Multi-sensor fused intelligent gas detection method, terminal and readable storage medium

Abstract

The invention discloses a multi-sensor fused intelligent gas detection method, a terminal and a readable storage medium, the method comprises the following steps: acquiring multi-sensor array signal data, and processing the sensor array signal data to obtain model input data; inputting the model input data into an analysis model, wherein the analysis model extracts signal features of the model input data; and the analysis model identifies the type and concentration of the gas according to the signal characteristics. The spatial information and the time information of the mixed gas sensor dimension are acquired through the sensor array comprising the plurality of sensor units, and then the relationship between the spatial information and the time information and the type and the concentration of the mixed gas is learned by adopting the analysis model, so that the complex space-time characteristics of the mixed gas are considered; therefore, the detection effect on the mixed gas is far better than that in the prior art.

Description

Technical Field

[0001] The present invention relates to the field of gas detection technology, and in particular to a multi-sensor fusion intelligent gas detection method, terminal and readable storage medium. Background Art

[0002] Gas detection is mainly used in the fields of toxic and harmful gas monitoring, food quality and freshness inspection, disease breath detection, etc. Among them, for mixed gases composed of multiple gases, such as volatile organic compounds (VOCs) mixed gases, volatile organic compounds are a class of organic compounds with a boiling point between 50°C and 250°C and are easily volatile at room temperature. Due to its high volatility and the toxicity of some gases, it has caused potential harm to the ecological environment and human health. In recent years, its monitoring and governance issues have been highly valued.

[0003] VOCs mainly come from industrial emissions (such as printing, coating, surface treatment, etc.), solvent use (such as paint, detergent, printing ink, etc.), chemical production (such as plastics, rubber, fertilizers, etc.), daily necessities (such as furniture, carpets, detergents, etc.), as well as automobile exhaust, building materials, etc. In addition to the wide range of sources, the composition of VOCs is also complex and diverse, among which the common ones are alcohols: methanol, ethanol, propanol; aldehydes: formaldehyde, acetaldehyde; ketones: acetone; hydrocarbons: methane, ethane, ethylene, propylene, benzene, toluene, etc. The hazards of excessive VOCs are mainly in the following aspects: First, it is harmful to human health. Since most VOCs are irritating, high concentrations of VOCs are highly irritating to the eyes, nose and throat, which may cause red eyes, tears, coughing and other reactions. If exposed to VOCs for a long time, it may cause dizziness, headache and fatigue. VOCs are also harmful to human internal organs. For example, long-term exposure to toluene and xylene will affect the nervous system, liver and kidneys, and long-term exposure to carbon tetrachloride may also cause damage to the heart. More seriously, some VOCs have been proven to be carcinogenic to humans, such as benzene, formaldehyde, dichlorodiethyl sulfide, vinyl chloride, etc. Long-term exposure to such gases may increase the risk of cancer. The second is the pollution of the atmospheric environment. VOCs have a variety of hazards to air quality, mainly including the following aspects: First, VOCs are important precursors to the formation of substances such as ozone (O3) and fine particulate matter (PM2.5). Among them, ozone is a strong oxidant, and excessive concentration will affect air quality, while fine particulate matter is the culprit for reducing air visibility. Second, VOCs can undergo complex chemical reactions in the atmosphere to generate some toxic substances, such as carcinogens, organic acids and aldehyde compounds. Third, the photochemical reaction of VOCs with nitrogen oxides will also produce photochemical smog, which will lead to environmental air pollution.

[0004] The third is that it endangers production safety. Some VOCs are flammable and explosive, such as methane, ethylene, ethanol, toluene, acetone, etc. When their concentration in the air reaches a certain limit, they are prone to explosion and fire when encountering fire sources or high temperatures. In addition, some VOCs are corrosive and may damage production equipment and facilities, thus bringing safety risks to people's production and manufacturing.

[0005] In general, VOCs have diverse sources and rich types. Their concentration and types in the environment fluctuate with time and location, which makes the detection of VOCs extremely complex. The existing gas detection methods do not take into account the complex spatiotemporal characteristics of VOCs, resulting in poor results for VOCs detection. Summary of the invention

[0006] The purpose of the present invention is to provide a multi-sensor fusion intelligent gas detection method, terminal and readable storage medium, aiming to solve the problem that the existing technology has poor mixed gas detection effect.

[0007] The technical solution adopted by the present invention to solve the technical problem is as follows: The present invention provides a multi-sensor fusion intelligent gas detection method, the multi-sensor fusion intelligent gas detection method comprising: Acquiring multi-sensor array signal data, and processing the sensor array signal data to obtain model input data; Inputting the model input data into an analysis model, wherein the analysis model extracts signal features of the model input data; The analysis model identifies gas type and concentration based on the signal characteristics.

[0008] Furthermore, the acquiring of multi-sensor array signal data and processing of the sensor array signal data to obtain model input data specifically includes: Acquire signals of a plurality of sensor units, perform signal preprocessing on the signals of the plurality of sensor units, and align the signals of the respective sensor units to obtain the multi-sensor array signal data; The multi-sensor array signal data is subjected to denoising and standardization processing to obtain the model input data.

[0009] Further, the analysis model includes a feature extraction network, and the feature extraction network includes a spatial extraction branch, a temporal extraction branch and a gating unit; The analysis model extracts the signal features of the model input data, specifically including: The spatial extraction branch extracts sensor dimension features of the model input data; The time extraction branch extracts the time dimension feature of the model input data; The gating unit fuses the sensor dimension feature and the time dimension feature to obtain the signal feature.

[0010] Further, the spatial extraction branch sequentially includes a spatial encoding network and one or more spatial encoder modules, each of the spatial encoder modules includes a multi-head attention layer and a spatial feedforward network layer, wherein the spatial encoding network includes a spatial linear embedding layer; The spatial extraction branch extracts the sensor dimension features of the model input data, specifically including: The spatial linear embedding layer projects the model input data into a high-dimensional space to obtain spatial data, and inputs the spatial data into the spatial encoder module; The multi-head attention layer extracts features of the input; The spatial feedforward network enhances the features extracted by the multi-head attention layer of the spatial encoder module; The output of the last spatial encoder module is used as the sensor dimension feature.

[0011] Furthermore, the spatial encoder module further includes a residual connection and a normalization layer after the multi-head attention layer and a residual connection and a normalization layer after the spatial feedforward network layer; The residual connection and normalization layer are used to perform residual connection and regularization processing on the features extracted by the multi-head attention layer or the features enhanced by the spatial feedforward network.

[0012] Further, the time extraction branch sequentially includes a time encoding network and one or more time encoder modules, each of the time encoder modules includes a masked multi-head attention layer and a time feedforward network layer, wherein the time encoding network includes a time linear embedding layer and a position embedding layer; The time extraction branch extracts the sensor dimension features of the model input data, specifically including: The time linear embedding layer projects the model input data into a high-dimensional space to obtain time intermediate data; The position embedding layer performs time step encoding on the time intermediate data to obtain time data, and inputs the time data into the time encoder module; The masked multi-head attention layer extracts input features; The temporal feedforward network enhances features extracted in the masked multi-head attention layer of the temporal encoder module; The output of the last time encoder module is used as the sensor dimension feature.

[0013] Furthermore, the temporal encoder module further comprises a residual connection and a normalization layer after the masked multi-head attention layer and a residual connection and a normalization layer after the temporal feedforward network layer; The residual connection and normalization layer are used to perform residual connection and regularization processing on the features extracted by the masked multi-head attention layer or the features enhanced by the time feedforward network.

[0014] Further, the analysis model includes a classification network, and the classification network includes a linear layer and an output layer; The analysis model identifies the gas type and concentration according to the signal characteristics, specifically including: The linear layer performs linear transformation on the signal features to obtain linear transformation data; The output layer converts the linear transformation data into gas type and concentration and outputs it.

[0015] In addition, to achieve the above-mentioned purpose, the present invention also provides a terminal, which includes: a memory, a processor, and a multi-sensor fusion intelligent gas detection program stored in the memory and executable on the processor, and when the multi-sensor fusion intelligent gas detection program is executed by the processor, the terminal is controlled to implement the steps of the multi-sensor fusion intelligent gas detection method as described above.

[0016] In addition, to achieve the above-mentioned purpose, the present invention also provides a readable storage medium, which stores a multi-sensor fusion intelligent gas detection program, and when the multi-sensor fusion intelligent gas detection program is executed by a processor, the steps of the multi-sensor fusion intelligent gas detection method as described above are implemented.

[0017] The present invention adopts the above technical solution to achieve the following effects: The present invention collects spatial information of mixed gas sensor dimensions through a sensor array including multiple sensor units. At the same time, the signals of multiple time steps of the sensor array naturally have time information. Therefore, the sensor array signal data includes spatial information and time information of the mixed gas. Then, an analytical model is used to learn the relationship between this spatial information and time information and the type and concentration of the mixed gas, that is, to learn the characteristics that the concentration and type of the mixed gas in the environment will fluctuate with changes in time and place, thereby taking into account the complex spatiotemporal characteristics of the mixed gas to identify the type and concentration of the mixed gas. Its detection effect on the mixed gas is much better than that of the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flowchart of the steps of a dimensionality reduction representation method of a three-dimensional property right body in a preferred embodiment of the present invention; Figure 2 It is a structural flow chart of the analysis model in a preferred embodiment of the present invention; Figure 3 It is a schematic diagram of the operating environment of a preferred embodiment of the terminal of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention clearer and more specific, the present invention is further described in detail with reference to the accompanying drawings and examples. It should be understood that the described embodiments are only part of the embodiments of the present disclosure, rather than all the embodiments. The specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention. All other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present disclosure.

[0020] Embodiment 1 See also Figure 1 Embodiment 1 of the present invention is a multi-sensor fusion intelligent gas detection method, which can be applied to a chip integrated with a multi-sensor array to directly output the gas type and concentration obtained by analysis, or can be run on a terminal or server. The terminal or server collects signal data through a multi-sensor array, analyzes the signal data to obtain the gas type and concentration, and displays them to the user accordingly. The terminal includes a mobile device, a user terminal, a handheld device, a computing device, a wearable device, etc. The server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud storage, big data, and artificial intelligence platforms.

[0021] A multi-sensor fusion intelligent gas detection method of this embodiment includes the following steps: S1. Acquire multi-sensor array signal data, and process the sensor array signal data to obtain model input data.

[0022] Among them, the multi-sensor array includes multiple sensor units distributed at different positions, each sensor unit includes multiple chemical sensors, which can detect all types of gases in the mixed gas that needs to be detected. In this embodiment, each sensor unit includes sensors for multiple gas components such as formaldehyde, toluene and methanol.

[0023] The processing of the sensor array signal data specifically includes sequentially performing signal preprocessing, noise removal and standardization on the sensor array signal data, so as to obtain cleaner and more reliable data and lay a foundation for subsequent signal recognition.

[0024] Among them, signal preprocessing is to align the time points of all sensor units. It selects a reference sensor unit, finds the characteristic time point sequence of all sensor units, and aligns the characteristic time point sequence of the remaining sensor units with the characteristic time point sequence of the quasi-sensor unit.

[0025] The standardization process is specifically based on the Z-score data normalization method, which standardizes the data and eliminates the impact of different scales and dimensions.

[0026] Finding out the characteristic time point sequence of all sensor units specifically involves analyzing the original data sequence of the corresponding sensor unit in the sensor array signal data, and obtaining the rising and falling time points of the original data sequence as the characteristic time point sequence.

[0027] S2. Input the model input data into an analysis model, and the analysis model extracts signal features of the model input data.

[0028] Specifically, in this embodiment, the powerful feature extraction and modeling capabilities of the Transformer model are used to develop a targeted gas component identification analysis model to perform adaptive fusion analysis on the multi-dimensional feature data obtained by the chemical sensor array, thereby achieving rapid and accurate detection of VOCs and harmful gases.

[0029] It uses the python programming language for programming and running in the PyCharm integrated development environment, and uses the mainstream pytorch deep learning development framework to implement data analysis and processing of gas sensor signals under different categories and the construction of related classification models.

[0030] Specifically, this embodiment uses the Transformer-based dual-tower GTN (Gated Transformer Networks) gas classification and recognition model as the analysis model. The present invention plans to use deep learning technology to improve the accuracy of VOCs detection. Specifically, the present invention will start with the establishment of a database, use gas sensors to collect a large amount of VOCs data, and then build a deep learning model. This model will automatically learn and identify the characteristics and behavior patterns of VOCs by learning a large amount of VOCs data, thereby achieving accurate and rapid detection and differentiation of VOCs.

[0031] It is worth noting that in other optional embodiments, classic machine learning algorithms such as SVM and random forest can also be used to identify gas components in combination with feature data obtained by the gas sensor array, and rapid detection of common gases can be achieved through feature engineering and model optimization. The classic algorithm solution is relatively simple, but has high requirements for feature extraction and model design, and generally has poor recognition performance and adaptability.

[0032] Deep learning models such as convolutional neural networks (CNN) and recurrent neural networks (RNN) can also be used to directly learn the identification features of gas components from sensor data. Compared with the classic solutions of traditional machine learning, deep learning solutions have stronger feature extraction capabilities and adaptability, but this solution usually requires a large amount of labeled data for training, has high requirements for hardware resources, and lacks interpretability of the model. This makes it difficult to design and improve the overall model, and difficult to train and implement.

[0033] Therefore, in this embodiment, it is preferred to develop a targeted gas component identification analysis model based on the powerful feature extraction and modeling capabilities of the Transformer model as a feature extraction and classification solution.

[0034] Please refer to Figure 2 The analysis model of this embodiment includes a feature extraction network and a classification network. In this step, the feature extraction network is used.

[0035] The feature extraction network adopts a dual-tower structure, including a spatial extraction branch, a temporal extraction branch and a gating unit, wherein the spatial extraction branch and the temporal extraction branch are both encoder blocks of a Transformer model to encode the input data into signal features.

[0036] The spatial extraction branch includes a spatial encoding network and multiple spatial encoder modules in sequence. Each spatial encoder module includes a multi-head attention layer, a residual connection and normalization layer, a spatial feedforward network layer, and a residual connection and normalization layer. Among them, the spatial encoding network is used to project the model input data into a high-dimensional space so that the subsequent feature extraction network can better model the complex relationship in the input data. The spatial encoding network includes a spatial linear embedding layer (Embedding (Linear)), which is used to project the model input data into a high-dimensional space as spatial data for subsequent extraction of spatial connections between various sensor units.

[0037] Among them, the multi-head attention layer is used to calculate the relationship between different input channels. It maps the input spatial data to different subspaces, extracts different connections between the spatial data in the subspaces, and finally merges the information of all subspaces to find the correlation relationships at different angles in the overall sequence of spatial data.

[0038] The spatial feedforward network layer includes two linear transformations and the ReLU activation function between the linear transformations. The role of the feedforward network layer is to further process and transform the spatial features extracted by the multi-head attention layer to further extract and fuse feature information and enhance the nonlinear ability and expression ability of the model.

[0039] The residual connection and normalization layer are used to perform subsequent processing for each step in the Transformer module, which includes residual connection and regularized normalization. The purpose of residual connection is to help avoid the gradient vanishing problem during deep network training, and the purpose of regularized normalization is to speed up the training process and improve the stability of the model.

[0040] The spatial extraction branch finally outputs the sensor dimension features of the model input data. The sensor dimension features can model the correlation and differences between different sensors and utilize the complementarity of multi-sensor information. Therefore, the features in the sensor dimension help to better distinguish different types of gases.

[0041] It can be seen that since the spatial extraction branch only extracts the features of the sensor dimension and does not consider the features of the time dimension, the position encoding of the time step can be omitted during spatial encoding, reducing the overall delay and required resources of the model.

[0042] The temporal extraction branch sequentially includes a temporal encoding network, and multiple temporal encoder modules, each of which includes a masked multi-head attention layer, a residual connection and normalization layer, a temporal feedforward network layer, and a residual connection and regularization layer.

[0043] The temporal encoding network includes a temporal linear embedding layer and a positional embedding layer. The temporal linear embedding layer is also used to project the model input data into a high-dimensional space. The difference is that the model input data that has passed the temporal linear embedding layer also needs to pass through the positional embedding layer. The positional embedding layer embeds the position information of the sequence data in the sequence, thereby encoding the time steps to form temporal data, which is used to maintain the order of the sequence data so as to subsequently extract the temporal features of the model input data.

[0044] Masked Multi-Head Attention For time steps, it calculates the relationship between different time points, and for channels, it calculates the relationship between different input channels. Masked attention is used in the prediction part of the model to ensure that future data is not seen when generating predictions for the current time point, which helps maintain the causal relationship of the time series.

[0045] The function of the temporal feedforward network layer is similar to that of the spatial feedforward network layer, including two linear transformations and the ReLU activation function between the linear transformations, which is used to further process and transform the temporal features extracted by the masked multi-head attention layer to further extract and fuse feature information and enhance the nonlinear ability and expression ability of the model.

[0046] Residual connections and normalization layers are also used for subsequent processing of each step in the Transformer module, including residual connections and regularized normalization. The purpose of residual connections is also to help avoid the gradient vanishing problem during deep network training. The purpose of regularized normalization is also to speed up the training process and improve the stability of the model.

[0047] The time extraction branch finally outputs the time dimension features of the model input data. This can capture the characteristics of the signal changing over time, including fast-changing instantaneous features and long-term trend features. The features in the time dimension can better model the dynamic behavior of gas signals.

[0048] The time dimension features and sensor dimension features of the model input data are sent to the gate control unit together. The gate control unit is used to control the passage of feature information to fuse the sensor dimension features and the time dimension features to obtain signal features. The gate mechanism transmits more important information through fusion. It accepts the input of the downstream linear layer and transforms the input of the linear layer using a sigmoid activation function to obtain a set of values ​​between 0 and 1. These values ​​are called "gating values". The gating mechanism can learn how to adaptively weigh the importance of time dimension features and sensor dimension features, and dynamically adjust the contribution of the two dimensional features in the final classification results. These "gating values" are used to perform element-by-element multiplication operations with the sensor dimension features and time dimension features of the upstream input, so that it can selectively control which information can pass and which information should be blocked, and can control the importance of each information. Through this gating mechanism, the model can dynamically learn which information is most important for the current task. This fusion mechanism can make full use of the advantages of the two dimensional features, so as to perform more targeted feature extraction and information transmission, which can help the model better capture the long-term dependencies in time series data and improve the accuracy and robustness of prediction.

[0049] It can be seen that adaptive data fusion based on the Transformer model can effectively improve the accuracy and robustness of gas composition identification.

[0050] S3. The analysis model identifies the gas type and concentration according to the signal characteristics.

[0051] In this step, the classification network in the analysis model is used. The classification network includes a linear layer and an output layer.

[0052] Among them, the linear layer is to complete a linear transformation and map the input signal features to a new dimensional space. It performs this linear transformation through a learnable weight matrix and bias vector. This linear transformation can help the model learn the linear relationship and pattern of the input data.

[0053] The output layer is specifically an activation function, the Softmax function, which converts the input data into a probability distribution and normalizes the output value of the model so that the sum of the probability output by the model is 1, which can be interpreted as the probability of each category. The Softmax layer is used as the output layer of multi-classification tasks, which can help the model output the probability prediction of each category, thereby making more reliable classification decisions.

[0054] Embodiment 2 See also Figure 3Based on the above method, the present invention also provides a terminal, which includes: a memory 10, a processor 20, and a multi-sensor fusion intelligent gas detection program stored in the memory 10 and executable on the processor 20. When the multi-sensor fusion intelligent gas detection program is executed by the processor 20, the terminal is controlled to implement the steps of the multi-sensor fusion intelligent gas detection method as described above.

[0055] In some embodiments, the memory 10 may be an internal storage unit of the terminal, such as a hard disk or memory of the terminal. In other embodiments, the memory 10 may also be an external storage device of the terminal, such as a plug-in hard disk, a smart memory card (SmartMediaCard, SMC), a secure digital (SecureDigital, SD) card, a flash card (FlashCard), etc. equipped on the terminal. Furthermore, the memory 10 may also include both an internal storage unit of the terminal and an external storage device. The memory 10 is used to store application software and various types of data installed on the terminal, such as the program code of the installation terminal. The memory 10 can also be used to temporarily store data that has been output or is to be output. In one embodiment, the memory 10 stores a multi-sensor fusion intelligent gas detection program, which can be executed by the processor 20, thereby realizing the multi-sensor fusion intelligent gas detection method in the present application.

[0056] In some embodiments, the processor 20 may be a central processing unit (CPU), a microprocessor or other data processing chip, used to run the program code or process data stored in the memory 10, such as executing the multi-sensor fusion intelligent gas detection method.

[0057] Embodiment 3 This embodiment provides a storage medium, wherein the readable storage medium stores a multi-sensor fusion intelligent gas detection program, and when the multi-sensor fusion intelligent gas detection program is executed by a processor, the steps of the multi-sensor fusion intelligent gas detection method as described above are implemented.

[0058] In summary, the present invention collects spatial information of mixed gas sensor dimensions through a sensor array including multiple sensor units. At the same time, the signals of multiple time steps of the sensor array naturally have time information. Therefore, the sensor array signal data includes spatial information and time information of the mixed gas. Then, an analytical model is used to learn the relationship between this spatial information and time information and the type and concentration of the mixed gas, that is, to learn the characteristics that the concentration and type of the mixed gas in the environment will fluctuate with changes in time and place, thereby taking into account the complex spatiotemporal characteristics of the mixed gas to identify the type and concentration of the mixed gas. Its detection effect on the mixed gas is much better than the existing technology.

[0059] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or terminal including the element.

[0060] Of course, those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware (such as a processor, a controller, etc.) through a computer program, and the program can be stored in a computer-readable storage medium, and the program can include the processes of the above-mentioned method embodiments when executed. The storage medium can be a memory, a disk, an optical disk, etc.

[0061] It should be understood that the application of the present invention is not limited to the above examples. For ordinary technicians in this field, improvements or changes can be made based on the above description. All these improvements and changes should fall within the scope of protection of the claims attached to the present invention.

Claims

1. A multi-sensor fusion intelligent gas detection method, characterized in that: The multi-sensor fusion intelligent gas detection method comprises: Acquiring multi-sensor array signal data, and processing the sensor array signal data to obtain model input data; Inputting the model input data into an analysis model, wherein the analysis model extracts signal features of the model input data; The analysis model identifies gas type and concentration based on the signal characteristics.

2. The multi-sensor fusion intelligent gas detection method according to claim 1, characterized in that: The acquiring of multi-sensor array signal data and processing of the sensor array signal data to obtain model input data specifically includes: Acquire signals of a plurality of sensor units, perform signal preprocessing on the signals of the plurality of sensor units, and align the signals of the respective sensor units to obtain the multi-sensor array signal data; The multi-sensor array signal data is subjected to denoising and standardization processing to obtain the model input data.

3. The multi-sensor fusion intelligent gas detection method according to claim 1, characterized in that: The analysis model includes a feature extraction network, and the feature extraction network includes a spatial extraction branch, a temporal extraction branch and a gating unit; The analysis model extracts the signal features of the model input data, specifically including: The spatial extraction branch extracts sensor dimension features of the model input data; The time extraction branch extracts the time dimension feature of the model input data; The gating unit fuses the sensor dimension feature and the time dimension feature to obtain the signal feature.

4. The multi-sensor fusion intelligent gas detection method according to claim 3 is characterized in that: The spatial extraction branch sequentially includes a spatial encoding network and one or more spatial encoder modules, each of which includes a multi-head attention layer and a spatial feedforward network layer, wherein the spatial encoding network includes a spatial linear embedding layer; The spatial extraction branch extracts the sensor dimension features of the model input data, specifically including: The spatial linear embedding layer projects the model input data into a high-dimensional space to obtain spatial data, and inputs the spatial data into the spatial encoder module; The multi-head attention layer extracts input features; The spatial feedforward network enhances the features extracted by the multi-head attention layer of the spatial encoder module; The output of the last spatial encoder module is used as the sensor dimension feature.

5. The multi-sensor fusion intelligent gas detection method according to claim 4, characterized in that: The spatial encoder module also includes a residual connection and a normalization layer after the multi-head attention layer and a residual connection and a normalization layer after the spatial feedforward network layer; The residual connection and normalization layer are used to perform residual connection and regularization processing on the features extracted by the multi-head attention layer or the features enhanced by the spatial feedforward network.

6. The multi-sensor fusion intelligent gas detection method according to claim 3, characterized in that: The time extraction branch sequentially includes a time encoding network and one or more time encoder modules, each of which includes a masked multi-head attention layer and a time feedforward network layer, wherein the time encoding network includes a time linear embedding layer and a position embedding layer; The time extraction branch extracts the sensor dimension features of the model input data, specifically including: The time linear embedding layer projects the model input data into a high-dimensional space to obtain time intermediate data; The position embedding layer performs time step encoding on the time intermediate data to obtain time data, and inputs the time data into the time encoder module; The masked multi-head attention layer extracts input features; The temporal feedforward network enhances features extracted in the masked multi-head attention layer of the temporal encoder module; The output of the last time encoder module is used as the sensor dimension feature.

7. The multi-sensor fusion intelligent gas detection method according to claim 6, characterized in that: The temporal encoder module further includes a residual connection and a normalization layer after the masked multi-head attention layer and a residual connection and a normalization layer after the temporal feedforward network layer; The residual connection and normalization layer are used to perform residual connection and regularization processing on the features extracted by the masked multi-head attention layer or the features enhanced by the time feedforward network.

8. The multi-sensor fusion intelligent gas detection method according to claim 1, characterized in that: The analysis model includes a classification network, and the classification network includes a linear layer and an output layer; The analysis model identifies the gas type and concentration according to the signal characteristics, specifically including: The linear layer performs linear transformation on the signal features to obtain linear transformation data; The output layer converts the linear transformation data into gas type and concentration and outputs it.

9. A terminal, characterized in that: The terminal includes: a memory, a processor, and a multi-sensor fusion intelligent gas detection program stored in the memory and executable on the processor. When the multi-sensor fusion intelligent gas detection program is executed by the processor, the terminal is controlled to implement the steps of the multi-sensor fusion intelligent gas detection method as described in any one of claims 1 to 8.

10. A readable storage medium, characterized in that: The readable storage medium stores a multi-sensor fusion intelligent gas detection program, and when the multi-sensor fusion intelligent gas detection program is executed by a processor, the steps of the multi-sensor fusion intelligent gas detection method according to any one of claims 1 to 8 are implemented.

Images