Research On Comprehensive Quality Evaluation And Alternative Measurement Method Of Cigarette Formula Module Based On Neural Network

The cigarette formula module is the main raw material of cigarette manufacturing enterprises,which is obtained by mixing flue-cured tobacco leaves of different production areas and different grades of quality according to the corresponding proportion using the theory of leaf group formula design.The quality of the formula module directly affects the quality of the finished cigarette products,while the comprehensive quality evaluation of the cigarette formula module involves many quality indicators such as chemical quality,sensory quality,physical quality and so on,which is a complex work.How to objectively and scientifically evaluate the comprehensive quality of the cigarette formula module,and then predict and substitute the comprehensive quality of the module,so as to provide a basis for the use of the cigarette formula module and improve the stability of the quality of cigarette products,has always been a key issue for cigarette manufacturing enterprises.This paper is based on the use data,chemical quality and sensory quality data of 276 cigarette formula modules of cigarette company A from 2016 to 2020.First,a comprehensive quality evaluation method based on the historical use data of cigarette formula modules is found by using the data mining method.The evaluation result of this method is obviously better than the current sensory quality evaluation result of cigarette company A,It can more accurately and carefully reflect the comprehensive quality level and economic value of the formula module;Then,the key indexes affecting the quality of the cigarette formula module were selected by using the statistical analysis method,and the BP neural network model of the key indexes and the comprehensive quality was established.After training,the accuracy of the prediction model reached more than 85%,which can better realize the comprehensive quality prediction;Finally,based on the research of simple correlation coefficient and European distance,the measurement method of cigarette formula module substitutability is improved,and the measurement method is applied to the SOM neural network model to achieve classification based on the similarity of comprehensive quality.The results show that the improved substitutability measurement method can accurately measure the substitutability between formula modules,and can be consistent with the actual replacement results in history,The classification results based on substitution are also more reasonable.