摘要
在复烤润叶机加水控制系统中,入口烟叶流量及其含水率的随机性以及润叶过程的不确定性,导致了难以应用变量间函数依赖关系实现对润后烟叶含水率精准的控制。鉴于贝叶斯网在表示和处理随机变量概率依赖关系方面的优势,通过对复烤润叶机组成的分析建立起复烤润叶机加水控制贝叶斯网结构,定性描述了润后烟叶含水率及其影响变量之间的概率依赖关系;利用实时贝叶斯网参数学习算法,根据生产过程中实时采集的数据学习贝叶斯网参数,以定量表示这些随机变量之间的影响程度,从而实现可测变量对受控变量的实时推算。实测数据表明,使用贝叶斯网方法后,润后烟叶含水率的加工能力指数Cpk从0.371提高到2.799,润后烟叶含水率波动明显降低,标准差Sigma从1.512降低到0.236,有效提高了润后烟叶含水率的控制精度。
In the moisture control system of ordering machine, the irregularity of tobacco input and its moisture content and the uncertainty of the ordering process make it impossible to control the moisture content in ordered tobacco leaf precisely with the functional relations between variables. Bayesian network had the advantages of expressing and processing the dependence relations between stochastic variables, the dependence relation between the moisture content in ordered leaf and control variables is qualitatively described by means of the Bayesian network structure established via the analysis of ordering machine. The degree of dependence between variables could be quantitated by the Bayesian network parameter learning algorithm with the reahime data collected during tobacco processing, and the controlled variable was estimable from measurable variables. The practical data indicated that by using Bayesian network method, the process capability index (Cpk) increased from 0. 371 to 2. 799, the fluctuation of moisture content in ordered leaf reduced obviously, and the standard error decreased from 1. 512 to 0. 236.
出处
《烟草科技》
EI
CAS
北大核心
2009年第7期18-22,共5页
Tobacco Science & Technology
关键词
复烤润叶机
加水流量
贝叶斯网
烟叶含水率
Ordering machine
Water input
Bayesian network
Moisture content in tobacco leaf