NDIR mixed gas detection method and device

Abstract

The invention discloses an NDIR (Non-Dispersive Infra-Red) mixed gas detection method and device, and belongs to the technical field of gas detection. Considering that the absorption peak wavelengths of different gases are different and the concentrations of the gases directly influence the corresponding absorption intensities, the mixed gas infrared spectrum carrying spectral value information under different wavelengths is directly obtained, and an NDIR mixed gas detection model is trained, so that the detection accuracy of the NDIR mixed gas is improved. The mapping relation between the infrared spectrum of the mixed gas to be detected and the classification result and concentration value of each component gas in the mixed gas to be detected is constructed, so that the detection of the NDIR mixed gas is realized; according to the invention, a multi-path optical filter is not needed to separate the mixed gas into gases with different peak absorption wavelengths, the sensitivity is relatively high, gases with different components and relatively close peak absorption wavelengths can be accurately distinguished, and multi-component gas detection can be accurately realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of gas detection, and more specifically, relates to a NDIR mixed gas detection method and device. Background Art

[0002] In recent years, with the rapid development of social economy and heavy industry, various toxic, harmful, flammable and explosive gases have been produced and spread invisibly. These gases can endanger personal safety and cause major accidents. Therefore, it is of great industrial significance to carry out high-precision type identification and concentration detection of these harmful gases.

[0003] The detection of multi-component mixed gases is a very difficult problem in gas detection and gas sensing. Currently, common gas detection methods mainly include electrochemical detection, catalytic combustion detection, and gas chromatography detection, but these methods all have some limitations. Compared with the above-mentioned gas detection methods, infrared spectroscopy has the advantages of high sensitivity and high precision. The non-dispersive inifrared (NDIR) detection method in infrared spectroscopy does not require the light source to be split, but uses a filter device to perform split processing at the detector end, which greatly reduces the volume of the entire gas detection system, making the system miniaturized and integrated. It can be used in various extreme environments to monitor the gas environment in real time. The equipment has good stability, can be used for a long time, has low maintenance costs, and has high industrial application value.

[0004] The existing NDIR mixed gas detection method uses a multi-channel filter to separate the mixed gas into gases with different peak absorption wavelengths, and then splits them for photoelectric conversion. By detecting the voltage on different optical paths, and establishing a mapping relationship between the voltage on each optical path and the corresponding gas concentration based on the Lambert-Beer law, the mixed gas detection is achieved. However, due to the large attenuation of the filter, the sensitivity to the lower-concentration component gas in the mixed gas is low, and it is impossible to accurately distinguish different components of gas with close peak absorption wavelengths, which affects the accuracy of detection. Summary of the invention

[0005] In view of the above defects or improvement needs of the prior art, the present invention provides a NDIR mixed gas detection method and device to solve the technical problem that the prior art cannot accurately realize multi-component gas detection.

[0006] In order to achieve the above object, in a first aspect, the present invention provides a NDIR mixed gas detection method, comprising:

[0007] The infrared spectrum of the mixed gas to be detected is input into the NDIR mixed gas detection model to obtain the classification results and concentration values ​​of each component gas in the mixed gas to be detected;

[0008] Among them, the NDIR mixed gas detection model is trained in the following way:

[0009] The pre-collected training set is input into the NDIR mixed gas detection model, and the NDIR mixed gas detection model is trained with the goal of minimizing the training loss;

[0010] The training set includes: infrared spectra of different mixed gases and corresponding labels; the labels include: the true category and true concentration value of each component gas in the mixed gas;

[0011] The training loss includes: the difference loss between the classification result of each component gas in the mixed gas output by the NDIR mixed gas detection model and the corresponding true category, and the difference loss between the concentration value of each component gas in the mixed gas output by the NDIR mixed gas detection model and the corresponding true concentration value.

[0012] Further preferably, the infrared spectrum of the mixed gas to be detected is obtained by an infrared detection device;

[0013] The infrared detection device includes: an air chamber, a light source and a spectrum detection module;

[0014] The gas chamber is provided with a vent hole for introducing the mixed gas to be detected;

[0015] The light source is used to provide infrared light;

[0016] The spectrum detection module is used to receive infrared light passing through the mixed gas, and then detect and obtain the infrared spectrum of the mixed gas to be detected.

[0017] Further preferably, the above-mentioned spectrum detection module is a metamaterial detector;

[0018] The metamaterial detector comprises: a substrate, and M metamaterial detection units disposed on the substrate and arranged in an array;

[0019] Each metamaterial detection unit includes: a thermoelectric conversion layer and an optical antenna array of a sub-wavelength structure arranged in sequence from bottom to top;

[0020] The optical antenna array is used to receive infrared light signals corresponding to the absorption peak wavelength;

[0021] The thermoelectric conversion layer is used to perform thermoelectric conversion on the infrared light signal received by the optical antenna array thereon to obtain the infrared spectrum value at the corresponding absorption peak wavelength;

[0022] Among the M optical antenna arrays included in the metamaterial detector, the absorption peaks of N optical antenna arrays are λ1,λ2,......,λ N ;0<N≤M;λ1,λ2,...,λ N The infrared spectrum values ​​under the above two together constitute the infrared spectrum of the mixed gas;

[0023] Among them, λ j It is the absorption peak wavelength of the jth single gas in the preset gas set, or the central wavelength of the jth sub-band of the preset band; j=1, 2, ..., N; the preset gas set is a set composed of N single gases of different preset types; the preset band is divided into N sub-bands.

[0024] Further preferably, the absorption peak is λ j The optical antennas in the optical antenna array are cylindrical, and the spacing between the optical antennas is λ j , the diameter of the optical antenna on it is λ j / 2.

[0025] Further preferably, the absorption peak is λ j The optical antennas in the optical antenna array are cross-shaped, and the spacing between the optical antennas is 5λ. j / 8, the lengths of the two arms of the optical antenna are 3λ j / 10 and λ j / 10.

[0026] Further preferably, the material of the thermoelectric conversion layer is a pyroelectric material.

[0027] Further preferably, the inner wall of the air chamber is provided with a diffuse reflection coating for diffusely reflecting the infrared light.

[0028] In a second aspect, the present invention provides an NDIR mixed gas detection device, comprising:

[0029] Infrared detection device, used for detecting the infrared spectrum of the mixed gas to be detected;

[0030] The gas detection module is used to execute the NDIR mixed gas detection method provided by the first aspect of the present invention.

[0031] Further preferably, the above-mentioned infrared detection device is the infrared detection device used in the NDIR mixed gas detection method provided in the first aspect of the present invention.

[0032] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects:

[0033] 1. The present invention provides a NDIR mixed gas detection method. Considering that the absorption peak wavelengths of different gases are different and the concentration of the gas will directly affect the corresponding absorption intensity, the present invention directly obtains the infrared spectrum of the mixed gas carrying the spectral value information at different wavelengths, and trains an NDIR mixed gas detection model to construct a mapping relationship between the infrared spectrum of the mixed gas to be detected and the classification results and concentration values ​​of each component gas in the mixed gas to be detected, thereby realizing the detection of the NDIR mixed gas; the present invention does not need to use a multi-channel filter to separate the mixed gas into gases with different peak absorption wavelengths, has high sensitivity, can accurately distinguish gases of different components with relatively close peak absorption wavelengths, and can accurately realize multi-component gas detection.

[0034] 2. Furthermore, currently, multi-component gas detection is generally completed using multiple discrete devices. If a single device is used to detect a gas, the more target gases there are, the more devices are required, the higher the cost, the larger the volume, and the extremely limited use. The present invention obtains the infrared spectrum of the mixed gas to be detected through an infrared detection device, wherein the metamaterial detector includes: a substrate, and M metamaterial detection units arranged on the substrate and arranged in an array; each metamaterial detection unit includes: a thermoelectric conversion layer and an optical antenna array of a subwavelength structure arranged in sequence from bottom to top, which can more sensitively perceive slight changes in gas concentration; at the same time, among the M optical antenna arrays contained in the metamaterial detector, N optical antenna arrays have different absorption peaks, namely λ1, λ2,......,λ N ; j It is the absorption peak wavelength of the jth single gas in the preset gas concentration, or the central wavelength of the jth sub-band of the preset band; each sub-wavelength structure of the optical antenna array can correspond to a gas under the absorption peak wavelength alone, the volume of the whole device is small, and one device can be used to complete the detection of the required range or band. There is no upper limit to the types of target gases, and there is no need to increase the volume exponentially when expanding the detection range. It has the advantages of flexible and wide application.

[0035] 3. Furthermore, the material of the thermoelectric conversion layer in the metamaterial detector provided by the present invention is pyroelectric material. Compared with photoelectric materials, pyroelectric materials have a wider detection wavelength range, which can cover the near-infrared and mid-infrared ranges and are suitable for mixed gas detection; at the same time, pyroelectric materials have high thermoelectric conversion capabilities, high specific detection rates, and high sensitivity, and are suitable for low-concentration gas detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A schematic diagram of a multi-component mixed gas distribution system provided by an embodiment of the present invention;

[0037] Figure 2 A schematic diagram of the structure of an infrared detection device provided by an embodiment of the present invention;

[0038] Figure 3 A schematic diagram of the absorption spectrum of a mixed gas of CO2 and N2O for a metamaterial detector provided in an embodiment of the present invention;

[0039] Figure 4 A schematic diagram of the structure of a four-channel metamaterial detector provided by an embodiment of the present invention;

[0040] Figure 5 A schematic diagram of using a neural network model as an NDIR mixed gas detection model provided by an embodiment of the present invention;

[0041] Figure 6 A graph showing the average predicted relative error of each component gas of the mixed gas provided by an embodiment of the present invention at different concentration values;

[0042] Figure 7 A schematic diagram of the NDIR mixed gas detection process provided by an embodiment of the present invention;

[0043] Figure 8 A schematic top view of a sixteen-channel metamaterial detector provided by an embodiment of the present invention;

[0044] Fig. 9 A schematic top view of a 10*10 channel metamaterial detector provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0046] In order to achieve the above object, in a first aspect, the present invention provides a NDIR mixed gas detection method, comprising:

[0047] The infrared spectrum of the mixed gas to be detected is input into the NDIR mixed gas detection model to obtain the classification results and concentration values ​​of each component gas in the mixed gas to be detected;

[0048] Among them, the NDIR mixed gas detection model is trained in the following way:

[0049] The pre-collected training set is input into the NDIR mixed gas detection model, and the NDIR mixed gas detection model is trained with the goal of minimizing the training loss;

[0050] The training set includes: infrared spectra of different mixed gases and corresponding labels; the labels include: the true category and true concentration value of each component gas in the mixed gas sample;

[0051] The training loss includes: the difference loss between the classification result of each component gas in the mixed gas output by the NDIR mixed gas detection model and the corresponding true category, and the difference loss between the concentration value of each component gas in the mixed gas output by the NDIR mixed gas detection model and the corresponding true concentration value.

[0052] It should be noted that the NDIR mixed gas detection model is a machine model, which can be a SVM model, a neural network model, a CNN model, a K-nearest neighbor regression model, a random forest model, etc., which is not limited here.

[0053] In an optional implementation manner, the infrared spectrum of the mixed gas to be detected is obtained by an infrared detection device;

[0054] The infrared detection device includes: an air chamber, a light source and a spectrum detection module;

[0055] The gas chamber is provided with a vent hole for introducing the mixed gas to be detected;

[0056] The light source is used to provide infrared light;

[0057] The spectrum detection module is used to receive infrared light passing through the mixed gas, and then detect and obtain the infrared spectrum of the mixed gas to be detected.

[0058] It should be noted that there are many spectral detection modules that can be used, such as infrared spectrometers, which are not limited here. However, considering that the infrared spectrometer is large in size and high in cost, preferably, in an optional implementation, the present invention provides a metamaterial detector for detecting mixed gases that fall within the detection object range of the metamaterial detector.

[0059] The metamaterial detector includes: a substrate, and M metamaterial detection units disposed on the substrate and arranged in an array;

[0060] Each metamaterial detection unit includes: a thermoelectric conversion layer and an optical antenna array of a sub-wavelength structure arranged in sequence from bottom to top;

[0061] The optical antenna array is used to receive infrared light signals corresponding to the absorption peak wavelength;

[0062] The thermoelectric conversion layer is used to perform thermoelectric conversion on the infrared light signal received by the optical antenna array thereon to obtain the infrared spectrum value at the corresponding absorption peak wavelength;

[0063] Among the M optical antenna arrays included in the metamaterial detector, the absorption peaks of N optical antenna arrays are λ1,λ2,......,λ N ;0<N≤M;λ1,λ2,...,λ N The infrared spectrum values ​​under the above two together constitute the infrared spectrum of the mixed gas;

[0064] The detection range of the metamaterial detector is a set of mixed gases formed by mixing multiple gases arbitrarily selected from a preset gas set, including range one or range two;

[0065] In the range below, the preset gas set is a set of single gases of N different preset types; j is the absorption peak wavelength of the jth single gas in the preset gas set;

[0066] In range 2, the preset gas set is a set consisting of single gases whose absorption peak wavelengths belong to N sub-bands of the preset band; j is the central wavelength of the jth sub-band of the preset band;

[0067] j=1,2,......,N.

[0068] It should be noted that, in an optional implementation, the preset gas set includes CO2, CF4, N2O, SF6, NO2, SO2, C4H 10 , CH4, C2H2, C2H4 and other single types of gases.

[0069] The preset band can be a full band, a full infrared band, a near infrared band, etc., and can be set according to user needs without limitation.

[0070] It should be noted that the optical antennas in the optical antenna array may have different shapes, preferably, may be cylinders or crosses.

[0071] When the optical antennas in a certain optical antenna array are cylindrical, the spacing between the optical antennas on the optical antenna array is λ j , the diameter of the optical antenna is λ j / 2, at this time, the absorption peak of the optical antenna array is λ j .

[0072] When the optical antennas in a certain optical antenna array are cross-shaped, the spacing between the optical antennas on the optical antenna array is 5λ j / 8, the lengths of the two arms of the optical antenna are 3λ j / 10 and λ j / 10, at this time, the absorption peak of the optical antenna array is λ j .

[0073] It should be noted that the material of the thermoelectric conversion layer may be a pyroelectric material, a thermopile material, etc. Pyroelectric materials have a higher specific detection rate and a wider detection wavelength range, so pyroelectric materials are preferably used. Among pyroelectric materials, lithium tantalate has the highest pyroelectric coefficient. In an optional embodiment, the material of the thermoelectric conversion layer is lithium tantalate.

[0074] In an optional embodiment, the inner wall of the air chamber is provided with a diffuse reflection coating for diffusely reflecting infrared light. Preferably, in an optional embodiment, the diffuse reflection coating material is gold.

[0075] In order to further illustrate the NDIR mixed gas detection method provided by the present invention, some specific embodiments are described in detail below:

[0076] Example 1

[0077] In this embodiment, a metamaterial detector is used as a spectrum detection module. After the incident wide-spectrum infrared light passes through the metamaterial detection unit with different structural parameters, a series of transmission spectrum data is obtained, which shows the selective absorption of infrared light by the multi-component gas itself, thereby constructing an NDIR mixed gas detection model representing the infrared spectrum of the mixed gas and the mapping relationship between the multi-component gas types and concentrations in the mixed gas.

[0078] Specifically, in this embodiment, the detection process of the NDIR mixed gas detection method includes:

[0079] 1) Construction of NDIR mixed gas detection model

[0080] 1.1) Get the training set:

[0081] A mixed gas composed of different types of component gases is obtained, and the types and concentrations of different component gases in the mixed gas are used as labels to establish the label part of the training set. Specifically, Figure 1 As shown, a multi-channel gas mass flow controller and a highly sealed gas mixing device are used to mix different types of component gases respectively, and then introduced into the reflective gas chamber with a gold-plated mirror on the inner wall of the infrared detection device. Each group of experiments is ventilated until the gas concentration in the gas chamber is stable. The infrared detection device in this embodiment is as shown in FIG. Figure 2 shown.

[0082] After the multi-component gas is introduced into the designed gas chamber according to a certain concentration ratio, a data acquisition card is connected to the metamaterial detector to read the absorption peaks of the metamaterial detector in real time, which are λ1, λ2, ..., λ N The amplified voltage value of the metamaterial detection unit, that is, the infrared spectrum value, λ1,λ2,......,λ NThe infrared spectrum values ​​under the above conditions together constitute the infrared spectrum of the mixed gas (i.e., wavelength-absorption intensity distribution data), which are used as training samples in the training set.

[0083] This embodiment uses a multi-channel gas mass flow controller and a gas mixing device with strong sealing properties to mix gas, and passes it into the inner wall gold-plated mirror reflection gas chamber. Each group of experiments is ventilated until the gas concentration in the gas chamber is stable. The gas mass flow meter is controlled by solenoid valve and software. Replacing flow meters with different ranges can prepare mixed gases of different concentrations. In this process, the standard component gas with a determined concentration is neutralized with pure nitrogen. According to the different volume percentages of the gas, the opening and closing and flow rate of the mass flow meter are controlled to obtain mixed gases with different component concentrations, wherein the gas concentration of each component is non-negative, and the sum of the flow rates of all gases is fixed to the maximum flow rate of the mass flow meter.

[0084] The preset band in this embodiment is the full infrared band to achieve full band coverage. The metamaterial detection unit has a micro-nano structure of a metal optical antenna array, which can achieve selective absorption of infrared light of a specific wavelength and obtain a narrow-band infrared transmission spectrum. Adjusting the size and layout of the optical antenna can change the peak absorption wavelength of the metamaterial detection unit. A large number of metamaterial detection units with different peak absorption wavelengths are integrated on a wafer to achieve full-band infrared spectrum resolution. This embodiment uses a metamaterial detector to divide the full-band infrared light into a series of tiny sub-bands, and detects changes in infrared light on a single sub-band. The detector is coated with a layer of thermoelectric conversion layer material with strong infrared light absorption ability on the base material to enhance the detection ability of weak changes, and then coated with a metal optical antenna structure to adjust the transmission wavelength of the detector. Several detectors are integrated on the same wafer, sharing electrodes, and the working mode is exactly the same.

[0085] After the infrared light emitted by the light source in the infrared detection device is absorbed by the mixed gas and selectively transmitted by the metamaterial detector, only a specific part of the wide-spectrum light input by the light source can be retained. Figure 3 The figure shows the spectrum of a mixed gas of CO2 and N2O. The corresponding results can be obtained by extending it to other gases.

[0086] The training set is preprocessed to obtain normalized sample data to complete the creation of the training set.

[0087] 1.2) Training of NDIR mixed gas detection model:

[0088] The training set is input into the NDIR mixed gas detection model, and the NDIR mixed gas detection model is trained with the goal of minimizing the training loss;

[0089] Among them, the training loss includes: the difference loss between the classification result of each component gas in the mixed gas output by the NDIR mixed gas detection model and the corresponding real category, and the difference loss between the concentration value of each component gas in the mixed gas output by the NDIR mixed gas detection model and the corresponding real concentration value.

[0090] The supervision mechanism is introduced to ensure that the NDIR mixed gas detection model is optimized in the direction of reducing the error. When the error increases continuously with the increase of training times, the training process will be terminated early.

[0091] The NDIR mixed gas detection model in this embodiment is a neural network model established using big data and gradient descent algorithm based on Python language and Keras framework. It aims to train a suitable neural network using the infrared spectrum of mixed gases under partial concentration ratio conditions, and can calculate the concentration value of each component for mixed gases in which the component gases change arbitrarily within a specific concentration range, and ensure that the relative error is at a low level, thereby achieving accurate inversion of the concentration of multi-component mixed gases in the target environment for online monitoring of actual mixed gas environments.

[0092] 2) NDIR mixed gas detection

[0093] The mixed gas to be detected is input into the gas chamber of the infrared detection device, and the infrared spectrum of the mixed gas to be detected is obtained by the metamaterial detector in the infrared detection device, and input into the NDIR mixed gas detection model to obtain the classification results and concentration values ​​of each component gas in the mixed gas to be detected.

[0094] It should be noted that the infrared spectrum of the mixed gas in the training set and the infrared light of the mixed gas to be detected can be obtained by the same device or different devices. In this embodiment, the same infrared detection device is used to obtain the infrared spectrum, which is only an embodiment.

[0095] Example 2

[0096] like Figure 4 As shown, this embodiment uses a four-channel metamaterial detector (M=4), and designs peak transmission wavelengths for four specific gas bands, each corresponding to a component gas. The metamaterial detector outputs spectral data of these four bands, and the micro-nano structures of the four detectors are different.

[0097] In this embodiment, a TO-8 tube shell is used to package the metamaterial detector, a PCB circuit board is used to simultaneously read the data of the four channels of the metamaterial detector, and a single power supply is used to power the circuit, thereby reducing environmental noise and electromagnetic interference and improving the metamaterial detector's ability to detect tiny concentration changes.

[0098] In this embodiment, a four-input four-output neural network model is constructed as the NDIR mixed gas detection model, such as Figure 5 As shown in the figure, the dimensions of both input and output are 4. The model contains three hidden layers, which can realize the fitting of complex multivariate nonlinear functions. A large amount of spectral data of mixed gases with different concentrations are collected to train the model. A monitoring mechanism is introduced to stop training when the error no longer decreases to prevent the model from overfitting.

[0099] The test set is fed into the trained model, and a graph is drawn to reflect the error between the predicted concentration and the actual concentration. The fitting effect is analyzed, and the average predicted relative error of each component gas of the mixed gas at different concentration values ​​is calculated. Figure 6 The results are shown. Figure 6 It can be seen that the average prediction relative errors of different gases at different concentration values ​​are all low, and become lower and lower with the increase of concentration and finally remain at a low error. It can be seen that the NDIR mixed gas detection model provided by the present invention has good multi-component gas concentration prediction ability.

[0100] Specifically, the detection flow chart of the NDIR mixed gas of this embodiment is as follows: Figure 7 As shown, Figure 7 , m=n=L=4.

[0101] Example 3

[0102] like Figure 8 As shown, this embodiment uses a sixteen-channel metamaterial detector (M=16), which can cover a preset band with a bandwidth range of 3-5um, and takes the characteristic absorption wavelengths of several target gases in the band as the main target; the preset band is evenly divided into 16 sub-bands, and the bandwidth of each sub-band is 150nm, that is, the bandwidth of the metamaterial detector of each channel is 150nm, and the output is full-band spectral data composed of sixteen channels. Each metamaterial detector in the metamaterial detector is prepared according to the optimal parameters obtained by comsol simulation design, and has the best absorption capacity and monochromaticity.

[0103] In this embodiment, a QFN tube shell is used to package the detector, and a PCB circuit board is used for power supply and data processing. The PCB board integrates a single-chip microcomputer chip and an AD sampling circuit. After the sixteen-channel data is read out, it is processed into spectral data by the single-chip microcomputer and saved in the host computer.

[0104] In this embodiment, a multi-input and multi-output neural network model is constructed as an NDIR mixed gas detection model. The input dimension can be selected according to the computing power limitation and accuracy requirement. The spectral data is sampled. The more sampling points, the higher the fitting accuracy. The model has multiple hidden layers, uses the ReLU function as the nonlinear function, and uses the Adam method as the training working mode. The output dimension is determined by the number of target gases, and the output data is the concentration of the target gas.

[0105] Example 4

[0106] like Fig. 9 As shown, a 10*10 channel metamaterial detector (M=100) is used, and the bandwidth range of the preset band that can be covered is 3-8um, and the characteristic absorption wavelengths of several target gases in this band are taken as the main targets; the preset band is evenly divided into 100 sub-bands, and the bandwidth of each sub-band is 50nm, that is, the bandwidth of the metamaterial detector of each channel is 50nm, and the output is the spectral data of the full band. The metamaterial detection unit in each metamaterial detector is prepared according to the optimal parameters obtained by the comsol simulation design, ensuring that the metamaterial detector covers the preset band and there is no large-scale overlap or deviation from the central wavelength.

[0107] In this embodiment, a QFN tube shell is used to package the detector, and a PCB circuit board is used for power supply. Data is scanned and collected through an AD conversion chip. The scanning speed used needs to be greater than the response speed of the metamaterial detector. The scanned data is transmitted to a host computer and processed by software to reconstruct the infrared spectrum of the mixed gas.

[0108] In this embodiment, a multi-input and multi-output neural network model is constructed as an NDIR mixed gas detection model. The input dimension can be selected according to the computing power limitation and accuracy requirement, and the spectral data is sampled. The more sampling points, the higher the fitting accuracy. In this embodiment, high requirements are placed on the accuracy of the data, and the number of sampling points should be at least 1024. The model has multiple hidden layers, uses the ReLU function as a nonlinear function, and uses the Adam method as a training working mode. In order to prevent overfitting, some neurons can be randomly set to zero during training to improve the flexibility of the algorithm. The output dimension is determined by the number of target gases, and the output data is the concentration of the target gas.

[0109] In summary, compared with traditional gas sensors, the present invention can be applied to the detection of complex multi-component mixed gases, has the advantages of low noise and high sensitivity, and can detect target gases with lower concentrations. Specifically:

[0110] The metamaterial detector described in the present invention uses a high-response substrate material and a composite metal optical antenna structure, which can sense slight changes in gas concentration, and each part of the optical antenna can correspond to a narrowband spectrum individually. The spectrum of the entire infrared band is detected through the detector array to monitor the spectral changes of the target gas.

[0111] The present invention trains the NDIR mixed gas detection model through a large amount of measured data, and predicts and calculates the concentration of each component gas; and makes up for the defect that the material itself cannot recognize the gas. The algorithm has a strong error correction ability for gas resolution and concentration calculation with overlapping spectra, which can achieve higher-precision detection.

[0112] In addition, currently, multi-component gas detection is generally completed using multiple discrete devices. If a single device is used to detect one gas, the more target gases there are, the more devices are required, the higher the cost, the larger the volume, and the extremely limited use. The detection device based on the metamaterial detector provided by the present invention can use one device to complete the full-band detection, there is no upper limit to the types of target gases, and when expanding the detection range, there is no need to increase the volume exponentially, which has the advantages of flexible and wide application.

[0113] In a second aspect, the present invention provides an NDIR mixed gas detection device, comprising:

[0114] Infrared detection device, used for detecting the infrared spectrum of the mixed gas to be detected;

[0115] The gas detection module is used to execute the NDIR mixed gas detection method provided by the first aspect of the present invention.

[0116] It should be noted that there are many spectral detection modules that can be used, such as infrared spectrometers, which are not limited here. However, considering that the infrared spectrometer is large in size and high in cost, preferably, in an optional embodiment, the infrared detection device is the infrared detection device used in the NDIR mixed gas detection method provided in the first aspect of the present invention.

[0117] The related technical solution is the same as the NDIR mixed gas detection method provided in the first aspect of the present invention, and will not be described in detail here.

[0118] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A NDIR mixed gas detection method, characterized in that: include: The infrared spectrum of the mixed gas to be detected is input into the NDIR mixed gas detection model to obtain the classification results and concentration values ​​of each component gas in the mixed gas to be detected; The NDIR mixed gas detection model is trained in the following way: Inputting the pre-collected training set into the NDIR mixed gas detection model, and training the NDIR mixed gas detection model with the goal of minimizing training loss; The training set includes: infrared spectra of different mixed gases and corresponding labels; the labels include: the true category and true concentration value of each component gas in the mixed gas; The training loss includes: the difference loss between the classification result of each component gas in the mixed gas output by the NDIR mixed gas detection model and the corresponding true category, and the difference loss between the concentration value of each component gas in the mixed gas output by the NDIR mixed gas detection model and the corresponding true concentration value.

2. The NDIR mixed gas detection method according to claim 1, characterized in that: The infrared spectrum of the mixed gas to be detected is obtained by an infrared detection device; Wherein, the infrared detection device comprises: an air chamber, a light source and a spectrum detection module; The air chamber is provided with a vent hole for introducing the mixed gas to be detected; The light source is used to provide infrared light; The spectrum detection module is used to receive infrared light passing through the mixed gas to be detected, and then detect and obtain the infrared spectrum of the mixed gas to be detected.

3. The NDIR mixed gas detection method according to claim 2, characterized in that: The spectrum detection module is a metamaterial detector; The metamaterial detector comprises: a substrate, and M metamaterial detection units disposed on the substrate and arranged in an array; Each of the metamaterial detection units comprises: a thermoelectric conversion layer and an optical antenna array of a sub-wavelength structure arranged in sequence from bottom to top; The optical antenna array is used to receive infrared light signals corresponding to the absorption peak wavelength; The thermoelectric conversion layer is used to perform thermoelectric conversion on the infrared light signal received by the optical antenna array thereon to obtain the infrared spectrum value at the corresponding absorption peak wavelength; Among the M optical antenna arrays included in the metamaterial detector, the absorption peaks of N optical antenna arrays are λ1, λ2, ..., λ N ;0<N≤M;λ1,λ2,...,λ N The infrared spectrum values ​​under the above two together constitute the infrared spectrum of the mixed gas; Among them, λ j It is the absorption peak wavelength of the jth single gas in the preset gas set, or the central wavelength of the jth sub-band of the preset band; j=1, 2, ..., N; the preset gas set is a set composed of single gases of N different preset types; the preset band is divided into N sub-bands.

4. The NDIR mixed gas detection method according to claim 3, characterized in that: The absorption peak is λ j The optical antennas in the optical antenna array are cylindrical, and the spacing between the optical antennas is λ j , the diameter of the optical antenna on it is λ j / 2.

5. The NDIR mixed gas detection method according to claim 3, characterized in that: The absorption peak is λ j The optical antennas in the optical antenna array are cross-shaped, and the spacing between the optical antennas is 5λ. j / 8, the lengths of the two arms of the optical antenna are 3λ j / 10 and λ j / 10.

6. The NDIR mixed gas detection method according to claim 3, characterized in that: The material of the thermoelectric conversion layer is a pyroelectric material.

7. The NDIR mixed gas detection method according to any one of claims 2 to 6, characterized in that: The inner wall of the air chamber is provided with a diffuse reflection coating for diffusely reflecting infrared light.

8. An NDIR mixed gas detection device, characterized in that: include: Infrared detection device, used for detecting the infrared spectrum of the mixed gas to be detected; A gas detection module, used to execute the NDIR mixed gas detection method described in claim 1.

9. The NDIR mixed gas detection device according to claim 8, characterized in that: The infrared detection device is the infrared detection device used in the NDIR mixed gas detection method described in any one of claims 2 to 7.

Images