Research And Application Of VOCs And Odors Detection System Based On Electronic Nose

Volatile organic compounds(VOCs)are ubiquitous compounds in the air,which pollute the air and harm human health.Odors is a microscopic mixture of molecules that can irritate and harm humans even at a low density.Detection of these ingredients is helpful to improve human life.Electronic nose is a gas analysis system that is inspired by the principle of human bionics to achieve VOCs and Odors detection effectively.The front-end detection sensor array resembles the human nasal cavity,and the back-end pattern recognition algorithm simulates the analysis of human brain.This study is aiming at constructing an improved nose system to detect the VOCs and Odors from industrial emissions.This paper first presents and summarizes what is VOCs and odors and the detection methods of the two substances,and then analyses recent decades domestic and abroad research about electronic nose detection system in recent decades.In order to make up for the current deficiencies in electronic nose research and achieve actual environmental testing accuracy.We propose an improved detection system.In this paper,based on the electronic nose system,the main achievements and implementation of our system are summarized as follows:1.In this paper,against to the problem existing in the current gas sampling,a system automatic injection structure design model is proposed and complete the core control part of the prototype.The automatic sample introduction structure uses a stepper motor to drive the slider to achieve positioning.A syringe is mounted above the slider.The DC motor drives the syringe by winding the building string to achieve the extraction of the gas in the sample bag in the corresponding position.Compared with the current manual micro injection method,it is highly efficient and saves labor costs.2.Completed the system's hardware circuit design,the preparation of the underlying driver,MATLAB-based GUI design and data set.3.The feature extraction methods such as average differential method,integral area under the curve,and maximum value were applied to the feature extraction of the system sensor array response curve,respectively,and used for principal component analysis(PCA)and linear discriminant analysis(LDA)dimension reduction,respectively.The comparative analysis shows that the feature extraction applying the maximum value to reduce the dimension of LDA is more distinguishable,In addition,the algorithm optimization of the sensor array was completed.Finally,we compare BP neural network?support vector machine(SVM)with particle swarm optimization optimized support vector machine(PSO-SVM).It is verified that the detection accuracy of improved PSO-SVM method is superior to other methods and reach up to 96.364%.4.The estimated density of e-nose gas using principal component regression(PCR)and SVM,extreme learning machine(ELM)and genetic algorithm optimization limit learning machine(GA-ELM)and the GA-ELM R2 value was obtained.Among the tested datasets,the value is all above 0.99 whereas while RMSE less than 0.002.Our optimum density prediction surpasses to other regression methods.5.According to the requirements of environmental monitoring,PID-200 photoion sensor and NH3/CR-200,CH2O/C-10 and H2S/C-50 electrochemical sensors are selected to build the electrochemical sensor gas chamber.We use response curve and curve fitting,the reproducibility curve three different ways to verify that the screened sensors are reasonable.In order to verify the feasibility of our proposed system,new sample set is collected from the ammonia gas,formaldehyde,and hydrogen sulfide gas,and the PCA is used to estimate the respective concentration level.The results show that our system distinguish the three gases densities clearly.