Research On Application Of Digital Cigarette Formula Maintenance Based On Computational Intelligence

The stable quality of cigarette products is the internal driving force for the innovation and development of tobacco enterprises,and the maintenance of cigarette formula is the key to ensure the quality of products.Traditional cigarette formula maintenance mainly relies on expert experience and manual evaluation,which is subjective and inefficient,wastes a lot of raw materials and seriously affects the long-term development of enterprises.Therefore,tobacco enterprises urgently need to combine intelligent methods to achieve efficient and objective digital quality evaluation and formula maintenance of tobacco leaves,so as to ensure the stability of product quality in the production process.Aiming at the problems encountered in the maintenance of tobacco formula,a fast and reliable digital formula maintenance method system was developed based on the near infrared spectrum data and leaf group formula maintenance data,combined with spectral similarity measurement and data mining technology.The specific research contents are as follows:(1)Aiming at the problem of "distance failure" in distance measurement caused by the nonlinearity,high redundancy,and high noise of near-infrared spectroscopy data,a rank-order distance local preserving projection(KRLPP)similarity measurement algorithm based on kernel mapping is proposed.Based on the classic local preserving projection algorithm(LPP),this method introduces kernel mapping to map the spectral data to a higher-dimensional space,ensuring the nonlinear characteristics of the manifold structure to the greatest extent;The Euclidean distance metric commonly used in the LPP algorithm is replaced with the rank-order distance,and the similarity relationship of the sample points is re-measured by sharing the information of local neighboring points,and finally the similarity measurement of tobacco leaves is realized in a low-dimensional space.In order to verify the effectiveness of the method,firstly,the dimensionality reduction effect was compared with PCA,LPP,and INLPP.The experimental results showed that the method was significantly better than other algorithms for distinguishing different parts of tobacco leaves.Then PCA,LPP,and KRLPP are used as the target tobacco leaves in the formula to find replacement tobacco leaves,and the quality evaluation and analysis of the recommended replacement tobacco leaves are carried out through two methods of intrinsic chemical composition and sensory evaluation.The experiments show that the KRLPP algorithm has the best effect on finding similar replacement tobacco leaves in formula maintenance,and the obtained replacement tobacco leaves are closer to the target tobacco leaves in terms of chemical composition and sensory quality.(2)In the maintenance of cigarette formula,it is necessary to comprehensively consider the influence of the grade,position,origin and other factors of each single cigarette.However,the current mining of formula maintenance rules is mostly limited to the single attribute of tobacco leaves,which leads to the mining rules have little guiding significance for the maintenance of cigarette formula.Aiming at this problem,this paper proposes a multi-dimensional FP growth association rule algorithm based on hash table,fully considering the influence of the inherent multi-dimensional attributes of single cigarette on the final maintenance results.When scanning the database,the hash table is used to store the information of the items in the header table of FP growth algorithm.It does not need to traverse the linear order table from the beginning every time,which effectively improves the efficiency of the algorithm.At the same time,the multi-dimensional association rule algorithm integrates a variety of key attributes that affect the overall formula,such as the grade,location,and origin of the single-material cigarette,and excavates expert experience such as the compatibility and substitution rules among the single-material cigarettes accumulated in the formula list,so that the mining can be replaced The rules are more guiding and reference value.Finally,the excavated single-material tobacco replacement law was combined with the KRLPP similarity measurement algorithm to formulate a formula maintenance design plan,which solved the problem of finding similar tobacco leaves and replacing single-material tobacco in formula maintenance.The method can be directly used to guide the design and maintenance of the leaf group formula,improve the work efficiency of formula personnel,ensure the stability of product quality,and enhance the comprehensive competitiveness of the tobacco enterprise's formula research and development.