Study On Aliasing Peak Identification And Quantitative Analysis Of Chemical Warfare Agents Fourier-transform Infrared Spectroscopy

Chemical weapons is a kind of sophisticated weapons with super-toxic and deadly.They not only can affect enemy muchly in a variety of exposure, but also can form atoxic gas widespread through the wind and rain which will casue larger damage to human and ecological environment.The primary measure to reduce this hazard is chemical weapons protection which include monitoring alarm, destruction, protection, decontamination, rescue. One of the most crucial steps is monitoring and alarm.Timely reports will provide large amounts of data intelligence for subsequent decontamination and protection.Also it will enhance the operating efficiency.The presence of interfering gases in the atmosphere will affect the identification of chemical agents.In this dissertation, it is investigated spectrum data processing method of nerve agents sarin on the basis of Fourier-transform spectroscopy detection principle and atmospheric infrared absorption spectrum.This paper mainly carries out baseline correction, feature extraction and quantitative analysis for infrared spectrum of sarin simulant Dimethyl methy lphosphonate.Firstly, in view of the slowly varying baseline drift during infrared spectrum measurement, the algorithm of adaptive iteratively reweighted penalized least squares is used to correct baseline.This method can obtain a better fit smoothness baseline,which makes the corrected spectrum information maximum approximation to the original spectrum,while the fidelity reached above 99.06%.Secondly,for the non-Gaussian,nonlinear and high-dimensional of mixed gases including Dimethyl methylphosphonate,anti-2-butene,1,3-butadiene,propylene and random noise, the study considers above-mentioned four gases and random noise signals as stay measuring component.Then using fast independent component analysis based on higher-order statistics separated the components. Also it enhances convergence rate of traditional fast independent component analysis.through improving iterative process.Thirdly, extrem learning machine was used to establish concentrations correction model of each component,and optimizated the parameters of quantitative model.Finally, the proposed methods according to theory were verified by experiments. Mixed gases comprising Dimethyl methylphosphonate, anti-2-butene,1,3-butadiene and propylene collected were collected by the experimental system.Quantitative analysis model were optimizated via parameter adjustment.And the feature extraction method can be validated against the precision of the quantitative model. Experimental results show that the optimized fast independent component analysis makes the separated similarity reach 0.9748 and the average speed increases 0.05 seconds.The precision of each components’quantitative models are reached above 0.9846 after parameter optimization.