CN118392800B - Gas station inspection monitoring method and system based on smell identification - Google Patents
Gas station inspection monitoring method and system based on smell identification Download PDFInfo
- Publication number
- CN118392800B CN118392800B CN202410562826.0A CN202410562826A CN118392800B CN 118392800 B CN118392800 B CN 118392800B CN 202410562826 A CN202410562826 A CN 202410562826A CN 118392800 B CN118392800 B CN 118392800B
- Authority
- CN
- China
- Prior art keywords
- model
- data
- odor
- combustible
- toxic gas
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000012544 monitoring process Methods 0.000 title claims abstract description 94
- 238000000034 method Methods 0.000 title claims abstract description 73
- 238000007689 inspection Methods 0.000 title claims abstract description 21
- 239000002341 toxic gas Substances 0.000 claims abstract description 167
- 230000007613 environmental effect Effects 0.000 claims abstract description 102
- 238000005070 sampling Methods 0.000 claims abstract description 78
- 239000007789 gas Substances 0.000 claims abstract description 50
- 238000004458 analytical method Methods 0.000 claims abstract description 36
- 238000012549 training Methods 0.000 claims abstract description 15
- 238000009826 distribution Methods 0.000 claims description 44
- 239000013598 vector Substances 0.000 claims description 42
- 230000004927 fusion Effects 0.000 claims description 35
- 230000007246 mechanism Effects 0.000 claims description 32
- 238000012545 processing Methods 0.000 claims description 24
- 238000009792 diffusion process Methods 0.000 claims description 20
- 238000005457 optimization Methods 0.000 claims description 17
- 238000005516 engineering process Methods 0.000 claims description 13
- 230000010354 integration Effects 0.000 claims description 12
- 238000000513 principal component analysis Methods 0.000 claims description 12
- 238000000605 extraction Methods 0.000 claims description 10
- 238000003860 storage Methods 0.000 claims description 9
- 238000013507 mapping Methods 0.000 claims description 8
- 238000010183 spectrum analysis Methods 0.000 claims description 8
- 238000004088 simulation Methods 0.000 claims description 6
- 238000010586 diagram Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 4
- 239000000446 fuel Substances 0.000 claims description 2
- 238000013135 deep learning Methods 0.000 abstract description 6
- 235000019645 odor Nutrition 0.000 description 100
- 230000008569 process Effects 0.000 description 19
- UHOVQNZJYSORNB-UHFFFAOYSA-N Benzene Chemical compound C1=CC=CC=C1 UHOVQNZJYSORNB-UHFFFAOYSA-N 0.000 description 15
- 230000008859 change Effects 0.000 description 6
- 230000004044 response Effects 0.000 description 6
- 230000008901 benefit Effects 0.000 description 4
- 238000007405 data analysis Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 4
- 230000003595 spectral effect Effects 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 239000003344 environmental pollutant Substances 0.000 description 3
- 238000003912 environmental pollution Methods 0.000 description 3
- 238000007726 management method Methods 0.000 description 3
- 231100000719 pollutant Toxicity 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000012732 spatial analysis Methods 0.000 description 2
- 239000012855 volatile organic compound Substances 0.000 description 2
- CTQNGGLPUBDAKN-UHFFFAOYSA-N O-Xylene Chemical compound CC1=CC=CC=C1C CTQNGGLPUBDAKN-UHFFFAOYSA-N 0.000 description 1
- 238000010521 absorption reaction Methods 0.000 description 1
- 238000003915 air pollution Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000004445 quantitative analysis Methods 0.000 description 1
- 230000001932 seasonal effect Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000004611 spectroscopical analysis Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000008096 xylene Substances 0.000 description 1
Abstract
The application relates to the technical field of deep learning, and discloses a gas station inspection monitoring method and system based on smell identification. The method comprises the following steps: detecting combustible and toxic gases and extracting environmental factors from the plurality of air sample data to obtain a plurality of first smell classification labels and first environmental factor data; model training is carried out through a memristor model and a transducer model, and an odor identification model is obtained; air and environment monitoring is carried out on a plurality of smell sampling points of a target gas station, so that air monitoring data and second environment factor data are obtained; carrying out combustible and toxic gas identification and content analysis through an odor identification model to obtain a combustible and toxic gas identification result and combustible and toxic gas content data of each odor sampling point; the application improves the recognition accuracy of the combustible and toxic gas smell of the gas station and further improves the inspection and anomaly monitoring accuracy of the gas station.
Description
Technical Field
The application relates to the technical field of deep learning, in particular to a gas station inspection monitoring method and system based on smell identification.
Background
Volatile organic compounds (combustible and toxic gases) are common harmful gases in gas stations, which not only pose a threat to human health, but also can cause air pollution and fire explosion accidents. Therefore, the method has important significance in effectively monitoring combustible and toxic gases and early warning leakage of the gas station.
However, existing methods for monitoring flammable and toxic gases rely mainly on conventional gas sensors, which, while providing real-time monitoring in some cases, are often limited by the sensitivity of the sensor, the monitoring range and environmental factors, and it is difficult to achieve accurate identification and quantitative analysis of flammable and toxic gases in a complex environment at a gas station.
Disclosure of Invention
The application provides a gas station inspection monitoring method and system based on smell identification, the application improves the recognition accuracy of the odor of the combustible and toxic gases in the gas station and further improves the inspection and anomaly monitoring accuracy of the gas station.
In a first aspect, the present application provides a gas station inspection and monitoring method based on smell identification, where the gas station inspection and monitoring method based on smell identification includes:
detecting combustible and toxic gases and extracting environmental factors from the plurality of air sample data to obtain a plurality of first odor classification labels and first environmental factor data of the combustible and toxic gases in each air sample data;
Inputting the air sample data, the first smell classification tags and the first environmental factor data into a preset memristor model and a transducer model for model training to obtain a smell recognition model;
air and environment monitoring is carried out on a plurality of smell sampling points of a target gas station, so that air monitoring data and second environment factor data of each smell sampling point are obtained;
Inputting the air monitoring data and the second environmental factor data into the odor identification model to identify and analyze the combustible and toxic gas content, and obtaining the combustible and toxic gas identification result and the combustible and toxic gas content data of each odor sampling point;
And carrying out content anomaly analysis and leakage source positioning on the combustible and toxic gas identification result and the combustible and toxic gas content data of each odor sampling point by adopting a triangulation method to obtain leakage source positioning information.
In a second aspect, the present application provides a gas station inspection monitoring system based on smell identification, the gas station inspection monitoring system based on smell identification includes:
The extraction module is used for detecting combustible and toxic gases and extracting environmental factors from the plurality of air sample data to obtain a plurality of first odor classification labels and first environmental factor data of the combustible and toxic gases in each air sample data;
the training module is used for inputting the air sample data, the first odor classification labels and the first environmental factor data into a preset memristor model and a transducer model for model training to obtain an odor identification model;
the monitoring module is used for carrying out air and environment monitoring on a plurality of smell sampling points of the target gas station to obtain air monitoring data and second environment factor data of each smell sampling point;
The analysis module is used for inputting the air monitoring data and the second environmental factor data into the odor identification model to identify flammable and toxic gases and analyze the flammable and toxic gas content, so as to obtain the flammable and toxic gas identification result and the flammable and toxic gas content data of each odor sampling point;
And the positioning module is used for carrying out content anomaly analysis and leakage source positioning on the combustible and toxic gas identification result and the combustible and toxic gas content data of each smell sampling point by adopting a triangulation method to obtain leakage source positioning information.
A third aspect of the present application provides a computer apparatus comprising: a memory and at least one processor, the memory having instructions stored therein; the at least one processor invokes the instructions in the memory to cause the computer device to perform the scent-identification-based fuel station patrol monitoring method described above.
A fourth aspect of the present application provides a computer readable storage medium having instructions stored therein which, when run on a computer, cause the computer to perform the above-described scent-based gas station inspection monitoring method.
According to the technical scheme provided by the application, the high-precision characteristic frequency identification and frequency domain characteristic extraction of the combustible and toxic gases are realized by utilizing the multi-frequency spectrum analysis technology and combining with Fourier transformation, so that the accuracy of the type and concentration detection of the combustible and toxic gases is greatly improved. By collecting environmental factors such as temperature, humidity and wind speed and performing feature mapping and principal component analysis, the influence of environmental variables on combustible and toxic gas monitoring data is effectively considered, and the comprehensiveness and reliability of data analysis are enhanced. The mixed network structure design integrating the memristor model and the transducer model not only plays the advantages of the memristor network in the aspect of processing time series data, but also utilizes the strong capability of the transducer network in the aspect of processing complex data, and realizes the deep learning and comprehensive analysis of odor data characteristics. Through the dynamic learning characteristic of the memristor model, the quick response and self-adaptive adjustment to the small change of the monitoring data are realized, and the monitoring system is ensured to be kept efficient and sensitive in a continuously-changing environment. The accurate positioning of the leakage source is carried out by combining the triangulation method with the combustible and toxic gas content data and the environmental factors, so that scientific basis is provided for quick identification and emergency treatment of the leakage source, and the environmental pollution and potential safety risk are greatly reduced. By means of abnormal analysis of the combustible and toxic gas content and timely generation of leakage source positioning information, early warning of potential leakage events is achieved, and powerful technical support is provided for timely risk management and emergency response of gas stations.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings may be obtained based on these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic diagram of an embodiment of a gas station inspection monitoring method based on smell recognition in an embodiment of the present application;
FIG. 2 is a schematic diagram of an embodiment of a gas station inspection monitoring system based on smell recognition in an embodiment of the present application.
Detailed Description
The embodiment of the application provides a gas station inspection monitoring method and system based on smell identification. The terms "first," "second," "third," "fourth" and the like in the description and in the claims and in the above drawings, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments described herein may be implemented in other sequences than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus.
For easy understanding, the following describes a specific flow of an embodiment of the present application, referring to fig. 1, and an embodiment of a gas station inspection monitoring method based on smell recognition in the embodiment of the present application includes:
Step 101, detecting combustible and toxic gases and extracting environmental factors from a plurality of air sample data to obtain a plurality of first odor classification labels and first environmental factor data of the combustible and toxic gases in each air sample data;
It can be appreciated that the implementation subject of the present application may be a gas station inspection monitoring system based on smell recognition, and may also be a terminal or a server, which is not limited herein. The embodiment of the application is described by taking a server as an execution main body as an example.
Specifically, a plurality of air sample data are acquired to capture the air conditions of the surrounding environment of the gas station. And (3) carrying out characteristic frequency identification on volatile organic compounds (combustible and toxic gases) in the air sample by adopting a multi-frequency spectrum analysis technology, separating various combustible and toxic gases in the air sample, and obtaining corresponding characteristic information of the combustible and toxic gases by generating a characteristic frequency set of data of each air sample. And carrying out Fourier transform processing on the characteristic frequency set of each air sample data, and converting the time sequence data into frequency domain characteristic data, so that the subsequent analysis and recognition are facilitated. The obtained frequency domain characteristic data is used for identifying the types of the combustible and toxic gases, the types of the combustible and toxic gases contained in each air sample are determined by analyzing the relation between the frequency domain characteristic data and the characteristic frequencies of the known combustible and toxic gases, and a plurality of first odor classification labels of each air sample data are generated. And extracting environmental factors including parameters such as temperature, humidity and wind speed from each air sample data, wherein the environmental factors are helpful for understanding the behavior patterns of combustible and toxic gases in the air sample data. And the environmental characteristics of each air sample data are obtained by carrying out characteristic mapping on the environmental parameters, so that the subsequent data analysis and model training are facilitated. And extracting main influencing factors from the multi-environment factors by carrying out Principal Component Analysis (PCA) on the environmental characteristics of each air sample data, so as to obtain first environmental factor data of each air sample data. The principal component analysis is a common data dimension reduction technology, and can extract the most representative factors from complex environmental characteristics, thereby simplifying the subsequent analysis process.
102, Inputting a plurality of air sample data, a plurality of first smell classification labels and first environmental factor data into a preset memristor model and a transducer model for model training to obtain a smell recognition model;
Specifically, a hybrid network construction of a preset memristor model and a transducer model is realized by creating a fusion layer. The characteristics and learning mechanisms of the two models are mutually complemented, and the advantage of the memristor model in terms of processing time sequence data is combined with the capability of the transducer model in terms of processing sequence through a fusion layer, so that an initial hybrid network architecture is obtained. Model parameter initialization is performed on the initial hybrid network architecture, including configuring proper memristor units and initial weights for a memristor model, and setting parameters of proper network layer numbers, head numbers, hidden layer dimensions and self-attention mechanisms for a transducer model. The method comprises the steps of inputting a plurality of air sample data and first environmental factor data into an initial hybrid network architecture for odor identification, and generating first odor molecule fingerprint prediction data of each air sample data by a model through processing the input data, wherein the data reflects understanding and prediction of the odor characteristics of combustible and toxic gases in each sample by the model. In order to optimize the model performance, a loss value between the model predicted odor molecule fingerprint and the actual first odor classification label is calculated, and the calculation result of the loss value directly influences the adjustment direction and the adjustment amplitude of the model parameters. Based on feedback of the loss value, the model is refined by updating the weight of the initial hybrid network architecture, so that the model is more close to the actual odor recognition requirement, continuous iteration is provided, prediction errors are reduced, and the recognition accuracy of the model is improved. Through a model integration technology, the target hybrid network architecture subjected to repeated iterative optimization is integrated into an odor identification model, and the model can accurately identify the odor characteristics of combustible and toxic gases in the gas station environment and effectively analyze the odor characteristics.
And updating the weight of the memristor model in the initial hybrid network architecture according to the loss value. Based on a loss value generated in the training process, parameters in the memristor model are adjusted to reduce a gap between a prediction result and an actual value, and a first model weight is obtained. And further updating the model parameters of the initial hybrid network architecture by using the updated first model weight to obtain the first hybrid network architecture. The method comprises the steps of adjusting a memristor model and optimizing the whole hybrid network structure, so that the model can accurately perform odor identification on the whole. And fixing the first model weight of the memristor model, and calculating the weight of the transducer model according to the loss value, wherein the transducer model can process sequence data, and particularly, the transducer model is in the aspects of understanding and processing long-distance dependency. And performing model optimization on the first hybrid network architecture through the second model weight to obtain a second hybrid network architecture. Through adjusting the parameters of the transducer model, the transducer model can work with the memristor model better, and the accuracy and efficiency of odor identification are improved. And fixing the weight of a second model of the transducer model, calculating the characteristic distribution parameters of the fusion layer according to the loss value, optimizing the characteristic distribution in the fusion layer, and ensuring that the outputs of different models can be combined in the most effective mode to obtain the final target hybrid network architecture. And carrying out multi-round double-layer iterative optimization and model integration on the target hybrid network architecture, and combining a plurality of optimized models into a single odor identification model. The iterative optimization and model integration method enhances the generalization capability of the model and also improves the accuracy and reliability of smell identification.
Step 103, air and environment monitoring is carried out on a plurality of smell sampling points of a target gas station, and air monitoring data and second environment factor data of each smell sampling point are obtained;
Specifically, environmental assessment of the gas station, including identification of possible sources of combustible and toxic gas emissions, odor propagation paths, and ranges of variation in environmental factors, ensures that the selected sampling points are capable of comprehensively representing the air quality of the gas station and capturing changes in critical environmental factors. And positioning and arranging the air quality monitoring equipment and the environmental parameter monitoring instrument with high precision on the sampling points. These devices are capable of continuously monitoring data on combustible and toxic gas content, particulate size, temperature, humidity, and wind speed in air. By setting the monitoring time period, the odor sampling points are continuously monitored at different times and under different environmental conditions. Including daily air quality monitoring, also taking into account seasonal changes, air and environmental monitoring under special climatic conditions, and air quality changes during special operation of the gas station. And sending the monitoring data to a data processing center in real time through a data transmission technology. And analyzing the air monitoring data and the environmental factor data through data analysis software and algorithms, including cleaning, correcting, statistical analysis and the like of the data, identifying the types and the concentrations of main pollutants in the monitoring area, and analyzing how environmental factors influence the distribution and the change of the pollutants to obtain the air monitoring data and the second environmental factor data of each odor sampling point.
104, Inputting the air monitoring data and the second environmental factor data into an odor identification model for identifying and analyzing the combustible and toxic gas content, and obtaining the combustible and toxic gas identification result and the combustible and toxic gas content data of each odor sampling point;
Specifically, the air monitoring data and the second environmental factor data are input into an odor identification model, which is composed of a memristor model, a fusion layer and a transducer model. Air monitoring data are processed through the memristor model, characteristic vectors of odors are extracted through time sequence data processing capacity of the air monitoring data, and time changes and modes in odor characteristics of flammable and toxic gases are captured. And performing feature weight distribution on the smell feature vector through the fusion layer to obtain a target feature vector, and ensuring that the model can be adjusted according to the importance of different features to obtain an optimized target feature vector. And inputting the target feature vector and the second environmental factor data into a transducer model for processing. The transducer model analyzes the comprehensive data through its self-attention mechanism to realize the prediction of the odor-classification labels, resulting in a plurality of second odor-classification labels. These class labels represent the results of the model's identification of different flammable and toxic gas species. Based on the second odor classification label generated by the model, the model performs the identification and content analysis of combustible and toxic gases in combination with the original air monitoring data and the second environmental factor data. Including the identification of the combustible and toxic gas species and calculating the combustible and toxic gas content data for each odor sampling point.
The target feature vector and the second environmental factor data are subjected to attention mechanism analysis through a self-attention mechanism layer in the transducer model. The self-attention mechanism layer can automatically distinguish and give different importance weights to different data parts, and generates attention mechanism vectors by calculating the correlation inside the data. These vectors are essentially a weighted representation of features that can highlight the most relevant information for flammable and toxic gas identification and content analysis, while suppressing those less relevant information portions. And carrying out odor molecular fingerprint prediction on the attention mechanism vector through a decision layer in a transducer model. And outputting corresponding second odor molecule fingerprint prediction data according to the input weighted feature vector through a network structure in deep learning. These predictive data represent an understanding of the model for the odor characteristics of combustible and toxic gases in the air sample at each sampling point. And performing odor classification tag matching on the second odor molecule fingerprint prediction data. And matching each prediction result to a corresponding odor classification label by comparing the prediction data with a known odor molecular fingerprint database to obtain a plurality of second odor classification labels.
And 105, carrying out content anomaly analysis and leakage source positioning on the combustible and toxic gas identification result and the combustible and toxic gas content data of each smell sampling point by adopting a triangulation method to obtain leakage source positioning information.
Specifically, by setting the safety threshold of the combustible and toxic gas content, the combustible and toxic gas identification results and the combustible and toxic gas content data of all the odor sampling points in the gas station are subjected to content anomaly analysis, and sampling points possibly having leakage risks are screened out according to preset safety standards. And classifying the distribution positions of the screened target sampling points, and systematically arranging the spatial positions of the points to generate a graph comprehensively reflecting the spatial distribution of the target sampling points. And calculating to obtain the geometric center of the spatial distribution of the target sampling points by using a spatial analysis method, and determining the central position of the geometric shape formed by all the target sampling points, wherein the central position can approximately reflect the position of the potential leakage source. Based on the space distribution geometric center, the flammable and toxic gas diffusion path simulation is carried out on the target sampling points by adopting a triangulation method, and the simulated diffusion data are obtained. The propagation paths of the combustible and toxic gases are approximately reconstructed by simulating the process of diffusing the combustible and toxic gases from the potential leakage sources to each sampling point and considering the influence of environmental factors such as wind speed, wind direction and the like. And comparing the simulated diffusion data with actual combustible and toxic gas content data, and positioning a leakage source by analyzing the matching degree of the simulated diffusion data and the actual combustible and toxic gas content data to obtain leakage source positioning information.
In the embodiment of the application, the high-precision characteristic frequency identification and the frequency domain characteristic extraction of the combustible and toxic gases are realized by utilizing the multi-frequency spectrum analysis technology and combining with Fourier transformation, and the accuracy of the type and concentration detection of the combustible and toxic gases is greatly improved. By collecting environmental factors such as temperature, humidity and wind speed and performing feature mapping and principal component analysis, the influence of environmental variables on combustible and toxic gas monitoring data is effectively considered, and the comprehensiveness and reliability of data analysis are enhanced. The mixed network structure design integrating the memristor model and the transducer model not only plays the advantages of the memristor network in the aspect of processing time series data, but also utilizes the strong capability of the transducer network in the aspect of processing complex data, and realizes the deep learning and comprehensive analysis of odor data characteristics. Through the dynamic learning characteristic of the memristor model, the quick response and self-adaptive adjustment to the small change of the monitoring data are realized, and the monitoring system is ensured to be kept efficient and sensitive in a continuously-changing environment. The accurate positioning of the leakage source is carried out by combining the triangulation method with the combustible and toxic gas content data and the environmental factors, so that scientific basis is provided for quick identification and emergency treatment of the leakage source, and the environmental pollution and potential safety risk are greatly reduced. By means of abnormal analysis of the combustible and toxic gas content and timely generation of leakage source positioning information, early warning of potential leakage events is achieved, and powerful technical support is provided for timely risk management and emergency response of gas stations.
In a specific embodiment, the process of executing step 101 may specifically include the following steps:
(1) Acquiring a plurality of air sample data, and carrying out characteristic frequency identification on combustible and toxic gases in the plurality of air sample data by adopting a multi-frequency spectrum analysis technology to generate a characteristic frequency set of each air sample data;
(2) Carrying out Fourier transform processing on the characteristic frequency set of each air sample data to obtain frequency domain characteristic data of each air sample data;
(3) Carrying out combustible and toxic gas type identification on the frequency domain characteristic data of each air sample data to obtain a plurality of first odor classification labels of the combustible and toxic gas in each air sample data;
(4) Extracting environmental factors from the plurality of air sample data to obtain the temperature, humidity and wind speed corresponding to each air sample data;
(5) Performing feature mapping on the temperature, the humidity and the wind speed corresponding to each air sample data to obtain the environmental features of each air sample data;
(6) And carrying out principal component analysis on the environmental characteristics of each air sample data to obtain first environmental factor data of each air sample data.
In particular, a plurality of air sample data is obtained from a gas station environment. Representative and comprehensive data is obtained using air sampling devices, such as air sampling pumps and sampling tubes, to capture air samples at different times and locations. Processing by multi-frequency spectrum analysis technology. Multi-frequency spectroscopic analysis identifies combustible and toxic gases in an air sample by measuring the absorption and scattering properties of the sample under different frequency light illumination. Each combustible and toxic gas molecule has its unique spectral characteristics, enabling the system to accurately identify the combustible and toxic gas species present in the sample by this technique. And carrying out Fourier transform processing on the characteristic frequency set of each air sample data, converting the time series data into frequency domain data, and better analyzing the characteristic frequencies of combustible and toxic gases in the sample. Fourier transforms are a fundamental tool in signal processing that can reveal periodic features and frequency domain structures in sample data. And (3) identifying the types of combustible and toxic gases in each air sample by analyzing the frequency domain characteristic data, so as to obtain a first odor classification label of each sample. The known flammable and toxic gas spectral databases are used for matching, and the types of the flammable and toxic gases contained in the sample are accurately identified by comparing the frequency domain characteristics with the spectral characteristics in the databases. For example, if the frequency domain characteristics of a sample match the spectral characteristics of benzene in the database, then the sample is labeled as a benzene-containing sample. And extracting environmental factors from the collected air sample data, wherein the environmental factors comprise parameters such as temperature, humidity, wind speed and the like corresponding to each sample. These environmental parameters are important factors for understanding and analyzing the behavior of flammable and toxic gases, such as changes in temperature and humidity, which can affect the volatility and diffusion rate of the flammable and toxic gases. And performing feature mapping on environmental factors such as temperature, humidity, wind speed and the like of each sample, and converting the actual measured value into an environmental feature vector which can be processed by the model. The absolute values of the environmental parameters are converted to relative values or other forms of numerical representations by data preprocessing and encoding techniques to facilitate subsequent analysis. And carrying out Principal Component Analysis (PCA) on the environmental characteristics of each air sample, extracting the most influential environmental factors, and obtaining first environmental factor data of each sample. PCA is a commonly used data dimension reduction technique that reduces the complexity of the data by finding the dominant direction of change in the data, while retaining the most important information.
In a specific embodiment, the process of executing step 102 may specifically include the following steps:
(1) Creating a fusion layer, and carrying out mixed network construction on a preset memristor model and a transducer model through the fusion layer to obtain an initial mixed network architecture;
(2) Initializing model parameters of the initial hybrid network architecture to obtain configuration and initial weight of memristor units of a memristor model and parameter configuration of the number of layers, the number of heads, hidden layer dimensions and a self-attention mechanism of a transducer network of the transducer model;
(3) Inputting a plurality of air sample data and first environmental factor data into an initial hybrid network architecture for odor identification, and obtaining first odor molecule fingerprint prediction data of each air sample data;
(4) Calculating loss values between the plurality of first smell classification tags and the first smell molecular fingerprint prediction data;
(5) And updating the weight of the initial hybrid network architecture according to the loss value to obtain a target hybrid network architecture, and carrying out model integration on the target hybrid network architecture to obtain an odor identification model.
Specifically, a fusion layer is created, integrating information from the memristor model and the transducer model. Memristor models are good at processing time series data, can effectively capture time series features in air sample data, and transducer models process complex sequence data relationships through a self-attention mechanism. And constructing an initial hybrid network architecture by designing a fusion layer. Initializing model parameters of the initial hybrid network architecture, configuring proper parameters and initial weights for memristor units of the memristor model, and simultaneously setting proper parameters of network layers, head numbers, hidden layer dimensions and self-attention mechanisms for the transducer model. And inputting the collected plurality of air sample data and the corresponding first environmental factor data into a hybrid network architecture for odor identification processing. The hybrid network model uses its internal data processing and analysis capabilities to analyze each air sample data and predict the first odor molecule fingerprint data for each sample. To optimize the model performance, a loss value between the model predicted first odor molecular fingerprint data and the actual first odor class label is calculated. The loss value reflects the gap between the model prediction result and the actual result and is key feedback information in the model training process. And the weight and the parameters of the model are properly adjusted and optimized by analyzing the loss value, so that the prediction error is reduced, and the accuracy of the model is improved. And updating the weight of the initial hybrid network architecture according to the loss value to obtain the optimized target hybrid network architecture. And then, carrying out model integration on the target architecture, and integrating a plurality of model versions which are subjected to repeated iterative training and optimization into a unified odor identification model with better performance. The integration process can further improve the generalization capability and the prediction accuracy of the model, and can ensure that combustible and toxic gases in the gas station environment can be effectively identified and analyzed in practical application.
In a specific embodiment, the performing step updates the weight of the initial hybrid network architecture according to the loss value to obtain a target hybrid network architecture, and performs model integration on the target hybrid network architecture, so that the process of obtaining the odor identification model may specifically include the following steps:
(1) Carrying out weight updating on the memristor model in the initial hybrid network architecture according to the loss value to obtain a first model weight;
(2) Updating model parameters of the initial hybrid network architecture according to the first model weight to obtain a first hybrid network architecture;
(3) Fixing a first model weight of a memristor model in an initial hybrid network architecture, and carrying out weight calculation on a transducer model in the initial hybrid network architecture according to a loss value to obtain a second model weight;
(4) Performing model optimization on the first hybrid network architecture through the second model weight to obtain a second hybrid network architecture;
(5) Fixing a second model weight of a transducer model in a second hybrid network architecture, and calculating characteristic distribution parameters of the fusion layer according to the loss value to obtain the characteristic distribution parameters of the fusion layer;
(6) Performing fusion layer optimization on the second hybrid network architecture according to the characteristic allocation parameters to obtain a target hybrid network architecture;
(7) And carrying out multi-round double-layer iterative optimization and model integration on the target hybrid network architecture to obtain the odor identification model.
Specifically, the memristor model is updated in weight according to the loss value. And adjusting weight parameters of the memristor model according to the gradient obtained by calculating the loss function through a back propagation algorithm so as to reduce the difference between model output and a real label and obtain a first model weight. For example, if the model predicted odor classification does not match the actual classification when processing a particular air sample data, the loss function value will increase and the model learning corrects for this deviation by adjusting the weights. And carrying out parameter updating on the whole hybrid network architecture according to the updated first model weight to form the first hybrid network architecture, ensuring that the memristor model plays a better role in the whole architecture, and enhancing the processing capacity of the model on time series data. And fixing the weight of the memristor model in the first hybrid network architecture, and updating the weight of the transducer model according to the loss value. The weight adjustment of the transducer model is performed independently of the memristor model, focusing on the ability of the optimization model to handle complex relationships, resulting in a second model weight. For example, a transducer model may need to adjust parameters of its self-attention mechanism to more accurately identify subtle feature differences in the scent sample. And optimizing the first hybrid network architecture through the second model weight to obtain a second hybrid network architecture. And the odor recognition capability of the whole network architecture is enhanced by comprehensively considering the adjustment results of the memristor model and the transducer model. The weights of the transform model in the second hybrid network architecture are fixed, focusing on the optimization of the fusion layer. And calculating characteristic distribution parameters according to the loss values, wherein the parameters determine how the characteristics from the memristor model and the transducer model are combined in the fusion layer so as to improve the overall prediction accuracy of the model and form an optimized fusion layer. And optimizing the second hybrid network architecture based on the feature allocation parameters of the fusion layer to form a target hybrid network architecture. And carrying out multi-round double-layer iterative optimization and model integration on the target hybrid network architecture. The model continuously learns through new data, adjusts internal parameters, and simultaneously combines model versions which perform optimally in a plurality of training periods by adopting a model integration technology, so that the accuracy and the robustness of the odor identification model are improved.
In a specific embodiment, the process of executing step 104 may specifically include the following steps:
(1) Inputting the air monitoring data and the second environmental factor data into an odor identification model, the odor identification model comprising: memristor model, fusion layer, and transducer model;
(2) Performing odor characteristic extraction on the air monitoring data through the memristor model to obtain an odor characteristic vector;
(3) The characteristic weight distribution is carried out on the smell characteristic vector through the fusion layer, so that a target characteristic vector is obtained;
(4) Performing smell classification label prediction on the target feature vector and the second environmental factor data through a transducer model to obtain a plurality of second smell classification labels;
(5) Performing flammable and toxic gas identification through a plurality of second odor classification labels to obtain flammable and toxic gas identification results of each odor sampling point;
(6) And based on the combustible and toxic gas identification result, carrying out combustible and toxic gas content analysis on the air monitoring data and the second environmental factor data to obtain combustible and toxic gas content data of each odor sampling point.
Specifically, the air monitoring data and the second environmental factor data are input into an odor identification model, the odor identification model comprising: memristor model, fusion layer, and transducer model. Odor characteristics in the air monitoring data are extracted through the memristor model. The memristor model is adapted to process time series data, enabling capture of time-varying patterns in the data stream. For example, if the air-monitoring data shows a sudden flammable and toxic gas concentration rise at a particular point in time, the memristor model can recognize this change and translate it into a vector of odor signatures. And carrying out feature weight distribution on the smell feature vectors through the fusion layer, carrying out feature weight distribution on the vectors, and adjusting the weights of the smell feature vectors according to the importance of each feature on the smell recognition task to generate target feature vectors. And performing smell classification label prediction on the target feature vector and the second environmental factor data through a transducer model. The transducer model utilizes a self-attention mechanism to effectively process the feature vector and the environmental factor data to conduct odor classification label prediction. The model is capable of predicting the type of combustible and toxic gases present in each air sample, generating a plurality of second odor classification labels. For example, by analyzing specific odor characteristics and environmental factors, the transducer model may predict that a sample will contain benzene and xylene, both of which are common components of flammable and toxic gases. Combustible and toxic gas identification is performed by a plurality of second odor classification tags. And determining specific combustible and toxic gas components in each odor sampling point according to the predicted odor classification labels to obtain a combustible and toxic gas identification result of each sampling point, and indicating specific pollutant types in the environment of the gas station. And analyzing the combustible and toxic gas content of the air monitoring data and the second environmental factor data based on the recognition result of the combustible and toxic gas. Each identified combustible and toxic gas was quantitatively analyzed to determine its content in each odor sampling point.
In a specific embodiment, the performing step performs smell classification tag prediction on the target feature vector and the second environmental factor data by using a transducer model, and the process of obtaining the plurality of second smell classification tags may specifically include the following steps:
(1) Performing attention mechanism analysis on the target feature vector and the second environmental factor data through a self-attention mechanism layer in the transducer model to obtain an attention mechanism vector;
(2) Performing odor molecule fingerprint prediction on the attention mechanism vector through a decision layer in a transducer model to obtain second odor molecule fingerprint prediction data;
(3) And performing odor classification tag matching on the second odor molecular fingerprint prediction data to obtain a plurality of second odor classification tags.
Specifically, the attention mechanism analysis is performed on the target feature vector and the second environmental factor data through a self-attention mechanism layer in the transducer model. By calculating the correlation among the parts in the data, the most important characteristics of the current task are automatically identified, and the attention mechanism vector is obtained. And carrying out odor molecular fingerprint prediction on the attention mechanism vector through a decision layer in a transducer model. The decision layer makes predictions based on the processed data. The decision layer calculates based on the attention mechanism vector, predicts the specific type and characteristics of the combustible and toxic gases in each air sample, and generates second odor molecule fingerprint prediction data. The fingerprint features that best match the current sample are found in the odor molecule database, and each predicted odor molecule fingerprint represents one or more flammable and toxic gases that may be present in the sample. Based on the second odor molecule fingerprint prediction data, the model performs odor classification tag matching. Comparing the predicted odor molecule fingerprints with a predefined odor class database to determine the odor class for which each odor molecule fingerprint matches best. For example, if the predicted odor molecule fingerprint matches a fingerprint of benzene in the database, the sample will be labeled as an odor classification tag containing benzene.
In a specific embodiment, the process of executing step 105 may specifically include the following steps:
(1) Based on the combustible and toxic gas content safety threshold, carrying out content anomaly analysis on the combustible and toxic gas identification result and the combustible and toxic gas content data of each odor sampling point to obtain a plurality of target sampling points;
(2) Classifying the distribution positions of the plurality of target sampling points to obtain a spatial distribution diagram of the plurality of target sampling points;
(3) Calculating a spatial distribution geometric center of the spatial distribution map to obtain the spatial distribution geometric center;
(4) Based on a space distribution geometric center, performing flammable and toxic gas diffusion path simulation on a plurality of target sampling points by adopting a triangulation method to obtain simulated diffusion data;
(5) And comparing the simulated diffusion data with the combustible and toxic gas content data and positioning the leakage source to obtain the positioning information of the leakage source.
Specifically, a safety threshold of the combustible and toxic gas content is defined, content anomaly analysis is carried out on the combustible and toxic gas identification result and the combustible and toxic gas content data of each sampling point, and sampling points with the combustible and toxic gas content exceeding the safety threshold are identified. The air quality monitoring equipment is adopted to continuously collect the combustible and toxic gas contents in the air sample, and then the data are compared with the safety threshold value to determine which sampling points show abnormally high combustible and toxic gas contents. The distribution position of the target sampling points is classified, and a map reflecting the spatial distribution of the points is drawn by using a Geographic Information System (GIS) technology or other spatial analysis tools. The spatial profile provides the exact location of the target sampling points and also shows the relative relationship between these points. For example, these target sampling points may be clustered around a particular area of a gas station, implying possible pollution source locations. And analyzing the spatial distribution map to calculate the spatial distribution geometric centers of the target sampling points. Considering the coordinates of all target sampling points, a point representing the center of its overall spatial distribution is found, and the geometric center point is generally considered as a preliminary indication of a possible source of leakage. Based on the space distribution geometric center, the target sampling point is subjected to flammable and toxic gas diffusion path simulation by adopting a triangulation method. The path of out-diffusion of flammable and toxic gases from the geometric center is simulated taking into account wind direction, wind speed, and other environmental factors. Simulated diffusion data is generated by simulation software, which demonstrates the possible paths of flammable and toxic gases from predicted leakage sources to various target sampling points. And comparing and analyzing the simulated diffusion data with actual combustible and toxic gas content data, and accurately leaking the position of the source. The comparison considers the consistency between the diffusion mode and the actual monitoring data, and the specific position of the leakage source is determined by analyzing the matching degree of the diffusion mode and the actual monitoring data. For example, if the simulation data shows that flammable and toxic gases diffuse north from a certain tank in the tank area and the sampling point on the north side is abnormally high, it can be inferred that this tank is a source of leakage.
The method for inspecting and monitoring the gas station based on the smell identification in the embodiment of the present application is described above, and the system for inspecting and monitoring the gas station based on the smell identification in the embodiment of the present application is described below, referring to fig. 2, an embodiment of the system for inspecting and monitoring the gas station based on the smell identification in the embodiment of the present application includes:
The extraction module 201 is configured to perform combustible and toxic gas detection and environmental factor extraction on the plurality of air sample data, so as to obtain a plurality of first odor classification labels and first environmental factor data of the combustible and toxic gases in each air sample data;
The training module 202 is configured to input a plurality of air sample data, a plurality of first smell classification tags, and first environmental factor data into a preset memristor model and a transducer model for model training, so as to obtain a smell recognition model;
the monitoring module 203 is configured to perform air and environment monitoring on a plurality of smell sampling points of the target gas station, so as to obtain air monitoring data and second environmental factor data of each smell sampling point;
The analysis module 204 is configured to input the air monitoring data and the second environmental factor data into the odor identification model to perform flammable and toxic gas identification and flammable and toxic gas content analysis, so as to obtain a flammable and toxic gas identification result and flammable and toxic gas content data of each odor sampling point;
And the positioning module 205 is configured to perform content anomaly analysis and leakage source positioning on the flammable and toxic gas identification result and the flammable and toxic gas content data of each smell sampling point by using a triangulation method, so as to obtain leakage source positioning information.
Through the cooperation of the components, the high-precision characteristic frequency identification and frequency domain characteristic extraction of the combustible and toxic gases are realized by utilizing a multi-frequency spectrum analysis technology and combining with Fourier transformation, and the accuracy of the type and concentration detection of the combustible and toxic gases is greatly improved. By collecting environmental factors such as temperature, humidity and wind speed and performing feature mapping and principal component analysis, the influence of environmental variables on combustible and toxic gas monitoring data is effectively considered, and the comprehensiveness and reliability of data analysis are enhanced. The mixed network structure design integrating the memristor model and the transducer model not only plays the advantages of the memristor network in the aspect of processing time series data, but also utilizes the strong capability of the transducer network in the aspect of processing complex data, and realizes the deep learning and comprehensive analysis of odor data characteristics. Through the dynamic learning characteristic of the memristor model, the quick response and self-adaptive adjustment to the small change of the monitoring data are realized, and the monitoring system is ensured to be kept efficient and sensitive in a continuously-changing environment. The accurate positioning of the leakage source is carried out by combining the triangulation method with the combustible and toxic gas content data and the environmental factors, so that scientific basis is provided for quick identification and emergency treatment of the leakage source, and the environmental pollution and potential safety risk are greatly reduced. By means of abnormal analysis of the combustible and toxic gas content and timely generation of leakage source positioning information, early warning of potential leakage events is achieved, and powerful technical support is provided for timely risk management and emergency response of gas stations.
The application also provides a computer device, which comprises a memory and a processor, wherein the memory stores computer readable instructions, and the computer readable instructions, when executed by the processor, cause the processor to execute the steps of the gas station inspection monitoring method based on smell identification in the above embodiments.
The application also provides a computer readable storage medium, which can be a nonvolatile computer readable storage medium, and can also be a volatile computer readable storage medium, wherein instructions are stored in the computer readable storage medium, and when the instructions run on a computer, the instructions cause the computer to execute the steps of the gas station inspection monitoring method based on smell identification.
It will be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, systems and units may refer to the corresponding processes in the foregoing method embodiments, which are not repeated herein.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a read-only memory (ROM), a random access memory (random access memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
The above embodiments are only for illustrating the technical solution of the present application, and not for limiting the same; although the application has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present application.
Claims (4)
1. The gas station inspection monitoring method based on smell identification is characterized by comprising the following steps of:
Detecting combustible and toxic gases and extracting environmental factors from the plurality of air sample data to obtain a plurality of first odor classification labels and first environmental factor data of the combustible and toxic gases in each air sample data; the method specifically comprises the following steps: acquiring a plurality of air sample data, and carrying out characteristic frequency identification on combustible and toxic gases in the plurality of air sample data by adopting a multi-frequency spectrum analysis technology to generate a characteristic frequency set of each air sample data; carrying out Fourier transform processing on the characteristic frequency set of each air sample data to obtain frequency domain characteristic data of each air sample data; carrying out combustible and toxic gas type identification on the frequency domain characteristic data of each air sample data to obtain a plurality of first odor classification labels of the combustible and toxic gas in each air sample data; extracting environmental factors from the plurality of air sample data to obtain the temperature, humidity and wind speed corresponding to each air sample data; performing feature mapping on the temperature, the humidity and the wind speed corresponding to each air sample data to obtain the environmental features of each air sample data; carrying out principal component analysis on the environmental characteristics of each air sample data to obtain first environmental factor data of each air sample data;
Inputting the air sample data, the first smell classification tags and the first environmental factor data into a preset memristor model and a transducer model for model training to obtain a smell recognition model; the method specifically comprises the following steps: creating a fusion layer, and carrying out mixed network construction on a preset memristor model and a transducer model through the fusion layer to obtain an initial mixed network architecture; initializing model parameters of the initial hybrid network architecture to obtain configuration and initial weight of memristor units of the memristor model and parameter configuration of the layer number, the head number, the hidden layer dimension and the self-attention mechanism of a transducer network of the transducer model; inputting the plurality of air sample data and the first environmental factor data into the initial hybrid network architecture for odor identification to obtain first odor molecular fingerprint prediction data of each air sample data; calculating loss values between the plurality of first smell classification tags and the first smell molecular fingerprint prediction data; carrying out weight updating on the memristor model in the initial hybrid network architecture according to the loss value to obtain a first model weight; updating model parameters of the initial hybrid network architecture according to the first model weight to obtain a first hybrid network architecture; fixing a first model weight of a memristor model in the initial hybrid network architecture, and carrying out weight calculation on a transducer model in the initial hybrid network architecture according to the loss value to obtain a second model weight; performing model optimization on the first hybrid network architecture through the second model weight to obtain a second hybrid network architecture; fixing a second model weight of a transducer model in the second hybrid network architecture, and calculating the characteristic distribution parameters of the fusion layer according to the loss value to obtain the characteristic distribution parameters of the fusion layer; performing fusion layer optimization on the second hybrid network architecture according to the characteristic allocation parameters to obtain a target hybrid network architecture; performing multi-round double-layer iterative optimization and model integration on the target hybrid network architecture to obtain an odor identification model;
air and environment monitoring is carried out on a plurality of smell sampling points of a target gas station, so that air monitoring data and second environment factor data of each smell sampling point are obtained;
Inputting the air monitoring data and the second environmental factor data into the odor identification model to identify and analyze the combustible and toxic gas content, and obtaining the combustible and toxic gas identification result and the combustible and toxic gas content data of each odor sampling point; the method specifically comprises the following steps: inputting the air monitoring data and the second environmental factor data into the scent identification model, the scent identification model comprising: memristor model, fusion layer, and transducer model; performing odor characteristic extraction on the air monitoring data through the memristor model to obtain an odor characteristic vector; performing feature weight distribution on the smell feature vector through the fusion layer to obtain a target feature vector; performing attention mechanism analysis on the target feature vector and the second environmental factor data through a self-attention mechanism layer in the transducer model to obtain an attention mechanism vector; performing odor molecular fingerprint prediction on the attention mechanism vector through a decision layer in the transducer model to obtain second odor molecular fingerprint prediction data; performing odor classification tag matching on the second odor molecular fingerprint prediction data to obtain a plurality of second odor classification tags; performing flammable and toxic gas identification through the plurality of second odor classification labels to obtain flammable and toxic gas identification results of each odor sampling point; based on the combustible and toxic gas identification result, carrying out combustible and toxic gas content analysis on the air monitoring data and the second environmental factor data to obtain combustible and toxic gas content data of each odor sampling point;
Carrying out content anomaly analysis and leakage source positioning on the combustible and toxic gas identification result and the combustible and toxic gas content data of each odor sampling point by adopting a triangulation method to obtain leakage source positioning information; the method specifically comprises the following steps: based on the combustible and toxic gas content safety threshold, carrying out content anomaly analysis on the combustible and toxic gas identification result and the combustible and toxic gas content data of each odor sampling point to obtain a plurality of target sampling points; the distribution position classification is carried out on the plurality of target sampling points, so that a spatial distribution diagram of the plurality of target sampling points is obtained; calculating the spatial distribution geometric center of the spatial distribution map to obtain the spatial distribution geometric center; based on the space distribution geometric center, performing flammable and toxic gas diffusion path simulation on the plurality of target sampling points by adopting a triangulation method to obtain simulated diffusion data; and comparing the simulated diffusion data with the combustible and toxic gas content data and positioning a leakage source to obtain leakage source positioning information.
2. The utility model provides a monitoring system is patrolled and examined to filling station based on smell discernment which characterized in that, monitoring system is patrolled and examined to filling station based on smell discernment includes:
The extraction module is used for detecting combustible and toxic gases and extracting environmental factors from the plurality of air sample data to obtain a plurality of first odor classification labels and first environmental factor data of the combustible and toxic gases in each air sample data; the method specifically comprises the following steps: acquiring a plurality of air sample data, and carrying out characteristic frequency identification on combustible and toxic gases in the plurality of air sample data by adopting a multi-frequency spectrum analysis technology to generate a characteristic frequency set of each air sample data; carrying out Fourier transform processing on the characteristic frequency set of each air sample data to obtain frequency domain characteristic data of each air sample data; carrying out combustible and toxic gas type identification on the frequency domain characteristic data of each air sample data to obtain a plurality of first odor classification labels of the combustible and toxic gas in each air sample data; extracting environmental factors from the plurality of air sample data to obtain the temperature, humidity and wind speed corresponding to each air sample data; performing feature mapping on the temperature, the humidity and the wind speed corresponding to each air sample data to obtain the environmental features of each air sample data; carrying out principal component analysis on the environmental characteristics of each air sample data to obtain first environmental factor data of each air sample data;
The training module is used for inputting the air sample data, the first odor classification labels and the first environmental factor data into a preset memristor model and a transducer model for model training to obtain an odor identification model; the method specifically comprises the following steps: creating a fusion layer, and carrying out mixed network construction on a preset memristor model and a transducer model through the fusion layer to obtain an initial mixed network architecture; initializing model parameters of the initial hybrid network architecture to obtain configuration and initial weight of memristor units of the memristor model and parameter configuration of the layer number, the head number, the hidden layer dimension and the self-attention mechanism of a transducer network of the transducer model; inputting the plurality of air sample data and the first environmental factor data into the initial hybrid network architecture for odor identification to obtain first odor molecular fingerprint prediction data of each air sample data; calculating loss values between the plurality of first smell classification tags and the first smell molecular fingerprint prediction data; carrying out weight updating on the memristor model in the initial hybrid network architecture according to the loss value to obtain a first model weight; updating model parameters of the initial hybrid network architecture according to the first model weight to obtain a first hybrid network architecture; fixing a first model weight of a memristor model in the initial hybrid network architecture, and carrying out weight calculation on a transducer model in the initial hybrid network architecture according to the loss value to obtain a second model weight; performing model optimization on the first hybrid network architecture through the second model weight to obtain a second hybrid network architecture; fixing a second model weight of a transducer model in the second hybrid network architecture, and calculating the characteristic distribution parameters of the fusion layer according to the loss value to obtain the characteristic distribution parameters of the fusion layer; performing fusion layer optimization on the second hybrid network architecture according to the characteristic allocation parameters to obtain a target hybrid network architecture; performing multi-round double-layer iterative optimization and model integration on the target hybrid network architecture to obtain an odor identification model;
the monitoring module is used for carrying out air and environment monitoring on a plurality of smell sampling points of the target gas station to obtain air monitoring data and second environment factor data of each smell sampling point;
The analysis module is used for inputting the air monitoring data and the second environmental factor data into the odor identification model to identify flammable and toxic gases and analyze the flammable and toxic gas content, so as to obtain the flammable and toxic gas identification result and the flammable and toxic gas content data of each odor sampling point; the method specifically comprises the following steps: inputting the air monitoring data and the second environmental factor data into the scent identification model, the scent identification model comprising: memristor model, fusion layer, and transducer model; performing odor characteristic extraction on the air monitoring data through the memristor model to obtain an odor characteristic vector; performing feature weight distribution on the smell feature vector through the fusion layer to obtain a target feature vector; performing attention mechanism analysis on the target feature vector and the second environmental factor data through a self-attention mechanism layer in the transducer model to obtain an attention mechanism vector; performing odor molecular fingerprint prediction on the attention mechanism vector through a decision layer in the transducer model to obtain second odor molecular fingerprint prediction data; performing odor classification tag matching on the second odor molecular fingerprint prediction data to obtain a plurality of second odor classification tags; performing flammable and toxic gas identification through the plurality of second odor classification labels to obtain flammable and toxic gas identification results of each odor sampling point; based on the combustible and toxic gas identification result, carrying out combustible and toxic gas content analysis on the air monitoring data and the second environmental factor data to obtain combustible and toxic gas content data of each odor sampling point;
The positioning module is used for carrying out content anomaly analysis and leakage source positioning on the combustible and toxic gas identification result and the combustible and toxic gas content data of each odor sampling point by adopting a triangulation method to obtain leakage source positioning information; the method specifically comprises the following steps: based on the combustible and toxic gas content safety threshold, carrying out content anomaly analysis on the combustible and toxic gas identification result and the combustible and toxic gas content data of each odor sampling point to obtain a plurality of target sampling points; the distribution position classification is carried out on the plurality of target sampling points, so that a spatial distribution diagram of the plurality of target sampling points is obtained; calculating the spatial distribution geometric center of the spatial distribution map to obtain the spatial distribution geometric center; based on the space distribution geometric center, performing flammable and toxic gas diffusion path simulation on the plurality of target sampling points by adopting a triangulation method to obtain simulated diffusion data; and comparing the simulated diffusion data with the combustible and toxic gas content data and positioning a leakage source to obtain leakage source positioning information.
3. A computer device, the computer device comprising: a memory and at least one processor, the memory having instructions stored therein;
The at least one processor invokes the instructions in the memory to cause the computer device to perform the scent-based gas station patrol monitoring method of claim 1.
4. A computer readable storage medium having instructions stored thereon, wherein the instructions when executed by a processor implement the scent-based fuel station inspection and monitoring method of claim 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410562826.0A CN118392800B (en) | 2024-05-08 | Gas station inspection monitoring method and system based on smell identification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410562826.0A CN118392800B (en) | 2024-05-08 | Gas station inspection monitoring method and system based on smell identification |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118392800A CN118392800A (en) | 2024-07-26 |
| CN118392800B true CN118392800B (en) | 2024-11-12 |
Family
ID=
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112433028A (en) * | 2020-11-09 | 2021-03-02 | 西南大学 | Electronic nose gas classification method based on memristor cell neural network |
| CN117407707A (en) * | 2023-09-07 | 2024-01-16 | 清华大学 | Odor identification model training method and device |
| CN117846706A (en) * | 2024-02-18 | 2024-04-09 | 北京市市政一建设工程有限责任公司 | Monitoring and alarming method and control device for harmful gas of underground tunnel structure |
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112433028A (en) * | 2020-11-09 | 2021-03-02 | 西南大学 | Electronic nose gas classification method based on memristor cell neural network |
| CN117407707A (en) * | 2023-09-07 | 2024-01-16 | 清华大学 | Odor identification model training method and device |
| CN117846706A (en) * | 2024-02-18 | 2024-04-09 | 北京市市政一建设工程有限责任公司 | Monitoring and alarming method and control device for harmful gas of underground tunnel structure |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ansari et al. | A novel technique of mixed gas identification based on the group method of data handling (GMDH) on time-dependent MOX gas sensor data | |
| CN116187861A (en) | Isotope-based water quality traceability monitoring method and related device | |
| GB2561879A (en) | Spectroscopic analysis | |
| CN118392800B (en) | Gas station inspection monitoring method and system based on smell identification | |
| CN117538492B (en) | On-line detection method and system for pollutants in building space | |
| Rivera Martinez et al. | Reconstruction of high-frequency methane atmospheric concentration peaks from measurements using metal oxide low-cost sensors | |
| CN118392800A (en) | Gas station inspection monitoring method and system based on smell identification | |
| Mohammad et al. | Air Pollution Comparison RFM Model Using Machine Learning Approach | |
| Morsi | Electronic noses for monitoring environmental pollution and building regression model | |
| Bax et al. | Odour nuisance monitoring | |
| CN117521873A (en) | PM2.5 inversion method and system based on comprehensive data processing and random forest | |
| Talebi-Kalaleh et al. | Damage Detection in Bridge Structures through Compressed Sensing of Crowdsourced Smartphone Data | |
| CN115526330A (en) | Organic matter navigation data calibration method and device, computer equipment and storage medium | |
| Valencia et al. | Application of Random Forest in a Predictive Model of PM 10 Particles in Mexico City. | |
| Jian et al. | Applications and prospects of machine learning in turbulent combustion and engines | |
| Siddaraju et al. | Digital twin for image-based particle pollutant matter prognosis | |
| Galatro et al. | Data Analytics for Process Engineers: Prediction, Control and Optimization | |
| Rusanen et al. | A novel probabilistic source apportionment approach: Bayesian auto-correlated matrix factorization | |
| CN118376865B (en) | Rapid detection method and system for direct current characteristics of lightning arrester | |
| Okafor et al. | ProxySense: A novel approach for gas concentration estimation using Long Short-Term Memory Recurrent Neural Network (LSTM-RNN) | |
| AL AZIES | Air Pollution in Jakarta, Indonesia Under Spotlight: An AI-Assisted Semi-Supervised Learning Approach | |
| CN118690123A (en) | Volatile organic compound source analysis system based on PMF model | |
| Liu et al. | VALAR: Streamlining Alarm Ranking in Static Analysis with Value-Flow Assisted Active Learning | |
| Bose et al. | Environmental Data Analysis for Air Quality Monitoring and Control | |
| Parkavi et al. | LSTM Based Air Quality Prediction Model In Smart Cities |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |