Research Of Key Technology And Prototype Realization Of Air Detection Equipment In Industrial Park

Air monitoring and industrial park mainly through a series of sensitive membrane materials chemical gas sensor array to collecting data,through the electronic circuit will time series signal transmission to the artificial intelligence algorithm of complex software system,the feature selection method based on manual design such as random forests(RF),principal component analysis(PCA),by machine learning algorithms such as support vector machine(SVM),the characteristics of training,finally using the model of training to identify unknown category and concentration of gas.The sensitive membrane material of traditional gas detection equipment vibrates due to the adsorption of gas molecules and produces a complex time series signal.In the identification process,the degradation of sensitive membrane material,external environmental factors,and the type and concentration of gas all affect the generation of the signal,resulting in serious sensor drift.However,some subjective factors in feature extraction methods cannot be accurate and comprehensive.It is also found that different algorithms are directly related to the accuracy of gas detection results.This paper studies the technology from three key aspects of sensor transformation,feature extraction and gas detection algorithm,and proposes an improvement scheme for the traditional industrial gas detection system.The infrared optical gas detection sensor array device is designed in the scheme.It is a physical analytical instrument based on the electromagnetic wave characteristic absorption principle of different gases to the infrared band.For different types of industrial gases,it is only necessary to replace the optical filter module and adjust the program parameters.Using infrared optical gas detection sensor array equipment can improve the original characterization of collected gas sample data.Secondly introduced deep learning algorithm as gas recognition algorithm,and aimed at the problem of insufficient gas feature extraction methods,unsupervised learning in the process of deep learning can be used to automatically obtain the original gas data characteristics,the nature of industrial gas sample data accuracy reached 93.2%,and can give accurate gas density;Finally,it is verified that the deep learning algorithm can effectively compensate the gas data drift of the sensor,and the experimental results verify the effectiveness of the method.