A detection and identification method for multiple types of industrial toxic gases

Abstract

The present application relates to the field of gas detection, specifically to a detection and identification method for various types of industrial toxic gases, specifically comprising the following steps: S1, selecting different concentrations and different types of industrial toxic gases as gas samples, placing the gas samples in a gas information extraction and collection bin to extract a gas feature vector; S2, processing the gas feature vector to generate a first standby feature vector; S3, training the extreme learning machine to obtain a target model; S4, constructing an industrial gas information knowledge base; The present application can handle the mixed situation of various types of industrial toxic gases, overcome the limitations of traditional single gas detection method, and the trained model parameters can be stored in the industrial gas information knowledge base, only forward calculation is needed during actual detection, the calculation burden is small and the speed is fast, and it is suitable for embedding into portable or fixed gas detection equipment.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of gas detection, in particular to a detection and identification method for multiple types of industrial toxic gases. BACKGROUND

[0002] In recent years, China's industrial production has developed rapidly. Based on the concept of sustainable development, the monitoring of industrial waste gas has become more important. Timely mastering the type and concentration information of waste gas discharged by factories is of great significance to the safety production of enterprises and the promotion of national environmental protection concept.

[0003] However, at present, in the monitoring of industrial waste gas, due to the high dimension and complexity of the collected gas data, the detection system cannot master the effective gas information in a short time, and thus it is difficult for the relevant departments to effectively master the waste gas emission of the relevant enterprises, greatly reducing the supervision intensity of industrial production. Although the current detection methods of industrial toxic gases in theory, such as LMBP, GABP, RBF neural network and SVM, cannot well complete the type judgment of industrial toxic gases, the classification and identification of gases need to be greatly improved. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a detection and identification method for multiple types of industrial toxic gases, which solves the technical problems in the background art.

[0005] To achieve the above purpose, the present application provides the following technical scheme:

[0006] A detection and identification method for multiple types of industrial toxic gases, specifically comprising the following steps:

[0007] S1, selecting industrial toxic gases of different concentrations and different types as gas samples, and placing the gas samples in a gas information extraction and collection bin to extract a gas feature vector;

[0008] S2, processing the gas feature vector to generate a first standby feature vector ;

[0009] S3, training the extreme learning machine to obtain a target model;

[0010] S4, constructing an industrial gas information knowledge base;

[0011] S5, passing the gas to be tested into the gas information extraction and collection bin to obtain a gas feature vector to be tested, finding the corresponding parameters of the target model in the industrial gas information knowledge base according to the gas feature vector to be tested, and then inputting the gas feature vector to be tested into the target model to output the type of industrial toxic gas of the gas to be tested.

[0012] Further, in step S1, the gas information extraction collection bin is internally provided with a metal oxide gas sensitive sensor array, a digital-analog circuit and a data processing device;

[0013] The metal oxide gas sensitive sensor array includes 10 metal oxide gas sensitive sensors, the detection gas sample cycle of the metal oxide gas sensitive sensor is 80 seconds, the detection interval is 1 second, and each metal oxide gas sensitive sensor obtains a resistance change signal in one detection cycle; the resistance change signal is a response curve composed of 80 data points arranged in time sequence;

[0014] The digital-analog circuit converts the resistance change signal into a digital signal;

[0015] The data processing device extracts P features from the digital signal of each metal oxide gas sensitive sensor, and the P features of the 10 metal oxide gas sensitive sensors together constitute a n-dimensional gas feature vector, n=10×P.

[0016] Further, the specific steps of detecting the resistance change signal are:

[0017] S11, detecting the temperature and humidity in the gas information extraction collection bin, and calculating the baseline resistance of the metal oxide gas sensitive sensor according to the temperature and humidity The calculation formula is:

[0018]

[0019] In the formula, denotes the baseline resistance of the metal oxide gas sensitive sensor under and ; R0 denotes the baseline resistance of the metal oxide gas sensitive sensor under the reference temperature and the reference humidity; denotes the baseline resistance of the metal oxide gas sensitive sensor under the reference temperature and the reference humidity; denotes the exponential function; denotes the temperature sensitivity coefficient of the metal oxide gas sensitive sensor; denotes the temperature and humidity coupling coefficient; and denote the reference temperature and the reference humidity, respectively; denotes the humidity dependence coefficient of the metal oxide gas sensitive sensor;

[0020] S12, calculating the sensitivity coefficient of the metal oxide gas sensitive sensor The calculation formula is:

[0021]

[0022] In the formula, denotes the previous factor; Activation energy of gas; Boltzmann constant;

[0023] S13, after the gas is introduced into the gas information extraction collection bin, the original resistance signal of the metal oxide gas sensitive sensor is detected, and the actual resistance signal is generated after compensation The calculation formula is:

[0024]

[0025] In the formula, The original resistance signal is represented;

[0026] S14, according to all actual resistance signals in a detection period The resistance change signal is constructed.

[0027] Further, in step S2, the following steps are specifically included:

[0028] S21, calculating the average feature vector of the kth industrial toxic gas according to the gas feature vector The calculation formula is:

[0029]

[0030] In the formula, The gas feature vector of the kth industrial toxic gas is represented; The total number of gas samples of the kth industrial toxic gas is represented; The kth industrial toxic gas is represented; , The total number of industrial toxic gas types is represented;

[0031] S22, calculating the overall average feature vector of all gas samples according to the average feature vector The calculation formula is:

[0032]

[0033] In the formula, The total number of gas samples of all types of industrial toxic gas is represented;

[0034] S23, calculating the inter-class scatter matrix And the overall average feature vector The intra-class scatter matrix The calculation formula is:

[0035]

[0036] ​​

[0037] S24, constructing a projection matrix according to the inter-class scatter matrix and the intra-class scatter matrix ;

[0038] S25, calculating a first standby eigenvector according to the projection matrix , and the calculation formula is:

[0039]

[0040] In the formula, the dimension of the first standby eigenvector is d, and d < n.

[0041] Further, in step S24, the following steps are specifically included:

[0042] S241, defining a discriminant criterion function , and the expression is:

[0043]

[0044] In the formula, represents a projection vector; represents a transposition operation of a matrix;

[0045] S242, deriving the discriminant criterion function and setting the derivative to 0 to obtain a generalized eigenvalue equation, and the expression is:

[0046]

[0047] In the formula, represents a generalized eigenvalue;

[0048] S243, judging whether the intra-class scatter matrix is invertible;

[0049] If yes, the generalized eigenvalue equation is converted into an ordinary eigenvalue equation, and the projection vector and the generalized eigenvalue are solved, and the expression of the ordinary eigenvalue equation is:

[0050]

[0051] If no, the generalized eigenvalue equation is converted into a regularization equation, and the projection vector and the generalized eigenvalue are solved, and the expression of the regularization equation is:

[0052] ​​​

[0053] In the formula, Represents a constant term; Represents the first identity matrix;

[0054] S244. Selecting generalized eigenvalues The largest d projection vectors And construct the projection matrix Its expression is:

[0055]

[0056] In the formula, This represents d projection vectors.

[0057] Furthermore, step S3 specifically includes the following steps:

[0058] S31. Randomly generate input weights and hidden layer bias Their expressions are as follows:

[0059]

[0060]

[0061] In the formula, This represents the weight vector of the i-th hidden layer; This represents the bias value of the i-th hidden layer; This indicates the total number of hidden layers; Represents the set of real numbers;

[0062] S32. Calculate the first feature vector to be used. The output vector after each hidden layer The calculation formula is as follows:

[0063]

[0064] In the formula, Indicates the activation function;

[0065] S33, Output vectors of N gas samples Stack them together to obtain the hidden layer output matrix Its expression is:

[0066]

[0067] S34. Using the type of industrial toxic gas for each gas sample as the sample label, assign a unique label code to each sample label using unique thermal coding.

[0068] S35, for N gas samples, define the output linear equation, the expression is:

[0069]

[0070] Wherein,

[0071]

[0072]

[0073]

[0074] In the formula, Indicates the implicit layer output matrix; Indicates the output weight matrix; Indicates the label encoding confidence matrix; Indicates the output vector of the Nth gas sample in the Lth implicit layer; Indicates the output weight of the Lth implicit layer to the mth label encoding; Indicates the confidence of the Nth gas sample belonging to the mth label encoding; Indicates the total number of label encodings;

[0075] S36, define the error function, the expression is:

[0076]

[0077] In the formula, Indicates all output vectors of the ith gas sample; Indicates all output weights of the jth implicit layer; Indicates the label encoding of the ith gas sample in the jth category of industrial toxic gas;

[0078] S37, calculate the output weight matrix using the regularized least squares solution formula , the regularized least squares solution formula is:

[0079]

[0080] In the formula, Indicates the second unit matrix; Indicates the regularization coefficient.

[0081] Further, in step S4, the industrial gas information knowledge base includes the projection matrix , the input weight , the implicit layer bias , and the output weight matrix .

[0082] Further, in step S5, specifically comprising the following steps:

[0083] S51, calculating a second to-be-used feature vector according to the to-be-tested gas feature vector , and the calculation formula is:

[0084]

[0085] In the formula, indicates the to-be-tested gas feature vector;

[0086] S52, inputting the second to-be-used feature vector into the hidden layer of the target model to obtain a to-be-tested output vector , and the calculation formula is:

[0087]

[0088] S53, calling an output weight matrix , and calculating the original output of the target model according to the to-be-tested output vector , and the calculation formula is:

[0089]

[0090] S54, selecting the industrial toxic gas type corresponding to the sample label with the highest confidence in the original output as the output of the target model.

[0091] Compared with the prior art, the present application provides a detection and identification method for multiple types of industrial toxic gases, which has the following beneficial effects:

[0092] 1. The present application can handle mixed situations of multiple types of industrial toxic gases, overcome the limitations of traditional single gas detection methods, and the trained model parameters can be stored in an industrial gas information knowledge base. In actual detection, only forward calculation is required, the calculation burden is small and the speed is fast, and it is suitable for embedding into portable or fixed gas detection equipment.

[0093] 2. In the present application, when extracting the first to-be-used feature vector, the concentration difference is ignored by reducing the intra-class scatter, and the difference between gas types is amplified by increasing the inter-class scatter, so that when the projection matrix projects the gas feature vector into a low-dimensional space, even if the concentration difference of the same type of gas is large, it will also be projected into the area where the corresponding type is located. The recognition accuracy of industrial toxic gases can be significantly improved.

[0094] ​3、The application considers that the resistance signal detected by the sensor is affected by temperature and humidity, and the resistance change signal obtained by calculation has higher precision compared with other methods that only consider the single influence of temperature and humidity, and can be used as the basis for improving the precision of identifying industrial toxic gases. BRIEF DESCRIPTION OF DRAWINGS

[0095] The drawings described herein are used to provide further understanding of the present application, and constitute a part of the present application. The illustrative embodiments of the present application and their description serve to explain the present application. They do not constitute an improper limitation on the present application. In the drawings:

[0096] Figure 1 A schematic diagram of a detection and identification method for various types of industrial toxic gases according to the present application;

[0097] Figure 2 A flowchart of detecting the resistance change signal according to the present application;

[0098] Figure 3 A flowchart of step S2 according to the present application;

[0099] Figure 4 A flowchart of step S24 according to the present application;

[0100] Figure 5 A flowchart of step S3 according to the present application. DETAILED DESCRIPTION

[0101] In order to make the above-mentioned objects, features and advantages of the present application more apparent, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. The realization process of how to apply technical means to solve technical problems and achieve technical effects of the present application can be fully understood and implemented.

[0102] Those skilled in the art can understand that all or part of the steps in the following embodiments can be completed by programs instructing related hardware, therefore, the present application can adopt a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt a computer program product in the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program codes.

[0103] In recent years, China's industrial production has developed rapidly, based on the concept of sustainable development, the monitoring of industrial waste gas has become more important. Timely grasp the type and concentration information of waste gas discharged by the factory is of great significance to the safety production of enterprises and the promotion of national environmental protection concept, at present, the method for detecting a kind of gas is very common, but the recognition effect of these methods for multiple gases is poor, therefore, please refer to Figures 1-5 The present application provides a kind of detection and identification method for multiple types of industrial toxic gas, specifically comprising the following steps:

[0104] S1, different concentrations of different types of industrial toxic gas are selected as gas sample, and the gas sample is placed in gas information extraction collection bin to extract gas feature vector;Specifically, in step S1, the interior of gas information extraction collection bin is provided with metal oxide gas sensitive sensor array, digital-analog circuit and data processing device;

[0105] Metal oxide gas sensitive sensor array includes 10 metal oxide gas sensitive sensors, the detection cycle of metal oxide gas sensitive sensor to gas sample is 80 seconds, and the detection interval is 1 second, and each metal oxide gas sensitive sensor obtains resistance change signal in a detection cycle;Resistance change signal is a response curve composed of 80 data points arranged in time sequence;

[0106] Digital-analog circuit converts resistance change signal into digital signal;Specifically, the AD conversion chip with model AD7705 is used in digital-analog circuit;

[0107] Data processing device extracts P features in the digital signal of each metal oxide gas sensitive sensor, and the P features of 10 metal oxide gas sensitive sensors together constitute an n-dimensional gas feature vector, n=10×P;Specifically, the features include steady state value (i.e. the final stable value of response curve), integral value (i.e. the area under the whole response curve), maximum value (i.e. the peak value of response curve), rising / falling slope (the slope of certain stage of curve) and the like.

[0108] When detecting the resistance change signal of each industrial toxic gas, it will be affected by temperature and humidity, which includes not only the individual influence of temperature and humidity on resistance change signal, but also the coupling influence of temperature and humidity on resistance change signal (temperature and humidity change will cause the shift of resistance change signal), therefore, in step S1, it specifically includes the following steps:

[0109] S11, detect the temperature and humidity in gas information extraction collection bin, and calculate the baseline resistance of metal oxide gas sensitive sensor according to the temperature and humidity , the calculation formula is as follows:

[0110]

[0111] In the formula, denotes the baseline resistance of the metal oxide gas sensor under and ; denotes the baseline resistance of the metal oxide gas sensor under a reference temperature and a reference humidity; denotes an exponential function; denotes the temperature sensitivity coefficient of the metal oxide gas sensor; denotes the temperature and humidity coupling coefficient; and respectively denote the reference temperature and the reference humidity; denotes the humidity dependence coefficient of the metal oxide gas sensor; specifically, and can be obtained through technical parameters provided by the manufacturer of the metal oxide gas sensor; and are 24 degrees Celsius and 50% respectively; The baseline resistance of the sensor can be measured at different temperature and humidity grid points, and then fitted using a nonlinear regression algorithm; dimensionless; and have the same dimension as ;

[0112] S12, calculating the sensitivity coefficient of the metal oxide gas sensor , the calculation formula of which is:

[0113]

[0114] In the formula, denotes the pre-factor; denotes the activation energy of the gas; denotes the Boltzmann constant; specifically, can be obtained through technical parameters provided by the manufacturer of the metal oxide gas sensor; can be found in the chemical manual or gas sensor research literature;

[0115] S13, after the gas is introduced into the gas information extraction collection bin, detecting the original resistance signal of the metal oxide gas sensor and generating an actual resistance signal after compensation , the calculation formula of which is:

[0116]

[0117] In the formula, denotes the original resistance signal.

[0118] S14, according to all the actual resistance signals in one detection period constructing the resistance change signal.

[0119] The application considers that the resistance signal detected by the sensor is affected by temperature and humidity, and the resistance change signal obtained by calculation has higher accuracy compared with other methods that only consider the individual influence of temperature and humidity, which can be used as the basis for improving the accuracy of identifying industrial toxic gases.

[0120] S2, processing the gas feature vector to generate a first standby feature vector Specifically, in step S2, the following steps are included:

[0121] S21, calculating the average feature vector of the kth industrial toxic gas according to the gas feature vector The calculation formula is:

[0122]

[0123] In the formula, indicates the gas feature vector of the kth industrial toxic gas; indicates the total number of gas samples of the kth industrial toxic gas; indicates the kth industrial toxic gas; , indicates the total number of types of industrial toxic gases;

[0124] S22, calculating the overall average feature vector of all gas samples according to the average feature vector The calculation formula is:

[0125]

[0126] In the formula, indicates the total number of gas samples of all types of industrial toxic gases;

[0127] S23, calculating the inter-class scatter matrix and the intra-class scatter matrix according to the average feature vector and the overall average feature vector The calculation formulas are respectively:

[0128]

[0129]

[0130] S24, constructing the projection matrix according to the inter-class scatter matrix and the intra-class scatter matrix ​​ Specifically, in step S24, the following steps are specifically included:

[0131] S241, defining a discriminant function The expression is:

[0132]

[0133] In the formula, represents a projection vector; represents a transpose operation of a matrix;

[0134] S242, deriving the discriminant function Derive the derivative and set the derivative to 0 to obtain a generalized eigenvalue equation, the expression of which is:

[0135]

[0136] In the formula, represents a generalized eigenvalue;

[0137] S243, judging whether the within-class scatter matrix is invertible;

[0138] If yes, convert the generalized eigenvalue equation into a normal eigenvalue equation and solve the projection vector and the generalized eigenvalue The expression of the normal eigenvalue equation is:

[0139]

[0140] If no, convert the generalized eigenvalue equation into a regularization equation and solve the projection vector and the generalized eigenvalue The expression of the regularization equation is:

[0141]

[0142] In the formula, represents a constant term; represents a first identity matrix; specifically, is ; is an n-row n-column matrix;

[0143] S244, selecting the d projection vectors with the largest generalized eigenvalue and constructing a projection matrix The expression is:

[0144]

[0145] In the formula, denotes d projection vectors.

[0146] S25, calculating the first standby feature vector according to the projection matrix , and its calculation formula is:

[0147]

[0148] wherein the first standby feature vector has a dimension of d, and d < n.

[0149] In the present application, when the first standby feature vector is extracted, the concentration difference is ignored by reducing the intra-class scatter, and the difference between gas types is amplified by increasing the inter-class scatter, so that when the projection matrix projects the gas feature vector into a low-dimensional space, even if the concentration difference of the same type of gas is large, it will be projected into the area where the corresponding type is located, which can significantly improve the recognition accuracy of industrial toxic gases.

[0150] S3, training the extreme learning machine to obtain a target model; specifically, in step S3, the following steps are specifically included:

[0151] S31, randomly generating input weights and hidden layer biases , whose expressions are respectively:

[0152]

[0153]

[0154] wherein, denotes the weight vector of the i-th hidden layer; denotes the bias value of the i-th hidden layer; denotes the total number of hidden layers; denotes the real set; specifically, each weight vector in is randomly sampled from a uniform distribution on the interval each bias value in is randomly sampled from a uniform distribution on the interval It should be noted that and do not participate in the subsequent weight update process.

[0155] S32, calculating the first standby feature vector through the output vector after each hidden layer , and its calculation formula is:

[0156]

[0157] wherein, represents an activation function;

[0158] S33, output vectors of N gas samples are stacked together to obtain an implicit layer output matrix stacked together to obtain an implicit layer output matrix , the expression of which is:

[0159]

[0160] S34, using one-hot encoding to assign a unique label code to each sample label with the industrial toxic gas type of each gas sample as the sample label;

[0161] S35, for N gas samples, define an output linear equation, the expression of which is:

[0162]

[0163] wherein,

[0164]

[0165]

[0166]

[0167] , wherein, represents an implicit layer output matrix; represents an output weight matrix; represents a label coding confidence matrix; represents the output vector of the Nth gas sample at the Lth implicit layer; represents the output weight of the Lth implicit layer to the mth label code; represents the confidence of the Nth gas sample belonging to the mth label code; represents the total number of label codes;

[0168] S36, define an error function, the expression of which is:

[0169]

[0170] , wherein, represents all output vectors of the ith gas sample; represents all output weights of the jth implicit layer; represents the label code of the ith gas sample in the jth category of industrial toxic gas;

[0171] S37, calculate the output weight matrix using the regularized least squares solution formula , the regularized least squares solution formula is:

[0172]

[0173] wherein, denotes a second unit matrix; denotes a regularization coefficient; specifically, is an L row and L column matrix; .

[0174] S4, constructing an industrial gas information knowledge base; specifically, in step S4, the industrial gas information knowledge base includes a projection matrix , an input weight , a hidden layer bias , and an output weight matrix .

[0175] S5, passing the to-be-tested gas into a gas information extraction and collection bin to obtain a to-be-tested gas feature vector, searching for corresponding parameters of a target model in the industrial gas information knowledge base according to the to-be-tested gas feature vector, and then inputting the to-be-tested gas feature vector into the target model to output an industrial toxic gas type of the to-be-tested gas; specifically, in step S5, the following steps are specifically included:

[0176] S51, calculating a second to-be-used feature vector according to the to-be-tested gas feature vector, and the calculation formula is:

[0177]

[0178] wherein, denotes a to-be-tested gas feature vector; specifically, the to-be-tested gas feature vector is obtained after processing by a metal oxide gas sensitive sensor array, a digital-analog circuit, and a data processing device; the calculation formula corresponding to step S25;

[0179] S52, inputting the second to-be-used feature vector into a hidden layer of the target model to obtain a to-be-tested output vector , and the calculation formula is:

[0180]

[0181] Specifically, the calculation formula corresponding to step S32;

[0182] S53, calling the output weight matrix , and calculating an original output of the target model according to the to-be-tested output vector , and the calculation formula is:

[0183]

[0184] ​​S54, selecting the original output corresponding to the highest confidence sample label as the output of the target model; specifically, the calculation formula corresponding to step S35.

[0185] The application can handle various types of industrial toxic gas mixture situations, overcome the limitations of traditional single gas detection methods, and the trained model parameters can be stored in an industrial gas information knowledge base. Only forward calculation is required during actual detection, the calculation burden is small, and it is suitable for embedding into portable or fixed gas detection equipment.

[0186] The above embodiments have been described in detail, and specific examples have been used to describe the principles and embodiments of the application. The above examples are only used to help understand the method and core idea of the application; at the same time, for those skilled in the art, according to the idea of the application, the specific embodiments and application scope will be changed, and the above description should not be understood as a limitation of the application.

Claims

1. A method for detecting and identifying various types of industrial toxic gases, characterized in that, Specifically, the following steps are included: S1. Select industrial toxic gases of different concentrations and types as gas samples, and place the gas samples in the gas information extraction and collection chamber to extract gas feature vectors. S2. Process the gas feature vector to generate the first feature vector to be used. ; S3. Train the extreme learning machine to obtain the target model; Step S3 specifically includes the following steps: S31. Randomly generate input weights and hidden layer bias Their expressions are as follows: ； ； In the formula, This represents the weight vector of the i-th hidden layer; This represents the bias value of the i-th hidden layer; This indicates the total number of hidden layers; Represents the set of real numbers; S32. Calculate the first feature vector to be used. The output vector after each hidden layer The calculation formula is as follows: ； In the formula, Indicates the activation function; S33, Output vectors of N gas samples Stack them together to obtain the hidden layer output matrix Its expression is: ； S34. Using the type of industrial toxic gas for each gas sample as the sample label, assign a unique label code to each sample label using unique thermal coding. S35. For N gas samples, define the output linear equation, whose expression is: ； in, ； ； ； In the formula, This represents the hidden layer output matrix; This represents the output weight matrix; Represents the label encoding confidence matrix; This represents the output vector of the Nth gas sample in the Lth hidden layer; This represents the output weight of the Lth hidden layer for encoding the mth label; This represents the confidence level that the Nth gas sample belongs to the mth tag code; Indicates the total number of tag codes; S36. Define the error function, its expression is: ； In the formula, Represents all output vectors for the i-th gas sample; This represents all output weights of the j-th hidden layer; This represents the label code for the i-th gas sample in the j-th category of industrial toxic gases; S37. Calculate the output weight matrix using the regularized least squares solution formula. The formula for the regularized least squares solution is: ； In the formula, Represents the second identity matrix; Represents the regularization coefficient; S4. Construct an industrial gas information knowledge base; In step S4, the industrial gas information knowledge base includes a projection matrix. Input weights Hidden layer bias and output weight matrix ; S5. Pass the gas to be tested into the gas information extraction and acquisition chamber to obtain the feature vector of the gas to be tested. Based on the feature vector of the gas to be tested, find the corresponding parameters of the target model in the industrial gas information knowledge base. Then, input the feature vector of the gas to be tested into the target model and output the industrial toxic gas type of the gas to be tested.

2. The method for detecting and identifying various types of industrial toxic gases according to claim 1, characterized in that, In step S1, the gas information extraction and acquisition chamber is equipped with a metal oxide gas sensor array, digital-analog circuit and data processing device. The metal oxide gas sensor array includes 10 metal oxide gas sensors. The detection period of the metal oxide gas sensors is 80 seconds, and the detection interval is 1 second. Each metal oxide gas sensor obtains a resistance change signal in one detection period. The resistance change signal is a response curve composed of 80 data points arranged in chronological order. Digital-to-analog circuits convert resistance change signals into digital signals; The data processing device extracts P features from the digital signal of each metal oxide gas sensor. The P features of the 10 metal oxide gas sensors together form an n-dimensional gas feature vector, n=10×P.

3. The method for detecting and identifying various types of industrial toxic gases according to claim 2, characterized in that, The specific steps for detecting resistance change signals are as follows: S11. Detect the temperature and humidity inside the gas information extraction chamber, and calculate the baseline resistance of the metal oxide gas sensor based on these values. The calculation formula is as follows: ； In the formula, This indicates that the metal oxide gas sensor is in and The baseline resistance below; This represents the baseline resistance of the metal oxide gas sensor at the reference temperature and reference humidity. Represents an exponential function; This represents the temperature sensitivity coefficient of a metal oxide gas sensor. Indicates the temperature and humidity coupling coefficient; and These represent the reference temperature and reference humidity, respectively. This represents the humidity dependence coefficient of the metal oxide gas sensor. S12. Calculate the sensitivity coefficient of the metal oxide gas sensor. The calculation formula is as follows: ； In the formula, Indicates the pre-exponential factor; Indicates the activation energy of a gas; Represents the Boltzmann constant; S13. After the gas is introduced into the gas information extraction and acquisition chamber, the original resistance signal of the metal oxide gas sensor is detected and compensated to generate the actual resistance signal. The calculation formula is as follows: ； In the formula, Represents the original resistance signal; S14. Based on all actual resistance signals in a detection cycle Construct a resistance change signal.

4. The method for detecting and identifying various types of industrial toxic gases according to claim 1, characterized in that, Step S2 specifically includes the following steps: S21. Calculate the average characteristic vector of the kth type of industrial toxic gas based on the gas characteristic vector. The calculation formula is as follows: ； In the formula, The gas feature vector representing the k-th type of industrial toxic gas; This represents the total number of gas samples of the k-th type of industrial toxic gas; This represents the kth type of industrial toxic gas; , This indicates the total number of types of industrial toxic gases; S22. Based on the average eigenvector Calculate the overall average eigenvector of all gas samples. The calculation formula is as follows: ； In the formula, This represents the total number of gas samples of all types of industrial toxic gases. S23. Based on the average eigenvector and the overall average eigenvector Calculate the inter-class scatter matrix and intra-class scatter matrix The calculation formulas are as follows: ； ； S24. Based on the inter-class scatter matrix and intra-class scatter matrix Constructing the projection matrix ; S25. According to the projection matrix Calculate the first potential feature vector The calculation formula is as follows: ； In the formula, the first feature vector to be used The dimension is d, where d < n.

5. The method for detecting and identifying various types of industrial toxic gases according to claim 4, characterized in that, Step S24 specifically includes the following steps: S241. Define the discrimination criterion function. Its expression is: ； In the formula, Represents the projection vector; This represents the matrix transpose operation; S242, Regarding the discrimination criterion function Taking the derivative and setting it to zero, we obtain the generalized eigenvalue equation, whose expression is: ； In the formula, Represents generalized eigenvalues; S243, Determine the within-class scatter matrix Is it reversible? If so, then transform the generalized eigenvalue equation into a common eigenvalue equation and solve for the projection vector. and generalized eigenvalues The expression for the ordinary eigenvalue equation is: ； If not, then the generalized eigenvalue equation is transformed into a regularized equation, and the projection vector is solved. and generalized eigenvalues The regularization equation is expressed as follows: ； In the formula, Represents a constant term; Represents the first identity matrix; S244. Selecting generalized eigenvalues The largest d projection vectors And construct the projection matrix Its expression is: ； In the formula, This represents d projection vectors.

6. The method for detecting and identifying various types of industrial toxic gases according to claim 1, characterized in that, Step S5 specifically includes the following steps: S51. Calculate the second potential feature vector based on the feature vector of the gas to be measured. The calculation formula is as follows: ； In the formula, Represents the characteristic vector of the gas to be measured; S52, the second feature vector to be used Input the hidden layer of the target model to obtain the output vector to be tested. The calculation formula is as follows: ； S53, Call the output weight matrix And based on the output vector to be tested Calculate the raw output of the target model The calculation formula is as follows: ； S54, Select Original Output The type of industrial toxic gas corresponding to the sample label with the highest confidence level is used as the output of the target model.

Images