Prediction models for harmful components in mainstream smoke of slim cigarette based on cigarette material parameters

In order to investigate the effects of cigarette material parameters (including filter ventilation, filter rod pressure drop, cigarette paper grammage, air permeability of cigarette paper, combustion improver content in cigarette paper, and K/Na ratio in combustion improver) on the deliveries of tar, nicotine, the 7 harmful smoke components, and the hazard index (H) of cigarette mainstream smoke, 50 cigarette samples were prepared according to a central composite design combined with orthogonal design methods. Ten multi-factor prediction models based on cigarette material parameters were established by linear regression and stepwise regression methods, and the optimal models were selected according to the statistical principle of minimum RMSECV (Root mean squares error of cross validation). The results showed that the ten models had good prediction accuracies, with the average relative deviations of prediction between 3.11% and 8.10%. The models were well adaptive for different cigarette material parameters.