Multi-channel NDIR infrared gas sensor detection data analysis system and method

Abstract

The invention discloses a multi-channel NDIR infrared gas sensor detection data analysis system and method, and relates to the technical field of gas detection.The method comprises the steps that infrared light intensity of each gas channel is collected to serve as original data, and a fluctuation value of the original data is calculated to serve as an original fluctuation value; acquiring historical temperature and humidity data, fitting a linear relationship between the historical temperature and humidity data and the historical infrared light intensity data, and calibrating the original data based on a correction coefficient obtained by fitting to obtain correction data and a corresponding calibration fluctuation value; calculating a variable coefficient of the correction data, and setting a detection threshold value for each gas channel in combination with the calibration fluctuation value so as to screen out effective original data; distributing an initial weight for each gas channel, dynamically adjusting the weight based on the variable coefficient of the effective data, and processing the adjusted weight through a smoothing algorithm to obtain a smooth weight; and converting the effective original data into gas concentration data, and carrying out weighted fusion on the gas concentration data and the smoothed weight so as to output final gas concentration data.

Description

Technical Field

[0001] The present invention relates to the technical field of gas detection, and in particular to a multi-channel NDIR infrared gas sensor detection data analysis system and method. Background Art

[0002] As a commonly used gas concentration detection device, NDIR gas sensors are widely used in industrial safety, environmental testing, medical health, agriculture and other fields. In practical applications, NDIR sensors usually collect infrared light intensity signals of different gases in parallel through multiple gas channels to achieve simultaneous detection of multiple gas concentrations. Currently, traditional multi-channel NDIR gas sensing systems are often significantly affected by environmental factors such as temperature and humidity in actual working environments. In addition, the response characteristics and sensitivity of different gas channels vary, making it difficult to obtain accurate concentration results. In existing technologies, correction methods for the effects of temperature and humidity are mostly based on static calibration or single-variable regression models, which lack real-time performance and adaptability. When processing multi-channel data, fixed weighting or simple averaging is often used, which fails to fully consider the dynamic stability and abnormal fluctuations of channel data, making it difficult to obtain more accurate and stable gas concentration output results. Therefore, a multi-channel NDIR infrared gas sensor detection data analysis system and method are urgently needed to solve the above problems. Summary of the Invention

[0003] The object of the present invention is to provide a multi-channel NDIR infrared gas sensor detection data analysis system and method to solve the problems raised in the prior art.

[0004] To achieve the above object, the present invention provides the following technical solution: a multi-channel NDIR infrared gas sensor detection data analysis method, the method comprising the following steps: S100 collects infrared light intensity data of each gas channel as raw data and calculates the raw data fluctuation value; Furthermore, the specific steps for calculating the original fluctuation value are: S101 collects several gas channels in the NDIR device. Each gas channel can be configured with a corresponding central wavelength filter according to the infrared absorption characteristics of different gases to achieve selective detection of multiple gases such as carbon dioxide, methane, and carbon monoxide. The gas channel set A={A1,A2,A3,...,Am} is constructed, where A1,A2,A3,... and Am represent the 1st, 2nd, 3rd,... and mth gas channels respectively. The infrared light intensity of m gas channels within the time period Tnow with a period of z is collected in real time. Data, take the collected infrared light intensity data as the raw data, for the i-th gas channel, i∈[1,m], collect n raw data in total, add time stamps to the n raw data, collect the raw data of the i-th gas channel in time axis order, and construct the raw data set Ii={Ii1,Ii2,Ii3,...,Iin} of the i-th gas channel, where Ii1,Ii2,Ii3,... and Iin represent the 1st, 2nd, 3rd,... and nth raw data collected corresponding to the i-th gas channel respectively; S102 calculates the original fluctuation value σi of the original data in the set Ii, and the calculation formula is: , where j∈[1,n], is the mean of the original data in set Ii; S200 collects historical temperature and humidity data, fits the linear relationship between it and historical infrared light intensity data, and calibrates the original data based on the correction coefficient obtained by fitting to obtain the corrected data and the corresponding calibration fluctuation value; Furthermore, the specific steps for obtaining the calibration fluctuation value are: S201 collects real-time temperature data W and real-time humidity data RH, aligns them with the time axis of the original data in set Ii, and then extracts the corresponding relationship data pairs between the historical temperature data and humidity data and the historical infrared intensity data of the i-th gas channel, respectively, denoted as (Wi, Ii) and (RHi, Ii), extracts k records in total, and constructs the corresponding data set WIi={(Wi1, Ii1),(Wi2, Ii2),(Wi3, Ii3),...,(Wik, Iik)} of temperature and original data, where (Wi1, Ii1),(Wi2, Ii2),(Wi3, Ii3),... and (Wik, Iik) represent the corresponding data pairs of the 1st, 2nd, 3rd,... and kth historical temperature data and original data, respectively. A corresponding data set RHIi={(RHif, Iif)|f=1,2,3,...,k} of humidity data and original data is also constructed. Temperature influence: Temperature changes can cause fluctuations in the infrared source's luminous intensity, resulting in a shift in the light intensity received by the detector. The molecular thermal motion rate of the measured gas changes with temperature, altering its infrared absorption characteristics (such as line width and absorption coefficient). Some optical devices (filters, detectors) are temperature sensitive, and rising temperatures can cause zero-point drift or nonlinear response. Influence of humidity: Water vapor itself has multiple infrared absorption peaks, some of which overlap or are adjacent to the absorption peaks of the target gas, which can easily cause interference absorption. In high humidity environments, micro-condensation is prone to form on the surface of the optical window, leading to infrared scattering and signal attenuation. Humidity also has a certain impact on gas density and mixing ratio. S202 fits the linear relationship between the historical temperature data and the historical infrared light intensity data of the i-th gas channel, and constructs the equation: Iif=a×Wif+b, where a is the temperature influence slope coefficient, which represents the degree of linear response of temperature change to infrared light intensity data, b is the temperature influence intercept, which represents the theoretical reference value of infrared light intensity when the temperature is zero, Iif is the f-th infrared light intensity data in the set WI, and Wif is the historical temperature data corresponding to the f-th infrared light intensity data. The values of a and b are obtained by solving the data pairs in the set WIi using the least squares method; S203 calculates the temperature correction coefficient CW for the j-th original data Iij in the set Ii, and the calculation formula is: , Where Wij is the real-time temperature data corresponding to the j-th original data. The temperature correction data Icalij is obtained by calculating the j-th original data after temperature correction. The calculation formula is: ; S204 fits the linear relationship between historical humidity data and historical infrared light intensity data, constructing the equation: Iif = c × RHif + d, where c is the humidity influence slope coefficient, which represents the linear response degree of humidity change to infrared light intensity data, and d is the humidity influence intercept, which represents the theoretical reference value of infrared light intensity when the humidity is zero. The values of c and d are obtained by solving the data pairs in the set RHIi using the least squares method; S205 calculates the humidity correction coefficient CRH for the j-th original data Iij in the set Ii, and the calculation formula is: , Where RHij is the real-time humidity data corresponding to the j-th original data. The corrected data Icalij* obtained by correcting the humidity data of the j-th original data is calculated using the following formula: ; S206 collects n corrected data to construct a corrected data set Ii*={Icali1*,Icali2*,Icali3*,...,Icalin*}, where Icali1*, Icali2*, Icali3*,..., Icalin* represent the first, second, third,..., and nth corrected original data in the set Ii, respectively, and calculates the standard deviation σcali of the data in the set Ii* as the calibrated fluctuation value; A temperature and humidity correction model is constructed based on the linear relationship between historical temperature and humidity data and infrared light intensity data. The infrared light intensity data is corrected point by point based on real-time temperature and humidity data, significantly reducing data measurement fluctuations caused by environmental factors and improving the quality of raw data and the accuracy of concentration estimation. S300 calculates the coefficient of variation of the corrected data, and sets the data detection threshold for each gas channel in combination with the calibration fluctuation value, and filters the original data to obtain valid original data; Furthermore, the specific steps to obtain valid original data are: S301 calculates the mean of the data in the set Ii* , based on the mean The coefficient of variation CVcali of the corrected data is calculated using the calibrated fluctuation value σcali. The calculation formula is: ; S302 uses a sliding window with a window length of y and a step size of u to calculate the original data of the i-th gas channel in q, q≥y consecutive time periods. The mean μCVcali and standard deviation σCVcali of the coefficient of variation within the sliding window are calculated, and the risk levels of the gases detected by the m gas channels are set respectively. The risk levels are set to D1 and D2. D1 is high risk, and the gases included are high-risk gases. The gas characteristics are toxic, flammable or acute hazards. D2 is low risk, and the gases included are low-risk gases. The gas characteristics are non-acute toxicity, and most of them are environmental or climate detection indicators. The abnormal threshold is set according to the risk level. When the risk level is D1, the threshold is set to THD1=μCVcali+ pD1×σCVcali. When the risk level is D2, the threshold is set as THD2=μCVcali+pD2×σCVcali, where λ is the weight influence coefficient, pD1 and pD2 are the risk sensitivity coefficients of D1 and D2 respectively. The risk sensitivity coefficient is determined by using the distribution law of the coefficient of variation in the historical correction data. The mean and standard deviation of the historical correction data are first calculated, and then the confidence level is set according to the risk level. The standard normal distribution quantile under the corresponding confidence level is directly used as the value of pD1 and pD2 to set the abnormal threshold; S303 compares the coefficient of variation CVcali of the corrected data of the i-th gas channel calculated within the time period Tnow with the corresponding abnormal threshold: if the gas detected by the i-th gas channel belongs to D1, when CVcali>THD1, the original data of the i-th gas channel is determined to be abnormal data; if the gas detected by the i-th gas channel belongs to D2, when CVcali>THD2, the original data of the i-th gas channel is determined to be abnormal data. When it is determined to be abnormal data, the abnormal data points in the set Ii are eliminated by the Z-sore anomaly detection method. For each data, the Z value is calculated by the Z-sore anomaly detection method, and the data that meets |Z|>ω is marked as abnormal data. The value of ω for D1 type gas is usually 2, and the value of ω for D2 type gas is usually 3. The remaining e original data are taken as valid original data e∈[1,n] to obtain the valid original data set Ii^; By using threshold detection and anomaly detection for gases, the interference noise in the original data can be effectively eliminated and the real data information can be retained. When the effective original data processed in this way participates in the subsequent concentration calculation and weighted fusion, the interference of noise on the data can be avoided, thereby outputting accurate concentration data that is closer to the actual gas content, improving the reliability and accuracy of the detection results; S400 assigns an initial weight to each gas channel, performs dynamic weight adjustment based on the coefficient of variation of the valid original data, and smoothes the adjusted weight value to obtain a smoothed weight; Furthermore, the specific steps to obtain the smoothed weight are: S401 assigns an initial weight Qi0 of 1 to the i-th gas channel, calculates the coefficient of variation CVcali^ of the data in the set Ii^, and dynamically adjusts the initial weight based on CVcali^ to obtain the adjusted weight Qi1. The adjustment formula is: , Where r is the weight sensitivity coefficient, r∈[0,1], and the value of r is optimized based on the mean square error (MSE); S402 uses exponentially weighted moving average to smooth the adjusted weight Qi1 to obtain the smoothed weight Qi2, which is expressed as follows: , Where α is the smoothing coefficient, α∈[0,1], Qi1 is the adjusted weight of the i-th gas channel in the real-time acquisition period Tnow, and Qiold is the smoothed weight of the i-th gas channel in the T(now-1)th period. When the formula is calculated for the first time, since there is no historical smoothing weight, the initial weight can be directly used as the first smoothing result; Through the "dynamic weight allocation + exponential smoothing" mechanism, channel volatility is reflected, effectively identifying and reducing the impact of unstable channels, improving the stability and credibility of raw data, and providing key support for building highly robust data analysis with low false alarm rates. S500 converts the effective raw data into gas concentration data, performs weighted fusion with the smoothed weights, and outputs the final gas concentration data; Furthermore, the specific steps for outputting the final gas concentration data are: S601 converts the g-th valid raw data in the i-th gas channel, g∈[1,e], into the corresponding gas concentration data Cig using the Lambert-Beer law. The conversion formula is: , Where ε is the molar absorptivity, ι is the optical path length, Io is the incident light intensity, and Icalig is the final raw data after the gth correction in the i-th gas channel; ε can be queried through the spectral database, ι is the inherent parameter of the device, and Io is the factory calibration value of the device; S602 performs weighted fusion on the calculated gas concentration data and the weight Qv2 after smoothing of the i-th gas channel to obtain the final gas concentration data Sig corresponding to the j-th valid original data. The fusion formula is: ; In order to better implement the above method, a multi-channel NDIR infrared gas sensor detection data analysis system is proposed. The system includes a data acquisition module, a data calibration module, a data screening module, a weight calculation module, and a concentration data output module. The data acquisition module is used to collect infrared light intensity data of each gas channel; The data calibration module is used to calibrate the infrared light intensity data using temperature and humidity data; The calibration module includes a temperature calibration unit and a humidity calibration unit; The temperature calibration unit is used to perform temperature compensation processing on the infrared light intensity data to eliminate the influence of temperature fluctuation on the infrared light intensity data; The humidity calibration unit is used to perform humidity compensation processing on the infrared light intensity data to eliminate the impact of humidity fluctuations on the infrared light intensity data; The data screening module is used to set the data detection threshold for each gas channel and screen the infrared light intensity data; The weight calculation module is used to assign initial weights to each gas channel and perform dynamic weight adjustment and smoothing on the initial weights; The concentration data output module is used to convert infrared light intensity data into gas concentration data and output the final gas concentration data; The concentration data output module includes a conversion unit and a weighted fusion unit; The conversion unit is used to convert the processed infrared light intensity data into corresponding gas concentration data; The weighted fusion unit is used to perform weighted fusion on the gas concentration data output by each gas channel with the corresponding smoothing weight, and output the final weighted gas concentration value.

[0005] Compared with the prior art, the present invention has the following beneficial effects: 1. Through temperature and humidity data, a linear fitting method based on historical correlation data is used to dynamically compensate for real-time infrared light intensity data, effectively reducing the interference of ambient temperature and humidity fluctuations on infrared signals and improving the stability and consistency of the original data.

[0006] 2. By building a dynamic relative volatility assessment model, we achieve dynamic adjustment and smoothing of channel weights, avoiding the channel error propagation problem in traditional fixed weight methods and improving the accuracy and robustness of multi-channel data fusion; 3. The infrared light intensity data after temperature and humidity calibration is used, and combined with dynamic smoothing weights for weighted fusion. While retaining the redundancy advantage of multiple channel detections, it effectively avoids the impact of abnormal channel data on the final concentration estimation, significantly improving the reliability and accuracy of gas concentration data output. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 This is a structural diagram of the multi-channel NDIR infrared gas sensor detection data analysis system of the present invention; Figure 2 Schematic diagram of the process of the multi-channel NDIR infrared gas sensor detection data analysis method of the present invention; Figure 3 Schematic diagram of an embodiment of the multi-channel NDIR infrared gas sensor detection data analysis method of the present invention. DETAILED DESCRIPTION

[0008] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative work shall fall within the scope of protection of the present invention.

[0009] Example: Figure 1-Figure 3 As shown, the present invention provides a technical solution, a multi-channel NDIR infrared gas sensor detection data analysis method, the method comprising the following steps: S100 collects infrared light intensity data of each gas channel as raw data and calculates the raw data fluctuation value; Furthermore, the specific steps for calculating the original fluctuation value are: S101 collects several gas channels in the NDIR device. Each gas channel can be configured with a corresponding central wavelength filter according to the infrared absorption characteristics of different gases to achieve selective detection of multiple gases such as carbon dioxide, methane, and carbon monoxide. The gas channel set A={A1,A2,A3,...,Am} is constructed, where A1,A2,A3,... and Am represent the 1st, 2nd, 3rd,... and mth gas channels respectively. The infrared light intensity of m gas channels within the time period Tnow with a period of z is collected in real time. Data, take the collected infrared light intensity data as the raw data, for the i-th gas channel, i∈[1,m], collect n raw data in total, add time stamps to the n raw data, collect the raw data of the i-th gas channel in time axis order, and construct the raw data set Ii={Ii1,Ii2,Ii3,...,Iin} of the i-th gas channel, where Ii1,Ii2,Ii3,... and Iin represent the 1st, 2nd, 3rd,... and nth raw data collected corresponding to the i-th gas channel respectively; S102 calculates the original fluctuation value σi of the original data in the set Ii, and the calculation formula is: , where j∈[1,n], is the mean of the original data in set Ii; S200 collects historical temperature and humidity data, fits the linear relationship between it and historical infrared light intensity data, and calibrates the original data based on the correction coefficient obtained by fitting to obtain the corrected data and the corresponding calibration fluctuation value; Furthermore, the specific steps for obtaining the calibration fluctuation value are: S201 collects real-time temperature data W and real-time humidity data RH, aligns them with the time axis of the original data in set Ii, and then extracts the corresponding relationship data pairs between the historical temperature data and humidity data and the historical infrared intensity data of the i-th gas channel, respectively, denoted as (Wi, Ii) and (RHi, Ii), extracts k records in total, and constructs the corresponding data set WIi={(Wi1, Ii1),(Wi2, Ii2),(Wi3, Ii3),...,(Wik, Iik)} of temperature and original data, where (Wi1, Ii1),(Wi2, Ii2),(Wi3, Ii3),... and (Wik, Iik) represent the corresponding data pairs of the 1st, 2nd, 3rd,... and kth historical temperature data and original data, respectively. A corresponding data set RHIi={(RHif, Iif)|f=1,2,3,...,k} of humidity data and original data is also constructed. Temperature influence: Temperature changes can cause fluctuations in the infrared source's luminous intensity, resulting in a shift in the light intensity received by the detector. The molecular thermal motion rate of the measured gas changes with temperature, altering its infrared absorption characteristics (such as line width and absorption coefficient). Some optical devices (filters, detectors) are temperature sensitive, and rising temperatures can cause zero-point drift or nonlinear response. Influence of humidity: Water vapor itself has multiple infrared absorption peaks, some of which overlap or are adjacent to the absorption peaks of the target gas, which can easily cause interference absorption. In high humidity environments, micro-condensation is prone to form on the surface of the optical window, leading to infrared scattering and signal attenuation. Humidity also has a certain impact on gas density and mixing ratio. S202 fits the linear relationship between the historical temperature data and the historical infrared light intensity data of the i-th gas channel, and constructs the equation: Iif=a×Wif+b, where a is the temperature influence slope coefficient, which represents the degree of linear response of temperature change to infrared light intensity data, b is the temperature influence intercept, which represents the theoretical reference value of infrared light intensity when the temperature is zero, Iif is the f-th infrared light intensity data in the set WI, and Wif is the historical temperature data corresponding to the f-th infrared light intensity data. The values of a and b are obtained by solving the data pairs in the set WIi using the least squares method; S203 calculates the temperature correction coefficient CW for the j-th original data Iij in the set Ii, and the calculation formula is: , Where Wij is the real-time temperature data corresponding to the j-th original data. The temperature correction data Icalij is obtained by calculating the j-th original data after temperature correction. The calculation formula is: ; S204 fits the linear relationship between historical humidity data and historical infrared light intensity data, constructing the equation: Iif = c × RHif + d, where c is the humidity influence slope coefficient, which represents the linear response degree of humidity change to infrared light intensity data, and d is the humidity influence intercept, which represents the theoretical reference value of infrared light intensity when the humidity is zero. The values of c and d are obtained by solving the data pairs in the set RHIi using the least squares method; S205 calculates the humidity correction coefficient CRH for the j-th original data Iij in the set Ii, and the calculation formula is: , Where RHij is the real-time humidity data corresponding to the j-th original data. The corrected data Icalij* obtained by correcting the humidity data of the j-th original data is calculated using the following formula: ; S206 collects n corrected data to construct a corrected data set Ii*={Icali1*,Icali2*,Icali3*,...,Icalin*}, where Icali1*, Icali2*, Icali3*,..., Icalin* represent the first, second, third,..., and nth corrected original data in the set Ii, respectively, and calculates the standard deviation σcali of the data in the set Ii* as the calibrated fluctuation value; A temperature and humidity correction model is constructed based on the linear relationship between historical temperature and humidity data and infrared light intensity data. The infrared light intensity data is corrected point by point based on real-time temperature and humidity data, significantly reducing data measurement fluctuations caused by environmental factors and improving the quality of raw data and the accuracy of concentration estimation. S300 calculates the coefficient of variation of the corrected data, and sets the data detection threshold for each gas channel in combination with the calibration fluctuation value, and filters the original data to obtain valid original data; Furthermore, the specific steps to obtain valid original data are: S301 calculates the mean of the data in the set Ii* , based on the mean The coefficient of variation CVcali of the corrected data is calculated using the calibrated fluctuation value σcali. The calculation formula is: ; S302 uses a sliding window with a window length of y and a step size of u to calculate the original data of the i-th gas channel in q, q≥y consecutive time periods. The mean μCVcali and standard deviation σCVcali of the coefficient of variation within the sliding window are calculated, and the risk levels of the gases detected by the m gas channels are set respectively. The risk levels are set to D1 and D2. D1 is high risk, and the gases included are high-risk gases. The gas characteristics are toxic, flammable or acute hazards. D2 is low risk, and the gases included are low-risk gases. The gas characteristics are non-acute toxicity, and most of them are environmental or climate detection indicators. The abnormal threshold is set according to the risk level. When the risk level is D1, the threshold is set to THD1=μCVcali+ pD1×σCVcali. When the risk level is D2, the threshold is set as THD2=μCVcali+pD2×σCVcali, where λ is the weight influence coefficient, pD1 and pD2 are the risk sensitivity coefficients of D1 and D2 respectively. The risk sensitivity coefficient is determined by using the distribution law of the coefficient of variation in the historical correction data. The mean and standard deviation of the historical correction data are first calculated, and then the confidence level is set according to the risk level. The standard normal distribution quantile under the corresponding confidence level is directly used as the value of pD1 and pD2 to set the abnormal threshold; S303 compares the coefficient of variation CVcali of the corrected data of the i-th gas channel calculated within the time period Tnow with the corresponding abnormal threshold: if the gas detected by the i-th gas channel belongs to D1, when CVcali>THD1, the original data of the i-th gas channel is determined to be abnormal data; if the gas detected by the i-th gas channel belongs to D2, when CVcali>THD2, the original data of the i-th gas channel is determined to be abnormal data. When it is determined to be abnormal data, the abnormal data points in the set Ii are eliminated by the Z-sore anomaly detection method. For each data, the Z value is calculated by the Z-sore anomaly detection method, and the data that meets |Z|>ω is marked as abnormal data. The value of ω for D1 type gas is usually 2, and the value of ω for D2 type gas is usually 3. The remaining e original data are taken as valid original data e∈[1,n] to obtain the valid original data set Ii^; By using threshold detection and anomaly detection for gases, the interference noise in the original data can be effectively eliminated and the real data information can be retained. When the effective original data processed in this way participates in the subsequent concentration calculation and weighted fusion, the interference of noise on the data can be avoided, thereby outputting accurate concentration data that is closer to the actual gas content, improving the reliability and accuracy of the detection results; S400 assigns an initial weight to each gas channel, performs dynamic weight adjustment based on the coefficient of variation of the valid original data, and smoothes the adjusted weight value to obtain a smoothed weight; Furthermore, the specific steps to obtain the smoothed weight are: S401 assigns an initial weight Qi0 of 1 to the i-th gas channel, calculates the coefficient of variation CVcali^ of the data in the set Ii^, and dynamically adjusts the initial weight based on CVcali^ to obtain the adjusted weight Qi1. The adjustment formula is: , Where r is the weight sensitivity coefficient, r∈[0,1], and the value of r is optimized based on the mean square error (MSE); S402 uses exponentially weighted moving average to smooth the adjusted weight Qi1 to obtain the smoothed weight Qi2, which is expressed as follows: , Where α is the smoothing coefficient, α∈[0,1], Qi1 is the adjusted weight of the i-th gas channel in the real-time acquisition period Tnow, and Qiold is the smoothed weight of the i-th gas channel in the T(now-1)th period. When the formula is calculated for the first time, since there is no historical smoothing weight, the initial weight can be directly used as the first smoothing result; Through the "dynamic weight allocation + exponential smoothing" mechanism, channel volatility is reflected, effectively identifying and reducing the impact of unstable channels, improving the stability and credibility of raw data, and providing key support for building highly robust data analysis with low false alarm rates. S500 converts the effective raw data into gas concentration data, performs weighted fusion with the smoothed weights, and outputs the final gas concentration data; Furthermore, the specific steps for outputting the final gas concentration data are: S601 converts the g-th valid raw data in the i-th gas channel, g∈[1,e], into the corresponding gas concentration data Cig using the Lambert-Beer law. The conversion formula is: , Where ε is the molar absorptivity, ι is the optical path length, Io is the incident light intensity, and Icalig is the final raw data after the gth correction in the i-th gas channel; ε can be queried through the spectral database, ι is the inherent parameter of the device, and Io is the factory calibration value of the device; S602 performs weighted fusion on the calculated gas concentration data and the weight Qv2 after smoothing of the i-th gas channel to obtain the final gas concentration data Sig corresponding to the j-th valid original data. The fusion formula is: ; In order to better implement the above method, a multi-channel NDIR infrared gas sensor detection data analysis system is proposed. The system includes a data acquisition module, a data calibration module, a data screening module, a weight calculation module, and a concentration data output module. The data acquisition module is used to collect infrared light intensity data of each gas channel; The data calibration module is used to calibrate the infrared light intensity data using temperature and humidity data; The calibration module includes a temperature calibration unit and a humidity calibration unit; The temperature calibration unit is used to perform temperature compensation processing on the infrared light intensity data to eliminate the influence of temperature fluctuation on the infrared light intensity data; The humidity calibration unit is used to perform humidity compensation processing on the infrared light intensity data to eliminate the impact of humidity fluctuations on the infrared light intensity data; The data screening module is used to set the data detection threshold for each gas channel and screen the infrared light intensity data; The weight calculation module is used to assign initial weights to each gas channel and perform dynamic weight adjustment and smoothing on the initial weights; The concentration data output module is used to convert infrared light intensity data into gas concentration data and output the final gas concentration data; The concentration data output module includes a conversion unit and a weighted fusion unit; The conversion unit is used to convert the processed infrared light intensity data into corresponding gas concentration data; The weighted fusion unit is used to perform weighted fusion on the gas concentration data output by each gas channel with the corresponding smoothing weight, and output the final weighted gas concentration value.

[0010] Example: This example is based on an NDIR infrared gas sensor with three gas channels. The channels are used to detect CO2, CH4, and CO respectively. The sampling period is 1 second and the time period is 60 seconds. A gas channel set A is constructed. Each channel collects 60 infrared light intensity data, such as CO2 channel data: I1={2200,2215,2232,...,2280}. The fluctuation value σ1≈25.7 of the data in the set is calculated. The temperature and humidity data, historical temperature and humidity data and their corresponding historical infrared light intensity data are collected in real time. The temperature and humidity correction process for the infrared light intensity data is carried out, and the original data is corrected in two steps. , the corrected data set I1* is obtained, and the corrected standard deviation σcal1≈13.5. The sliding window (length 10, step size 5) is used to calculate the CV mean and standard deviation: μCV=0.56%, σCV=0.12%, threshold (risk level D1): TH=0.56+2×0.12=0.80%. If the current CV is 0.83%, it is judged as abnormal, and the abnormal points are removed. 54 valid data are retained, with an initial weight of 1, which is adjusted to 0.97 according to CV. After exponential smoothing (α=0.6), a smoothing weight of 0.94 is obtained. The infrared light intensity data is converted to obtain the CO2 concentration data of approximately 9.5 ppm. After multiplying by the weight, the final output concentration is 8.93 ppm.

[0011] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art will be able to modify the technical solutions described in the aforementioned embodiments or substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A multi-channel NDIR infrared gas sensor detection data analysis method, characterized by: S100 collects infrared light intensity data of each gas channel as raw data and calculates the raw data fluctuation value; S200 collects historical temperature and humidity data, fits the linear relationship between it and historical infrared light intensity data, and calibrates the original data based on the correction coefficient obtained by fitting to obtain the corrected data and the corresponding calibration fluctuation value; S300 calculates the coefficient of variation of the corrected data, and sets the data detection threshold for each gas channel in combination with the calibration fluctuation value, and filters the original data to obtain valid original data; S400 assigns an initial weight to each gas channel, performs dynamic weight adjustment based on the coefficient of variation of the valid original data, and smoothes the adjusted weight value to obtain a smoothed weight; S500 converts the effective raw data into gas concentration data, performs weighted fusion with the smoothed weights, and outputs the final gas concentration data.

2. The multi-channel NDIR infrared gas sensor detection data analysis method according to claim 1, characterized in that: The specific steps for calculating the fluctuation value of the infrared light intensity raw data in S100 are: S101 collects several gas channels in the NDIR device and constructs a gas channel set A={A1,A2,A3,...,Am}, where A1,A2,A3,... and Am represent the 1st, 2nd, 3rd,... and mth gas channels respectively. Infrared light intensity data of the m gas channels in a time period Tnow with a period of z are collected in real time. The collected infrared light intensity data are used as raw data. For the i-th gas channel, i∈[1,m], a total of n raw data are collected, and the n raw data are timestamped. The raw data of the i-th gas channel are collected in time sequence to construct a raw data set Ii={Ii1,Ii2,Ii3,...,Iin} of the i-th gas channel, where Ii1,Ii2,Ii3,... and Iin represent the 1st, 2nd, 3rd,... and nth raw data collected corresponding to the i-th gas channel respectively. S102 calculates the original fluctuation value σi of the original data in the set Ii, and the calculation formula is: ， where j∈[1,n], is the mean of the original data in set Ii.

3. The multi-channel NDIR infrared gas sensor detection data analysis method according to claim 2, characterized in that: The specific steps of obtaining the calibration fluctuation value in S200 are: S201 collects real-time temperature data W and real-time humidity data RH, aligns them with the time axis of the original data in set Ii, and then extracts the corresponding relationship data pairs between the historical temperature data and humidity data and the historical infrared intensity data of the i-th gas channel, respectively, denoted as (Wi, Ii) and (RHi, Ii), extracts k records in total, and constructs the corresponding data set WI of temperature and original data = {(Wi1, Ii1), (Wi2, Ii2), (Wi3, Ii3), ..., (Wik, Iik)}, where (Wi1, Ii1), (Wi2, Ii2), (Wi3, Ii3), ... and (Wik, Iik) represent the corresponding data pairs of the 1st, 2nd, 3rd, ... and kth historical temperature data and original data, respectively, and constructs the corresponding data set RHIi of humidity data and original data = {(RHif, Iif)|f=1, 2, 3, ..., k}; S202 fits the linear relationship between the historical temperature data and the historical infrared light intensity data of the i-th gas channel, and constructs the equation: Iif=a×Wif+b, where a is the temperature influence slope coefficient, b is the temperature influence intercept, Iif is the f-th infrared light intensity data in the set WI, and Wif is the historical temperature data corresponding to the f-th infrared light intensity data. The values of a and b are obtained by combining the data pairs in the set WI with the least squares method; S203 calculates the temperature correction coefficient CW for the j-th original data Iij in the set Ii, and the calculation formula is: ， Where Wij is the real-time temperature data corresponding to the j-th original data. The temperature correction data Icalij is obtained by calculating the j-th original data after temperature correction. The calculation formula is: ； S204 fits the linear relationship between historical humidity data and historical infrared light intensity data, constructing the equation: Iif=c×RHif+d, where c is the slope coefficient of humidity influence and d is the intercept of humidity influence. The values of c and d are obtained by combining the data pairs in the set RHIi with the least squares method; S205 calculates the humidity correction coefficient CRH for the j-th original data Iij in the set Ii, and the calculation formula is: , Where RHij is the real-time humidity data corresponding to the j-th original data. The corrected data Icalij* obtained by correcting the humidity data of the j-th original data is calculated using the following formula: ； S206 collects n corrected data and constructs a corrected data set Ii*={Icali1*,Icali2*,Icali3*,...,Icalin*}, where Icali1*, Icali2*, Icali3*,..., Icalin* represent the original data in the first, second, third,...th and nth corrected set Ii respectively, and calculates the standard deviation σcali of the data in the set Ii* as the calibrated fluctuation value.

4. The multi-channel NDIR infrared gas sensor detection data analysis method according to claim 3, characterized in that: The specific steps in S300 are: S301 calculates the mean of the data in the set Ii* , based on the mean The coefficient of variation CVcali of the corrected data is calculated using the calibrated fluctuation value σcali. The calculation formula is: ； S302 uses a sliding window with a window length of y and a step size of u to calculate the original data of the i-th gas channel in q, q≥y consecutive time periods. The mean μCVcali and standard deviation σCVcali of the coefficient of variation within the sliding window are calculated, and the risk levels of the gases detected by the m gas channels are set respectively, and the risk levels are set to D1 and D2. The abnormal threshold is set according to the risk level. When the risk level is D1, the threshold is set to THD1=μCVcali+pD1×σCVcali, and when the risk level is D2, the threshold is set to THD2=μCVcali+pD2×σCVcali, where λ is the weight influence coefficient, pD1 and pD2 are the risk sensitivity coefficients of D1 and D2 respectively; S303 compares the coefficient of variation CVcali of the corrected data of the i-th gas channel calculated in the time period Tnow with the corresponding abnormal threshold: if the gas detected by the i-th gas channel belongs to D1, when CVcali>THD1, the original data of the i-th gas channel is determined to be abnormal data; If the gas detected by the i-th gas channel belongs to D2, when CVcali>THD2, the raw data of the i-th gas channel is judged to be abnormal data. When it is judged to be abnormal data, the abnormal data points in the set Ii are eliminated through Z-sore anomaly detection, and the remaining e raw data are taken as valid raw data e∈[1,n] to obtain the valid raw data set Ii^.

5. The multi-channel NDIR infrared gas sensor detection data analysis method according to claim 4, characterized in that: The specific steps of obtaining the smoothed weight in S400 are: S401 assigns an initial weight Qi0 of 1 to the i-th gas channel, calculates the coefficient of variation CVcali^ of the data in the set Ii^, and dynamically adjusts the initial weight based on CVcali^ to obtain the adjusted weight Qi1. The adjustment formula is: ， Where r is the weight sensitivity coefficient; S402 uses exponentially weighted moving average to smooth the adjusted weight Qi1 to obtain the smoothed weight Qi2, which is expressed as follows: ， Where α is the smoothing coefficient, α∈[0,1], Qi1 is the adjusted weight of the i-th gas channel in the real-time acquisition time period Tnow, and Qiold is the smoothed weight of the i-th gas channel in the T(now-1)th time period.

6. The multi-channel NDIR infrared gas sensor detection data analysis method according to claim 6, characterized in that: The specific steps for outputting the final gas concentration data in S500 are: S601 converts the g-th valid raw data in the i-th gas channel, g∈[1,e], into the corresponding gas concentration data Cig. The conversion formula is: ， Where ε is the molar absorptivity, ι is the optical path length, Io is the incident light intensity, and Icalig is the final raw data after the gth correction in the i-th gas channel; S602 performs weighted fusion on the calculated gas concentration data and the weight Qv2 after smoothing of the i-th gas channel to obtain the final gas concentration data Sig corresponding to the j-th valid original data. The fusion formula is: 。 7. A multi-channel NDIR infrared gas sensor detection data analysis system, applied to the multi-channel NDIR infrared gas sensor detection data analysis method according to any one of claims 1 to 6, characterized in that: The multi-channel NDIR infrared gas sensor detection data analysis system includes a data acquisition module, a data calibration module, a data screening module, a weight calculation module, and a concentration data output module; The data acquisition module is used to collect infrared light intensity data of each gas channel; The data calibration module is used to calibrate the infrared light intensity data using temperature and humidity data; The data screening module is used to set a data detection threshold for each gas channel and screen the infrared light intensity data; The weight calculation module is used to assign initial weights to each gas channel and perform dynamic weight adjustment and smoothing on the initial weights; The concentration data output module is used to convert infrared light intensity data into gas concentration data and output the final gas concentration data.

8. The multi-channel NDIR infrared gas sensor detection data analysis system according to claim 7, characterized in that: The calibration module includes a temperature calibration unit and a humidity calibration unit; The temperature calibration unit is used to perform temperature compensation processing on the infrared light intensity data to eliminate the influence of temperature fluctuation on the infrared light intensity data; The humidity calibration unit is used to perform humidity compensation processing on the infrared light intensity data to eliminate the influence of humidity fluctuation on the infrared light intensity data.

9. The multi-channel NDIR infrared gas sensor detection data analysis system according to claim 8, characterized in that: The concentration data output module includes a conversion unit and a weighted fusion unit; The conversion unit is used to convert the processed infrared light intensity data into corresponding gas concentration data; The weighted fusion unit is used to perform weighted fusion on the gas concentration data output by each gas channel and the corresponding smoothing weight, and output a final weighted gas concentration value.

Images