Mine geological environment intelligent monitoring and early warning method and system

Abstract

An intelligent monitoring and early warning method for a mine geological environment relates to the technical field of geological disaster early warning and comprises the following steps: step 1, dividing monitoring points and collecting historical geological data, historical environmental data and historical meteorological data of a mine, step 2, preprocessing and summarizing the data, and performing preprocessing and feature extraction; the system comprises an information collection module, an edge calculation module, a model training module and an alarm module, the monitoring point data is collected, the early warning model is constructed, the early warning model is trained and deployed, and the energy consumption control and the alarm information transmission are carried out according to the early warning result. The system comprises an information collection module, an edge calculation module, a model training module and an alarm module. The power consumption of the monitoring point is determined according to the result of the early warning model, and alarm information is transmitted to the outside, thereby realizing prediction of mine geological disasters, improving the accuracy of the prediction result, controlling the power consumption of the monitoring point, prolonging the service life of the monitoring point, and promoting the stability of the monitoring system.

Description

Technical Field

[0001] The present invention relates to the technical field of geological disaster early warning, and in particular to an intelligent monitoring and early warning method and system for a mine geological environment. Background Art

[0002] At present, the monitoring and early warning technology of the mine geological environment has achieved certain development and is widely used in the prevention of geological disasters such as landslides and mudslides. However, the existing technical solutions still have obvious deficiencies in data transmission, accuracy of early warning models and system energy consumption management. The existing monitoring systems usually have large data transmission delays and limited coverage. At the same time, the accuracy of traditional early warning models is not high, and it is difficult to adapt to changes in the geological environment in real time. In addition, the high energy consumption of sensors and data processing equipment also limits the long-term stable operation of the system. Therefore, the existing technology cannot meet the comprehensive needs of high efficiency and low energy consumption for intelligent monitoring and early warning of the mine geological environment.

[0003] According to a method and system for intelligent monitoring and early warning of mine geological environment disclosed in publication number CN202410001432.8, an intelligent monitoring model of mine geological environment is constructed, and the monitoring parameters of various target environmental factors are determined based on the intelligent monitoring model of mine geological environment; an intelligent environmental monitoring sensor network is deployed in the target monitoring area according to the monitoring parameters, and the target environmental factor data in the target monitoring area is collected by using the intelligent environmental monitoring sensor network; a mine geological disaster risk assessment model is constructed, and the current geological disaster risk in the target monitoring area is assessed by combining the mine geological disaster risk assessment model with the target environmental factor data.

[0004] According to the technical documents that have been made public, although the assessment of mining geological disaster risks has been achieved, the monitored data is not specific enough and has not been targeted. The traditional prediction methods used have questionable prediction accuracy and do not fully consider energy consumption needs. Summary of the invention

[0005] 1. Technical issues to be solved

[0006] In view of the shortcomings of the prior art, the present invention provides a method and system for intelligent monitoring and early warning of the mining geological environment. By collecting monitoring point data, monitoring and early warning are performed through edge computing nodes, the central processing unit is constructed and deployed, and energy consumption control is implemented according to the prediction results, so as to improve the stability and sustainability of the monitoring system.

[0007] (II) Technical solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0009] A method for intelligent monitoring and early warning of mine geological environment, comprising the following steps:

[0010] Step 1: Divide the monitoring points and collect the historical geological data, historical environmental data and historical meteorological data of the mine;

[0011] Step 2: Preprocess the data, extract features from the data, collect the types of geological disasters and organize them into labels to form a data set;

[0012] Step 3: Train the model through decision tree, judge the node by information gain, deploy the model to the node and make real-time prediction;

[0013] Step 4: Control the energy consumption of the monitoring point and transmit the warning information of the monitoring point according to the results of the early warning model.

[0014] Furthermore, monitoring points are divided in the mine to be monitored, and historical geological data, historical environmental data and historical meteorological data of the mine are collected at the monitoring points. The historical geological data are the vibration frequency and amplitude of the mine rock mass, the historical environmental data are the temperature and humidity in the monitoring point area, and the historical meteorological data are the wind speed and precipitation in the mine area. The collected data are timestamped to form data set D.

[0015] Furthermore, edge computing nodes are evenly arranged between monitoring points, and each edge computing node is connected to at least two monitoring points. The edge computing node collects and summarizes the data of the monitoring points connected to it, supplements the missing data with the next bit of data, and then performs noise filtering. The formula is:

[0016]

[0017] X(t j-a ) represents the time t j-a The original data at the time point, a represents the index variable at the time point, a∈N, 0≤a≤N, X filtered (t j ) represents the filtered data;

[0018] N is the size of the filter window, where the filter window size for amplitude data is 30, the filter window size for vibration frequency, wind speed and precipitation data is 10, and the filter window size for temperature and humidity data is 5;

[0019] Then the data of the same type are normalized to form a data set DD.

[0020] Further, features are extracted from the data set DD, each data type is marked as a single feature, and then cross features are extracted. The cross feature is the product of a single feature and another single feature. The number of cross features is more than one and includes any two types of data. The correlation between the two data types within the cross feature is determined, and the two types of data values ​​are sorted respectively. Each specific data value generates a ranking in this type of data. When the same data value appears, the ranking is the middle value of the ranking of these same values. The Spearman rank correlation coefficient is used to determine the correlation between the two data types within the cross feature. The formula is as follows:

[0021]

[0022] r s is the correlation coefficient value, d i is the ranking difference between the two types of data at the i-th position, and n is the number of data points;

[0023] When the correlation coefficient value r s When the absolute value of is greater than 0.5, the two data types within the cross feature are considered to have high correlation and are retained. s When the absolute value of is less than or equal to 0.5, it is considered that the two data types within this cross feature have no high correlation and are discarded;

[0024] All single features and retained cross features are aggregated to form the data set D°;

[0025] Statistics of historical disaster data were collected, and geological disaster labels were added to the dataset DD according to the timestamps of the collected historical data. The geological disaster type labels were set as follows: no geological disaster, landslide, mud-rock flow, and flood.

[0026] Furthermore, 80% of the data set D° is divided into a training set and 20% is a test set. An early warning model is created in the training set by building a decision tree. A feature A in the data set D° is taken and the information gain is set. The information gain is set according to the following formula:

[0027]

[0028] IG(D°,A) is the information gain of feature A for data set D°, H(D°) is the entropy of data set D°, representing the uncertainty of data set D°, Values(A) is all the data values ​​contained in feature A, D° v are all subsets containing feature A data value v, |D°| represents the number of samples in data set D°, |D° v | represents the subset D° v The number of samples, H(D° v ) is a subset D°v The entropy of .

[0029] Furthermore, the formula for entropy is as follows:

[0030]

[0031] H(U) is the entropy of the data set U, where U can refer to any data set, k is the number of geological disaster categories, k=4, p i is the probability of geological disaster category i, which means the proportion of the number of samples of a certain geological disaster in the total number of samples in the data set U.

[0032] Furthermore, the information gain of each feature of the dataset D° is calculated, and the dataset D° is split into several subsets D° by the feature A that maximizes the information gain. v , with feature A as the node, for each subset D° v Recursively call the splitting process to form a branch structure, and set the stopping condition to be that all samples are correctly classified according to the labels. At this point, the early warning model is obtained;

[0033] The early warning model is tested in the test set and judged by the accuracy. The correct samples are judged as the proportion of all samples in the test set. When the accuracy is higher than 95%, the early warning model is considered to be usable and deployed to each edge computing center. Otherwise, the early warning model is considered to be unusable and the parameters of the early warning model are adjusted for retraining.

[0034] Furthermore, real-time geological data, real-time environmental data and real-time meteorological data of each monitoring point are continuously collected, the collected real-time data are preprocessed in the same way as when training the model, and features are extracted according to the content of the data set D°. The early warning model predicts whether a geological disaster will occur based on the monitored real-time data. If a geological disaster will not occur, the monitoring frequency of the monitoring points in this area will be reduced to 50% of the maximum monitoring frequency. If a geological disaster will occur, the highest monitoring frequency will be restored to improve monitoring accuracy and early warning will be issued at the same time. Monitoring is continued at the highest monitoring frequency. If the early warning model determines that a geological disaster will occur within 1 day, an alarm will be issued.

[0035] The present invention also includes a mine geological environment intelligent monitoring system:

[0036] Information collection module, used to divide monitoring points and collect historical geological data, historical environmental data and historical meteorological data of the mine;

[0037] The edge computing module is used to pre-process the information collected by the information collection module, receive the early warning model and make real-time predictions;

[0038] Model training module, used to aggregate data from edge computing modules, train early warning models, and deploy them to edge computing modules;

[0039] The alarm module is used to receive instructions from the edge computing module and issue an alarm to the outside world.

[0040] (III) Beneficial effects

[0041] The present invention collects monitoring point data, constructs an early warning model, determines the power consumption of the monitoring point according to the results of the early warning model, and transmits alarm information to the outside world, thereby realizing the prediction of mining geological disasters and improving the accuracy of the prediction results. At the same time, the power consumption of the monitoring point is controlled, the working life of the monitoring point is extended, and the stability of the monitoring system is promoted. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic diagram of the process of the present invention;

[0043] Figure 2 It is a schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0044] 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.

[0045] See also Figure 1 , a mine geological environment intelligent monitoring and early warning method, the steps include:

[0046] Step 1: Divide the monitoring points and collect the historical geological data, historical environmental data and historical meteorological data of the mine;

[0047] The step 1 includes the following contents:

[0048] Step 101: Divide monitoring points in the mine to be monitored, and collect historical geological data, historical environmental data and historical meteorological data of the mine at the monitoring points. The historical geological data are the vibration frequency and amplitude of the rock mass of the mine. The vibration frequency can indicate changes in the rock mass structure or stress concentration areas. When the frequency changes, it may indicate the expansion of cracks inside the rock mass or the formation of new cracks. The amplitude indicates the intensity of the vibration. When the amplitude increases, it usually indicates that the activity of the rock mass is enhanced, which may indicate an increased risk of rock fracture or slippage. The historical environmental data are the temperature and humidity in the monitoring point area. Temperature changes may affect the stability of the rock mass. For example, temperature increases may cause rock expansion and increase stress. Temperature changes may also affect the rock The moisture content in the rock mass can change its physical properties. Abnormal temperature changes may indicate chemical reactions or fires inside the rock mass. Humidity has a significant impact on the stability of the rock mass. High humidity may increase the moisture content in the rock, reduce its strength and friction, and increase the risk of landslides or collapses. Monitoring humidity changes can help identify the impact of groundwater infiltration or precipitation on the rock mass. The historical meteorological data are wind speed and precipitation in the mining area. Long-term high wind speeds will blow away the soil in the gaps between the rocks on the surface of the mine, which will affect the stability of the rocks in the mine. Large amounts of precipitation in a short period of time can easily cause mudslides. Long-term continuous rainfall will also form flowing water in the gullies of the mine, affecting mine safety. The collected data are timestamped to form data set D.

[0049] By collecting specific data from each monitoring point, a data foundation is built for the training model. At the same time, by collecting data at the first time, it is helpful to make judgments on the monitoring points at the first time.

[0050] Step 2: Preprocess the data, extract features from the data, collect the types of geological disasters and organize them into labels to form a data set;

[0051] The step 2 includes the following contents:

[0052] Step 201: Edge computing nodes are evenly set between monitoring points. Each edge computing node is connected to at least two monitoring points. The edge computing nodes perform preliminary processing and analysis on the data and upload only valuable information to the database, which can reduce the pressure on network bandwidth. The edge computing nodes collect and summarize the data of the monitoring points connected to them, supplement the missing data with the next bit of data, and then perform noise filtering to smooth data changes, improve data accuracy and reliability, and provide higher quality data input for model training. The formula is:

[0053]

[0054] X(t j-a ) represents the time t j-aThe original data at the time point, a represents the index variable at the time point, a∈N, 0≤a≤N, X filtered (t j ) represents the filtered data;

[0055] N is the size of the filter window, where the filter window size for amplitude data is 30, the filter window size for vibration frequency, wind speed and precipitation data is 10, and the filter window size for temperature and humidity data is 5;

[0056] Step 202: Then, the data of the same type are normalized to form a data set DD.

[0057] By preprocessing the data, the difficulty of model training can be reduced.

[0058] Step 3: Train the model through decision tree, judge the node by information gain, deploy the model to the node and make real-time prediction;

[0059] The step 3 comprises the following steps:

[0060] Step 301: extract features from the data set DD, mark each data type as a single feature, and then extract cross features, where the cross feature is the product of a single feature and another single feature. The number of cross features is more than one and includes any two types of data. Determine the correlation between the two data types within the cross feature, sort the two types of data values ​​respectively, and each specific data value generates a ranking in the data of this type. When the same data value appears, the ranking is the middle value of the ranking of these same values. The Spearman rank correlation coefficient is used to determine the correlation between the two data types within the cross feature. The formula is as follows:

[0061]

[0062] r s is the correlation coefficient value, d i is the ranking difference between the two types of data at the i-th position, and n is the number of data points;

[0063] When the correlation coefficient value r s When the absolute value of is greater than 0.5, the two data types within the cross feature are considered to have high correlation and are retained. s When the absolute value of is less than or equal to 0.5, it is considered that the two data types within this cross feature have no high correlation and are discarded;

[0064] All single features and retained cross features are aggregated to form the data set D°;

[0065] Statistics of historical disaster data were collected, and geological disaster labels were added to the dataset DD according to the timestamps of the collected historical data. The geological disaster type labels were set as follows: no geological disaster, landslide, mud-rock flow, and flood.

[0066] By extracting features from the original data, we can understand the relationship between the data more clearly. At the same time, introducing various types of geological disasters can facilitate all-round prediction of monitoring points and help take targeted preventive measures.

[0067] Step 302: 80% of the data set D° is divided into a training set and 20% is divided into a test set. An early warning model is created in the training set by constructing a decision tree. A feature A in the data set D° is taken and the information gain is set. The formula is as follows:

[0068]

[0069] IG(D°,A) is the information gain of feature A for data set D°, H(D°) is the entropy of data set D°, representing the uncertainty of data set D°, Values(A) is all the data values ​​contained in feature A, D° v are all subsets containing feature A data value v, |D°| represents the number of samples in data set D°, |D° v | represents the subset D° v The number of samples, H(D° v ) is a subset D° v The entropy of .

[0070] Step 303: The entropy is based on the following formula:

[0071]

[0072] H(U) is the entropy of the data set U, where U can refer to any data set, k is the number of geological disaster categories, k=4, p i is the probability of geological disaster category i, which means the proportion of the number of samples of a certain geological disaster in the total number of samples in the data set U.

[0073] Step 304: Calculate the information gain of each feature of the dataset D°, and split the dataset D° into several subsets D° by the feature A that maximizes the information gain. v , taking feature A as a node, for each subset D° v The splitting process is called recursively to form a branch structure, and the stopping condition is set to that all samples are correctly classified according to the labels. At this time, the early warning model is obtained.

[0074] By calculating the entropy of the data set, quantifying the uncertainty of the data set, and then selecting the optimal features for splitting through information gain, the purity of the data set is gradually increased, and finally an efficient decision tree model is constructed.

[0075] Step 304: Test the early warning model in the test set and judge it by accuracy. The correct samples are judged as the proportion of all samples in the test set as the accuracy rate. When the accuracy rate is higher than 95%, the early warning model is considered to be usable and deployed to each edge computing center. Otherwise, the early warning model is considered to be unusable and the parameters of the early warning model are adjusted to retrain.

[0076] Deploying early warning models at edge nodes can enable prediction and analysis while data is being collected, significantly shortening the response time for early warnings. This can also improve the robustness and fault tolerance of the system. Even if problems occur with the central server, the edge nodes can still operate independently and provide prediction and early warning services.

[0077] Step 4: Control the energy consumption of the monitoring point and transmit the warning information of the monitoring point according to the results of the early warning model;

[0078] The step 4 includes the following contents:

[0079] Real-time geological data, real-time environmental data and real-time meteorological data of each monitoring point are continuously collected. The collected real-time data are preprocessed in the same way as when training the model, and features are extracted according to the content of the data set D°. The early warning model predicts whether a geological disaster will occur based on the real-time data monitored. If a geological disaster will not occur, the monitoring frequency of the monitoring points in this area will be reduced to 50% of the maximum monitoring frequency. If a geological disaster will occur, it will be restored to the highest monitoring frequency to improve monitoring accuracy and issue an early warning at the same time. Continuous monitoring is carried out at the highest monitoring frequency. If the early warning model determines that a geological disaster will occur within 1 day, an alarm will be issued.

[0080] See also Figure 2 , the present invention also includes a mine geological environment intelligent monitoring system:

[0081] The information collection module is used to divide the monitoring points and collect the historical geological data, historical environmental data and historical meteorological data of the mine. For example, seismic sensors, temperature and humidity sensors and meteorological sensors are installed in mine tunnels, pillars and guardrails, mine boundaries, geological faults and landslide-prone areas for initial data collection, and the working frequency is controlled according to the instructions of the edge computing module;

[0082] The edge computing module is used to pre-process the data and upload the pre-processed data to the network. The edge computing nodes are evenly distributed in the information collection module, and the early warning model is deployed to directly process the collected data on-site and complete the geological disaster prediction of the connected area on-site.

[0083] The model training module aggregates all collected data and trains the model, such as training a decision tree in the central processor, and then deploys the trained warning model to the edge computing module;

[0084] The alarm module, upon receiving information about possible geological disasters, such as information from the edge computing module, issues an early warning to the outside world.

[0085] The above formulas are all dimensionless and numerical calculations. The formula is a formula that is closest to the actual situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0086] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0087] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for intelligent monitoring and early warning of mine geological environment, characterized in that the steps include: Step 1: Divide the monitoring points and collect the historical geological data, historical environmental data and historical meteorological data of the mine; Step 2: Preprocess the data, extract features from the data, collect the types of geological disasters and organize them into labels to form a data set; Step 3: Train the model through decision tree, judge the node by information gain, deploy the model to the node and make real-time prediction; Step 4: Control the energy consumption of the monitoring point and transmit the warning information of the monitoring point according to the results of the early warning model.

2. The method for intelligent monitoring and early warning of mine geological environment according to claim 1, characterized in that: Monitoring points are divided in the mine to be monitored, and historical geological data, historical environmental data and historical meteorological data of the mine are collected at the monitoring points. The historical geological data are the vibration frequency and amplitude of the mine rock mass, the historical environmental data are the temperature and humidity in the monitoring point area, and the historical meteorological data are the wind speed and precipitation in the mine area. The collected data are timestamped to form data set D.

3. The method for intelligent monitoring and early warning of mine geological environment according to claim 2 is characterized in that: Edge computing nodes are evenly set between monitoring points. Each edge computing node is connected to at least two monitoring points. The edge computing node collects and summarizes the data of the monitoring points connected to it, supplements the missing data with the next bit of data, and then performs noise filtering. The formula is: X(t j-a ) represents the time t j-a The original data at the time point, a represents the index variable at the time point, a∈N, 0≤a≤N, X filtered (t j ) represents the filtered data; N is the size of the filter window, where the filter window size of the amplitude data is 30, the filter window size of the vibration frequency, wind speed and precipitation data is 10, and the filter window size of the temperature and humidity data is 5. Then, the data of the same type are normalized to form the data set DD.

4. The method for intelligent monitoring and early warning of mine geological environment according to claim 3 is characterized in that: Extract features from the data set DD, mark each data type as a single feature, and then extract cross features, which are the product of a single feature and another single feature. The number of cross features is more than one and includes any two types of data. Determine the correlation between the two data types within the cross feature, sort the two types of data values ​​respectively, and each specific data value generates a ranking in this type of data. When the same data value appears, the ranking is the middle value of the ranking of these same values. The Spearman rank correlation coefficient is used to determine the correlation between the two data types within the cross feature. The formula is as follows: r s is the correlation coefficient value, d i is the ranking difference between the two types of data at the i-th position, and n is the number of data points; When the correlation coefficient value r s When the absolute value of is greater than 0.5, the two data types within the cross feature are considered to have high correlation and are retained. s When the absolute value of is less than or equal to 0.5, it is considered that the two data types within this cross feature have no high correlation and are discarded; All single features and retained cross features are aggregated to form the data set D°; Statistics of historical disaster data were collected, and geological disaster labels were added to the dataset DD according to the timestamps of the collected historical data. The geological disaster type labels were set as follows: no geological disaster, landslide, mud-rock flow, and flood.

5. The method for intelligent monitoring and early warning of mine geological environment according to claim 4 is characterized in that: 80% of the data set D° is divided into a training set and 20% is divided into a test set. An early warning model is created in the training set by building a decision tree. A feature A in the data set D° is taken and the information gain is set. The formula is as follows: IG(D°,A) is the information gain of feature A for data set D°, H(D°) is the entropy of data set D°, representing the uncertainty of data set D°, Values(A) is all the data values ​​contained in feature A, D° v are all subsets containing feature A data value v, |D°| represents the number of samples in data set D°, |D° v | represents the subset D° v The number of samples, H(D° v ) is a subset D° v The entropy of .

6. The method for intelligent monitoring and early warning of mine geological environment according to claim 5, characterized in that: The entropy is based on the following formula: H(U) is the entropy of the data set U, where U can refer to any data set, k is the number of geological disaster categories, k=4, p i is the probability of geological disaster category i, which means the proportion of the number of samples of a certain geological disaster in the total number of samples in the data set U.

7. The method for intelligent monitoring and early warning of mine geological environment according to claim 6 is characterized in that: Calculate the information gain of each feature of the dataset D° and split the dataset D° into several subsets D° by the feature A that maximizes the information gain. v , taking feature A as a node, for each subset D° v Recursively call the splitting process to form a branch structure, and set the stopping condition to be that all samples are correctly classified according to the labels. At this point, the early warning model is obtained; The early warning model is tested in the test set and judged by the accuracy. The correct samples are judged as the proportion of all samples in the test set. When the accuracy is higher than 95%, the early warning model is considered to be usable and deployed to each edge computing center. Otherwise, the early warning model is considered to be unusable and the parameters of the early warning model are adjusted for retraining.

8. The method for intelligent monitoring and early warning of mine geological environment according to claim 7 is characterized in that: Real-time geological data, real-time environmental data and real-time meteorological data of each monitoring point are continuously collected. The collected real-time data are preprocessed in the same way as when training the model, and features are extracted according to the content of the data set D°. The early warning model predicts whether a geological disaster will occur based on the real-time data monitored. If a geological disaster will not occur, the monitoring frequency of the monitoring points in this area will be reduced to 50% of the maximum monitoring frequency. If a geological disaster will occur, it will be restored to the highest monitoring frequency to improve monitoring accuracy and issue an early warning at the same time. Continuous monitoring is carried out at the highest monitoring frequency. If the early warning model determines that a geological disaster will occur within 1 day, an alarm will be issued.

9. An intelligent monitoring system for mine geological environment, characterized in that: The monitoring system is used to execute a mine geological environment intelligent monitoring and early warning method according to any one of claims 1 to 8, comprising: Information collection module, used to divide monitoring points and collect historical geological data, historical environmental data and historical meteorological data of the mine; The edge computing module is used to pre-process the information collected by the information collection module, receive the early warning model and make real-time predictions; Model training module, used to aggregate data from edge computing modules, train early warning models, and deploy them to edge computing modules; The alarm module is used to receive instructions from the edge computing module and issue an alarm to the outside world.

Images