Digital mine safety production monitoring method and system based on Internet of Things

Abstract

The invention discloses a digital mine safety production monitoring method and system based on the Internet of Things, and relates to the technical field of mine safety. According to the invention, a multi-dimensional monitoring system is constructed by integrating multi-source data, and three-dimensional perception and visualization of a safety state are realized by combining a sensor topological map and a three-dimensional point cloud model; the improved LoRaWAN is adopted to guarantee the real-time performance of data transmission; based on the optimized CNN sensor data prediction model, triggering primary early warning to actively cut off an equipment power supply through residual calculation; calculating a risk value by applying a D-S evidence theory, and realizing historical case intelligent matching of second-level early warning and dynamic escape path planning of third-level early warning by combining a real-time three-dimensional stress cloud picture generated by finite element analysis; the method has the capabilities of adaptive optimization of model parameters and dynamic updating of a case library, realizes accurate risk positioning, multi-level differentiation early warning and scientific decision while improving the communication efficiency through three-dimensional visualization, and constructs a mine safety management full-closed-loop ecology.

Description

Technical Field

[0001] The present invention belongs to the technical field of mine safety. Specifically, it particularly relates to a digital mine safety production monitoring method and system based on the Internet of Things. Background Art

[0002] With the continuous development of the mining industry, the mine safety production faces challenges in various aspects such as complex environments, technical equipment, and management levels. It is still necessary to improve the safety production technology level to ensure the healthy development of mining enterprises and the lives of workers.

[0003] Chinese Patent CN108757040B discloses a mine safety monitoring system; it connects the upper and lower parts of the mine through a wireless communication network, and sets fixed monitoring devices, mobile monitoring devices, and camera devices to monitor the environmental parameters and personnel location information in the mining area in real time, realizing the linkage management of the monitoring system and improving the overall automation and informatization level; by obtaining various types of environmental data in the mining area in real time, determining the environmental safety levels of corresponding types and the ranges of dangerous areas of corresponding types, and when the environmental safety level is at a dangerous level, controlling multiple infrared sensors in the area where the monitoring sensor group is located to turn on and monitor the approaching signals of personnel and send alarm signals to the fixed monitoring devices in the corresponding areas.

[0004] In the prior art, the monitoring dimension of the mine environment is single, only monitoring environmental parameters and personnel locations, lacking real-time monitoring of the mechanical state of the mine and equipment operation parameters, and unable to comprehensively reflect the comprehensive safety state of the mine; the communication and data processing capabilities are insufficient, and the problems of signal attenuation, electromagnetic interference, and node conflicts in the complex mine environment are not solved, which may lead to data transmission delays or losses; the risk warning mechanism is simple, only triggering alarms according to the environmental safety level, not realizing multi-level warning hierarchical response, and lacking the intelligent decision-making ability of matching historical cases and dynamically planning escape routes; lacking predictive analysis ability, unable to predict the trends of sensor data and discover potential risks in advance. Summary of the Invention

[0005] (1) Technical Problems to be Solved

[0006] Aiming at the problems in the related technologies, the present invention provides a digital mine safety production monitoring method based on the Internet of Things. Through multi-dimensional data fusion and precise monitoring, efficient communication and predictive maintenance, multi-level warning and intelligent decision-making, and three-dimensional visualization and mechanical analysis, the present invention solves the problems of single monitoring dimension of the mine environment, insufficient communication and data processing capabilities, simple risk warning mechanism, and lack of predictive analysis.

[0007] (2) Technical Solutions

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

[0009] S1. Set data related to mine safety and deploy sensors; construct a sensor topology map and a 3D point cloud model of the mine; set response actions for first-level, second-level, and third-level warnings;

[0010] S2. Improve the traditional LoRaWAN in combination with the mine environment to obtain an improved LoRaWAN; construct a CNN sensor data prediction model, and optimize the parameters of the CNN sensor data prediction model in combination with an optimization algorithm to obtain an optimized sensor data prediction model;

[0011] S3. Collect original sensor data through sensors, and combine with the improved LoRaWAN to obtain real-time multimodal data; input the real-time multimodal data into the optimized CNN sensor data prediction model to obtain predicted sensor data; calculate the residuals between the real-time multimodal data and the predicted sensor data to obtain a set of residual values;

[0012] S4. Judge whether to execute a first-level warning according to the set of residual values; obtain a fused risk value using the D-S evidence theory; use finite element analysis, and combine with the sensor topology map and the 3D point cloud model of the mine to obtain a stress concentration coefficient, and judge whether to execute second-level and third-level warnings according to the fused risk value and the stress concentration coefficient.

[0013] Preferably, the S1 includes the following steps:

[0014] S11. Set a mine environment data set a = {a1, a2,..., a i ,... a m1}, a mine mechanics data set b = {b1, b2,..., b i ,... b m2}, a mine positioning label data set c = {c1, c2,..., c i ,... c m3} and a device status parameter set d = {d1, d2,..., d i ,... d m4};

[0015] Among them, a i represents the i-th environmental data related to mine safety, and m1 represents the total number of types of environmental data related to mine safety; b i represents the mechanical data related to mine safety, and m2 represents the total number of types of mechanical data related to mine safety; c i represents the positioning label data related to mine safety, and m3 represents the total number of types of mechanical data related to mine safety;

[0016] di represents the \(i\)th device parameter for reflecting the device status, \(m4\) represents the total number of device parameters; mine environmental data such as CH4, CO concentration and temperature and humidity data, mine mechanics data such as roof displacement and rock stratum strain data, mine positioning tag data such as personnel and equipment coordinates, and device parameters such as device vibration frequency and pressure;

[0017] S12. Deploy sensors at key mine locations according to the mine environmental data set, mine mechanics data set, mine positioning tag data set and device status parameter set;

[0018] S13. Register the deployed sensor positions into three-dimensional coordinates to obtain a sensor topology map with spatial attributes; Scan the mine using a mobile laser scanner to obtain a three-dimensional point cloud model of the mine;

[0019] S14. Set the response action for the first-level warning as stopping the power operation of relevant devices;

[0020] Set the similarity threshold as \(u\); Set the response action for the second-level warning as matching historical cases from the risk knowledge base and using the solutions of historical cases with similarity ≥ similarity threshold \(u\) to solve. If there are no historical cases with similarity ≥ similarity threshold \(u\), generate a risk report and send it to the safety officer, who conducts risk analysis and makes decisions;

[0021] Set the response action for the third-level warning as obtaining real-time personnel positioning and real-time three-dimensional stress cloud map and planning an escape route for personnel;

[0022] The above steps form a sensor topology map with spatial attributes and a three-dimensional point cloud model of the mine by setting the mine environmental, mechanics, positioning and device status data sets and deploying sensors at key locations; Define the response actions for the three-level warning, including cutting off the device power for the first-level warning, matching historical cases and generating risk reports for the second-level warning, and planning escape routes for the third-level warning; Realize the comprehensive monitoring of the mine safety status; The three-level warning mechanism can not only quickly respond to potential dangers and cut off the device power to prevent the expansion of accidents, but also provide solutions through historical case matching or generate risk reports for safety officers to analyze and make decisions when there are no matching cases; In case of emergency, it can use real-time positioning and stress cloud map to plan escape routes, significantly improving the mine's emergency response ability and personnel safety guarantee level.

[0023] Preferably, the said S2 includes the following steps:

[0024] S21. Improve the traditional LoRaWAN to obtain an improved LoRaWAN;

[0025] S22. Build a CNN sensor data prediction model, collect historical data, and adjust the parameters of the CNN sensor data prediction model in combination with an optimization algorithm to obtain an optimized CNN sensor data prediction model;

[0026] The above steps summarize the improvement of the traditional LoRaWAN protocol, forming an improved LoRaWAN, and building a CNN sensor data prediction model; by collecting historical data and combining it with an optimization algorithm to adjust the model parameters, an optimized CNN sensor data prediction model is finally obtained; the improved LoRaWAN is improved according to the real-time environment of the mine and is more suitable for the transmission environment of the mine; the optimized CNN sensor data prediction model improves the prediction accuracy of abnormal data through precise parameter adjustment, provides more reliable technical support for mine safety, and enhances the active prevention and control ability of the system.

[0027] Preferably, the S21 includes the following steps:

[0028] S211. Adjust the SF value according to the real-time mine signal-to-noise ratio to balance the relationship between roadway attenuation and electromagnetic interference;

[0029] S212. For the problem of high-density deployment conflicts inside the mine, use a time slot allocation algorithm to divide the time frame into fixed time slots and a contention window, and give priority to transmitting abnormal data; the time slot allocation algorithm formula is as follows,

[0030]

[0031] where T represents the time slot length, that is, the time period allocated to each node in the TDMA frame; M represents the frame length, that is, the duration of the entire TDMA frame; N represents the number of active nodes, that is, the number of nodes currently participating in communication; P represents the received power, that is, the signal power received by the node; Q represents the noise power spectral density, that is, the noise power per unit bandwidth;

[0032] S213. The node dynamically selects the next hop through a routing cost function and combines the link quality and remaining power of the mine; the routing cost function formula is as follows,

[0033]

[0034] where C represents the routing cost, which is used to evaluate the cost or cost of selecting a certain path; α, β, and λ all represent weight coefficients, which are used to adjust the importance of different factors in routing selection; RSSI represents the received signal strength indication, that is, the received signal strength; PLR represents the packet loss rate, that is, the proportion of data packets lost during transmission; ER represents the remaining power, that is, the remaining battery power of the node;

[0035] S214. Obtain the improved LoRaWAN through S211, S212, and S213;

[0036] The above steps adapt to the change of signal-to-noise ratio in the mine by adjusting the spreading factor (SF value) in real time, optimize the time frame division by using the time sequence allocation algorithm, give priority to the transmission of abnormal data, and dynamically select the communication path by using the routing cost function, taking into account the signal strength, packet loss rate, and remaining battery power of the node; effectively improve the performance and reliability of the mine communication system; adjusting the SF value in real time balances signal attenuation and electromagnetic interference, ensuring the stability of data transmission; the time sequence allocation algorithm solves the conflict problem under high-density deployment, ensuring the timely transmission of abnormal data; the introduction of the routing cost function enables the node to intelligently select the path according to the link quality and power status, reducing the cost and delay of data transmission, and overall improving the communication efficiency of the system and the response speed to mine safety risks.

[0037] Preferably, the S22 includes the following steps:

[0038] S221. Build a CNN sensor data prediction model and set the initial learning rate of the CNN sensor data prediction model to p1;

[0039] S222. Collect historical normal working condition data of mine sensors and historical abnormal working condition data of mine sensors to obtain training data; collect abnormal mine data corresponding to the first-level, second-level, and third-level early warnings to obtain test data;

[0040] S223. Set the training error threshold q1 and the maximum number of training iterations; use the training data to repeatedly train the CNN sensor data prediction model; after each round of training, calculate the training error q2 of the CNN sensor data prediction model and adjust the initial learning rate according to the training error;

[0041] When q2 ≤ q1 or the maximum number of iterations is reached, obtain the trained CNN sensor data prediction model with the learning rate of q2;

[0042] S224. Use the test data to test the trained CNN sensor data prediction model, and after the test is completed, obtain the optimized CNN sensor data prediction model;

[0043] The above steps include setting the initial learning rate of the model, collecting historical normal and abnormal working condition data of the mine as training data, and abnormal data corresponding to each level of early warning as test data; setting the training error threshold and the maximum number of iterations, repeatedly training the model using the training data, and adjusting the learning rate according to the training error; when the error requirement is met or the maximum number of iterations is reached, obtaining the trained model, and finally testing the model using the test data to obtain an optimized CNN sensor data prediction model; improving the optimized CNN sensor data prediction model, specifically for the accurate prediction of sensor data in the mine, providing strong support for timely early warning and taking emergency measures, and effectively improving the intelligent level of mine safety production.

[0044] Preferably, the S224 includes the following steps:

[0045] S2241. Set the test accuracy threshold r1, test the trained CNN sensor data prediction model using the test data, and obtain the test accuracy r2;

[0046] S2242. If r2 ≥ r1, then use the trained CNN sensor data prediction model as the optimized CNN sensor data prediction model; otherwise, use the hippopotamus algorithm to find the learning rate of the trained CNN sensor data prediction model, obtain the global optimal solution, use the global optimal solution as the learning rate of the trained CNN sensor data prediction model, and obtain the optimized CNN sensor data prediction model;

[0047] By setting the test accuracy threshold through the above steps, test and verify the trained CNN sensor data prediction model; if the test accuracy meets the preset threshold, directly use this model as the optimized model; if not, use the optimization algorithm to optimize the learning rate of the model and find the global optimal solution, so as to obtain a more accurate optimized CNN sensor data prediction model; ensuring that the prediction performance of the model meets the expected standard.

[0048] Preferably, using the hippopotamus algorithm to find the learning rate of the trained CNN sensor data prediction model and obtain the global optimal solution in S2242 includes the following steps:

[0049] S22421. Construct a hippopotamus population w, set the scale of the hippopotamus population as z, then the hippopotamus population w = {w1, w2,..., w i ,...w z}, where w i represents the i-th hippopotamus in the hippopotamus population;

[0050] S22422. Randomly generate the initial position group x = {x1, x2,..., x i,...x z}, where x i represents the initial position of the \(i\)-th hippopotamus in the hippopotamus population;

[0051] S22423. Start the iterative operation, and set the maximum number of optimization iterations to \(\varepsilon\); during each round of iteration, update the positions of each hippopotamus in the initial position set of the hippopotamus population according to the test accuracy; and obtain the global best hippopotamus individual and the global best objective in the hippopotamus population during each round of iteration.

[0052] S22424. When \(r2 \geq r1\) or the maximum number of optimization iterations is reached, the hippopotamus population stops the iterative operation and obtains the global optimal solution.

[0053] The above steps optimize the learning rate of the CNN sensor data prediction model through an optimization algorithm. First, construct a hippopotamus population and randomly generate initial positions, then update the hippopotamus positions through iterative operations, and find the global best hippopotamus individual and objective according to the test accuracy. When the test accuracy meets the preset threshold or the maximum number of iterations is reached, stop the iteration and obtain the global optimal solution of the learning rate; the introduction of the hippopotamus algorithm significantly improves the optimization effect of the CNN sensor data prediction model.

[0054] Preferably, the said S3 includes the following steps:

[0055] S31. Collect various data in the mine environment dataset, mine mechanics dataset, mine positioning label dataset, and equipment status parameter set through sensors to obtain the original sensor data; input the original sensor data into the improved LoRaWAN for transmission to obtain the time-stamp-aligned real-time multimodal data.

[0056] S32. Input the time-stamp-aligned real-time multimodal data into the optimized CNN sensor data prediction model to obtain the predicted sensor data.

[0057] S33. Set the residual value threshold to \(e\), calculate the residuals between the data in the predicted sensor data and the data in the time-stamp-aligned real-time multimodal data to obtain the residual value set.

[0058] The above steps collect the multivariate data of the mine through sensors and use the improved LoRaWAN for efficient transmission to ensure the time-stamp alignment and real-time nature of the data; input the time-stamp-aligned real-time multimodal data into the optimized CNN sensor data prediction model for prediction; calculate the residuals between the predicted data and the real-time data and set the residual value threshold to obtain the residual value set; realizing the real-time and efficient collection and processing of mine data, providing a solid data foundation for safety production monitoring.

[0059] Preferably, the said S4 includes the following steps:

[0060] S41. Determine whether there is a residual value ≥ the residual value threshold e in the residual value set. If so, execute a first-level warning, and collect the sensor data corresponding to the residual values ≥ the residual value threshold e in the residual value set to obtain an abnormal data set; otherwise, maintain the current state;

[0061] S42. Set the first-level risk value threshold as f and the second-level risk value threshold as g. Collect geological exploration data and equipment log data, combine the geological exploration data and equipment log data with the abnormal data in the abnormal data set to obtain combined abnormal data, and use the D-S evidence theory to calculate the combined abnormal data to obtain a fusion risk value h;

[0062] S43. Set the stress concentration coefficient threshold as k; extract the mechanical data in the real-time multimodal data to obtain real-time mechanical data, superimpose the real-time mechanical data with the mine three-dimensional point cloud model to obtain superimposed mechanical data, use finite element analysis to calculate the superimposed mechanical data to obtain a stress concentration coefficient, and combine the stress concentration coefficient with the sensor topology map to obtain a three-dimensional stress cloud map;

[0063] S44. If the fusion risk value h ≥ the first-level risk value threshold f, execute a second-level warning. If the fusion risk value h ≥ the second-level risk value threshold g or the stress concentration coefficient j ≥ the stress concentration coefficient threshold k, execute a third-level warning;

[0064] The above steps determine whether to trigger a first-level warning by judging whether the residual value set exceeds the preset threshold, and collect the corresponding abnormal data; combine geological exploration and equipment log data, calculate the fusion risk value using the D-S evidence theory; extract real-time mechanical data and superimpose it with the mine three-dimensional model, perform finite element analysis to obtain the stress concentration coefficient, and generate a three-dimensional stress cloud map; according to the comparison results of the fusion risk value and the stress concentration coefficient with their respective thresholds, execute the corresponding second-level or third-level warning; realize the precise identification and hierarchical warning of mine safety risks; improve the comprehensiveness and accuracy of risk assessment; provide an intuitive mechanical analysis basis for early warning; the hierarchical warning mechanism ensures differential responses for different risk levels, effectively improving the early warning ability and emergency response efficiency of the mine safety monitoring system.

[0065] An Internet of Things-based digital mine safety production monitoring system for implementing the above-mentioned Internet of Things-based digital mine safety production monitoring method, including a data collection and initialization module, a communication and prediction model optimization module, a real-time data processing and anomaly detection module, and a risk assessment and early warning trigger module;

[0066] The data collection and initialization module is used to define mine safety-related data sets; deploy sensors and construct a topology map and a three-dimensional point cloud model; set key sub-modules for response action rules for third-level early warnings:

[0067] The communication and prediction model optimization module is used to improve the traditional LoRaWAN protocol and construct and optimize a CNN sensor data prediction model to obtain an optimized CNN sensor data prediction model;

[0068] The real-time data processing and anomaly detection module is used to collect multi-modal data through sensors and align them through the improved LoRaWAN transmission; use the optimized CNN sensor data prediction model data to calculate the residuals between the real-time data and the predicted values to obtain a set of residual values;

[0069] The risk assessment and early warning trigger module is used to determine whether to trigger a first-level early warning based on the set of residual values; fuse the D-S evidence theory and finite element analysis to calculate the risk value and stress concentration coefficient, and determine whether to trigger second- and third-level early warnings.

[0070] (III) Beneficial effects

[0071] The present invention has the following beneficial effects:

[0072] The present invention constructs a multi-dimensional data monitoring system by integrating mine environmental data, mechanical data, equipment operation parameters, and personnel positioning information; combines a sensor topology map with a mine three-dimensional point cloud model to realize three-dimensional perception and visual display of the mine safety state, comprehensively covering environmental, equipment, personnel, and geomechanical risks, and significantly improving the monitoring accuracy and coverage.

[0073] In view of the complex mine environment, the improved LoRaWAN protocol effectively solves the problems of signal attenuation and node conflict by dynamically adjusting communication parameters, timing allocation algorithms, and routing optimization, ensuring the real-time and reliable data transmission; at the same time, based on the CNN sensor data prediction model and optimization algorithm optimization, the system can predict the abnormal trend of data in advance, trigger a first-level early warning through residual calculation, and realize the transformation from passive response to active prevention and control.

[0074] The present invention directly cuts off the power supply of the equipment through the first-level early warning to prevent the expansion of the accident; the second-level early warning matches historical cases through a risk knowledge base to assist in rapid decision-making; the third-level early warning combines real-time personnel positioning and a three-dimensional stress cloud map to dynamically plan the optimal escape route; through the multi-source data fusion of the D-S evidence theory and the mechanical calculation of the finite element analysis, the system can quantify the risk value and stress concentration coefficient, providing a scientific basis for differential emergency response.

[0075] Through the adaptive optimization of the prediction model parameters by the optimization algorithm and the continuous update of the historical case library, the system has the ability of dynamic learning and iteration; the three-dimensional visualization interface is linked with the real-time data, which not only supports accurate risk positioning, but also provides data support for the long-term safety planning of the mine; compared with the traditional solution, the present invention has achieved breakthroughs in communication efficiency, early warning intelligence, decision-making science and system adaptability, and constructed a full-closed-loop ecosystem for mine safety management.

[0076] Of course, it is not necessary for any product implementing the present invention to achieve all the above-mentioned advantages simultaneously. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] In order to more clearly illustrate the technical solutions of the embodiments of the invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0078] Figure 1 It is a schematic flow chart of a method for monitoring the safe production of a digital mine based on the Internet of Things according to the present invention;

[0079] Figure 2 It is a schematic flow chart of deploying sensors and setting three-level early warning response actions in a method for monitoring the safe production of a digital mine based on the Internet of Things according to the present invention;

[0080] Figure 3 It is a schematic flow chart of obtaining an improved LoRaWAN in a method for monitoring the safe production of a digital mine based on the Internet of Things according to the present invention;

[0081] Figure 4 It is a schematic flow chart of obtaining an optimized CNN sensor data prediction model in a method for monitoring the safe production of a digital mine based on the Internet of Things according to the present invention;

[0082] Figure 5 It is a schematic diagram of the modules of a system for monitoring the safe production of a digital mine based on the Internet of Things according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0083] The following will clearly and completely describe the technical solutions in the embodiments of the invention with reference to the drawings in the embodiments of the invention. Obviously, the described embodiments are only some embodiments of the invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the invention without creative efforts belong to the scope of protection of the invention.

[0084] In the description of the present invention, it should be understood that the terms "opening", "upper", "lower", "top", "middle", "inner", etc. indicating orientation or positional relationships are only for convenience in describing the invention and simplifying the description, rather than indicating or implying that the components or elements referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation on the invention.

[0085] Embodiment 1:

[0086] Please refer to Figure 1 , Figure 2 , Figure 3 and Figure 4 , the present invention discloses a digital mine safety production monitoring method based on the Internet of Things, including the following steps:

[0087] S1. Set data related to mine safety and deploy sensors; construct a sensor topology map and a three-dimensional point cloud model of the mine; set response actions for first-level, second-level, and third-level early warnings;

[0088] The said S1 includes the following steps:

[0089] S11. Set a mine environmental data set a = {a1, a2,..., a i ,... a m1}, a mine mechanics data set b = {b1, b2,..., b i ,... b m2}, a mine positioning tag data set c = {c1, c2,..., c i ,... c m3}, and a device status parameter set d = {d1, d2,..., d i ,... d m4};

[0090] Among them, a i represents the i-th environmental data related to mine safety, and m1 represents the total number of types of environmental data related to mine safety; b i represents the mechanics data related to mine safety, and m2 represents the total number of types of mechanics data related to mine safety; c i represents the positioning tag data related to mine safety, and m3 represents the total number of types of mechanics data related to mine safety;

[0091] d i represents the i-th device parameter for reflecting the device status, and m4 represents the total number of device parameters; mine environmental data such as CH4, CO concentration, and temperature and humidity data, mine mechanics data such as roof displacement and strata strain data, mine positioning tag data such as personnel and equipment coordinates, and device parameters such as device vibration frequency and pressure;

[0092] S12. Deploy sensors at key mine location points according to the mine environment dataset, mine mechanics dataset, mine positioning tag dataset, and equipment status parameter set;

[0093] S13. Register the deployed sensor positions into three-dimensional coordinates to obtain a sensor topology map with spatial attributes; Scan the mine using a mobile laser scanner to obtain a three-dimensional point cloud model of the mine;

[0094] S14. Set the response action for level-one warning as stopping the power operation of relevant equipment;

[0095] Set the similarity threshold as u; Set the response action for level-two warning as matching historical cases from the risk knowledge base and using the solutions of historical cases with similarity ≥ similarity threshold u to solve the problem. If there are no historical cases with similarity ≥ similarity threshold u, generate a risk report and send it to the safety officer, who will conduct risk analysis and make a decision;

[0096] Set the response action for level-three warning as obtaining real-time personnel positioning and real-time three-dimensional stress cloud map and planning an escape route for personnel;

[0097] S2. Improve the traditional LoRaWAN in combination with the mine environment to obtain an improved LoRaWAN; Build a CNN sensor data prediction model and optimize the parameters of the CNN sensor data prediction model in combination with an optimization algorithm to obtain an optimized sensor data prediction model;

[0098] The S2 includes the following steps:

[0099] S21. Improve the traditional LoRaWAN to obtain an improved LoRaWAN;

[0100] The S21 includes the following steps:

[0101] S211. Adjust the SF value according to the real-time mine signal-to-noise ratio to balance the relationship between roadway attenuation and electromagnetic interference;

[0102] S212. For the problem of high-density deployment conflict inside the mine, use the time slot allocation algorithm to divide the time frame into fixed time slots and a contention window, and give priority to propagating abnormal data; The time slot allocation algorithm formula is as follows,

[0103]

[0104] where, T represents the time slot length; N represents the number of active nodes; P represents the received power, that is, the signal power received by the node; Q represents the noise power spectral density;

[0105] S213. The node dynamically selects the next hop through the routing cost function and in combination with the link quality and remaining power of the mine; The routing cost function formula is as follows,

[0106]

[0107] Among them, C represents the routing cost, α, β, and λ all represent weight coefficients, RSSI represents the received signal strength indication, PLR represents the packet loss rate, and ER represents the remaining power;

[0108] S214. Obtain the improved LoRaWAN through S211, S212, and S213;

[0109] S22. Improve the D-S evidence theory to obtain the improved D-S evidence theory;

[0110] S22. Construct a CNN sensor data prediction model, collect historical data, and combine with an optimization algorithm to adjust the parameters of the CNN sensor data prediction model to obtain an optimized CNN sensor data prediction model;

[0111] The said S22 includes the following steps:

[0112] S221. Construct a CNN sensor data prediction model, and set the initial learning rate of the CNN sensor data prediction model to p1;

[0113] S222. Collect historical normal working condition data of mine sensors and historical abnormal working condition data of mine sensors to obtain training data; collect abnormal mine data corresponding to the first-level, second-level, and third-level early warnings to obtain test data;

[0114] S223. Set the training error threshold q1 and the maximum number of training iterations; use the training data to repeatedly train the CNN sensor data prediction model; after each round of training, calculate the training error q2 of the CNN sensor data prediction model, and adjust the initial learning rate according to the training error;

[0115] When q2 ≤ q1 or the maximum number of iterations is reached, obtain the trained CNN sensor data prediction model with the learning rate of q2;

[0116] S224. Use the test data to test the trained CNN sensor data prediction model, and after the test is completed, obtain the optimized CNN sensor data prediction model;

[0117] The said S224 includes the following steps:

[0118] S2241. Set the test accuracy threshold r1, use the test data to test the trained CNN sensor data prediction model, and obtain the test accuracy r2;

[0119] S2242. If r2 ≥ r1, then use the trained CNN sensor data prediction model as the optimized CNN sensor data prediction model; otherwise, use the hippopotamus algorithm to find the learning rate of the trained CNN sensor data prediction model, obtain the global optimal solution, use the global optimal solution as the learning rate of the trained CNN sensor data prediction model, and obtain the optimized CNN sensor data prediction model.

[0120] The steps of using the hippopotamus algorithm to find the learning rate of the trained CNN sensor data prediction model and obtain the global optimal solution in the above S2242 include the following:

[0121] S22421. Construct a hippopotamus population w, and set the size of the hippopotamus population to z. Then the hippopotamus population w = {w1, w2,..., w i ,... w z}, where w i represents the i-th hippopotamus in the hippopotamus population;

[0122] S22422. Randomly generate the initial position group x = {x1, x2,..., x i ,... x z} of the hippopotamus population according to the learning rate, where x i represents the initial position of the i-th hippopotamus in the hippopotamus population;

[0123] S22423. Start the iterative operation, and set the maximum number of optimization iterations to ε; in each round of iteration, update the positions of each hippopotamus in the initial position set of the hippopotamus population according to the test accuracy; and obtain the global best hippopotamus individual and the global best objective in the hippopotamus population in each round of iteration.

[0124] S22424. When r2 ≥ r1 or the maximum number of optimization iterations is reached, the hippopotamus population stops the iterative operation and obtains the global optimal solution;

[0125] S3. Collect the original sensor data through the sensor, and combine it with the improved LoRaWAN to obtain the real-time multimodal data; input the real-time multimodal data into the optimized CNN sensor data prediction model to obtain the predicted sensor data; calculate the residuals between the real-time multimodal data and the predicted sensor data to obtain the residual value set.

[0126] The above S3 includes the following steps:

[0127] S31. Collect the data in the mine environment dataset, mine mechanics dataset, mine positioning label dataset, and equipment status parameter set through the sensor to obtain the original sensor data; input the original sensor data into the improved LoRaWAN for transmission to obtain the real-time multimodal data with timestamp alignment;

[0128] S32. Input the real-time multimodal data into the optimized CNN sensor data prediction model to obtain the predicted sensor data;

[0129] S33. Set the residual value threshold as e, calculate the residuals between the data in the predicted sensor data and the data in the real-time multimodal data to obtain a set of residual values;

[0130] S4. Judge whether to execute a primary warning according to the set of residual values; obtain the fused risk value using the D-S evidence theory; use finite element analysis and combine the sensor topology map with the mine 3D point cloud model to obtain the stress concentration coefficient, and judge whether to execute secondary and tertiary warnings according to the fused risk value and the stress concentration coefficient;

[0131] The said S4 includes the following steps:

[0132] S41. Judge whether there is a residual value in the set of residual values that is ≥ the residual value threshold e. If so, execute a primary warning and collect the sensor data corresponding to the residual values in the set of residual values that are ≥ the residual value threshold e to obtain an abnormal data set; otherwise, maintain the current state;

[0133] S42. Set the primary risk value threshold as f and the secondary risk value threshold as g. Collect geological exploration data and equipment log data, combine the geological exploration data and equipment log data with the abnormal data in the abnormal data set to obtain combined abnormal data, and use the D-S evidence theory to calculate the combined abnormal data to obtain the fused risk value h;

[0134] S43. Set the stress concentration coefficient threshold as k; extract the mechanical data in the real-time multimodal data to obtain real-time mechanical data, superimpose the real-time mechanical data on the mine 3D point cloud model to obtain superimposed mechanical data, use finite element analysis to calculate the superimposed mechanical data to obtain the stress concentration coefficient, and combine the stress concentration coefficient with the sensor topology map to obtain a 3D stress cloud map;

[0135] S44. If the fused risk value h ≥ the primary risk value threshold f, execute a secondary warning. If the fused risk value h ≥ the secondary risk value threshold g or the stress concentration coefficient j ≥ the stress concentration coefficient threshold k, execute a tertiary warning.

[0136] Embodiment 2:

[0137] Please refer to Figure 5 , an Internet of Things-based digital mine safety production monitoring system for implementing the above-mentioned Internet of Things-based digital mine safety production monitoring method, including a data acquisition and initialization module, a communication and prediction model optimization module, a real-time data processing and anomaly detection module, and a risk assessment and warning trigger module;

[0138] The data acquisition and initialization module is used to define datasets related to mine safety; deploy sensors and construct a topological map and a three-dimensional point cloud model; set the key sub-module of the response action rules for three-level early warning:

[0139] The communication and prediction model optimization module is used to improve the traditional LoRaWAN protocol and construct and optimize a CNN sensor data prediction model to obtain an optimized CNN sensor data prediction model;

[0140] The real-time data processing and anomaly detection module is used to collect multi-modal data through sensors and align them through the improved LoRaWAN transmission; use the data of the optimized CNN sensor data prediction model to calculate the residuals between the real-time data and the predicted values to obtain a set of residual values;

[0141] The risk assessment and early warning trigger module is used to determine whether to trigger a first-level early warning according to the set of residual values; fuse the D-S evidence theory and finite element analysis to calculate the risk value and the stress concentration coefficient, and determine whether to trigger the second- and third-level early warnings.

[0142] In the description of this specification, the description referring to terms such as "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0143] The preferred embodiments of the invention disclosed above are only used to help explain the invention. The preferred embodiments do not elaborate all the details, nor do they limit the invention to the specific embodiments described. Obviously, many modifications and variations can be made according to the content of this specification. These embodiments are selected and specifically described in this specification to better explain the principle and practical application of the invention, so that those skilled in the art can understand and utilize the invention well.

Claims

1. A method for monitoring the safe production of a digital mine based on the Internet of Things, characterized in that, It includes the following steps: S1. Set data related to mine safety and deploy sensors; construct a sensor topology map and a 3D point cloud model of the mine; set response actions for first-level, second-level, and third-level warnings; S2. Improve the traditional LoRaWAN in combination with the mine environment to obtain an improved LoRaWAN; construct a CNN sensor data prediction model, and optimize the parameters of the CNN sensor data prediction model in combination with an optimization algorithm to obtain an optimized sensor data prediction model; S3. Collect original sensor data through sensors and combine it with the improved LoRaWAN to obtain real-time multimodal data; input the real-time multimodal data into the optimized CNN sensor data prediction model to obtain predicted sensor data; Calculate the residuals between the real-time multimodal data and the predicted sensor data to obtain a set of residual values; S4. Determine whether to execute a first-level warning according to the set of residual values; Obtain a fused risk value using the D-S evidence theory; Use finite element analysis and combine the sensor topology map and the 3D point cloud model of the mine to obtain a stress concentration coefficient, and determine whether to execute second-level and third-level warnings according to the fused risk value and the stress concentration coefficient.

2. The method for monitoring the safe production of a digital mine based on the Internet of Things according to claim 1, wherein, The S1 includes the following steps: S11. Set the mine environment data set a = {a1, a2,..., a i ,... a m1}, the mine mechanics data set b = {b1, b2,..., b i ,... b m2}, the mine positioning label data set c = {c1, c2,..., c i ,... c m3}, and the equipment status parameter set d = {d1, d2,..., d i ,... d m4}; Among them, a i represents the i-th environmental data related to mine safety, and m1 represents the total number of types of environmental data related to mine safety; b i represents the mechanical data related to mine safety, and m2 represents the total number of types of mechanical data related to mine safety; c i represents the positioning tag data related to mine safety, and m3 represents the total number of types of mechanical data related to mine safety; d i represents the i-th equipment parameter used to reflect the equipment state, and m4 represents the total number of equipment parameters; S12. Deploy sensors at key mine location points according to the mine environment dataset, mine mechanics dataset, mine positioning label dataset, and equipment status parameter set; S13. Register the deployed sensor positions into three-dimensional coordinates to obtain a sensor topology map with spatial attributes; scan the mine using a mobile laser scanner to obtain a 3D point cloud model of the mine; S14. Set the response action for the first-level warning as stopping the power operation of relevant equipment; Set the similarity threshold as u; set the response action for the second-level warning as matching historical cases from the risk knowledge base and using the solutions of historical cases with a similarity ≥ similarity threshold u to solve the problem. If there are no historical cases with a similarity ≥ similarity threshold u, generate a risk report and send it to the safety officer, who will conduct a risk analysis and make a decision; Set the response action for the third-level warning as obtaining real-time personnel positioning and real-time three-dimensional stress cloud map and planning an escape route for personnel.

3. A method for monitoring the safe production of a digital mine based on the Internet of Things according to claim 1, characterized in that, The S2 includes the following steps: S21. Improve the traditional LoRaWAN to obtain an improved LoRaWAN; S22. Construct a CNN sensor data prediction model, collect historical data, and adjust the parameters of the CNN sensor data prediction model in combination with an optimization algorithm to obtain an optimized CNN sensor data prediction model.

4. A safety production monitoring method for digital mines based on the Internet of Things according to claim 3, characterized in that, The S21 includes the following steps: S211. Adjust the SF value according to the real-time mine signal-to-noise ratio to balance the relationship between roadway attenuation and electromagnetic interference; S212. For the problem of high-density deployment conflicts inside the mine, use a time slot allocation algorithm to divide the time frame into fixed time slots and a contention window, and give priority to propagating abnormal data; the time slot allocation algorithm formula is as follows, where T represents the time slot length, N represents the number of active nodes, P represents the received power, that is, the signal power received by the node, and Q represents the noise power spectral density; S213. The node dynamically selects the next hop through a routing cost function, in combination with the link quality and remaining power of the mine; the formula of the routing cost function is as follows. Among them, C represents the routing cost, α, β, and λ all represent weight coefficients, RSSI represents the received signal strength indication, PLR represents the packet loss rate, and ER represents the remaining power. S214. The improved LoRaWAN is obtained through S211, S212, and S213.

5. The method for monitoring the safe production of a digital mine based on the Internet of Things according to claim 3, characterized in that, The S22 includes the following steps: S221. Construct a CNN sensor data prediction model, and set the initial learning rate of the CNN sensor data prediction model to p1. S222. Collect historical normal working condition data and historical abnormal working condition data of mine sensors to obtain training data; collect abnormal mine data corresponding to the first-level, second-level, and third-level early warnings to obtain test data. S223. Set the training error threshold q1 and the maximum number of training iterations; use the training data to repeatedly train the CNN sensor data prediction model; after each round of training, calculate the training error q2 of the CNN sensor data prediction model, and adjust the initial learning rate according to the training error. When q2 ≤ q1 or the maximum number of iterations is reached, a trained CNN sensor data prediction model with a learning rate of q2 is obtained. S224. Use the test data to test the trained CNN sensor data prediction model. After the test is completed, an optimized CNN sensor data prediction model is obtained.

6. The method for monitoring the safe production of a digital mine based on the Internet of Things according to claim 5, wherein The S224 includes the following steps: S2241. Set the test accuracy threshold r1, use the test data to test the trained CNN sensor data prediction model, and obtain the test accuracy r2. S2242. If r2 ≥ r1, then use the trained CNN sensor data prediction model as the optimized CNN sensor data prediction model; otherwise, use the hippopotamus algorithm to find the learning rate of the trained CNN sensor data prediction model to obtain the global optimal solution, and use the global optimal solution as the learning rate of the trained CNN sensor data prediction model to obtain the optimized CNN sensor data prediction model.

7. A method for monitoring the safe production of a digital mine based on the Internet of Things according to claim 6, characterized in that, The steps of using the hippopotamus algorithm to find the learning rate of the trained CNN sensor data prediction model to obtain the global optimal solution in the S2242 include: S22421. Construct a hippopotamus population w. Set the size of the hippopotamus population to z. Then the hippopotamus population w = {w1, w2,..., w i ,... w z}, where w i represents the i-th hippopotamus in the hippopotamus population; S22422. Randomly generate an initial position group x = {x1, x2,..., x i ,...x z} of the hippopotamus population according to the learning rate, where x i represents the initial position of the i-th hippopotamus in the hippopotamus population; S22423. Start the iterative operation, and set the maximum number of optimization iterations to ε; in each round of iteration, update the positions of each hippopotamus in the initial position set of the hippopotamus population according to the test accuracy; and obtain the global best hippopotamus individual and the global best target in the hippopotamus population in each round of iteration. S22424. When r2 ≥ r1 or the maximum number of optimization iterations is reached, the hippopotamus population stops the iterative operation to obtain the global optimal solution.

8. A method for monitoring the safe production of a digital mine based on the Internet of Things according to claim 1, characterized in that, The S3 includes the following steps: S31. Collect various data in the mine environment dataset, mine mechanics dataset, mine positioning label dataset, and equipment status parameter set through sensors to obtain raw sensor data; input the raw sensor data into the improved LoRaWAN for transmission to obtain timestamp-aligned real-time multimodal data. S32. Input the real-time multimodal data into the optimized CNN sensor data prediction model to obtain the predicted sensor data; S33. Set the residual value threshold as e, calculate the residuals between the data in the predicted sensor data and the data in the real-time multimodal data to obtain a set of residual values.

9. A safety production monitoring method for digital mines based on the Internet of Things according to claim 1, characterized in that, The said S4 includes the following steps: S41. Determine whether there is a residual value in the set of residual values that is ≥ the residual value threshold e. If so, execute a primary warning, and collect the sensor data corresponding to the residual values in the set of residual values that are ≥ the residual value threshold e to obtain an abnormal data set; otherwise, maintain the current state; S42. Set the primary risk value threshold as f and the secondary risk value threshold as g. Collect geological exploration data and equipment log data, combine the geological exploration data and equipment log data with the abnormal data in the abnormal data set to obtain combined abnormal data, and use the D-S evidence theory to calculate the combined abnormal data to obtain a fusion risk value h; S43. Set the stress concentration coefficient threshold as k; extract the mechanical data in the real-time multimodal data to obtain real-time mechanical data, superimpose the real-time mechanical data on the mine three-dimensional point cloud model to obtain superimposed mechanical data, use finite element analysis to calculate the superimposed mechanical data to obtain a stress concentration coefficient, and combine the stress concentration coefficient with the sensor topology map to obtain a three-dimensional stress cloud map; S44. If the fusion risk value h ≥ the primary risk value threshold f, execute a secondary warning. If the fusion risk value h ≥ the secondary risk value threshold g or the stress concentration coefficient j ≥ the stress concentration coefficient threshold k, execute a tertiary warning.

10. A digital mine safety production monitoring system based on the Internet of Things is used to implement a digital mine safety production monitoring method as described in any one of claims 1-9, and is characterized in that, The said system includes a data acquisition and initialization module, a communication and prediction model optimization module, a real-time data processing and anomaly detection module, and a risk assessment and warning trigger module.

Images