Gas identification method and device

Abstract

The invention discloses a gas identification method and device, relates to the technical field of gas identification, and aims to solve the problem that the existing gas identification method cannot accurately identify gas types in a vacuum environment. The method comprises the steps of obtaining measurement data of a vacuum environment; converting the first response data into a two-dimensional image by adopting a Grubrum angle field, and processing the first response data by adopting a sectional interpolation method to obtain target response data; determining the vacuum degree of the vacuum environment according to the first response data and the second response data; based on the vacuum degree, a target method corresponding to the vacuum degree is selected to recognize the gas in the vacuum environment, and the gas type in the vacuum environment is obtained; the target method comprises a gas recognition model or a difference method. According to the gas recognition method, the Pirani sensor and the absolute pressure sensor are adopted for gas recognition in the vacuum environment, and the precision is high.

Description

Technical Field

[0001] The present invention relates to the technical field of gas recognition, and particularly to a gas recognition method and device. Background Art

[0002] In many critical fields such as modern industrial production and scientific research, the accuracy and reliability of gas recognition in a vacuum environment are of crucial importance. In numerous industries ranging from semiconductor manufacturing, high-vacuum physical experiments to aerospace, food preservation, and pharmaceutical production, accurately identifying the types and characteristics of gases in a vacuum environment plays an indispensable role in ensuring product quality, optimizing process flows, ensuring the safe operation of equipment, and promoting the progress of scientific research.

[0003] Traditional methods usually use gas sensors for gas recognition. When facing changing vacuum conditions, due to low sensitivity to certain gas components or being easily interfered by environmental factors such as temperature, humidity, and pressure changes, it is often impossible to collect sufficiently accurate information, resulting in deviations in measurement results, thereby affecting the accurate discrimination of gas types. In addition, as the usage time increases, the performance of the sensor often drifts, further reducing the measurement accuracy and making it difficult to meet the requirements of long-term stable and accurate gas recognition.

[0004] Therefore, there is an urgent need for a gas recognition method and device. Summary of the Invention

[0005] The purpose of the present invention is to provide a gas recognition method and device for solving the problem that the existing gas recognition methods cannot accurately identify the gas types in a vacuum environment.

[0006] To achieve the above purpose, the present invention provides the following technical solutions:

[0007] In the first aspect, the present invention provides a gas recognition method, including:

[0008] Obtaining measurement data of a vacuum environment; the measurement data includes first response data measured by a Pirani sensor and second response data measured by an absolute pressure sensor;

[0009] Converting the first response data into a two-dimensional image by using Gramian angular field and processing the first response data by using piecewise interpolation method to obtain target response data;

[0010] Determining the vacuum degree of the vacuum environment according to the first response data and the second response data;

[0011] Based on the vacuum degree, selecting a target method corresponding to the vacuum degree to recognize the gas in the vacuum environment to obtain the gas types in the vacuum environment; the target method includes a gas recognition model or a difference method.

[0012] Optionally, based on the vacuum degree, select the target method corresponding to the vacuum degree to identify the gas in the vacuum environment, and the types of gases in the vacuum environment obtained include:

[0013] When the vacuum degree is within the first vacuum range, input the two-dimensional image into the gas recognition model for gas recognition to obtain the types of gases in the vacuum environment; the first vacuum range is greater than 0 and less than 1000 Pa;

[0014] When the vacuum degree is within the second vacuum range, calculate the difference between the target response data and the second response data, and determine the gas type corresponding to the numerical range in the mapping relation table where the difference is located as the gas type in the vacuum range; the second vacuum range is greater than or equal to 1000 Pa and less than the atmospheric pressure; the mapping relation table is a correspondence table between gas types and numerical ranges.

[0015] Optionally, the method for determining the vacuum degree of the vacuum environment according to the first response data and the second response data includes:

[0016] When the first response data and the second response data are within the first vacuum range, determine the vacuum degree corresponding to the first response data as the vacuum degree of the vacuum environment;

[0017] When the first response data and the second response data are within the second vacuum range, determine the vacuum degree corresponding to the second response data as the vacuum degree of the vacuum environment.

[0018] Optionally, the gas recognition model includes: four convolutional layers, a max pooling layer, and three fully connected layers, and the number of output channels of the convolutional layers increases from top to bottom; inputting the two-dimensional image into the gas recognition model for gas recognition, the types of gases in the vacuum environment obtained include:

[0019] Use four convolutional layers to extract local features from the two-dimensional image to obtain a two-dimensional feature map;

[0020] Use the max pooling layer to perform downsampling on the two-dimensional feature map, and select the maximum value in each pooling area for output to obtain multi-scale local features;

[0021] Perform pooling flattening on the multi-scale local features to obtain multiple one-dimensional feature vectors;

[0022] Concatenate multiple one-dimensional feature vectors to obtain a concatenated feature vector;

[0023] Use a fully connected layer to perform dimensionality reduction on the concatenated feature vector, and use the softmax function for classification in the last fully connected layer to obtain the gas type.

[0024] Optionally, before selecting a target method to identify the gas in the vacuum environment based on the degree of vacuum and obtaining the types of gases in the vacuum environment, the following steps are further included:

[0025] Obtain training data; the training data includes two-dimensional images corresponding to different types of gases;

[0026] Input the training data into a convolutional neural network model to obtain predicted gas types;

[0027] Compare the predicted gas types with the actual gas types to obtain a comparison result;

[0028] Adjust the parameters of the convolutional neural network model according to the comparison result to obtain a gas recognition model.

[0029] Optionally, the step of converting the first response data into a two-dimensional image by using the Gramian angular field includes:

[0030] Calculate the cosine similarity between each group of data points in the first response data, and each group of data points includes any two data points in the first response data;

[0031] Map the cosine similarity to the interval [0, Π] to obtain a plurality of mapped points, and convert the mapped points into angles. The plurality of angles form a two-dimensional image in matrix form.

[0032] Compared with the prior art, a gas recognition method provided by the present invention converts the first response data into a two-dimensional image by using the Gramian angular field, and processes the first response data by using a piecewise interpolation method to obtain target response data, which can reflect the sensitive characteristics of the Pirani sensor to gases with different thermal conductivities; determine the degree of vacuum in the vacuum environment according to the first response data measured by the Pirani sensor and the second response data of the absolute pressure sensor; based on the degree of vacuum, select a target method corresponding to the degree of vacuum to identify the gas in the vacuum environment and obtain the types of gases in the vacuum environment; this method utilizes the gas sensitivity of the Pirani sensor and combines the data measured by the absolute pressure sensor to accurately identify the types of gases in different vacuum environments; determining the degree of vacuum in the vacuum environment through the measurement data of the two sensors can avoid the problem that the sensor performance drifts over time, resulting in a decrease in measurement accuracy, and can determine the degree of vacuum stably and accurately for a long time, and further identify the types of gases stably and accurately for a long time. In addition, selecting different methods for gas recognition based on the degree of vacuum can accurately identify gases in different vacuum ranges.

[0033] Second aspect, the present invention provides a gas identification device for implementing the above-mentioned gas identification method. The gas identification device at least includes: a Pirani sensor, an absolute pressure sensor, a calibration fusion module, a data processing module, and a gas identification module;

[0034] Both the absolute pressure sensor and the Pirani sensor are used to measure the pressure in a vacuum environment;

[0035] The data processing module is used to convert the first response data measured by the Pirani sensor into a two-dimensional image using the Gram angular field, and process the first response data using the piecewise interpolation method to obtain target response data;

[0036] The calibration fusion module is used to determine the vacuum degree of the vacuum environment according to the first response data and the second response data measured by the absolute pressure sensor;

[0037] The gas identification module is used to select a target method to identify the gas in the vacuum environment based on the vacuum degree, and obtain the gas type in the vacuum environment; the target method includes a gas identification model or a difference method.

[0038] Optionally, the gas identification module includes:

[0039] A first gas identification unit, configured to input the two-dimensional image into the gas identification model for gas identification when the vacuum degree is within a first vacuum range, and obtain the gas type in the vacuum environment; the first vacuum range is greater than 0 and less than 1000 Pa;

[0040] A second gas identification unit, configured to calculate the difference between the target response data and the second response data when the vacuum degree is within a second vacuum range, and determine the gas type corresponding to the numerical range in the mapping relation table where the difference is located as the gas type in the vacuum range; the second vacuum range is greater than or equal to 1000 Pa and less than the atmospheric pressure; the mapping relation table is a corresponding relation table between gas types and numerical ranges.

[0041] Optionally, the gas identification device further includes a conditioning chip, which is communicatively connected to the absolute pressure sensor, and the conditioning chip is used to convert the electrical signal of the absolute pressure sensor into a digital signal.

[0042] Optionally, the gas identification device further includes: a first inner shell and a second inner shell. Both the first inner shell and the second inner shell are provided with ventilation holes. The conditioning chip and the absolute pressure sensor are arranged in the first inner shell, and the Pirani sensor is arranged in the second inner shell. The first inner shell and the second inner shell are connected to the data processing module through pins.

[0043] Compared with the prior art, a gas identification device provided by the present invention can reflect the sensitive characteristics of a Pirani sensor to gases with different thermal conductivities by processing the first response data through the segmented difference method of the data processing module. The calibration and fusion module can determine the vacuum degree of the vacuum environment according to the first response data and the second response data, which can avoid the problem that the sensor performance drifts over time, resulting in a decrease in the measurement accuracy. It can determine the vacuum degree stably and accurately for a long time, and then identify the gas type stably and accurately for a long time. It not only improves the measurement accuracy of the vacuum degree, but also solves the problem of sensor performance drift under long-term operation, greatly improving the stability and adaptability of the device; the gas identification module is used to identify the gas in the vacuum environment by selecting a target method based on the vacuum degree, and can accurately identify gases in different vacuum ranges. Description of the Drawings

[0044] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention. In the drawings:

[0045] Figure 1 It is a flowchart of a gas identification method provided by the present invention;

[0046] Figure 2 It is a schematic structural diagram of an experimental device provided by the present invention;

[0047] Figure 3 It is the response data of the Pirani sensor under different pressures and different gas types provided by the present invention;

[0048] Figure 4 It is the response data of the absolute pressure sensor under different pressures and different gas types provided by the present invention;

[0049] Figure 5 It is a comparison diagram of the response data of the Pirani sensor and the absolute pressure sensor provided by the present invention;

[0050] Figure 6 It is a schematic diagram of the training probability and the test probability provided by the present invention;

[0051] Figure 7 It is a schematic diagram of the training loss and the test loss provided by the present invention;

[0052] Figure 8 It is a schematic structural diagram of a gas identification device provided by the present invention.

[0053] Reference Signs:

[0054] 1 - Pirani sensor, 2 - absolute pressure sensor, 3 - conditioning chip, 4 - microprocessor, 41 - calibration and fusion module, 42 - gas identification module, 43 - data processing module, 44 - communication interface, 21 - constant temperature control device, 22 - gas cylinder, 23 - vacuum control device, 24 - data acquisition device. Specific embodiments

[0055] For the convenience of clearly describing the technical solutions of the embodiments of the present invention, in the embodiments of the present invention, terms such as "first" and "second" are used to distinguish identical or similar items with basically the same functions and roles. For example, the first threshold and the second threshold are only used to distinguish different thresholds, and do not limit their order. Those skilled in the art can understand that terms such as "first" and "second" do not limit the quantity and execution order, and terms such as "first" and "second" do not necessarily limit being different.

[0056] It should be noted that in the present invention, words such as "exemplary" or "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as "exemplary" or "for example" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly speaking, using words such as "exemplary" or "for example" aims to present related concepts in a specific way.

[0057] In the present invention, "at least one" means one or more, and "a plurality" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone, where A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects before and after. "At least one (item)" or its similar expression below refers to any combination of these items, including any combination of single item (item) or plural items (items). For example, at least one (item) of a, b or c can represent: a, b, c, the combination of a and b, the combination of a and c, the combination of b and c, or the combination of a, b and c, where a, b and c can be single or multiple.

[0058] Before introducing the embodiments of the present invention, the following interpretations are made for the relevant nouns involved in the embodiments of the present invention:

[0059] Piecewise interpolation method: It is a numerical calculation method mainly used for interpolation between given discrete data points to improve the accuracy and smoothness of interpolation. By dividing the interpolation interval into several segments and using different interpolation polynomials for interpolation in each segment, the Runge phenomenon that may occur in high-degree polynomial interpolation can be avoided. Piecewise linear interpolation uses a linear function for interpolation in each small interval, while piecewise quadratic interpolation uses a quadratic polynomial.

[0060] The Softmax function is an activation function used in machine learning for multi-class classification problems. It converts a real vector into a probability distribution, where each element value is between 0 and 1 and the sum is 1. This function is commonly used in the output layer of neural networks to ensure that the output can be interpreted as the probability of a class.

[0061] GAE, namely Gramian Angular Field, is a method for converting one-dimensional time series data into a two-dimensional image representation. This method treats the data points in the time series as points in a vector space, calculates the cosine values of the angles between these points, and then maps these cosine values to the pixels of a two-dimensional image, thereby generating an image that can reflect the dynamic and periodic characteristics of the time series.

[0062] The traditional method of using gas sensors for gas identification is vulnerable to environmental interference, resulting in false alarms and thus affecting the accurate discrimination of gas types.

[0063] To solve the above problems, the present invention provides a gas identification method and device, which make full use of the response differences of Pirani sensors and absolute pressure sensors to gas characteristics under different vacuum degrees, and combine a gas identification model based on deep learning and the difference method to achieve high-accuracy identification of gas types. The following will be described in conjunction with the accompanying drawings.

[0064] See Figure 1 , the present invention provides a gas identification method, including the following steps:

[0065] Step 101: Obtain measurement data of the vacuum environment;

[0066] The measurement data includes first response data measured by a Pirani sensor and second response data measured by an absolute pressure sensor;

[0067] A Pirani sensor is a vacuum degree measurement element based on the principle of heat conduction. Its working principle is to determine the vacuum degree by detecting the heat exchange between a heating element and the surrounding gas. Different types of gases have different thermal conductivities at the same vacuum degree, so a Pirani sensor can be used for gas identification.

[0068] The absolute pressure sensor is a mechanically sensitive structure. When an external pressure acts on the element, the absolute pressure sensor will undergo a slight deformation, which will cause a change in the electrical properties of the sensitive structure inside the element. This change amount is proportional to the pressure, so pressure measurement can be achieved.

[0069] Step 102: Convert the first response data into a two-dimensional image using Gramian Angular Field, and process the first response data using the piecewise interpolation method to obtain target response data;

[0070] Specifically, the conversion of the first response data into a two-dimensional image using the Gram angular field includes:

[0071] Calculating the cosine similarity between each group of data points in the first response data, where each group of data points includes any two data points in the first response data;

[0072] Mapping the cosine similarity into the interval [0, Π] to obtain a plurality of mapped points, and converting the mapped points into angles, and the plurality of angles form a two-dimensional image in matrix form.

[0073] Processing the first response data using the Gram angular field GAF helps to capture the spatial correlation in the time series data, enabling the convolutional neural network to better learn the internal pattern of the data.

[0074] The piecewise interpolation method can reflect the sensitive characteristics of the Pirani sensor to gases with different thermal conductivities.

[0075] Step 103: Determine the vacuum degree of the vacuum environment according to the first response data and the second response data;

[0076] Specifically, when the vacuum degree corresponding to the first response data and the vacuum degree corresponding to the second response data are within the first vacuum range, the vacuum degree corresponding to the first response data is determined as the vacuum degree of the vacuum environment; the first vacuum range is greater than 0 and less than 1000 Pa;

[0077] When the vacuum degree corresponding to the first response data and the vacuum degree corresponding to the second response data are within the second vacuum range, the vacuum degree corresponding to the second response data is determined as the vacuum degree of the vacuum environment. The second vacuum range is greater than or equal to 1000 Pa and less than the atmospheric pressure.

[0078] As an alternative method, when the first response data and the second response data are within different vacuum ranges, the vacuum degree corresponding to the second response data is determined as the vacuum degree of the vacuum environment. Or when the vacuum degrees corresponding to the first response data and the second response data are both near 1000 Pa, the vacuum degree corresponding to the second response data is determined as the vacuum degree of the vacuum environment.

[0079] Since the Pirani sensor has a higher vacuum degree measurement accuracy than the absolute pressure sensor in the first vacuum range, and the absolute pressure sensor has a higher vacuum degree measurement accuracy than the Pirani sensor in the second vacuum range, the above method for determining the vacuum degree can improve the accuracy of the measured vacuum degree.

[0080] Step 104: Based on the vacuum degree, select the target method corresponding to the vacuum degree to identify the gas in the vacuum environment, and obtain the gas type in the vacuum environment;

[0081] The target method includes a gas recognition model or a difference method.

[0082] As an alternative, step 104 can be implemented based on the following steps:

[0083] Step 1041: When the vacuum degree is within the first vacuum range, input the two-dimensional image into the gas recognition model for gas recognition to obtain the gas types in the vacuum environment;

[0084] Utilizing the characteristic that the Pirani sensor is sensitive to gas types, taking the data measured by the Pirani gas sensor as the input of the gas recognition model can effectively perform gas recognition. The gas recognition model is a model established based on a deep learning algorithm. The deep learning algorithm can automatically learn complex feature patterns and internal laws from a large amount of data without relying on manually designed feature extraction methods. By constructing a deep neural network model for the sensor measurement data and using intelligent algorithms for multi-level abstraction and analysis, deeper measurement information can be mined, and more accurate gas type recognition can be achieved in gas recognition.

[0085] Step 1042: When the vacuum degree is within the second vacuum range, calculate the difference between the target response data and the second response data, and determine the gas type corresponding to the numerical range in the mapping relation table where the difference is located as the gas type in the vacuum range; the mapping relation table is a corresponding relation table between gas types and numerical ranges.

[0086] Since the Pirani sensor is sensitive to gas types, there are differences in the response data of different gases measured at the same vacuum degree, while the absolute pressure sensor is not sensitive to gas types and the response data of different gases measured at the same vacuum degree is the same. Therefore, by calculating the difference between the first response data measured by the Pirani sensor and the second response data measured by the absolute pressure sensor, different gas types can be identified according to the numerical range where the difference is located.

[0087] The present invention selects different gas recognition methods in different vacuum ranges, which can ensure high-precision gas type recognition. At the same time, the present invention is applicable to the gas type recognition in a vacuum environment with continuously changing vacuum degree.

[0088] The gas recognition model in the above steps includes four convolutional layers, a max pooling layer, and three fully connected layers. The number of output channels of the convolutional layers increases from top to bottom, being 16, 32, 64, and 128 in sequence. At the same time, each convolutional layer uses a convolutional kernel of size 4×4. This size of convolutional kernel can capture features in a larger range while maintaining computational efficiency. The max pooling layer uses a pooling kernel of 3×3, which means that within each 3×3 region, only the maximum value is retained as the representative of that region. The first fully connected layer from top to bottom contains 512 neurons, the second fully connected layer contains 256 neurons, and the third fully connected layer contains the number of gas types, such as four gas types including nitrogen, helium, argon, and carbon dioxide.

[0089] Specifically, step 1041 can be implemented based on the following steps:

[0090] Use four convolutional layers to extract local features from the two-dimensional image to obtain a two-dimensional feature map;

[0091] Use the max pooling layer to perform downsampling on the two-dimensional feature map and select the maximum value in each pooling region for output to obtain multi-scale local features;

[0092] Perform pooling flattening on the multi-scale local features to obtain multiple one-dimensional feature vectors;

[0093] Concatenate multiple one-dimensional feature vectors to obtain a concatenated feature vector; this can achieve the fusion of local features and global features.

[0094] Use a fully connected layer to perform dimensionality reduction on the concatenated feature vector and use the softmax function for classification in the last fully connected layer to obtain the gas type.

[0095] To improve the stability and convergence speed of the model, layer normalization and the exponential linear unit activation function are used in the convolutional neural network. Layer normalization makes the model more stable during training by normalizing the input of each layer. The ELU activation function can introduce non-linearity and provide a smooth output in the negative value region, helping the model better fit complex data distributions.

[0096] As can be seen from the above method, after multiple convolutional and pooling operations, the CNN can extract multi-level and multi-scale features from the input GAF image. These features include not only local details, such as the change in sensor readings at a single time point, but also global information, such as the overall behavior of the gas response during the entire measurement process. Converting the multi-scale local features into a one-dimensional feature vector can preserve the spatial structure information of the original features and the feature information at different scales. By concatenating the flattened features, the fusion of local features and global features can be achieved. Applying the softmax function for gas type classification, the output values are given in the form of probabilities. In this way, the model can effectively identify and classify different gas types.

[0097] As an alternative way, before performing gas identification in step 104, it is necessary to construct and train a gas identification model and establish a mapping relation table. Specifically, it is achieved through the following steps:

[0098] See Figure 2 , first, data collection is carried out. Specifically, a composite sensor composed of a Pirani sensor and an absolute pressure sensor is placed in the thermostat of the temperature control device 21. There is a set target temperature environment in the thermostat. The temperature control device 21 is connected to different gas cylinders 22, such as helium gas cylinders, carbon dioxide gas cylinders, nitrogen gas cylinders, argon gas cylinders, etc., for inputting gas into the thermostat. The temperature control device 21 is connected to a vacuum control device 23, and the vacuum control device 23 is used to control the vacuum environment of the thermostat. Each time a type of gas is input into the thermostat, then the vacuum control device 23 controls the vacuum environment in the thermostat to change from high vacuum to low vacuum. When the vacuum environment pressure is stable, the data collection device 24 collects the response data of the Pirani sensor and the absolute pressure sensor corresponding to the target vacuum point. Each vacuum point is collected and tested more than five times. The collected data is converted into a response curve that changes with the vacuum value from high to low. For the intuitiveness of data observation, the abscissa representing the vacuum degree takes the logarithm with base 10. The response curves of the Pirani sensor to four gases are as Figure 3 shown. The data measured by the Pirani sensor for different gases at different vacuum degrees are different. The response curves of the absolute pressure sensor to four gases are as Figure 4 shown. The response of the absolute pressure sensor to the four gases at different vacuum degrees is the same. As Figure 5 shown, the absolute pressure sensor is not sensitive to gas types, but has high measurement accuracy for vacuum degree in the second vacuum range. The Pirani sensor is highly sensitive to gas types and has high measurement accuracy for vacuum degree in the first vacuum range.

[0099] Then, data processing is carried out. The piecewise interpolation method is used to process the data corresponding to the Pirani sensor within the second vacuum range. The GAF algorithm is used to process the data corresponding to the Pirani sensor within the first vacuum range to obtain two-dimensional images corresponding to different gases.

[0100] Finally, a mapping relationship table and a training gas recognition model are constructed. Among them, the mapping relationship table is the corresponding relationship among gas types, vacuum points, and numerical ranges. Calculate the difference between the data corresponding to the two sensors of each gas at different vacuum points, sort the obtained differences from largest to smallest, and the adjacent two differences are the maximum and minimum values of the numerical range corresponding to the gas. Generate a mapping relationship table according to the numerical ranges corresponding to each gas at different vacuum points.

[0101] The steps for constructing and training the gas recognition model are as follows: Divide the two-dimensional images corresponding to different gases into training data and test data according to a certain ratio; First, obtain the training data;

[0102] Input the training data into the convolutional neural network model to obtain the predicted gas type;

[0103] Compare the predicted gas type with the actual gas type to obtain a comparison result;

[0104] Adjust the parameters of the convolutional neural network model according to the comparison result until the accuracy converges to a satisfactory value to obtain the gas recognition model.

[0105] Then, use the test data to test the trained gas recognition model. Input the test data into the model to obtain the prediction result, compare it with the actual gas type, and calculate evaluation indicators such as accuracy and loss value. If the expected value is not reached, further optimize the model architecture or adjust the training parameters until the performance of the gas recognition model meets the requirements. The model training results and test results are as Figure 6 shown. As the training progresses, the accuracy of the model's prediction result gradually approaches 1, and the test result shows that the accuracy is close to 1. As Figure 7 shown. As the training progresses, the loss value gradually decreases and gradually stabilizes at around 0.7.

[0106] As an alternative approach, the data from the absolute pressure sensor and the Pirani sensor can also be jointly used as the input of a convolutional neural network to train a gas recognition model, and then gas recognition can be performed based on the trained gas recognition model. Specifically, the response data of the Pirani sensor and the response data of the absolute pressure sensor are collected in different gas environments and at different vacuum points. The collected data is converted into a response curve that changes from high to low with the vacuum value. For the response curves of different gases corresponding to the Pirani sensor, the peak value of the curve slope change, the area under the curve, and the data fluctuation range within a specific time period are extracted. For the response curves of different gases corresponding to the absolute pressure sensor, the curve slope is extracted. The extracted feature values are standardized, and the standardized data is randomly divided into a training set and a test set according to a certain ratio, such as 70% and 30%. A neural network model is constructed using the CNN algorithm. The training set is input into the CNN model, and continuous iterative training is performed until the accuracy converges to a satisfactory value to obtain the gas recognition model. The model is tested using the test set, the model is corrected based on the test results, and gas recognition is performed based on the corrected gas recognition model.

[0107] See Figure 8 , based on the above method, the present invention also provides a gas recognition device for implementing the above gas recognition method, such as Figure 8 shown, the gas recognition device includes a composite sensor and a microprocessor 4. The composite sensor includes a Pirani sensor 1, an absolute pressure sensor 2, and a conditioning chip 3. The microprocessor 4 includes a calibration fusion module 41, a data processing module 43, and a gas recognition module 42;

[0108] Among them, the conditioning chip 3 is communicatively connected to the absolute pressure sensor 2. The absolute pressure sensor 2 and the conditioning chip 3 are arranged in the first inner shell, and the Pirani sensor 1 is arranged in the second inner shell. Both the first inner shell and the second inner shell are provided with air holes to ensure that the internal sensors can sense changes in the external gas environment. The first inner shell and the second inner shell are connected to the data processing module through pins. The first inner shell and the second inner shell can be To housings.

[0109] The absolute pressure sensor 2 and the Pirani sensor 1 in the above structure are both used to measure the pressure in the vacuum environment; the conditioning chip 3 is used to convert the electrical signal of the absolute pressure sensor into a digital signal to reduce noise interference.

[0110] The data processing module 43 is used to convert the first response data measured by the Pirani sensor 1 into a two-dimensional image using the Gramian angular field and process the first response data using the piecewise interpolation method to obtain the target response data;

[0111] Specifically, the converting the first response data into a two-dimensional image using the Gramian angular field includes:

[0112] Calculate the cosine similarity between each group of data points in the first response data, where each group of data points includes any two data points in the first response data;

[0113] Map the cosine similarity to the interval [0, Π] to obtain multiple mapping points, and convert the mapping points into angles. The multiple angles form a two-dimensional image in matrix form.

[0114] The calibration and fusion module 41 is used to determine the vacuum degree of the vacuum environment according to the first response data and the second response data measured by the absolute pressure sensor 2; and send the determined vacuum degree to the gas identification module 42 and output it through the communication interface 44.

[0115] The gas identification module 42 is used to select a target method to identify the gas in the vacuum environment based on the vacuum degree, obtain the gas type in the vacuum environment, and output the determined gas type through the communication interface 44; the target method includes a gas identification model or a difference method.

[0116] The gas identification module 42 in the above structure specifically includes:

[0117] The first gas identification unit is used to input the two-dimensional image into the gas identification model for gas identification when the vacuum degree is within the first vacuum range to obtain the gas type in the vacuum environment; the first vacuum range is greater than 0 and less than 1000 pa;

[0118] The second gas identification unit is used to calculate the difference between the target response data and the second response data when the vacuum degree is within the second vacuum range, and determine the gas type corresponding to the numerical range in the mapping relation table where the difference is located as the gas type in the vacuum range; the second vacuum range is greater than or equal to 1000 pa and less than the atmospheric pressure; the mapping relation table is a corresponding relation table between gas types and numerical ranges.

[0119] Among them, the gas identification model includes: four convolutional layers, a max pooling layer, and three fully connected layers. The number of output channels of the convolutional layer increases from top to bottom; the above first gas identification unit is specifically used for: using four convolutional layers to perform local feature extraction on the two-dimensional image to obtain a two-dimensional feature map;

[0120] Using the max pooling layer to perform downsampling processing on the two-dimensional feature map, and selecting the maximum value in each pooling area for output to obtain multi-scale local features;

[0121] Performing pooling flattening processing on the multi-scale local features to obtain multiple one-dimensional feature vectors;

[0122] Concatenate multiple one-dimensional feature vectors to obtain a concatenated feature vector;

[0123] The fully connected layer is used to perform dimensionality reduction processing on the spliced feature vector, and the softmax function is used for classification in the last fully connected layer to obtain the gas type.

[0124] The calibration fusion module 41 in the above structure may specifically include:

[0125] The first execution unit is configured to determine the vacuum degree corresponding to the first response data as the vacuum degree of the vacuum environment when the first response data and the second response data are within the first vacuum range;

[0126] The second execution unit is configured to determine the vacuum degree corresponding to the second response data as the vacuum degree of the vacuum environment when the first response data and the second response data are within the second vacuum range.

[0127] Optionally, the above Pirani sensor and absolute pressure sensor are both built-in with a calibration module for calibrating the sensor to ensure the accuracy of the sensor.

[0128] Optionally, the data processing module is further configured to perform preliminary processing on the first response data and the second response data to convert them into data that can be used by the gas recognition model and the calibration fusion module.

[0129] Before using this device to measure the vacuum degree and identify gases, it is necessary to construct a mapping relationship table and a gas recognition model. The method is the same as the relevant content of the above gas recognition method and will not be elaborated here.

[0130] This gas recognition device integrates the pressure information provided by the absolute pressure sensor and the response characteristics of the Pirani sensor due to the difference in the thermal conductivity of different gases. It determines the vacuum degree according to the data measured by the two sensors, and identifies the gas type through the gas recognition model of the deep learning algorithm and the difference method, realizing the simultaneous measurement of the vacuum degree and the identification of the gas type. It avoids the situation that the traditional measurement of the vacuum degree and gas identification uses two separate sensors, resulting in a large overall volume and high power consumption, which limits its application range.

[0131] The above mainly introduces the solution provided by the embodiment of the present invention from the perspective of the interaction between each module. It can be understood that in order to implement the above functions, it includes the corresponding hardware structure and / or software module for executing each function. Those skilled in the art should easily realize that, combined with the units and algorithm steps of each example described in the embodiments disclosed in this article, the present invention can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in the way of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0132] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present invention are executed in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a terminal, a user device, or other programmable devices. The computer program or instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer program or instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center in a wired or wireless manner. The computer-readable storage medium can be any available medium that can be accessed by a computer, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium, such as a floppy disk, a hard disk, or a magnetic tape; it can also be an optical medium, such as a digital video disc (DVD); or it can be a semiconductor medium, such as a solid state drive (SSD).

[0133] Although the present invention has been described in conjunction with various embodiments, however, in the process of implementing the claimed invention, those skilled in the art can understand and implement other variations of the disclosed embodiments by viewing the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "one" does not exclude a plurality. A single processor or other unit can implement several functions recited in the claims. Certain measures are recited in mutually different dependent claims, but this does not mean that these measures cannot be combined to produce good results.

[0134] Although the present invention has been described in connection with specific features and their embodiments, it is obvious that various modifications and combinations can be made without departing from the spirit and scope of the present invention. Accordingly, the present specification and the drawings are merely exemplary illustrations of the invention defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present invention. Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these changes and modifications.

Claims

1. A gas identification method, characterized in that: include: Acquire measurement data of a vacuum environment; the measurement data includes first response data measured by a Pirani sensor and second response data measured by an absolute pressure sensor; The first response data is converted into a two-dimensional image by using a Gram angle field, and the first response data is processed by using a piecewise interpolation method to obtain target response data; determining the vacuum degree of the vacuum environment according to the first response data and the second response data; Based on the vacuum degree, a target method corresponding to the vacuum degree is selected to identify the gas in the vacuum environment to obtain the type of gas in the vacuum environment; the target method includes a gas identification model or a difference method.

2. The gas identification method according to claim 1, characterized in that: Based on the vacuum degree, a target method corresponding to the vacuum degree is selected to identify the gas in the vacuum environment, and the gas types in the vacuum environment include: When the vacuum degree is within a first vacuum range, the two-dimensional image is input into the gas identification model for gas identification to obtain the type of gas in the vacuum environment; the first vacuum range is greater than 0 and less than 1000 Pa; When the vacuum degree is within the second vacuum range, the difference between the target response data and the second response data is calculated, and the gas type corresponding to the numerical range in the mapping relationship table where the difference is located is determined as the gas type in the vacuum range; the second vacuum range is greater than or equal to 1000 Pa and less than atmospheric pressure; the mapping relationship table is a correspondence table between gas types and numerical ranges.

3. The gas identification method according to claim 2, characterized in that: Determining the vacuum degree of the vacuum environment according to the first response data and the second response data includes: When the first response data and the second response data are within the first vacuum range, determining the vacuum degree corresponding to the first response data as the vacuum degree of the vacuum environment; When the first response data and the second response data are within the second vacuum range, the vacuum degree corresponding to the second response data is determined as the vacuum degree of the vacuum environment.

4. The gas identification method according to claim 2, characterized in that: The gas recognition model includes: four convolutional layers, a maximum pooling layer and three fully connected layers, and the number of output channels of the convolutional layers increases from top to bottom; the two-dimensional image is input into the gas recognition model for gas recognition, and the gas types in the vacuum environment include: Using four convolutional layers to extract local features from the two-dimensional image to obtain a two-dimensional feature map; The maximum pooling layer is used to downsample the two-dimensional feature map, and the maximum value in each pooling area is selected for output to obtain multi-scale local features; Performing pooling and flattening processing on the multi-scale local features to obtain multiple one-dimensional feature vectors; Concatenate multiple one-dimensional feature vectors to obtain a concatenated feature vector; A fully connected layer is used to reduce the dimension of the concatenated feature vector, and a softmax function is used in the last fully connected layer to perform classification to obtain the gas type.

5. The gas identification method according to claim 2, characterized in that: The method of selecting a target method to identify the gas in the vacuum environment based on the vacuum degree to obtain the type of gas in the vacuum environment also includes: Acquire training data; the training data includes two-dimensional images corresponding to different types of gases; Inputting the training data into a convolutional neural network model to obtain predicted gas types; Comparing the predicted gas type with the actual gas type to obtain a comparison result; The parameters of the convolutional neural network model are adjusted according to the comparison result to obtain a gas recognition model.

6. The gas identification method according to claim 1, characterized in that: The converting the first response data into a two-dimensional image by using the Gram angle field comprises: Calculating the cosine similarity between groups of data points in the first response data, each group of data points including any two data points in the first response data; The cosine similarity is mapped to the interval [0, Π] to obtain a plurality of mapping points, the mapping points are converted into angles, and the plurality of angles form a two-dimensional image in a matrix form.

7. A gas identification device, characterized in that: Used to implement the gas identification method according to any one of claims 1 to 6, the gas identification device comprises at least: a Pirani sensor, an absolute pressure sensor, a calibration fusion module, a data processing module and a gas identification module; The absolute pressure sensor and the Pirani sensor are both used to measure the pressure of a vacuum environment; The data processing module is used to convert the first response data measured by the Pirani sensor into a two-dimensional image using the Gram angle field, and process the first response data using a piecewise interpolation method to obtain target response data; The calibration fusion module is used to determine the vacuum degree of the vacuum environment according to the first response data and the second response data measured by the absolute pressure sensor; The gas identification module is used to select a target method to identify the gas in the vacuum environment based on the vacuum degree to obtain the type of gas in the vacuum environment; the target method includes a gas identification model or a difference method.

8. The gas identification device according to claim 7, characterized in that: The gas identification module comprises: A first gas identification unit is used to input the two-dimensional image into the gas identification model to perform gas identification to obtain the type of gas in the vacuum environment when the vacuum degree is within a first vacuum range; the first vacuum range is greater than 0 and less than 1000 Pa; The second gas identification unit is used to calculate the difference between the target response data and the second response data when the vacuum degree is within a second vacuum range, and determine the gas type corresponding to the numerical range in the mapping relationship table where the difference is located as the gas type in the vacuum range; the second vacuum range is greater than or equal to 1000 Pa and less than atmospheric pressure; the mapping relationship table is a correspondence table between gas types and numerical ranges.

9. The gas identification device according to claim 7, characterized in that: The gas identification device further comprises a conditioning chip, wherein the conditioning chip is communicatively connected with the absolute pressure sensor, and the conditioning chip is used for converting the electrical signal of the absolute pressure sensor into a digital signal.

10. The gas identification device according to claim 9, characterized in that: The gas identification device also includes: a first inner shell and a second inner shell, the first inner shell and the second inner shell are both provided with air holes, the conditioning chip and the absolute pressure sensor are arranged in the first inner shell, the Pirani sensor is arranged in the second inner shell, and the first inner shell and the second inner shell are connected to the data processing module through pins.

Images