Coal mine dust concentration real-time monitoring method based on big data analysis

Abstract

The invention discloses a coal mine dust concentration real-time monitoring method based on big data analysis, and relates to the technical field of coal mine dust concentration monitoring, and the method comprises the following steps: obtaining various parameter information generated in the data fusion process of a big data system in real time, including transient response information and time synchronization information, and analyzing and processing the obtained transient response information and the time synchronization information. According to the method, the transient deviation index and the phase error index are introduced, the data fusion process is intelligently evaluated in combination with the machine learning model, the error problem caused by the sensor response time difference is solved, and multi-level intelligent sensing is achieved. The system accurately judges the risk level, reduces false alarm and missing alarm, takes maintenance, adjustment or emergency intervention measures corresponding to different risks through a grading early warning mechanism, remarkably improves the coal mine monitoring safety, and reduces the accident risk.

Description

Technical Field

[0001] The present invention relates to the technical field of coal mine dust concentration monitoring, and in particular to a real-time monitoring method for coal mine dust concentration based on big data analysis. Background Art

[0002] Real-time monitoring of dust concentration in coal mines based on big data analysis refers to the use of big data technology and sensor equipment to continuously and automatically monitor dust concentration in coal mine environments, and to achieve accurate early warning and optimized management through data analysis. In coal mine work scenarios, changes in dust concentration will directly affect the health and safety of workers, and may also cause serious accidents such as explosions. The real-time monitoring system collects data such as dust concentration, temperature and humidity, and wind speed in the environment through various sensors arranged in the mining area, generating a large amount of dynamic information. Big data analysis processes, screens, and predicts trends of these data to help managers detect abnormal concentrations or potential risks in a timely manner.

[0003] The core of this technology is to use the computing power of the big data platform to conduct a comprehensive analysis of historical data and real-time data, and use machine learning algorithms to predict the changing trend of dust concentration. For example, the system can detect a sudden increase in dust concentration in a certain area, and automatically determine whether there are hidden dangers based on historical data and external environmental conditions, so as to issue an alarm in advance and guide relevant personnel to ventilate or stop work for inspection. This technical means can not only improve the safety and efficiency of coal mine production, but also reduce the lag and errors caused by traditional manual detection, providing scientific support for the safety management of coal mining enterprises.

[0004] The prior art has the following deficiencies:

[0005] To ensure comprehensive monitoring, coal mines usually deploy multiple sensors at different locations. However, due to the different response speeds of each sensor, some devices can respond quickly to environmental changes, while others have a certain delay. This time difference may cause errors in the data fusion process of the big data system. If the sensor data cannot be synchronized, the system may not be able to capture the dangerous peak of dust concentration in time. For example, some sensors may have detected that the concentration exceeds the critical value, but the delayed data has not been updated, causing the system to misjudge that the concentration is still within the safe range. When the combustible dust in the coal mine mixes with the oxygen in the air to reach the explosion limit, if timely warning and measures are not taken, it may cause a dust explosion, causing equipment damage and serious casualties.

[0006] 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

[0007] The purpose of the present invention is to provide a real-time monitoring method for coal mine dust concentration based on big data analysis. By introducing transient deviation index and phase error index as key indicators, and combining machine learning models to intelligently evaluate the data fusion process, the scheme effectively solves the error problem caused by differences in sensor response time, and realizes the transformation from simple alarm to multi-level intelligent perception. The system can accurately judge the risk level, reduce false alarms and missed alarms, and ensure the real-time and accuracy of data processing. Based on the multi-level early warning mechanism, the system prompts daily maintenance through the first-level early warning when the risk is low, and timely adjusts the second-level early warning when the risk is medium to prevent the problem from escalating. The third-level early warning quickly initiates emergency intervention and evacuation measures when the risk is high. This hierarchical response system greatly improves the safety of coal mine dust monitoring and reduces the risk of accidents to solve the problems in the above-mentioned background technology.

[0008] In order to achieve the above object, the present invention provides the following technical solution: a real-time monitoring method for coal mine dust concentration based on big data analysis, comprising the following steps:

[0009] Acquire various parameter information generated in the data fusion process of the big data system in real time, including transient response information and time synchronization information, and analyze and process the acquired transient response information and time synchronization information;

[0010] The analyzed and processed transient response information and time synchronization information are input into the pre-learned machine learning model, and the abnormal fusion index is generated through the machine learning model. The abnormal fusion index is used to intelligently evaluate the data fusion process of the big data system.

[0011] Based on the evaluation results of the machine learning model, the data fusion process of the big data system is divided into a normal fusion process or an abnormal fusion process, and the abnormal state of data fusion in the big data system is intelligently perceived;

[0012] For the abnormal fusion process, continuously obtain the data fusion parameter information of the big data system, and further divide the abnormal fusion process into low-risk fusion process, medium-risk fusion process and high-risk fusion process;

[0013] Different countermeasures should be taken for different levels of integration process.

[0014] Preferably, the transient response information generated during the data fusion process of the big data system includes a transient deviation index, and the time synchronization information generated during the data fusion process of the big data system includes a phase error index. After acquisition, the transient deviation index and the phase error index are calibrated as Transient DI and Phase EI .

[0015] Preferably, the logic for obtaining the transient deviation index is as follows:

[0016] Under the detection window, multiple sensors arranged at different locations in the coal mine collect dust concentration data, and the dust concentration data output by each sensor is calibrated as X i (t), where: X i (t)={X1(t)、X2(t)、X3(t)、……、X n (t)},X i (t) is the dust concentration value output by the i-th sensor at time t, and n is the total number of sensors;

[0017] Quantify the degree of deviation of each sensor's response from the reference model. The calculation expression is as follows:

[0018] , where CRD i is the response offset integral, which represents the cumulative response offset of the i-th sensor, α is the sensitivity control parameter, which is used to control the amplification or compression effect of the instantaneous deviation, t1 is the start time of the detection window, t2 is the end time of the detection window, and R(t) is the reference threshold of the dust concentration value;

[0020] The entropy weight of each sensor is calculated to dynamically measure the contribution of each sensor and reduce the negative impact of a single abnormal sensor on the result. The calculation expression is as follows:

[0021] , in the formula, EW i is the entropy weight, i.e. the entropy weight of the i-th sensor, P ij is the normalized response intensity, which represents the normalized response intensity of the i-th sensor at time point j, P ij The calculation expression is as follows:

[0023] , where X i (t j ) is the dust concentration value output by the i-th sensor at time point j, X k (t j ) is the dust concentration value output by the kth sensor at time point j, X min (j) is the minimum dust concentration value among all sensors at time point j;

[0025] Comprehensive response deviation integral CRD i and entropy weight EW i , generate the transient deviation index, the calculation expression is as follows:

[0026] , where Transient DI is the transient deviation index, CRD krepresents the cumulative response offset of the kth sensor.

[0028] Preferably, the phase error index acquisition logic is as follows:

[0029] In the detection window, the time series data of dust concentration is collected from sensors distributed at different locations, and each sensor S z The generated time series data is labeled as y z (t), where S z ={S1, S2, S3, ..., S n}, n is the total number of sensors, y z (t) represents the measurement data of the zth sensor at time t, where t represents time, indicating that this is a measurement value that changes over time;

[0030] To analyze the phase difference, the time series data y z (t) Perform discrete Fourier transform to convert it from the time domain to the frequency domain. For each sensor’s time series data y z (t), the Fourier transform formula is:

[0031] , where y z (f) is the frequency domain representation of the zth sensor at frequency f, O is the total number of time points, each time t corresponds to a sample, and j is the imaginary unit;

[0033] From the frequency domain representation y of each sensor z (f) Extract the phase information and calculate the expression as follows: z (f) = arg(y z (f)), where φ z (f) is the phase angle, indicating the sensor S z The phase angle at frequency f, arg(y z (f)) is the complex number y z (f) the angle of argument;

[0034] Then select a main sensor S1 as the reference point and calculate the phase difference between the remaining sensors and the main sensor. The calculation expression is as follows: Δφ z (f) = φ z (f)-φ1(f), where Δφ z (f) is the phase difference between the zth sensor and the reference master sensor S1 at frequency f, φ1(f) is the phase angle of the reference master sensor S1 at frequency f;

[0035] By the phase difference Δφ z (f), calculate sensor S z The time delay between the sensor and the main sensor is calculated as follows:

[0036] , where Δt z represents the time delay, which is the sensor S z For the time delay of the reference main sensor S1;

[0038] Considering the phase difference φ between all sensors and the reference sensor z (f) and the corresponding frequency f, generate the phase error index, and the calculation expression is as follows:

[0039] , where w z is the weight of the zth sensor, f peak is the selected frequency, cos(Δφ z (f peak )) is a cosine function based on the phase difference, which is used to evaluate the synchronization of each sensor, and ∈ is a small positive number used to avoid the denominator being zero.

[0041] Preferably, the transient deviation index Transient DI and phase error index Phase EI The data is input into the pre-learned machine learning model, and the abnormal fusion index AFI is generated by the machine learning model according to the following formula:

[0042] , where k1 and k2 are transient deviation indexes. DI and phase error index Phase EI The preset proportional coefficient, and k1 and k2 are both greater than 0.

[0044] Preferably, the abnormal fusion index generated in the detection window when the big data system performs data fusion is compared and analyzed with the preset abnormal fusion index reference threshold, and the data fusion process of the big data system is divided. The specific division results are as follows:

[0045] If the abnormal fusion index is greater than or equal to the preset abnormal fusion index reference threshold, the data fusion process of the big data system is classified as an abnormal fusion process;

[0046] If the abnormal fusion index is less than the preset abnormal fusion index reference threshold, the data fusion process of the big data system is classified as a normal fusion process.

[0047] Preferably, after the data fusion process of the big data system is divided into an abnormal fusion process, a plurality of abnormal fusion indexes subsequently generated by the data fusion process of the big data system are continuously obtained to establish a data set for comprehensive analysis;

[0048] The average value and standard deviation of several anomaly fusion indexes in the data set are calculated, and the calculated average value and standard deviation of the anomaly fusion index are compared and analyzed with the preset reference threshold of the average value of the anomaly fusion index and the preset reference threshold of the standard deviation of the anomaly fusion index, respectively, to further divide the data anomaly fusion process of the big data system, and the division results are as follows;

[0049] If the average value of the abnormal fusion index is less than the preset average value reference threshold of the abnormal fusion index, the data abnormal fusion process of the big data system is further divided into a low-risk fusion process;

[0050] If the average value of the anomaly fusion index is less than the preset average value reference threshold of the anomaly fusion index and the standard deviation of the anomaly fusion index is greater than or equal to the preset standard deviation reference threshold of the anomaly fusion index, the data anomaly fusion process of the big data system is further divided into a medium-risk fusion process;

[0051] If the average value of the anomaly fusion index is greater than or equal to the preset average value reference threshold of the anomaly fusion index and the standard deviation of the anomaly fusion index is greater than or equal to the preset standard deviation reference threshold of the anomaly fusion index, the data anomaly fusion process of the big data system is further divided into a high-risk fusion process.

[0052] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0053] The present invention introduces the transient deviation index and phase error index as key indicators in data fusion, and combines the machine learning model to intelligently evaluate and classify the fusion process. This solution effectively solves the error problem caused by differences in sensor response time. The big data system can accurately judge the risk level in different scenarios. It no longer relies on the traditional static threshold judgment method, but generates an abnormal fusion index by dynamically fusing different parameter information. The system can not only detect drastic fluctuations in dust concentration, but also intelligently judge the risk level of abnormal data, realizing the transition from simple alarm to multi-level intelligent perception. This greatly improves the responsiveness of the system, reduces the possibility of false alarms and missed alarms, and ensures the real-time and accuracy of data processing.

[0054] The present invention adopts a multi-level early warning mechanism and takes corresponding countermeasures in combination with different risk levels, thereby significantly improving the safety of coal mine dust monitoring. In the case of low-risk fusion, the system prompts technicians to perform daily maintenance through a first-level early warning to ensure that the system continues to operate smoothly; in the case of medium-risk fusion, a second-level early warning is used to make timely adjustments to potential sensor failures or data anomalies to prevent the problem from further deteriorating; and in the case of high-risk fusion, a third-level early warning mechanism can quickly initiate emergency intervention, organize personnel evacuation, and take necessary repair measures to prevent accidents. This hierarchical response system for different risk levels ensures that the most effective response can be made according to the actual situation when problems arise, greatly reducing the risk of accidents in coal mine operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] 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.

[0056] Figure 1 This is a method flow chart of a real-time monitoring method for coal mine dust concentration based on big data analysis of the present invention. DETAILED DESCRIPTION

[0057] 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.

[0058] The present invention provides Figure 1 A method for real-time monitoring of coal mine dust concentration based on big data analysis is shown, comprising the following steps:

[0059] Acquire various parameter information generated in the data fusion process of the big data system in real time, including transient response information and time synchronization information, and analyze and process the acquired transient response information and time synchronization information;

[0060] The transient response information generated during the data fusion process of the big data system includes the transient deviation index, and the time synchronization information generated during the data fusion process of the big data system includes the phase error index. After acquisition, the transient deviation index and the phase error index are calibrated as Transient DI and Phase EI ;

[0061] The transient deviation index is used to measure the difference in the responses of multiple sensors to drastic environmental changes in a short period of time. It reflects the immediate response capability and synchronization of each sensor in an emergency (such as a sharp increase in dust concentration).

[0062] In the data fusion process of the big data system, the change of the transient deviation index may cause errors in the data fusion process because it reflects the difference in the response of multiple sensors in response to drastic changes in the environment. When the dust concentration fluctuates sharply, the sensitivity and response speed of different sensors are inconsistent, which may cause the data to deviate from the actual situation in a short period of time. If the transient deviation index increases, it means that there are significant differences in the response between sensors, some devices quickly feedback changes, while others lag behind or fail to detect abnormalities. This inconsistency will make it difficult for the big data system to accurately reflect the real-time environment during data fusion, causing the system to misjudge the current dust concentration level or trend. For example, in the case of a sharp increase in dust concentration, if some sensors lag in response, the system may underestimate the actual risk, delay key warnings, and lead to major safety hazards such as dust explosions. In addition, even if some sensors capture instantaneous changes, the fusion algorithm cannot effectively handle these deviations, which may cause false alarms, trigger unnecessary shutdowns or control operations, and reduce the operating efficiency of the system. Therefore, the change of the transient deviation index directly affects the accuracy and stability of data fusion, and these deviations must be compensated in time through algorithm optimization, dynamic calibration and other means to ensure the reliability of data fusion results.

[0063] The logic for obtaining the transient deviation index is as follows:

[0064] Under the detection window, multiple sensors arranged at different locations in the coal mine collect dust concentration data, and the dust concentration data output by each sensor is calibrated as X i (t), where: X i (t)={X1(t)、X2(t)、X3(t)、……、X n (t)},X i (t) is the dust concentration value output by the i-th sensor at time t, and n is the total number of sensors;

[0065] Quantify the degree of deviation of each sensor's response from the reference model. The calculation expression is as follows:

[0066] , where CRD i is the response offset integral, which represents the cumulative response offset of the i-th sensor, α is the sensitivity control parameter, which is used to control the amplification or compression effect of the instantaneous deviation, t1 is the start time of the detection window, t2 is the end time of the detection window, and R(t) is the reference threshold of the dust concentration value;

[0068] Quantifying the deviation of each sensor's response from the system reference model means calculating and evaluating the degree of difference between each sensor's measurement and the normal value expected by the system. The system reference model is usually based on historical data, theoretical models, or data from calibrated reference sensors, representing the normal trend of dust concentration changes in the environment. By comparing with the reference model, it can be determined whether there are anomalies, delays, or errors in the response of a sensor. The purpose of this process is to find out which sensors show inconsistencies or deviations from the normal state during the monitoring process, so as to make timely corrections, eliminations, or trigger early warnings to ensure the accuracy of data fusion and the reliability of system operation.

[0069] In order to ensure that the dust concentration monitoring system in coal mines and other environments can accurately identify risks and issue warnings in a timely manner, it is critical to set reference thresholds reasonably. These reference thresholds are usually obtained in the following ways and are continuously adjusted and optimized according to different environmental needs:

[0070] 1. Regulations and industry standards

[0071] Source: Health and safety standards issued by the government or industry organizations (such as the Occupational Disease Prevention and Control Law of the People's Republic of China, OSHA standards, etc.).

[0072] Purpose: These standards set dust concentration limits in different working environments, such as the maximum allowable concentration of inhalable dust or coal dust. Companies can use these standards to establish initial thresholds for their systems.

[0073] For example, in a coal mine environment, the dust concentration in certain areas must not exceed 10 mg / m 3 ; For workers who are exposed for a long time, a lower threshold may be set.

[0074] 2. Site survey and benchmark testing

[0075] Source: Before the monitoring system is officially put into use, on-site testing and surveys are carried out to obtain dust concentration conditions at different times and in different areas.

[0076] Function: Through field data, we can understand the concentration range of each area under normal conditions and identify high-risk areas. Field benchmark data can also be used to establish preliminary reference thresholds under different working conditions.

[0077] Adjustment method: If the baseline concentration in a certain area is high, a lower alarm threshold needs to be set to ensure that the cumulative effect is dealt with in a timely manner.

[0078] 3. Historical data analysis

[0079] Source: Based on the company's historical monitoring data, the dust concentration trend in different time periods and production stages is analyzed.

[0080] Function: Through historical data, periodic fluctuations and abnormal conditions of dust concentration can be identified, and dynamic thresholds can be set based on this. For example, if the dust concentration peaks in a certain period of time, the peak threshold needs to be set reasonably according to the trend.

[0081] 4. Laboratory testing and simulation

[0082] Source: The changes in dust concentration under different environments are simulated in the laboratory to test the impact of dust on sensors and the environment.

[0083] Role: Laboratory testing can help determine the impact of changes in dust concentration on safety under extreme environmental conditions (such as high temperature and high humidity), thereby optimizing threshold settings. For example, simulating the reading deviation of dust concentration sensors under different humidity conditions.

[0084] 5. Risk assessment and expert opinion

[0085] Source: Assessment by a team of environmental monitoring experts, safety engineers or occupational health experts.

[0086] Function: By analyzing the possible risks of dust explosions, occupational diseases, etc., the expert team can provide reasonable threshold recommendations for the system. For example, for areas where explosions may occur, a lower trigger threshold can be set to prevent them in advance.

[0087] 6. Real-time monitoring and self-learning optimization

[0088] Source: Dynamically adjust thresholds through the machine learning algorithm and self-learning function of the intelligent monitoring system.

[0089] Function: During the monitoring process, the system continuously adjusts the threshold according to environmental changes and actual feedback. For example, if the reading of a sensor is close to the set threshold for a long time, but no actual problem occurs, the system can gradually relax the threshold to reduce false alarms.

[0090] 7. Custom thresholds under different working conditions

[0091] Source: According to the specific production process or working conditions, the dust concentration threshold in a specific scenario is set.

[0092] Function: The amount of dust generated under different working conditions is different, so the threshold needs to be flexibly adjusted according to the actual situation. For example, during the equipment maintenance stage, instantaneous high-concentration dust may be generated. At this time, a temporary threshold needs to be set to avoid false alarms.

[0093] Obtaining the reference threshold of dust concentration is a complex process, which requires a combination of regulatory standards, field data, historical analysis, and expert evaluation. As the system runs deeper, the threshold needs to be dynamically optimized based on real-time monitoring data. Reasonable threshold settings can not only detect potential risks in a timely manner, but also reduce false alarms and missed alarms, ensuring production safety and personnel health in high-risk environments such as coal mines.

[0094] The entropy weight of each sensor is calculated to dynamically measure the contribution of each sensor and reduce the negative impact of a single abnormal sensor on the result. The calculation expression is as follows:

[0095] , in the formula, EW i is the entropy weight, i.e. the entropy weight of the i-th sensor, P ij is the normalized response intensity, which represents the normalized response intensity of the i-th sensor at time point j, P ij The calculation expression is as follows:

[0097] , where X i (t j ) is the dust concentration value output by the i-th sensor at time point j, X k (t j ) is the dust concentration value output by the kth sensor at time point j, X min (j) is the minimum dust concentration value among all sensors at time point j;

[0099] Calculating the entropy weight of each sensor means dynamically assigning weights to each sensor data by analyzing its volatility and information content, so as to reasonably reflect its importance in data fusion. The entropy weight method determines which sensors provide more useful information in a certain period of time by normalizing the changes in the data of each sensor and calculating the information entropy. If the data of a sensor changes greatly and is valuable, its entropy weight will be higher; on the contrary, if the data of a sensor changes little or the information is redundant, its entropy weight will be lower. In this way, the system can pay more attention to the sensor data that has an important response to environmental changes.

[0100] An important role of this entropy weight allocation mechanism is to reduce the negative impact of a single abnormal sensor on the overall monitoring results. In actual operation, sometimes a sensor will generate abnormal data due to failure, error or environmental interference. If all sensors are given the same weight indiscriminately, the data of abnormal sensors may seriously affect the final judgment results. Through the entropy weight method, the system will identify these abnormal sensors and reduce their weights, reduce their interference with the overall results, and ensure the accuracy of data fusion and the reliability of monitoring.

[0101] Comprehensive response deviation integral CRD i and entropy weight EW i , generate the transient deviation index, the calculation expression is as follows:

[0102] , where Transient DI is the transient deviation index, CRD k represents the cumulative response offset of the kth sensor;

[0104] From the transient deviation index calculation expression, it can be seen that the larger the transient deviation index performance value generated under the detection window when the big data system performs data fusion, the more obvious the difference in the response of each sensor in response to environmental mutations, thereby increasing the probability of errors in the data fusion process. A high transient deviation index means that the feedback between sensors is inconsistent. Some devices may capture changes quickly, while others lag behind or fail to detect changes. This will cause the fusion algorithm to fail to correctly reflect the real-time situation, increasing the risk of misjudgment and missed judgment. Conversely, a small transient deviation index performance value indicates that the responses between sensors are more synchronized and consistent, the data fusion process is more reliable, and the probability of error is reduced accordingly.

[0105] The phase error index is used to evaluate the time synchronization between multiple sensor data to ensure that they report consistent data states in the same time period. It quantifies the phase difference between different sensors. Especially when monitoring dynamic environments such as dust concentration, the time alignment of data is crucial.

[0106] During the data fusion process of the big data system, if the phase error index changes, it will directly lead to errors in the fusion results. This is because the phase error index reflects the time synchronization deviation between different sensors, that is, whether the sensors can report consistent data status in the same time period. When the phase of the sensor data is inconsistent, for example, the response of one sensor lags behind or advances that of other sensors, the system will fuse the data at the wrong time point. This time misalignment can lead to misjudgment of the dust concentration change trend. For example, when the dust concentration in a certain area exceeds the standard for a short time but quickly returns to normal, if the delayed data of some sensors are mixed with the real-time data, the system may mistakenly believe that the concentration is still rising or has returned to a safe state, triggering wrong decisions or delaying emergency response. Especially in the case of drastic fluctuations in dust concentration or explosion criticality, the error in time synchronization will directly affect the system's early warning capability and risk judgment, and may cause major safety hazards. Therefore, it is crucial to dynamically monitor and adjust the phase error index to ensure the accuracy of data fusion, reduce false alarms or missed alarms, and improve the efficiency and reliability of the system's response to emergencies.

[0107] The logic of obtaining the phase error index is as follows:

[0108] In the detection window, the time series data of dust concentration is collected from sensors distributed at different locations, and each sensor S z The generated time series data is labeled as y z (t), where S z ={S1, S2, S3, ..., S n}, n is the total number of sensors, y z (t) represents the measurement data of the zth sensor at time t, where t represents time, indicating that this is a measurement value that changes over time (i.e., time series data);

[0109] To analyze the phase difference, the time series data y z (t) Perform discrete Fourier transform to convert it from the time domain to the frequency domain. For each sensor’s time series data y z (t), the Fourier transform formula is:

[0110] , where y z (f) is the frequency domain representation of the zth sensor at frequency f, O is the total number of time points, each time t corresponds to a sample, and j is the imaginary unit;

[0112] Phase difference refers to the time or angle offset between the fluctuations of two periodic signals at the same frequency. It indicates the leading or lagging relationship of one signal relative to another. In the formula, the phase difference is usually expressed as an angle (radian). For example, the phases of the two signals are φ1 and φ2, and the phase difference between them is Δφ=φ2-φ1. In a multi-sensor system, multiple sensors need to capture environmental changes synchronously, and the phase difference is used to measure whether different sensors are accurately synchronized. If the phases between sensors are not synchronized, the system may misjudge trends or ignore key changes.

[0113] Analyzing phase differences is critical to ensuring the time synchronization and accurate fusion of data. In a sensor network, each sensor may be time-misaligned due to hardware differences, response speed, or network latency. If these phase differences are not taken into account, the results of data fusion may be inaccurate. For example, in coal mine monitoring, some sensors may detect a peak in dust concentration while other sensor data are updated later. By analyzing the phase difference, the time offset between different sensors can be quantified to ensure the correctness of data fusion and prevent misjudgment and underreporting.

[0114] By converting time series data to the frequency domain through discrete Fourier transform (DFT), the periodic characteristics and phase relationship of the data can be analyzed more intuitively. In the time domain, the frequency and phase information of the signal are not easy to observe directly, but after DFT conversion, the amplitude and phase of each frequency component can be clearly presented. In the frequency domain, the system can accurately calculate the phase difference of different sensors at the same frequency and perform time synchronization analysis. In addition, frequency domain analysis can also help filter noise, extract the main characteristic frequencies, and improve the data fusion accuracy and real-time performance of multi-sensor systems.

[0115] From the frequency domain representation y of each sensor z (f) Extract the phase information and calculate the expression as follows: z (f) = arg(y z (f)), where φ z (f) is the phase angle, indicating the sensor S z The phase angle at frequency f, arg(y z (f)) is the complex number y z (f) the angle of argument;

[0116] Then select a main sensor S1 as the reference point and calculate the phase difference between the remaining sensors and the main sensor. The calculation expression is as follows: Δφ z (f) = φ z (f)-φ1(f), where Δφ z (f) is the phase difference between the zth sensor and the reference master sensor S1 at frequency f, φ1(f) is the phase angle of the reference master sensor S1 at frequency f;

[0117] The phase angle of a sensor at a certain frequency refers to the initial position of the periodic signal corresponding to the frequency or the offset angle relative to time, usually expressed in radians or degrees. In the frequency domain after Fourier transform, the phase angle describes the relative position of the signal on a certain frequency component, that is, the leading or lagging situation of the signal waveform relative to the reference time. The role of the phase angle is mainly reflected in the synchronization and analysis of multi-sensor data. If the phase angles of each sensor are consistent, it means that they synchronously measure the same environmental phenomenon at the same frequency; on the contrary, if there is a difference in the phase angle, it indicates that there is a response delay or asynchrony between the sensors, and correction is required to ensure the accurate fusion of the monitoring data.

[0118] The argument of a complex number is the rotation angle of the complex number relative to the positive real axis on a two-dimensional plane, indicating the direction of the complex number. For a complex number z = a + bi, its position in the complex plane can be expressed in polar coordinate form:

[0119] z=r·ejθ

[0120] Where θ is the argument of the complex number, that is, the angle with the positive real axis, describing its phase direction. The role of the argument of the complex number is particularly critical in Fourier transform because it determines the phase position of each frequency component. In multi-sensor signal processing, the argument of the complex number is used to calculate the phase difference of the sensor at different frequencies to ensure that the system can accurately synchronize this data. If the argument is not aligned, errors may occur after signal fusion, affecting the accurate monitoring and analysis of environmental changes.

[0121] By the phase difference Δφ z (f), calculate sensor S z The time delay between the sensor and the main sensor is calculated as follows:

[0122] , where Δt z represents the time delay, which is the sensor S z For the time delay of the reference main sensor S1;

[0124] Considering the phase difference φ between all sensors and the reference sensor z (f) and the corresponding frequency f, generate the phase error index, and the calculation expression is as follows:

[0125] , where w z is the weight of the zth sensor, which is usually set according to the location or importance of the sensor, f peak is the main frequency selected, usually the maximum response frequency, cos(Δφ z (f peak )) is a cosine function based on the phase difference, which is used to evaluate the synchronization of each sensor. The closer the value is to 1, the smaller the phase difference and the lower the error. ∈ is a small positive number used to avoid the denominator being zero.

[0127] It can be seen from the calculation expression of the phase error index that the larger the phase error index performance value generated under the detection window when the big data system performs data fusion, the more obvious the time synchronization deviation between multiple sensors is, which leads to a greater probability of error in the data fusion process. The phase error index reflects whether the responses of different sensors to the same event within the detection window are consistent. If the response of some sensors lags behind or advances that of other sensors, the fused data may lose the accuracy of the timing, causing the system to be unable to correctly judge the actual trend of environmental changes. When the phase error index performance value is larger, this time misalignment will increase the risk of false alarms and missed alarms in the key monitoring links of the system, especially in the scenario of drastic changes in dust concentration or safety warning. On the contrary, if the phase error index performance value is smaller, it means that the responses of each sensor are highly synchronized, and the results of data fusion are more accurate and reliable, which can effectively reduce the errors in monitoring and improve the judgment and response capabilities of the system.

[0128] The analyzed and processed transient response information and time synchronization information are input into the pre-learned machine learning model, and the abnormal fusion index is generated through the machine learning model. The abnormal fusion index is used to intelligently evaluate the data fusion process of the big data system.

[0129] Transient deviation index DI and phase error index Phase EI The data is input into the pre-learned machine learning model, and the abnormal fusion index AFI is generated by the machine learning model according to the following formula:

[0130] , where k1 and k2 are transient deviation indexes. DI and phase error index Phase EI The preset proportional coefficient, and k1 and k2 are both greater than 0.

[0132] It can be seen from the calculation expression of the anomaly fusion index that the larger the transient deviation index performance value and the phase error index performance value generated under the detection window when the big data system performs data fusion, that is, the larger the anomaly fusion index performance value generated under the detection window when the big data system performs data fusion, the greater the probability of anomalies occurring when the big data system performs data fusion, and vice versa.

[0133] Based on the evaluation results of the machine learning model, the data fusion process of the big data system is divided into a normal fusion process or an abnormal fusion process, and the abnormal state of data fusion in the big data system is intelligently perceived;

[0134] The abnormal fusion index generated in the detection window when the big data system performs data fusion is compared and analyzed with the pre-set abnormal fusion index reference threshold, and the data fusion process of the big data system is divided. The specific division results are as follows:

[0135] If the abnormal fusion index is greater than or equal to the preset abnormal fusion index reference threshold, the data fusion process of the big data system is classified as an abnormal fusion process;

[0136] If the abnormal fusion index is less than the preset abnormal fusion index reference threshold, the data fusion process of the big data system is classified as a normal fusion process;

[0137] For the abnormal fusion process, continuously obtain the data fusion parameter information of the big data system, and further divide the abnormal fusion process into low-risk fusion process, medium-risk fusion process and high-risk fusion process;

[0138] When the data fusion process of the big data system is divided into an abnormal fusion process, several abnormal fusion indexes subsequently generated by the data fusion process of the big data system are continuously obtained to establish a data set for comprehensive analysis;

[0139] The average value and standard deviation of several anomaly fusion indexes in the data set are calculated, and the calculated average value and standard deviation of the anomaly fusion index are compared and analyzed with the preset reference threshold of the average value of the anomaly fusion index and the preset reference threshold of the standard deviation of the anomaly fusion index, respectively, to further divide the data anomaly fusion process of the big data system, and the division results are as follows;

[0140] If the average value of the abnormal fusion index is less than the preset average value reference threshold of the abnormal fusion index, the data abnormal fusion process of the big data system is further divided into a low-risk fusion process;

[0141] If the average value of the anomaly fusion index is less than the preset average value reference threshold of the anomaly fusion index and the standard deviation of the anomaly fusion index is greater than or equal to the preset standard deviation reference threshold of the anomaly fusion index, the data anomaly fusion process of the big data system is further divided into a medium-risk fusion process;

[0142] If the average value of the abnormal fusion index is greater than or equal to the preset average value reference threshold of the abnormal fusion index and the standard deviation of the abnormal fusion index is greater than or equal to the preset standard deviation reference threshold of the abnormal fusion index, the data abnormal fusion process of the big data system is further divided into a high-risk fusion process;

[0143] Different responses are adopted for different levels of integration process;

[0144] During the low-risk fusion process, although the abnormal fusion index in the system exceeds the threshold of normal fusion, the overall performance is stable and there is no obvious sudden risk. This indicates that there may be slight fluctuations or occasional errors in the data fusion process, but the impact on the operation of the system is small. In view of this situation, it is recommended to take regular monitoring measures and strengthen daily inspections of sensors and data fusion processes. At the same time, a first-level warning can be set to remind technicians to pay attention to the system status and record relevant data for future analysis. At this stage, no emergency action is required, but it is necessary to ensure that the system continues to operate stably.

[0145] The medium-risk fusion process indicates that the system has experienced significant fluctuations. It may be that some sensors have temporarily failed or the data has changed abnormally, causing the data fusion results to deviate from the normal state. At this time, the data and sensor nodes should be checked in detail to identify the specific problems and make appropriate adjustments or maintenance. At the same time, the system should enter the second-level warning state and send an alert to the technical and management teams to explain the potential risk areas and possible causes. If necessary, local shutdowns or system load reduction operations can be adopted to reduce risks. In addition, the data fusion algorithm should be optimized based on the analysis results to prevent similar situations from happening again.

[0146] A high-risk fusion process means that there are serious errors in the sensors or data fusion algorithms within the system, which may cause the system to misjudge key data and cause major hidden dangers, such as dust explosions and other dangerous situations. At this time, emergency intervention measures should be initiated immediately, including suspending operations in high-risk areas, comprehensively troubleshooting sensor failures, and analyzing whether the data fusion model has defects. The system should enter the third-level warning state, issue an emergency alert to all relevant personnel and departments, arrange necessary personnel evacuation, and immediately organize an emergency response team to conduct on-site inspections and repairs. After the problem is resolved, the system status should be re-evaluated, and monitoring of high-risk areas should be strengthened to ensure the safety of subsequent operations.

[0147] The present invention introduces the transient deviation index and phase error index as key indicators in data fusion, and combines the machine learning model to intelligently evaluate and classify the fusion process. This solution effectively solves the error problem caused by differences in sensor response time. The big data system can accurately judge the risk level in different scenarios. It no longer relies on the traditional static threshold judgment method, but generates an abnormal fusion index by dynamically fusing different parameter information. The system can not only detect drastic fluctuations in dust concentration, but also intelligently judge the risk level of abnormal data, realizing the transition from simple alarm to multi-level intelligent perception. This greatly improves the responsiveness of the system, reduces the possibility of false alarms and missed alarms, and ensures the real-time and accuracy of data processing.

[0148] The present invention adopts a multi-level early warning mechanism and takes corresponding countermeasures in combination with different risk levels, thereby significantly improving the safety of coal mine dust monitoring. In the case of low-risk fusion, the system prompts technicians to perform daily maintenance through a first-level early warning to ensure that the system continues to operate smoothly; in the case of medium-risk fusion, a second-level early warning is used to make timely adjustments to potential sensor failures or data anomalies to prevent the problem from further deteriorating; and in the case of high-risk fusion, a third-level early warning mechanism can quickly initiate emergency intervention, organize personnel evacuation, and take necessary repair measures to prevent accidents. This hierarchical response system for different risk levels ensures that the most effective response can be made according to the actual situation when problems arise, greatly reducing the risk of accidents in coal mine operations.

[0149] 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. A real-time monitoring method for coal mine dust concentration based on big data analysis, characterized in that: The following steps are involved: Acquire various parameter information generated in the data fusion process of the big data system in real time, including transient response information and time synchronization information, and analyze and process the acquired transient response information and time synchronization information; The analyzed and processed transient response information and time synchronization information are input into the pre-learned machine learning model, and the abnormal fusion index is generated through the machine learning model. The abnormal fusion index is used to intelligently evaluate the data fusion process of the big data system. Based on the evaluation results of the machine learning model, the data fusion process of the big data system is divided into a normal fusion process or an abnormal fusion process, and the abnormal state of data fusion in the big data system is intelligently perceived; For the abnormal fusion process, continuously obtain the data fusion parameter information of the big data system, and further divide the abnormal fusion process into low-risk fusion process, medium-risk fusion process and high-risk fusion process; Different countermeasures should be taken for different levels of integration process.

2. The method for real-time monitoring of coal mine dust concentration based on big data analysis according to claim 1 is characterized in that: The transient response information generated during the data fusion process of the big data system includes the transient deviation index, and the time synchronization information generated during the data fusion process of the big data system includes the phase error index. After acquisition, the transient deviation index and the phase error index are calibrated as Transient DI and Phase EI .

3. The method for real-time monitoring of coal mine dust concentration based on big data analysis according to claim 2 is characterized in that: The logic for obtaining the transient deviation index is as follows: Under the detection window, multiple sensors arranged at different locations in the coal mine collect dust concentration data, and the dust concentration data output by each sensor is calibrated as X i (t), where: X i (t)={X1(t)、X2(t)、X3(t)、……、X n (t)},X i (t) is the dust concentration value output by the i-th sensor at time t, and n is the total number of sensors; Quantify the degree of deviation of each sensor's response from the reference model. The calculation expression is as follows: , where CRD i is the response offset integral, which represents the cumulative response offset of the i-th sensor, α is the sensitivity control parameter, which is used to control the amplification or compression effect of the instantaneous deviation, t1 is the start time of the detection window, t2 is the end time of the detection window, and R(t) is the reference threshold of the dust concentration value; The entropy weight of each sensor is calculated to dynamically measure the contribution of each sensor and reduce the negative impact of a single abnormal sensor on the result. The calculation expression is as follows: , in the formula, EW i is the entropy weight, i.e. the entropy weight of the i-th sensor, P ij is the normalized response intensity, which represents the normalized response intensity of the i-th sensor at time point j, P ij The calculation expression is as follows: , where X i (t j ) is the dust concentration value output by the i-th sensor at time point j, X k (t j ) is the dust concentration value output by the kth sensor at time point j, X min (j) is the minimum dust concentration value among all sensors at time point j; Comprehensive response deviation integral CRD i and entropy weight EW i , generate the transient deviation index, the calculation expression is as follows: , where Transient DI is the transient deviation index, CRD k represents the cumulative response offset of the kth sensor.

4. The method for real-time monitoring of coal mine dust concentration based on big data analysis according to claim 2 is characterized in that: The logic for obtaining the phase error index is as follows: In the detection window, the time series data of dust concentration is collected from sensors distributed at different locations, and each sensor S z The generated time series data is labeled as y z (t), where S z ={S1, S2, S3, ..., S n }, n is the total number of sensors, y z (t) represents the measurement data of the zth sensor at time t, where t represents time, indicating that this is a measurement value that changes over time; To analyze the phase difference, the time series data y z (t) Perform discrete Fourier transform to convert it from the time domain to the frequency domain. For each sensor’s time series data y z (t), the Fourier transform formula is: , where y z (f) is the frequency domain representation of the zth sensor at frequency f, O is the total number of time points, each time t corresponds to a sample, and j is the imaginary unit; From the frequency domain representation y of each sensor z (f) Extract the phase information and calculate the expression as follows: z (f) = arg(y z (f)), where φ z (f) is the phase angle, indicating the sensor S z The phase angle at frequency f, arg(y z (f)) is the complex number y z (f) the angle of argument; Then select a main sensor S1 as the reference point and calculate the phase difference between the remaining sensors and the main sensor. The calculation expression is as follows: Δφ z (f) = φ z (f)-φ1(f), where Δφ z (f) is the phase difference between the zth sensor and the reference master sensor S1 at frequency f, φ1(f) is the phase angle of the reference master sensor S1 at frequency f; By the phase difference Δφ z (f), calculate sensor S z The time delay between the sensor and the main sensor is calculated as follows: , where Δt z represents the time delay, which is the sensor S z For the time delay of the reference main sensor S1; Considering the phase difference φ between all sensors and the reference sensor z (f) and the corresponding frequency f, generate the phase error index, and the calculation expression is as follows: , where w z is the weight of the zth sensor, f peak is the selected frequency, cos(Δφ z (f peak )) is a cosine function based on the phase difference, which is used to evaluate the synchronization of each sensor, and ∈ is a small positive number used to avoid the denominator being zero.

5. The method for real-time monitoring of coal mine dust concentration based on big data analysis according to claim 2 is characterized in that: Transient deviation index DI and phase error index Phase EI The data is input into the pre-learned machine learning model, and the abnormal fusion index AFI is generated by the machine learning model according to the following formula: , where k1 and k2 are transient deviation indexes. DI and phase error index Phase EI The preset proportional coefficient, and k1 and k2 are both greater than 0.

6. The method for real-time monitoring of coal mine dust concentration based on big data analysis according to claim 5 is characterized in that: The abnormal fusion index generated in the detection window when the big data system performs data fusion is compared and analyzed with the pre-set abnormal fusion index reference threshold, and the data fusion process of the big data system is divided. The specific division results are as follows: If the abnormal fusion index is greater than or equal to the preset abnormal fusion index reference threshold, the data fusion process of the big data system is classified as an abnormal fusion process; If the abnormal fusion index is less than the preset abnormal fusion index reference threshold, the data fusion process of the big data system is classified as a normal fusion process.

7. A method for real-time monitoring of coal mine dust concentration based on big data analysis according to claim 6, characterized in that: When the data fusion process of the big data system is divided into an abnormal fusion process, several abnormal fusion indexes subsequently generated by the data fusion process of the big data system are continuously obtained to establish a data set for comprehensive analysis; The average value and standard deviation of several anomaly fusion indexes in the data set are calculated, and the calculated average value and standard deviation of the anomaly fusion index are compared and analyzed with the preset reference threshold of the average value of the anomaly fusion index and the preset reference threshold of the standard deviation of the anomaly fusion index, respectively, to further divide the data anomaly fusion process of the big data system, and the division results are as follows; If the average value of the abnormal fusion index is less than the preset average value reference threshold of the abnormal fusion index, the data abnormal fusion process of the big data system is further divided into a low-risk fusion process; If the average value of the anomaly fusion index is less than the preset average value reference threshold of the anomaly fusion index and the standard deviation of the anomaly fusion index is greater than or equal to the preset standard deviation reference threshold of the anomaly fusion index, the data anomaly fusion process of the big data system is further divided into a medium-risk fusion process; If the average value of the anomaly fusion index is greater than or equal to the preset average value reference threshold of the anomaly fusion index and the standard deviation of the anomaly fusion index is greater than or equal to the preset standard deviation reference threshold of the anomaly fusion index, the data anomaly fusion process of the big data system is further divided into a high-risk fusion process.

Images