Intelligent mine safety monitoring system and method

Abstract

The invention relates to the technical field of mine monitoring, and provides an intelligent mine safety monitoring system and method, and the system comprises a path network modeling device, a multi-source risk monitoring device, a personnel positioning and tracking device, a dynamic path updating device, and an intelligent evacuation guiding device. When any path risk level exceeds a preset safety threshold value, triggering an emergency evacuation instruction: calculating and generating a personalized optimal evacuation path for each person based on the currently updated three-dimensional topology model and the person position information; outputting evacuation information; the evacuation information comprises an emergency evacuation alarm of a whole mine broadcast level; and a real-time dynamic pointing instruction facing each node, the pointing instruction being generated according to the personalized optimal evacuation path and being presented through a physical indication device deployed at the node or a terminal device carried by a person. According to the technical scheme disclosed by the invention, workers in the mine can be timely guided to be evacuated.

Description

Technical Field

[0001] The present application relates to the field of mine monitoring technology, and more specifically, to a smart mine safety monitoring system and method. Background Art

[0002] The contents of this section merely provide background information related to this application and may not constitute prior art.

[0003] Underground mines typically have a complex, three-dimensional tunnel network. This network performs multiple critical functions, including personnel passage, material transportation, and ventilation. However, when emergencies such as fire, flooding, and roof collapse require large-scale evacuation, the following challenges arise:

[0004] Limitations of spatial cognition: Underground workers generally lack a global understanding of the vast and dynamically changing topology of the tunnel network. Their evacuation route selection relies primarily on fragmented personal spatial memories and a limited static node identification system.

[0005] Convergence of group behavior: The aforementioned cognitive limitations can easily lead to a phenomenon in which evacuees, under conditions of panic and information asymmetry, tend to choose similar routes. Large numbers of people spontaneously flock to remembered or seemingly familiar escape routes.

[0006] Bottleneck effect at key nodes: Due to geological conditions, functional requirements and cost constraints, there are generally a large number of narrow sections (such as connecting tunnels, air doors, ladder rooms, maintenance passages, etc.) in mine tunnel design. Their effective traffic cross-sections are extremely small, often allowing only one person to pass through in sequence, and the throughput capacity is extremely limited.

[0007] Congestion and a sudden drop in efficiency: When evacuation flows converge at these narrow nodes, they instantly form a serious bottleneck. Crowd density rises dramatically, flow rates plummet, and even stagnate. This congestion not only significantly reduces the actual efficiency of the node but also triggers a chain reaction, blocking upstream lanes and potentially causing secondary disasters such as stampedes.

[0008] In summary, the current mine management system cannot reasonably guide the escape direction of personnel inside the mine, making it impossible for personnel inside the mine to evacuate the mine quickly and efficiently when an emergency occurs. Summary of the Invention

[0009] In view of this, the purpose of this application is to provide a smart mine safety monitoring system and monitoring method. The smart mine safety monitoring system and monitoring method disclosed in this application can be implemented.

[0010] The purpose of this application is achieved through the following technical solutions:

[0011] A smart mine safety monitoring system, comprising:

[0012] A path network modeling device is used to construct and store a three-dimensional topological model of an underground mine tunnel network, wherein the three-dimensional topological model includes nodes, path connection relationships, and dynamic traffic efficiency parameters of each path under preset conditions;

[0013] Multi-source risk monitoring devices are distributed and deployed along each path in the mine to collect and process indicator data in real time. The indicator data includes at least vibration information that characterizes rock mass stability and gas concentration information that characterizes environmental safety, and generates a real-time risk level for each path.

[0014] Personnel location tracking device, used to obtain and update the number of people entering the mining area and their location information in real time, and generate personnel location information;

[0015] A dynamic path updating device is connected to the multi-source risk monitoring device and the path network modeling device to receive real-time risk level and index data of each path and dynamically update the three-dimensional topology model accordingly;

[0016] The intelligent evacuation guidance device is in communication with the dynamic path updating device, the multi-source risk monitoring device, and the personnel positioning and tracking device, and is configured to:

[0017] Real-time monitoring of the risk level output by multi-source risk monitoring devices;

[0018] When the risk level of any path exceeds the preset safety threshold, an emergency evacuation command is triggered:

[0019] Based on the currently updated 3D topological model and personnel location information, a personalized optimal evacuation path is calculated and generated for each person;

[0020] Output evacuation information;

[0021] The evacuation information includes:

[0022] Mine-wide broadcast-level emergency evacuation alarm;

[0023] Real-time dynamic directional instructions for each node, the directional instructions are generated according to the personalized optimal evacuation path and presented through a physical indicator device deployed at the node or a terminal device carried by personnel.

[0024] The smart mine safety monitoring system provided in this application significantly improves evacuation efficiency and personnel safety in mine emergencies through its integrated and intelligent design. Its technical effects are mainly reflected in:

[0025] Accurate and dynamic risk perception and path assessment: Through distributed multi-source risk monitoring devices, the system realizes refined, real-time monitoring and risk assessment of key safety indicators (such as rock stability and gas environment) across the entire mine, providing high-confidence input for subsequent decision-making.

[0026] The dynamic path update device continuously refreshes the traffic status (such as whether it is unobstructed, blocked, or dangerous) and traffic efficiency parameters (such as speed and capacity) of each path in the three-dimensional topology model based on real-time risk data, ensuring that the network information relied on for evacuation path planning always reflects the current actual situation, effectively overcoming the lag of static maps or simple signs in disaster environments.

[0027] Intelligent, personalized evacuation decision-making and guidance: When a disaster triggers, the intelligent evacuation guidance device can comprehensively utilize a real-time updated 3D topological model (including path status and efficiency) and accurate personnel location distribution information to generate a personalized optimal evacuation path for each underground occupant. This effectively solves the problem of blindly following and assuming the same path as traditional evacuations due to a lack of information.

[0028] Traditional mine path risk monitoring methods typically focus solely on a single risk factor, such as vibration or gas concentration, and fail to comprehensively consider the impact of multiple risk factors on mine path safety. Furthermore, independent warnings are generated for different types of monitoring data, lacking unified integration and risk level determination standards. This results in an inability to accurately reflect the true risk status of mine paths, making it prone to delayed warnings or misjudgments. This makes it difficult to meet the demand for real-time, accurate monitoring of path risks in complex mine environments, and fails to provide a comprehensive and effective decision-making basis for mine safety management.

[0029] In some possible embodiments, the multi-source risk monitoring device includes:

[0030] The vibration information monitoring module is distributed and deployed on each path in the mine to monitor the vibration information of each path and generate the first warning level based on the vibration information;

[0031] Gas concentration monitors are distributed and deployed along each route in the mine to monitor gas concentration information along each route and generate a second warning level based on the gas concentration information, which includes both combustible and hazardous gas concentrations.

[0032] The risk level generator receives the first warning level and the second warning level in real time, and takes the highest level of the first warning level and the second warning level as the risk level of the path.

[0033] This solution uses distributed deployment of vibration information monitoring modules and gas concentration monitors to obtain real-time vibration information and gas concentration information for each mine path, and generates the first warning level and the second warning level, thus achieving comprehensive monitoring of multi-dimensional risk factors of mine paths. The risk level generator receives two types of warning levels in real time, and uses the highest level as the final risk level to build a unified and scientific risk assessment system. This method effectively integrates multi-source monitoring data, avoids the one-sidedness caused by single monitoring, and can more accurately and promptly reflect the actual risk status of mine paths, significantly improving the accuracy and effectiveness of risk warnings, providing reliable data support and decision-making basis for mine safety management, and helping mines to carry out risk prevention and control work efficiently.

[0034] Existing rock wall vibration monitoring technologies often only monitor and analyze vibration information in a single direction when assessing the impact of vibration on the rock wall. This fails to fully integrate lateral and longitudinal vibration data from the rock wall, resulting in an inaccurate assessment of the actual vibration impact on the rock wall. Furthermore, the lack of extraction and accumulation of analysis of the temporal characteristics of vibration information makes it difficult to accurately capture the micro-cracks gradually generated within the rock mass due to long-term vibration and their development and changes. This can easily lead to monitoring omissions or delayed warnings, hindering the timely and effective assessment and warning of potential rock wall risks.

[0035] In some possible embodiments, the vibration information monitoring module includes:

[0036] Vibration sensor, used to obtain transverse vibration information P and longitudinal vibration information S of the rock wall;

[0037] The vibration information processor receives the lateral vibration information P and the longitudinal vibration information S to generate the vibration index information h t , collect vibration index information h t Generate vibration index sequence H, h t Represents the vibration index information at time t;

[0038] The vibration warning generator extracts the time series features of the vibration index sequence H and generates a first warning level based on the time series features.

[0039] This solution uses vibration sensors to acquire transverse and longitudinal vibration information (P and S) from the rock wall. This information is then integrated into a vibration information processor to generate a vibration index sequence (H) containing multi-dimensional vibration information. This allows for a comprehensive and accurate assessment of the vibration impact on the rock wall, effectively addressing the one-sided nature of single-direction monitoring. A vibration early warning generator extracts the temporal characteristics of the vibration index sequence (H) to analyze the long-term accumulation of microcracks within the rock mass. When the accumulated microcrack value exceeds a preset value, a timely and accurate alarm is issued. This significantly improves the accuracy of monitoring and early warning capabilities for potential rock wall risks, providing more reliable technical support for rock wall stability assessment.

[0040] In some possible embodiments, h t ={E, V, L}; E represents the vibration energy parameter, V represents the deformation parameter, and L represents the energy release parameter;

[0041]

[0042] ρ represents the rock density, v p，s represents the average velocity of the transverse vibration wave and the longitudinal vibration wave, R represents the distance from the crack source, and t s represents the duration of vibration, μ represents the displacement function of the vibration sensor, and ∫dt represents the integral operation over time;

[0043]

[0044] M represents the fracture energy parameter, which is used to measure the total mechanical energy released during the fault rupture process, U represents the rock mass shear stiffness, and E represents the vibration energy parameter;

[0045] c represents the normalization constant, d represents the scaling exponent, M represents the crack energy parameter, and E represents the vibration energy parameter.

[0046] The vibration energy parameter can quantify the destructive strength of the rock wall when the crack breaks, and can serve as an early signal of mine collapse. The deformation parameter can measure the deformation range of the rock mass, and then detect abnormal rise when the mine is about to collapse, so that an early warning can be issued. The energy release parameter can measure the force accumulation of the rock mass, and then describe the internal structure of the rock mass. In this way, E, V, and L in this application can respectively describe the actual situation of the rock wall from three dimensions: the early stage of rock wall collapse, the period when the rock wall is about to collapse, and the loose internal structure of the rock wall. It can not only provide an early warning in the early stage of rock wall collapse, but also issue early warning information when it is about to collapse.

[0047] When analyzing rock wall vibration data, traditional vibration monitoring systems often fail to effectively separate key parameters and extract in-depth features from the vibration information. Furthermore, they struggle to integrate features from different dimensions. For example, they are unable to specifically extract the time series characteristics of the vibration energy, deformation, and energy release parameters in the vibration index sequence, which reflect the different stages of rock wall collapse. This results in inadequate information utilization and makes it difficult to identify subtle trends within the rock mass.

[0048] In some possible embodiments, the vibration warning generator includes:

[0049] There are three information processing units, each for inputting E, V, and L in the vibration index sequence H to extract a first extraction feature, a second extraction feature, and a third extraction feature, respectively;

[0050] The information fusion unit connects the first extracted features, the second extracted features, and the third extracted features, and performs global pooling to generate a temporal feature.

[0051] The vibration warning generator in this solution uses an information processing unit to extract the first, second, and third extracted features from the E, V, and L parameters in the vibration index sequence H, respectively. This enables precise analysis and feature mining of key rock wall vibration parameters, fully leveraging the risk information inherent in each parameter. The information fusion unit further connects and globally pools the three types of time series features, integrating the scattered features into a complete and comprehensive time series feature set. This effectively incorporates multi-dimensional information, including early stage rock wall collapse, impending collapse, and internal structural changes.

[0052] When processing multi-parameter time series data (such as E, V, and L), traditional LSTM networks extract feature vectors that differ significantly in scale and dimension. Directly concatenating these misaligned feature vectors leads to weight imbalance during information fusion, making it impossible to effectively capture temporal correlations between parameters, thereby reducing the accuracy of predicting internal structural changes and collapse risk. Furthermore, traditional LSTMs are susceptible to vanishing / exploding gradients, making it difficult to learn long-term dependencies, further limiting the accuracy of time series feature extraction.

[0053] The information processing unit includes: input layer, LSTM layer, normalization layer, and fully connected layer;

[0054] The input layer is used to input a parameter in the vibration index information;

[0055] LSTM layer, with 128 built-in LSTM cores. Each LSMT unit includes an input gate, a forget gate, a candidate memory unit, and an output gate.

[0056] Input Gate:

[0057] Forget gate: f t =σ(W if x t +b if +W hf h t-1 +b hf );

[0058] Candidate memory cells:

[0059] Output gate: o t =σ(W iox t +b io +W ho h t-1 +b ho );

[0060] The update memory of each LSTM core is updated as follows:

[0061]

[0062] The hidden state output of each LSTM core is: h t =o t ⊙tanh(c t );

[0063] Where: σ is the sigmoid activation function, ⊙ represents element-by-element multiplication, λ is the L2 regularization coefficient, t represents the current time step, T represents the total length of the sequence, and x t represents the input vector at time t, n represents the feature dimension, h t-1 Indicates the hidden state of the previous moment, c t-1 Indicates the memory state of the previous moment, W ii 、W if 、W ig 、W io Represents x t The input gate weight matrix, forget gate weight matrix, candidate memory unit weight matrix and output gate weight matrix, i, f, g, o represent the input gate, forget gate, candidate memory unit and output gate respectively, W hi 、W hf 、W hg 、W ho They represent the input gate recurrent weight, forget gate recurrent weight, candidate memory unit recurrent weight, and output gate recurrent weight, respectively. hi 、b hf 、b hg 、b ho Represent the input gate bias term, forget gate bias term, candidate memory unit bias term, and output gate bias term respectively;

[0064] i t =σ(·) represents the input gate vector, f t =σ(·) represents the forget gate vector, Represents the candidate memory unit vector tanh: represents the hyperbolic tangent function, o t =σ(·) represents the output gate vector; c t represents the output variable updated by the LSTM core, h t The feature vector representing the output of the output gate;

[0065] Normalization layer, normalizes the feature vector of each LSTM layer to obtain a normalized vector

[0066]

[0067] μ t ,σ t are the mean and standard deviation of the time step t within the batch, γ represents the scaling parameter, β represents the offset parameter, and ∈ is a numerical stability constant;

[0068] The fully connected layer normalizes the vector Connect and use ReLU as the activation function to output the extracted feature z:

[0069]

[0070] W fc1 is the output weight matrix, b fc1 is the output bias term.

[0071] In the technical solution provided by this application, in order to accurately extract the timing features in E, V, and L, a normalization layer is set in each LSMT unit. The normalization layer will standardize the feature vector output by each LSTM core, and adjust the length of the feature vector under the influence of the scaling parameter and the offset parameter, so that the feature vectors extracted from different weight information can be aligned with each other, so that when the feature vectors output by the LSTM core are subsequently connected, the connection accuracy is high.

[0072] This solution introduces a normalization layer after the LSTM unit. By adaptively adjusting the scaling parameter (γ) and the offset parameter (β), it unifies the feature vectors extracted by different parameters into the same distribution space, significantly improving feature alignment accuracy. The normalization process effectively alleviates the vanishing gradient problem, enabling the network to learn more complex temporal dependencies. Furthermore, the application of the ReLU activation function in the fully connected layer enhances nonlinear expression capabilities and further explores the implicit correlations between parameters.

[0073] In some possible embodiments, the gas concentration monitor generates a second warning level based on the growth rates of the combustible gas concentration and the harmful gas concentration.

[0074] The winding and deep tunnels cause multiple reflections and refractions of signals, triggering a multipath effect. The mountain rock's strong absorption of electromagnetic waves causes significant signal attenuation. The dust and mist generated by blasting operations further interfere with signal transmission. These factors reduce positioning accuracy, or even completely eliminate it, making it impossible to meet the real-time, precise monitoring of personnel locations required for mine safety management.

[0075] In some possible embodiments, the personnel location tracking device includes:

[0076] Entrance and exit people counter, used to obtain the number of people entering and leaving the mine to generate the total number of people in the mine;

[0077] There are multiple personnel monitors, which are arranged at the entrance and exit of each path, and are used to obtain the number of people entering the path and the number of people leaving the path to generate the total number of people in the path;

[0078] The personnel distribution information generator generates the number of people in each path in the three-dimensional topological model and the personnel location information based on the total number of people in the mine and the total number of people in each path.

[0079] In the technical solution provided in this application, by installing a large number of personnel monitors and entrance and exit population counters, the number of people in each path is counted, and the personnel distribution information in the mine can also be obtained.

[0080] Existing evacuation plans typically involve sending fixed evacuation routes directly to personnel, or updating them in real time. These routes are difficult to memorize, and when routes are updated, personnel may not be able to detect them in time and take the wrong route, hindering evacuation. Furthermore, route updates can cause panic, making evacuation even more difficult.

[0081] In some possible embodiments, the intelligent evacuation guidance device includes:

[0082] Information collection module, which updates the 3D topology model and personnel location information in real time;

[0083] An evacuation information generation unit creates an evacuation plan based on the 3D topology model and personnel location information. The evacuation plan includes the number of people passing through each node in different directions.

[0084] There are multiple evacuation information indication modules, which are respectively arranged at each node, receive the evacuation plan in real time, and send the passing direction to the people passing through the node based on the evacuation plan;

[0085] Broadcast module, used to send emergency evacuation alerts to the entire mining area;

[0086] The evacuation information generation unit updates the evacuation plan based on the real-time updated three-dimensional topology model and personnel location information.

[0087] The technical solution provided by this application continuously updates the evacuation plan based on the three-dimensional topological model and personnel location information, thereby ensuring that the evacuation work can be carried out in an orderly manner. In addition, when updating the evacuation plan, it is updated directly at each node, so internal personnel are unaware of route changes. On the one hand, this can prevent personnel from failing to evacuate as required due to not noticing the change in the evacuation route during evacuation, and on the other hand, it can also prevent personnel from panicking due to knowing that the route has changed.

[0088] A smart mine safety monitoring method uses the aforementioned smart mine safety monitoring system to monitor mines.

[0089] The beneficial effects of this application are:

[0090] (1) Accurate and dynamic risk perception and path assessment: Through distributed multi-source risk monitoring devices, the system realizes refined and real-time monitoring and risk assessment of key safety indicators (such as rock stability and gas environment) across the entire mine, providing high-confidence input for subsequent decision-making.

[0091] (2) The dynamic path updating device continuously updates the traffic status (such as whether it is unobstructed, blocked, or dangerous) and traffic efficiency parameters (such as speed and capacity) of each path in the three-dimensional topology model based on real-time risk data, ensuring that the network information relied on by evacuation path planning always reflects the current actual situation, effectively overcoming the lag of static maps or simple signs in disaster environments.

[0092] (3) Intelligent and personalized evacuation decision-making and guidance: When a disaster is triggered, the intelligent evacuation guidance device can comprehensively use the real-time updated three-dimensional topological model (including path status and efficiency) and accurate personnel location distribution information to generate a personalized optimal evacuation path for each underground personnel. This effectively solves the problem of blindly following and path convergence caused by lack of information in traditional evacuation. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 This is a structural diagram of the smart mine safety monitoring system.

[0094] Figure 2 A more specific structural diagram of the smart mine safety monitoring system.

[0095] Figure 3 This is a structural diagram of the information processing unit. DETAILED DESCRIPTION

[0096] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific implementation methods. The same figure marks in the accompanying drawings represent the same components. It should be noted that the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the described embodiments of this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0097] Compared to the embodiments shown in the drawings, feasible embodiments within the scope of protection of the present application may have fewer components, other components not shown in the drawings, different components, differently arranged components, or differently connected components, etc. In addition, two or more components in the drawings may be implemented in a single component, or a single component shown in the drawings may be implemented as multiple separate components.

[0098] Unless otherwise defined, technical or scientific terms used herein shall have the ordinary meanings understood by persons of ordinary skill in the art to which this application belongs. The terms "first," "second," and similar terms used in this specification and claims do not denote any order, quantity, or importance, but are merely used to distinguish different components.

[0099] Example 1:

[0100] refer to Figure 1 and Figure 2 The first embodiment of the present application discloses a smart mine safety monitoring system, comprising a path network modeling device, a multi-source risk monitoring device, a personnel positioning and tracking device, a dynamic path updating device, and an intelligent evacuation guidance device. The multi-source risk monitoring device is signal-connected to the dynamic path updating device, which is signal-connected to the path network modeling device. The intelligent evacuation guidance device is signal-connected to the path network modeling device and the personnel positioning and tracking device, respectively. Connections are made using a wired method such as optical fiber or broadband.

[0101] A path network modeling device is used to construct and store a three-dimensional topological model of an underground mine tunnel network. The three-dimensional topological model includes nodes, path connection relationships, and dynamic traffic efficiency parameters of each path under preset conditions.

[0102] The path network modeling device defines lane intersections, entrances, exits, and forks as nodes. The lanes connecting these nodes are considered paths. Using topological principles, it clearly depicts how these nodes connect to form a complete lane network.

[0103] Dynamic traffic efficiency parameters are calculated by comprehensively considering a variety of preset conditions. For example, during normal operating hours, these preset conditions include factors such as mine car traffic frequency, personnel density, tunnel slope, and ventilation conditions. These can be obtained in advance through experiments. In the event of an emergency, such as a small localized collapse or equipment failure that blocks traffic, the path network modeling device can quickly switch to the preset conditions, reassessing and updating the traffic efficiency parameters for the corresponding paths.

[0104] For example, before a collapse occurs, visibility in a mine is high, leading to high throughput efficiency. Conversely, after a collapse, the efficiency decreases due to the dust. The throughput efficiency parameter is essentially the maximum number of people that can pass through per unit time.

[0105] Multi-source risk monitoring devices are distributed throughout the mine's pathways, collecting and processing real-time indicator data. This data includes at least vibration information that indicates rock mass stability and gas concentration information that indicates environmental safety, generating a real-time risk level for each pathway. The multi-source risk monitoring devices primarily monitor tunnel collapse and gas hazards. Mine collapse is primarily detected by monitoring acoustic signals within the tunnel, while gas hazards are monitored by monitoring the concentrations of combustible and hazardous gases.

[0106] The multi-source risk monitoring device includes a vibration information monitoring module, a gas concentration monitor, and a risk level generator. The vibration information monitoring module and the gas concentration monitor are respectively connected to the risk level generator, which is in turn connected to the dynamic path update device.

[0107] The vibration information monitoring module is distributed and deployed on various paths in the mine. It is used to monitor the vibration information of each path and generate the first warning level based on the vibration information; the gas concentration monitor is distributed and deployed on various paths in the mine. It is used to monitor the gas concentration information of each path and generate the second warning level based on the gas concentration information. The gas concentration information includes the concentration of combustible gases and the concentration of harmful gases; the risk level generator receives the first warning level and the second warning level in real time, and takes the highest level of the first warning level and the second warning level as the risk level of the path.

[0108] The vibration information monitoring module and the gas concentration monitor are respectively corresponding to the collapse and gas hazard monitoring of an area (path or tunnel).

[0109] Specifically, the gas concentration monitor generates a second warning level based on the growth rate of combustible and hazardous gas concentrations. Mine tunnels typically have well-developed exhaust systems, and gas hazards typically occur when large amounts of fuel gas or flammable gas leaks occur. Therefore, the growth rate of combustible and hazardous gas concentrations is monitored. A higher growth rate corresponds to a higher second warning level. For example, there are five second warning levels, each corresponding to a range of growth rates for combustible and hazardous gas concentrations.

[0110] Mine tunnel collapses are not sudden events, but rather a gradual process. Essentially, they occur when, for some reason, small cracks continuously develop within the rock mass above the tunnel. As these cracks increase in number, the rock mass's internal structure changes, and the rock mass loses its structural integrity, leading to collapse. When cracks appear within the rock mass, they generate acoustic signals. Receive these acoustic signals to analyze the location and size of the cracks within the rock mass, thereby monitoring whether tunnel collapse is imminent.

[0111] Specifically, the vibration information monitoring module includes a vibration sensor, a vibration information processor, and a vibration warning generator. The vibration sensor, the vibration information processor, and the vibration warning generator are connected in sequence. The vibration sensor is used to obtain the lateral vibration information P and the longitudinal vibration information S of the rock wall. The lateral vibration information P is the transverse wave received by the vibration sensor, and the longitudinal vibration information S is the longitudinal wave received by the vibration sensor. The vibration information processor receives the lateral vibration information P and the longitudinal vibration information S to generate the vibration index information h t , collect vibration index information h t Generate vibration index sequence H, h t Indicates the vibration index information at time t.

[0112] h t ={E, V, L}; E represents the vibration energy parameter, V represents the deformation parameter, and L represents the energy release parameter;

[0113]

[0114] ρ represents the rock density, v p，s represents the average velocity of the transverse vibration wave and the longitudinal vibration wave, R represents the distance from the crack source, and t s represents the duration of vibration, μ represents the displacement function of the vibration sensor, ∫dt represents the integral operation of time, and the crack source distance is calculated based on the time difference of the arrival of the sound wave. That is, the vibration sensor can be an array of multiple sensors, and the crack source distance is calculated based on the time when each sensor receives the sound wave signal.

[0115]

[0116] M represents the crack energy parameter, which is used to measure the total mechanical energy released during the fault rupture process. U represents the shear stiffness of the rock mass. E represents the vibration energy parameter. The crack energy parameter is used to reflect the magnitude of the permanent static displacement caused by the crack source and is generated based on the spectrum energy of the longitudinal vibration information S.

[0117] c represents the normalization constant, d represents the scaling exponent, M represents the crack energy parameter, and E represents the vibration energy parameter.

[0118] Generally speaking, only when cracks occur in the rock mass can the vibration sensor receive the lateral vibration information P and the longitudinal vibration information S of the rock wall. However, cracks do not always occur in the rock mass, so the vibration index information h in different time periods is t The vibration index information h generated by different crack sizes is not the same. t The values ​​of the cracks are different, and the cracks have a cumulative effect on the collapse. Therefore, it is necessary to combine the vibration index information h of different time periods. t Collect them as the vibration index sequence H, and then conduct a comprehensive analysis on the vibration index sequence H.

[0119] Specifically, the vibration warning generator extracts the temporal characteristics of the vibration index sequence H and generates a first-level warning based on these characteristics. By extracting the temporal characteristics of the index sequence H, this solution allows for a global view of the accumulation of cracks within the rock mass and accurately identifies the timing of tunnel collapse.

[0120] The vibration warning generator includes an information processing unit and an information fusion unit, which are signal-connected. The information processing unit is provided with three units, each of which is used to input E, V, and L in the vibration index sequence H to extract the first extracted feature, the second extracted feature, and the third extracted feature, respectively. The information fusion unit connects the first extracted feature, the second extracted feature, and the third extracted feature and performs global pooling to generate a time series feature.

[0121] refer to Figure 3 ,The information processing unit includes : an input layer, an LSTM layer, a normalization layer, and a fully connected layer. ,The input layer, LSTM layer, normalization layer, and the fully connected layer are connected in sequence;

[0122] The input layer is used to input a parameter in the vibration index information;

[0123] LSTM layer, with 128 built-in LSTM cores. Each LSMT unit includes an input gate, a forget gate, a candidate memory unit, and an output gate.

[0124] Input Gate:

[0125] Forget gate: f t =σ(W if x t +b if +W hf h t-1 +b hf );

[0126] Candidate memory cells:

[0127] Output gate: o t =σ(W io x t +b io +W ho h t-1 +b ho );

[0128] The update memory of each LSTM core is updated as follows:

[0129]

[0130] The hidden state output of each LSTM core is: h t =o t ⊙tanh(c t );

[0131] Where: σ is the sigmoid activation function, ⊙ represents element-by-element multiplication, λ is the L2 regularization coefficient, t represents the current time step, T represents the total length of the sequence, and x t represents the input vector at time t, n represents the feature dimension, h t-1 Indicates the hidden state of the previous moment, c t-1 Indicates the memory state of the previous moment, W ii 、W if 、W ig 、W io Represents x t The input gate weight matrix, forget gate weight matrix, candidate memory unit weight matrix and output gate weight matrix, i, f, g, o represent the input gate, forget gate, candidate memory unit and output gate respectively, W hi 、W hf 、W hg 、W ho They represent the input gate recurrent weight, forget gate recurrent weight, candidate memory unit recurrent weight, and output gate recurrent weight, respectively. hi 、b hf 、b hg 、b ho Represent the input gate bias term, forget gate bias term, candidate memory unit bias term, and output gate bias term respectively;

[0132] i t =σ(·) represents the input gate vector, f t =σ(·) represents the forget gate vector, Represents the candidate memory unit vector tanh: represents the hyperbolic tangent function, o t =σ(·) represents the output gate vector; c t represents the output variable updated by the LSTM core, h t The feature vector representing the output of the output gate;

[0133] Normalization layer, normalizes the feature vector of each LSTM layer to obtain a normalized vector

[0134]

[0135] μ t ,σ t are the mean and standard deviation of the time step t within the batch, γ represents the scaling parameter, β represents the offset parameter, and ∈ is a numerical stability constant;

[0136] The fully connected layer normalizes the vector Connect and use ReLU as the activation function to output the extracted feature z:

[0137]

[0138] W fc1 is the output weight matrix, b fc1 is the output bias term.

[0139] After outputting the extracted feature z, the first extracted feature, the second extracted feature, and the third extracted feature are connected and globally pooled to generate a time series feature. Then, different time series features are mapped to the corresponding first warning level.

[0140] A dynamic path updating device is connected to the multi-source risk monitoring device and the path network modeling device to receive real-time risk level and index data of each path and dynamically update the three-dimensional topology model accordingly;

[0141] The dynamic path updating device is connected to the multi-source risk monitoring device and the path network modeling device, receives the real-time risk level and index data of each path, and dynamically updates the three-dimensional topology model based on the data.

[0142] When updating the 3D topology model, the dynamic efficiency parameters are primarily updated. For example, if a path is detected to have a high risk level or high concentration of hazardous gases, the passability of that path is likely to be poor, and therefore the dynamic efficiency parameters need to be lowered. If a path is detected to have collapsed or the concentration of hazardous gases exceeds a preset safety value, the dynamic efficiency parameters for that path are set to 0, indicating that the area is impassable.

[0143] The personnel location tracking device is used to obtain and update the number of people entering the mining area and their location information in real time, generating personnel location information. When tracking personnel, it is difficult to accurately locate the personnel location through wireless networks.

[0144] The personnel positioning and tracking device includes: an entrance and exit people counter, a personnel monitor, and a personnel distribution information generator. The entrance and exit people counter and the personnel monitor are respectively connected to the personnel distribution information generator by signal, and the personnel distribution information generator is connected to the intelligent evacuation guidance device by signal.

[0145] Entrance and exit crowd counters are used to collect data on the number of people entering and exiting the mine, generating the total number of people within the mine. Multiple personnel monitors are provided, located at the entrance and exit of each path, to collect data on the number of people entering and exiting a path, generating the total number of people within that path. Both the entrance and exit crowd counters and personnel monitors use infrared technology to identify the number of people passing through. A personnel distribution information generator generates information on the number of people on each path within the three-dimensional topological model, based on the total number of people within the mine and the total number of people on each path, as well as information on the location of people.

[0146] Although the personnel positioning tracking device cannot obtain the exact location information of each person, it can obtain the approximate location information of each person. Based on the approximate location information of each person, evacuation routes can also be designed. In addition, personnel will be monitored by the personnel monitors on the nodes during the movement process, so that the location information of the personnel will not be missed.

[0147] The intelligent evacuation guidance device is in communication with the dynamic path updating device, the multi-source risk monitoring device, and the personnel positioning and tracking device, and is configured to:

[0148] Real-time monitoring of the risk level output by multi-source risk monitoring devices;

[0149] When the risk level of any path exceeds the preset safety threshold, an emergency evacuation command is triggered:

[0150] Based on the currently updated 3D topological model and personnel location information, a personalized optimal evacuation path is calculated and generated for each person;

[0151] Output evacuation information;

[0152] Evacuation information includes:

[0153] Mine-wide broadcast-level emergency evacuation alarm;

[0154] Real-time dynamic directional instructions for each node, the directional instructions are generated according to the personalized optimal evacuation path and presented through a physical indicator device deployed at the node or a terminal device carried by personnel.

[0155] Specifically, the intelligent evacuation guidance device includes an information collection module, an evacuation information generation unit, an evacuation information indication module, and a broadcast module. The information collection module is signal-connected to the path network modeling device and the personnel location tracking device. The information collection module is signal-connected to the evacuation information generation unit. The evacuation information indication module and the broadcast module are each signal-connected to the evacuation information generation unit.

[0156] The information collection module updates the 3D topology model and occupant location information in real time. The evacuation information generation unit creates an evacuation plan based on the 3D topology model and occupant location information. The plan includes the number of people passing through each node in different directions. The evacuation information indication module and the evacuation information generation unit are connected via a wired signal.

[0157] Multiple evacuation information indicator modules are provided, each deployed at each node. They receive evacuation plans in real time and, based on the plans, send directions to personnel passing through the node. The evacuation information indicator module can be a sign placed at each node. For example, if a node has three escape directions, A, B, and C, and all personnel needing to pass through the node flee in direction A, the sign for direction A will light up, and personnel passing through the node will automatically evacuate in direction A.

[0158] In addition, the evacuation information indication module can be a wireless signal transmitter that can send low-frequency signals. When a worker reaches a node with a handheld pointing device, it will automatically receive the low-frequency signal and then generate a corresponding direction indication. The broadcast module is used to send an emergency evacuation alarm to the entire mining area. During testing, the emergency evacuation alarm can be ensured to be heard throughout the mining area. In this way, workers will start to evacuate after hearing the emergency evacuation alarm. When passing each node, they can evacuate according to the pointing direction on the node. The evacuation information generation unit updates the evacuation plan based on the real-time updated three-dimensional topological model and personnel location information. Workers evacuate at different speeds, so the evacuation plan needs to be constantly updated.

[0159] Example 2: Example 2 provides a method for generating an evacuation plan based on Example 1. The specific method for generating an evacuation plan includes the following steps:

[0160] Step 1: Collect personnel location information and 3D topology models in real time and set preset conditions;

[0161] The default conditions are:

[0162] (1) The personnel travel speed is a constant value;

[0163] (2) The length of each path is a fixed value;

[0164] (3) The locations of exits and entrances are constant, and all personnel are required to reach at least one exit or entrance;

[0165] (4) The number of people arriving at each node per unit time cannot exceed the dynamic traffic efficiency parameter;

[0166] (5) The length of each person's evacuation path cannot exceed the preset value. If it exceeds the preset value, a penalty parameter needs to be set;

[0167] Step 2: Set constraints based on pre-set conditions;

[0168] The constraints are:

[0169] represents the time when the total number of people with a proportion a is evacuated from the mine;

[0170] x ik Indicates whether the kth person passes the rth node in unit time, x rk =1 means pass, x rk Equal to 0 means failed, M i represents the dynamic traffic efficiency parameter of the i-th node, K represents the total number of people, and this constraint requires that the number of people passing through each node per unit time must be less than the dynamic traffic efficiency parameter;

[0171] L represents the number of nodes passed through. This constraint states that each person will not pass through the same node twice;

[0172] Step 3: Randomly generate a route a for each person, collect all routes a to generate a feasible solution AW;

[0173] Step 4: Delete the feasible solutions AW that do not meet the constraints and construct the path matrix RW;

[0174] The path matrix RW is sorted from large to small based on the personnel route overlap factor c;

[0175] in, Indicates the positive and negative signs. If the value in is not greater than 0, Take 0, when If the value in is greater than 0, is the absolute value symbol;

[0176] When most of the people in the feasible solution AW choose the same route, the The value of is large, and the corresponding route overlap factor c is large. When most of the feasible solutions AW choose different routes, then If it is smaller than L, the value of c is very small;

[0177] Step 5: Set the fitness function f(x), the maximum number of iterations, and randomly generate several particles in the path matrix RW. The fitness value of each particle needs to be greater than the preset value;

[0178]

[0179] Among them, j represents the node index near the high risk level, and J represents the number of nodes near the high risk level;

[0180] Step 6: Update the particle position based on the following conditions;

[0181] Step 61: Calculate the steering vector B;

[0182] b represents the number of iterations, represents the current position of particle s, represents the speed of particle s, Maxib represents the maximum number of iterations;

[0183] Step 62: Calculate the distance vector D α 、D β 、D δ ;

[0184] Among them, C1, C2, C3 represent the disturbance parameters, C l =3×rand[0.1]; Indicates the position with the best fitness value among all particles, Indicates the position with the second best fitness value among all particles, Indicates the position with the third best fitness value among all particles, l represents the index of the perturbation parameter, and rand[0.1] represents the generation of a random number between 0 and 1. In this scheme, as long as the random number generated is greater than one-third, the perturbation parameter will be greater than 1. Therefore, when calculating the distance vector in practice, the position of the distance vector will be increased as much as possible, so that particles are more inclined to diffuse to the bottom of the path matrix RW during the diffusion process, and in practice, they can quickly converge to the optimal fitness function f(x).

[0185] Step 63: Update particle positions

[0186]

[0187] Repeat the above steps until the maximum number of iterations is reached or the fitness function f(x) reaches a preset value.

[0188] Example 3: A smart mine safety monitoring method, using the smart mine safety monitoring system described in Example 1 to monitor the mine.

[0189] The above are merely preferred embodiments of the present application and are not intended to limit the present application. Those skilled in the art will readily appreciate that various modifications and variations are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A smart mine safety monitoring system, characterized in that: include: A path network modeling device is used to construct and store a three-dimensional topological model of an underground mine tunnel network, wherein the three-dimensional topological model includes nodes, path connection relationships, and dynamic traffic efficiency parameters of each path under preset conditions; Multi-source risk monitoring devices are distributed and deployed along each path in the mine to collect and process indicator data in real time. The indicator data includes at least vibration information that characterizes rock mass stability and gas concentration information that characterizes environmental safety, and generates a real-time risk level for each path. Personnel location tracking device, used to obtain and update the number of people entering the mining area and their location information in real time, and generate personnel location information; A dynamic path updating device is connected to the multi-source risk monitoring device and the path network modeling device to receive real-time risk level and index data of each path and dynamically update the three-dimensional topology model accordingly; The intelligent evacuation guidance device is in communication with the dynamic path updating device, the multi-source risk monitoring device, and the personnel positioning and tracking device, and is configured to: Real-time monitoring of the risk level output by multi-source risk monitoring devices; When the risk level of any path exceeds the preset safety threshold, an emergency evacuation command is triggered: Based on the currently updated 3D topological model and personnel location information, a personalized optimal evacuation path is calculated and generated for each person; Output evacuation information; Evacuation information includes: Mine-wide broadcast-level emergency evacuation alarm; Real-time dynamic directional instructions for each node, the directional instructions are generated according to the personalized optimal evacuation path and presented through a physical indicator device deployed at the node or a terminal device carried by personnel.

2. The intelligent mine safety monitoring system according to claim 1, characterized in that: Multi-source risk monitoring devices include: The vibration information monitoring module is distributed and deployed on each path in the mine to monitor the vibration information of each path and generate the first warning level based on the vibration information; Gas concentration monitors are distributed and deployed along each route in the mine to monitor gas concentration information along each route and generate a second warning level based on the gas concentration information, which includes both combustible and hazardous gas concentrations. The risk level generator receives the first warning level and the second warning level in real time, and takes the highest level of the first warning level and the second warning level as the risk level of the path.

3. The intelligent mine safety monitoring system according to claim 2, characterized in that: The vibration information monitoring module includes: Vibration sensor, used to obtain transverse vibration information P and longitudinal vibration information S of the rock wall; The vibration information processor receives the lateral vibration information P and the longitudinal vibration information S to generate the vibration index information h t , collect vibration index information h t Generate vibration index sequence H, h t Represents the vibration index information at time t; The vibration warning generator extracts the time series features of the vibration index sequence H and generates a first warning level based on the time series features.

4. The intelligent mine safety monitoring system according to claim 3, characterized in that: h t ={E, V, L}; E represents the vibration energy parameter, V represents the deformation parameter, and L represents the energy release parameter; ρ represents the rock density, v p，s represents the average velocity of the transverse vibration wave and the longitudinal vibration wave, R represents the distance from the crack source, and t s represents the duration of vibration, μ represents the displacement function of the vibration sensor, and ∫dt represents the integral operation over time; M represents the fracture energy parameter, which is used to measure the total mechanical energy released during the fault rupture process, U represents the rock mass shear stiffness, and E represents the vibration energy parameter; c represents the normalization constant, d represents the scaling exponent, M represents the crack energy parameter, and E represents the vibration energy parameter.

5. The intelligent mine safety monitoring system according to claim 4, characterized in that: The vibration warning generator includes: There are three information processing units, each for inputting E, V, and L in the vibration index sequence H to extract a first extraction feature, a second extraction feature, and a third extraction feature, respectively; The information fusion unit connects the first extracted features, the second extracted features, and the third extracted features, and performs global pooling to generate a temporal feature.

6. The intelligent mine safety monitoring system according to claim 5, characterized in that: The information processing unit includes: input layer, LSTM layer, normalization layer, and fully connected layer; The input layer is used to input a parameter in the vibration index information; LSTM layer, with 128 built-in LSTM cores. Each LSMT unit includes an input gate, a forget gate, a candidate memory unit, and an output gate. Input Gate: Forget gate: f t =σ(W if x t +b if +W hf h t-1 +b hf ); Candidate memory cells: Output gate: o t =σ(W io x t +b io +W ho h t-1 +b ho ); The update memory of each LSTM core is updated as follows: The hidden state output of each LSTM core is: h t =o t ⊙tanh(c t ); Where: σ is the sigmoid activation function, ⊙ represents element-by-element multiplication, λ is the L2 regularization coefficient, t represents the current time step, T represents the total length of the sequence, and x t represents the input vector at time t, n represents the feature dimension, h t-1 Indicates the hidden state of the previous moment, c t-1 Indicates the memory state of the previous moment, W ii 、W if 、W ig 、W io Represents x t The input gate weight matrix, forget gate weight matrix, candidate memory unit weight matrix and output gate weight matrix, i, f, g, o represent the input gate, forget gate, candidate memory unit and output gate respectively, W hi 、W hf 、W hg 、W ho They represent the input gate recurrent weight, forget gate recurrent weight, candidate memory unit recurrent weight, and output gate recurrent weight, respectively. hi 、b hf 、b hg 、b ho Represent the input gate bias term, forget gate bias term, candidate memory unit bias term, and output gate bias term respectively; i t =σ(·) represents the input gate vector, f t =σ(·) represents the forget gate vector, Represents the candidate memory unit vector tanh: represents the hyperbolic tangent function, o t =σ(·) represents the output gate vector; c t represents the output variable updated by the LSTM core, h t The feature vector representing the output of the output gate; Normalization layer, normalizes the feature vector of each LSTM layer to obtain a normalized vector μ t ,σ t are the mean and standard deviation of the time step t within the batch, γ represents the scaling parameter, β represents the offset parameter, and ∈ is a numerical stability constant; The fully connected layer normalizes the vector Connect and use ReLU as the activation function to output the extracted feature z: W fc1 is the output weight matrix, b fc1 is the output bias term.

7. The intelligent mine safety monitoring system according to claim 2, characterized in that: The gas concentration monitor generates a second warning level based on the growth rate of the combustible gas concentration and the harmful gas concentration.

8. The intelligent mine safety monitoring system according to claim 1, characterized in that: Personnel location tracking devices include: Entrance and exit people counter, used to obtain the number of people entering and leaving the mine, and generate the total number of people in the mine; There are multiple personnel monitors, which are arranged at the entrance and exit of each path, and are used to obtain the number of people entering the path and the number of people leaving the path to generate the total number of people in the path; The personnel distribution information generator generates the number of people in each path in the three-dimensional topological model and the personnel location information based on the total number of people in the mine and the total number of people in each path.

9. The intelligent mine safety monitoring system according to claim 1, characterized in that: Intelligent evacuation guidance device includes: Information collection module, which updates the 3D topology model and personnel location information in real time; An evacuation information generation unit creates an evacuation plan based on the 3D topology model and personnel location information. The evacuation plan includes the number of people passing through each node in different directions. There are multiple evacuation information indication modules, which are respectively arranged at each node, receive the evacuation plan in real time, and send the passing direction to the people passing through the node based on the evacuation plan; Broadcast module, used to send emergency evacuation alerts to the entire mining area; The evacuation information generation unit updates the evacuation plan based on the real-time updated three-dimensional topology model and personnel location information.

10. A smart mine safety monitoring method, characterized in that: The smart mine safety monitoring system according to any one of claims 1 to 9 is used to monitor mines.

Images