Mixed precursor chemical gas identification method based on feature construction

Abstract

The invention provides a mixed precursor chemical gas identification method based on feature construction, and relates to the technical field of mixed gas identification, and the method comprises the steps: determining a division criterion of a transient response stage and a steady-state response stage through a working mode of fixed temperature modulation as an inverted step wave; and taking the average value in the steady-state response stage as the characteristic of the structure, establishing a sample library of the structure characteristic and the mixed gas concentration, and fitting the relationship between the structure characteristic and the mixed gas concentration by using a regression model with a simple structure, so that the function of identifying the mixed precursor chemical gas by using the simple model is realized. The current situation that the model for identifying the mixed precursor chemical gas is more complex and the algorithm scale is more huge is changed, and the working efficiency of mixed gas detection is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of mixed gas identification, and in particular to a mixed precursor chemical gas identification method based on feature structure. Background Art

[0002] The raw materials of synthetic drugs are called precursor chemicals. For example, toluene, acetone, ether and 2-butanone are important raw materials for synthesizing heroin. Rapid and accurate detection of precursor chemical gases can effectively control the spread of drugs from the source, which has important social significance. Semiconductor gas sensors have the advantages of simple structure, high sensitivity and rapid response, and are the most widely used technology in the field of gas detection. Temperature modulation can obtain feature-rich timing signals by applying a periodically changing voltage waveform to the gas sensor, significantly improving the cross-sensitivity characteristics of the gas sensor. In the field of gas detection, constructing a data set based on temperature-modulated timing signals, constructing a precursor chemical gas identification model through machine learning and deep learning, and realizing gas concentration prediction are strategies widely adopted by laboratory researchers and engineering application personnel.

[0003] Although temperature modulation can effectively improve the selectivity of gas sensors, there is no relatively fixed working mode. Experienced operators usually adjust the experimental parameters in a laboratory environment, including heating waveform, cycle, high and low voltage, duty cycle and other parameters, and then construct a data set, which greatly increases the workload. In actual gas environments, multiple gases generally exist at the same time. Increasing the size of the data set and using intelligent recognition algorithms with stronger fitting capabilities can improve the recognition accuracy of mixed gases, but it will lead to overfitting. Dimensionality reduction algorithms such as feature extraction and feature selection can alleviate the overfitting phenomenon, but it will further increase the complexity of the precursor chemical gas recognition model.

[0004] Under the existing technology, when predicting gas concentration in a complex gas environment, the algorithm scale of the gas identification model is increased, and the requirement for the data set scale is also increased. The present invention proposes a simple and effective feature construction method to assist operators in designing a simple gas identification model to complete the quantitative identification of mixed precursor chemical gases. Summary of the invention

[0005] In view of the problem in the prior art that the concentration identification of mixed precursor chemical gases mainly relies on increasing the scale of algorithm models, the present invention provides a mixed precursor chemical gas identification method based on feature construction. The present invention defines the steady-state response stage of the inverted step wave, takes the average value of the time series response signal of the steady-state response stage as the identification space, and fits the relationship between the constructed features and the concentration of the mixed precursor chemical gases, thereby realizing the function of quantitatively identifying the mixed precursor chemical gases with a simple model.

[0006] A method for identifying mixed precursor chemical gases based on feature construction comprises the following steps:

[0007] Step 1: Determine the temperature modulation mode of the gas sensor and collect the corresponding timing response signals under different gas conditions;

[0008] The temperature modulation mode is specifically as follows: the cyclic heating waveform is an inverted step wave, and the period is based on the stable response at each stage of the step wave;

[0009] Step 2: construct features of the time series response signal to obtain constructed feature data;

[0010] The characteristic construction is specifically as follows: the response of each stage of the time series response signal is divided into a transient response stage and a steady-state response stage, the average value of the steady-state response stage is defined as a steady-state response value, and the steady-state response value is used as a characteristic of the corresponding stage;

[0011] Step 2.1: Calculate the absolute value of the first-order derivative of the timing response signal. The formula is as follows:

[0012]

[0013] Where f'(x) is the first-order derivative of the timing response signal, f(x) is the timing response signal, x is the sampling time, and h represents the sampling interval;

[0014] Step 2.2: When from a certain time x i When the average value of the absolute value of the derivative is less than the set threshold, the time x i The end of the stage is divided into the steady-state response stage of the corresponding stage, and the rest of the stage is divided into the transient response stage;

[0015] Step 2.3: Count the length of the steady-state response phase corresponding to each stage, select the minimum length as the unified division standard, update the division result of each stage, and calculate the corresponding steady-state response value as the characteristic of this stage;

[0016] Step 3: Repeat step 2 for all acquired timing response signals;

[0017] Step 4: Test the gas concentration of the existing precursor chemical samples to obtain actual mixed gas concentration test data;

[0018] Step 5: Establish a sample library based on the characteristic data in step 2 and the mixed gas concentration test data in step 4;

[0019] Step 6: Construct a mixed precursor chemical gas recognition model based on the regression model, use the sample data in the sample library to train and test the recognition model, and save the parameters of the recognition model; save the relationship between the characteristic data and the mixed gas concentration test data in the form of a function;

[0020] Step 6.1: Shuffle all the data in the sample library, sort them randomly, randomly select some samples as training data sets, and the remaining samples as test data sets;

[0021] Step 6.2: Standardize the characteristic data of the training data set and the mixed gas concentration test data, and apply the standardization rules to the test data set;

[0022] The standardization process is normalization process, and the normalization formula is as follows:

[0023]

[0024] Among them, P' is the normalized data, P is the original data, and P min is the minimum value of the input data, P max is the maximum value of the input data;

[0025] Step 6.3: Fit the data in the training data set through the regression model. The output of the regression model is denormalized according to the concentration test data standardization rule in step 6.2, and the fitting results of the training set and the test set are tested. When the average test error of the mixed gas concentration is less than 10ppm, the parameters of the recognition model are saved;

[0026] Step 6.4: Save the parameters of the identification model in step 6.3 in the form of a function, and obtain the functional relationship S as follows:

[0027] Concentration=S(Features,Parameters)

[0028] Where Concentration represents the predicted concentration of mixed precursor chemical gas, Features represents the constructed features, and Parameters represents the parameters of the concentration identification model;

[0029] Step 7: Obtain the time series response signal to be detected under temperature modulation of the gas sensor, perform feature construction on the time series response signal according to step 2, and obtain feature data;

[0030] Step 8: Use the characteristic data obtained in step 7 as input, calculate the concentration of the mixed gas through the function obtained in step 6.4 and output it to complete the identification of the mixed precursor chemical gas.

[0031] The beneficial effects of adopting the above technical solution are:

[0032] The present invention provides a mixed precursor chemical gas identification method based on feature construction. According to the combination of historical data, actual condition experiments and theoretical analysis, the feature construction can effectively reduce the algorithm scale of the mixed gas identification model. By determining the temperature modulation scheme with a fixed working mode, a feature closely related to the concentration can be constructed; since the average value is taken in the steady-state response stage, noise interference can be reduced; at the same time, the data dimension of the time series response signal is effectively reduced, and the constructed features are used as input to perform concentration prediction modeling based on a regression model with a simple structure, thereby completing the function of identifying mixed precursor chemical gases.

[0033] Compared with the existing methods, the mixed precursor chemical gas identification method based on feature construction of the present invention has the following advantages: through the feature construction technology, a mixed precursor chemical gas identification model is established. The simple model can be used to identify the concentration of the mixed gas, which changes the current situation that the algorithm scale of the mixed gas concentration identification model is becoming increasingly large. At the same time, it can reduce noise interference and adapt to more practical application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is an overall flow chart of identifying mixed precursor chemical gases in an embodiment of the present invention.

[0035] Figure 2 This is the temperature modulation mode of the gas sensor in the embodiment of the present invention.

[0036] Figure 3 It is a timing response signal of the mixed precursor chemical gas combination 1 to be tested in the embodiment of the present invention.

[0037] Figure 4 It is the timing response signal of the mixed precursor chemical gas combination 2 to be tested in the embodiment of the present invention. DETAILED DESCRIPTION

[0038] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0039] A mixed precursor chemical gas identification method based on feature construction, such as Figure 1 As shown, the following steps are included:

[0040] Step 1: Determine the temperature modulation mode of the gas sensor and collect the corresponding timing response signals under different gas conditions;

[0041] The temperature modulation mode is specifically as follows: the cyclic heating waveform is an inverted step wave, and the period is based on the stable response at each stage of the step wave;

[0042] The temperature modulation mode of the gas sensor in this embodiment is as follows: Figure 2 As shown;

[0043] Step 2: construct features of the time series response signal to obtain constructed feature data;

[0044] The characteristic construction is specifically as follows: the response of each stage of the time series response signal is divided into a transient response stage and a steady-state response stage, the average value of the steady-state response stage is defined as a steady-state response value, and the steady-state response value is used as a characteristic of the corresponding stage;

[0045] Step 2.1: Calculate the absolute value of the first-order derivative of the timing response signal. The formula is as follows:

[0046]

[0047] Where f'(x) is the first-order derivative of the timing response signal, f(x) is the timing response signal, x is the sampling time, and h represents the sampling interval;

[0048] Step 2.2: When from a certain time x i When the average value of the absolute value of the derivative is less than the set threshold, the time x i The end of the stage is divided into the steady-state response stage of the corresponding stage, and the rest of the stage is divided into the transient response stage;

[0049] Step 2.3: Count the length of the steady-state response phase corresponding to each stage, select the minimum length as the unified division standard, update the division result of each stage, and calculate the corresponding steady-state response value as the characteristic of this stage;

[0050] In this embodiment, after testing several common precursor chemical gases, the first 15 seconds of each stage is divided into a transient response stage, and the last 5 seconds is divided into a steady-state response stage, and the corresponding steady-state response value is calculated;

[0051] Step 3: Repeat step 2 for all acquired timing response signals;

[0052] Step 4: Test the gas concentration of the existing precursor chemical samples to obtain actual mixed gas concentration test data;

[0053] In this embodiment, two mixed gases are used respectively, and the mixed gas 1 is acetone-toluene. The concentration test data results are shown in Table 1:

[0054] Table 1 Concentration of mixed gas 1

[0055]

[0056]

[0057] Mixed gas 2 is 2-butanone-ether, and the concentration test data results are shown in Table 2:

[0058] Table 2 Concentration of mixed gas 2

[0059]

[0060] Step 5: Establish a sample library based on the characteristic data in step 2 and the mixed gas concentration test data in step 4;

[0061] Step 6: Construct a mixed precursor chemical gas recognition model based on the regression model, use the sample data in the sample library to train and test the recognition model, and save the parameters of the recognition model; save the relationship between the characteristic data and the mixed gas concentration test data in the form of a function;

[0062] Step 6.1: Shuffle all the data in the sample library, sort them randomly, randomly select some samples as training data sets, and the remaining samples as test data sets;

[0063] In this embodiment, the first three quarters of the samples are taken as training data, and the last quarter is taken as test data;

[0064] Step 6.2: Standardize the characteristic data of the training data set and the mixed gas concentration test data, and apply the standardization rules to the test data set;

[0065] The standardization process is normalization process, and the normalization formula is as follows:

[0066]

[0067] Among them, P' is the normalized data, P is the original data, and P min is the minimum value of the input data, P max is the maximum value of the input data;

[0068] Step 6.3: Fit the data in the training data set through the regression model. The output of the regression model is denormalized according to the concentration test data standardization rule in step 6.2, and the fitting results of the training set and the test set are tested. When the average test error of the mixed gas concentration is less than 10ppm, the parameters of the recognition model are saved;

[0069] Step 6.4: Save the parameters of the identification model in step 6.3 in the form of a function, and obtain the functional relationship S as follows:

[0070] Concentration=S(Features,Parameters)

[0071] Where Concentration represents the predicted concentration of mixed precursor chemical gas, Features represents the constructed features, and Parameters represents the parameters of the concentration identification model;

[0072] In this embodiment, for mixed gas 1 and mixed gas 2, the regression models use convolutional neural network and BP neural network respectively. The convolutional neural network model includes an input layer, an output layer, a convolutional layer and a fully connected layer; wherein the dimension of the input layer is the size of the input vector, 5×1, the dimension of the convolutional layer is 2×1×16, the number of nodes in the fully connected layer is 64, the number of nodes in the output layer is 2, and the activation functions of the convolutional layer and the fully connected layer are both ReLU functions;

[0073] The BP neural network model includes an input layer, an output layer and a hidden layer; the input layer has 5 nodes, the hidden layer has 2 nodes, the activation function is the hyperbolic tangent function, the output layer has 2 nodes, and the activation function is a linear function;

[0074] Step 7: Obtain the time series response signal to be detected under temperature modulation of the gas sensor, perform feature construction on the time series response signal according to step 2, and obtain feature data;

[0075] The timing response signal of the mixed gas 1 to be tested in this embodiment is as follows: Figure 3 As shown, the result of feature construction is [21.47, 20.04, 7.71, 3.53, 1.14];

[0076] The timing response signal of the mixed gas 2 to be tested in this embodiment is as follows: Figure 4 As shown, the result of feature construction is [18.31, 16.07, 10.11, 3.91, 1.19];

[0077] Step 8: Use the characteristic data obtained in step 7 as input, calculate the concentration of the mixed gas through the function obtained in step 6.4 and output it to complete the identification of the mixed precursor chemical gas.

[0078] The identification results and actual test results of the mixed precursor chemical gas in this embodiment are as follows:

[0079] Table 3 Concentration values ​​of mixed gas 1

[0080]

[0081] Table 4 Concentration values ​​of mixed gas 2

[0082]

[0083] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with the technical features with similar functions disclosed in the embodiments of the present disclosure (but not limited to) to form a technical solution.

Claims

1. A method for identifying mixed precursor chemicals gas based on feature construction, characterized in that: The following steps are involved: Step 1: Determine the temperature modulation mode of the gas sensor and collect the corresponding timing response signals under different gas conditions; Step 2: construct features of the time series response signal to obtain constructed feature data; Step 3: Repeat step 2 for all acquired timing response signals; Step 4: Test the gas concentration of the existing precursor chemical samples to obtain actual mixed gas concentration test data; Step 5: Establish a sample library based on the characteristic data in step 2 and the mixed gas concentration test data in step 4; Step 6: Construct a mixed precursor chemical gas recognition model based on the regression model, use the sample data in the sample library to train and test the recognition model, and save the parameters of the recognition model; save the relationship between the characteristic data and the mixed gas concentration test data in the form of a function; Step 7: Obtain the time series response signal to be detected under temperature modulation of the gas sensor, perform feature construction on the time series response signal according to step 2, and obtain feature data; Step 8: Use the characteristic data obtained in step 7 as input, calculate the concentration of the mixed gas through the function obtained in step 6.4 and output it to complete the identification of the mixed precursor chemical gas.

2. The method for identifying mixed precursor chemicals based on feature construction according to claim 1 is characterized in that: The temperature modulation mode described in step 1 is specifically as follows: the cyclic heating waveform is an inverted step wave, and the period is based on the criterion of a stable response in each stage of the step wave.

3. The method for identifying mixed precursor chemicals based on feature construction according to claim 1 is characterized in that: The feature construction described in step 2 is specifically as follows: the response of each stage of the timing response signal is divided into a transient response stage and a steady-state response stage, the average value of the steady-state response stage is defined as a steady-state response value, and the steady-state response value is used as the feature of the corresponding stage.

4. The method for identifying mixed precursor chemicals based on feature construction according to claim 1, characterized in that: The step 2 specifically includes the following steps: Step 2.1: Calculate the absolute value of the first-order derivative of the time series response signal; Step 2.2: When from a certain time x i When the average value of the absolute value of the derivative is less than the set threshold, the time x i The end of the stage is divided into the steady-state response stage of the corresponding stage, and the rest of the stage is divided into the transient response stage; Step 2.3: Count the length of the steady-state response phase corresponding to each stage, select the minimum length as the unified division standard, update the division result of each stage, and calculate the corresponding steady-state response value as the feature of this stage.

5. The method for identifying mixed precursor chemicals gas based on feature construction according to claim 4 is characterized in that: The absolute value of the first-order derivative of the timing response signal described in step 2.1 is as follows: Where f'(x) is the first-order derivative of the timing response signal, f(x) is the timing response signal, x is the sampling time, and h represents the sampling interval.

6. The method for identifying mixed precursor chemicals based on feature construction according to claim 1, characterized in that: The step 6 specifically comprises the following steps: Step 6.1: Shuffle all the data in the sample library, sort them randomly, randomly select some samples as training data sets, and the remaining samples as test data sets; Step 6.2: Standardize the characteristic data of the training data set and the mixed gas concentration test data, and apply the standardization rules to the test data set; Step 6.3: Fit the data in the training data set through the regression model. The output of the regression model is denormalized according to the concentration test data standardization rule in step 6.2, and the fitting results of the training set and the test set are tested. When the average test error of the mixed gas concentration is less than 10ppm, the parameters of the recognition model are saved; Step 6.4: Save the parameters of the model identified in step 6.3 in the form of a function.

7. The method for identifying mixed precursor chemicals based on feature construction according to claim 6 is characterized in that: The standardization process described in step 6.2 is normalization, and the normalization formula is as follows: Among them, P' is the normalized data, P is the original data, and P min is the minimum value of the input data, P max is the maximum value of the input data.

8. The method for identifying mixed precursor chemicals based on feature construction according to claim 6 is characterized in that: The function S described in step 6.4 is as follows: Concentration=S(Features,Parameters) Where Concentration represents the predicted concentration of mixed precursor chemical gas, Features represents the constructed features, and Parameters represents the parameters of the concentration identification model.

Images