基于傅里叶变换红外光谱技术的非线性气体定量分析模型研究-安全技术及工程专业论文.docx

AbstractAbstract Abstract Abstract Quantitative analysis of toxic and hazard gases is very useful for fire rescuing，fire monitoring，gas leaking diagnosis and research on material characteristics．The current gas monitoring technology has both advantages and disadvantages on consideration of applied scope， sensitivity，reliability，interference immunity，service life，convenience， and economy．With high sensitivity，wide applied scope，fast response， long service life and convenience，Fourier Transform Infrared Spectroscopy(FTIR)was chosen as the ideal technology in the field of gas monitoring．The foundation of the study is to obtain accurate spectrum data by designing reliable experimental system and strictly controlling the process of the experiment．Meanwhile，to establish the new nonlinear gas quantitative analysis models is the key point of our research． The experimental system was established by FTIR，gas compounding equipment and other devices．Eight different kinds of classic toxic gases with low concentration，including CO．C02，NO，N02。S02．HCI．HBI。and HCN，were chosen for quantitative analysis．To remove the interference from vapor and carbon dioxide，appropriate spectrum region was chosen． Analyzing models were established after data pre·treating，standards selecting and parameters setting．The R2 value，root·mean standard error of cal ibration(RMSEC)and root-mean standard error of predi ction (RMSEP)of each component were then calcul ated． To study the relationship between the spectrum data and gas concentration data，two new non．1inear methods，including Polynomial partial least square(PPLS)and partial least square with BP artificial neural net(PLS-BP)were created．To compare the fitting capacity and predicting performance among the non—linear models，least square with support vector regression(LS·SVR)was imported to the field of gases quantitative analysis．Compared with two classical methods，PLS and classical least square(CLS)，the characteristics，prediction performance and