Mine ventilation fault diagnosis and early warning method

Abstract

The application discloses a mine ventilation fault diagnosis and early warning method and relates to the technical field of mine ventilation; the method comprises the following steps: acquiring real-time original ventilation data; processing the original ventilation data by using a preset feature extraction method to obtain ventilation data stream parameters; integrating differentiated features of ventilation system deviation to construct a historical feature library; the ventilation system deviation comprises control adjustment deviation, equipment mechanical characteristic deviation and ventilation network structure characteristic deviation; based on the real-time ventilation data stream parameters and the historical feature library, the source of periodic airflow pulsation is diagnosed to obtain a source diagnosis report; based on the source diagnosis report, the periodic airflow pulsation is processed and graded early warned according to the source; the application realizes accurate identification of periodic airflow pulsation, graded early warning of risks and targeted development of processing strategies.

Description

Technical Field

[0001] This invention relates to the field of mine ventilation technology, and more specifically, to a method for diagnosing and warning of mine ventilation faults. Background Technology

[0002] With the continuous development of the mining industry, the requirements for underground operation safety and continuous mine operation are also gradually increasing. As a core facility ensuring underground operation safety in the mining field, the mine ventilation system is a crucial link supporting safe mine operation. Its operational stability and control effectiveness directly affect the dilution effect of harmful gases, dust removal efficiency, and temperature and humidity control levels underground, thus impacting the safety of underground personnel, normal equipment operation, and continuous mining operations. Currently, mainstream mines generally adopt a monitoring system integrating wind speed sensors, gas sensors, and dust sensors, combined with main ventilation fans, local ventilation fans, and dampers to form an automated control architecture. Based on the core principle of static parameter matching, data is collected in real time by wind speed sensors, preprocessed by edge computing nodes, and uploaded to the control platform. The PLC controller then performs closed-loop adjustment according to preset wind speed thresholds. Simultaneously, dust sensors assist in confirming the dust-carrying effect of airflow, achieving underground airflow speed control, harmful gas concentration monitoring, and abnormal alarms, meeting the simple ventilation safety management needs of daily mining operations.

[0003] However, when faced with the complex production environment and dynamic changes in operating conditions in mines, such as adjustment deviations caused by control signal transmission delays, fluctuations in mechanical characteristics due to aging of ventilation equipment components, and airflow turbulence caused by changes in the local structural resistance of the underground ventilation network, existing mine ventilation control technologies based on static parameter matching face the problem of difficulty in eliminating periodic airflow pulsations. This can easily lead to instantaneous exceedances in dust concentration and the risk of static electricity accumulation, which may result in accelerated mechanical wear of ventilation equipment, interruption of underground production, or even triggering explosions and other safety production problems. Therefore, there is a need for a mine ventilation control technology and method that can accurately identify the source of periodic airflow pulsations and eliminate them at their root, achieving stable airflow regulation under different mine operating conditions to ensure the safety of underground operations and the continuous and stable operation of mining.

[0004] Currently, existing mine ventilation control technologies and methods mainly revolve around static parameter monitoring and basic deviation correction. Examples include closed-loop control schemes for underground wind speed, abnormal alarm technologies for harmful gas concentrations, and auxiliary monitoring methods for dust concentration. While these methods can control basic underground ventilation parameters, research and practical application have revealed a lack of precise mechanisms for identifying the sources of periodic airflow pulsations. They cannot effectively distinguish between three core pulsation causes: control deviation, equipment mechanical characteristic deviation, and ventilation network structural characteristic deviation. Specifically, control deviation, caused by signal delay, unoptimized algorithms, and limited data, leads to controller misjudgments and periodic adjustment cycles. Equipment mechanical characteristic deviation, caused by component aging and poor power matching, results in periodic fluctuations in airflow parameters with equipment operation. Ventilation network structural characteristic deviation, caused by changes in local structural resistance, leads to periodic airflow disturbances.

[0005] In view of this, the present invention proposes a method for diagnosing and warning of mine ventilation failures to solve the above problems. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art and achieve the above objectives, the present invention provides the following technical solution: a method for diagnosing and warning of mine ventilation faults, comprising:

[0007] Obtain real-time raw ventilation data;

[0008] The raw ventilation data is processed using a preset feature extraction method to obtain ventilation data stream parameters;

[0009] Integrate the differential characteristics of ventilation system deviations to construct a historical feature database; ventilation system deviations include control and regulation deviations, equipment mechanical characteristic deviations, and ventilation network structural characteristic deviations.

[0010] Based on real-time ventilation data stream parameters and historical feature database, the source of periodic airflow pulsation is diagnosed and a source diagnosis report is obtained.

[0011] Based on the source diagnosis report, periodic airflow pulsations are processed by source and classified for early warning.

[0012] Furthermore, methods for constructing historical feature databases include:

[0013] Historical operational data of the mine ventilation system during periods of deviation and normal operation are collected in advance. The historical operational data during periods of deviation include complete data when the deviation occurs, with the deviation marked. The historical operational data during periods of normal operation include complete data when the ventilation system has no deviation. The historical operational data includes airflow parameters, fan vibration parameters, damper opening fluctuations, roadway resistance changes, dust concentration, and electrostatic voltage.

[0014] Historical operational data is preprocessed by using linear interpolation to fill in missing values ​​and standardizing the data to eliminate dimensional differences. For historical operational data during periods of ventilation system deviation, the data is processed according to the specific ventilation system deviation to obtain the correlation strength threshold, time lag range, similarity threshold, and electrostatic voltage threshold for each deviation. Data statistics and feature extraction are performed on historical operational data during normal operation periods to obtain the parameter stability range. Finally, the ventilation system deviation, correlation strength threshold, time lag range, similarity threshold, electrostatic voltage threshold, and parameter stability range are entered into the historical feature database to obtain the constructed historical feature database.

[0015] Furthermore, the calculation method for the association strength threshold includes:

[0016] The source parameters of ventilation system deviations are defined as fan vibration parameters, damper opening fluctuations, and roadway resistance changes; among them, the source parameter corresponding to control and regulation deviations is damper opening fluctuations, the source parameter corresponding to equipment mechanical characteristic deviations is fan vibration parameters, and the source parameter corresponding to ventilation network structural characteristic deviations is roadway resistance changes.

[0017] The damper opening fluctuation and airflow parameters are collected at preset sampling intervals to form time series of source parameters and airflow parameters of control and regulation deviation. The Pearson correlation coefficient of the time series of source parameters and airflow parameters is used as the correlation strength to obtain a set of correlation coefficients. Descriptive statistical methods are used to obtain the correlation strength threshold of control and regulation deviation.

[0018] Based on the method of obtaining the correlation strength threshold of the control and regulation deviation, the correlation strength thresholds of the equipment mechanical characteristic deviation and the ventilation network structural characteristic deviation are calculated respectively.

[0019] Furthermore, the methods for calculating the time lag range include:

[0020] The time series of damper opening fluctuations and airflow parameters are obtained. The time offset between the time series of damper opening fluctuations and airflow parameters is used as the lag value. The lag correlation strength corresponding to different lag values ​​is calculated using a cross-correlation function. The lag value corresponding to the maximum lag correlation strength is taken as the target lag value. The cross-correlation function is calculated for all airflow parameters corresponding to damper opening fluctuations in historical operating data to obtain the target lag value set. Descriptive statistical methods are used to obtain the time lag range of control and regulation deviations.

[0021] Based on the method of obtaining the time lag range of the control adjustment deviation, the time lag ranges of the equipment mechanical characteristics and the ventilation network structural characteristics are calculated respectively.

[0022] Furthermore, the methods for calculating the similarity threshold include:

[0023] The pre-processed dust concentration is uniformly resampled with a preset sampling frequency, such as 1Hz. The wavelet threshold denoising method is used to remove high-frequency interference signals to obtain the dust concentration time domain signal.

[0024] Extract the time-domain features of the dust concentration time-domain signal; use Fast Fourier Transform to convert the dust concentration time-domain signal into a dust concentration frequency-domain signal, and extract the frequency-domain features of the dust concentration frequency-domain signal.

[0025] Extract dust concentrations from the historical feature database that have been marked with ventilation system deviations. Classify the ventilation system deviations and statistically analyze the time-domain and frequency-domain distributions of dust concentrations when each ventilation system deviation occurs. Calculate the feature standard values.

[0026] Extract dust concentrations from the historical feature database that have already labeled ventilation system deviations, and hide the corresponding ventilation system deviation labels for the dust concentrations. Extract the temporal and frequency domain features of the dust concentration data, and calculate the cosine similarity between the feature values ​​of the temporal and frequency domain features and the feature standard values ​​to obtain the similarity results. Recover the hidden ventilation system deviation labels, and match the calculated similarity results with the deviation labels to obtain a similarity threshold.

[0027] Furthermore, the calculation method for the electrostatic voltage threshold includes:

[0028] Extract the electrostatic voltages of ventilation system deviations marked in the historical feature database, classify them according to ventilation system deviations, and obtain the control and regulation deviation voltage set, the equipment mechanical characteristic deviation voltage set, and the ventilation network structural characteristic deviation voltage set;

[0029] Taking control and regulation deviation as an example, the electrostatic voltages of the control and regulation deviation voltage set are sorted in ascending order to obtain the voltage sorting set; the voltage mean and voltage standard deviation of the electrostatic voltages in the voltage sorting set are calculated; the product of the preset skewness adaptation coefficient and the voltage standard deviation is calculated to obtain the voltage fluctuation; the difference between the voltage mean and the voltage fluctuation is used as the lower limit of the electrostatic voltage threshold, and the sum of the voltage mean and the voltage fluctuation is used as the upper limit of the electrostatic voltage threshold.

[0030] Based on the method for obtaining the electrostatic voltage threshold of the control and regulation deviation, the electrostatic voltage thresholds of the equipment mechanical characteristics and the ventilation network structural characteristics are calculated respectively.

[0031] Furthermore, methods for obtaining source diagnostic reports include:

[0032] Using the same method as the historical feature database, the ventilation data stream parameters are preprocessed to obtain real-time source parameters, real-time airflow parameters, real-time dust concentration, and real-time electrostatic voltage.

[0033] Using real-time source parameters as rows and real-time airflow parameters as columns, the correlation coefficient between real-time source parameters and real-time airflow parameters is obtained through the Pearson correlation coefficient calculation method, and an association matrix is ​​constructed.

[0034] Extract the parameter pairs in the correlation matrix whose absolute values ​​of each correlation coefficient are greater than the corresponding ventilation system deviation correlation strength threshold, and form a strong correlation set corresponding to the candidate deviation;

[0035] Perform time lag verification on strongly correlated sets to obtain a valid candidate deviation set;

[0036] The valid candidate biases are fused and determined to obtain the intersection of the candidate biases;

[0037] The intersection of candidate deviations is screened with electrostatic voltage assistance to obtain the final source of periodic airflow pulsation and obtain a source diagnosis report.

[0038] Furthermore, methods for verifying time lag include:

[0039] For each strongly correlated parameter pair in the strongly correlated set, the target lag is calculated using a cross-correlation function; candidate deviations whose time lag is not within the corresponding range are eliminated to form a set of valid candidate deviations; if the set of valid candidate deviations is empty, it is determined that no periodic airflow pulsation source matching the time lag characteristic has been detected, the current diagnostic process ends and the process returns to the step of obtaining real-time raw ventilation data; if the set of valid candidate deviations is not empty, the process proceeds to the fusion determination step.

[0040] Furthermore, the methods for determining fusion include:

[0041] The real-time dust concentration is subjected to wavelet threshold denoising and fixed-interval resampling to calculate time-domain and frequency-domain features. The time-domain and frequency-domain features of the real-time dust concentration are used to form a real-time target vector. A standard vector is extracted from the historical feature database, consisting of the time-domain and frequency-domain features of the dust concentration corresponding to the ventilation system deviation. The cosine similarity between the real-time target vector and the standard vector is calculated to obtain the real-time dust concentration similarity. The real-time dust concentration similarity is compared with a similarity threshold to obtain the corresponding ventilation system deviation, forming a dust matching candidate set.

[0042] The intersection of the effective candidate deviation set and the dust matching candidate set is denoted as the candidate deviation intersection. The real-time source parameters corresponding to the ventilation system deviations in the candidate deviation intersection are extracted. Combined with the parameter stability range in the historical feature database, it is determined whether the real-time source parameters exceed the parameter stability range. If they do not exceed the range, the corresponding ventilation system deviations are eliminated to obtain the candidate deviation intersection. If there is a unique ventilation system deviation in the candidate deviation intersection, the corresponding deviation is determined to be the final source of the periodic airflow pulsation. If there are multiple ventilation system deviations in the candidate deviation intersection, the final source of the periodic airflow pulsation is obtained by electrostatic voltage-assisted screening.

[0043] Furthermore, electrostatic voltage-assisted screening methods include:

[0044] Extract the electrostatic voltage threshold corresponding to each ventilation system deviation from the candidate deviation intersection set from the historical feature database;

[0045] If the real-time electrostatic voltage matches only the electrostatic voltage threshold range of a single ventilation system deviation, then the corresponding ventilation system deviation is the ultimate source of periodic airflow pulsations. If multiple matching ventilation system deviations exist, the electrostatic voltages corresponding to the ventilation system deviations are extracted from the historical feature database and standardized to obtain a voltage standard curve. The real-time electrostatic voltage and voltage standard curve are resampled with the same number of sampling points to obtain a real-time voltage sequence and a standard voltage sequence. The difference between the real-time voltage and the standard voltage at the corresponding sampling point is calculated to obtain a voltage deviation sequence. Each value in the voltage deviation sequence is normalized, and the arithmetic mean of the normalized voltage deviation sequence is calculated to obtain the voltage deviation degree. The ventilation system deviation with the smallest voltage deviation degree is selected and determined as the ultimate source of periodic airflow pulsations.

[0046] Compared with the prior art, the technical effects and advantages of the mine ventilation fault diagnosis and early warning method of the present invention are as follows:

[0047] This invention acquires real-time raw ventilation data, processes it using a preset feature extraction method to obtain ventilation data stream parameters, and integrates the differentiated features of three types of ventilation system deviations—control and regulation deviations, equipment mechanical characteristic deviations, and ventilation network structural characteristic deviations—to construct a historical feature database. Based on the real-time ventilation data stream parameters and the historical feature database, it diagnoses the sources of periodic airflow pulsations and generates a source diagnosis report. Finally, based on the report, it performs source-specific processing and graded early warning for periodic airflow pulsations. This invention establishes a precise diagnostic model and source-specific control mechanism for the sources of periodic airflow pulsations through multi-dimensional coupled analysis of real-time ventilation data, the three types of deviation features, and historical operational data. This solves the problem that mine ventilation control technology cannot accurately identify the sources of periodic airflow pulsations, achieving accurate identification of periodic airflow pulsations, graded early warning of risks, and targeted formulation of handling strategies. This improves underground operation safety, ventilation equipment operation reliability, and mine operation stability, meeting the development needs of safe and intelligent operation in the mining industry. Attached Figure Description

[0048] Figure 1 This is a flowchart of the mine ventilation fault diagnosis and early warning method according to an embodiment of the present invention;

[0049] Figure 2 This is a flowchart illustrating the construction of the historical feature library according to an embodiment of the present invention;

[0050] Figure 3 This is a flowchart illustrating the process of obtaining a source diagnostic report according to an embodiment of the present invention. Detailed Implementation

[0051] The technical solutions of the embodiments of the present invention will be described in detail, clearly, and completely below with reference to the accompanying drawings. It should be particularly noted that the specific embodiments described below are only used to better illustrate and explain the technical solutions of the present invention, and are intended to enable those skilled in the art to better understand and implement the present invention, and should not be construed as limiting the scope of protection of the present invention. Without departing from the spirit and substance of the present invention, those skilled in the art can modify, adjust, or make equivalent substitutions based on the content disclosed in the present invention, and these should all be considered within the scope of protection of the present invention.

[0052] Example 1:

[0053] Please see Figure 1 As shown in the figure, this embodiment discloses a method for diagnosing and warning of mine ventilation faults, including:

[0054] Obtain real-time raw ventilation data;

[0055] The raw ventilation data is processed using a preset feature extraction method to obtain ventilation data stream parameters;

[0056] Integrate the differential characteristics of ventilation system deviations to construct a historical feature database; ventilation system deviations include control and regulation deviations, equipment mechanical characteristic deviations, and ventilation network structural characteristic deviations.

[0057] Based on real-time ventilation data stream parameters and historical feature database, the source of periodic airflow pulsation is diagnosed and a source diagnosis report is obtained.

[0058] Based on the source diagnosis report, periodic airflow pulsations are processed by source and classified for early warning.

[0059] Acquire real-time raw ventilation data. Raw ventilation data includes airflow parameters, fan parameters, damper parameters, tunnel structure parameters, dust concentration, and electrostatic voltage; among which, airflow parameters include wind speed and wind pressure, fan parameters include fan speed, vibration acceleration, and current; damper parameters include damper opening and actuator current; tunnel structure parameters include tunnel resistance and cross-sectional changes.

[0060] By installing intrinsically safe ultrasonic wind speed sensors in the straight sections of roadways, near the face of mining operations, and at specific locations in return airways, and by installing differential pressure wind pressure sensors at the air inlets and outlets of ventilation fans, on both sides of main ventilation doors, and at branch nodes of the ventilation network in the mining area, airflow speed and pressure data are obtained.

[0061] Non-contact speed sensors, such as Hall sensors, installed at the fan main shaft coupling, mine explosion-proof vibration sensors on the bearing housing, and current sensors, such as current transformers, integrated in the fan control cabinet, are used to monitor the real-time speed, vibration acceleration, and current of the main fan.

[0062] By installing an intrinsically safe rotary encoder for mining applications at the damper shaft and connecting a non-contact current sensor, such as a current clamp, in series in the damper actuator power circuit, the damper opening degree and actuator current are obtained. The intrinsically safe rotary encoder for mining applications calculates the damper opening degree by measuring the shaft angle; for example, 0 degrees represents fully closed and 100 degrees represents fully open.

[0063] Intrinsically safe laser rangefinders are installed in roadways prone to roof falls, such as the return airway of a longwall face. These laser rangefinders reflect cross-sectional changes in real time and calculate roadway resistance by linking wind speed and wind pressure data.

[0064] The roadway resistance is calculated using the Darcy-Weisbach modified formula. The proportionality coefficient between resistance loss and the square of air volume is defined as roadway resistance. Therefore, in ventilation engineering, the Darcy-Weisbach modified formula can be expressed as the pressure difference between the two ends of the roadway equals the product of the roadway resistance and the square of the air volume. Air volume is the volume of air passing through the roadway cross-section per unit time, equal to the product of the average cross-sectional velocity and the roadway cross-sectional area. Substituting the air volume into the Darcy-Weisbach modified formula yields the roadway resistance. The pressure difference between the two ends of the roadway is obtained using a mine differential pressure wind pressure sensor; the average wind speed of the roadway cross-section is measured using a mine differential pressure wind pressure sensor; and the roadway cross-sectional area is calculated by measuring the roadway width and height using a laser rangefinder sensor.

[0065] Mining laser scattering dust sensors are installed at the breathing zone height of workers in the mining face and at the return airway to collect dust concentration data; intrinsically safe electrostatic sensors are installed on parts of the roadway such as metal supports and fan casings that are prone to static electricity accumulation to collect electrostatic voltage data.

[0066] The raw ventilation data is processed using a preset feature extraction method to obtain ventilation data stream parameters.

[0067] The ventilation data stream parameters include wind speed fluctuations, wind pressure fluctuations, fan vibration parameters, damper opening fluctuations, roadway resistance changes, dust concentration curves, and electrostatic voltage changes. The preset feature extraction method is pre-set during the method design phase based on the mine ventilation mechanism, the physical meaning of ventilation parameters, the operating characteristics of ventilation equipment, and industry-standard calculation criteria. This method is used to extract parameters from the raw ventilation data that characterize changes and anomalies in ventilation status. The pre-setting process is guided by the needs of ventilation safety analysis, selecting feature extraction methods that have clear physical meaning, are computationally stable, and are suitable for time-series data.

[0068] The preset feature extraction methods include, but are not limited to: sliding window statistical feature extraction method, differential feature extraction method, RMS calculation method, peak feature extraction method, and multi-parameter correlation discrimination method. Different types of raw ventilation data will use feature extraction methods that match their variation patterns and engineering significance.

[0069] After collecting the raw ventilation data, the raw ventilation data is preprocessed by noise reduction filtering, missing value completion, and redundancy compression at the downhole edge computing node.

[0070] The ventilation data stream parameters are obtained by calculating the processed raw ventilation data.

[0071] Wind speed fluctuation is used to quantify the overall change in wind speed within a roadway over a short period of time, employing a pre-defined sliding window statistical feature extraction method. The sliding window size (e.g., 10 seconds) is determined based on the sampling frequency and operating condition characteristics of the ventilation data. Preprocessed wind speed data within the sliding window is extracted, and the mean wind speed within the window is calculated. The sum of squared deviations between each preprocessed wind speed data point and the mean wind speed within the window is calculated to obtain the sum of squared wind speed deviations. The variance of the wind speed within the window is calculated by dividing the sum of squared wind speed deviations by the window size. The square root of the variance is then used to obtain the wind speed fluctuation.

[0072] The calculation process for wind pressure fluctuation is the same as that for wind speed fluctuation, except that the pre-processed wind speed in the formula is replaced with the pre-processed wind pressure difference between the two ends of the roadway.

[0073] The fan vibration parameters are extracted using a pre-defined effective value feature extraction method and a multi-parameter correlation discrimination method. The effective vibration value is the core of the fan vibration parameters, reflecting the vibration energy level of the main ventilation fan mechanical system and serving as a key indicator for judging faults such as bearing wear and impeller imbalance. The calculation process is based on the original vibration acceleration signal, combined with correction and verification using the main ventilation fan speed and current operating data.

[0074] Preprocessed data of vibration acceleration, rotational speed and current are obtained; the sign of vibration acceleration represents the vibration direction. After processing the vibration acceleration by squaring, integrating, averaging and taking the square root, the effective value of vibration is obtained, which is proportional to the vibration energy.

[0075] The fan vibration parameters are corrected by combining the speed and the current is used to verify the abnormality of the fan vibration parameters, so as to avoid misjudgment based on a single vibration signal.

[0076] The abnormality of fan vibration parameters is determined by using a speed correction threshold. The ratio of the fan speed to the rated speed is calculated to obtain the fan speed ratio. The product of the fan speed ratio and the vibration safety threshold is then calculated to obtain the speed correction threshold. The vibration safety threshold at the rated speed is set according to industry equipment standards, while the rated speed of the fan is set by the parameters provided by the equipment.

[0077] The essence of fan vibration is forced vibration of the mechanical system. The normal vibration level of a fan is directly related to the impeller excitation frequency, which in turn is proportional to the fan speed. When the speed changes, if a fixed vibration threshold is still used, the excitation frequency will increase as the speed increases, and the vibration amplitude will naturally increase, easily triggering false exceedances; conversely, the impeller excitation frequency will decrease as the speed decreases, and the vibration amplitude will decrease, potentially masking minor faults. Therefore, the speed correction threshold should be dynamically adjusted to match the normal vibration level at the current speed.

[0078] The source of abnormal vibration can be distinguished by current signal, providing a basis for fault location. When the fan current is greater than the rated current, it indicates that the fan is overloaded. In this case, the increased vibration may be caused by load problems such as excessive ventilation resistance rather than mechanical failure. When the fan current is less than the rated current, it indicates that the increased vibration is caused by mechanical problems, such as bearing wear or impeller imbalance.

[0079] The damper opening fluctuation is used to quantify the damper's adjustment stability over a short period of time. The stability of the damper opening directly affects the airflow distribution in the roadway. Frequent fluctuations in the damper opening may cause local airflow to fluctuate, affecting gas concentration control or dust diffusion. The calculation process for damper opening fluctuation is consistent with that for air velocity fluctuation, using the sliding window standard deviation method to obtain the damper opening fluctuation.

[0080] The actuator current helps determine whether the damper opening fluctuation is abnormal, avoiding single misjudgments. When the opening fluctuation and the actuator current fluctuate synchronously, it indicates that the damper opening fluctuation may be caused by frequent start-stop of the actuator, such as control logic parameter delay or sensor signal interference. Optimize the control program or check the signal line. If the opening fluctuates but the current does not, it indicates that the damper opening fluctuation is caused by encoder or data transmission delay issues.

[0081] The changes in roadway resistance are extracted using a pre-defined differential feature extraction method. Roadway resistance change is a core indicator reflecting sudden changes in roadway ventilation status. By quantifying the instantaneous changes in roadway resistance, timely warnings can be issued for roadway blockages, roof falls, and spalling.

[0082] Obtain the roadway resistance values ​​at continuous times from the original ventilation data, extract the roadway resistance at time t and time t-1; calculate the difference between the roadway resistance at time t and time t-1 to obtain the roadway resistance difference; take the absolute value of the roadway resistance difference to obtain the instantaneous resistance change.

[0083] The dust concentration curve employs a pre-defined time-series characteristic representation method. The dust concentration curve is a crucial indicator for determining the source of periodic airflow pulsations. By extracting the dust concentration time series from the pre-processed raw ventilation data and plotting time on the horizontal axis and concentration on the vertical axis, a continuously changing dust concentration curve is generated. The curve clearly displays the instantaneous fluctuations, stable periods, and exceedances in the concentration.

[0084] The electrostatic voltage change value was extracted using a pre-defined sliding window peak feature extraction method. The electrostatic voltage change value is a core indicator for assessing the risk of electrostatic discharge in roadways, used to quantify the fluctuation range of static electricity accumulation. Electrostatic voltage data was extracted from the pre-processed raw ventilation data, and sliding window peak processing was used to determine the window range, consistent with the window size for fluctuations in wind speed and damper opening. The maximum and minimum electrostatic voltage values ​​within the window were calculated. The electrostatic voltage change value was obtained by subtracting the minimum electrostatic voltage value from the maximum value.

[0085] Integrate the differentiated characteristics of ventilation system deviations to construct a historical feature database; ventilation system deviations include control and regulation deviations, equipment mechanical characteristic deviations, and ventilation network structural characteristic deviations.

[0086] Different sources of periodic airflow pulsations in mines exhibit differentiated correlations with ventilation data stream parameters. Control and regulation deviations first trigger fluctuations in damper parameters, which then synchronously propagate to abnormal airflow parameters; deviations in equipment mechanical characteristics first cause abnormal fan parameters, which then lag behind in affecting abnormal airflow parameters; deviations in ventilation network structural characteristics first lead to changes in roadway structural parameters, which then delay inducing abnormal airflow parameters. Based on these differentiated correlation patterns, the source parameters of deviations are defined as fan vibration parameters, damper opening fluctuations, and roadway resistance changes. Specifically, control and regulation deviations correspond to damper opening fluctuations, equipment mechanical characteristic deviations correspond to fan vibration parameters, and ventilation network structural characteristic deviations correspond to roadway resistance changes.

[0087] Construct a historical feature library that includes the correlation strength threshold, time lag range, electrostatic voltage threshold, and parameter stability range of source parameters and ventilation system deviations, as well as the stability range of various ventilation data stream parameters.

[0088] Please see Figure 2 As shown, the methods for constructing a historical feature database include:

[0089] Historical operational data of the mine ventilation system during periods of deviation and normal operation are collected in advance, and the data is combined with maintenance fault records to indicate ventilation system deviations. Historical operational data during periods of deviation include all monitoring data on airflow parameters, fan vibration parameters, damper opening fluctuations, roadway resistance changes, dust concentration, and electrostatic voltage at the time of the deviation, with the ventilation system deviation indicated. Historical operational data during normal operation includes complete data when the ventilation system has no deviation, serving as the benchmark for calculating parameter stability ranges. Historical operational data includes airflow parameters, fan vibration parameters, damper opening fluctuations, roadway resistance changes, dust concentration, and electrostatic voltage.

[0090] Historical operational data is preprocessed, and missing values ​​are filled in using linear interpolation. All target parameters are standardized to eliminate dimensional differences, such as by mapping them to the [0,1] interval through min-max standardization.

[0091] Historical operational data for the periods when deviations occurred were processed according to the ventilation system deviations to obtain the correlation strength threshold, time lag range, similarity threshold, and electrostatic voltage threshold between the source parameters and the ventilation system deviations. Data statistics and feature extraction were performed on historical operational data during normal operation periods to obtain the parameter stability range. The ventilation system deviations, correlation strength thresholds, time lag ranges, similarity thresholds, electrostatic voltage thresholds, and parameter stability ranges were then integrated and entered into a historical feature database, stored using a lightweight database such as SQLite. A periodic update mechanism was set up, for example, recalculating thresholds and standards every six months based on newly added fault cases to optimize the accuracy of the historical feature database, resulting in a well-constructed historical feature database.

[0092] The methods for calculating the association strength threshold include:

[0093] The correlation strength threshold is used to characterize the degree of correlation between the deviation source parameter and the airflow parameter, and is calculated separately for each ventilation system deviation. For historical operating data with labeled ventilation system deviations, the corresponding deviation source parameters and airflow parameters are selected.

[0094] The damper opening fluctuation and airflow parameters are collected at preset sampling intervals to form a time series of damper opening fluctuations. Time series of airflow parameters Where n is the length of the time series, , These are the source parameters and airflow parameters for the i-th timestamp, respectively. , These are the source parameters and airflow parameters for the nth timestamp, respectively.

[0095] Methods for calculating the correlation strength of X and Y using the Pearson correlation coefficient include:

[0096] ;

[0097] r is the Pearson correlation coefficient, with a value range of [-1, 1]. The Pearson correlation coefficient is calculated for all airflow parameters corresponding to damper opening fluctuations in historical operational data to obtain a set of correlation coefficients.

[0098] In mine ventilation systems, many key parameters, such as air volume, air pressure, and regulating valve opening, exhibit direct proportional or linear relationships with the generated electrostatic voltage signals. For example, control deviations often lead to linear changes in ventilation parameters, which in turn cause a linear response in the electrostatic voltage. Therefore, the Pearson correlation coefficient, which measures the degree and direction of the linear relationship between two variables, is used to calculate the correlation strength between source parameters and airflow parameters.

[0099] Using descriptive statistical methods, such as the upper quartile method, the correlation coefficient set is sorted in ascending order to obtain the correlation coefficient sorting set; the correlation coefficient position index of the upper quartile in the correlation coefficient sorting set is calculated; the result of the correlation coefficient position index is rounded up to obtain the specific correlation coefficient position number; the correlation coefficient corresponding to the correlation coefficient position number is the correlation strength threshold for controlling the adjustment bias.

[0100] Based on the method of obtaining the correlation strength threshold of the control and regulation deviation, the correlation strength thresholds of the equipment mechanical characteristic deviation and the ventilation network structural characteristic deviation are calculated respectively.

[0101] Methods for calculating the time lag range include:

[0102] The time lag range is used to quantify the time interval between the abnormal transmission of deviation source parameters to airflow parameters under different ventilation system deviations, and is calculated separately for each ventilation system deviation. For historical operating data with known ventilation system deviations, the corresponding deviation source parameters and airflow parameters are selected.

[0103] Obtain the time series of damper opening fluctuations Time series of airflow parameters Methods that use the time offset between the time series of damper opening fluctuations and the time series of airflow parameters as lag values, and employ cross-correlation functions to analyze the lag relationship between damper opening fluctuations and airflow parameters include:

[0104] ;

[0105] Where k is the lag value; CCF(k) is the lag correlation strength; , These are the means of X and Y, respectively; , Let X and Y be the standard deviations, respectively.

[0106] Calculate the hysteresis correlation strength corresponding to different hysteresis values, and take the hysteresis value corresponding to the maximum hysteresis strength as the target hysteresis value; calculate the cross-correlation function of the airflow parameters corresponding to all damper opening fluctuations in historical operation data to obtain the target hysteresis value set.

[0107] Using descriptive statistical methods, the lower quartile and upper quartile methods are used to construct the time lag range for controlling the adjustment deviation. The target lag set is sorted in ascending order to obtain the lag sort set. The quartile lag position index and the upper quartile lag position index are calculated respectively. The results of the quartile lag position index and the upper quartile lag position index are rounded up to obtain the specific quartile lag position number and the upper quartile lag position number. The target lag corresponding to the quartile lag position number is the lower limit of the time lag range, and the target lag corresponding to the upper quartile lag position number is the upper limit of the time lag range.

[0108] Based on the method of obtaining the time lag range of the control adjustment deviation, the time lag ranges of the equipment mechanical characteristics and the ventilation network structural characteristics are calculated respectively.

[0109] The methods for calculating the similarity threshold include:

[0110] By quantifying the characteristics of dust concentration, similarity thresholds corresponding to deviations in different ventilation systems are established, providing a basis for diagnosing the source of airflow pulsation.

[0111] The preprocessed dust concentration was uniformly resampled, and the wavelet threshold denoising method was used to remove high-frequency interference signals to obtain the time domain signal of dust concentration.

[0112] Temporal features are extracted from the dust concentration time-domain signal. These features include the number of peaks, peak duration, mean dust concentration, variance of dust concentration, and peak dust concentration.

[0113] Methods for obtaining temporal features include:

[0114] Calculate the peak concentration of the dust concentration time-domain signal that exceeds the dust safety threshold per unit time, and obtain the number of peaks; calculate the time from when a single peak exceeds the dust safety threshold to when it falls below the dust safety threshold, and obtain the peak duration; calculate the mean and variance of dust concentration; calculate the difference between the maximum and minimum dust concentration, and obtain the peak dust concentration.

[0115] The dust concentration time-domain signal is converted into the dust concentration frequency-domain signal using Fast Fourier Transform, and the frequency domain features of the dust concentration frequency-domain signal are extracted. The frequency domain features include peak frequency, spectral bandwidth, and dominant frequency amplitude.

[0116] Methods for obtaining frequency domain features include:

[0117] Calculate the frequency with the highest spectral energy in the dust concentration frequency domain signal to obtain the peak frequency; calculate the energy value corresponding to the peak frequency to obtain the main frequency amplitude.

[0118] Extract dust concentrations from the historical feature database that have been marked with ventilation system deviations. Classify these deviations and statistically analyze the time-domain and frequency-domain distributions of dust concentrations when each deviation occurs. Calculate the midpoint value of all features to obtain the standard value of each feature. For example, if the distribution range of the number of peaks is 8 to 12, then the midpoint value of the number of peaks is 10.

[0119] Extract dust concentrations from the historical feature database that have already labeled ventilation system deviations, and hide the corresponding ventilation system deviation labels for the dust concentrations. Extract the temporal and frequency domain features of the dust concentration data, and calculate the cosine similarity between the feature values ​​of the temporal and frequency domain features and the feature standard values ​​to obtain the similarity results. Recover the hidden ventilation system deviation labels, and match the calculated similarity results with the deviation labels to obtain a similarity threshold.

[0120] Methods for calculating cosine similarity include:

[0121] The time-domain and frequency-domain features of dust concentration are combined to form a vector to be determined, and the standard values ​​of the ventilation system deviation are combined to form a standard vector. The vector dimension is consistent with the total number of extracted time-domain and frequency-domain features.

[0122] Normalize the vector to be calculated and the standard vector, such as using Min-Max normalization, to eliminate the influence of differences in the magnitude of different features.

[0123] Calculate the dot product of the vector to be found and the standard vector, calculate the product of the vector magnitudes of the vector to be found and the standard vector, and divide the dot product by the product of the vector magnitudes to obtain the cosine similarity.

[0124] Methods for matching the calculated similarity results with the deviation annotations include:

[0125] The three types of ventilation system deviations—control and regulation deviations, equipment mechanical characteristic deviations, and ventilation network structural characteristic deviations—are addressed separately.

[0126] For each ventilation system deviation category, all dust concentration samples labeled as the corresponding category are collected from historical data, and the cosine similarity results of these dust concentration samples and the corresponding categories are selected to form a set of similarity results for the corresponding categories.

[0127] Perform statistical analysis on the similarity result set, and set the lowest percentile in the similarity result set as the similarity threshold for the corresponding category; for example, set the fifth percentile of the similarity result set as the similarity threshold.

[0128] The process of matching similarity results with deviation labels yields three similarity thresholds corresponding to control and regulation deviations, equipment mechanical characteristic deviations, and ventilation network structural characteristic deviations.

[0129] The methods for calculating the electrostatic voltage threshold include:

[0130] The electrostatic voltage threshold is used to quantify the reasonable boundary of electrostatic voltage under different ventilation system deviations. It is the core benchmark for judging whether the real-time electrostatic voltage matches the specific ventilation system deviation and is calculated separately according to the ventilation system deviation.

[0131] By using the electrostatic voltages of ventilation system deviations already marked in the historical feature database, and classifying them according to ventilation system deviations, we obtain a set of historical voltage data, including a set of control and regulation deviation voltages, a set of equipment mechanical characteristic deviation voltages, and a set of ventilation network structural characteristic deviation voltages.

[0132] The electrostatic voltages of the control and regulation deviation voltage set are sorted in ascending order to obtain a voltage sorting set. The mean voltage and standard deviation of the electrostatic voltages in the voltage sorting set are calculated. The voltage fluctuation is obtained by calculating the product of the preset skewness adaptation coefficient and the voltage standard deviation. The difference between the mean voltage and the voltage fluctuation is used as the lower limit of the electrostatic voltage threshold, and the sum of the mean voltage and the voltage fluctuation is used as the upper limit of the electrostatic voltage threshold. The skewness adaptation coefficient is set according to the distribution characteristics of historical electrostatic voltage data. For example, the skewness adaptation coefficient can be set based on the skewness statistics of the data distribution. When the data distribution is right-skewed, the skewness adaptation coefficient can be set to 1.2 to expand the upper limit boundary, and when it is left-skewed, it can be set to 0.8 to tighten the upper limit boundary. The specific coefficient can be modified based on historical data analysis.

[0133] Based on the method for obtaining the electrostatic voltage threshold of the control and regulation deviation, the electrostatic voltage thresholds of the equipment mechanical characteristics and the ventilation network structural characteristics are calculated respectively.

[0134] Method for calculating the parameter stability range during normal operation:

[0135] The parameter stability range during normal operation is determined by statistical analysis of historical operating data of the ventilation system when there is no deviation, combined with industry standards, to establish the normal fluctuation range of parameters, providing a benchmark for subsequent judgment of whether real-time data is abnormal.

[0136] Arrange each type of parameter in the historical operational data during normal operation periods by timestamp to obtain normal parameter data sequences. For each normal parameter data sequence, calculate the normal mean of the parameter, reflecting the central level of the parameter during normal operation; calculate the normal standard deviation of the parameter, reflecting the degree of fluctuation of the parameter during normal operation.

[0137] The stable range of a parameter is determined based on a first confidence level. The corresponding standard normal variable value is obtained by querying the standard normal distribution table based on the first confidence level. The product of the standard normal variable value and the normal standard deviation is calculated to obtain the normal fluctuation. The difference between the normal mean and the normal fluctuation value is used as the lower limit of the stable range of the parameter, and the sum of the normal mean and the normal fluctuation value is used as the upper limit of the stable range of the parameter. The first confidence level is set based on industry standard requirements, the sample size, and the distribution of the parameter data.

[0138] The calculated stable range of parameters is compared with the relevant industry standards for coal mine ventilation. If the upper limit of the stable range of a certain parameter exceeds the normal threshold specified in the standard, the upper limit of the stable range of that parameter is adjusted to the industry standard threshold; if the lower limit is lower than the normal threshold specified in the standard, it is adjusted to the industry standard threshold.

[0139] The parameter stability range during normal operation includes the airflow parameter stability range, the source parameter stability range, the electrostatic voltage safety range, and the dust concentration safety range.

[0140] Based on real-time ventilation data stream parameters and historical feature database, the source of periodic airflow pulsations is diagnosed, and a source diagnosis report is obtained.

[0141] Please see Figure 3 As shown, methods for obtaining source diagnostic reports include:

[0142] Using a method consistent with the historical feature database, ventilation data stream parameters are preprocessed to obtain real-time source parameters, real-time airflow parameters, real-time dust concentration, and real-time electrostatic voltage. The correlation strength threshold, time lag range, electrostatic voltage threshold, and parameter stability range between source parameters and ventilation system deviations are extracted from the historical feature database.

[0143] The correlation between real-time source parameters and real-time airflow parameters is quantified. Using real-time source parameters as rows and real-time airflow parameters as columns, the correlation coefficient is obtained through the Pearson correlation coefficient calculation method, and a correlation matrix is ​​constructed.

[0144] For each correlation coefficient in the correlation matrix, determine whether the absolute value of the correlation coefficient is greater than the correlation strength threshold of the corresponding ventilation system deviation. When the absolute value of the correlation coefficient is greater than the correlation strength threshold of the corresponding ventilation system deviation, the parameter pair consisting of the real-time source parameter and the real-time airflow parameter corresponding to the correlation coefficient is included in the strong correlation set corresponding to the ventilation system deviation, forming a strong correlation set corresponding to the candidate ventilation system deviation. If the strong correlation set is empty, it means that there is no obvious correlation anomaly in the real-time data, and it is determined that the ventilation system is operating without deviation. If the strong correlation set is not empty, the strong correlation set is verified for time lag to obtain a valid candidate deviation set. The valid candidate deviations are fused and determined to obtain the intersection of candidate deviations. If there is a unique ventilation system deviation in the intersection of candidate deviations, it is determined that the deviation is the final source of the periodic airflow pulsation. If there are multiple ventilation system deviations in the intersection of candidate deviations, the intersection of candidate deviations is screened with electrostatic voltage assistance to obtain the final source of the periodic airflow pulsation and obtain a source diagnosis report.

[0145] Methods for verifying time lag include:

[0146] For each strongly correlated parameter pair in the strongly correlated set, the target lag is calculated using a cross-correlation function. The target lag is compared with the time lag range of the corresponding ventilation system deviation, and candidate ventilation system deviations whose time lag is not within the corresponding range are eliminated to form a valid candidate deviation set. If the valid candidate deviation set is empty, it is determined that no periodic airflow pulsation source matching the time lag characteristics has been detected, the current diagnostic process ends, and the process returns to the step of obtaining real-time raw ventilation data; if the valid candidate deviation set is not empty, the process proceeds to the fusion determination step.

[0147] The methods for determining fusion include:

[0148] Wavelet threshold denoising and fixed-interval resampling are performed on the real-time dust concentration to calculate its time-domain and frequency-domain features. The time-domain and frequency-domain features of the real-time dust concentration are combined to form a real-time target vector. The standard vector in the historical feature library is extracted, and the cosine similarity between the real-time target vector and the standard vector is calculated to obtain the real-time dust concentration similarity. The real-time dust concentration similarity is combined with the similarity threshold to obtain the corresponding ventilation system deviation, forming a dust matching candidate set.

[0149] The intersection of the effective candidate deviation set and the dust matching candidate set is denoted as the candidate deviation intersection. The real-time source parameters corresponding to the ventilation system deviations in the candidate deviation intersection are extracted. Combined with the parameter stability range in the historical feature library, it is determined whether the real-time source parameters exceed the parameter stability range. If they do not exceed the range, the corresponding ventilation system deviations are removed to obtain the candidate deviation intersection.

[0150] Electrostatic voltage-assisted screening methods include:

[0151] The preprocessed real-time electrostatic voltage is extracted. The electrostatic voltage threshold corresponding to each ventilation system deviation in the intersection of candidate deviations is extracted from the historical feature database. If the real-time electrostatic voltage change matches only the electrostatic voltage threshold range of one ventilation system deviation, then that ventilation system deviation is the ultimate source of the periodic airflow pulsation. If multiple matching ventilation system deviations exist, the electrostatic voltages corresponding to the ventilation system deviations are extracted from the historical feature database and standardized to obtain a voltage standard curve. The real-time electrostatic voltage and voltage standard curve are resampled with the same number of sampling points to obtain a real-time voltage sequence and a standard voltage sequence. The difference between the real-time voltage and the standard voltage at the corresponding sampling point is calculated to obtain the voltage deviation. The voltage deviation sequence is normalized. This normalization process includes: using the difference between the maximum and minimum voltage values ​​in the standard voltage sequence corresponding to the ventilation system deviation as a normalization factor, dividing each voltage deviation value by the normalization factor to obtain a normalized voltage deviation sequence, thus eliminating the influence of voltage amplitude differences under different ventilation system deviations; calculating the arithmetic mean of the normalized voltage deviation sequence to obtain the voltage deviation degree of the corresponding ventilation system deviation; comparing the voltage deviation degrees corresponding to each ventilation system deviation, selecting the ventilation system deviation with the smallest voltage deviation degree, and determining it as the ultimate source of periodic airflow pulsations.

[0152] Methods for obtaining voltage standard curves include:

[0153] Extract the historical voltage data set, calculate the mean electrostatic voltage change within the set according to the timestamp alignment principle, arrange the mean electrostatic voltage change of all timestamps by timestamp, and obtain the voltage standard curve with timestamp as the horizontal axis and mean electrostatic voltage change as the vertical axis.

[0154] Output source diagnostic report, including final source, strong correlation set, effective candidate deviation set, dust matching candidate set, intersection of candidate deviations and voltage deviation degree.

[0155] Based on the source diagnosis report, periodic airflow pulsations are processed by source and classified for early warning.

[0156] Based on the final source determination conclusion of the source diagnosis report, the corresponding problem links are located, such as the damper control system, main ventilation fan equipment, and ventilation tunnel structure; combined with the dust matching candidate set, the treatment priority is determined, and deviations that directly aggravate dust exceeding the standard and static electricity accumulation are treated first.

[0157] The core monitoring object is the set of strongly correlated parameters. For example, when the diagnosis is control and regulation deviation, the focus is on tracking the strongly correlated parameter pairs of damper opening fluctuation. When the diagnosis is equipment mechanical characteristic deviation, the focus is on monitoring the strongly correlated parameter pairs of fan vibration parameters. This ensures that the processing operation focuses on parameters directly related to the pulsation. Based on the effective candidate deviation set, source parameters are processed separately.

[0158] During the source-specific treatment process, electrostatic voltage is used as a real-time auxiliary indicator. Changes in electrostatic voltage are monitored synchronously during the treatment process. Combined with the safe range of electrostatic voltage, the effect of the treatment measures on mitigating the risk of pulsation is initially judged. Finally, the fluctuation range of airflow parameters before and after treatment is compared. Combined with the safe range of dust concentration, it is determined whether the airflow parameters are within the stable range of airflow parameters to confirm whether the airflow pulsation has been reduced after treatment. If the airflow parameters and dust concentration of the ventilation system are still not within the stable range of parameters, the strongly correlated parameter pairs are re-examined based on the correlation matrix, and the treatment method is optimized until the pulsation is effectively controlled.

[0159] By defining predefined warning levels, and based on real-time monitoring data from pulsation source diagnostic reports, parameter stability ranges, and processing procedures, tiered warnings and dynamic adjustments are triggered, while corresponding warning response measures are taken to ensure that risks are controllable.

[0160] Source separation processing methods include:

[0161] If the ultimate source is determined to be control adjustment deviation, the control system parameter calibration method is adopted, such as optimizing PID adjustment parameters to correct the lag and overshoot problems of damper opening fluctuation;

[0162] If the ultimate cause is determined to be a deviation in the mechanical characteristics of the equipment, mechanical component inspection and maintenance methods should be adopted, such as adjusting the fan bearing clearance, repairing impeller wear, and eliminating periodic fluctuations in speed and vibration.

[0163] If the ultimate cause is determined to be a deviation in the structural characteristics of the ventilation network, network structure optimization methods should be adopted, such as clearing blockages in the roadway, reinforcing the support of the roof fall section, and stabilizing the roadway resistance and cross-sectional dimensions.

[0164] The methods for classifying warning levels include:

[0165] The warning levels are divided based on the logic of the transmission path, scope of impact, risk reversibility, and degree of safety threat of ventilation system deviations. Control and regulation deviations are classified as Level 3 warnings because the risk is local, controllable, and reversible. Its core characteristic is that it first causes fluctuations in damper parameters, which then lead to local fluctuations in airflow parameters. The impact is limited to a single roadway, the treatment cost is low and it is reversible, and the impact on safety risks is small. Equipment mechanical characteristic deviations are classified as Level 2 warnings because the risk is diffuse, requires intervention, and is semi-reversible. Its essence is that abnormal fan parameters cause system-level airflow fluctuations, affecting the entire mining area or the entire mine. Treatment requires shutdown intervention, but performance can be restored. The risk is more diffuse than that of control and regulation deviations. Ventilation network structural characteristic deviations are classified as Level 1 warnings because the risk is global, fatal, and irreversible. Changes in roadway structural parameters induce global airflow fluctuations, which can easily lead to irreversible risks. The treatment cost is high and some damage is irreversible, directly triggering safety risks.

[0166] The methods for tiered early warning and dynamic adjustment include:

[0167] Once the source diagnostic report clearly identifies the final ventilation system deviation, the real-time monitoring data is simultaneously compared with the corresponding level's judgment criteria. If the criteria are met, the corresponding level's warning is immediately triggered.

[0168] During the source-based processing, corresponding real-time ventilation data stream parameters are collected, such as damper opening fluctuations during control adjustment deviation processing, fan vibration parameters during equipment mechanical characteristic deviation processing, and roadway resistance during ventilation network structural characteristic deviation processing. If the real-time ventilation data stream parameters meet the lower-level warning conditions within the preset period, the warning level is lowered; if the data continues to exceed the standard, the warning level is raised.

[0169] Once the source is processed and the real-time monitoring data returns to a stable range of parameters, and there are no abnormalities within a continuous preset period, the warning is lifted.

[0170] Early warning response measures include:

[0171] The three-level early warning response system remotely optimizes the PID adjustment parameters of the damper through downhole edge nodes, tracks the strong correlation between damper opening fluctuations and air pressure parameters in real time, and pushes monitoring data to the regional monitoring center in real time.

[0172] In response to the Level II early warning, dispatchers adjusted the wind turbine bearing clearance and cleaned the dust from the impeller on-site. During this process, the wind turbine current was kept within a safe current range, and the progress was pushed to both the regional center and the ground control center in real time. The safe current range was set to prevent the fault from worsening, protect the core components of the wind turbine, and ensure the safety of maintenance operations.

[0173] In the first-level early warning response, mining operations in the affected area shall be immediately stopped, and the blockage in the roadway shall be cleared or the section of roof collapse shall be reinforced. If the electrostatic voltage exceeds the safe range, a local power outage shall be triggered, and parameters and emergency progress shall be pushed to the mine-level dispatch room and safety supervision department in real time until the risk is eliminated.

[0174] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0175] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for diagnosing and predicting ventilation failures in mines, characterized in that, include: Obtain real-time raw ventilation data; The raw ventilation data is processed using a preset feature extraction method to obtain ventilation data stream parameters; Integrate the differential characteristics of ventilation system deviations to construct a historical feature database; ventilation system deviations include control and regulation deviations, equipment mechanical characteristic deviations, and ventilation network structural characteristic deviations. Based on real-time ventilation data stream parameters and historical feature database, the source of periodic airflow pulsation is diagnosed and a source diagnosis report is obtained. Historical operational data of the mine ventilation system during periods of deviation and normal operation are collected in advance. The historical operational data during periods of deviation include complete data when the deviation occurs, with the deviation marked. The historical operational data during periods of normal operation include complete data when the ventilation system has no deviation. The historical operational data includes airflow parameters, fan vibration parameters, damper opening fluctuations, roadway resistance changes, dust concentration, and electrostatic voltage. Historical operational data is preprocessed by using linear interpolation to fill in missing values ​​and standardizing the data to eliminate dimensional differences. For historical operational data during periods of ventilation system deviation, the data is processed according to the specific deviation to obtain the correlation strength threshold, time lag range, similarity threshold, and electrostatic voltage threshold for each deviation. Data statistics and feature extraction are performed on historical operational data during normal operation periods to obtain parameter stability ranges. Finally, the ventilation system deviation, correlation strength threshold, time lag range, similarity threshold, electrostatic voltage threshold, and parameter stability range are entered into a historical feature library to obtain the constructed historical feature library. The source parameters of ventilation system deviations are defined as fan vibration parameters, damper opening fluctuations, and roadway resistance changes; among them, the source parameter corresponding to control and regulation deviations is damper opening fluctuations, the source parameter corresponding to equipment mechanical characteristic deviations is fan vibration parameters, and the source parameter corresponding to ventilation network structural characteristic deviations is roadway resistance changes. The damper opening fluctuation and airflow parameters are collected at preset sampling intervals to form time series of source parameters and airflow parameters of control and regulation deviation. The Pearson correlation coefficient of the time series of source parameters and airflow parameters is used as the correlation strength to obtain a set of correlation coefficients. Descriptive statistical methods are used to obtain the correlation strength threshold of control and regulation deviation. Based on the method of obtaining the correlation strength threshold of the control and regulation deviation, the correlation strength thresholds of the equipment mechanical characteristic deviation and the ventilation network structural characteristic deviation are calculated respectively. The time series of damper opening fluctuation and airflow parameters are obtained. The time offset interval between the time series of damper opening fluctuation and airflow parameters is used as the lag value. The lag correlation strength corresponding to different lag values ​​is calculated using the cross-correlation function. The lag value corresponding to the maximum lag correlation strength is taken as the target lag value. Cross-correlation functions were performed on all damper opening fluctuations in historical operating data to obtain the target lag set; descriptive statistical methods were used to obtain the time lag range of control and regulation deviations. Based on the method of obtaining the time lag range of the control adjustment deviation, the time lag ranges of the equipment mechanical characteristic deviation and the ventilation network structural characteristic deviation are calculated respectively. The preprocessed dust concentration was uniformly resampled, and the wavelet threshold denoising method was used to remove high-frequency interference signals to obtain the time domain signal of dust concentration. Extract the time-domain features of the dust concentration time-domain signal; use Fast Fourier Transform to convert the dust concentration time-domain signal into a dust concentration frequency-domain signal, and extract the frequency-domain features of the dust concentration frequency-domain signal. Extract dust concentrations from the historical feature database that have been marked with ventilation system deviations. Classify the ventilation system deviations and statistically analyze the time-domain and frequency-domain distributions of dust concentrations when each ventilation system deviation occurs. Calculate the feature standard values. Extract dust concentrations from the historical feature database that have been labeled with ventilation system deviations, and hide the corresponding ventilation system deviation labels for the dust concentrations; extract the temporal and frequency domain features of the dust concentration data, and calculate the cosine similarity between the feature values ​​of the temporal and frequency domain features and the feature standard values ​​to obtain the similarity results; restore the hidden ventilation system deviation labels, and match the calculated similarity results with the deviation labels to obtain the similarity threshold; Extract the electrostatic voltages of ventilation system deviations marked in the historical feature database, classify them according to ventilation system deviations, and obtain the control and regulation deviation voltage set, the equipment mechanical characteristic deviation voltage set, and the ventilation network structural characteristic deviation voltage set; Sort the electrostatic voltages of the control regulation deviation voltage set in ascending order to obtain the voltage sorting set; calculate the voltage mean and voltage standard deviation of the electrostatic voltages in the voltage sorting set; The voltage fluctuation is obtained by calculating the product of the preset skewness adaptation coefficient and the voltage standard deviation. The difference between the average voltage and the voltage fluctuation is used as the lower limit of the electrostatic voltage threshold, and the sum of the average voltage and the voltage fluctuation is used as the upper limit of the electrostatic voltage threshold. Based on the method for obtaining the electrostatic voltage threshold of the control and regulation deviation, the electrostatic voltage thresholds of the equipment mechanical characteristic deviation and the ventilation network structural characteristic deviation are calculated respectively. Based on the source diagnosis report, periodic airflow pulsations are processed by source and classified for early warning.

2. The method for diagnosing and warning of mine ventilation faults according to claim 1, characterized in that, Methods for obtaining source diagnostic reports include: Using the same method as the historical feature database, the ventilation data stream parameters are preprocessed to obtain real-time source parameters, real-time airflow parameters, real-time dust concentration, and real-time electrostatic voltage. Using real-time source parameters as rows and real-time airflow parameters as columns, the correlation coefficient between real-time source parameters and real-time airflow parameters is obtained through the Pearson correlation coefficient calculation method, and an association matrix is ​​constructed. Extract the parameter pairs in the correlation matrix whose absolute values ​​of each correlation coefficient are greater than the corresponding correlation strength threshold of the ventilation system deviation, and form a set of strong correlations corresponding to the ventilation system deviation; Perform time lag verification on strongly correlated sets to obtain a valid candidate deviation set; The effective candidate deviation set is fused and determined to obtain the intersection of the candidate deviations; The intersection of candidate deviations is screened with electrostatic voltage assistance to obtain the final source of periodic airflow pulsation and obtain a source diagnosis report.

3. The method for diagnosing and warning of mine ventilation faults according to claim 2, characterized in that, Methods for verifying time lag include: For each strongly correlated parameter pair in the strongly correlated set, the target lag is calculated using a cross-correlation function; candidate deviations whose time lag is not within the corresponding range are eliminated to form a set of valid candidate deviations; if the set of valid candidate deviations is empty, it is determined that no periodic airflow pulsation source matching the time lag characteristic has been detected, the current diagnostic process ends and the process returns to the step of obtaining real-time raw ventilation data; if the set of valid candidate deviations is not empty, the process proceeds to the fusion determination step.

4. The method for diagnosing and warning of mine ventilation faults according to claim 2, characterized in that, The methods for determining fusion include: The real-time dust concentration is denoised using wavelet thresholding and resampled at fixed intervals to calculate its time-domain and frequency-domain features. These features are then combined to form a real-time target vector. A standard vector is extracted from the historical feature database, composed of the time-domain and frequency-domain features of the dust concentration corresponding to ventilation system deviations. The cosine similarity between the real-time target vector and the standard vector is calculated to obtain the real-time dust concentration similarity. Finally, the real-time dust concentration similarity is compared with a similarity threshold to obtain the corresponding ventilation system deviation, forming a dust matching candidate set. The intersection of the effective candidate deviation set and the dust matching candidate set is denoted as the candidate deviation intersection. The real-time source parameters corresponding to the ventilation system deviations in the candidate deviation intersection are extracted. Combined with the parameter stability range in the historical feature library, it is determined whether the real-time source parameters exceed the parameter stability range. If they do not exceed the range, the corresponding ventilation system deviations are removed to obtain the candidate deviation intersection. If there is a unique ventilation system deviation in the candidate deviation intersection, the corresponding deviation is determined to be the final source of the periodic airflow pulsation.

5. The method for diagnosing and warning of mine ventilation faults according to claim 2, characterized in that, Electrostatic voltage-assisted screening methods include: Extract the electrostatic voltage threshold corresponding to each ventilation system deviation from the candidate deviation intersection set from the historical feature database; If the real-time electrostatic voltage matches only the electrostatic voltage threshold range of a single ventilation system deviation, then the corresponding ventilation system deviation is the ultimate source of periodic airflow pulsations. If multiple matching ventilation system deviations exist, the electrostatic voltages corresponding to the ventilation system deviations are extracted from the historical feature database and standardized to obtain a voltage standard curve. The real-time electrostatic voltage and voltage standard curve are resampled with the same number of sampling points to obtain a real-time voltage sequence and a standard voltage sequence. The difference between the real-time voltage and the standard voltage at the corresponding sampling point is calculated to obtain a voltage deviation sequence. Each value in the voltage deviation sequence is normalized, and the arithmetic mean of the normalized voltage deviation sequence is calculated to obtain the voltage deviation degree. The ventilation system deviation with the smallest voltage deviation degree is selected and determined as the ultimate source of periodic airflow pulsations.

Images