Study On Volatile Organic Compounds Monitoring Method Based On Photoionization Detector

In the background of the continuous development of Volatile organic organic compounds(VOCs)monitoring technologies and the maturing of VOCs monitoring technologies in complex environments,the application of methods such as machine learning in the Volatile organic compounds monitoring process is increasing.When detecting VOCs in humidity condition by Photoionization detector(PID),the PID photocathode surface ionization barrier and uneven distribution result in low frequency noise of the PID signal.The accuracy of traditional PID calibration model is low and the robustness of traditional PID based VOCs monitoring method is poor.In this study,software and hardware development,wavelet packet transformation,signal feature extraction,principal component analysis,genetic algorithm,support vector machine regression and other methods were used to study the volatile organic compounds monitoring method based on PID.The correction method of PID for different VOCs is analyzed,and the principle of selecting photoionization detector is described.The problems of PID monitoring VOCs in humidity environment are analyzed,and the research scheme is designed.A test platform for volatile organic compounds based on photoionization detector was built,including a system for preparing low concentration VOCs with high concentration standard VOCs gas.Hardware development module and software development module of the test platform were designed.So as to realize the acquisition,storage and processing of PID response VOCs signal in the laboratory environment.In this thesis,PID monitoring VOCs in soil-water-gas is studied.Due to high humidity in soil-water environment,PID monitoring VOCs will cause barriers and uneven distribution on the surface of PID photocathode,resulting in a large number of low frequency noises in PID signal.In order to reduce the influence of low frequency noise on PID signal generated in humidity environment,this study constructed a denoising method combining wavelet packet and energy adaptive weight,and compared this denoising method with traditional wavelet packet denoising method.It is verified that this method can reduce the mean square error of PID signal by 51% and increase the signal to noise ratio of PID signal by 9% compared with the traditional wavelet packet denoising in humidity environment.The reasons for the low accuracy of the traditional SVM regression model were analyzed,and the characteristics extraction and principal component analysis of PID signals were carried out to reduce the data dimension.The optimal SVM parameters were selected by using the genetic algorithm,and the PID calibration model combining support vector machine regression with signal time-domain feature extraction,signal frequency domain feature extraction,principal component analysis and genetic algorithm was constructed.The R2 of this model can reach 99.99%;The influence of the number of samples on the accuracy of the model is analyzed.Aiming at the problem of weak robustness of traditional PID calibration model,a PID monitoring VOCs method combining wavelet packet energy adaptive soft threshold weight denoising with PID signal feature extraction in time domain and frequency domain-PCA-GA-SVM regression was constructed to realize VOCs monitoring in humidity environment,which verified the robustness of the monitoring method proposed in this study.There are 36 figures,21 tables,and 85 references in the thesis.