Study On Quantitative Model Of Compound Odor Based On Multi-parameter And Multi-correlation

With the rapid development of China’s economy,the problem of environmental pollution has become more serious.As a special kind of environmental pollution,odor pollution has become one of the seven major public nuisances in the world,seriously affecting the ecological environment and human health.At present,the commonly used odor detection methods include olfactory test method and instrumental analysis method.The olfactory test method detects malodorous gas according to the olfactory sense of the olfactory person,and has the disadvantages of strong subjectivity,slow detection speed and poor realtime performance;the instrument analysis method needs special use.The equipment is tested,which is expensive,has poor real-time performance,and cannot simultaneously test the chemical composition,content and malodor of the complex malodor.Therefore,it is of great practical significance to develop a system that can quickly and accurately detect complex malodorous gases and realize online automatic measurement.Through the research on the existing odor detection methods at home and abroad,combined with the actual needs,a set of compound odor online detection system with electronic nose as the core was developed,and the gas path design,sensor array formation,odor signal collection and pattern recognition were completed.The design of the algorithm successfully achieved the quantitative detection of the complex malodorous gas.The odor detection results are affected by a variety of parameters,of which the effective signal,eigenvalue extraction and pattern recognition are the two most critical parts.Effective feature extraction can greatly improve the performance of the malodor detection system and effectively reduce the difficulty of pattern recognition.In this paper,feature extraction is performed by segmentation fitting response curve,and feature selection is carried out by combining principal component analysis(PCA)and linear discriminant analysis(LDA).The pattern recognition algorithm appropriately processes the information after feature extraction to obtain accurate malodor concentration information.The reasonable pattern recognition algorithm can improve the pattern recognition accuracy.This paper compares the two pattern recognition algorithms of partial least squares and BP neural network.The experimental results show that BP neural network has higher precision for the prediction of malodorous gas concentration,and PLS can’t establish a good regression prediction model for low concentration malodorous gas.In order to evaluate the impact of odor on the human environment,it is necessary to establish a response model for odor evaluation indicators and malodor concentrations.The factors affecting the odor evaluation index are not only related to the malodor concentration,but also related to the malodorous component.Therefore,qualitative analysis must be performed to obtain malodorous components before quantitative analysis.Two qualitative identification methods of support vector machine(SVM)and BP neural network are compared.The results show that the combination of LDA and SVM/BP network can identify malodorous gases 100%.