Methane electronic nose system based on semiconductor material and application

Abstract

The application belongs to the technical field of sensor devices and intelligent manufacturing, and specifically discloses a methane electronic nose system based on semiconductor materials and application. The methane electronic nose system comprises a central control device; two or more sets of sensing element combinations connected with the central control device; the sensing element combination comprises a Pd-SnO2 sensing element, a Pd4Au1-SnO2 sensing element, a Pd2Au1-SnO2 sensing element and a Pd1Au1-SnO2 sensing element; a temperature control device is arranged on each set of the sensing element combination, so that each sensing element can work at a set temperature and obtain a first signal corresponding to a to-be-detected gas; and the central control device is used for obtaining the concentration of methane in the to-be-detected gas according to the first signal collected by each sensing element and a second model. The semiconductor material used in the application is simpler and easier to synthesize, thereby reducing the production cost.

Description

Technical Field

[0001] This application belongs to the field of sensor devices and intelligent manufacturing technology, and more specifically, relates to a methane electronic nose system and its application based on semiconductor materials. Background Technology

[0002] Methane is an important industrial energy source, widely used in metallurgy, petrochemicals, and residential heating. However, as a colorless, odorless, and highly flammable gas, methane has a lower explosive limit of 4.4-5% in ambient air, making it extremely prone to explosion and posing a significant safety hazard. In recent years, there have been frequent household gas explosions, highlighting the increasing importance of monitoring methane leaks. Therefore, developing high-performance CH4 sensors for real-time, rapid detection is crucial. Current methane detection technologies mainly include catalytic combustion methane sensors, infrared methane sensors, and semiconductor methane sensors. Catalytic combustion methane sensors offer good selectivity for methane, but their catalysts are susceptible to poisoning and have a short lifespan. Infrared methane sensors offer high detection accuracy, but their high equipment cost limits widespread application. Semiconductor methane sensors are inexpensive, but individual semiconductor methane sensors are susceptible to interference from common indoor gases such as acetic acid, ethanol, and formaldehyde, leading to false positives and false negatives. Furthermore, electronic nose systems utilizing semiconductor sensor arrays and intelligent recognition algorithms can effectively improve the selectivity and accuracy of target gas detection in complex atmospheric environments.

[0003] Currently, electronic nose systems mainly consist of sensor arrays with cross-sensitivity, signal acquisition modules, and simple pattern recognition algorithms, making them difficult to effectively adapt to atmospheres with significant differences. Therefore, for methane gas detection in the presence of interfering gases such as acetic acid, ethanol, and formaldehyde, it is necessary to rationally combine multiple sensor arrays with cross-sensitivity, signal acquisition and processing units, complex atmosphere training sets, and intelligent recognition algorithms to achieve accurate real-time online detection.

[0004] Patent CN119064426.A discloses "a V2C / V2O5 / NiO composite gas-sensitive material and its preparation method." This method involves partially oxidizing V2C MXene to derive V2O5 nanobulbs onto the layered structure of V2C MXene, obtaining uniformly distributed "instant noodle" structured nano-V2O5. NiO nanoparticles are then obtained through heat treatment of Ni-MOF, synthesizing the V2C / V2O5 / NiO composite gas-sensitive structure. The methane sensor prepared from this invention using the V2C / V2O5 / NiO composite film exhibits advantages such as good reversibility, fast response / recovery speed, and good selectivity. It can detect methane gas concentrations in the range of 500 ppm to 4000 ppm, effectively detecting even low concentrations of methane gas. Furthermore, it operates at low detection temperatures, effectively overcoming the shortcomings of traditional methane gas sensors, such as slow response speed, poor sensitivity at low concentrations, and poor durability under high-temperature operating conditions. However, the material preparation method is relatively complex and costly, making stable mass production difficult. Summary of the Invention

[0005] In view of the shortcomings of the prior art, the purpose of this application is to select a lower-cost semiconductor material to form the sensing element combination, thereby reducing the cost of the electronic nose system and further improving the detection sensitivity of methane.

[0006] To achieve the above objectives, this application provides a methane electronic nose system based on semiconductor materials, including a central control device; and two or more sets of sensing element combinations respectively connected to the central control device; the sensing element combinations include Pd-SnO2 sensing elements, Pd4Au1-SnO2 sensing elements, Pd2Au1-SnO2 sensing elements and Pd1Au1-SnO2 sensing elements. Each set of the aforementioned sensing element assembly is equipped with a temperature control device, enabling each sensing element to operate at a set temperature and obtain a second signal corresponding to the gas to be measured; the second signal is eight or more of the following parameters: baseline resistance, minimum response, response time, recovery time, area integral of the response curve, area integral of the recovery curve, maximum derivative of the response curve, minimum derivative of the response curve, peak value of the spectrum, center frequency, bandwidth, spectral entropy, and spectral kurtosis. The central control device is used to obtain the concentration of methane in the gas to be tested based on the second signal and the second model collected by each sensing element.

[0007] Preferably, the set temperature of the temperature control device includes a first temperature and a second temperature, wherein the first temperature is between 185°C and 195°C, and the second temperature is between 225°C and 235°C.

[0008] Preferably, the second signal is obtained by feature extraction and normalization of the first signal, wherein the first signal is the response curve of the sensing element to the resistance of the gas to be measured at the set temperature.

[0009] As a further preferred embodiment, the first signal is obtained by cleaning and smoothing the original data.

[0010] As a further preferred embodiment, the second signal is the baseline resistance, minimum response, response time, recovery time, area integral of the response curve, area integral of the recovery curve, maximum derivative of the response curve, and minimum derivative of the response curve.

[0011] Preferably, the method for obtaining the second model is as follows: S21. Using at least six different methane concentrations as calibration gases as test gases, collect the second signal corresponding to the calibration gases; S22. After normalizing the second signal in step S21, train it in a support vector machine with the concentration of methane in the corresponding calibration gas to obtain the second model.

[0012] As a further preferred embodiment, in step S21, each of the different methane concentrations of the calibration gas corresponds to at least two different humidity levels.

[0013] Preferably, the central control device is also used to obtain the gas type of the gas to be tested based on the second signal collected by each sensing element and the first model.

[0014] As a further preferred embodiment, the gas types include methane, ethanol, formaldehyde, acetic acid, a mixture of methane and ethanol, a mixture of methane and formaldehyde, and a mixture of methane and acetic acid.

[0015] As a further preferred embodiment, the method for obtaining the first model includes the following steps: S11. Using at least six different concentrations of calibration gases of different gas types as test gases, and collecting the second signal corresponding to the calibration gases; S12. After the second signal in step S11 is reduced in dimension by linear discriminant analysis, it is trained in a support vector classifier according to the gas type of the calibration gas corresponding to the second signal to obtain the first model.

[0016] This application also provides the application of the above-mentioned methane electronic nose system in methane detection.

[0017] In summary, compared with the prior art, the technical solution conceived in this application, by systematically integrating the sensing element combination with the model constructed by the intelligent recognition algorithm, can achieve methane detection, and specifically has the following advantages: 1. The semiconductor materials used in the sensing elements are simpler and easier to synthesize, making methane sensors easier to mass-produce and reducing production costs; 2. Due to the adoption of array construction technology that is more sensitive to methane response, namely four sensing elements and two set temperatures; when collecting the first signal of the calibration gas, humidity, gas type and gas concentration are taken into account. It has been verified that the methane sensor system of this application can also identify methane under interfering gases such as acetic acid, ethanol and formaldehyde. 3. The detection limit for methane gas has been further reduced to 100 ppb. Attached Figure Description

[0018] Figure 1 This is a flowchart of the preparation of the semiconductor gas-sensitive material in Embodiment 1 of this application; Figure 2 This is a connection diagram of the methane electronic nose system provided in Embodiment 1 of this application when collecting raw data; Figure 3a These are the response curves of different semiconductor materials to different concentrations of methane in Example 1 of this application; Figure 3b These are the response curves of different semiconductor materials to a mixture of ethanol and methane of different concentrations in Example 1 of this application; Figure 4 This is a schematic diagram of the creation process of the second model and the first model provided in Embodiment 1 of this application; Figure 5 This is the decision boundary diagram of Embodiment 1 of this application; Figure 6 This is the confusion matrix diagram of Embodiment 1 of this application; Figure 7 This is a schematic diagram showing the actual concentration and predicted concentration of Example 1 of this application; Figure 8 This is the response value of the Pd2Au1-SnO2 sensing element in Example 1 of this application to different concentrations of methane. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0020] In the description of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0021] Furthermore, throughout this specification, references to "an embodiment"; "an embodiment," "an example," or similar language indicate that a particular feature, structure, or characteristic described in connection with that embodiment is included in at least one embodiment of this application. Therefore, the appearance of the phrase "in one embodiment;" throughout this specification, and similar language, may, but not necessarily, refer to the same embodiment.

[0022] This application provides a methane electronic nose system based on semiconductor materials for detecting methane concentration. The methane electronic nose system (i.e., a methane sensor system) includes a central control unit; and two or more sets of sensing elements connected to the central control unit; the sensing element sets include Pd-SnO2 sensing elements, Pd4Au1-SnO2 sensing elements, Pd2Au1-SnO2 sensing elements, and Pd1Au1-SnO2 sensing elements. Each set of the sensing element assembly is equipped with a temperature control device, enabling each sensing element to operate at a set temperature and obtain a first signal corresponding to the gas to be measured, the first signal being a resistance signal; in one embodiment, there are two sets of sensing element assemblies, and the set temperature corresponding to the temperature control device on them is any temperature between 185℃ and 195℃, and any temperature between 225℃ and 235℃, such as 190℃ and 230℃.

[0023] The central control device is used to obtain a second signal by performing feature extraction and normalization on the first signal collected by each sensing element, and then obtain the concentration of methane in the gas to be tested based on the second signal and the second model, and obtain the type of gas to be tested based on the second signal and the first model; the first signal is the response curve of the sensing element to the resistance of the gas to be tested at the set temperature, which is obtained by cleaning and smoothing the original data; The second signal is eight or more of the following: baseline resistance, minimum response, response time, recovery time, area integral of the response curve, area integral of the recovery curve, maximum derivative of the response curve, minimum derivative of the response curve, peak value of the spectrum, center frequency, bandwidth, spectral entropy, and spectral kurtosis; the gas type includes methane, ethanol, formaldehyde, acetic acid, a mixture of methane and ethanol, a mixture of methane and formaldehyde, and a mixture of methane and acetic acid.

[0024] The methods for obtaining the second model and the first model are as follows: S1. Collect the raw signals of the calibration gases, clean and smooth them to obtain the first signal, i.e., the response curve of the resistance of each calibration gas; then, obtain the second signal through feature extraction and normalization; the calibration gases include single-component or two-component gases; the single-component gases include methane, ethanol, formaldehyde, and acetic acid; the two-component gases include mixtures of methane and ethanol, mixtures of methane and formaldehyde, and mixtures of methane and acetic acid; each calibration gas has at least two humidity levels, for example, the humidity can be set to 0% and 50%; each component of each gas type has at least five or more concentration gradients; for single-component gases, taking methane as an example, six or more concentration gradients from 50 ppm to 2000 ppm can be set; for two-component gases, taking a mixture of methane and ethanol as an example, five or more concentration gradients can be set for each component (i.e., at least 5*5=25 gas concentrations). S2. Normalize the second signal of the calibration gas and establish a second model with the corresponding methane concentration using support vector classifier (SVC) and support vector regression (SVR); at the same time, reduce the dimensionality of the second signal of the calibration gas using linear discriminant analysis (LDA) and establish a first model with the corresponding gas type using support vector classifier (SVC).

[0025] This application also provides the application of the above-mentioned methane electronic nose system in actual methane detection; when used to detect the gas to be tested, the response curve of the resistance of each sensing element to the gas to be tested at the same set temperature as in step S2 above is first obtained at at least two different set temperatures, and then the second signal of the gas to be tested is obtained; then the second signal is input into the second model and the first model respectively, and the concentration and type of the gas to be tested can be obtained.

[0026] The following is an example.

[0027] Example 1 Step 1: Synthesizing Semiconductor Materials This embodiment uses a simple hydrothermal method to synthesize SnO2 nanoflowers (the synthesis process is referenced in "Enhanced methane sensing performance of SnO2 nanoflowers based sensors decorated with Au nanoparticles"). Taking Pd4Au1-SnO2 as an example, as follows... Figure 1As shown in step a, firstly, 1.33 g of SnCl4·5H2O was mixed with 30 mL of 1.26 M NaOH solution and stirred for 10 min. Then, 30 mL of anhydrous ethanol was added to the reaction system and stirred to obtain a white suspension. Next, the resulting white solution was transferred to a 100 mL Teflon-lined stainless steel autoclave and stored at 200 °C for 24 hours.

[0028] The synthesized product after heat treatment should be allowed to cool naturally to room temperature, such as Figure 1 As shown in step b. Then, after washing repeatedly with deionized water and ethanol, a white precipitate was obtained, which was dried at 60°C for 12 h to obtain the SnO2 precursor. Finally, calcination in air at 350°C for 2 h yielded the SnO2 nanoflower sample.

[0029] This embodiment uses an in-situ ascorbic acid reduction method to prepare PdAu-SnO2 nanoflowers with different Pd / Au atomic ratios, and uses them to construct differentiated sensor arrays. First, 100 mg of SnO2 nanoflower precursor powder was dispersed in 30 mL of deionized water and sonicated for 5 minutes. Then, 100 μl of palladium chloride (PdCl2, 50 mM) and 25 μl of tetrachlorogold(III) trihydrate (HAuCl4·xH2O, 50 mM) solution were added, and the mixture was stirred for 5 minutes. Next, 3 mL of ascorbic acid (C6H8O6, 100 mM) solution was added. After stirring at room temperature for 2.5 h, the mixture was washed several times with deionized water and ethanol, and the synthesized sample was dried at 60 °C for 12 h. The synthesized sample was then calcined in air at 350 °C with a heating rate of 5 °C / min for 2 h. Figure 1 Step c involves conversion and annealing. Then, the mixture is treated in H2 / Ar (1:10) at a heating rate of 250℃ and 5℃ / min for 2 hours. Figure 1 The reduction in step d yields a purer product, ultimately resulting in the product shown below. Figure 1 The PdAu-SnO2 nanoflowers shown in step e, with the small image in the upper left corner representing the Pd and Au nanocomposite.

[0030] The same method can be used to synthesize Pd-SnO2, Pd2Au1-SnO2 and Pd1Au1-SnO2. The difference is that the volumes of PdCl2 and HAuCl4·xH2O added in step 1d are adjusted according to Table 1. By adjusting the proportion of Pd, sensing elements with different sensitivities to methane can be obtained.

[0031] Table 1 Different parameters for the fabrication of sensing elements

[0032] After obtaining different semiconductor samples, TO-type sensors were fabricated from these samples. Different methane concentrations were detected within a humidity range of 0%–90% and an operating temperature range of 150℃–250℃, yielding differentiated response results. To improve the detection accuracy of the electronic nose system in practical applications, this embodiment selected Pd-SnO2, Pd4Au1-SnO2, Pd2Au1-SnO2, and Pd1Au1-SnO2 sensors with significantly different operating temperatures of 190℃ and 230℃. An array of eight sensors was constructed to form a methane electronic nose system. Figure 2 As shown, the leftmost vertical structures represent gas cylinders containing methane (CH4), ethanol (C2H5OH), formaldehyde (HCOH), acetic acid (CH3COOH), and air, respectively. These cylinders are connected to square flow meters via solenoid valves. A computer is connected to the control terminals of each square flow meter to set the flow rate of each gas. The computer is also connected to a multi-channel testing system. Below the multi-channel testing system is a knob for adjusting the voltage. By adjusting the voltage, the temperature control device located on the gas chamber (sensor array) can be adjusted. By controlling the flow meters and the multi-channel testing system, the computer can control the composition of the gas in the gas chamber and the operating temperature of the sensor array. The humidity generator can regulate the humidity of the gas. The entire device is used to collect raw data for testing and classifying methane, ethanol, formaldehyde, acetic acid, and their mixtures at different humidity levels.

[0033] Among them, arrays of different semiconductor gas-sensitive materials combined with different operating temperatures can amplify the differences in response to different gases, thus solving the problem of inaccurate identification caused by cross-interference. The signal characteristics of semiconductor gas sensors mainly include the response to the target gas, the integral area, and the response / recovery time. These are the main basis for gas identification and parameters reflecting the material properties.

[0034] Step 2: Collect raw data First, raw data from the sensors was collected. In this embodiment, raw data from 228 gas samples were collected. For ease of description, methane, ethanol, formaldehyde, and acetic acid are represented by A, B, C, and D, respectively, with the numbers following the letters indicating the concentration of the corresponding gas. At the same time, we set two humidity levels, 0% and 50%, which are represented by Roman numerals I and II, respectively. For example, a gas containing 50 ppm methane, 1 ppm ethanol, and 50% humidity can be represented as A50B1II. The original data included 48 gas samples with single components: methane gas at different humidity levels with concentrations of 50 ppm, 100 ppm, 200 ppm, 500 ppm, 1000 ppm, and 2000 ppm; ethanol, formaldehyde, and acetic acid gas at different humidity levels with concentrations of 1 ppm, 2 ppm, 5 ppm, 10 ppm, 20 ppm, and 50 ppm; and 180 gas samples with mixed components. For example, a mixture of methane and ethanol was sampled as follows: A50B1, A100 B1, A200 B1, A500 B1, A1000 B1, A2000 B1, A50B2, A100 B2, A200 B2, A500 B2, A1000 B2, A2000 B2, A50B1, A100B5, A200 B5, A500 B5, A1000... There are 60 types of gases in total: B5, A2000 B5, A50B10, A100 B10, A200 B10, A500 B10, A1000B10, A2000 B10, A50B20, A100 B20, A200 B20, A500 B20, A1000 B20, and A2000 B20 (Roman numerals are omitted because each sample has two humidity levels). The mixing method for methane and other gases is the same, except that ethanol gas is replaced with formaldehyde and acetic acid gas of the same concentration.

[0035] Samples were taken for each gas (sampling frequency f). s =2Hz, response curve time is approximately 900 seconds, recovery curve response time is approximately 1800 seconds, T total =2700 s), the concentration design in the experiment must maintain an arithmetic sequence. Due to factors such as long-term acquisition and hardware instability, data loss may occur, so linear interpolation is used to supplement the missing data.

[0036] Missing data points The data point preceding the missing data point The next data point after the missing data point k: Data points that need to be smoothed N: Average smoothing radius Sample mean Sample standard deviation The value after W: Projection matrix (8) w: Weight vector (9, 10, 11) b: Bias coefficient ||w||: Represents the L2 norm of the weight vector w. Error tolerance : The slack variable of the i-th sample express x The corresponding data, x Indicates the nth point, Indicates the inserted data point. This represents the smoothed data. This represents the total sampling time for a sample. n Represents the nth data point Indicates the baseline resistance. Indicates the first i The resistance value corresponding to each point This indicates searching for the minimum value. This indicates searching for the maximum value. s represents the sampling frequency. This indicates that the input value corresponding to the minimum value of the function is being searched. This represents the derivative value at the i-th data point. This represents the j-th feature of the i-th sample. To represent the transpose of a matrix, This represents the category label of the i-th sample. As shown in equation 1:

[0037] After obtaining the complete original data, an average smoothing method is used to select the values ​​before and after the original data points. The average of the data points was used to replace the original data points to obtain smoothed data, as shown in Figure 3. Figure 3a The graph shows the response curves of different semiconductor materials to methane concentrations of 50 ppm, 100 ppm, 200 ppm, 500 ppm, 1000 ppm, and 2000 ppm at 0% humidity (i.e., gas numbers A50I, A100I, A200I, A500I, A1000I, and A2000I, respectively). It can be seen from the graph that the higher the methane concentration, the smaller the decrease in resistance, and the steeper the decrease. Figure 3b The response curves of different semiconductor materials to mixed gases of methane and 1 ppm ethanol at 0% humidity (i.e., gas numbers A50B1I, A100B1I, A200B1I, A500B1I, A1000B1I, and A2000B1I) are shown, and their trends are similar to those of the mixed gases at different concentrations of methane and 1 ppm ethanol. Figure 3a Same. See Equation 2:

[0038] Step 3: Feature Extraction Next, features are extracted from the smoothed data. The extracted features and their corresponding calculation methods are shown in Table 2 below.

[0039] Table 2. Extracted features and their calculation methods

[0040] After obtaining a 228*(8*8) feature matrix (228 samples, 8 sensors, 8 features), due to the different scales among the features, a standardization method is used to unify the data scale, as shown in Equations 3 (calculating the feature mean), 4 (calculating the standard deviation), and 5 (standardization processing):

[0041]

[0042]

[0043] Step 4: Building the Model This concludes the data preprocessing section. The following section introduces the machine learning algorithm. This application utilizes machine learning algorithms to extract features of the sensor array under different atmospheres and uses these features to build classification and concentration regression models, such as... Figure 4 As shown, the model specifically includes a classification model (i.e., the first model) and a concentration regression model (i.e., the second model).

[0044] First, this application implements basic gas category classification, which includes the seven organic gases mentioned above (methane, ethanol, formaldehyde, acetic acid, methane + ethanol, methane + formaldehyde, methane + acetic acid). First, a linear discriminant analysis algorithm, namely LDA dimensionality reduction, is used to supervise the reduction of the 64 features corresponding to the 228 samples to 2 dimensions. LDA constructs two scatter matrices: (1) the inter-class matrix (S... B (2) Calculate the distance between the means of each class; (3) Calculate the distance between the means of each class; (4) Calculate the distance between the means of each class; (5) Calculate the distance between the means of each class; (6) Calculate the distance between the means of each class; (7) Calculate the distance between the means of each class; (8) Calculate the distance between the means of each class; (9) Calculate the distance between the means of each class; (10) Calculate W The algorithm returns the distance between the mean of each class and the data within that class, calculated using formulas shown in Equations 6 and 7. Finally, the optimal linear projection matrix is ​​obtained using Equation 8, LDA.

[0045]

[0046]

[0047] Where k is the number of categories. It is a category The mean vector, It is the mean vector of the entire data. It is a category The number of samples, It is a category A set of.

[0048] The dimensionality-reduced data was then fed into a Support Vector Classifier (SVC) for training to improve its ability to recognize seven gas classes, thus obtaining a classification model. After cross-validation, the model achieved an accuracy of 97.10% in the seven classifications, and the support vector decision boundary was plotted based on the support vectors. Figure 5 The x and y axes are LDA1 and LDA2, respectively, representing the two dimensions of linear discriminant analysis. Figure 5 The results show that, apart from two samples being misclassified, most of the points representing each gas sample were divided into different regions, with clear separation. Figure 6 The confusion matrix diagram shows that one methane sample was misidentified as a mixture of methane and ethanol, and another sample from a methane and ethanol mixture was misidentified as acetic acid (marked by the red circle). This is consistent with the results of the decision boundary diagram. This indicates that the gas sensor array based on PdAu-SnO2 material exhibits good selectivity for the seven selected gas types.

[0049] Next, we will discuss the effectiveness of concentration regression. To unify the feature scales of classification and regression, we directly input the 64 features before dimensionality reduction into the SVC and Support Vector Regression (SVR) models for training, and then establish classification and regression models. Both are based on Support Vector Machine (SVM) methods, and both seek an optimal hyperplane in the feature space. SVC separates the classes and maximizes the margin between data points on both sides of the hyperplane; SVR minimizes the error between the predicted and actual values, finding a hyperplane that makes most data points fall within a width of . The function in the regression band is the concentration regression model. Equation 9 is the objective function, Equation 10 is the SVC constraint, and Equation 11 are the SVR constraints:

[0050]

[0051]

[0052] in, Let be the normal vector of the hyperplane. For sample number, for Sample characteristics, For sample labels, For the hyperplane bias term, For sample size, The relaxation coefficient is... The hyperplane spacing is used.

[0053] After training, a high classification accuracy (92.22%) and regression coefficient were obtained. (0.9434), and a scatter plot of the actual concentration and the predicted concentration was drawn. Figure 7 As can be seen from the figure, the actual concentration of methane is well distributed on both sides of the regression curve formed near the predicted concentration. These results demonstrate the excellent selectivity of the PdAu-SnO2 material sensor array for the detected gas.

[0054] At 230°C and 50% relative humidity, the response to 500 ppm methane was 5.93, with a detection limit as low as 100 ppb. Figure 8 As shown, where x is the methane concentration, y is the response value, and R... 2 represents the regression coefficients of the fitted curve, and LOD represents the theoretical lower limit of detection for the fitted curve.

[0055] Comparative Example 1 The experiment was repeated using the same steps as in Example 1, with the difference that the sensing element combination contained only a single Pd-SnO2 material, and the temperature control device was set to a fixed temperature of 230°C, without employing a multi-temperature combination strategy. Therefore, the sensor corresponding to the collected raw data was only 1 / 8 of that in Example 1. Only a single humidity level (50%) was used during raw data acquisition, and mixed gas samples were not included (only pure methane gas was tested). The results showed that the false detection rate significantly increased and the classification accuracy dropped to 78.125% in the presence of interfering gases such as ethanol and formaldehyde. The R-value for methane concentration regression... 2 The coefficient was only 0.712, the detection limit was 500 ppb, and the sensitivity dropped significantly. Due to the lack of multi-temperature and multi-material synergy, the sensor did not respond to low concentrations of methane (<200 ppb).

[0056] Comparative Example 2 The experiment was repeated using the same steps as in Example 1, with the difference that in step one, only Pd-SnO2 and unmodified SnO2 were used instead of Au-modified elements in the sensor combination consisting of Pd-SnO2, Pd4Au1-SnO2, Pd2Au1-SnO2, and Pd1Au1-SnO2. Furthermore, the temperature control device was set to a single temperature of 230°C. Therefore, the collected raw data corresponded to only 1 / 4 of the sensors used in Example 1. The results showed that the classification accuracy for the mixed gas was only 80.6%, the confusion matrix indicated a misclassification rate of up to 40% for the methane and ethanol mixed gas samples, and a large deviation between the predicted and actual methane concentration values ​​(R0). 2=0.621), with a detection limit of 1 ppm, failing to reach a sensitivity of 100 ppb. The sensor exhibited poor stability during long-term testing, with a significant drift in the response curve and a recovery time increased by more than 30%. Therefore, the embodiments of this application enhance material selectivity by optimizing the palladium modification ratio and heating temperature.

[0057] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A methane electronic nose system based on semiconductor materials, characterized in that, Includes a central control device; and two or more sets of sensing element combinations respectively connected to the central control device; the sensing element combinations include Pd-SnO2 sensing elements, Pd4Au1-SnO2 sensing elements, Pd2Au1-SnO2 sensing elements and Pd1Au1-SnO2 sensing elements. Each set of the aforementioned sensing element assembly is equipped with a temperature control device, enabling each sensing element to operate at a set temperature and obtain a second signal corresponding to the gas to be measured; the second signal is eight or more of the following parameters: baseline resistance, minimum response, response time, recovery time, area integral of the response curve, area integral of the recovery curve, maximum derivative of the response curve, minimum derivative of the response curve, peak value of the spectrum, center frequency, bandwidth, spectral entropy, and spectral kurtosis. The central control device is used to obtain the concentration of methane in the gas to be measured based on the second signal collected by each sensing element and the second model, wherein, The method for obtaining the second model is as follows: S21. Using at least six different methane concentrations as calibration gases as test gases, collect the second signal corresponding to the calibration gases; S22. After normalizing the second signal in step S21, train it in a support vector machine with the concentration of methane in the corresponding calibration gas to obtain the second model.

2. The methane electronic nose system as described in claim 1, characterized in that, The temperature setting of the temperature control device includes a first temperature and a second temperature, wherein the first temperature is between 185°C and 195°C, and the second temperature is between 225°C and 235°C.

3. The methane electronic nose system as described in claim 1, characterized in that, The second signal is obtained by feature extraction and normalization of the first signal, which is the response curve of the sensing element to the resistance of the gas to be measured at the set temperature.

4. The methane electronic nose system as described in claim 3, characterized in that, The second signal is the baseline resistance, minimum response, response time, recovery time, area integral of the response curve, area integral of the recovery curve, maximum derivative of the response curve, and minimum derivative of the response curve.

5. The methane electronic nose system as described in claim 4, characterized in that, In step S21, each of the different methane concentrations of the calibration gas corresponds to at least two different humidity levels.

6. The methane electronic nose system as described in claim 1, characterized in that, The central control device is also used to obtain the gas type of the gas to be tested based on the second signal collected by each sensing element and the first model.

7. The methane electronic nose system as described in claim 6, characterized in that, The types of gases include methane, ethanol, formaldehyde, acetic acid, a mixture of methane and ethanol, a mixture of methane and formaldehyde, and a mixture of methane and acetic acid.

8. The methane electronic nose system as described in claim 7, characterized in that, The method for obtaining the first model includes the following steps: S11. Using at least six different concentrations of calibration gases of different gas types as test gases, and collecting the second signal corresponding to the calibration gases; S12. After the second signal in step S11 is reduced in dimension by linear discriminant analysis, it is trained in a support vector classifier according to the gas type of the calibration gas corresponding to the second signal to obtain the first model.

9. The application of the methane electronic nose system according to any one of claims 1-8 in methane detection.

Images