Research On The Key Technology Of Artificial Olfactory Detection Of Mixed Gas

Gas detection is an important means to prevent various safety accidents.In daily life and industrial production,a large amount of flammable,explosive,toxic and harmful gases are often produced.It will bring great safety hazards to the health of people who live or work there for a long time if the leakage cannot be found in time and corresponding measures are taken.At the same time,it is likely to cause an explosion accident and bring catastrophic consequences if such gas accumulates to a certain concentration.Over the years,those security incidents have emerged one after another.Therefore,it has widely application value on how to realize the detection of leaking gas quickly and accurately.In this paper,a detection method using a combination of sensor arrays and pattern recognition algorithms is proposed.The main research contents can be illustrated as follows:(1)In view of the cross-sensitivity problem of metal oxide semiconductor gas sensors,four sensors with different sensitivities are selected to form a sensor array.At the same time,related matching circuits are designed and a data acquisition system is built.The data collection is completed by configuring different concentrations of methane,carbon monoxide,ethylene pure gas and mixed gas in the experiment,which provides reliable data source for subsequent research.(2)Focusing on the problem of base value interference caused by the inherent reasons of the sensor itself,the paper preprocesses the output response signal of the sensor array to eliminate the interference of base value.The feature extraction and analysis of the data are carried out by the Kernel Principal Component Analysis to realize the characteristic analysis of the mixed gas.(3)Qualitative identification of mixed gas is a key issue for artificial olfactory gas detection and analysis.Due to traditional methods such as Principal Component Analysis and Support Vector Machine require manual design of feature extraction functions,and the quality of feature extraction function design will directly affect the accuracy of subsequent classification.In order to improve the accuracy of artificial olfactory detection and recognition of mixed gas,a gas recognition method based on improved one-dimensional convolutional neural network is proposed in this paper.This method further strengthens the adaptive feature extraction ability of the convolutional neural network from the original data by improving the structure of the network model.The experimental results show that the recognition accuracy rate of the proposed method can reach 99.98%.Compared with the traditional algorithm,the proposed method has higher accuracy and model generalization ability.(4)On the basis of qualitative analysis,this paper uses convolutional neural network combined with Relevance Vector Machine to make quantitative estimation of the gas with the characteristics of it such as high sparsity,few parameters and fast convergence.The upper computer is programmed by Lab VIEW,and the gas response data is processed by using the trained network classification model.The results of qualitative and quantitative analysis are visualized on the upper computer interface.