基于RBFNN的硫熏强度软测量建模

Soft-Sensing Modeling of the Intensity of Sulfitation Based on RBF Neural Network

硫熏强度是亚法糖厂澄清工段非常重要的一个工艺指标,硫熏强度过低会影响澄清效果,过高会造成成品糖二氧化硫残留过高。由于目前尚缺乏合适的硫熏强度在线测试仪,人工化验滞后时间较长,难以根据该指标及时指导生产。为此,提出了一种基于糖厂澄清过程大量离/在线历史数据的硫熏强度软测量方法,分别建立基于径向基函数神经网络(RBFNN)、BP神经网络方法和广义动态模糊神经网络(GDFNN)的硫熏强度软测量模型。通过对模型的性能进行对比分析,说明了基于RBFNN硫熏强度软测量模型的优越性。

Intensity of sulfitation is a key technique index in the clarification process of sugar mill with sulfitation process. It will lead to bad clarification effect when the intensity of sulfitation is too low, and the high intensity of sulfitation will bring high sulfur dioxide residue in the finished sugar. Due to lacking of effective on-line instrument to measure the intensity of sulfitation and a long delay time of manual measuring, it is hard to guide the operation timely, according to this technique index. So a soft-sensor method for the intensity of sulfitation is put forward, which is based on a lot of off-line and on-line history data. Several soft-sensing models are established respectively by radial basis function neural network（RBFNN）, BP neural network and genetic dynamic fuzzy neural network（ GDFNN）. The effectiveness of the soft-sensing models based on RBFNN is tested by comparing analysis of the model＇ s performance with the other two models.