An intelligent decision-making method and system for mine risks

Abstract

The present application provides an intelligent decision-making method and system for mine risks, relating to the technical field of mine risk assessment. The method includes: collecting real-time data, historical data, types of historical data, and historical risk strategies of a mine; constructing a risk decision-making model, training the risk decision-making model by using the historical data and historical risk strategies to obtain a trained risk decision-making model; and inputting the real-time data into the trained risk decision-making model to obtain a real-time risk strategy. The present application discriminates abnormal real-time data by constructing a risk decision-making model, updates the conditional probability table of the Bayesian network sub-model based on the conditional probability of the abnormal real-time data, so as to optimize the Bayesian network model, improve the adaptability of the Bayesian network model structure to the dynamic change environment of mine risks, and further improve the accuracy of mine risk assessment results.

Description

Technical Field

[0001] The present application relates to the technical field of mine risk assessment, and in particular, to an intelligent decision-making method and system for mine risks. Background Art

[0002] Mining is an important pillar industry of the national economy. However, its operating environment is complex and harsh, facing multiple risks such as geological disasters (such as collapses, gas explosions), equipment failures, human operation errors, and tailings pond dam failures. Traditional mine risk management relies on manual experience judgment and static data records, with problems such as lagging risk identification, strong subjectivity in assessment, and extensive control measures. At the same time, as the mining depth increases, mines face extreme geological conditions such as high ground stress and high gas pressure. At the same time, the dynamic operation scenario exacerbates the randomness, coupling, and complexity of risk factors. In addition, derivative risks such as tailings pond dam failures and environmental pollution further increase the difficulty of mine risk management. In this context, there is an urgent need to accurately predict and real-time dynamically control mine risks.

[0003] In related technologies, a risk evolution topology structure is constructed using historical data to train a Bayesian network model, and the trained model is used to evaluate mine risks. However, the risk environment of mines is dynamically changing. With the progress of mine operations, equipment aging, etc., the relationships between various risk factors may also change, resulting in changes in the mutual influence relationships between various mine risks. When the mine risk environment changes, the original risk evolution topology structure may not be able to accurately describe the relationships between current risk factors, making the Bayesian network model unable to adapt to this dynamic change in a timely manner, resulting in the risk assessment results not matching the actual situation and affecting the accuracy of mine risk assessment results. Summary of the Invention

[0004] In order to improve the accuracy of mine risk assessment results, the present application provides an intelligent decision-making method and system for mine risks.

[0005] In a first aspect, the present application provides an intelligent decision-making method for mine risks, adopting the following technical solution:

[0006] An intelligent decision-making method for mine risks, the method includes:

[0007] Data collection: Collect real-time data, historical data, types of historical data, and historical risk strategies of the mine;

[0008] Model construction: Construct a risk decision-making model, where the risk decision-making model includes an identification sub-model, a Bayesian network sub-model, and a decision-making sub-model;

[0009] Model training: Use historical data and historical risk strategies to train a risk decision-making model, obtaining a trained recognition sub-model, a trained Bayesian network sub-model, and a trained decision-making sub-model;

[0010] First recognition: Input real-time data into the trained recognition sub-model to obtain abnormal real-time data, denoted as first data, and integrate the types of data with abnormalities in the first data into a first combination;

[0011] First calculation: Based on historical data, the first data, and the first combination, calculate the target conditional probability according to the conditional probability calculation formula;

[0012] First discrimination: Determine whether the target conditional probability belongs to the conditional probability table of the trained Bayesian network sub-model:

[0013] If so, perform the steps of the first evaluation;

[0014] If not, perform the steps of the first optimization;

[0015] First optimization: Add the target conditional probability to the conditional probability table of the trained Bayesian network sub-model to obtain a new Bayesian network sub-model, and update the new Bayesian network sub-model as the trained Bayesian network sub-model;

[0016] First evaluation: Input the first data into the trained Bayesian network sub-model to obtain second data;

[0017] Strategy acquisition: Input the second data into the trained decision-making sub-model to obtain a real-time risk strategy.

[0018] By adopting the above technical solution, based on the constructed risk decision-making model, use the recognition sub-model to judge whether there is abnormal real-time data. Based on the abnormal real-time data, obtain an abnormal data combination. Based on the abnormal data combination, calculate the conditional probability of the abnormal data combination. Determine whether the architecture of the Bayesian network model has changed by discriminating whether the conditional probability of the abnormal data combination belongs to the conditional probability table of the trained Bayesian network sub-model. Update the conditional probability table of the Bayesian network sub-model based on the discrimination result to optimize the architecture of the Bayesian network model, enabling it to adapt to the new risk environment and improving the accuracy of risk assessment. This application uses the conditional probability of the abnormal data combination to judge whether the architecture of the Bayesian network model needs to be updated, and then dynamically optimizes the Bayesian network structure, which helps to improve the adaptability of the Bayesian network model structure to the dynamic change environment of mine risks, improves the real-time performance, effectiveness, and accuracy of the Bayesian network model, further improves the adaptability and precision of mine risk assessment, and enhances the immediacy of mine safety management.

[0019] Optionally, after the step of performing the first recognition and before the step of performing the first calculation, the method further includes:

[0020] First clustering: Using a clustering algorithm to perform clustering analysis on historical data to obtain several clustering clusters of historical data;

[0021] Second calculation: Calculating the Euclidean distance between the first data and the clustering center of each clustering cluster, denoted as the first distance;

[0022] First judgment: Judging whether the first distances are all less than a preset distance threshold;

[0023] If so, perform the step of first evaluation;

[0024] If not, perform the step of first calculation.

[0025] By adopting the above technical solution, using a clustering algorithm to perform clustering analysis on historical data, and further judging the similarity between real-time data and historical data through the Euclidean distance, it avoids triggering model updates due to short-term fluctuations or noises, and improves the accuracy of abnormal data prediction. In addition, by comparing the Euclidean distance between the real-time data and the clustering center of each historical clustering cluster through the clustering algorithm to determine whether to trigger model updates, it helps to reduce the possibility of data anomaly misjudgment, helps to improve the adaptability of the dynamic change of the Bayesian network model, improves the prediction accuracy of abnormal data, and at the same time, helps to optimize computing resources, and improves the stability and adaptability of mine risk assessment.

[0026] Optionally, after the step of performing the first judgment and before the step of first calculation, the method further includes:

[0027] First processing: Extracting the data in the historical data that is of the same type as in the first combination, denoted as the first extracted data, and extracting the data in the first data that is of the same type as in the first combination, denoted as the second extracted data;

[0028] Third calculation: Using the local outlier factor algorithm to calculate the local reachability density of the first extracted data and the second extracted data, and calculating the local outlier factor of the second extracted data based on the calculated local reachability density;

[0029] Second judgment: Judging whether the local outlier factors of the second extracted data are all not lower than a preset local outlier factor threshold:

[0030] If so, no processing is performed;

[0031] If not, denoting the type of the second extracted data with a local outlier factor higher than the preset local outlier factor threshold as the second combination, and updating the second combination to the first combination.

[0032] By adopting the above technical solution, the local outlier factor of historical data and real-time data is calculated through the local outlier factor algorithm to judge the local anomaly degree of the real-time data, and the type of abnormal data is updated based on the discrimination result, which helps to accurately identify the dynamic change factors in the mine risk environment, reduces the possibility of misjudgment caused by single or small amounts of abnormal data, helps to improve the pertinence and accuracy of the Bayesian network structure adjustment, and improves the accuracy of the risk assessment result. In addition, the LOF algorithm is used to pre-screen the types of abnormal data, and only the categories of data with a high degree of outlier are updated, reducing the possibility of updating the Bayesian network model for all abnormal data, reducing unnecessary computational overhead, optimizing the allocation of computing resources, improving the operating efficiency of the model, and improving the accuracy and efficiency of mine risk assessment.

[0033] Optionally, based on the trained Bayesian network sub-model, the risk category of historical data at the same timestamp is obtained. After performing the first discrimination step and before performing the first optimization step, it further includes:

[0034] First screening: Sort the first distances in descending order, and denote the smallest first distance as the minimum distance;

[0035] Second screening: Denote the clustering cluster corresponding to the minimum distance as the first clustering cluster, and denote the risk category of the historical data corresponding to the first clustering cluster as the first label;

[0036] First annotation: Annotate the first label as the risk category of the first data to obtain the annotated first data, denoted as the new first data.

[0037] By adopting the above technical solution, based on cluster analysis and Euclidean distance sorting, the risk category most similar to the first data is matched from the historical data as the risk category of the first data, improving the rationality of the risk category of abnormal data. In addition, by dynamically annotating the risk category of abnormal data, it helps to dynamically adjust the risk category mapping based on the annotated risk category of the first data when the risk category evolves due to changes in the mine operation environment, so as to optimize the structure of the Bayesian network model, helps to improve the ability of the Bayesian network model to adapt to changes in the risk environment, enables the Bayesian network model to reflect the latest risk factor relationship, and improves the accuracy of the risk assessment result.

[0038] Optionally, the risk decision model further includes a semi-Markov sub-model. After performing the first annotation step and before performing the first optimization step, it further includes:

[0039] First extension: Define the state space of the semi-Markov sub-model based on the first data, historical data, and the risk category of the historical data;

[0040] Fourth calculation: Use the maximum likelihood estimation method to calculate the residence time of each state in the state space, and record the calculated residence time as the first time;

[0041] Matrix construction: Based on the first time, the first data, and historical data, use the maximum likelihood estimation method to calculate the transition probability of each state in the state space, and construct a transition probability matrix according to the calculated transition probability.

[0042] By adopting the above technical solution, a semi-Markov submodel is introduced on the basis of the Bayesian network model. Based on the semi-Markov submodel, the residence time distribution of different risk states is characterized, which improves the description accuracy of the evolution of complex risk states. At the same time, by constructing a state transition probability matrix, the evolution law between different risk states is established, which helps the Bayesian network model to have a time-series adaptive ability, optimizes the time-series correlation between different types of data, helps the Bayesian network model to be adjusted according to the dynamic changes of the mine operation environment, and improves the risk prediction accuracy of the Bayesian network model.

[0043] Optionally, based on the trained Bayesian network submodel, obtain the edges between the nodes in the Bayesian network submodel. After performing the matrix construction step and before performing the first optimization step, it further includes:

[0044] Third judgment: Based on the risk categories of the first combination and the first data, judge whether there is an edge between the nodes corresponding to the first combination in the Bayesian network submodel:

[0045] If so, do nothing;

[0046] If not, based on the first combination, add an edge between the nodes corresponding to the first combination in the Bayesian network submodel;

[0047] Probability update: Based on the first data, historical data, and the transition probability matrix, use the maximum likelihood estimation method to calculate the conditional probability between each node in the Bayesian network submodel, update the conditional probability table according to the calculation results, obtain the updated Bayesian network submodel, and use the updated Bayesian network submodel as the new trained Bayesian network submodel;

[0048] In the first evaluation step, input the first data into the new trained Bayesian network submodel.

[0049] By adopting the above technical solution, dynamically increasing the edges between nodes based on the first combination helps to update the topological structure of the Bayesian network model in real time, ensuring that the network structure of the Bayesian network model can reflect the correlation relationship between the types of the latest data. In addition, updating the conditional probability table based on the state transition probability matrix of the semi-Markov submodel improves the accuracy of probability inference of the Bayesian network model, enabling the Bayesian network model to not only adapt to newly emerged state data, but also predict the evolution direction of risks based on the evolution trend of historical states combined with time factors, improving the accuracy of risk assessment of the Bayesian network model.

[0050] Optionally, after the step of obtaining the policy, the following steps are further included:

[0051] Second processing: Based on the second data, select the risk policy corresponding to the risk category of the second data from the preset policy library, denoted as the first risk policy;

[0052] Third processing: Based on the second data, use the Monte Carlo simulation method to simulate the first risk policy and the real-time risk policy to obtain the risk index sequences of each risk policy;

[0053] Fourth processing: Define the reference risk index sequence of the risk policy based on the ideal state;

[0054] Fifth calculation: Use the grey relational analysis method to calculate the correlation degree between the risk index sequence of each risk policy and the reference risk index sequence, and denote the calculated correlation degree as the index correlation degree;

[0055] Fifth processing: Sort the obtained index correlation degrees in ascending order, and select the risk policy with the highest index correlation degree as the new real-time risk policy.

[0056] By adopting the above technical solution, using the Monte Carlo simulation to simulate the first risk policy and the real-time risk policy to obtain the risk index sequence, constructing the reference risk index based on the ideal state, using the grey relational analysis method to analyze the correlation degree between the risk index sequence and the reference risk index sequence, quantifying the advantages and disadvantages between different risk policies, and selecting the policy that best conforms to the current mine risk state, which helps to dynamically optimize the risk policy, making the mine risk policy always match the latest risk environment, improving the scientificity, accuracy and adaptability of risk policy selection, and enhancing the dynamic response ability of mine safety management.

[0057] Optionally, the second data includes the risk level; after the step of obtaining the policy, the following steps are further included:

[0058] First setting: Set the data of the mine based on the real-time risk policy as the test data;

[0059] Second evaluation: Input the test data into the trained Bayesian network sub-model to obtain the third data;

[0060] Fourth judgment: Judge whether the risk level of the third data is less than the risk level of the second data:

[0061] If so, execute the real-time risk strategy;

[0062] If so, execute the steps of the first setting;

[0063] First setting: Determine the optimization objectives, which include maximizing safety, minimizing resources, and minimizing costs;

[0064] Function construction: Based on the optimization objectives, construct the objective function;

[0065] First definition: Use the optimization algorithm to optimize the model parameters of the trained decision sub-model to obtain the first model parameters, and use the first model parameters as the new model parameters of the trained decision sub-model;

[0066] Second definition: Select the risk strategy corresponding to the risk category of the historical data from the preset strategy library as the new historical risk strategy;

[0067] Second optimization: Based on the historical data, the new historical risk strategy, and the new model parameters, train the trained decision sub-model to obtain a new trained decision sub-model, update the new trained decision sub-model as the trained decision sub-model, and then execute the step of policy acquisition.

[0068] By adopting the above technical solutions, the selected risk strategy is tested. By introducing the risk level comparison, the effectiveness of the risk strategy is ensured. At the same time, based on the optimization algorithm and optimization objectives, the parameters of the decision sub-model are optimized, enhancing the real-time performance and accuracy of the decision sub-model. In addition, adopting the cyclic optimization mechanism helps to improve the adaptability of the decision sub-model to the dynamic changing environment of the mine, improves the accuracy of the risk assessment of the risk decision model, and enhances the scientificity and reliability of the risk strategy of the risk decision model.

[0069] In a second aspect, the present application provides a mine risk intelligent decision-making system, adopting the following technical solutions:

[0070] A mine risk intelligent decision-making system, the system includes:

[0071] A processor and a memory,

[0072] The memory stores program codes;

[0073] When the processor calls the program codes in the memory, it executes the steps of the method described in the first aspect above.

[0074] By adopting the above technical solution, based on the constructed risk decision-making model, it is determined whether there is abnormal real-time data. Based on the abnormal real-time data, an abnormal data combination is obtained. Based on the abnormal data combination, the conditional probability of the abnormal data combination is calculated to determine whether the architecture of the Bayesian network model needs to be updated, thereby dynamically optimizing the structure of the Bayesian network model, which helps to improve the adaptability of the Bayesian network model structure to the dynamically changing environment of mine risks, improves the real-time performance, effectiveness and accuracy of the Bayesian network model, further improves the adaptability and accuracy of mine risk assessment, and enhances the immediacy of mine safety management.

[0075] In summary, the present application includes at least one of the following beneficial technical effects:

[0076] 1. Based on the constructed risk decision-making model, the recognition sub-model is used to determine whether there is abnormal real-time data. Based on the abnormal real-time data, an abnormal data combination is obtained. Based on the abnormal data combination, the conditional probability of the abnormal data combination is calculated. By determining whether the conditional probability of the abnormal data combination belongs to the conditional probability table of the trained Bayesian network sub-model, it is determined whether there is a change in the architecture of the Bayesian network model. Based on the determination result, the conditional probability table of the Bayesian network sub-model is updated to realize the optimization of the architecture of the Bayesian network model, so that it can adapt to the new risk environment and improve the accuracy of risk assessment. The present application uses the conditional probability of the abnormal data combination to determine whether the architecture of the Bayesian network model needs to be updated, thereby dynamically optimizing the Bayesian network structure, which helps to improve the adaptability of the Bayesian network model structure to the dynamically changing environment of mine risks, improves the real-time performance, effectiveness and accuracy of the Bayesian network model, further improves the adaptability and accuracy of mine risk assessment, and enhances the immediacy of mine safety management.

[0077] 2. The clustering algorithm is used to perform clustering analysis on historical data. The Euclidean distance is further used to judge the similarity between the real-time data and the historical data, avoiding triggering model updates due to short-term fluctuations or noises, and improving the accuracy of abnormal data prediction. In addition, by comparing the Euclidean distance between the real-time data and the clustering center of each historical clustering cluster through the clustering algorithm to determine whether to trigger model updates, it helps to reduce the possibility of occurrence of data anomaly misjudgment, helps to improve the adaptability of the dynamic change of the Bayesian network model, improves the prediction accuracy of abnormal data, and at the same time, helps to optimize computing resources, and improves the stability and adaptability of mine risk assessment.

[0078] 3. Introduce a semi-Markov submodel based on the Bayesian network model. Characterize the residence time distribution of different risk states based on the semi-Markov submodel, improve the description accuracy of the evolution of complex risk states. At the same time, by constructing a state transition probability matrix, establish the evolution law between different risk states, help the Bayesian network model have a time-series adaptive ability, optimize the time-series correlation between data types, help the Bayesian network model adjust according to the dynamic changes of the mine operation environment, and improve the risk prediction accuracy of the Bayesian network model.

[0079] 4. Dynamically increase the edges between nodes based on the first combination, which helps to update the topological structure of the Bayesian network model in real time and ensure that the network structure of the Bayesian network model can reflect the correlation relationship between the latest data types. In addition, update the conditional probability table based on the state transition probability matrix of the semi-Markov submodel, improve the accuracy of probability inference of the Bayesian network model, enable the Bayesian network model to not only adapt to newly emerging state data, but also predict the evolution direction of risks based on the evolution trend of historical states combined with time factors, and improve the accuracy of risk assessment of the Bayesian network model. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 is the flowchart of Embodiment 1 of the present application;

[0081] Figure 2 is the flowchart of S41 data discrimination in Embodiment 1 of the present application;

[0082] Figure 3 is the flowchart of S61 model optimization in Embodiment 2 of the present application;

[0083] Figure 4 is the flowchart of S91 strategy optimization in Embodiment 3 of the present application;

[0084] Figure 5 is the flowchart of S10 strategy verification in Embodiment 4 of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0085] The following will Figures 1 to 5 further describe the present application in detail.

[0086] Embodiment 1: This embodiment discloses an intelligent decision-making method for mine risks, as Figure 1As shown in the figure, the method includes: collecting real-time data, historical data, types of historical data, and historical risk strategies of a mine; constructing a risk decision-making model, which includes an identification sub-model, a Bayesian network sub-model, and a decision-making sub-model; training the risk decision-making model with historical data and historical risk strategies to obtain a trained identification sub-model, a trained Bayesian network sub-model, and a trained decision-making sub-model, inputting the real-time data into the trained identification sub-model to obtain abnormal real-time data, and integrating the types of data that are abnormal in the abnormal real-time data into a first combination; calculating the target conditional probability of the first combination, and determining whether the target conditional probability belongs to the conditional probability table of the trained Bayesian network sub-model. If so, update the Bayesian network sub-model; if not, do not update the Bayesian network sub-model. Then, input the abnormal real-time data into the Bayesian network sub-model to obtain second data; input the second data into the trained decision-making sub-model to obtain a real-time risk strategy. This embodiment includes the following steps:

[0087] S1 Data collection: Collect real-time data, historical data, types of historical data, risk category labels of historical data, and historical risk strategies of a mine.

[0088] The real-time data includes equipment health data, mine environment data, employee operation data, and production management data. The equipment health data includes data such as equipment vibration data, equipment temperature data, equipment pressure data, equipment working status data, and equipment current and voltage data; the mine environment data includes temperature data, humidity data, wind speed data, wind direction data, precipitation data, gas concentration data, carbon dioxide concentration data, soil humidity data, gas temperature data in the mine, mine air flow data, blasting vibration data, mine pressure data, mine water level data, mine geological data, and smoke concentration data, etc.; the employee operation data includes employee daily operation record data and employee location data; the production management data includes supplier information data, logistics transportation data, procurement data, inventory data, production progress data, ore extraction volume data, ore extraction quality data, mining machinery load data, and production efficiency data. The real-time risk data can be obtained through channels such as mine site monitoring systems, various sensors, and equipment health monitoring platforms.

[0089] Historical data includes historical equipment health data, historical mine environment data, historical employee operation data, and historical production management data. Historical equipment health data includes data such as historical equipment vibration data, historical equipment temperature data, historical equipment pressure data, historical equipment working status data, and historical equipment current and historical voltage data; historical mine environment data includes historical temperature data, historical humidity data, historical wind speed data, historical wind direction data, historical precipitation data, historical gas concentration data, historical carbon dioxide concentration data, historical soil humidity data, historical gas temperature data in the mine, historical mine air flow data, historical blasting vibration data, historical pressure data in the mine, historical mine water level data, historical mine geology data, and historical smoke concentration data, etc.; historical employee operation data includes historical employee daily operation record data and historical employee location data; historical production management data includes historical supplier information data, historical logistics transportation data, historical procurement data, historical inventory data, historical production progress data, historical ore extraction volume data, historical ore extraction quality data, historical mining machinery load data, and historical production efficiency data.

[0090] The types of historical data include equipment vibration, equipment temperature, equipment pressure, equipment working status, equipment current and voltage, temperature, humidity, wind speed, wind direction, precipitation, gas concentration, carbon dioxide concentration, soil humidity, gas temperature in the mine, mine air flow, blasting vibration, pressure in the mine, mine water level, mine geology, smoke concentration, employee daily operation records, employee location, supplier information, logistics transportation, procurement, inventory, production progress, ore extraction volume, ore extraction quality, mining machinery load, and production efficiency, etc.

[0091] In this embodiment, the real-time data is a set of equipment health data, mine environment data, employee operation data, and production management data at the same timestamp, and the historical data is actually a historical data set. The historical data contains historical data at several different times, and each historical data is a set of historical equipment health data, historical mine environment data, historical employee operation data, and historical production management data at the same timestamp. For easy understanding, some historical data is listed as an example in Table 1.

[0092] Table 1 Historical Data Table of the Mine

[0093] Number Time Air Temperature (°C) Humidity (%) Wind Speed (m / s) Precipitation (mm / h) Gas Concentration (%) Carbon Dioxide Concentration (ppm) 1 2024.12.1 24 75 3.2 0 1.2 0.05 2 2024.12.4 20 68 2.5 5 1.0 0.04 3 2024.12.8 22 65 4.0 1 1.3 0.06

[0094] The risk category labels of historical data include equipment risk - mechanical and electrical accidents, equipment risk - roof accidents, equipment risk - gas accidents, equipment risk - transportation accidents, equipment risk - blasting accidents, environmental risk - water damage accidents, environmental risk - fire accidents, environmental risk - air pollution, environmental risk - dust and geological disasters, personnel risk - object strikes, personnel risk - mechanical and electrical injuries, personnel risk - vehicle injuries, personnel risk - falls from height, personnel risk - electric shock injuries, personnel risk - lifting injuries, personnel risk - scalding injuries, personnel risk - poisoning and asphyxiation, production management risk - production scheduling errors, production management risk - insufficient safety management, production management risk - quality management problems, and production management risk - non - compliance problems.

[0095] The historical risk strategy is the risk - solving strategy implemented when historical data is in mine risk.

[0096] S2 Model construction: Construct a risk decision - making model, which includes an identification sub - model, a semi - Markov sub - model, a Bayesian network sub - model, and a decision - making sub - model.

[0097] The identification sub - model is an isolation forest model, and the decision - making sub - model uses a long - short - term memory network model as the basic model.

[0098] S3 Model training: Use historical data and historical risk strategies to train the risk decision - making model, and obtain the trained identification sub - model, the trained semi - Markov sub - model, the trained Bayesian network sub - model, and the trained decision - making sub - model.

[0099] Use historical data to train the identification sub - model to obtain the trained identification sub - model. Use historical data and the risk category labels of historical data to train the semi - Markov sub - model and the Bayesian network sub - model respectively to obtain the trained semi - Markov sub - model and the trained Bayesian network sub - model. Use historical data and the corresponding historical risk strategies to train the decision - making sub - model to obtain the trained decision - making sub - model.

[0100] S4 First identification: Input real - time data into the trained identification sub - model to obtain abnormal real - time data, denoted as the first data, and integrate the types of data with abnormalities in the first data into the first combination. For example, if the type of real - time data is (temperature, humidity, wind speed, wind direction, precipitation, gas concentration, carbon dioxide concentration), and the real - time data is (15℃, 65%, 2.3m / s, 45 o , 0.5mm, 0.3%, 400ppm), input the real - time data into the trained identification sub - model, and output real - time data (15℃, 65%, 2.3m / s, 45 o, 0.5mm, 0.3%, 400ppm), the real-time data marked with an abnormal label is recorded as the first data. Then, the types of abnormal data in the first data are integrated into the first combination, that is, (precipitation, gas concentration, carbon dioxide concentration) is recorded as the first combination. For the sake of simplicity in narration, only some types of real-time data are listed here. In this embodiment, the numerical setting for the wind direction is that the north is 0 degrees and increases in the clockwise direction.

[0101] In this embodiment, when using historical data to train the recognition sub-model, each historical data is used as a feature to train the isolation forest, and the first isolation forest model is obtained. Then, the data of each type in the historical data is extracted and used to train the isolation forest respectively, and several second isolation models are obtained. All the trained second isolation forest models are used as a recognition sub-model.

[0102] Before the first recognition in S4, it also includes: inputting the real-time data as a data feature into the first isolation forest model: if the abnormal label of the output real-time data is "-1", then this real-time data is recorded as the new real-time data, and S4 first recognition is executed. If the abnormal label of the output real-time data is "1", then this real-time data is recorded as the new first data, and S8 first evaluation is executed.

[0103] S41 Data discrimination: includes S411 first clustering, S412 second calculation, S413 first judgment, S414 first processing, S415 third calculation, S416 second judgment, as Figure 2 shown.

[0104] S411 First clustering: Use a clustering algorithm (such as K-means, hierarchical clustering, DBSCAN and other clustering algorithms) to perform clustering analysis on the historical data to obtain several clustering clusters of historical data. In this embodiment, the DBSCAN clustering algorithm is used to perform clustering analysis on the historical data to obtain several clustering clusters of historical data.

[0105] S412 Second calculation: Use the Euclidean distance method to calculate the Euclidean distance between the first data and the clustering center of each clustering cluster, which is recorded as the first distance. Among them, the first distance is a set of first distances, including the Euclidean distance between the first data and the clustering center of each clustering cluster.

[0106] The Euclidean distance formula is:

[0107] .

[0108] Among them, represents the Euclidean distance between the first data and the clustering center of the th clustering cluster, represents the th data in the first data, Indicates the th data of the cluster center of the th cluster.

[0109] The DBSCAN clustering algorithm is used to perform clustering analysis on historical data, obtaining cluster A, cluster B, and cluster C. Calculate the average value of all historical data within each cluster as the cluster center of each cluster. For example, cluster A contains 5 historical data. If the categories of these 5 historical data are (equipment temperature and equipment pressure), which are data 1 (20, 30), data 2 (22, 31), data 3 (21, 29), data 4 (19, 28), and data 5 (23, 32) respectively, then the cluster center of cluster A is ( , ). For simplicity of narration, only some types of historical data are listed here.

[0110] The Euclidean distance method is used to calculate the Euclidean distance between the first data and the cluster centers of each cluster, obtaining the corresponding first distance A, first distance B, and first distance C. For example, the first data is (25, 35). From the above example, the cluster center of cluster A is (21, 30), then , and the first distance A is obtained as .

[0111] S413 First judgment: Judge whether the first distances are all less than the preset distance threshold.

[0112] If so, execute S8 First evaluation.

[0113] If not, execute S5 First calculation.

[0114] S414 First processing: Extract the data in the historical data that is consistent with the types in the first combination, denoted as the first extracted data. Extract the data in the first data that is consistent with the types in the first combination, denoted as the second extracted data. For example, if the first combination is (rainfall, gas concentration, carbon dioxide concentration), then extract the data on rainfall, gas concentration, and carbon dioxide concentration in the historical data, denoted as the first extracted data of the historical data.

[0115] S415 Third calculation: Use the local outlier factor algorithm to calculate the local reachability density of the first extracted data and the second extracted data, and calculate the local outlier factor of the second extracted data based on the calculated local reachability density.

[0116] Define the data range of each type of data in the second extracted data. Based on the defined data range, screen the historical data in the first extracted data, and screen out the historical data in the first extracted data that satisfies that the data of each type is within the data range of each type of defined data. Denote the screened first extracted data as the new first extracted data.

[0117] The Euclidean distance method is used to calculate the distance between each first extraction data and the second extraction data, and the reachable distance between the current first extraction data and the second extraction data is calculated according to the reachable distance formula. The local reachable density of the second extraction data is calculated according to the local reachable density formula.

[0118] The reachable distance formula is as follows:

[0119] 。

[0120] Among them, represents the reachable distance between the th first extraction data and the second extraction data, represents the second extraction data, represents the th first extraction data, represents the maximum Euclidean distance from the second extraction data to the first extraction data, represents the Euclidean distance from the second extraction data to the th first extraction data.

[0121] The local reachable density calculation formula is:

[0122] 。

[0123] Among them, represents the number of first extraction data, represents the first extraction data within the data range of each type of data of the defined second extraction data, represents the sum of the reachable distances between the second extraction data and all first extraction data.

[0124] Based on the above steps, the local reachable density of all first extraction data is calculated, and the local outlier factor of the second extraction data is calculated according to the local outlier factor calculation formula.

[0125] The local outlier factor calculation formula is: 。

[0126] Among them, represents the local reachable density of the th first extraction data.

[0127] S416 Second judgment: Judge whether the local outlier factor of the second extraction data is not lower than the preset local outlier factor threshold.

[0128] If so, no processing is performed.

[0129] Otherwise, record the types of the second extraction data with a local outlier factor higher than the preset local outlier factor threshold as the second combination, and update the second combination to the first combination. For example, according to the above calculation, the local outlier factors of rainfall and gas concentration in the second extraction data are not lower than the preset local outlier factor threshold, and the local outlier factor of carbon dioxide concentration is not higher than the preset local outlier factor threshold. Then, update (rainfall, gas concentration) to the first combination.

[0130] S5 First calculation: Calculate the joint frequency and marginal probability of the first combination based on historical data and real-time data, and calculate the target conditional probability through the calculation formula of conditional probability.

[0131] In this embodiment, the calculation formula of conditional probability is:

[0132] 。

[0133] Among them, represents the joint probability of the first combination, represents the marginal probability of the first combination. represents the number of types in the first combination, represents the th category of data in the first combination. Example, the types of real-time data are (temperature, humidity, gas concentration, carbon dioxide concentration), and the real-time data is (15°C, 65%, 0.3%, 400 ppm). Among them, the anomaly label of the real-time data is (1, -1, -1, 1). So the first data is (15°C, 65%, 0.3%, 400 ppm) marked with the anomaly label (1, -1, -1, 1). So the first combination is (humidity, gas concentration). Based on historical data, count the probability of the temperature being 15°C, denoted as P1, the probability of the humidity being 65%, denoted as P2, the probability of the gas concentration being 0.3%, denoted as P3, and the probability of the carbon dioxide concentration being 400 ppm, denoted as P4. Then calculate the joint probability of the first combination , ,calculate the marginal probability of the first combination , ,then the target conditional probability of the first combination 。

[0134] S6 First discrimination: Based on the obtained target conditional probability, traverse the conditional probability table of the trained Bayesian network sub-model, and determine whether the target conditional probability belongs to the conditional probability table of the trained Bayesian network sub-model.

[0135] If so, execute S8 First evaluation.

[0136] If not, execute S7 First optimization.

[0137] S7 First Optimization: Add the target conditional probability to the conditional probability table of the trained Bayesian network sub-model to obtain a new conditional probability table. Based on the obtained new conditional probability table, obtain a new Bayesian network sub-model, and update the new Bayesian network sub-model as the trained Bayesian network sub-model.

[0138] S8 First Evaluation: Input the first data into the trained Bayesian network sub-model to obtain second data, where the second data includes a risk level.

[0139] S9 Policy Acquisition: Input the second data into the trained decision sub-model to obtain a real-time risk policy. After obtaining the real-time risk policy, display the obtained real-time risk policy to the administrator, and execute the obtained real-time risk policy after the administrator confirms it.

[0140] In this embodiment, based on the constructed risk decision model, use the identification sub-model to determine whether there is abnormal real-time data. Based on the abnormal real-time data, obtain an abnormal data combination. Calculate the conditional probability of the abnormal data combination based on the abnormal data combination. Determine whether the conditional probability of the abnormal data combination belongs to the conditional probability table of the trained Bayesian network sub-model to determine whether there is a change in the architecture of the Bayesian network model. Update the conditional probability table of the Bayesian network sub-model based on the determination result to optimize the architecture of the Bayesian network model, enabling it to adapt to the new risk environment and improving the accuracy of risk assessment. This application uses the conditional probability of the abnormal data combination to determine whether the architecture of the Bayesian network model needs to be updated, and then dynamically optimizes the Bayesian network structure, which helps to improve the adaptability of the Bayesian network model structure to the dynamic change environment of mine risks, improves the real-time performance, effectiveness, and accuracy of the Bayesian network model, further improves the adaptability and precision of mine risk assessment, and enhances the immediacy of mine safety management.

[0141] Embodiment 2: The difference from Embodiment 1 is that:

[0142] The risk decision model further includes a semi-Markov sub-model. After performing S6 First Discrimination and before performing S7 First Optimization, it further includes: S61 Model Optimization, and S61 Model Optimization includes S611 First Screening, S612 Second Screening, S613 First Annotation, S614 First Extension, S615 Fourth Calculation, S616 Matrix Construction, S617 Third Judgment, and S618 Probability Update, as Figure 3 shown.

[0143] S611 First Screening: Sort the first distances in descending order, and record the smallest first distance as the minimum distance.

[0144] S612 Second Screening: Take the cluster corresponding to the minimum distance as the cluster closest to the first data, denoted as the first cluster, and denote the risk category of the historical data corresponding to the first cluster as the first label.

[0145] S613 First Annotation: Take the first label as the risk category of the first data for annotation, obtaining the annotated first data, denoted as the new first data. Use the risk category of the cluster closest to the first data as the risk category of the first data.

[0146] S614 First Expansion: Define the state space of the semi-Markov submodel based on the first data, historical data, and the risk categories of the historical data.

[0147] S615 Fourth Calculation: Obtain the state sequence observed at each moment, and use the maximum likelihood estimation method to calculate the residence time of each state in the state space. Denote the calculated residence time as the first time. The residence time is the duration of each state.

[0148] S616 Matrix Construction: Based on the first time, the first data, and the historical data, use the maximum likelihood estimation method to calculate the transition probability between each state in the state space. For example, count the number of times state b transfers to state j, denoted as T, and count the total number of transfers of state b, denoted as K. Then the probability that state b transfers to state j is , and then optimize the transition probability through the maximum likelihood estimation method to obtain the transition probability that state b transfers to state j. Based on the calculated transition probabilities between all states, construct the transition probability matrix W, where W(b, j) represents the transition probability that state b transfers to state j.

[0149] S617 Third Judgment: Based on the categories within the first combination as the parent variable nodes of the Bayesian network submodel, and take the risk category of the first data as the child variable node of the Bayesian network submodel, and judge whether there is an edge between the nodes corresponding to the first combination in the Bayesian network submodel.

[0150] If so, no processing is required.

[0151] If not, then based on the first combination, in the Bayesian network submodel, add the corresponding edge between the parent variable nodes of the Bayesian network submodel.

[0152] For example, the first combination is (rainfall, gas concentration). Then based on the first combination, check whether there is an edge between the node of rainfall and the node of gas concentration in the Bayesian network submodel. If there is, no processing is required. If not, then in the Bayesian network submodel, add the corresponding edge between the node of rainfall and the node of gas concentration in the Bayesian network submodel.

[0153] S618 Probability Update: Based on the first data, historical data, and the transition probability matrix, the maximum likelihood estimation method is used to calculate the conditional probability between each pair of nodes in the Bayesian network sub-model. According to the calculation results, the conditional probability table is updated to obtain an updated Bayesian network sub-model, and the updated Bayesian network sub-model is used as the new trained Bayesian network sub-model.

[0154] Define a likelihood function based on each node in the Bayesian network sub-model, which represents the probability that the sub-variable node takes different states given the state of the parent variable node. For example, let the sub-variable node be E and the parent variable node be , denote the historical data and the first data as the node sample data set. There are H node samples in the node sample data set. Then the state of the parent variable node in the c-th sample is , and the state of the sub-variable node is , then the likelihood function . Among them, is the number of parent variable nodes associated with the sub-variable node. By finding the extreme value of the likelihood function, the maximum likelihood estimation value of the conditional probability between each pair of nodes is obtained under the given node sample data. Based on the obtained maximum likelihood estimation value of the conditional probability between each pair of nodes, it is filled into the conditional probability table to update the conditional probability table, obtaining an updated conditional probability table. Based on the updated conditional probability table, an updated Bayesian network sub-model is obtained, and the updated Bayesian network sub-model is used as the new trained Bayesian network sub-model.

[0155] In S8 First Evaluation, the first data is input into the new trained Bayesian network sub-model.

[0156] In this embodiment, a semi-Markov sub-model is introduced based on the Bayesian network model. The residence time distribution of different risk states is characterized based on the semi-Markov sub-model, which improves the description accuracy of the evolution of complex risk states. At the same time, by constructing a state transition probability matrix, the evolution law between different risk states is established, which helps the Bayesian network model to have a time-series adaptive ability, optimizes the time-series correlation between different types of data, and helps the Bayesian network model to be adjusted according to the dynamic changes of the mine operation environment, improving the accuracy of risk prediction of the Bayesian network model. Adding edges between nodes dynamically based on the first combination helps to update the topological structure of the Bayesian network model in real time, ensuring that the network structure of the Bayesian network model can reflect the correlation relationship between the latest types of data. In addition, updating the conditional probability table based on the state transition probability matrix of the semi-Markov sub-model improves the accuracy of probability inference of the Bayesian network model, enabling the Bayesian network model to not only adapt to newly emerging state data but also predict the evolution direction of risks based on the evolution trend of historical states combined with time factors, improving the accuracy of risk assessment of the Bayesian network model.

[0157] Example 3: The difference from Example 1 is that:

[0158] As Figure 4 shown, after obtaining the S9 strategy, it further includes: S91 strategy optimization, and the S91 strategy optimization includes S911 second processing, S912 third processing, S913 fourth processing, S914 fifth calculation, and S915 fifth processing.

[0159] S911 second processing: Based on the second data, select the risk strategy corresponding to the risk category of the second data from the preset strategy library, denoted as the first risk strategy. The preset strategy library is a pre-set library containing risk strategies that can solve all mine risks.

[0160] S912 third processing: Based on the second data, use the Monte Carlo simulation method to simulate the first risk strategy and the real-time risk strategy, and obtain the risk index sequence of each risk strategy.

[0161] Using the value of the second data as the numerical range, adopt the Monte Carlo simulation method to construct mine scenarios in different situations, simulate the first risk strategy and the real-time risk strategy respectively based on the mine scenarios in different situations, and obtain the risk index sequences of the first risk strategy and the real-time risk strategy based on the simulation results. The risk index sequence includes cost-benefit, resource consumption, and recovery time. Denote the risk index sequence of the first risk strategy as the first sequence, and denote the risk index sequence of the real-time risk strategy as the second sequence.

[0162] S913 fourth processing: Define the reference risk index sequence based on the risk strategy under theoretical or ideal conditions. The reference risk index sequence is the minimum cost-benefit, minimum resource consumption, and minimum recovery time. Under ideal conditions, the values of cost-benefit, resource consumption, and recovery time in the reference risk index sequence are basically 0.

[0163] S914 fifth calculation: Based on the reference risk index sequence, calculate the difference between the first sequence and each index in the reference risk index sequence, denoted as the first difference sequence, and calculate the difference between the second sequence and each index in the reference risk index sequence, denoted as the second difference sequence. Calculate the correlation coefficient between the first difference sequence and the reference risk index sequence to obtain the correlation coefficient of each index, denoted as the first coefficient, and then calculate the average value of each correlation coefficient in the first coefficient to obtain the grey correlation degree, denoted as the first index correlation degree. Calculate the correlation coefficient between the second difference sequence and the reference risk index sequence to obtain the correlation coefficient of each index, denoted as the second coefficient, and then calculate the average value of each correlation coefficient in the second coefficient to obtain the grey correlation degree, denoted as the second index correlation degree.

[0164] S915 Fifth Processing: Sort the obtained first index correlation degree and second index correlation degree in ascending order, and select the risk strategy with the highest index correlation degree as the new real-time risk strategy.

[0165] In this embodiment, Monte Carlo simulation is used to simulate the first risk strategy and the real-time risk strategy to obtain a risk index sequence, and a reference risk index is constructed based on the ideal state. The grey relational analysis method is used to analyze the correlation degree between the risk index sequence and the reference risk index sequence, quantify the advantages and disadvantages between different risk strategies, and select the strategy that best matches the current mine risk state, which helps to dynamically optimize the risk strategy, make the mine risk strategy always match the latest risk environment, improve the scientificity, accuracy and adaptability of the risk strategy selection, and enhance the dynamic response ability of mine safety management.

[0166] Embodiment 4: The difference from Embodiment 1 is that:

[0167] The second data includes the risk level. After executing the S9 strategy to obtain, it further includes: S10 strategy verification, and S10 strategy verification includes S101 First Setting, S102 Second Evaluation, S103 Fourth Judgment, S104 First Setting, S105 Function Construction, S106 First Definition, S107 Second Definition and S108 Second Optimization, as Figure 5 shown.

[0168] S101 First Setting: Set the data of the mine based on the real-time risk strategy as the test data. According to the obtained real-time risk strategy, set the data of the mine after executing the real-time risk strategy as the test data.

[0169] S102 Second Evaluation: Input the test data into the trained Bayesian network sub-model for evaluation to obtain the third data.

[0170] S103 Fourth Judgment: Judge whether the risk level of the third data is less than the risk level of the second data.

[0171] If so, execute the obtained real-time risk strategy.

[0172] If so, execute S104 First Setting.

[0173] S104 First Setting: Determine the optimization objective of the decision sub-model, and the optimization objective of the decision sub-model includes safety maximization, resource minimization and cost minimization.

[0174] S105 Function Construction: Based on the optimization objective, construct the objective function.

[0175] The objective function is: 。

[0176] Where, represents the objective function, represents the risk strategy, represents the safety maximization optimization objective, represents the resource minimization optimization objective, represents the cost minimization optimization objective, and and represent the weights of each optimization objective, .

[0177] S106 First definition: Use an optimization algorithm to optimize the model parameters of the trained decision sub-model to obtain the first model parameters, and use the first model parameters as the new model parameters of the trained decision sub-model.

[0178] In this embodiment, a particle swarm optimization algorithm, a genetic algorithm, and a simulated annealing algorithm can be used to optimize the model parameters of the decision sub-model. For example, layer by layer optimize the model parameters of the decision sub-model through the particle swarm optimization algorithm, the genetic algorithm, and the simulated annealing algorithm in sequence, record the optimized model parameters as the first model parameters, and use the first model parameters as the new model parameters of the trained decision sub-model.

[0179] S107 Second definition: Select the risk strategy corresponding to the risk category of the historical data from the preset strategy library as the new historical risk strategy for the historical data.

[0180] The preset strategy library is a pre-set library containing risk strategies that can solve all mine risks. Select the corresponding risk strategy from the preset strategy library for each historical data according to the risk category of each historical data as the new risk strategy of the historical data to obtain the new historical risk strategy.

[0181] S108 Second optimization: Based on the historical data, the new historical risk strategy, and the new model parameters, perform model training on the trained decision sub-model to obtain the newly trained decision sub-model, update the newly trained decision sub-model as the trained decision sub-model, and then execute the S9 policy acquisition. After executing the S9 policy acquisition, execute the S10 policy verification until the risk level of the third data is less than the risk level of the second data. If the risk level of the third data is still less than the risk level of the second data within the preset loop time, send the real-time data and all the output real-time risk strategies to the administrator, and the administrator formulates the real-time risk strategy for the real-time data.

[0182] In this embodiment, the selected risk strategy is tested. By introducing a comparison of risk levels, the effectiveness of the risk strategy is ensured. At the same time, based on the optimization algorithm and the optimization objective, the parameters of the decision sub-model are optimized, enhancing the real-time performance and accuracy of the decision sub-model. In addition, the use of a cyclic optimization mechanism helps to improve the adaptability of the decision sub-model to the dynamic changing environment of the mine, improves the accuracy of the risk assessment of the risk decision model, and enhances the scientificity and reliability of the risk strategy of the risk decision model.

[0183] Embodiment 5: This embodiment discloses an intelligent risk decision-making system for mines, and the system includes:

[0184] a processor and a memory,

[0185] wherein program code is stored in the memory;

[0186] when the processor calls the program code in the memory, it executes the steps of the method described in the above embodiment.

[0187] In this embodiment, based on the constructed risk decision model, it is determined whether there is abnormal real-time data. Based on the abnormal real-time data, an abnormal data combination is obtained. Based on the abnormal data combination, the conditional probability of the abnormal data combination is calculated to determine whether the architecture of the Bayesian network model needs to be updated, and then the structure of the Bayesian network model is dynamically optimized. This helps to improve the adaptability of the Bayesian network model structure to the dynamic changing environment of mine risks, improves the real-time performance, effectiveness and accuracy of the Bayesian network model, further improves the adaptability and accuracy of mine risk assessment, and enhances the immediacy of mine safety management.

[0188] The above are all preferred embodiments of this application. The protection scope of this application is not limited accordingly. Therefore, all equivalent changes made according to the structure, shape and principle of this application should be covered within the protection scope of this application.

Claims

1. An intelligent decision-making method for mine risks, characterized in that, Including: Data collection: Collect real-time data, historical data, types of historical data, and historical risk strategies of the mine; Model construction: Construct a risk decision-making model, which includes an identification sub-model, a Bayesian network sub-model, and a decision-making sub-model; the identification sub-model is an isolation forest model, and the decision-making sub-model is a long short-term memory network model; Model training: Use historical data and historical risk strategies to train the risk decision-making model to obtain a trained identification sub-model, a trained Bayesian network sub-model, and a trained decision-making sub-model; including: using historical data to train the identification sub-model to obtain a trained identification sub-model, using historical data and risk category labels of historical data to train a semi-Markov sub-model and a Bayesian network sub-model respectively to obtain a trained semi-Markov sub-model and a trained Bayesian network sub-model, using historical data and corresponding historical risk strategies to train the decision-making sub-model to obtain a trained decision-making sub-model; First identification: Input real-time data into the trained identification sub-model to obtain abnormal real-time data, denoted as the first data, and integrate the types of data with anomalies in the first data into the first combination; First calculation: Based on historical data, the first data, and the first combination, calculate the target conditional probability according to the conditional probability calculation formula; First discrimination: Determine whether the target conditional probability belongs to the conditional probability table of the trained Bayesian network sub-model: If so, execute the steps of the first evaluation; If not, execute the steps of the first optimization; First optimization: Add the target conditional probability to the conditional probability table of the trained Bayesian network sub-model to obtain a new Bayesian network sub-model, and update the new Bayesian network sub-model to the trained Bayesian network sub-model; First evaluation: Input the first data into the trained Bayesian network sub-model to obtain the second data; Strategy acquisition: Input the second data into the trained decision-making sub-model to obtain the real-time risk strategy.

2. The intelligent decision-making method for mine risks according to claim 1, wherein After executing the steps of the first identification and before executing the steps of the first calculation, it also includes: First clustering: Use a clustering algorithm to perform clustering analysis on historical data to obtain several clustering clusters of historical data; Second calculation: Calculate the Euclidean distance between the first data and the clustering center of each clustering cluster, denoted as the first distance; First judgment: Determine whether the first distance is less than the preset distance threshold; If so, execute the steps of the first evaluation; If not, execute the steps of the first calculation.

3. The intelligent decision-making method for mine risks according to claim 2, wherein After executing the steps of the first judgment and before the steps of the first calculation, it also includes: First processing: Extract the data in historical data that is of the same type as the type in the first combination, denoted as the first extracted data, and extract the data in the first data that is of the same type as the type in the first combination, denoted as the second extracted data; Third calculation: Use the local outlier factor algorithm to calculate the local reachability density of the first extracted data and the second extracted data, and calculate the local outlier factor of the second extracted data based on the calculated local reachability density; Second judgment: Determine whether the local outlier factors of the second extracted data are all not lower than the preset local outlier factor threshold: If so, do not perform any processing; Otherwise, record the type of the second extraction data that is higher than the preset local outlier factor threshold as the second combination, and update the second combination to the first combination.

4. The intelligent decision-making method for mine risks according to claim 2, wherein, Based on the trained Bayesian network sub-model, obtain the risk categories of historical data at the same timestamp. After performing the first discrimination step and before performing the first optimization step, it further includes: First screening: Sort the first distances in descending order, and record the smallest first distance as the minimum distance. Second screening: Record the clustering cluster corresponding to the minimum distance as the first clustering cluster, and record the risk category of the historical data corresponding to the first clustering cluster as the first label. First annotation: Annotate the risk category of the first label as the first data to obtain the annotated first data, denoted as the new first data.

5. The intelligent decision-making method for mine risks according to claim 4, characterized in that The risk decision model further includes a semi-Markov sub-model. After performing the first annotation step and before performing the first optimization step, it further includes: First expansion: Define the state space of the semi-Markov sub-model based on the first data, historical data, and the risk categories of historical data. Fourth calculation: Use the maximum likelihood estimation method to calculate the residence time of each state in the state space, and record the calculated residence time as the first time. Matrix construction: Based on the first time, the first data, and historical data, use the maximum likelihood estimation method to calculate the transition probability of each state in the state space, and construct a transition probability matrix according to the calculated transition probability.

6. The intelligent decision-making method for mine risks according to claim 5, wherein Based on the trained Bayesian network sub-model, obtain the edges between nodes in the Bayesian network sub-model. After performing the matrix construction step and before performing the first optimization step, it further includes: Third judgment: Based on the first combination and the risk category of the first data, judge whether there is an edge between the nodes corresponding to the first combination in the Bayesian network sub-model: If so, do not process. Otherwise, based on the first combination, add an edge between the nodes corresponding to the first combination in the Bayesian network sub-model. Probability update: Based on the first data, historical data, and the transition probability matrix, use the maximum likelihood estimation method to calculate the conditional probability between each node in the Bayesian network sub-model, update the conditional probability table according to the calculation result, obtain the updated Bayesian network sub-model, and use the updated Bayesian network sub-model as the new trained Bayesian network sub-model. In the first evaluation step, input the first data into the new trained Bayesian network sub-model.

7. The intelligent decision-making method for mine risks according to claim 4, characterized in that, After performing the policy acquisition step, it further includes: Second processing: Based on the second data, select the risk policy corresponding to the risk category of the second data from the preset policy library, denoted as the first risk policy. Third processing: Based on the second data, use the Monte Carlo simulation method to simulate the first risk policy and the real-time risk policy to obtain the risk index sequence of each risk policy. Fourth processing: Define the reference risk index sequence of the risk policy based on the ideal state. Fifth calculation: Use the grey relational analysis method to calculate the correlation degree between the risk index sequence of each risk policy obtained and the reference risk index sequence, and record the calculated correlation degree as the index correlation degree. Fifth processing: Sort the obtained index correlation degrees in ascending order, and select the risk strategy with the highest index correlation degree as the new real-time risk strategy.

8. The intelligent decision-making method for mine risks according to claim 4, characterized in that, The second data includes a risk level; After the step of obtaining the strategy, it further includes: First setting: Set the data of the mine based on the real-time risk strategy as test data; Second evaluation: Input the test data into the trained Bayesian network sub-model to obtain the third data; Fourth judgment: Judge whether the risk level of the third data is less than the risk level of the second data: If so, execute the real-time risk strategy; If so, execute the steps of the first setting; First setting: Determine the optimization objectives, which include maximizing safety, minimizing resources, and minimizing costs; Function construction: Based on the optimization objectives, construct the objective function; First definition: Use the optimization algorithm to optimize the model parameters of the trained decision sub-model to obtain the first model parameters, and use the first model parameters as the new model parameters of the trained decision sub-model; Second definition: Select the risk strategy corresponding to the risk category of the historical data from the preset strategy library as the new historical risk strategy; Second optimization: Train the trained decision sub-model based on the historical data, the new historical risk strategy, and the new model parameters to obtain the newly trained decision sub-model, update the newly trained decision sub-model as the trained decision sub-model, and then execute the step of obtaining the strategy.

9. An intelligent decision-making system for mine risks, characterized in that, It includes: A processor and a memory, The memory stores program codes; When the processor calls the program codes in the memory, it executes the steps of the method described in any one of claims 1-8.

Images