CN116304912A - Sensor gas concentration detection method based on deep learning transducer neural network - Google Patents

Abstract

The invention discloses a sensor gas concentration detection method based on a deep learning transformer neural network, which belongs to the technical field of sensor intelligent detection, including data set preparation and preprocessing; construction of a Transformer neural network model; training of a Transformer neural network model; The trained Transformer neural network model estimates the actual environmental parameter values and gas concentration values. The algorithm uses the model trained by the existing data set to realize the rapid detection of the gas concentration in the environment by the gas sensor under the condition of fast and few samples; it has high precision, strong versatility, good robustness, and strong real-time performance, etc. Advantages, it can overcome the problems existing in traditional methods and achieve better gas concentration detection results.

Description

A sensor gas concentration based on deep learning transformer neural network Detection method

technical field

The invention belongs to the technical field of sensor intelligent detection, and in particular relates to a sensor gas concentration detection method based on a deep learning transformer neural network.

Background technique

In the sealing scene where a certain amount of gas has been injected, the gas sensor is used to detect the gas concentration in the environment; the gas sensor basically has a preheating step, and the gas sensor of the gas sensor reacts fully with the gas to produce its own properties Changes, the detection circuit is designed through different gas sensor characteristics, and finally the physical or chemical signal is converted into an electrical signal; from the above reaction characteristics, it can be concluded that the measurement cycle of the gas sensor is slow, and there are great difficulties in ensuring the accuracy.

The traditional gas sensor measurement method is to use the lower computer to collect the amplitude of the electrical signal sent by the gas sensor circuit, and set the stable value or peak value in the amplitude time series as the current ambient gas concentration value; as shown in Figure 1, in the carbon nanometer The tube gas sensor measures the conductance amplitude change amplitude value caused by the SOF ₂ and SO ₂ F ₂ gas concentration. It can be seen that there are irregular peak changes in the entire reaction process. The experiment process lasted 85 hours. The traditional method of collecting time-series amplitude peaks can only reduce errors by extending the experiment period infinitely. This process has problems such as time-consuming and deviation of experimental results.

Based on the traditional measurement scheme, a method of measuring the peak value of the amplitude based on mathematical characteristics such as time-series derivatives is proposed. This method estimates the actual gas concentration value based on the analysis of the first-order derivative value of the amplitude-time series curve. Change the original amplitude peak estimation method, and use the measured speed value to further estimate the actual value. This method reduces the experimental period to a certain extent, but discards the integrity of the time series curve, and cannot accurately estimate the continuous change curve of a class of time series acceleration values. In addition, during the dynamic measurement process, the gas concentration is easily affected by the historical concentration state. When the test gas is injected in the state of existing gas in a closed state, the process derivative of the concentration measurement curve is affected by many aspects. It may appear that the same concentration has different process derivatives, or the process derivatives are similar but the actual concentration does not match. Therefore, this method is still There are certain problems and limitations.

Contents of the invention

In order to overcome the current problems in the field of gas sensor detection, such as long detection time period, great historical influence, complex data processing, inaccurate experimental results, and low detection accuracy, the present invention provides a sensor gas sensor based on deep learning transformer neural network. Concentration detection method, the algorithm uses the model trained by the existing data set to realize the rapid detection of the gas concentration in the environment by the gas sensor under the condition of fast and few samples.

The present invention realizes through following technical scheme:

A sensor gas concentration detection method based on a deep learning transformer neural network, specifically comprising the following steps:

Step 1: Data set preparation and preprocessing;

Collect the data of the gas sensor, and preprocess the data, the preprocessing includes cleaning, denoising and standardization, and obtain the gas concentration series data with time dimension after preprocessing;

Step 2: Construction of the Transformer neural network model;

Segment the data set, bring the training set into the embedding layer (embedding), adjust the hyperparameters of the encoder (Encoder) and decoder (Decoder) modules in the model, and refer to the mean square error (MSE) and mean absolute error (MAE) indicators. Optimization functions such as grid search evaluate the optimal combination of hyperparameters, and use the best MSE and MAE to characterize model performance;

Step 3: Training of the Transformer neural network model;

Step 4: Use the trained Transformer neural network model to estimate the actual environmental parameter values and gas concentration values.

Further, in step 1, the data includes sequence values of gas concentration and time values, and the sequence values including concentration and time in the sensor are converted into a format for the Transformer model, so as to perform model training tasks, specifically including the following:

A1. Discretize time series: discretize continuous time series data into data with a fixed time interval of 10 minutes;

A 2. Sequence standardization: perform mean normalization on discretized time series to make them have similar statistical characteristics;

A3. Construct the input sequence: convert the mean-normalized time series data into an input sequence, that is, input a fixed period of data as a sequence into the Transformer neural network model;

A4. Batching and padding: When the length of the input sequence is insufficient, a padding operation is performed to ensure the consistency of the length of the input sequence.

Further, in step 2, the Encoder-Decoder model and the embedding layer of the Transformer neural network are used to perform sequential processing on the data; wherein, the embedding layer is used to convert the data collected by the sensor into a vector form that the neural network can process; The Encoder module is used to convert the input sequence into a set of hidden representations; the Decoder module is used to generate the output of the current time step according to the hidden representation provided by the Encoder module and the output generated before.

Further, the embedding layer is composed of a position encoder and an input embedding, the position encoder is used to add position information to the input data at each time point, so that the model can learn the order of the time series; the input embedding is used to convert each time point The input data of is converted into a fixed-dimensional vector representation, which is convenient for subsequent attention mechanism, encoder and decoder to process.

Further, the Encoder module includes:

Multi-Head Attention, a multi-head attention mechanism, is used to weight and aggregate the input sequence so that the Encoder module can better utilize the information of the input sequence;

Feedforward Neural Network Position-wise Feed-Forward Network: It is used to weight the output of the above multi-head attention mechanism to generate a set of hidden representations.

Further, the Decoder module includes:

Self-attention mechanism Masked Multi-Head Attention, used to calculate the relationship between the output of the current time step and the output generated before, and interact the hidden representation provided by the Encoder module with the output of the current time step;

The multi-head attention mechanism Multi-Head Attention is used to weight the hidden representation provided by the Encoder module so that the Decoder module can better utilize the information of the input sequence;

The feedforward neural network Position-wise Feed-Forward Network is used to weight and aggregate the outputs of the above two attention mechanisms to generate the output of the current time step.

Further, a LayerNormalization module is provided between each component of the Encoder module and the Decoder module for better signal transmission and preventing gradient disappearance during model training.

Further, the second step of building a model specifically includes the following contents:

B1: data set segmentation;

Segment the sequence data obtained in step 1, divide the 24-hour data into time windows of fixed length, each time window length is 10mins to 30mins; and use the ratio of 70/15/15 to divide the data set; Use the first 70% of the time window as the training set, the middle 15% as the validation set, and the remaining 15% as the test set;

B2: Set the hyperparameter range;

First, determine the length of the input sequence. According to the sampling frequency of the data and the application scenario, select the data points from 0 to 24 hours as an input sequence length to comprehensively record the change value of the ambient gas concentration; then, determine the batch size and hidden layer Number, set the batch to 32, 64 or 128; determine the number of hidden layers as 5-6 layers; finally, determine the number of heads as 6-8;

B3: grid search;

Use the grid search method to search for the optimal hyperparameter combination within the hyperparameter range;

B4: random search;

Use the random search method to randomly search for the optimal hyperparameter combination within the hyperparameter range;

B5: Bayesian optimization;

Use the Bayesian optimization method to find the optimal hyperparameter combination within the hyperparameter range;

B6: Evaluate model performance;

The Transformer neural network model is trained separately using the optimal hyperparameter combinations obtained by the above methods, and the optimal hyperparameter combination is evaluated using the mean square error (MSE) and mean absolute error (MAE) on the test set, and the optimal hyperparameter combination is used. The MSE and MAE corresponding to the optimal parameter model represent the performance of the model.

Further, step three is specifically as follows:

Select the extracted parameter value and its corresponding time value as the analysis feature quantity of the transformer neural network, that is, Q _i =[t _i , v _i ] ^T , Q _i represents the data parameter sent by the lower computer at a certain moment, t _i , v _i is the concentration and time value of the corresponding gas respectively; the training set is imported into the neural network through embedding, and after a series of interspersed, normalized, and attention mechanism training processes, the parameter relationship between the curve feature changes is obtained, that is, the model Training is complete.

The present invention uses a sensor gas concentration detection method based on a deep learning transformer neural network, which has the following advantages compared to traditional methods:

1. Improve the precision and accuracy of gas concentration detection: the existing gas concentration detection methods rely on the processing of sensor measurement data and model fitting, and there are problems such as insufficient model complexity and insufficient model generalization ability; and transformer neural network Through the training of a large amount of data and the optimization of the model, the characteristics of the data can be better mined, and the precision and accuracy of the detection can be improved;

2. Can handle the concentration detection of multiple gases: the existing gas concentration detection method is modeled and processed for specific gases, and it is difficult to handle the detection of multiple gases; while the transformer neural network method does not depend on a specific physical model, it can handle The detection of multiple gases has better versatility and scalability;

3. Can adapt to different environments and conditions: the existing gas concentration detection method is sensitive to environmental changes, sensor drift and other issues, and it is difficult to deal with complex environments and conditions; while the transformer neural network method is self-adaptive learning and optimization , can better adapt to different environments and conditions, and improve the robustness and stability of detection;

4. Online detection and real-time monitoring can be realized: the existing gas concentration detection method requires offline processing and model training, and cannot realize real-time monitoring and online detection; while the deep learning method has faster training speed and lower computational complexity, Real-time monitoring and online detection can be realized, which has better practicality and application value;

In summary, the sensor gas concentration detection method based on the deep learning transformer neural network of the present invention has the advantages of high precision, strong versatility, good robustness, and strong real-time performance. It can overcome the problems existing in traditional methods and achieve better Gas concentration detection effect.

Description of drawings

In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings that need to be used in the description of the specific embodiments or the prior art. Throughout the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, elements or parts are not necessarily drawn in actual scale.

Figure 1: The response curve of the gas sensor to 1 μL SO2F2;

Fig. 2: In the present invention, the overall architecture diagram of sequence processing based on the transformer model;

Fig. 3: the sensor testing time of the present invention narration and voltage change graph;

Figure 4: Two-dimensional diagram of sequence information in the present invention.

Detailed ways

In order to clearly and completely describe the technical solution of the present invention and its specific working process, in conjunction with the accompanying drawings, the specific implementation of the present invention is as follows:

Example 1

A sensor gas concentration detection method based on a deep learning transformer neural network provided in this embodiment specifically includes the following steps:

Step 1: Data set preparation and preprocessing;

Collect the data of the gas sensor, and preprocess the data, the preprocessing includes cleaning, denoising and standardization, and obtain the gas concentration series data with time dimension after preprocessing;

In this embodiment, specifically data processing includes the following steps:

A. Discretize time series: discretize continuous time series data into data with a fixed time interval of 10mins;

B. Sequence standardization: perform mean normalization processing on the discretized time series to make them have similar statistical characteristics;

C. Construct the input sequence: convert the mean-normalized time series data into an input sequence, that is, input data of a fixed length of time as a sequence into the Transformer model;

D. Batching and filling: For the case where the length of the input sequence is insufficient, a filling operation is performed to ensure the consistency of the length of the input sequence;

Step 2: Construction of the Transformer neural network model;

Segment the data set, bring the training set into the embedding layer (embedding), adjust the hyperparameters of the encoder (Encoder) and decoder (Decoder) modules in the model, and refer to the mean square error (MSE) and mean absolute error (MAE) indicators. Optimization functions such as grid search evaluate the optimal combination of hyperparameters, and use the best MSE and MAE to characterize model performance;

For gases with different characteristics, the detection parameters are quite different. In this embodiment, the model construction of nitrogen dioxide gas is taken as an example. The specific steps are as follows:

① Data segmentation, train the Transformer neural network model on the two-dimensional time series data obtained in step 1, and perform data segmentation, and divide the 24-hour data into fixed-length time windows, each time window length is 10mins to 30mins; and Use the ratio of 70/15/15 to divide the data set; use the first 70% of the time window as the training set, the middle 15% as the verification set, and the remaining 15% as the test set;

②Set the hyperparameter range. First, determine the length of the input sequence. According to the sampling frequency of the data and the application scenario, select the data points from 0 to 24 hours as an input sequence length to comprehensively record the change value of the ambient gas concentration; then, Determine the batch size, and set the batch size to 32, 64, or 128 according to the computing power of the GPU or TPU to avoid problems such as insufficient video memory; then, determine the number of hidden layers, and select the optimal one according to the complexity of the data and the limitation of computing resources The number of hidden layers, choose the number of hidden layers of 5-6 layers to improve the expression ability of the model; finally, determine the number of heads, the multi-head self-attention mechanism is one of the key components of the transformer network, in order to enable the model to better capture different time Dependency between steps, set the number of heads to 6-8;

③ Grid search, use the grid search method to search for the optimal hyperparameter combination within the hyperparameter range; the grid search method is an exhaustive search method that traverses all possible parameter combinations and selects the optimal parameter combination; the grid search method is implemented using the GridSearchCV library in Python;

④ Random search, using the random search method to randomly search for the optimal hyperparameter combination within the hyperparameter range. The random search method randomly selects some parameter combinations within the range of hyperparameters and selects the optimal parameter combination. The random search method can be implemented using the RandomizedSearchCV library in Python;

⑤Bayesian optimization, using the Bayesian optimization method to find the optimal hyperparameter combination within the hyperparameter range. Bayesian optimization methods search for the optimal combination of hyperparameters by establishing the connection between the prior distribution of hyperparameters and the posterior distribution of the model. The Bayesian optimization method is implemented using the BayesianOptimization library in Python;

⑥Evaluate model performance: Use the optimal hyperparameter combinations obtained by the above methods to train the model, and use the mean square error (MSE) and mean absolute error (MAE) to evaluate the optimal hyperparameter combination on the test set, and use The MSE and MAE corresponding to the optimal parameter model represent the performance of the model.

When using the Transformer neural network for time series processing in the gas sensor, the Encoder-Decoder model is usually used, where the Encoder module is responsible for converting the input sequence into a set of hidden representations, and the Decoder module is responsible for generating the output sequence. The overall framework of the scheme is shown in Figure 2. The time-amplitude sequence is provided by the lower computer, directly enters the transformer model through the embedding module, and is processed by the Encoder and Decoder modules to output the gas concentration value.

The embedding module, the embedding layer (embedding) is mainly used to convert the data collected by the sensor (such as a time series signal) into a vector form that can be processed by the neural network, that is, the data collected by the sensor is represented as a vector form, so that the neural network network for processing and learning. Specifically, the embedding layer maps the discrete values input by the host computer of the gas sensor to a continuous vector representation. In the Transformer neural network, the embedding layer usually consists of two parts: position encoder and input embedding. Among them, the position encoder is used to add position information to the input data at each time point, so that the model can learn the order of the time series. The input embedding is used to convert the input data at each time point into a fixed-dimensional vector representation for subsequent processing by the attention mechanism, encoder and decoder.

The Encoder module, the function of the Encoder module is to convert the input sequence into a set of hidden representations, so that the Decoder module can better utilize the information of the input sequence. In gas sensor applications, the Encoder module can be used to convert the gas concentration at the current time step into a set of hidden representations for use by the Decoder module. The Encoder module usually consists of the following components:

①Multi-Head Attention: The first component of the Encoder module is the multi-head attention mechanism, which is used to weight the input sequence so that the Encoder module can better utilize the information of the input sequence.

②Position-wise Feed-Forward Network: The second component of the Encoder module is a feed-forward neural network, which is used to weight and aggregate the output of the above-mentioned multi-head attention mechanism in order to generate a set of hidden representations.

③Layer Normalization: A LayerNormalization module is added between each component of the Encoder module to better perform signal transmission and prevent the gradient disappearance problem during model training.

The signal transmission relationship between the various components in the Encoder module is as follows:

Input sequence→Multi-Head Attention→Position-wise Feed-Forward Network→Hidden representation sequence

The Multi-Head Attention component in the Encoder module is used to extract information from the input sequence and generate a set of hidden representations for use by the Decoder module. The Position-wise Feed-Forward Network component is used to further process this information and generate the final sequence of hidden representations. Finally, the hidden representation sequence can be used in the Decoder module to generate the output sequence.

The Decoder module, the function of the Decoder module is to generate the output of the current time step according to the hidden representation provided by the Encoder module and the previously generated output. In a gas sensor application, the Decoder module is used to predict the gas concentration at the next time step. The Decoder module usually consists of the following components:

①Masked Multi-Head Attention: The first component of the Decoder module is the self-attention mechanism, which is used to calculate the relationship between the output of the current time step and the previously generated output, and to combine the hidden representation provided by the Encoder module with the current time step. output for interaction;

②Multi-Head Attention: The second component of the Decoder module is the multi-head attention mechanism, which is used to weight the hidden representation provided by the Encoder module so that the Decoder module can better utilize the information of the input sequence;

③Position-wise Feed-Forward Network: The third component of the Decoder module is the feed-forward neural network, which is used to weight the outputs of the above two attention mechanisms to generate the output of the current time step;

④Layer Normalization: A LayerNormalization module is added between each component of the Decoder module for better signal transmission and to prevent the gradient disappearance problem during model training.

The signal transmission relationship between the various components in the Decoder module is as follows:

Input sequence→Masked Multi-Head Attention→Multi-Head Attention→Position-wise Feed-Forward Network→Output sequence The Masked Multi-HeadAttention and Multi-Head Attention components in the Decoder module are used to extract information from the hidden representation provided by the Encoder module , and interact with the output of the current time step to generate the output of the current time step. The Position-wise Feed-Forward Network component is used to further process this information and generate the final output. Finally, the output sequence can be used to predict the gas concentration for the next time step.

Step 3: Training of the Transformer neural network model;

Select the extracted parameter value and its corresponding time value as the analysis feature quantity of the transformer neural network, as shown in accompanying drawing 3 and accompanying drawing 4, that is, Q _i =[t _i , v _i ] ^T , and Q _i means that at a certain time The data parameters sent by the lower computer, t _i and v _i are the concentration and time value of the corresponding gas respectively; the characteristic parameters of the time series curve are obtained as the training set, which is imported into the neural network through embedding, and after a series of interspersed, normalized, The attention mechanism training process obtains the parameter relationship between the curve feature changes, that is, the model training is completed;

Step 4: Use the trained Transformer neural network model to estimate the actual environmental parameter values and gas concentration values.

Example 2

A practical example of gas sensors using the Transformer model to detect the concentration of nitrogen dioxide gas:

First, collecting nitrogen dioxide gas sample data, placing the gas sensor in the monitoring area and collecting standard concentration nitrogen dioxide gas concentration data, the data includes the concentration of nitrogen dioxide gas and sequence values of time values;

Next, preprocess the collected data. Clean, denoise, and normalize the data; use a moving average to smooth the data and least squares to remove noise.

Then, the processed data is fed into the Transformer model for training. During the training process, the model will learn how to associate the input gas concentration data with the time series, and establish the weight parameters of the model;

After the training is completed, deploy the model to the actual system; when the concentration of nitrogen dioxide gas needs to be detected, the sensor only needs to transmit a small amount of collected sensor timing data to the Transformer model for inference;

Finally, the NO2 gas concentration is quickly assessed based on the model's output; if the concentration exceeds a safe threshold, the system can trigger an alarm or take other necessary actions.

The preferred embodiment of the present invention has been described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the specific details of the above embodiment, within the scope of the technical concept of the present invention, various simple modifications can be made to the technical solution of the present invention, These simple modifications all belong to the protection scope of the present invention.

In addition, it should be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable way if there is no contradiction. The combination method will not be described separately.

In addition, various combinations of different embodiments of the present invention can also be combined arbitrarily, as long as they do not violate the idea of the present invention, they should also be regarded as the disclosed content of the present invention.

Claims (9)

1. The sensor gas concentration detection method based on the deep learning transducer neural network is characterized by comprising the following steps of:

step one: preparing and preprocessing a data set;

collecting data of a gas sensor, and preprocessing the data, wherein the preprocessing comprises cleaning, denoising and standardization, and the preprocessing is performed to obtain gas concentration sequence data with a time dimension;

step two: constructing a transducer neural network model;

cutting the data set, taking the training set into an embedded layer embedding, adjusting super parameters of an Encoder Encoder module and a Decoder module in the model, evaluating an optimal super parameter combination by using optimization functions such as grid search and the like with reference to mean square error and average absolute error indexes, and characterizing the performance of the model by using optimal MSE and MAE;

step three: training a transducer neural network model;

step four: and estimating the actual environment parameter value and the gas concentration value by using the trained transducer neural network model.

2. The method for detecting the concentration of a sensor gas based on a deep learning transducer neural network according to claim 1, wherein in the first step, the data includes a sequence of values of concentration and time values of the gas, and the sequence of values including concentration and time in the sensor is converted into a format for a transducer model, thereby performing a model training task, specifically including the following:

a1, discretizing a time sequence: discretizing the continuous time series data into data fixed at time intervals of 10 min;

a2, sequence standardization: carrying out mean value normalization processing on the discretized time sequence to enable the discretized time sequence to have similar statistical characteristics;

a3, constructing an input sequence: converting the time sequence data after mean normalization into an input sequence, namely, taking a fixed time length of data as a sequence and inputting the sequence into a transducer neural network model;

a4, batch and filling: and for the condition that the input sequence length is insufficient, performing filling operation to ensure the consistency of the input sequence length.

3. The method for detecting the concentration of the sensor gas based on the deep learning transducer neural network according to claim 1, wherein in the second step, the data are processed in a time sequence by adopting an Encoder-Decoder model and an embedding layer of the transducer neural network; the embedded layer is used for converting data acquired by the sensor into a vector form which can be processed by the neural network; the Encoder module is configured to convert an input sequence into a set of hidden representations; the Decoder module is used for generating the output of the current time step according to the hidden representation provided by the Encoder module and the output generated before.

4. A method for detecting sensor gas concentration based on deep learning transducer neural network according to claim 3, wherein the embedding layer is composed of a position encoder and an input embedding, the position encoder is used for adding position information to the input data at each time point so as to facilitate model learning of the sequence of time series; input embedding is used to convert the input data at each point in time into a vector representation of fixed dimensions for subsequent attention mechanisms, encoders and decoders to handle.

5. The sensor gas concentration detection method based on deep learning transducer neural network of claim 3, wherein the Encoder module comprises:

the Multi-Head Attention mechanism is used for carrying out weighted convergence on the input sequence so that the Encoder module can better utilize the information of the input sequence;

the Feed Forward neural Network Position-wise Feed-Forward Network is used to weight aggregate the outputs of the multi-headed attention mechanism described above to generate a set of hidden representations.

6. The sensor gas concentration detection method based on deep learning transducer neural network according to claim 3, wherein the Decoder module comprises:

a self-Attention mechanism mask Multi-Head Attention for calculating a relationship between the output of the current time step and the previously generated output, and for interacting the hidden representation provided by the Encoder module with the output of the current time step;

the Multi-Head Attention mechanism is used for carrying out weighted aggregation on the hidden representations provided by the Encoder module so that the Decoder module can better utilize the information of the input sequence;

the feedforward neural Network Position-wise Feed-Forward Network is used for carrying out weighted aggregation on the outputs of the two attention mechanisms so as to generate the output of the current time step.

7. The method for detecting the concentration of the sensor gas based on the deep learning transducer neural network according to claim 3, wherein Layer Normalization modules are arranged between the components of the Encoder module and the Decoder module and are used for better signal transmission and preventing gradient disappearance problems during model training.

8. The method for detecting the concentration of the sensor gas based on the deep learning transducer neural network as claimed in claim 1, wherein the construction model in the second step specifically comprises the following steps:

b1: cutting the data set;

data segmentation is carried out on the sequence data obtained in the step one, 24 hours of data are segmented into time windows with fixed lengths, and the length of each time window is 10-30 mins; and dividing the data set by a ratio of 70/15/15; taking the first 70% of time window as a training set, 15% in the middle as a verification set and the remaining 15% as a test set;

b2: setting an over-parameter range;

firstly, determining the length of an input sequence, selecting data points from 0 to 24 hours as one input sequence length according to the sampling frequency of data and an application scene, and comprehensively recording the concentration change value of the ambient gas; then, determining the batch size and the number of hidden layers, and setting the batch as 32, 64 or 128; determining the number of hidden layers to be 5-6; finally, determining the number of heads to be 6-8;

b3: searching grids;

searching an optimal super-parameter combination in a super-parameter range by adopting a grid searching method;

b4: randomly searching;

randomly searching the optimal super-parameter combination in the super-parameter range by adopting a random searching method;

b5: bayesian optimization;

searching an optimal super-parameter combination in a super-parameter range by adopting a Bayes optimization method;

b6: evaluating the performance of the model;

respectively training a transducer neural network model by adopting the optimal super-parameter combinations obtained by the method, evaluating the optimal super-parameter combinations on a test set by using mean square error and average absolute error, and characterizing the model performance by using MSE and MAE corresponding to the optimal parameter models.

9. The method for detecting the concentration of the sensor gas based on the deep learning transducer neural network as claimed in claim 1, wherein the third step is as follows:

selecting the extracted parameter value and the corresponding time value as analysis characteristic quantity of a transducer neural network, namely Q _i ＝[t _i ，v _i ] ^T ,Q _i Representing data parameters, t, sent by a lower computer at a certain moment _i 、v _i The concentration and time value of the corresponding gas; will trainThe training set is led into the nerve network in an ebedding mode, and is subjected to a series of interpenetration, normalization and attention mechanism training processes to obtain the parameter relation between curve characteristic changes, namely model training is completed.

Images