Oxide semiconductor gas sensor array, preparation method, electronic nose and application

Abstract

The present application belongs to the field of smart sensors and relates to an oxide semiconductor gas sensor array, a preparation method, an electronic nose, and an application. The preparation method includes: preparing a mixed solution comprising 2-methylimidazole, a metal compound, and deionized water; adding a noble metal source to the mixed solution so that the molar ratio of metal atoms to noble metal atoms is a preset ratio; inserting a ceramic substrate into the dispersed mixed solution and soaking and growing it at a preset ambient temperature, wherein the portion of the ceramic substrate where no material is grown is covered with an insulating medium; after the soaking is completed, the ceramic substrate is removed and the sediment on its surface is removed to obtain a precursor of a MOF nanosheet doped with a noble metal grown on the surface of its interdigitated test electrode; the precursor is calcined in air to convert the precursor into a metal oxide nanosheet modified with a noble metal. Based on the above-mentioned nanosheets, a sensor array and an electronic nose are constructed, and the present application can achieve highly selective identification and quantitative detection of formaldehyde and ethanol gases.

Description

Technical Field

[0001] The present application belongs to the field of smart sensor technology, and more specifically, relates to oxide semiconductor gas sensor arrays, preparation methods, electronic noses, and applications. Background Art

[0002] With the increasing demand for monitoring various harmful gases and controlling air pollutants, there is an urgent need to develop more efficient gas sensors for rapid and accurate gas detection. The electronic nose is an artificial olfactory system that simulates the biological olfactory system, uses a gas sensor array, and combines pattern recognition technology to identify gases or odors. It has a wide range of applications in public safety, the food industry, environmental monitoring, and medical testing. However, traditional semiconductor gas sensors have high operating temperatures and high power consumption, and traditional pattern recognition algorithms are insufficient in analyzing sensor signal characteristics, which restricts the accuracy of qualitative identification. Quantitative detection is still imperfect, making it impossible to accurately detect target gases in complex atmospheres. Summary of the Invention

[0003] In response to the shortcomings of the existing technology, the purpose of this application is to provide an oxide semiconductor gas sensor array, preparation method, electronic nose and application to solve the problem of how to accurately identify and quantitatively detect target gases in complex atmospheres.

[0004] To achieve the above objectives, the present application provides a method for preparing an oxide semiconductor gas sensor array, the method comprising the following steps:

[0005] S1 prepares a mixed solution, wherein the mixed solution comprises 2-methylimidazole, a metal compound and deionized water;

[0006] S2: adding a noble metal source to the mixed solution so that the molar ratio of metal atoms to noble metal atoms in the mixed solution is a preset ratio; and then performing ultrasonic treatment to obtain a dispersed mixed solution;

[0007] S3 inserting a ceramic substrate into the dispersed mixed solution and soaking and growing the substrate at a preset ambient temperature, wherein a portion of the ceramic substrate where no material is grown is covered with an insulating medium; after the soaking is completed, the ceramic substrate is removed and the sediment on its surface is removed to obtain a precursor having noble metal-doped metal-organic framework nanosheets grown on the surface of the interdigitated test electrode;

[0008] S4: placing the precursor in an air atmosphere, heating it to a preset temperature at a preset heating rate, and calcining it to convert the precursor into a noble metal-modified metal oxide nanosheet;

[0009] S5 repeats steps S1-S4, and only the amount of the noble metal source added is adjusted during each repetition, while other conditions remain unchanged, thereby preparing metal oxide nanosheet sensor arrays modified with different amounts of noble metals.

[0010] Furthermore, the mixed solution contains 2-methylimidazole with a concentration of 2 mol / L to 10 mol / L and a metal compound aqueous solution with a concentration of 2 mol / L to 10 mol / L.

[0011] Furthermore, in step S2, the molar ratio of metal atoms to noble metal atoms in the mixed solution is 0-3.5%.

[0012] Furthermore, the metal compound includes cobalt nitrate hexahydrate, nickel chloride hexahydrate, ferric chloride hexahydrate and vanadium chloride; preferably, the noble metal source includes potassium tetrachloropalladate, chloroauric acid hydrate, silver nitrate or ruthenium chloride.

[0013] Furthermore, in step S3, the ambient temperature is preset to be 25° C. to 30° C., and the immersion growth time is 3 hours to 5 hours.

[0014] Furthermore, in step S4, the preset heating rate is 5°C / min-10°C / min, the preset temperature is 350°C, and the calcination time is 2 hours-4 hours.

[0015] According to another aspect of the present invention, an oxide semiconductor gas sensor array is further disclosed. The oxide semiconductor gas sensor array is prepared by any of the above methods for preparing an oxide semiconductor gas sensor array.

[0016] According to another aspect of the present invention, an electronic nose is also disclosed, comprising a plurality of groups of oxide semiconductor gas sensor arrays as described above, wherein different amounts of a noble metal source are added during the preparation of each group of the oxide semiconductor gas sensor arrays, so that the molar ratio of metal atoms to noble metal atoms in the corresponding mixed liquid is different.

[0017] According to another aspect of the present invention, there is also disclosed an application of the aforementioned electronic nose in gas detection, wherein the application method comprises the following steps:

[0018] S1 obtains resistance signal data of each oxide semiconductor gas sensor array in a target gas of preset concentration under preset humidity and outputs a resistance curve;

[0019] S2 extracts the dynamic response characteristics of all resistance curves and constructs a multidimensional feature matrix based on the dynamic response characteristics;

[0020] S3 performs dimensionality reduction processing and concentration quantification on the multidimensional feature matrix by a linear discriminant analysis method;

[0021] S4 performs power function fitting on the dimension-reduced data in step S3 and the quantitative concentration to obtain a concentration detection model of the target gas by the gas sensor array;

[0022] S5 detects the gas concentration in the environment to be tested using the concentration detection model, and further detects the type and concentration of the gas to be tested.

[0023] Furthermore, in step S1, the preset humidity is 30%-70% relative humidity (RH); the preset concentration is 0.1ppm-20ppm; preferably, the preset humidity and the preset concentration respectively include multiple groups of different values; more preferably, the target gas includes at least two types.

[0024] The above technical solution conceived by this application has the following advantages compared with the existing technology:

[0025] 1. The method for preparing an oxide semiconductor gas sensor array provided in this application, by in situ growing sensitive materials on interdigitated electrodes, can still well maintain the microstructure of the nanomaterial after calcination in an air atmosphere, ensuring a large number of adsorption sites, and can also avoid the problem of poor consistency of device surface materials caused by traditional film-forming methods such as drop coating and spin coating; at the same time, by changing the content of precious metals to regulate the gas-sensitive activity of cobalt oxide and construct a sensor array, combined with the coordinated optimization of the feature extraction algorithm, the recognition accuracy of the target gas is improved.

[0026] 2. In the process of preparing the oxide semiconductor gas sensor array, the mixed solution is ultrasonically treated to efficiently disperse the mixed liquid material; the part of the ceramic substrate where no material is grown is covered with an insulating medium, that is, the part of the interdigitated electrode is exposed, and the mixed solution is inserted and grown at 25°C-30°C for 3-5 hours, and the resulting device has stronger stability; after cleaning the deposits on the surface of the device, a metal-organic framework nanosheet modified with precious metals grown on the surface of the device's interdigitated electrodes can be obtained.

[0027] 3. The electronic nose prepared in this application includes multiple groups of gas sensor arrays. The amount of precious metal doped during the preparation process of each group of gas sensor arrays is different, which makes the surface activity of the gas sensors different, which is conducive to constructing a sensor array that can respond to the unique gas sensitivity of gases through different sensors.

[0028] 4. The electronic nose disclosed in this application is composed of multiple groups of different sensor arrays. At a specific operating temperature, the sensor arrays collect resistance data at different humidity and different concentrations of target gases (formaldehyde, ethanol), extract kinetic response characteristics from them, construct a multidimensional feature matrix, and combine it with a machine learning algorithm to obtain a gas detection model. This gas detection model makes up for the shortcomings of poor selectivity and difficulty in quantification, and realizes accurate identification and quantitative detection of target gases over a wide humidity range. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is a flow chart of the method for preparing a gas sensor array provided in this application;

[0030] Figure 2 This is a schematic diagram of the classification of formaldehyde and ethanol concentration monitoring provided in Example 3 of the present application;

[0031] Figure 3 This is a function fitting diagram of formaldehyde concentration and one-dimensional signal projection provided in Example 3 of the present application;

[0032] Figure 4 This is a schematic diagram of the classification of formaldehyde and ethanol concentration monitoring provided in Example 6 of the present application;

[0033] Figure 5 This is a function fitting diagram of formaldehyde concentration and one-dimensional signal projection provided in Example 6 of the present application. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0035] This application provides a method for preparing an oxide semiconductor gas sensor array, such as Figure 1 As shown, the preparation method comprises the following steps:

[0036] S1: preparing a mixed solution comprising 2-methylimidazole, a metal compound (including but not limited to cobalt nitrate hexahydrate, nickel chloride hexahydrate, ferric chloride hexahydrate, and vanadium chloride), and deionized water;

[0037] S2. adding a certain amount of a noble metal source (including but not limited to potassium tetrachloropalladate, chloroauric acid hydrate, silver nitrate, and ruthenium chloride) to the mixed solution so that the molar ratio of noble metal atoms to metal atoms in the mixed solution reaches a preset ratio, and ultrasonically treating the mixed solution to which the noble metal source has been added to obtain a dispersed mixed solution;

[0038] S3: Covering the portion of the ceramic substrate where no material is grown, that is, exposing the interdigitated electrode portion, with an insulating medium, and then inserting the ceramic substrate into the dispersion mixture and soaking and growing it at a preset ambient temperature for a certain period of time, removing the soaked ceramic substrate and removing the sediment on its surface to obtain a precursor of a metal-organic framework (MOF) nanosheet modified with a noble metal grown on the surface of the interdigitated test electrode;

[0039] S4: placing the precursor in an air atmosphere, heating it to a preset calcination temperature at a certain heating rate, and calcining it for a period of time to convert the precursor into a noble metal-modified metal oxide nanosheet;

[0040] S5 repeats steps S1-S4. During the repetition, the amount of the noble metal source added in step S2 is adjusted to prepare sensor arrays of metal oxide nanosheets modified with different amounts of noble metals.

[0041] In a preferred embodiment, the mixed solution comprises 2-methylimidazole at a concentration of 2 mol / L to 10 mol / L and an aqueous solution of a metal compound at a concentration of 2 mol / L to 10 mol / L, such as 2 mol / L, 3 mol / L, 4 mol / L, 5 mol / L, 6 mol / L, 7 mol / L, 8 mol / L, 9 mol / L, and 10 mol / L 2-methylimidazole, and 2 mol / L, 3 mol / L, 4 mol / L, 5 mol / L, 6 mol / L, 7 mol / L, 8 mol / L, 9 mol / L, and 10 mol / L aqueous solution of a metal compound. The specific ratio will be arbitrarily selected and combined within the above range according to actual needs. However, if the concentration exceeds 2 mol / L to 10 mol / L, the sensitivity and accuracy of the sensor array of metal oxide nanosheets modified with different amounts of noble metals will be reduced.

[0042] In a preferred embodiment, in step S2, the molar ratio of noble metal atoms to metal atoms in the mixed solution is preferably 0%-3.5%, such as 0.5%, 1.5%, 2%, 3.5%, etc.

[0043] In a preferred embodiment, in step S2, the ultrasonic treatment time is 5-15 minutes, such as 5 minutes, 6 minutes, 7 minutes, 8 minutes, 9 minutes, 10 minutes, 11 minutes, 12 minutes, 13 minutes, 14 minutes, 15 minutes, etc.

[0044] In a preferred embodiment, in step S3, the preset ambient temperature is 25°C to 30°C, such as 25°C, 26°C, 27°C, 28°C, 29°C, 30°C, etc.; the immersion growth time is preferably 3 hours to 5 hours, such as 3 hours, 3.5 hours, 4 hours, 4.5 hours, 5 hours, etc.

[0045] In a preferred embodiment, in step S4, the high-temperature calcination temperature is preferably 350°C; the heating rate during high-temperature calcination is preferably 5°C / min-10°C / min, such as 5°C / min, 6°C / min, 7°C / min, 8°C / min, 9°C / min or 10°C / min; the high-temperature calcination time is preferably 2 hours to 4 hours, such as 2 hours, 2.5 hours, 3 hours, 3.5 hours, 4 hours, etc.

[0046] In all the aforementioned embodiments, when steps S1 to S4 are repeated to prepare metal oxide nanosheet sensor arrays modified with different amounts of noble metals, other conditions remain unchanged except for adjusting the amount of the noble metal source.

[0047] According to another aspect of the present application, an oxide semiconductor gas sensor array is further provided. The gas sensor array is prepared by any of the previous oxide semiconductor gas sensor preparation methods.

[0048] According to another aspect of the present application, an electronic nose is also provided, which includes multiple groups of gas sensor arrays as described above. During the preparation process of each group of gas sensor arrays, different amounts of precious metal sources are added to make the molar ratio of metal atoms and precious metal atoms in the mixed solution different, while other conditions remain unchanged.

[0049] According to another aspect of the present application, there is also provided an application of an electronic nose as described in any one of the preceding items in gas detection, wherein the gas detection method comprises the following steps:

[0050] S1 obtains resistance signal data of each oxide semiconductor gas sensor array in a target gas of a specific concentration under specific humidity conditions and outputs a resistance curve;

[0051] S2 extracts the dynamic response characteristics of all resistance curves and constructs a multidimensional feature matrix based on the dynamic response characteristics;

[0052] S3 uses linear discriminant analysis to reduce the dimensionality of the multidimensional feature matrix and quantify the concentration;

[0053] S4 performs power function fitting on the dimension-reduced data in step S3 and the quantitative concentration to obtain a concentration detection model of the target gas by the gas sensor array;

[0054] S5 uses the concentration detection model to detect the gas concentration in the test environment, and then detects the type and concentration of the gas to be tested.

[0055] In a preferred embodiment, the specific humidity is 30%-70% RH (Relative Humidity), such as 30%, 35%, 40%, 45%, 50%, 55%, 60%, 65%, 70%, etc.; the specific concentration is preferably 0.1ppm-20ppm, such as 0.1ppm, 0.2ppm, 0.4ppm, 0.7ppm, 0.9ppm, 1ppm, 2ppm, 3ppm, 4ppm, 5ppm, 6ppm, 7ppm, 8ppm, 9ppm, 10ppm, 11ppm, 12ppm, 13ppm, 14ppm, 15ppm, 16ppm, 17ppm, 18ppm, 19ppm, 20ppm, etc.; more preferably, the specific humidity and the specific concentration are divided into multiple groups of different values, that is, the same specific humidity corresponds to different specific concentrations, and the same specific concentration corresponds to different specific humidity; more preferably, the target gases in step S1 include at least two types, and in other embodiments, may also include three, four, five, etc.

[0056] In order to illustrate the present application, the aforementioned method provided in the present application is described in detail below in conjunction with examples, but the following examples should not be understood as limiting the scope of protection of the present application.

[0057] Example 1

[0058] like Figure 1 FIG. 1 is a flow chart of a method for preparing an oxide semiconductor gas sensor array provided in this embodiment. The method includes the following steps:

[0059] S1. Prepare a mixed solution comprising 2-methylimidazole, cobalt nitrate hexahydrate and deionized water; the mixed solution comprises 50 ml of deionized water, 50 mL of 2-methylimidazole with a molar concentration of 4 mol / L and 50 mL of cobalt nitrate hexahydrate with a molar concentration of 4 mol / L, that is, a mixed solution of 16.42 g of 2-methylimidazole and 72.76 g of cobalt nitrate hexahydrate and 50 mL of deionized water.

[0060] S2. According to the preset atomic molar ratio of palladium and cobalt in the mixed solution, a corresponding amount of potassium tetrachloropalladate is added to the mixed solution, and ultrasonic treatment is performed for 10 minutes to fully disperse the mixed solution; the preset atomic molar ratio of palladium and cobalt in the mixed solution is 0%-3.5%, and the atomic molar ratio of palladium and cobalt is preferably any value of 0%, 0.5%, 1.5%, 2.5% or 3.5%.

[0061] S3. Covering the portion of the ceramic substrate where no material is grown, that is, exposing the interdigitated electrode portion of the ceramic substrate, with an insulating dielectric, and then inserting the ceramic substrate into the dispersed mixed solution and soaking and growing it at an ambient temperature of 25° C. to 30° C. for 3 to 5 hours; after the soaking, removing the ceramic substrate and removing the sediment on its surface to obtain a precursor having blue Pd-modified Co-MOF nanosheets grown on the surface of the interdigitated test electrode;

[0062] S4. The precursor is placed in an air atmosphere and calcined at 350°C at a heating rate of 5°C / min-10°C / min for 2-4 hours to convert the precursor into PdO-modified Co₃O₄ nanosheets. The resulting Co₃O₄ gas sensor better maintains the nanomaterial's microstructure, resulting in higher recognition accuracy and device stability.

[0063] Example 2

[0064] Five gas sensor arrays were prepared using the method disclosed in Example 1. The specific preparation method is as follows:

[0065] Five mixed solutions were taken, each containing an equal amount of a mixed solution of 16.42 g (4 mol / L, 50 mL) of 2-methylimidazole, 72.76 g (4 mol / L, 50 mL) of cobalt nitrate hexahydrate, and 50 mL of deionized water. Corresponding amounts of potassium tetrachloropalladate were added to the five mixed solutions according to the atomic molar ratios of palladium to cobalt in the mixed solutions being 0%, 0.5%, 1.5%, 2.5%, and 3.5%, respectively. All five mixed solutions were then ultrasonically treated for 10 minutes to fully disperse the materials in each mixed solution, thereby obtaining five mixed solutions having atomic molar ratios of palladium to cobalt of 0%, 0.5%, 1.5%, 2.5%, and 3.5%, respectively.

[0066] A gas-sensitive material is grown in situ on a ceramic wafer device. The ceramic wafer device is 26 mm long and 2 mm wide. The front of the device has interdigitated fingers and test electrodes made of gold, and the back of the device has heating electrodes made of platinum. The portion of the ceramic substrate (i.e., the ceramic wafer device) where no material is grown is covered with tape, exposing the interdigitated electrode portion. Five identically treated ceramic wafer devices are then inserted into the five ultrasonically treated solutions described above and grown at room temperature of 25°C to 30°C for 3 to 5 hours. In this embodiment, a growth time of 4 hours is preferred for optimal growth. Too low a temperature or too short a growth time is not conducive to obtaining a precursor with optimal performance.

[0067] After the immersion was completed, all the devices were taken out and the ceramic wafer devices were rinsed with deionized water to remove surface deposits, obtaining five blue Pd-modified Co-MOF nanosheet precursors grown on the surface of the interdigitated electrode part of the ceramic wafer devices.

[0068] The obtained Pd-modified Co-MOF nanosheet precursor was placed in an air atmosphere and heated to 350°C at a heating rate of 5°C / min-10°C / min for 2-4 hours to convert the precursor into PdO-modified Co3O4 nanosheets, thereby obtaining five Co3O4 nanosheets with different properties.

[0069] In other embodiments, the content of potassium tetrachloropalladate can also be adjusted to make the molar ratio of palladium atoms to cobalt atoms in the prepared mixed solution different from the molar ratio of palladium atoms to cobalt atoms in the above five mixed solutions, thereby preparing other gas sensor arrays with different performances.

[0070] Example 3

[0071] The five Co3O4 nanosheets with different characteristics prepared in Example 2 are used to form an electronic nose for gas detection. Specifically, the five Co3O4 nanosheets with different characteristics form a gas sensor array. The electronic nose in this embodiment includes five groups of oxide semiconductor gas sensor arrays. The steps of using the electronic nose for detection include:

[0072] S1. Provide 25 groups of formaldehyde and 25 groups of ethanol as gases to be detected, respectively. The formaldehyde to be detected and the ethanol to be detected respectively contain 5 groups of gases with different humidity conditions, and the humidity of each group of gases is 30% RH, 40% RH, 50% RH, 60% RH and 70% RH. Each group of gases with the same humidity has 5 groups of different gas concentrations, and the gas concentrations are set to 0.1ppm, 1ppm, 5ppm, 10ppm and 20ppm respectively; specifically, the gas under 30% RH contains 5 groups, and the concentration of each group of gases is 0.1ppm, 1ppm, 5ppm, 10ppm and 20ppm respectively; the gas under 40% RH contains 5 groups , the concentration of each gas group is 0.1ppm, 1ppm, 5ppm, 10ppm and 20ppm respectively; the gas under 50% RH contains 5 groups, the concentration of each gas group is 0.1ppm, 1ppm, 5ppm, 10ppm and 20ppm respectively; the gas under 60% RH contains 5 groups, the concentration of each gas group is 0.1ppm, 1ppm, 5ppm, 10ppm and 20ppm respectively; the gas under 70% RH contains 5 groups, the concentration of each gas group is 0.1ppm, 1ppm, 5ppm, 10ppm and 20ppm respectively. The above humidity conditions and gas concentrations are cross-combined to form a total of 25 test environments;

[0073] The Co3O4 gas sensor arrays prepared above were placed in 25 formaldehyde environments and 25 ethanol environments, respectively. The resistance signal data of each Co3O4 gas sensor array in the 25 formaldehyde and ethanol environments were obtained through the acquisition circuit and the host computer system, and the corresponding resistance curves were output. The specific process of performing the acquisition through the acquisition circuit and the host computer system is well known to those skilled in the art and will not be repeated here.

[0074] S2. Using Python programming software to extract kinetic response characteristics from the 50 sets of resistance curves collected in step S1, each Co3O4 gas sensor uses three characteristics: sensitivity, integrated area, and response time, and constructs a multidimensional feature matrix based on these kinetic response characteristics;

[0075] S3. Perform dimensionality reduction and concentration quantification on the multidimensional feature matrix using linear discriminant analysis method;

[0076] S4, performing power function fitting on the dimensionality reduction data in step S3 and the quantitative concentration to obtain a concentration detection model of the target gas by the Co3O4 gas sensor array;

[0077] S5. Use the concentration detection model to detect the gas concentration in the test environment, and then provide the type and concentration of the gas to be tested.

[0078] The specific steps of constructing a multidimensional feature matrix and obtaining a concentration detection model in steps S2-S4 are:

[0079] (1) The total of 15 features of the five gas sensor arrays are spliced ​​in the horizontal direction to form a 1*15 row vector, and then the row vectors of the five concentrations of the two gas samples measured at five humidity levels are spliced ​​in the vertical direction to construct a multidimensional feature matrix of size 50*15.

[0080] (2) The multidimensional feature matrix is ​​reduced in dimension by linear discriminant analysis. The specific dimensionality reduction process is as follows: At this time, the five concentration values ​​of the two gases are used as the dimensionality reduction target, so the target classification number C = 10. After extracting the three dynamic characteristics of each gas sensor array, each gas detection sample has a 15-dimensional feature vector. Therefore, it is necessary to select a basis vector of the same size of 15 dimensions (K = 15) for projection. The basic form of the basis vector is:

[0081] W=[w1|w2|…|w15] (1)

[0082] Among them, W is the basis vector set, w1 is the first basis vector;

[0083] By projection:

[0084] y=W Tx (2)

[0085] Among them, x is the feature matrix set of all samples;

[0086] The result of projecting the sample point on the K-dimensional vector is expressed as:

[0087] [y1,y2,...,y15] (3)

[0088] The degree of hashing of the sample points in each class relative to the center point in the class is expressed as:

[0089]

[0090]

[0091] Among them, S w is the sum of the hashing degrees of all sample points in each class relative to the center point in the class, μ i is the mean size of the i-th category sample, and c is the total number of categories of the sample.

[0092] The degree of hashing of the sample mean point of each class relative to the center point of all samples is expressed as:

[0093]

[0094] Among them, S B is the sum of the hashing degrees of the mean points of various samples relative to the center points of all samples, N i is the number of samples in the i-th category, ω i represents the i-th class sample;

[0095] After projection, the mean value of the i-th sample point projected on a certain basis vector is:

[0096]

[0097] Comprehensive projection vector and And update:

[0098]

[0099] Where W is the basis vector matrix, is the sum (variance) of the hash matrices within each class after projection, It is the sum (mean) of the hash matrices of the projections of the centers of each class relative to the center of the entire sample. The central idea of ​​linear discriminant analysis is to maximize the distance between classes and minimize the distance within a class. The distance after the projection of the class mean is used as a way to measure the degree of separation between classes.

[0100] Therefore, in order to minimize the sum of the intra-class hash matrices and maximize the sum of the inter-class hash matrices, we define:

[0101]

[0102] Among them, J(w) is the ratio of the defined between-class hash matrix sum to the within-class hash matrix sum.

[0103] Since the numerator and denominator of the above formula are both hash matrices, the determinant is taken to convert the matrix into a real number. Since the value of the determinant is actually the product of the matrix eigenvalues, and an eigenvalue can represent the degree of divergence on the eigenvector, the determinant is used for calculation.

[0104] Taking the derivative of J(w), we get:

[0105] S B w i =λS w w i (11)

[0106] This formula is equivalent to:

[0107]

[0108] Among them, λ is the matrix Finally, it comes down to finding the eigenvalue of the matrix The eigenvalues ​​of the first two largest eigenvalues ​​are taken, and the eigenvectors w1 and w2 corresponding to the first two largest eigenvalues ​​are taken, and the projection formula y = W is used. T The x obtained is the projection direction 1 and the projection direction 2. There is a certain relationship between the projection direction 1 or the projection direction 2 and the concentration of the gas. These relationship forms the basis for fitting the dimensionality reduction signal and the concentration. The projection vector with the same trend of concentration change for each type of gas is extracted and defined as the one-dimensional output signal of the electronic nose. The one-dimensional output signal and the concentration are then fitted with a power function using Origin software. The fitted power function form is:

[0109] y=a*x b (13)

[0110] x is the gas concentration, y is the one-dimensional signal after dimensionality reduction, and a and b are function coefficients.

[0111] Thus, a concentration detection model that can be used for target gas concentration detection in actual environments is obtained. The concentration detection model can first identify the gas type through the linear discriminant analysis algorithm, that is, the one-dimensional data after dimensionality reduction is brought into the fitted power function, and the concentration of the corresponding gas can be detected.

[0112] In other embodiments, more types of gas sensor arrays may be used in combination with machine learning algorithms for gas detection, so as to obtain a more accurate gas detection model through a larger amount of sample data.

[0113] Figure 2 This is a schematic diagram of the formaldehyde and ethanol concentration monitoring classification provided in this embodiment. By constructing a differentiated sensor array, different concentrations of formaldehyde and ethanol are detected, and feature extraction is performed on the detected data. The extracted features are normalized and displayed in the form of a heat map. The constructed multidimensional feature matrix is ​​reduced to projection direction 1 and projection direction 2 by linear discriminant analysis. Figure 2 As shown in the figure, when the number of target categories is 2, linear discriminant analysis can only reduce the target matrix to one-dimensional space. The numbers in the figure represent the specific values ​​after reducing to one-dimensional space. When the number of target categories is greater than 2, it can be reduced to two-dimensional space. The numbers in the figure represent the specific values ​​of projection direction 1 and projection direction 2 after reducing to two-dimensional space.

[0114] Figure 3 This is a function fitting diagram of formaldehyde concentration and one-dimensional signal projection provided in this embodiment, in which the projection direction that has a certain relationship with the concentration is defined as a one-dimensional output signal, which can be used for quantitative analysis of gas concentration. Since there is a certain relationship between the projection direction 1 after dimensionality reduction and the gas concentration, the projection direction 1 is selected as the one-dimensional output signal in this embodiment, and R is obtained. 2 Around 0.9 (respectively 0.902, 0.928, 0.940, 0.924, 0.895), close to 1, the function fitting relationship with an error accuracy within 5% can be obtained by (true value - fitting value) / range, where R 2 Refers to the degree of fit of the regression line to the observed value. The gray triangles in the figure represent different depths of different fitting lines. 2 Therefore, some of the output images overlap with each other.

[0115] Example 4

[0116] This is a schematic diagram of a process for preparing an oxide semiconductor gas sensor array provided in this embodiment. The preparation method includes the following steps:

[0117] S1. Prepare a mixed solution comprising 2-methylimidazole, nickel chloride hexahydrate and deionized water; the mixed solution comprises 50 ml of deionized water, 50 mL of 2-methylimidazole with a molar concentration of 4 mol / L and 50 mL of nickel chloride hexahydrate with a molar concentration of 4 mol / L, i.e., a mixed solution of 16.42 g of 2-methylimidazole and 47.6 g of nickel chloride hexahydrate and 50 mL of deionized water.

[0118] S2. According to the preset atomic molar ratio of ruthenium and nickel in the mixed solution, a corresponding amount of ruthenium chloride is added to the mixed solution, and ultrasonic treatment is performed for 10 minutes to fully disperse the mixed solution; the preset atomic molar ratio of ruthenium and nickel in the mixed solution is 0%-3.5%, and the atomic molar ratio of ruthenium and nickel is preferably any value of 0%, 0.5%, 1.5%, 2.5% or 3.5%.

[0119] S3. Covering the portion of the ceramic substrate where no material is grown, that is, exposing the interdigitated electrode portion, with an insulating dielectric, and then inserting it into the dispersed mixed solution and soaking and growing it at an ambient temperature of 25° C. to 30° C. for 3 to 5 hours; after the soaking is completed, removing the ceramic substrate and removing the sediment on its surface to obtain a precursor having Ru-doped Ni-MOF nanosheets grown on the surface of the interdigitated test electrode;

[0120] S4. The precursor is placed in an air atmosphere and calcined at 350°C at a heating rate of 5°C / min-10°C / min for 2-4 hours to convert the precursor into Ru-doped NiO nanosheets. The resulting NiO gas sensor can better maintain the nanomaterial's microstructure, resulting in higher recognition accuracy and device stability.

[0121] Example 5

[0122] Five gas sensor arrays were prepared using the method disclosed in Example 4. The specific preparation method is as follows:

[0123] Take five equal parts of a mixed solution containing 16.42 g (4 mol / L, 50 mL) of 2-methylimidazole, 47.6 g (4 mol / L, 50 mL) of nickel chloride hexahydrate, and 50 mL of deionized water. According to the atomic molar ratio of ruthenium to nickel in the mixed solution being 0%, 0.5%, 1.5%, 2.5% and 3.5%, a corresponding amount of ruthenium chloride is added to the five mixed solutions, and all of them are ultrasonically treated for 10 minutes to fully disperse the materials in each mixed solution, thereby obtaining five mixed solutions with atomic molar ratios of ruthenium to nickel being 0%, 0.5%, 1.5%, 2.5% and 3.5%, respectively.

[0124] Gas-sensing materials were grown in situ on a ceramic wafer device. The device was 26 mm long and 2 mm wide. The front of the device contained gold interdigitated electrodes and a test electrode, while the back of the device contained platinum heating electrodes. The portion of the ceramic substrate where material was not being grown, exposing the interdigitated electrodes, was covered with tape. The substrate was then inserted into the five ultrasonically treated solutions and grown at room temperature (25°C to 30°C) for 3 to 5 hours. In this example, 4 hours was preferred for optimal growth. Too low a temperature or too short a growth time was not conducive to obtaining the optimal precursor. The devices were then removed from the immersion chamber and rinsed with deionized water to remove surface deposits. Five Ni-MOF nanosheets with varying Ru doping concentrations were grown on the interdigitated electrodes. The resulting Ru-doped Ni-MOF nanosheet precursors were then placed in air and calcined at a heating rate of 5°C / min-10°C / min to 350°C for 2-4 hours to convert the precursor into Ru-doped NiO nanosheets, yielding five NiO nanosheets with different properties. The content of ruthenium chloride can also be adjusted to prepare a mixed solution with a ruthenium and nickel atomic molar ratio different from the above five ratios, thereby preparing other gas sensor arrays with different performances.

[0125] In other preferred embodiments, the above preparation method is the same, the only difference being that the metal compound is any one of cobalt nitrate hexahydrate, nickel chloride hexahydrate, ferric chloride hexahydrate or vanadium chloride, and the precious metal source is any one of potassium tetrachloropalladate, chloroauric acid hydrate, silver nitrate and ruthenium chloride. In more preferred embodiments, other applicable metal compounds and precious metal sources other than the metal compounds and precious metal sources listed above may also be selected.

[0126] Example 6

[0127] Using the preparation method of Example 5, an electronic nose comprising an array of five gas sensors with different characteristics was prepared for gas detection. The detection steps included:

[0128] S1 follows the combination method in step S1 of Example 3, with humidity conditions of 30% RH, 40% RH, 50% RH, 60% RH and 70% RH, and gas concentrations of 0.1 ppm, 1 ppm, 5 ppm, 10 ppm and 20 ppm, to provide 25 different combinations of formaldehyde and ethanol at different humidity and concentrations. Each NiO gas sensor array is then placed in 25 groups of formaldehyde to be tested and 25 groups of ethanol to be tested, and then the resistance signal data of each NiO gas sensor array in the 25 environments is obtained through the acquisition circuit and the host computer system, and the corresponding resistance curve is output; each NiO gas sensor array is then placed in humidity conditions of 30% RH, 40% RH, 50% RH, 60% RH and 70% RH, and each humidity condition corresponds to 5 groups of ethanol with gas concentrations of 0.1 ppm, 1 ppm, 5 ppm, 10 ppm and 20 ppm, and the resistance signal data of each NiO gas sensor array in 25 formaldehyde environments and 25 ethanol environments is obtained through the acquisition circuit and the host computer system, and the corresponding resistance curve is output;

[0129] S2 uses Python programming software or other software that can achieve the same function to extract the kinetic response characteristics of all resistance curves collected in step S1. Each NiO gas sensor array uses three characteristics: sensitivity, integrated area, and response time, and constructs a multidimensional feature matrix based on these kinetic response characteristics;

[0130] S3 uses linear discriminant analysis to reduce the dimensionality of the multidimensional feature matrix and quantify the concentration;

[0131] S4 performs power function fitting on the dimension-reduced data in step S8 and the quantitative concentration to obtain a concentration detection model of the target gas by the NiO gas sensor array;

[0132] S5 uses the concentration detection model to detect the gas concentration in the test environment, and then gives the type and concentration of the gas to be tested.

[0133] The specific steps of constructing a multidimensional feature matrix and obtaining a concentration detection model in steps S2-S4 of this embodiment are the same as the method of constructing a multidimensional feature matrix and obtaining a concentration detection model in Example 3, and will not be repeated here.

[0134] Figure 4 This is a schematic diagram of the formaldehyde and ethanol concentration monitoring classification provided in this embodiment. By constructing a differentiated sensor array, different concentrations of formaldehyde and ethanol are detected, and feature extraction is performed on the detected data. The extracted features are normalized and displayed in the form of a heat map. The constructed multi-dimensional feature matrix is ​​reduced to projection direction 1 and projection direction 2 by the linear discriminant analysis method. Figure 4As shown in the figure, when the number of target categories is 2, linear discriminant analysis can only reduce the target matrix to one-dimensional space. The numbers after the two projection directions 1 in the figure represent the specific values ​​after reducing to one-dimensional space. When the number of target categories is greater than 2, it can be reduced to two-dimensional space. The numbers after projection direction 1 and projection direction 2 in the figure represent the specific values ​​of projection direction 1 and projection direction 2 after reducing to two-dimensional space.

[0135] Figure 5 This is a function fitting diagram of formaldehyde concentration and one-dimensional signal projection provided in this embodiment, in which the projection direction that has a certain relationship with the concentration is defined as a one-dimensional output signal, which can be used for quantitative analysis of gas concentration. Since there is a certain relationship between the projection direction 1 after dimensionality reduction and the gas concentration, the projection direction 1 is selected as the one-dimensional output signal in this embodiment, and R is obtained. 2 All of them are around 0.9 (respectively 0.90, 0.91, 0.90, 0.89, 0.90), all close to 1. The function fitting relationship with an error accuracy within 5% can be obtained by (true value - fitting value) / range, where R 2 Refers to the degree of fit of the regression line to the observed value. The different depths of gray marks in the figure represent different fitting lines due to R 2 Therefore, the labels in the output graphics partially overlap with each other.

[0136] This application prepares oxide semiconductor gas sensor arrays by regulating the gas-sensing activity of metal oxides by varying the noble metal content. The in-situ growth of sensitive materials on interdigitated electrodes effectively maintains the nanomaterial's microstructure, ensuring a large number of adsorption sites and device consistency during subsequent film formation. In terms of application, by constructing an electronic nose composed of differentiated sensors, feature sets are extracted from the sensor array's output response curves for different gases and then subjected to dimensionality reduction processing. This, combined with machine learning algorithms, improves gas identification accuracy and detection precision.

[0137] This application improves the accurate identification and concentration detection of gases by preparing sensor arrays, regulating signal feature engineering, and combining machine learning algorithms. It has the characteristics of high sensitivity, high specificity, and excellent long-term stability.

[0138] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A method for preparing an oxide semiconductor gas sensor array, characterized in that: The preparation method comprises the following steps: S1: preparing a mixed solution comprising 2-methylimidazole, a metal compound, and deionized water, wherein the metal compound is any one of cobalt nitrate hexahydrate, nickel chloride hexahydrate, ferric chloride hexahydrate, and vanadium chloride; S2: adding a noble metal source selected from potassium tetrachloropalladate, chloroauric acid hydrate, silver nitrate, and ruthenium chloride to the mixed solution so that the molar ratio of metal atoms to noble metal atoms in the mixed solution is 0-3.5:100; and then performing ultrasonic treatment to obtain a dispersed mixed solution; S3 inserting a ceramic substrate into the dispersed mixed solution and soaking and growing the substrate at a temperature of 25° C. to 30° C. for 3 to 5 hours, wherein a portion of the ceramic substrate where no material is grown is covered with an insulating medium; after the soaking is completed, the ceramic substrate is removed and the sediment on its surface is removed to obtain a precursor having noble metal-doped metal-organic framework nanosheets grown on the surface of the interdigitated test electrode; S4: placing the precursor in an air atmosphere, heating it to a preset temperature of 350° C. at a preset heating rate of 5° C. / min-10° C. / min, and calcining it for 2 hours-4 hours to convert the precursor into a noble metal-modified metal oxide nanosheet; S5 repeats steps S1-S4, and only the amount of the noble metal source added is adjusted during each repetition, while other conditions remain unchanged, thereby preparing metal oxide nanosheet sensor arrays modified with different amounts of noble metals.

2. The method for preparing an oxide semiconductor gas sensor array according to claim 1, wherein: The mixed solution contains 2-methylimidazole with a concentration of 2 mol / L to 10 mol / L and a metal compound aqueous solution with a concentration of 2 mol / L to 10 mol / L.

3. An oxide semiconductor gas sensor array, characterized in that: The oxide semiconductor gas sensor array is prepared by the method for preparing an oxide semiconductor gas sensor array according to any one of claims 1 to 2.

4. An electronic nose, characterized in that: The electronic nose comprises a plurality of oxide semiconductor gas sensor arrays as described in claim 3 , wherein different amounts of precious metal sources are added during the preparation of each oxide semiconductor gas sensor array, so that the molar ratio of metal atoms to precious metal atoms in the corresponding mixed solution is different.

5. An application of the electronic nose according to claim 4 in gas detection, characterized in that: The application method includes the following steps: S1 obtains resistance signal data of each oxide semiconductor gas sensor array in a target gas with a preset concentration of 0.1 ppm to 20 ppm at a relative humidity of 30% to 70%, and outputs a resistance curve; the relative humidity and the preset concentration each include multiple sets of different values, and the target gas includes at least two types; S2 extracts the dynamic response characteristics of all resistance curves and constructs a multidimensional feature matrix based on the dynamic response characteristics; S3 performs dimensionality reduction processing and concentration quantification on the multidimensional feature matrix by a linear discriminant analysis method; S4 performs power function fitting on the dimension-reduced data in step S3 and the quantitative concentration to obtain a concentration detection model of the target gas by the gas sensor array; S5 detects the gas concentration in the environment to be tested using the concentration detection model, and further detects the type and concentration of the gas to be tested.

Images