Establishment and Application of Multivariate Prediction Models for Moisture Content in Input Tobacco to Tobacco Dryer

To stabilize the moisture content in input tobacco to a tobacco dryer during primary processing, the major factors influencing its moisture content were determined by Pearson correlation analysis method, a prediction model for the moisture content was established by neural network algorithm and multivariate regression analysis. The reference value of water addition rate in conditioning process could be calculated on the basis of given moisture content in input tobacco to tobacco dryer via the prediction models. The predicted values were compared with the measured values, the results showed that when the moisture content in input tobacco was set at 19.2%, the mean value of moisture content in input tobacco obtained by this method was 19.21%, better than 19.09% before improvement; furthermore the standard deviation of error decreased from 0.43% to 0.26%. The deviation of moisture content in input tobacco to tobacco dryer between batches was narrowed.