Coal mine dust diffusion intelligent monitoring method based on distributed sensor network

Abstract

The invention discloses a coal mine dust diffusion intelligent monitoring method based on a distributed sensing network, and relates to the technical field of coal mine dust diffusion intelligent monitoring, and the method comprises the following steps: deploying a plurality of sensing nodes in a coal mine, each node having a function of collecting environmental data, a distributed sensing network is formed through a wireless or wired communication module, and when all nodes are started, a unique node identifier is issued through a central controller, and self-inspection is carried out. Through bidirectional delay calculation and a progressive fine tuning strategy, high-precision time synchronization is achieved, and it is ensured that timestamps of data such as dust concentration and airflow are accurate and reliable. And data jump caused by large correction is avoided through gradual fine adjustment, data continuity is guaranteed, and the precision of dust diffusion analysis and an early warning model is improved. A redundant node management and automatic switching mechanism improves the fault processing capability of the system, avoids a monitoring blind area, accelerates maintenance response through fault information reporting, and ensures stable operation of a monitoring network.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring of coal mine dust diffusion, and in particular to an intelligent monitoring method of coal mine dust diffusion based on a distributed sensor network. Background Art

[0002] Intelligent monitoring of coal mine dust diffusion based on distributed sensor networks refers to the deployment of multiple interconnected sensor nodes in the mine to collect data such as dust concentration and airflow direction in the coal mine in real time, and transmit these data to the central control system. Each node of the distributed sensor network can operate independently and work together to ensure that even if an individual node fails, the overall system can still monitor effectively. Based on the Internet of Things technology and data analysis algorithms, the monitoring system can realize dynamic data collection and real-time processing, and analyze the trend and distribution of dust diffusion through cloud platforms or edge computing. The intelligent part is reflected in the fact that it can issue early warnings based on monitoring results, trigger alarms when dust concentration exceeds the standard, and propose corresponding emergency treatment measures. This monitoring system improves the safety of mines, reduces the risk of explosions and occupational diseases, and provides technical support for safe production in coal mines.

[0003] The prior art has the following deficiencies:

[0004] In the distributed sensor network for intelligent monitoring of dust diffusion in coal mines, if the time synchronization failure of the sensor node is not discovered in time, serious consequences may occur. Each node must synchronize time accurately to ensure that the dust concentration data accurately matches the information such as airflow and temperature on the time axis. However, network delays, communication failures, or node clock drift may cause time synchronization deviations. Timestamp errors can cause the system to misjudge the timing and diffusion path of dust concentration changes, which may miss the critical early warning window and cause high-concentration dust to fail to alarm in time. If mine operators rely on erroneous data to make judgments, ventilation measures may be delayed, increasing the risk of explosion or leading to occupational disease hazards, posing a serious threat to personnel safety and production operations.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The purpose of the present invention is to provide an intelligent monitoring method for coal mine dust diffusion based on a distributed sensor network. Through a two-way delay calculation method and a progressive fine-tuning strategy, the synchronization errors of network delays and clock drifts are effectively eliminated, ensuring the accurate consistency of the time of each sensor node and improving the reliability of the data. Progressive fine-tuning avoids time jumps and data mutations caused by large corrections, ensures the continuity of high-frequency acquisition data, and contributes to the accuracy of dust diffusion trend analysis and early warning models. In addition, the system adopts redundant node management and automatic switching mechanism. When a node is abnormal, the standby node can be enabled in time to take over the task to ensure uninterrupted monitoring, and improve the operation and maintenance efficiency through fault information reporting to solve the problems in the above-mentioned background technology.

[0007] In order to achieve the above object, the present invention provides the following technical solution: a method for intelligent monitoring of coal mine dust diffusion based on a distributed sensor network, comprising the following steps:

[0008] Multiple sensor nodes are deployed underground in coal mines. Each node has the function of collecting environmental data and forms a distributed sensor network through wireless or wired communication modules. When all nodes are started, a unique node identifier is issued through the central controller and self-checks are performed to ensure the normal operation of the hardware.

[0009] Each sensor node is synchronized with the central server through the network time protocol or time synchronization protocol. During the synchronization process, the communication delay is considered and corrected to ensure that the error of the timestamp is within the allowable range;

[0010] The system periodically detects node time deviations. When it detects that the clock drift of a node exceeds a set threshold, it immediately triggers the synchronization compensation mechanism.

[0011] Each sensor node collects dust concentration, airflow velocity and temperature data at a set frequency and marks the timestamp. When the data is uploaded to the central server, a dynamic screening algorithm is used to filter out data with abnormal timestamps or data that obviously does not conform to historical rules, thereby reducing misjudgments caused by synchronization problems.

[0012] The server uses a time series prediction algorithm to process uploaded data and predict the future dust diffusion path based on the changing trend of dust concentration and airflow direction. When the prediction results show that the dust concentration in a certain area exceeds the safety threshold in the future, an early warning is issued;

[0013] An adaptive time correction algorithm is used to regularly adjust the node clock deviation, and the accuracy of time synchronization is verified by the redundant node comparison method. When abnormalities are detected in the timestamps of multiple nodes, the abnormal nodes are automatically disabled and the backup nodes are started to ensure the continuity of monitoring and the reliability of data.

[0014] Preferably, the node identifier is used to uniquely identify each node in the distributed sensor network, and the node identifier includes a MAC address, an IP address, a serial number, or a device ID;

[0015] The MAC address is the physical address of the network interface, in a wireless or wired communication module;

[0016] IP addresses are used for dynamic network configuration;

[0017] The serial number or device ID is preset by the manufacturer and is used to distinguish between similar sensor devices.

[0018] Preferably, the specific steps of node clock drift detection and synchronization compensation are as follows:

[0019] The synchronization request is broadcasted from the central server to all sensor nodes through periodic tasks. After receiving the request, the node compares the current time with the reference time provided by the server and calculates the deviation value;

[0020] A time deviation threshold is preset in the system and the threshold is dynamically adjusted according to the actual needs of the mine environment;

[0021] When detecting that the node drift exceeds the threshold, the server immediately triggers the synchronization compensation mechanism;

[0022] After completing the time compensation, poll the data of the node and its surrounding redundant nodes again to ensure that the time synchronization compensation takes effect.

[0023] Preferably, the specific steps of filtering abnormal timestamps and historical data anomalies by dynamic screening algorithm are as follows:

[0024] Each sensor node collects data at a set frequency and adds a timestamp to each piece of data. When the data is uploaded to the central server, the timestamp of each record is first checked to ensure that it fits within the expected range of the upload time;

[0025] Compare the timestamp of the node with the upload time of the neighboring nodes to ensure that the time problem of a single node does not affect the overall judgment, and provide cross-validation for the node status;

[0026] A dynamic screening algorithm is used to further verify the uploaded data. Specifically, the server compares the current data with the historical data of the node over the past period of time to determine whether the changes are within the normal range;

[0027] When timestamp or data anomalies are detected, multi-level screening is automatically performed: data with slight anomalies will be marked as "to be observed", and the system will retain it and wait for further verification with future data; data with significant anomalies will be directly discarded and error logging will be triggered.

[0028] Preferably, the specific steps of coal mine dust diffusion path prediction and early warning based on ARIMA or LSTM model are as follows:

[0029] The dust concentration C uploaded by each node t , air velocity V t and airflow direction θ t The data is cleaned and normalized, and all data is processed using a sliding window mechanism so that the model can capture historical change patterns. The input feature matrix is ​​defined as follows: C t-k 、V t-k represents the dust concentration and airflow velocity of the first k steps, μ C , σ C represents the mean and standard deviation of dust concentration, which is used for standardization; the mean and standard deviation of air flow velocity are μ V , σ V , cos(θ t-k )、sin(θ t-k ) represents the vectorization of the airflow direction, which is convenient for capturing periodic changes. n is the size of the sliding window, which determines how much historical information the model captures;

[0030] Based on historical data, an LSTM network is constructed to capture the time dependency of dust diffusion. The core formula of LSTM is as follows: t =σ(W f ·[h t-1 , X t ]+b f ),i t =σ(W i ·[h t-1 , X t ]+b i ), h t =σ(W h ·[C t ,h t-1 ]+b h ), In the formula, f t is the forget gate, which is used to decide whether to keep the state of the previous moment. t is the input gate, which is used to control the degree of update of the cell state by the current input. is the candidate cell state at the current moment, C t is the final cell state, combining the past and current input information, h t is a hidden state used to capture long-term and short-term dependencies, σ is the Sigmoid activation function, and tanh is the hyperbolic tangent activation function.

[0031] Preferably, using the trained LSTM model, the system inputs the data X in the current time window t , predict the dust concentration after the future time step n, the prediction expression is: In order to further predict the dust diffusion path, the system combines the air flow speed and direction to calculate the dust distribution at the future moment. The calculation expression is: In the formula, is the predicted future dust concentration, V t+n is the future airflow velocity, cos(θ t+n -θ t ) is the angle between the future and current airflow directions, which is used to calculate the effective component of the dust propagation direction;

[0032] The predicted dust dispersion With the set safety threshold P safe For comparison, if An alarm is sounded to notify mine managers to take ventilation or evacuation measures.

[0033] Preferably, an adaptive time correction algorithm is used to regularly adjust the node clock deviation, and the accuracy of time synchronization is verified by a redundant node comparison method. The specific steps are as follows:

[0034] The system regularly broadcasts time synchronization signals to all nodes and collects the time information returned by each node. In order to eliminate the impact of network delay, a two-way delay calculation method is used, that is, the total time taken for the signal to go from the server to the node and back is measured, and half of it is included in the correction error;

[0035] After detecting that a node has time deviation, two time correction methods are implemented: immediate correction and gradual fine-tuning. Instant correction is used in the case of large drift. The server immediately sends the precise time to the node and forces synchronization. For small deviations, the system uses gradual fine-tuning to gradually adjust the node's clock.

[0036] Redundant nodes in each monitoring area are selected for comparative analysis. The timestamps of each node are verified against each other, and the server calculates the time difference between these nodes. When the timestamps of multiple nodes are consistent and the drift is within the threshold, the synchronization is considered successful. If the timestamp of a node is not within the threshold when compared with other redundant nodes, the node is marked as abnormal and the next step of processing is triggered.

[0037] When it is detected that the timestamp of a node is continuously abnormal and cannot meet the standard even after multiple corrections, the node will be marked as unavailable and its data upload function will be automatically disabled. At the same time, a backup node in the same monitoring area will be enabled to take over the monitoring task of the disabled node to ensure uninterrupted data monitoring.

[0038] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0039] The present invention can effectively eliminate the synchronization errors caused by network delays and clock drifts by adopting a two-way delay calculation method and a progressive fine-tuning strategy, so that the time of all sensor nodes is kept within an accurate range. This ensures that the timestamps of environmental data such as dust concentration, airflow velocity and temperature are consistent with the actual acquisition time, avoiding data dislocation and misjudgment caused by inaccurate time. In addition, the adaptive time correction algorithm can dynamically adjust the synchronization frequency according to historical drift trends, so that the system can maintain high-precision time synchronization during long-term operation. This precise time management improves the reliability of monitoring data and provides a solid data foundation for dust diffusion path analysis and real-time early warning.

[0040] The present invention avoids data jumps and sudden anomalies caused by large time corrections through gradual fine-tuning. The system gradually approaches the target time through multiple rounds of small adjustments, thereby ensuring the order and continuity of data upload. It is particularly suitable for dust concentration monitoring with high frequency collection, avoiding the analysis difficulties caused by data breaks or mutations. Continuous data sequences not only help the monitoring system to more accurately analyze the trend of dust diffusion, but also improve the accuracy of the early warning model, providing a more reliable reference for timely adoption of safety measures such as ventilation.

[0041] The present invention enables the system to have excellent fault handling capabilities through redundant node management and automatic switching mechanism. When the timestamp or data of a node is continuously abnormal, the system can automatically disable the faulty node and quickly enable the spare node to take over the monitoring task to ensure the continuous operation of the monitoring network. This redundant management avoids the monitoring blind spots or data loss that may be caused by single point failures. In addition, the system can also report fault information in a timely manner, helping maintenance personnel to quickly locate problems and improve maintenance efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0043] Figure 1 The present invention is a method flow chart of a method for intelligent monitoring of coal mine dust diffusion based on a distributed sensor network. DETAILED DESCRIPTION

[0044] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0045] The present invention provides Figure 1 The method for intelligent monitoring of coal mine dust diffusion based on a distributed sensor network includes the following steps:

[0046] Multiple sensor nodes are deployed underground in coal mines. Each node has the function of collecting environmental data such as dust concentration, airflow direction, and temperature, and forms a distributed sensor network through wireless or wired communication modules. When all nodes are started, a unique node identifier is issued through the central controller, and self-checks are performed to ensure the normal operation of the hardware;

[0047] Node identification is used to uniquely identify each node in a distributed sensor network. Common node identifications include MAC address, IP address, serial number, or device ID. MAC address is the physical address of the network interface and is commonly used in wireless or wired communication modules. IP address is suitable for dynamic network configuration, which is convenient for remote access and data transmission management. Serial number or device ID is usually preset by the manufacturer to distinguish similar sensor devices and bind to specific nodes when the system is initialized. Node identification should be unique and stable to avoid duplication to ensure the accuracy of data source and the effectiveness of subsequent node management. In addition, the hash value generated by combining MAC address and device ID is also a security-enhancing identification method to prevent malicious nodes from impersonating or data tampering.

[0048] Each sensor node synchronizes time with the central server through the Network Time Protocol (NTP) or a dedicated time synchronization protocol. During the synchronization process, communication delays are taken into account and corrected to ensure that the timestamp error is within the allowable range. The system periodically detects node time deviations. When it detects that the clock drift of a node exceeds the set threshold, the synchronization compensation mechanism is immediately triggered;

[0049] The specific steps of node clock drift detection and synchronization compensation are as follows:

[0050] The system broadcasts synchronization requests from the central server to all sensor nodes through periodic tasks (such as every 10 minutes or 1 hour). After receiving the request, the node compares its current time with the reference time provided by the server and calculates the deviation value. To prevent network delays from affecting the detection results, the system records the sending and returning time of the synchronization signal, uses polling time synchronization or a two-way delay calibration algorithm to calculate the network delay and remove it from the time difference. This timing detection mechanism ensures that the system can detect potential clock drift trends in advance.

[0051] A time deviation threshold (such as ±5 milliseconds or 1 second) is preset in the system, and the threshold is dynamically adjusted according to the actual needs of the mine environment. The server compares the time difference of each node with the set threshold one by one. If the deviation of all nodes is within the allowable range, the synchronization is considered normal and no further operation is required. If the drift of a node exceeds the threshold, the system records the identification and drift amount of the abnormal node for subsequent processing. The setting of the dynamic threshold can be adjusted according to the ambient temperature or the frequency of communication to improve the adaptability of the system.

[0052] When the node drift is detected to exceed the threshold, the server immediately triggers the synchronization compensation mechanism. The compensation method can be direct time correction or incremental adjustment:

[0053] In direct time correction, the server sends the calibration time to the abnormal node, requiring the node to immediately update its internal clock to the accurate time.

[0054] In the incremental adjustment method, in order to avoid drastic time jumps affecting data continuity, the system gradually approaches the node time to the correct time (such as adjusting 10 milliseconds each time). This method is particularly suitable for monitoring systems with high continuity requirements.

[0055] After completing the time compensation, the system will poll the data of the node and its surrounding redundant nodes again to ensure that the compensation of time synchronization is effective. If the node still has drift problems after multiple compensations, the system will record the abnormality and mark it as an unstable node. In addition, the system will enable the redundant node to replace the monitoring responsibilities of the node to avoid single point failures affecting the overall monitoring. This process ensures the accuracy of the data and the stability of the monitoring system, and avoids misjudgment of dust diffusion data due to time drift.

[0056] Each sensor node collects dust concentration, air flow velocity and temperature data at a set frequency and marks the timestamp. When the data is uploaded to the central server, a dynamic screening algorithm is used to filter out data with abnormal timestamps or data that obviously does not conform to historical rules, thereby reducing misjudgments caused by synchronization problems;

[0057] The specific steps to filter abnormal timestamps and historical data anomalies through dynamic filtering algorithms are as follows:

[0058] Each sensor node collects data such as dust concentration, airflow velocity, and temperature at a set frequency, and adds a timestamp to each piece of data. When the data is uploaded to the central server, the system first checks the timestamp of each record to ensure that it fits within the expected range of the upload time. For example, if a node should upload data every 10 seconds, but the upload interval is greater than the specified range, or the timestamp appears in the future (beyond the server time), the data will be preliminarily marked as abnormal. This preliminary check is the first step to ensure that the basic time synchronization is correct.

[0059] To further determine the credibility of the data, the system will compare the timestamp of the node with the upload time of the neighboring nodes. For example, if multiple nodes in a certain area are monitored simultaneously, when the timestamp of one node deviates from other nodes by more than a certain threshold (such as an offset within 5 seconds), the system will determine that the timestamp is abnormal. This process uses the redundancy of the distributed network to ensure that the time problem of a single node does not affect the overall judgment, while providing cross-validation for the node status.

[0060] The system uses a dynamic screening algorithm to further verify the uploaded data, especially the dust concentration, airflow velocity and temperature data. Specifically, the server compares the current data with the historical data of the node in the past period of time (such as 24 hours or a week) to determine whether the change is within a reasonable range. For example, if the dust concentration recorded by a node suddenly jumps by more than 50 times, and other adjacent nodes do not have similar situations, the system will mark it as historical abnormal data. The screening algorithm usually combines the sliding window mechanism and statistical models (such as mean, standard deviation) to analyze data trends in order to eliminate outliers.

[0061] When timestamp or data anomalies are detected, the system automatically performs multi-level screening: slightly abnormal data will be marked as "to be observed", and the system will retain it and wait for future data to further verify; while significantly abnormal data will be directly discarded and trigger error log records. In addition, the system also has a data correction mechanism: if the timestamp has only a slight deviation (such as an error within a few milliseconds), the system can correct it based on the time and trend data of nearby nodes to ensure the continuity and consistency of the data. Through this multi-level screening and correction, the system greatly reduces the probability of misjudgment caused by synchronization problems and provides reliable data support for dust diffusion monitoring.

[0062] The server uses a time series prediction algorithm (such as ARIMA or LSTM model) to process the uploaded data and predict the future dust diffusion path based on the changing trend of dust concentration and airflow direction. When the prediction results show that the dust concentration in a certain area may exceed the safety threshold in the future, an early warning will be issued;

[0063] The specific steps of coal mine dust diffusion path prediction and early warning based on ARIMA or LSTM model are as follows:

[0064] In order to ensure the quality of the data input to the model, the system collects the dust concentration (C t ), air flow velocity (V t ) and airflow direction (θ t ) data is cleaned and normalized. All data are processed using a sliding window mechanism, allowing the model to capture historical change patterns. The input feature matrix is ​​defined as follows: C t-k 、V t-k represents the dust concentration and airflow velocity of the first k steps, μ C , σ C It represents the mean and standard deviation of dust concentration, which is used for standardization; the mean and standard deviation of air flow velocity are μ V , σ V , cos(θ t-k )、sin(θ t-k ) represents the vectorization of the airflow direction, which is convenient for capturing periodic changes. n is the size of the sliding window, which determines how much historical information (such as 24-hour data) the model captures.

[0065] These features constitute the input sequence of the model and ensure that the data dimension is consistent to fit the LSTM model.

[0066] The system builds an LSTM (Long Short-Term Memory) network based on historical data to capture the time dependency of dust diffusion. The LSTM core formula is as follows: t =σ(W f ·[h t-1 , X t ]+b f ),i t =σ(W i ·[h t-1 , X t ]+b i ), h t =σ(W h ·[C t ,h t-1 ]+b h ), In the formula, f t is the forget gate, which is used to decide whether to keep the state of the previous moment. t is the input gate, which is used to control the degree of update of the cell state by the current input. is the candidate cell state at the current moment, C t is the final cell state, combining the past and current input information, h tis the hidden state, used to capture long-term and short-term dependencies, σ is the Sigmoid activation function, and tanh is the hyperbolic tangent activation function;

[0067] The LSTM model adjusts the parameter matrix (such as W) through the back propagation algorithm. f , W i , W c ) to minimize the prediction error. After training, the model can predict the dust concentration and airflow change trend in the future.

[0068] Using the trained LSTM model, the system inputs the data X in the current time window t , predict the dust concentration after the future time step n, the prediction expression is: In order to further predict the dust diffusion path, the system combines the air flow speed and direction to calculate the dust distribution at the future moment. The calculation expression is: In the formula, is the predicted future dust concentration, V t+n is the future airflow velocity, cos(θ t+n -θ t ) is the angle between the future and current airflow directions, which is used to calculate the effective component of the dust propagation direction.

[0069] The system will predict the amount of dust dispersion With the set safety threshold P safe For comparison, if An alarm is sounded;

[0070] If the prediction results show that the dust concentration will exceed the safety threshold at some point in the future, the system will immediately issue an early warning to notify mine managers to take ventilation or evacuation measures. Through the rolling prediction mechanism, the system can update the prediction results in real time and provide accurate decision support for mine safety.

[0071] In order to deal with the problem of node clock drift or network delay in long-term operation, the system uses an adaptive time correction algorithm to regularly adjust the node clock deviation, and verifies the accuracy of time synchronization through redundant node comparison method. When abnormal timestamps of multiple nodes are detected, the abnormal nodes are automatically disabled and the backup nodes are started to ensure the continuity of monitoring and the reliability of data;

[0072] Adopt the adaptive time correction algorithm to regularly adjust the node clock deviation, and verify the accuracy of time synchronization through the redundant node comparison method. The specific steps are as follows:

[0073] The system broadcasts time synchronization signals to all nodes regularly and collects the time information returned by each node. In order to eliminate the impact of network delay, the system uses a two-way delay calculation method, which measures the total time it takes for the signal to go from the server to the node and back, and counts half of it as the correction error. In addition, the system applies an adaptive model (such as Kalman filtering) to predict future deviations based on the historical clock deviation trend of the node. This adaptive detection can not only identify the current time deviation, but also predict the clock drift trend in long-term operation to prevent the accumulation of synchronization problems.

[0074] After detecting that a node has time deviation, the system performs two time correction methods: immediate correction and gradual fine-tuning. Instant correction is used for large drifts (such as more than a few seconds). The server immediately sends the precise time to the node and forces synchronization. For small deviations (such as milliseconds), the system uses gradual fine-tuning to gradually adjust the node's clock to avoid the impact of data jumps caused by large corrections. Through this flexible correction strategy, the system can ensure accurate clock synchronization while maintaining data continuity.

[0075] To verify the accuracy of time synchronization, the system selects redundant nodes in each monitoring area for comparative analysis. The timestamps of each node are checked against each other, and the server calculates the time difference between these nodes. When the timestamps of multiple nodes are consistent and the drift is within the threshold, the system considers synchronization successful. If the timestamp of a node deviates too much from that of other redundant nodes, the system marks the node as abnormal and triggers the next step of processing. This redundant node comparison method ensures that even if a clock problem occurs on a node, the time synchronization of the overall monitoring remains consistent.

[0076] When it is detected that the timestamp of a node is continuously abnormal and still cannot meet the standard even after multiple corrections, the system will mark the node as unavailable and automatically disable its data upload function. At the same time, the system will enable the backup node in the same monitoring area to take over the monitoring task of the disabled node to ensure uninterrupted data monitoring. At this time, the status information of the disabled node will be sent to the management platform to prompt maintenance personnel to check and repair it. Through this automated node management process, the system can respond quickly when faced with node failures or drifts, ensuring the continuity of the monitoring system and the reliability of the data.

[0077] Implementation method 1: Bidirectional delay calculation to correct clock deviation

[0078] In a distributed sensor network, the time synchronization between nodes and servers is often affected by network delays, especially in complex coal mine environments, where the transmission of network signals is prone to fluctuations due to obstacles, wireless interference, and other factors. If these delay factors are not considered, the system clock deviation will accumulate, resulting in inaccurate timestamps for data such as dust concentration, airflow, and temperature, making the monitoring system lose its timeliness. To this end, this implementation method uses a two-way delay calculation method to ensure the accuracy of time synchronization.

[0079] Specifically, when the server sends a synchronization signal to the node, the initial time T1T_1T1 is recorded. The node records the current time T2T_2T2 at the moment of receiving the signal, and returns the time together with the synchronization response signal to the server. When the server receives the response returned by the node, it records the final time T3T_3T3. By calculating the formula ΔT=(T3-T1) / 2\Delta T=(T_3-T_1) / 2ΔT=(T3-T1) / 2, the system can estimate the average delay of the network. This process effectively eliminates the influence of external factors such as network congestion, making the synchronization result more accurate.

[0080] In addition, this method can be further combined with a dynamic delay adjustment mechanism. During long-term operation, the system will analyze the delay data of each round of synchronization, calculate the fluctuation range of network delay, and adjust the synchronization period. For example, if the network delay changes frequently during a period of time, the system will shorten the synchronization interval and increase the synchronization frequency to ensure the accuracy of the clock. In special cases, such as when the network delay is too large, the system can also automatically determine whether to retransmit to avoid synchronization failure.

[0081] Through this two-way delay calculation, not only higher synchronization accuracy is achieved, but also the communication robustness in the complex environment of coal mines is improved, the timestamp of each node data is guaranteed to be accurate, and the data foundation is laid for subsequent dust diffusion monitoring.

[0082] Implementation method 2: Gradual fine-tuning to ensure data continuity

[0083] In the coal mine monitoring system, the accuracy of clock correction directly affects the reliability of data. However, if the node clock is adjusted significantly during the time synchronization process, it is easy to cause time jumps in the monitoring data (that is, the timestamp of the data sequence changes suddenly), which will cause data missing or data order confusion. This situation is particularly unfavorable for dynamically changing dust concentration monitoring. Therefore, this implementation adopts a gradual fine-tuning strategy to ensure data continuity.

[0084] The core of gradual fine-tuning is to achieve gradual synchronization of time through small-step adjustments. For example, when the system detects that the clock of a certain node deviates from the server time by 100 milliseconds, it will not correct to the target time all at once, but will only adjust 5 milliseconds each time and gradually approach the correct time during multiple rounds of synchronization. This method avoids false alarms caused by sudden changes in the monitoring system and ensures that the order and trend of the data are not affected. In addition, this small adjustment can effectively reduce the risk of sensor data loss. After multiple rounds of synchronization, the time of each node can be kept consistent within the millisecond range.

[0085] To further improve the correction efficiency, the system will formulate a dynamic adjustment plan based on the historical drift data of the node. For example, some nodes may have frequent clock drift due to high usage frequency or low battery. The system will record the amplitude and interval of each correction, and use adaptive algorithms (such as Kalman filter) to optimize the future fine-tuning frequency to avoid unnecessary adjustments. In this way, the system can minimize the consumption of communication resources while ensuring time synchronization.

[0086] The gradual fine-tuning strategy improves the clock synchronization accuracy while maintaining the continuity of monitoring data. Even under high-frequency sampling conditions, the data of each node can be uploaded in a reasonable time series, providing reliable guarantee for trend analysis and real-time warning of dust diffusion.

[0087] Implementation method 3: Redundant node switching and faulty node isolation

[0088] In a distributed sensor network, any abnormality of a node will affect the monitoring results. Therefore, this implementation ensures the stability of the system by deploying redundant nodes, and when a node abnormality is detected, the backup node is switched in time to achieve uninterrupted monitoring tasks.

[0089] In the system, each important monitoring area is equipped with several redundant nodes. These nodes are usually in standby mode with low power consumption and are activated only when other nodes are detected to be abnormal or faulty. When the system finds that the timestamp drift of a node exceeds the allowable range and cannot be restored after multiple corrections, the system will automatically disable the node and mark it as a faulty node. At this time, the redundant node will start immediately to take over the monitoring task to ensure continuous data upload. The activation process of the redundant node is automatically completed by the server without manual intervention, thereby improving the response speed of the system.

[0090] At the same time, in order to further ensure the reliability of monitoring data, the system adopts a dynamic redundant switching strategy. During the switching process, the server will verify whether the clock of the backup node has been synchronized and whether its data collection capability is normal. Once it is confirmed that the backup node is in an available state, the system will officially complete the switch. During this process, the server will also report detailed information of the faulty node to the management platform, including node identification, fault type and fault time, to remind the operation and maintenance personnel to conduct inspections and repairs.

[0091] In addition, the system also supports an automated redundant node health check mechanism. Under normal operating conditions, the standby node will regularly communicate with the server to detect whether its status is normal. Once a problem is found in the redundant node itself, the system will issue a warning in advance to avoid the standby node being unavailable at a critical moment. Through this multi-level redundancy management and fault isolation mechanism, the system can ensure that even when an abnormality occurs in an individual node, the entire monitoring network can still operate stably without monitoring blind spots or data loss.

[0092] By adopting a two-way delay calculation method and a progressive fine-tuning strategy, the system can effectively eliminate the synchronization errors caused by network delays and clock drift, so that the time of all sensor nodes is kept within an accurate range. This ensures that the timestamps of environmental data such as dust concentration, airflow velocity and temperature are consistent with the actual collection time, avoiding data dislocation and misjudgment caused by inaccurate time. In addition, the adaptive time correction algorithm can dynamically adjust the synchronization frequency according to historical drift trends, so that the system can maintain high-precision time synchronization during long-term operation. This precise time management improves the reliability of monitoring data and provides a solid data foundation for dust diffusion path analysis and real-time early warning.

[0093] The present invention avoids data jumps and sudden anomalies caused by large time corrections through gradual fine-tuning. For example, even if the clock drift of the monitoring node has a large error, the system will gradually approach the target time through multiple rounds of small adjustments, thereby ensuring the order and continuity of data upload. This strategy is particularly suitable for high-frequency dust concentration monitoring, avoiding analysis difficulties caused by data breaks or mutations. Continuous data sequences not only help the monitoring system to more accurately analyze the trend of dust diffusion, but also improve the accuracy of the early warning model, providing a more reliable reference for timely adoption of safety measures such as ventilation.

[0094] The present invention uses redundant node management and automatic switching mechanisms, and the system has excellent fault handling capabilities. When the timestamp or data of a node is continuously abnormal, the system can automatically disable the faulty node and quickly enable the backup node to take over the monitoring task to ensure the continuous operation of the monitoring network. This redundant management avoids the monitoring blind spots or data loss that may be caused by single point failures. In addition, the system can also report fault information in a timely manner, helping maintenance personnel to quickly locate problems and improve maintenance efficiency.

[0095] Overall, this solution enhances the stability and adaptability of the monitoring system in complex mining environments, providing a higher level of protection for safe production in coal mines.

[0096] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. An intelligent monitoring method for coal mine dust diffusion based on a distributed sensor network, characterized in that: The following steps are involved: Multiple sensor nodes are deployed underground in coal mines. Each node has the function of collecting environmental data and forms a distributed sensor network through wireless or wired communication modules. When all nodes are started, a unique node identifier is issued through the central controller and self-checks are performed to ensure the normal operation of the hardware. Each sensor node is synchronized with the central server through the network time protocol or time synchronization protocol. During the synchronization process, the communication delay is considered and corrected to ensure that the error of the timestamp is within the allowable range; The system periodically detects node time deviations. When it detects that the clock drift of a node exceeds a set threshold, it immediately triggers the synchronization compensation mechanism. Each sensor node collects dust concentration, airflow velocity and temperature data at a set frequency and marks the timestamp. When the data is uploaded to the central server, a dynamic screening algorithm is used to filter out data with abnormal timestamps or data that obviously does not conform to historical rules, thereby reducing misjudgments caused by synchronization problems. The server uses a time series prediction algorithm to process uploaded data and predict the future dust diffusion path based on the changing trend of dust concentration and airflow direction. When the prediction results show that the dust concentration in a certain area exceeds the safety threshold in the future, an early warning is issued; An adaptive time correction algorithm is used to regularly adjust the node clock deviation, and the accuracy of time synchronization is verified by the redundant node comparison method. When abnormalities are detected in the timestamps of multiple nodes, the abnormal nodes are automatically disabled and the backup nodes are started to ensure the continuity of monitoring and the reliability of data.

2. According to claim 1, a method for intelligent monitoring of coal mine dust diffusion based on a distributed sensor network is characterized in that: The node identifier is used to uniquely identify each node in the distributed sensor network. The node identifier includes a MAC address, an IP address, a serial number, or a device ID; The MAC address is the physical address of the network interface, in a wireless or wired communication module; IP addresses are used for dynamic network configuration; The serial number or device ID is preset by the manufacturer and is used to distinguish between similar sensor devices.

3. The method for intelligent monitoring of coal mine dust diffusion based on a distributed sensor network according to claim 1 is characterized in that: The specific steps of node clock drift detection and synchronization compensation are as follows: The synchronization request is broadcasted from the central server to all sensor nodes through periodic tasks. After receiving the request, the node compares the current time with the reference time provided by the server and calculates the deviation value; A time deviation threshold is preset in the system and the threshold is dynamically adjusted according to the actual needs of the mine environment; When detecting that the node drift exceeds the threshold, the server immediately triggers the synchronization compensation mechanism; After completing the time compensation, poll the data of the node and its surrounding redundant nodes again to ensure that the time synchronization compensation takes effect.

4. The method for intelligent monitoring of coal mine dust diffusion based on a distributed sensor network according to claim 1 is characterized in that: The specific steps to filter abnormal timestamps and historical data anomalies through dynamic filtering algorithms are as follows: Each sensor node collects data at a set frequency and adds a timestamp to each piece of data. When the data is uploaded to the central server, the timestamp of each record is first checked to ensure that it fits within the expected range of the upload time; Compare the timestamp of the node with the upload time of the neighboring nodes to ensure that the time problem of a single node does not affect the overall judgment, and provide cross-validation for the node status; A dynamic screening algorithm is used to further verify the uploaded data. Specifically, the server compares the current data with the historical data of the node over the past period of time to determine whether the changes are within the normal range; When timestamp or data anomalies are detected, multi-level screening is automatically performed: data with slight anomalies will be marked as "to be observed", and the system will retain it and wait for further verification with future data; data with significant anomalies will be directly discarded and error logging will be triggered.

5. The method for intelligent monitoring of coal mine dust diffusion based on a distributed sensor network according to claim 1 is characterized in that: The specific steps of coal mine dust diffusion path prediction and early warning based on ARIMA or LSTM model are as follows: The dust concentration C uploaded by each node t , air velocity V t and airflow direction θ t The data is cleaned and normalized, and all data are processed using a sliding window mechanism so that the model can capture historical change patterns. The input feature matrix is ​​defined as follows: C t-k 、V t-k represents the dust concentration and airflow velocity of the first k steps, μ C , σ C It represents the mean and standard deviation of dust concentration, which is used for standardization; the mean and standard deviation of air flow velocity are μ V , σ V , cos(θ t-k )、sin(θ t-k ) represents the vectorization of the airflow direction, which is convenient for capturing periodic changes. n is the size of the sliding window, which determines how much historical information the model captures; Based on historical data, an LSTM network is constructed to capture the time dependency of dust diffusion. The core formula of LSTM is as follows: t =σ(W f ·[h t-1 , X t ]+b f ),i t =σ(W i ·[h t-1 , X t ]+b i ), h t =σ(W h ·[C t ,h t-1 ]+b h ), In the formula, f t is the forget gate, which is used to decide whether to keep the state of the previous moment. t is the input gate, which is used to control the degree of update of the cell state by the current input. is the candidate cell state at the current moment, C t is the final cell state, combining the past and current input information, h t is a hidden state used to capture long-term and short-term dependencies, σ is the Sigmoid activation function, and tanh is the hyperbolic tangent activation function.

6. The method for intelligent monitoring of coal mine dust diffusion based on a distributed sensor network according to claim 5 is characterized in that: Using the trained LSTM model, the system inputs the data X in the current time window t , predict the dust concentration after the future time step n, the prediction expression is: In order to further predict the dust diffusion path, the system combines the air flow speed and direction to calculate the dust distribution at future times. The calculation expression is: ×cos(θ t+n -θ t ), where is the predicted future dust concentration, V t+n is the future airflow velocity, cos(θ t+n -θ t ) is the angle between the future and current airflow directions, which is used to calculate the effective component of the dust propagation direction; The predicted dust dispersion With the set safety threshold P safe For comparison, if >P safe , an alarm is issued to notify mine managers to take ventilation or evacuation measures.

7. The method for intelligent monitoring of coal mine dust diffusion based on a distributed sensor network according to claim 1 is characterized in that: Adopt the adaptive time correction algorithm to regularly adjust the node clock deviation, and verify the accuracy of time synchronization through the redundant node comparison method. The specific steps are as follows: The system regularly broadcasts time synchronization signals to all nodes and collects the time information returned by each node. In order to eliminate the impact of network delay, a two-way delay calculation method is used, that is, the total time taken for the signal to go from the server to the node and back is measured, and half of it is included in the correction error; After detecting that a node has time deviation, two time correction methods are implemented: immediate correction and gradual fine-tuning. Instant correction is used in the case of large drift. The server immediately sends the precise time to the node and forces synchronization. For small deviations, the system uses gradual fine-tuning to gradually adjust the node's clock. Redundant nodes in each monitoring area are selected for comparative analysis. The timestamps of each node are verified against each other, and the server calculates the time difference between these nodes. When the timestamps of multiple nodes are consistent and the drift is within the threshold, the synchronization is considered successful. If the timestamp of a node is not within the threshold when compared with other redundant nodes, the node is marked as abnormal and the next step of processing is triggered. When it is detected that the timestamp of a node is continuously abnormal and cannot meet the standard even after multiple corrections, the node will be marked as unavailable and its data upload function will be automatically disabled. At the same time, a backup node in the same monitoring area will be enabled to take over the monitoring task of the disabled node to ensure uninterrupted data monitoring.

Images