Ambient gas interference correction method and device for quartz enhanced photoacoustic spectrometry detection system

Abstract

The invention provides a quartz enhanced photoacoustic spectrometry detection system environment gas interference correction method which can be applied to the technical field of photoacoustic spectrometry gas detection. The method comprises the following steps: analyzing vibration-translation relaxation time and infrared absorption peak intensity of each gas to be detected; the to-be-detected gas with the vibration-translation relaxation time exceeding the time threshold value and the absorption peak intensity larger than the intensity threshold value is selected as comparison gas; dividing the laser of the comparison gas into a first laser beam and a second laser beam according to a preset proportion; using the first laser beam to excite a sound wave signal and detecting by a quartz tuning fork to generate a tuning fork electric signal; the second laser beam is enhanced and absorbed by a multi-reflection absorption cell and then is detected by a photoelectric detector to generate a light intensity electric signal; establishing a neural network model taking the tuning fork electric signal, the light intensity electric signal and the environment gas component vector as input and taking the to-be-detected gas concentration vector as output; and correcting the concentration of the to-be-detected gas under the interference of the environmental gas in real time through the trained neural network model.

Description

Technical Field

[0001] The present invention relates to the technical field of photoacoustic spectroscopy gas detection, and more particularly to a method and device for correcting environmental gas interference in a quartz-enhanced photoacoustic spectroscopy detection system. Background Art

[0002] Quartz-enhanced photoacoustic spectroscopy (QEPAS), a highly sensitive gas detection technology based on the photoacoustic effect, has a unique operating mechanism and significant advantages. In this technique, a gas absorbs laser light, becomes excited, and then de-excites by releasing heat through a non-radiative recombination process. When the laser wavelength is periodically modulated, the gas undergoes periodic temperature fluctuations based on the modulation frequency, which in turn induces periodic pressure fluctuations, or acoustic waves.

[0003] This technology uses a quartz tuning fork to detect acoustic signals. By inverting the detected signals, the concentration of the gas to be measured can be accurately determined. QEPAS technology has many outstanding features: first, there is no background absorption, which makes the detection results more accurate and reliable; second, the acoustic signal is proportional to the intensity of the excitation light, which means that the signal-to-noise ratio of the photoacoustic signal can be improved by increasing the light source power, thereby effectively improving the sensitivity of the system. Theoretically, its detection limit can reach 10 -13 Thirdly, since the detection of acoustic wave signals is not related to the wavelength of the excitation light source, it has obvious advantages in multi-component gas detection scenarios and can accurately detect multiple gas components at the same time.

[0004] Quartz tuning forks play a key role in QEPAS technology. Their extremely high Q factor effectively filters out interference from non-resonant frequencies, ensuring a purer detected signal. Furthermore, quartz tuning forks respond only to differential-mode vibrations, while ambient noise typically induces common-mode noise. This gives them excellent noise immunity, enabling stable operation in complex environments and minimizing the impact of external interference on detection results.

[0005] However, in practical applications, QEPAS technology faces significant challenges in complex environments. Most test environments involve complex atmospheres, some even containing unknown gas compositions. The composition and concentration of gases in complex atmospheres are constantly changing, significantly impacting the effective VT (vibration-translation) relaxation rate, the speed of sound, and the frequency of the acoustic wave.

[0006] Specifically, changes in the effective VT relaxation rate alter the energy transfer process between gas molecules, thereby affecting the generation mechanism of sound waves. Changes in the speed of sound can lead to changes in the characteristics of sound waves during propagation. The deviation of the sound wave frequency is a key issue, as it can cause the sound wave frequency to deviate from the resonant frequency of the absorption cell and the response frequency of the quartz tuning fork. When this frequency deviation occurs, the photoacoustic conversion efficiency is significantly reduced, resulting in a weakening of the detected sound wave signal intensity, seriously affecting the accurate detection of gas concentrations by QEPAS technology and reducing the reliability and accuracy of the technology in practical applications. Therefore, how to effectively correct the interference of complex environmental gases on the detection results of QEPAS technology has become an important issue that needs to be addressed. Summary of the Invention

[0007] (1) Technical issues to be resolved

[0008] To address these issues, the present invention provides a method and device for correcting ambient gas interference in a quartz-enhanced photoacoustic spectroscopy (QEPAS) detection system. This method continuously collects tuning fork electrical signals, light intensity signals, and ambient gas composition vectors, and inputs this real-time data into a trained neural network model. Based on this input data, the model quickly and accurately outputs the corrected concentration of the gas being measured, enabling real-time, precise detection of the gas under ambient gas interference. This method corrects the effects of complex atmosphere relaxation characteristics on acoustic wave detection, improves the sensitivity of multiple gas detection, and promotes the application of QEPAS technology in various gas detection applications.

[0009] (2) Technical solution

[0010] In response to the above technical problems, embodiments of the present invention provide a method and device for correcting ambient gas interference in a quartz-enhanced photoacoustic spectroscopy detection system.

[0011] According to a first aspect of the present invention, a method for correcting environmental gas interference in a quartz-enhanced photoacoustic spectroscopy detection system is provided, comprising: analyzing the vibration-translational relaxation time and infrared absorption peak intensity of each gas to be measured; selecting a gas to be measured that satisfies both the vibration-translational relaxation time exceeding a time threshold and the absorption peak intensity exceeding an intensity threshold as a comparison gas; dividing the laser of the comparison gas into a first laser beam and a second laser beam according to a preset ratio; utilizing the first laser beam to excite an acoustic wave signal and detecting it with a quartz tuning fork to generate a tuning fork electrical signal; subjecting the second laser beam to enhanced absorption in a multi-reflection absorption cell and then detecting it with a photodetector to generate a light intensity electrical signal; establishing a neural network model with the tuning fork electrical signal, the light intensity electrical signal and the environmental gas component vector as input and the concentration vector of the gas to be measured as output; and correcting the concentration of the gas to be measured under environmental gas interference in real time using the trained neural network model.

[0012] In some exemplary embodiments, when analyzing the vibration-translation relaxation time of each gas to be measured, each gas to be measured is in a pure environment without other gases.

[0013] In some exemplary embodiments, the laser light for comparison gas is divided into the first laser beam and the second laser beam according to a preset ratio, including: using a semi-transparent semi-reflective mirror or an optical fiber coupler to divide the laser light for comparison gas into the first laser beam and the second laser beam according to a ratio of 90% and 10%.

[0014] In some exemplary embodiments, the multiple reflection absorption cell reflects the laser light path multiple times, so that the second laser beam interacts fully with the gas to enhance the absorption effect.

[0015] In some exemplary embodiments, when establishing a neural network model, standard concentration gas and a gas distributor are used to generate a variety of environmental gas samples with different compositions and concentrations. The types and quantities of environmental gas samples are optimized according to the gas composition of the actual application environment to improve the training efficiency and generalization ability of the neural network model.

[0016] In some exemplary embodiments, when correcting the concentration of the gas to be measured under the interference of ambient gas in real time, the tuning fork electrical signal, light intensity electrical signal and ambient gas composition vector are continuously collected, and the collected real-time data are input into the trained neural network model to output the corrected concentration of the gas to be measured.

[0017] In some exemplary embodiments, the time threshold is 10 times the minimum vibration-translation relaxation time of each gas to be measured.

[0018] In some exemplary embodiments, the neural network model includes: an input preprocessing unit, which is used to perform phase-locked amplification on the tuning fork electrical signal, and the detection frequency is the resonant frequency of the quartz tuning fork; a feature fusion layer, which is used to normalize and splice the tuning fork electrical signal, light intensity electrical signal and ambient gas component vector extracted by the phase-locked amplification; a convolutional neural network module, which is used to extract the joint features of the normalized and spliced ​​multiple signals; a long and short-term memory module, which is used to capture the temporal change characteristics of the ambient gas components; and a regression output layer, which is used to output the concentration values ​​of each gas to be measured through an activation function.

[0019] In some exemplary embodiments, the training of the neural network model adopts a transfer learning strategy, including: using a simulated data set generated by a gas distribution instrument to train a convolutional neural network module and a long short-term memory module; loading data collected on site, freezing the weights of the convolutional neural network module, and only optimizing the weights of the regression output layer.

[0020] According to a second aspect of the present invention, there is provided an environmental gas interference correction device for a quartz enhanced photoacoustic spectroscopy detection system, comprising: a testing module for analyzing the vibration-translational relaxation time and infrared absorption peak intensity of each gas to be tested; a selection module for selecting a gas to be tested that satisfies both the vibration-translational relaxation time exceeding a time threshold and the absorption peak intensity exceeding an intensity threshold as a comparison gas; a beam splitting module for splitting the laser of the comparison gas into a first laser beam and a second laser beam according to a preset ratio; a first generating module for exciting an acoustic wave signal with the first laser beam and detecting it with a quartz tuning fork to generate a tuning fork electrical signal; a second generating module for enhancing the absorption of the second laser beam by multiple reflections in an absorption cell and detecting it with a photodetector to generate a light intensity electrical signal; a model building module for establishing a neural network model with the tuning fork electrical signal, the light intensity electrical signal and the environmental gas component vector as input and the concentration vector of the gas to be tested as output; and a concentration testing module for correcting the concentration of the gas to be tested under environmental gas interference in real time through the trained neural network model.

[0021] (3) Beneficial effects

[0022] It can be seen from the above technical solutions that the method and device for correcting ambient gas interference in a quartz-enhanced photoacoustic spectroscopy detection system provided by the embodiments of the present invention have at least the following beneficial effects:

[0023] (1) When there are many types of gases and their concentrations change in real time, the parameters affecting the multi-component gases and their concentration changes increase exponentially, making it difficult to analyze them using physical models. Neural networks are a powerful tool for complex spectral recognition due to their learnability and strong generalization capabilities.

[0024] (2) A photoelectric detector is used to detect the light intensity changes caused by laser infrared absorption. It does not go through the photoacoustic process and is not affected by the VT relaxation characteristics during the acoustic wave generation process. It is used to correct the interference of the ambient mixed gas introduced by the acoustic wave generation process, and can include the correction of the interference of unknown gases therein. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The above contents and other objects, features and advantages of the present invention will become more apparent through the following description of the embodiments of the present invention with reference to the accompanying drawings, in which:

[0026] Figure 1 A flowchart schematically illustrates a method for correcting environmental gas interference in a quartz-enhanced photoacoustic spectroscopy detection system according to an embodiment of the present invention;

[0027] Figure 2 The block diagram schematically shows an environmental gas interference correction device for a quartz-enhanced photoacoustic spectroscopy detection system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0028] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings.

[0029] The terms used herein are only for describing specific embodiments and are not intended to limit the present invention. The terms "comprise," "include," etc. used herein indicate the presence of features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0030] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0031] Figure 1 The flowchart schematically shows a method for correcting environmental gas interference in a quartz-enhanced photoacoustic spectroscopy detection system according to an embodiment of the present invention.

[0032] like Figure 1 As shown, a method for correcting environmental gas interference in a quartz-enhanced photoacoustic spectroscopy detection system according to an embodiment of the present invention includes operations S110 to S170.

[0033] In operation S110, the vibration-translation relaxation time and infrared absorption peak intensity of each gas to be tested are analyzed. When analyzing the vibration-translation relaxation time of each gas to be tested, it is necessary to place it in a pure environment without other gases.

[0034] In some exemplary embodiments, step S110 specifically includes: establishing an experimental platform comprising a gas absorption cell, a tunable laser source, and a signal detection and analysis system. The gas absorption cell is used to contain the gas to be tested. The tunable laser source can output laser light of a specific wavelength to excite gas molecules. The signal detection and analysis system is used to record and analyze the relevant signals. The gas to be tested is introduced into the gas absorption cell, ensuring that the gas is in a pure environment, that is, free of other gas components that may interfere with the detection. The tunable laser source is used to output laser light of a specific wavelength to illuminate the gas to be tested, stimulating its infrared absorption. The signal detection device records the process by which the gas molecules deexcite from a vibrational state to a translational state through non-radiative recombination and release of heat energy. The energy change efficiency of the gas molecules during this process is analyzed, thereby inverting the vibrational-translational relaxation time of each gas to be tested. The wavelength of the tunable laser source is adjusted to scan across the infrared absorption band of the gas to be tested. A photodetector is used to record the gas's absorption of the laser light at different wavelengths, and the infrared absorption spectrum of the gas is plotted using data processing and analysis software. The infrared absorption peak position of each gas to be measured is determined according to the absorption spectrum, and the absorption peak intensity is calculated.

[0035] In operation S120 , a gas to be tested that satisfies both the vibration-translation relaxation time exceeding a time threshold and the absorption peak intensity exceeding an intensity threshold is selected as a comparison gas.

[0036] Optionally, the time threshold is 10 times the minimum vibration-translation relaxation time of each gas to be measured.

[0037] Optionally, the intensity threshold may be set according to the sensitivity and noise level of the detection system.

[0038] While gases with longer relaxation times are more susceptible to interference from changes in ambient gas composition and concentration, these interferences can be effectively compensated and corrected through the proper design of the detection system and signal processing algorithms. Furthermore, the strong signal generated by the higher infrared absorption peak intensity can enhance the detection system's ability to identify the target gas, further strengthening the system's anti-interference capabilities.

[0039] In operation S130 , laser light for comparing gas is divided into a first laser beam and a second laser beam according to a preset ratio.

[0040] For example, a semi-transparent, semi-reflective mirror or fiber coupler can be used to split the laser light from the comparison gas into a first laser beam and a second laser beam at a ratio of 90% to 10%. Semi-transparent, semi-reflective mirrors offer advantages such as simple structure and low cost, making them suitable for applications where beam splitting accuracy is not critical. Fiber couplers, on the other hand, offer advantages such as precise splitting ratios and low insertion loss, making them suitable for applications with high requirements for laser transmission and beam splitting. This 90% to 10% ratio ensures that the first laser beam has sufficient energy to generate a clear and detectable acoustic signal while also enabling the second laser beam to provide accurate light intensity information.

[0041] In operation S140 , an acoustic wave signal is excited by the first laser beam and detected by the quartz tuning fork to generate a tuning fork electrical signal.

[0042] In some exemplary embodiments, a focusing lens is used to focus the first laser beam onto a specific location within the gas absorption cell, ensuring that the laser energy is concentrated in the gas region to be measured, thereby improving the efficiency of the laser-gas interaction. The laser irradiation angle is adjusted to ensure that the laser can fully excite the gas molecules to vibrate. After absorbing the laser energy, the gas molecules de-excite and release heat energy through a non-radiative recombination process, causing periodic temperature changes in the gas medium and, in turn, generating periodic pressure fluctuations, i.e., an acoustic signal. The frequency of the acoustic signal is related to the modulation frequency of the laser. A quartz tuning fork is placed in a suitable position so that it can sensitively detect the tiny vibrations caused by the acoustic signal. The quartz tuning fork has an extremely high Q value, effectively filtering out interference from non-resonant frequencies and responding only to acoustic signals of a specific frequency. When the frequency of the acoustic signal matches the resonant frequency of the quartz tuning fork, the quartz tuning fork vibrates, and the piezoelectric effect converts the mechanical vibrations into an electrical signal, i.e., the tuning fork electrical signal.

[0043] Using a quartz tuning fork to detect acoustic signals offers the advantages of high sensitivity and selectivity. The high Q factor of the quartz tuning fork enables accurate detection of acoustic signals of specific frequencies, effectively suppressing interference from ambient noise and other non-target signals, thereby improving the accuracy and reliability of acoustic signal detection. The generated electrical signal from the tuning fork directly reflects the absorption of laser light by gas molecules and the intensity of the acoustic signal.

[0044] In operation S150 , the second laser beam is enhanced and absorbed by the multi-reflection absorption cell and then detected by a photodetector to generate a light intensity electrical signal.

[0045] In some exemplary embodiments, the multiple reflection absorption cell increases the absorption optical path so that the second laser beam interacts fully with the gas to enhance the absorption effect.

[0046] In the embodiments of this application, the use of a multi-reflection absorption cell enhances the gas's absorption of laser light, improving the sensitivity of light intensity detection. By increasing the interaction length between the laser and the gas, significant changes in light intensity can be generated even at low gas concentrations. This enables the photodetector to detect weaker gas absorption signals, improving the detection system's ability to detect low-concentration gases. The resulting light intensity electrical signal is closely correlated with the gas concentration, providing an important basis for subsequent concentration inversion.

[0047] In operation S160 , a neural network model is established with the tuning fork electrical signal, the light intensity electrical signal, and the ambient gas composition vector as inputs and the measured gas concentration vector as output.

[0048] When establishing the neural network model, standard concentration gas and gas distribution instruments are used to generate a variety of environmental gas samples with different compositions and concentrations. The types and quantities of environmental gas samples are optimized according to the gas composition of the actual application environment to improve the training efficiency and generalization ability of the neural network model.

[0049] In some exemplary embodiments, the neural network model includes: an input preprocessing unit for performing phase-locked amplification on the tuning fork electrical signal, and the detection frequency is the resonant frequency of the quartz tuning fork; a feature fusion layer for normalizing and splicing the tuning fork electrical signal, light intensity electrical signal and ambient gas component vector extracted by the phase-locked amplification; a convolutional neural network module for extracting the joint features of the normalized and spliced ​​multiple signals; a long short-term memory module for capturing the temporal variation characteristics of the ambient gas components; and a regression output layer for outputting the concentration values ​​of each gas to be measured through an activation function.

[0050] The neural network model training adopts the transfer learning strategy, including: using the simulated data set generated by the gas distribution instrument to train the convolutional neural network module and the long short-term memory module; loading the data collected on site, freezing the weights of the convolutional neural network module, and only optimizing the weights of the regression output layer.

[0051] The neural network model possesses powerful nonlinear mapping and adaptive learning capabilities, automatically learning the complex relationship between input data (tuning fork electrical signals, light intensity electrical signals, and ambient gas composition vectors) and output data (measured gas concentration vectors). Compared to traditional physical models, the neural network model does not require the establishment of precise physical equations, can better adapt to complex environments and gas characteristics, and improves the accuracy and reliability of concentration inversion.

[0052] By incorporating the ambient gas component vector as input, the neural network model can account for the impact of ambient gases on detection results, thereby automatically correcting for ambient gas interference. The combined use of convolutional neural network modules and long-short-term memory modules enables the model to simultaneously extract both spatial and temporal features of the input data, further improving the model's ability to process complex signals and the accuracy of concentration predictions.

[0053] In operation S170 , the concentration of the gas to be measured under the interference of the ambient gas is corrected in real time using the trained neural network model.

[0054] When correcting the concentration of the gas to be measured under the interference of ambient gas in real time, the tuning fork electrical signal, light intensity electrical signal and ambient gas composition vector are continuously collected, and the collected real-time data are input into the trained neural network model to output the corrected concentration of the gas to be measured.

[0055] The trained neural network model can quickly and accurately output the concentration of the gas being measured based on the input data. Because the model has learned the impact of ambient gas interference on the detection results during training, the output concentration results have automatically corrected the errors caused by ambient gas interference.

[0056] The embodiments of the present application realize real-time detection and correction of the concentration of the gas to be measured, can promptly reflect the changes in gas concentration, and provide real-time and accurate data support for environmental monitoring, industrial control and other fields. In practical applications, gas leaks, abnormal concentrations and other situations can be discovered in a timely manner, and corresponding measures can be taken to deal with them to ensure production safety and environmental quality. The automatic correction function of the neural network model effectively improves the adaptability and accuracy of the detection system in complex environments. Environmental gas interference can be corrected without human intervention, which reduces the impact of human factors on the test results and improves the reliability and stability of the test.

[0057] Figure 2 The block diagram schematically shows an environmental gas interference correction device for a quartz-enhanced photoacoustic spectroscopy detection system according to an embodiment of the present invention.

[0058] like Figure 2 The quartz enhanced photoacoustic spectroscopy detection system environmental gas interference correction device 800 of the embodiment shown includes a testing module 810, a selection module 820, a beam splitting module 830, a first generation module 840, a second generation module 850, a model building module 860 and a concentration testing module 870.

[0059] The testing module 810 is used to analyze the vibration-translation relaxation time and infrared absorption peak intensity of each gas to be tested.

[0060] The selection module 820 is configured to select a gas to be tested that satisfies both the vibration-translation relaxation time exceeding a time threshold and the absorption peak intensity exceeding an intensity threshold as a comparison gas.

[0061] The beam splitting module 830 is used to split the laser beam for comparing the gas into a first laser beam and a second laser beam according to a preset ratio.

[0062] The first generating module 840 is configured to generate a tuning fork electrical signal by exciting the acoustic wave signal with the first laser beam and detecting the signal with the quartz tuning fork.

[0063] The second generating module 850 is used to generate a light intensity electrical signal by detecting the second laser beam through a photoelectric detector after enhanced absorption by the multi-reflection absorption cell.

[0064] The model building module 860 is used to establish a neural network model with the tuning fork electrical signal, the light intensity electrical signal and the ambient gas composition vector as input and the measured gas concentration vector as output.

[0065] The concentration test module 870 is used to correct the concentration of the gas to be tested under the interference of environmental gases in real time through the trained neural network model.

[0066] According to embodiments of the present invention, any multiple modules among the testing module 810, the selection module 820, the beam splitting module 830, the first generation module 840, the second generation module 850, the model building module 860, and the concentration testing module 870 may be combined into a single module, or any one of these modules may be split into multiple modules. Alternatively, at least part of the functionality of one or more of these modules may be combined with at least part of the functionality of other modules and implemented in a single module. According to embodiments of the present invention, at least one of the testing module 810, the selection module 820, the beam splitting module 830, the first generation module 840, the second generation module 850, the model building module 860, and the concentration testing module 870 may be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or may be implemented in hardware or firmware by any other reasonable means of circuit integration or packaging, or may be implemented in any one of software, hardware, and firmware, or any suitable combination thereof. Alternatively, at least one of the testing module 810, the selection module 820, the beam splitting module 830, the first generation module 840, the second generation module 850, the model building module 860, and the concentration testing module 870 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is executed.

[0067] It will be understood by those skilled in the art that the features described in the various embodiments of the present invention may be combined and / or coupled in various ways, even if such combinations or couplings are not explicitly described in the present invention. In particular, the features described in the various embodiments of the present invention may be combined and / or coupled in various ways without departing from the spirit and teachings of the present invention. All such combinations and / or couplings fall within the scope of the present invention.

Claims

1. A method for correcting ambient gas interference in a quartz-enhanced photoacoustic spectroscopy detection system, characterized in that: include: Analyze the vibration-translation relaxation time and infrared absorption peak intensity of each gas to be tested; Selecting a gas to be tested that satisfies both the vibration-translation relaxation time exceeding the time threshold and the absorption peak intensity exceeding the intensity threshold as a comparison gas; Splitting the laser light of the comparison gas into a first laser beam and a second laser beam according to a preset ratio; Using the first laser beam to excite an acoustic wave signal and detecting it with a quartz tuning fork to generate a tuning fork electrical signal; After the second laser beam is enhanced and absorbed by a multi-reflection absorption cell, it is detected by a photodetector to generate a light intensity electrical signal; Establishing a neural network model with the tuning fork electrical signal, the light intensity electrical signal and the ambient gas composition vector as input and the measured gas concentration vector as output; The trained neural network model is used to correct the concentration of the gas to be measured under the interference of environmental gases in real time.

2. The method according to claim 1, characterized in that When analyzing the vibration-translation relaxation time of each gas to be measured, each gas to be measured is in a pure environment without other gases.

3. The method according to claim 1, characterized in that The laser beam of the comparison gas is divided into a first laser beam and a second laser beam according to a preset ratio, comprising: The laser light of the comparison gas is divided into a first laser beam and a second laser beam in a ratio of 90% to 10% by using a semi-transparent and semi-reflective mirror or an optical fiber coupler.

4. The method according to claim 1, wherein The multiple reflection absorption cell increases the absorption optical path so that the second laser beam fully interacts with the gas, thereby enhancing the absorption effect.

5. The method according to claim 1, wherein When establishing the neural network model, standard concentration gas and a gas distributor are used to generate a variety of environmental gas samples with different compositions and concentrations. The types and quantities of environmental gas samples are optimized according to the gas composition of the actual application environment to improve the training efficiency and generalization ability of the neural network model.

6. The method according to claim 1, wherein When correcting the concentration of the gas to be measured under the interference of ambient gas in real time, the tuning fork electrical signal, light intensity electrical signal and ambient gas composition vector are continuously collected, and the collected real-time data are input into the trained neural network model to output the corrected concentration of the gas to be measured.

7. The method according to claim 1, characterized in that The time threshold is 10 times the minimum vibration-translation relaxation time of each gas to be measured.

8. The method according to claim 1, characterized in that The neural network model includes: An input preprocessing unit, configured to perform phase-locked amplification on the tuning fork electrical signal, wherein the detection frequency is the resonant frequency of the quartz tuning fork; The feature fusion layer is used to normalize and splice the tuning fork electrical signal, light intensity electrical signal, and ambient gas component vector extracted by phase-locked amplification; Convolutional neural network module, used to extract joint features of multiple signals after normalization and splicing; Long short-term memory module, used to capture the temporal variation characteristics of ambient gas composition; The regression output layer is used to output the concentration value of each gas to be measured through the activation function.

9. The method according to claim 8, characterized in that The training of the neural network model adopts a transfer learning strategy, including: Using a simulated data set generated by a gas distribution instrument to train the convolutional neural network module and the long short-term memory module; The data collected on site is loaded, the weights of the convolutional neural network module are frozen, and only the weights of the regression output layer are optimized.

10. A device for correcting environmental gas interference in a quartz-enhanced photoacoustic spectroscopy detection system, characterized in that: include: The test module is used to analyze the vibration-translation relaxation time and infrared absorption peak intensity of each gas to be tested; A selection module is used to select a gas to be tested that satisfies both the vibration-translation relaxation time exceeding the time threshold and the absorption peak intensity exceeding the intensity threshold as a comparison gas; A beam splitting module, configured to split the laser light of the comparison gas into a first laser beam and a second laser beam according to a preset ratio; a first generating module, configured to generate a tuning fork electrical signal by exciting an acoustic wave signal with the first laser beam and detecting the signal with a quartz tuning fork; a second generating module, configured to generate a light intensity electrical signal by detecting the second laser beam through a photoelectric detector after enhanced absorption by the multi-reflection absorption cell; A model building module, for establishing a neural network model with the tuning fork electrical signal, the light intensity electrical signal and the ambient gas composition vector as input and the measured gas concentration vector as output; The concentration test module is used to correct the concentration of the gas to be tested under the interference of environmental gases in real time through the trained neural network model.

Images