Abstract: In order to promote the control precision of moisture content in strips out of loosening and conditioning process step, a precise control model was developed following the statistical regression analysis of historical data of the said process step, and further revised for self-adaptive optimization with self-learning algorithm. The control system was validated with the data of on-line monitoring samples of "Hongqiqu (Tianxingjian) "cigarette brand in Nanyang Cigarette Factory. The results showed that:the control precision of moisture content in output strips was significantly improved, process deviation, standard deviation and extreme difference decreased by 0.24%, 0.078% and 0.34%, respectively; process capability index increased by 0.54. It indicated that the level of process control was effectively improved. This method provides a support for the promotion of intra-batch quality consistency in tobacco primary processing.