the partial autocorrelation diagram after removing the influence of all previous lag periods, and determine the order (q) of the moving average term by finding the partial autocorrelation coefficient that is significantly non-zero.

[0082] If the data is not stationary, it needs to be differentiated. The number of differences d is determined by observing the trend and seasonality of the data. The differentiated data should meet the stationary requirements, that is, its mean, variance and autocovariance do not change over time.

[0083] S212: Based on the least squares method, estimate the parameters of the autoregressive integrated moving average model to obtain an initial autoregressive integrated moving average model; in this embodiment, the parameters of the autoregressive integrated moving average model include an autoregressive coefficient, a moving average coefficient, and an intercept term.

[0084] S213: Perform residual analysis on the initial autoregressive integrated moving average model, and adjust the initial autoregressive integrated moving average model according to the residual analysis results to obtain the autoregressive integrated moving average optimization model; specifically, use the initial autoregressive integrated moving average model to fit the data, calculate the residual sequence, and check the properties of the residual sequence, including its mean, variance, autocorrelation, and heteroskedasticity. If the residual sequence shows significant autocorrelation or heteroskedasticity, it means that the model may not fully fit the data. Regarding the adjustment of the model based on the residual analysis results, for example, if the residual sequence has autocorrelation, it may be necessary to increase the order of the moving average term; if the residual sequence has heteroskedasticity, it may be necessary to consider using weighted least squares or generalized autoregressive conditional heteroskedasticity (GARCH) model.

[0085] S214: Extracting long-term trends and cyclical patterns in security management data based on an autoregressive integrated moving average optimization model.

[0086] S22: Using a long short-term memory network, predict the multidimensional time series of security management data, obtain data anomaly patterns, and use the data anomaly patterns as anomaly pattern features; the long short-term memory network used in this embodiment is a network improved by an autoencoder, and the specific improvement process is as follows:

[0087] The first part of the network is composed of a combination of multiple long short-term memory networks and autoencoders. The specific number is determined according to the dimension of the input multidimensional time series. The first part reconstructs and fits the sub-time series of the multidimensional time series, and uses the reconstruction error of the obtained time series as the local feature; the second part is a fully connected network autoencoder, which is used to concatenate the reconstruction errors of each sequence. The multidimensional vector reconstruction error obtained after concatenation is the global feature of the multidimensional time series.

[0088] A further implementation method is that the method for obtaining the data anomaly pattern includes:

[0089] S221: extracting time series of different types of data in the security management data as sub-time series, and obtaining a multidimensional time series based on the sub-time series;

[0090] S222: Calculate the low-dimensional compressed representation of the sub-time series using the LSTM encoder;

[0091] S223: Process the low-dimensional compressed representation based on the LSTM decoder to obtain a reconstructed time series;

[0092] S224: Calculate the reconstruction error between the sub-time series and the corresponding reconstructed time series; the calculation formula is as follows:

[0093] In the formula, n represents the length of the time series, X j is the sub-time series, X' j To reconstruct the time series.

[0094] S225: concatenate all reconstruction error series slices to obtain a multi-dimensional vector, and calculate the reconstruction error of the multi-dimensional vector; the calculation formula is as follows:

[0095] In the formula, m represents the dimension of the input multidimensional time series, X i Represents a multidimensional vector, X i ' represents the reconstruction of multidimensional vector.

[0096] S226: sorting the reconstruction errors of the multidimensional vectors to obtain anomaly scores of the security management data;

[0097] Specifically, based on the reconstruction error of the time series in step S224 and the reconstruction error of the multidimensional vector in step S225 and the preset regularization parameter, the loss function of the improved long short-term memory network is obtained. Based on the loss function value, the model parameters are optimized using the back propagation algorithm. Iterate steps S221-S226 until the loss function converges, and then execute step S227.

[0098] S227: Based on the anomaly score, a data anomaly pattern is obtained. The anomaly score is the multidimensional vector reconstruction error. When the multidimensional vector reconstruction error is greater than a preset threshold, the data is determined to be abnormal.

[0099] S23: Extract statistical features of the safety management data, and combine the time series features and the abnormal pattern features to obtain a data feature vector. In this embodiment, the statistical features include mean, variance, maximum, minimum, median, quartile, etc. In particular, the abnormal pattern features include the above-mentioned recognition results of abnormal behaviors of operators.

[0100] S3: Use artificial bee colony algorithm to establish a safety risk assessment model, and evaluate the safety status of the filling mining area based on data feature vectors and historical accident data to obtain safety level assessment results;

[0101] A further implementation method is that the method of establishing a security risk assessment model using an artificial bee colony algorithm includes:

[0102] Randomly generate initial nectar sources and initialize model parameters; the model parameters include the maximum number of nectar sources to be mined, the number of mining times, and the maximum number of outer cycles;

[0103] Calculate the fitness function based on the model parameters and the initial nectar source;

[0104] Randomly generate new nectar sources and calculate the fitness function value of the new nectar source based on the fitness function;

[0105] When the fitness function value of the new nectar source meets the preset fitness function value, the nectar source is updated;

[0106] Based on the updated honey source, the model parameters are updated. When the model parameters meet the preset threshold, the optimal model parameters are obtained.

[0107] Based on the optimal model parameters, a security risk assessment model is constructed. Specifically, the optimal model parameters are substituted into the covariance function of the Gaussian process, the Gaussian kernel function is determined, and the prior distribution of the optimal model parameters is calculated. Based on the prior distribution, the joint prior distribution is calculated to obtain the target equation. Based on the target equation, the construction of the security risk assessment model is completed.

[0108] The risk assessment model calculates the safety level of the filling mining area based on the input data feature vector and the level classification of historical accidents.

[0109] The security level is divided according to actual needs, for example, into three levels: high, medium, and low, or more detailed levels.

[0110] S4: Based on the safety level assessment results, establish emergency plans and operation adjustment plans, and complete the safety management of filling mining based on intelligent monitoring.

[0111] Obtain the real-time assessment result data generated by the security risk assessment system, compare the assessment result with the preset security warning threshold, and determine whether it exceeds the threshold range. If the assessment result exceeds the warning threshold, the warning information generation module is triggered to automatically generate the warning information content that meets the requirements based on the preset warning information template and assessment result data. Call the SMS and voice call sending interface to send the generated warning information in the form of text and voice to the pre-configured on-site operator contact information. Push the warning information to the security management platform, and display the key information such as the severity of the warning event, the time of occurrence, and the associated risk factors through the visual interface.

[0112] On the other hand, the present invention also provides a backfill mining safety management system based on intelligent monitoring, the system is used to implement any one of the methods, including:

[0113] Data acquisition module, used to collect and pre-process the safety management data of the filling mining area; the safety management data includes environmental data, operating condition data and filling material transportation data;

[0114] The feature extraction module is used to perform time series analysis and anomaly detection on the pre-processed security management data to obtain data feature vectors;

[0115] The risk assessment module is used to establish a safety risk assessment model using an artificial bee colony algorithm, and to assess the safety status of the filling mining area based on data feature vectors and historical accident data to obtain safety level assessment results;

[0116] The emergency dispatch module is used to establish emergency plans and operation adjustment plans based on the safety level assessment results, and complete the filling mining safety management based on intelligent monitoring.

[0117] A further implementation is that the feature extraction module comprises:

[0118] A time series analysis unit is used to perform time series modeling on the security management data using an autoregressive integrated moving average model to obtain data trends and data periodicity patterns; and to obtain time series characteristics of the security management data based on the data trends and data periodicity patterns;

[0119] An anomaly detection unit, used to use a long short-term memory network to predict the multi-dimensional time series of security management data, obtain data anomaly patterns, and use the data anomaly patterns as anomaly pattern features;

[0120] The feature vector acquisition unit is used to extract the statistical features of the security management data and obtain the data feature vector by combining the time series features and the abnormal pattern features.

[0121] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A backfill mining safety management method based on intelligent monitoring, characterized in that: The method comprises: Collect and pre-process the safety management data of the filling mining area; the safety management data includes environmental data, operating condition data and filling material transportation data; Performing time series analysis and anomaly detection on the preprocessed security management data to obtain a data feature vector; An artificial bee colony algorithm is used to establish a safety risk assessment model, and the safety status of the filling mining area is evaluated based on the data feature vector and historical accident data to obtain a safety level assessment result; Based on the safety level assessment results, an emergency plan and operation adjustment plan are established to complete the safety management of filling mining based on intelligent monitoring.

2. The method according to claim 1, characterized in that The environmental data include stress and strain, displacement, temperature, humidity, oxygen concentration, harmful gas concentration, dust concentration and geological pressure in the filling mining area; The operation condition data includes personnel activity data, equipment operation status and filling operation progress; wherein the personnel activity data includes personnel location, operation time and operation type; The filling material conveying data includes the filling material type, quality, quantity, conveying speed, conveying pressure and material humidity.

3. The method according to claim 1, characterized in that Methods for obtaining data feature vectors include: Using an autoregressive integrated moving average model to perform time series modeling on the security management data to obtain data trends and data periodicity patterns; Obtaining time series characteristics of security management data based on the data trend and the data periodicity pattern; Using a long short-term memory network, predicting the multidimensional time series of the security management data, obtaining a data anomaly pattern, and using the data anomaly pattern as an anomaly pattern feature; The statistical features of the security management data are extracted, and combined with the time series features and the abnormal pattern features to obtain the data feature vector.

4. The method according to claim 3, characterized in that Methods for obtaining data trends and periodic patterns include: The order of the autoregressive integrated moving average model is obtained by analyzing the autocorrelation function and the partial autocorrelation function; wherein the order includes the order of the autoregressive term, the number of differences, and the order of the moving average term; Based on the least square method, the parameters of the autoregressive integrated moving average model are estimated to obtain the initial autoregressive integrated moving average model; Performing residual analysis on the initial autoregressive integrated moving average model, and adjusting the initial autoregressive integrated moving average model according to the residual analysis result to obtain an autoregressive integrated moving average optimization model; Long-term trends and periodic patterns in the security management data are extracted based on the autoregressive integrated moving average optimization model.

5. The method according to claim 3, characterized in that: Methods for obtaining data anomaly patterns include: Extracting time series of different types of data in the security management data as sub-time series, and obtaining a multidimensional time series based on the sub-time series; Calculating a low-dimensional compressed representation of the sub-time series using an LSTM encoder; Processing the low-dimensional compressed representation based on an LSTM decoder to obtain a reconstructed time series; Calculating a reconstruction error between the sub-time series and the corresponding reconstructed time series; All reconstruction error series slices are concatenated to obtain a multi-dimensional vector, and the reconstruction error of the multi-dimensional vector is calculated; sorting the reconstruction errors of the multidimensional vectors to obtain anomaly scores of the security management data; Based on the anomaly score, the data anomaly pattern is obtained.

6. The method according to claim 3, characterized in that: Methods for establishing a security risk assessment model using artificial bee colony algorithms include: Randomly generate an initial nectar source and initialize model parameters; wherein the model parameters include the maximum number of nectar sources to be mined, the number of mining times, and the maximum number of outer cycles; Calculating a fitness function based on the model parameters and the initial nectar source; Randomly generate a new nectar source, and calculate the fitness function value of the new nectar source based on the fitness function; When the fitness function value of the new nectar source satisfies the preset fitness function value, the nectar source is updated; Based on the updated honey source, the model parameters are updated, and when the model parameters meet the preset threshold, the optimal model parameters are obtained; Based on the optimal model parameters, the security risk assessment model is constructed.

7. A backfill mining safety management system based on intelligent monitoring, the system is used to implement the method according to any one of claims 1 to 4, characterized in that: include: Data collection module, used to collect and pre-process the safety management data of the filling mining area; The safety management data includes environmental data, operating condition data and filling material delivery data; A feature extraction module, used to perform time series analysis and anomaly detection on the pre-processed security management data to obtain a data feature vector; A risk assessment module is used to establish a safety risk assessment model using an artificial bee colony algorithm, and to assess the safety status of the backfill mining area based on the data feature vector and historical accident data to obtain a safety level assessment result; The emergency dispatch module is used to establish emergency plans and operation adjustment plans based on the safety level assessment results, and complete the backfill mining safety management based on intelligent monitoring.

8. The system according to claim 7, characterized in that The feature extraction module comprises: A time series analysis unit, configured to use an autoregressive integrated moving average model to perform time series modeling on the security management data to obtain data trends and data periodicity patterns; and to obtain time series characteristics of the security management data according to the data trends and the data periodicity patterns; an anomaly detection unit, configured to use a long short-term memory network to predict the multidimensional time series of the security management data, obtain a data anomaly pattern, and use the data anomaly pattern as an anomaly pattern feature; The feature vector acquisition unit is used to extract the statistical features of the security management data, and obtain the data feature vector in combination with the time series features and the abnormal pattern features.

Images