Study On Cigarette Manufacturing Process Parameters And Harmful Constituents In Mainstream Smoke By Near Infrared Spectroscopy Combined With Chemometric Methods

As a non-destructive new analytical technique, Near-infrared (NIR) Spectroscopy, a newhigh-efficiency analytical technique, has been receiving considerable attention in Tobaccoindustry due to its high measuring speed, simultaneous determination of components, zeropollution, very low running costs and requirement of less or even no sample preparationswhich make them highly suitable for in-line process monitoring. However a number of issuesarise with respect to Near Infrared Spectroscopic instrumentation including (i) the resultingspectral data comprises hundreds of measurement values per spectrum,(ii) there is a lack ofselectivity with respect to the spectroscopic measurements. Therefore, the successfulapplication of Near Infrared Spectroscopy to the characterization of complex systems such ascigarette mainstream smoke can only be achieved by incorporating chemometric methods (e.g.PCR and PLS) in the extraction of chemical information from the spectral measurements. Themain aim of this thesis is to investigate the application potential of Near InfraredSpectroscopy combined with chemometric methods to the study of cigarette manufacturingprocess parameters and harmful constituents in mainstream smoke. The main achievements ofthis thesis are as follows:1..A robust calibration model has been established by PLS between the near infraredspectral measurement of cigarette samples and the contents of Phenols in mainstreamsmoke. The root mean square error of cross validation of the established calibrationmodel is5.1μg/cig. The differences between the predicted values obtained by thecalibration model and the expected values are in the range of±13μg/cig. The effectivepredictive range of the established calibration model is in37.0~101.1μg/cig with arelative prediction error of about6.7%.2. A robust calibration model has also been established by PLS between the near infraredspectral measurement of cigarette samples and the contents of NH3in mainstream smoke.Experimental results showed that the contents of NH3predicted by the establishedcalibration model are in good agreement with the corresponding expected values. Theroot mean square error of cross validation of the established calibration model is4.2μg/cig. The differences between the predicted values obtained by the calibration modeland the expected values are in the range of±10μg/cig. The effective predictive range of the established calibration model is in5.6~39.7μg/cig.3. The application potential of a newly developed chemometric method, Opticalpath-Length Estimation and Correction (OPLEC), in the analysis of the near infraredspectral measurements has been investigated. For the first time, a robust calibrationmodel with good predicting performance between the near infrared spectralmeasurements of tobacco leaf slices and cigarette pressure drop has been established byOPLEC.4. OPLEC has been successfully used to build a robust calibration model between the nearinfrared spectral measurements of tobacco powder samples and their heat of combustion.The obtained calibration model can provided quite accurate predictions for the heat ofcombustion values of tobacco powder samples (mean relative prediction error=0.7%).The correlation analysis revealed that heat of combustion of tobacco powder samples ispositively correlated with the chemical constituents in the mainstream smoke (MSS) suchas yields of tar, nicotine, carbon monoxide (CO) and total particulate matter (TPM),which may suggest its potential use in the evaluation of the usability of tobacco.The techniques developed in this thesis has features of good accuracy, high speed, zeropollution and low running costs, and therefore can be applied to in-situ real time monitoringof cigarette manufacturing process with a view to ensure the production of cigarette productswith stable quality.