A coal mine underground gas sensor sampling signal processing method
By employing hybrid sampling and cascaded filtering techniques in the signal processing of underground gas sensors in coal mines, the signal interference problem under complex working conditions in coal mines has been solved, achieving high reliability, low latency, and high accuracy in gas concentration monitoring.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-04-23
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies cannot effectively process gas sensor signals under complex working conditions in coal mines, leading to false alarms and malfunctions, and failing to achieve highly reliable, low-latency, and high-precision gas concentration monitoring.
A five-stage cascaded filtering architecture is constructed by combining hybrid sampling and random sampling, including IIR notch filtering, amplitude limiting filtering, median filtering, Kalman filtering, and wavelet transform filtering, to accurately remove power frequency interference, impulse noise, and broadband random noise in a layered manner.
It effectively eliminates power frequency interference, impulse noise, and broadband random noise, ensuring the accuracy and real-time nature of carbon dioxide concentration changes and improving the reliability and accuracy of monitoring.
Smart Images
Figure CN122409959A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a method for processing sampling signals from underground gas sensors in coal mines, belonging to the field of coal mine safety monitoring technology. Background Technology
[0002] Currently, coal is my country's primary energy source, and safe production in coal mines is of paramount importance to the industry's development. Real-time monitoring of the concentrations of toxic, harmful, and asphyxiating gases such as methane, carbon monoxide, and carbon dioxide underground is a core technological means to prevent gas explosions, poisoning, asphyxiation, and the spread of fires. As monitoring terminals, the accuracy and reliability of gas sensor signal acquisition directly determine the effectiveness of early warning systems.
[0003] Coal mines operate under harsh conditions characterized by strong interference, high noise levels, and non-stationarity. Sensor sampling signals commonly suffer from the following problems: 1. The underground power supply system is 50Hz AC. The operation of high-voltage cables, frequency converters, and motors generates 50Hz fundamental and 100Hz harmonic interference, causing periodic fluctuations in the sampled values and severely masking the true gas concentration signal.
[0004] 2. The start-up, shutdown, and impact of equipment such as coal mining machines, tunneling machines, pneumatic picks, hydraulic supports, and circuit breakers generate spike pulse noise with sudden amplitude changes and short durations, which can easily cause false alarms and malfunctions in the system.
[0005] 3. Pneumatic equipment operation, dust impact, and pipeline vibration generate non-stationary random noise of 1Hz~10kHz. Traditional Fourier transform is only applicable to stationary signals and cannot achieve effective separation.
[0006] 4. Existing sensors mostly use fixed-period, equally spaced sampling, and have not been optimized for downhole co-frequency interference, which easily leads to defects such as signal aliasing, poor dispersion, and synchronous noise amplification.
[0007] 5. Conventional filtering uses only a single algorithm or a simple combination, which cannot simultaneously meet the four major requirements of power frequency suppression, pulse elimination, random noise cancellation, and preservation of true signal fidelity. It is prone to problems such as signal lag, distortion, and incomplete filtering.
[0008] While existing technologies employ single or combined filtering methods such as notch filtering, median filtering, Kalman filtering, and wavelet filtering to process industrial sensor signals, none of them have been designed in an integrated, hierarchical, and parameterized manner to address the mixed noise characteristics and slowly varying carbon dioxide concentration in underground coal mines. They cannot form a complete anti-interference solution from sampling mechanism to filtering process, making it difficult to meet the requirements of high reliability, low latency, and high precision gas monitoring in underground coal mines.
[0009] Therefore, developing a gas sensor sampling signal processing method that is adapted to the complex working conditions of underground coal mines, optimizes the sampling mechanism, ensures accurate cascade filtering, and maintains high signal fidelity has become a pressing technical challenge in this field. Summary of the Invention
[0010] The technical problem to be solved by this invention is to overcome the shortcomings of the prior art and provide a method for processing sampling signals from underground gas sensors in coal mines. By optimizing the sampling timing, using a combination of hybrid sampling and random sampling, and constructing a five-level cascaded filtering architecture, the method achieves layered and accurate removal of power frequency interference, impulse noise, and broadband random noise, while ensuring that the actual changes in carbon dioxide concentration are not delayed or distorted, thereby improving the reliability and accuracy of monitoring.
[0011] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: A method for processing sampling signals from a gas sensor in an underground coal mine includes the following steps: Step S1: Perform sensor data sampling control and sample the data signal output by the sensor; Step S2: Perform IIR notch filtering on the sensor sampled data to remove power frequency interference introduced by the coal mine power grid; Step S3: Perform amplitude limiting filtering on the extreme impulse noise in the sensor sampling data; Step S4: Perform median filtering on the smoothed residual pulses in the sensor sampling data after amplitude limiting filtering; Step S5: Perform Kalman filtering on the residual low-frequency random noise in the sensor sampling data after median filtering; Step S6: Remove broadband random noise from the sensor sampling data after Kalman filtering by wavelet transform filtering.
[0012] Furthermore, in step S1, sensor data sampling control is performed and the data signal output by the sensor is sampled, specifically including the following steps: The ADC sampling is triggered by a timer inside the MCU processor, with a sampling interval of 1ms. In addition, dense sampling was performed near the sampling point, once every 10us for a total of 10 times; The data signal output by the sensor is sampled to obtain the sensor sampling data.
[0013] Furthermore, in step S2, the sensor sampling data is subjected to IIR notch filtering to remove power frequency interference introduced by the coal mine power grid, specifically including the following steps: Design a 50Hz notch filter to remove 50Hz power frequency interference from sensor sampling data. The expression for the 50Hz notch filter is as follows: ; Design a 100Hz notch filter to remove 100Hz harmonic interference from sensor sampling data. The expression for the 100Hz notch filter is as follows: .
[0014] Furthermore, in step S3, the extreme impulse noise in the sensor sampling data is subjected to amplitude limiting filtering, which specifically includes the following steps: An amplitude threshold is set based on the maximum rate of change of the concentration of the detected gas. If the difference between the current sampled value and the previous filtered value is greater than the amplitude threshold, it is determined to be a broadband spike pulse, and the previous filtered value is used to replace the current sampled value; if the difference between the current sampled value and the previous filtered value is less than or equal to the amplitude threshold, the current sampled value is retained.
[0015] Furthermore, in step S4, median filtering is performed on the smoothed residual pulses in the sensor sampling data after amplitude limiting filtering, specifically including the following steps: Construct a 3-point sliding window [x (n-1) ,x (n) ,x (n+1) The median of the three data points is taken as the current output, and residual pulses are removed to smooth them out.
[0016] Furthermore, in step S5, Kalman filtering is performed on the residual low-frequency random noise in the sensor sampling data after median filtering, specifically including the following steps: Establish the state transition equation, the expression of which is as follows: X (n) =A・X (n-1) +W (n) Where A=1, W (n) This is process noise; Establish the observation equation, the expression of which is as follows: Z (n) =H・X (n) +V (n) Where H=1, V (n) To observe noise; Residual low-frequency random noise is removed through prediction, updating, and gain calculation iterations.
[0017] Furthermore, in step S6, the broadband random noise in the sensor sampling data after Kalman filtering is removed by wavelet transform filtering, specifically including the following steps: The db4 wavelet basis function is used to decompose the sensor sampled data signal to obtain the first high-frequency layer d1, the second high-frequency layer d2, the third high-frequency layer d3, and the low-frequency layer a3; Soft thresholding is performed on the first high-frequency layer d1, the second high-frequency layer d2, and the third high-frequency layer d3. Coefficients smaller than the threshold are set to zero, while large coefficients are retained. The low-frequency layer a3 is retained; The sensor sampling data signal is then reconstructed through inverse wavelet transform.
[0018] By adopting the above technical solution, the present invention has the following beneficial effects: 1. This invention employs 1ms equal-interval sampling and 10us×10 times dense sampling, combined with mixed sampling and random sampling, to effectively avoid co-channel interference and signal aliasing, improve data discreteness and sampling integrity, and reduce noise introduction from the source.
[0019] 2. This invention addresses downhole 50 / 100Hz power frequency, frequency conversion pulses, and non-stationary broadband random noise by achieving layered separation and step-by-step purification: IIR notch filtering suppresses power frequency interference with almost no attenuation of useful signals; amplitude limiting filtering removes extreme pulse noise; median filtering quickly eliminates residual low-frequency random noise; Kalman filtering removes smooth residual pulses; and wavelet transform filtering removes broadband random noise, achieving a separation effect superior to traditional Fourier transform.
[0020] 3. This invention constrains the filtering threshold and window with the rate of change of gas concentration, which not only achieves strong noise suppression, but also does not lose the true concentration change information, avoids response lag caused by excessive filtering, and meets the needs of rapid early warning for safety monitoring.
[0021] 4. This invention forms a complete closed loop from sampling triggering, timing control, sampling method to filtering process. It is specially customized for the complex environment of underground coal mines and can be directly embedded into existing mine gas sensor controllers. It requires little hardware modification and is easy to implement in engineering.
[0022] 5. This invention effectively reduces false alarm rate and false alarm rate, improves signal-to-noise ratio and concentration measurement accuracy, and provides reliable data support for coal mine underground fire early warning, gas control and ventilation optimization. Attached Figure Description
[0023] Figure 1 This is a flowchart of the coal mine underground gas sensor sampling signal processing method of the present invention. Detailed Implementation
[0024] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings.
[0025] like Figure 1As shown in the figure, this embodiment provides a method for processing sampling signals from a gas sensor in a coal mine, including the following steps: Step S1: Perform sensor data sampling control and sample the data signal output by the sensor. Specifically: This embodiment uses an STM32 series processor, and triggers ADC sampling through the timer inside the MCU processor. The sampling interval is 1ms to ensure the discreteness of the data.
[0026] In addition, dense sampling was performed 10 times every 10µs near the sampling point to increase local data density.
[0027] The sensor output signal is sampled to obtain sensor sampling data. However, the sensor sampling data signal contains environmental interference and electronic noise, so it is necessary to filter and reduce the noise. Actual noise in coal mines is mostly mixed noise; this embodiment achieves comprehensive noise reduction through cascaded filtering.
[0028] Step S2: Perform IIR notch filtering on the sensor sampled data to remove power frequency interference introduced by the coal mine power grid. Specifically: The 50Hz power frequency and 100Hz harmonic interference introduced by the coal mine power grid can cause periodic fluctuations in the sampled values, especially when the sensor is close to high-voltage cables (such as the power cable of the coal mining machine). Precise suppression of specific frequencies is necessary. This embodiment designs a second-order IIR notch filter to address the 50Hz power frequency and 100Hz harmonic interference. The IIR notch filter, through a "pole-zero configuration," achieves deep attenuation (≥40dB) at the center frequencies of 50Hz and 100Hz, while offering almost no attenuation for other frequency signals (especially low-frequency carbon dioxide concentration signals).
[0029] The standard form of the second-order notch transfer function is: in, , The center frequencies are 50Hz and 100Hz. To determine the sampling rate, the sampling interval in this embodiment is 1ms. That is, 1000Hz; ρ is the pole radius (0 < ρ < 1), which determines the notch bandwidth and depth.
[0030] Design a 50Hz notch filter to remove 50Hz power frequency interference from sensor sampling data. The expression for the 50Hz notch filter is as follows: Take ρ=0.98 (bandwidth approximately 6Hz, satisfying ≥40dB attenuation); get: ; Design a 100Hz notch filter to remove 100Hz harmonic interference from sensor sampling data. The expression for the 100Hz notch filter is as follows: Take ρ = 0.98; , , get: .
[0031] Step S3: Perform amplitude limiting and filtering on the extreme impulse noise in the sensor sampling data. Specifically: The impact noise from pneumatic picks, the pulse interference from circuit breakers, and the broadband spike pulses from frequency converters can cause false alarms. These broadband spike pulses need to be quickly eliminated while retaining the true concentration changes.
[0032] An amplitude threshold is set based on the maximum rate of change of the concentration of the detected gas. If the difference between the current sampled value and the previous filtered value is greater than the amplitude threshold, it is determined to be a broadband spike pulse, and the previous filtered value is used to replace the current sampled value; if the difference between the current sampled value and the previous filtered value is less than or equal to the amplitude threshold, the current sampled value is retained.
[0033] Amplitude limiting filtering can quickly eliminate excessively large amplitude pulses (i.e., broadband spike pulses), preventing the median filtering window from being occupied by extreme values in the next step.
[0034] Step S4: Perform median filtering on the smoothed residual pulses in the sensor sampling data after amplitude limiting filtering. Specifically: Construct a 3-point sliding window [x (n-1) ,x (n) ,x (n+1) The median of the three data points is taken as the current output. Residual small-amplitude pulses are removed (i.e., residual pulses are smoothed out) without affecting the actual gradual change in gas concentration.
[0035] Step S5: Perform Kalman filtering on the residual low-frequency random noise in the sensor sampling data after median filtering. Specifically: Even after median filtering, low-frequency random noise remains in the sensor sampling data. Kalman filtering, through iterative system state estimation and measurement updates, smooths the sampling data signal in the sense of minimum mean square error.
[0036] The state transition equation is established, and its expression is as follows: X (n) =A・X (n-1) +W (n) Where A=1, W (n) This is process noise; Establish the observation equation, the expression of which is as follows: Z (n) =H・X (n) +V (n) Where H=1, V (n) To observe noise; By iteratively predicting, updating, and calculating gain, minimum mean square error smoothing is achieved, and residual low-frequency random noise is removed.
[0037] Step S6: Remove broadband random noise from the sensor sampling data after Kalman filtering using wavelet transform filtering. Specifically: The mid-to-high frequency broadband noise (1~10kHz) of pneumatic equipment and the low-frequency random noise (10~100Hz) of dust impact are non-stationary random noises. Traditional Fourier filtering cannot effectively separate them (the useful signal and noise frequencies may overlap), and separation can only be achieved through joint time-domain and frequency-domain analysis.
[0038] The db4 wavelet basis function is used to decompose the sensor sampled data signal to obtain the first high-frequency layer d1, the second high-frequency layer d2, the third high-frequency layer d3, and the low-frequency layer a3; Soft thresholding is performed on the first high-frequency layer d1, the second high-frequency layer d2 and the third high-frequency layer d3. Coefficients smaller than the threshold are set to zero, while large coefficients are retained to avoid over-smoothing. The low-frequency layer a3 is directly retained (containing the useful signal); The sensor sampling data signal is then reconstructed using inverse wavelet transform to obtain a pure gas concentration signal.
[0039] This embodiment uses Stein unbiased risk estimation to adaptively calculate the threshold. : in, is the noise standard deviation of the high-frequency layer coefficients, estimated by the median absolute deviation; N is the number of sampling points, with 32 sampling points processed each time, N=32.
[0040] The specific embodiments described above further illustrate the technical problems, technical solutions, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are 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 processing sampling signals from a gas sensor in an underground coal mine, characterized in that, Includes the following steps: Step S1: Perform sensor data sampling control and sample the data signal output by the sensor; Step S2: Perform IIR notch filtering on the sensor sampled data to remove power frequency interference introduced by the coal mine power grid; Step S3: Perform amplitude limiting filtering on the extreme impulse noise in the sensor sampling data; Step S4: Perform median filtering on the smoothed residual pulses in the sensor sampling data after amplitude limiting filtering; Step S5: Perform Kalman filtering on the residual low-frequency random noise in the sensor sampling data after median filtering; Step S6: Remove broadband random noise from the sensor sampling data after Kalman filtering by wavelet transform filtering.
2. The method for processing sampling signals from underground gas sensors in coal mines according to claim 1, characterized in that, In step S1, sensor data sampling control is performed and the data signal output by the sensor is sampled, specifically including the following steps: The ADC sampling is triggered by a timer inside the MCU processor, with a sampling interval of 1ms. In addition, dense sampling was performed near the sampling point, once every 10us for a total of 10 times; The data signal output by the sensor is sampled to obtain the sensor sampling data.
3. The method for processing sampling signals from underground gas sensors in coal mines according to claim 1, characterized in that, In step S2, the sensor sampled data is subjected to IIR notch filtering to remove power frequency interference introduced by the coal mine power grid. This specifically includes the following steps: Design a 50Hz notch filter to remove 50Hz power frequency interference from sensor sampling data. The expression for the 50Hz notch filter is as follows: ； Design a 100Hz notch filter to remove 100Hz harmonic interference from sensor sampling data. The expression for the 100Hz notch filter is as follows: 。 4. The method for processing sampling signals from underground gas sensors in coal mines according to claim 1, characterized in that, In step S3, the extreme impulse noise in the sensor sampling data is subjected to amplitude limiting filtering, which specifically includes the following steps: An amplitude threshold is set based on the maximum rate of change of the concentration of the detected gas. If the difference between the current sampled value and the previous filtered value is greater than the amplitude threshold, it is determined to be a broadband spike pulse, and the previous filtered value is used to replace the current sampled value; if the difference between the current sampled value and the previous filtered value is less than or equal to the amplitude threshold, the current sampled value is retained.
5. The method for processing sampling signals from underground gas sensors in coal mines according to claim 1, characterized in that, In step S4, median filtering is performed on the smoothed residual pulses in the sensor sampling data after amplitude limiting filtering, specifically including the following steps: Construct a 3-point sliding window [x (n-1) ,x (n) ,x (n+1) The median of the three data points is taken as the current output, and residual pulses are removed to smooth them out.
6. The method for processing sampling signals from underground gas sensors in coal mines according to claim 1, characterized in that, In step S5, Kalman filtering is performed on the residual low-frequency random noise in the sensor sampling data after median filtering, specifically including the following steps: Establish the state transition equation, the expression of which is as follows: X (n) =A・X (n-1) +W (n) Where A=1, W (n) This is process noise; Establish the observation equation, the expression of which is as follows: Z (n) =H・X (n) +V (n) Where H=1, V (n) To observe noise; Residual low-frequency random noise is removed through prediction, updating, and gain calculation iterations.
7. The method for processing sampling signals from underground gas sensors in coal mines according to claim 1, characterized in that, In step S6, broadband random noise in the sensor sampling data after Kalman filtering is removed by wavelet transform filtering, specifically including the following steps: The db4 wavelet basis function is used to decompose the sensor sampled data signal to obtain the first high-frequency layer d1, the second high-frequency layer d2, the third high-frequency layer d3, and the low-frequency layer a3; Soft thresholding is performed on the first high-frequency layer d1, the second high-frequency layer d2, and the third high-frequency layer d3, and coefficients smaller than the threshold are set to zero. The low-frequency layer a3 is retained; The sensor sampling data signal is then reconstructed through inverse wavelet transform.