Research On Optimization Of Tobacco Redrying Process Parameters Based On RBFNN Fusion Optimization Algorith

In the tobacco regrilling process,the setting of tobacco regrilling process parameters is greatly influenced by factors such as ambient temperature、humidity,and the moisture content of tobacco leaves.In the complex and changing characteristics of the incoming tobacco,it is difficult to precisely adapt the process parameters to the fluctuation of the incoming tobacco by manual experience alone,which cannot guarantee the export quality of tobacco regrilling.In this paper,the Radial Basis Function Neural Network(RBFNN)fused with Particle Swarm Optimization(PSO)is proposed to provide control parameters for tobacco regrilling production.This paper mainly includes the following three research contents.(1)Selection of key parameters affecting the quality of tobacco regrilling and data pre-processing.Since the regrilling production data were collected directly from the PLC,there were missing and abnormal values,so this paper firstly preprocessed the data.Considering that the regrilling process variables are highly coupled and the drying characteristics of tobacco leaves are affected by multiple factors,this paper analyzed the data to select the physical characteristics parameters that affect the drying of tobacco leaves,and finally,the key process parameters that affect the drying of tobacco leaves were selected using gray correlation analysis.(2)Process optimization was carried out for tobacco leaves with different incoming characteristics to meet the standard roaster outlet moisture and temperature values required by different principals.First,RBFNN was used to fit multivariate relationships with roaster outlet moisture and temperature to obtain a tobacco quality prediction model that meets the demand.Then,an improved Particle Swarm Optimization(IPSO)algorithm was used to optimize the process with the standard roaster exit moisture and temperature as the optimization objectives.To address the problem that the traditional PSO easily falls into local optimum in the late stage of optimization,two improvement strategies are adopted in this paper: firstly,a new inertia weight update method is introduced,which adopts a larger inertia weight in the early stage to speed up the search rate,and the inertia weight gradually becomes smaller as the iteration progresses,which facilitates the overall local optimization search;secondly,a decay factor is introduced to increase the search diversity and improve the algorithm’s search accuracy.Finally,the method is compared with the common intelligent optimization algorithm,and the experimental results show that the improved optimization algorithm has improved in convergence speed and accuracy.(3)To address the problem of slow convergence of the optimization algorithm,the IPSO fusion Self-Adaptive Penalty Function(SAPF)method is used.In the actual regrilling production,it is important to shorten the optimization algorithm’s search time to improve the efficiency of tobacco regrilling.From the high-dimensional nature of the process operating conditions,it is clear that the difficulty of this experiment lies in the search for the optimal solution of the optimization problem under the complex nonlinear constraints of the multi-process.To address this problem,this paper adopts a penalty function to transform the constrained optimization problem into an unconstrained optimization problem,and constructs a method of adaptive penalty function by introducing the information of the proportion of feasible particles in the candidate solution into the penalty function calculation during the update of the solution.At the same time,in order to increase the degree of integration between the variation of the fitness function and the particle updating process,this paper adopts the method of SAPF fusion IPSO to optimize the solution of the process parameters.Finally,in order to verify the effectiveness of the method in this paper,a re-baking experiment was conducted.The experimental results show that the improved optimization algorithm can reach full convergence at 39 iterations,and the fluctuation range values of baking machine outlet moisture and temperature are reduced by 7.5% and 11.8%,respectively,compared with the real-time values in the field,which has good application prospects and promotion value.