Optimization Of Tobacco Shred Manufacturing Key Process Parameters Based On DOE And BP Neural Networks

Tobacco shred manufacturing line is one of the most important aspects in cigarette production process. Each processing parameter settings in this line will significantly affect the physical quality, chemical composition and sensory quality of the cigarette. This paper were based on a school-enterprise cooperation project between K cigarette factory and Kunming university of science and technology named "adding spices process multiphase flow numerical modeling and applied research", and used the tobacco shred manufacturing line of Z brand cigarettes that the drying process was using drum device as the research object to study, the four major processing operations of loose resurgence, adding spices, cutting tobacco shred and drying. And we used the methods of DOE (Design of Experiments), AHP comprehensive evaluation method and BP (Back Propagation) Neural Networks to study and optimize the process parameters.This thesis includes two parts, the one is basic experimental studies, and the other is experimental study of modeling and simulation and parameter optimization. For the basic experimental studies, firstly, we were exhaustively analysed the tobacco shred manufacturing process, and combined with causal analysis and Delphi method to determine the critical process parameters from many process parameters, and gave the values for each parameter level. Then, we arranged each parameter combination test by uniform design method, examined the quality index values of sensory quality, chemical component, and smoke routine, analyzed the relationship between parameter level combination and tobacco quality, and determined the quality indicators of parameter optimization by compositing the statistical analysis results, the function of tobacco quality index values and process parameters, as well as the study of corresponding literature of the tobacco qualities. Finally, we composited the four quality indicators which were selected out before to a comprehensive score according to the AHP comprehensive evaluation method, so the multi-objective parameter optimization problem were transformed into a single objective parameter optimization problem. After the key process parameters, key quality indicators, and the comprehensive evaluation method of these indicators had be identified, we were carried on the study of modeling and simulation experiment, in this part, we should establish the BP neural network model for the process parameters and the quality indicators, and put the experimental data into the neural network to train the network and test its availability. And then we redesigned and rearranged the experiments by Taguchi method, as well as predicted the value of all the key quality indicators by the established BP neural network, and combined with SNR (Signal-to-Noise Ratio) analysis tool to optimize the process parameters.From this study we got the optimal combination of process parameters were 65 ℃, 70℃,35kg/h,100℃ and 0.25m/s, the corresponding tobacco qualities such as the whole wire cut tobacco was 85.87%, filling value was 4.74cm3/g, the total sugar content was 22.9%, reducing sugar content was 19.21%, and the composite score was 74%, which was the maximum of all test results. The conclusions show that the study had achieved the desired effect, the theories and methods of this study were reasonable and feasible, and this research had a certain practicality.