The segmentation method model based on cigarette density for burning cone drop rate prediction of super slim cigarettes

Abstract

The phenomenon of burning cone drop is a serious quality defect in super slim cigarettes. In order to monitor the burning cone drop, a segmentation model conveniently and effectively about cigarette packing density was established and used for predicting the burning cone drop rate. The method is a non-destructive testing (NDT), which is applied by using cigarette density and its change rate.

The model of cigarette density segmentation method is based on a straight line ρ=A×ρ’+B in the Cartesian coordinate system with abscissa ρ′ and ordinate ρ. The straight line split the coordinate system into two regions. By measuring the density of the super slim cigarettes, the position of the point (ρ’, ρ) can be determined, then the burning cone drop tendency can be predicted.

When verifying the accuracy of the model, it is found that the predicted value and the measured value are basically near the 1:1 line, the coefficient of determination (R²) and the index of agreement (D index) between predicted and observed values are 0.98 and 0.99 respectively, and the normalized root mean square error (nRMSE) is 11.8%. The coefficient of variation of the predicted burning cone drop rate is 6.5% for 10 consecutive replicates.

The cigarette density segmentation method model is suitable for predicting burning cone drop of the super slim cigarette samples with varying packing densities or cut tobacco distribution. The model has a good effect on the burning cone drop prediction under certain conditions, and it has an important guiding significance for super slim cigarette production and quality inspection.