能处理定性参数输入的推广模糊神经网络软测量建模方法

Extended FNN Soft Sensor Model for Handling Qualitative Inputs

针对生产过程中存在多种类属型数据和混合型数据,而大多数软测量方法只能处理数值型数据的问题,提出了一种基于粗糙集方法的推广模糊神经网络软测量建模方法,该方法既可以接受定量参数输入,也可以接受定性参数输入。首先建立模糊-清晰混合规则的定义,对具有混合类型属性的样本集进行离散化处理后,利用粗糙集的约简算法进行规则提取,获得最小决策集。由得到的混合决策规则构建推广模糊神经网络,使用样本集训练网络参数。最后将该方法应用于蒸发器的污垢热阻值估计,取得了良好的效果。

An extended FNN soft sensor model was developed for handling quantitative and quahtative inputs （include categorical variables and hybrid variables） in process. Fuzzy-crisp hybrid rule was defined firstly. Rough set theory is used to obtain the reductive hybrid rules from sample sets that have been preprocessed using discretization method. The inference system comprised by hybrid rules was represented via an extend fuzzy neural network（FNN）, which was trained by sample sets. The approach is used to build a soft sensor model for estimating the fouling resistance of evaporator and has a good performance.