摘要
在总结现有各种软测量技术的基础上,指出统计分析方法中的部分最小二乘法(PLS)和神经网络方法中的径向基函数网络(RBFN),是经实际应用证明有效的方法。提出了一种将PLS和RBFN结合的方法(PLS-RBFN),并将PLS,RBFN,PLS-RBFN3种算法分别用于加氢裂化分馏塔航煤干点软测量模型的建立,其泛化结果表明基于PLS-RBFN算法建立的软测量模型具有更好的预测精度。
Some of the softsensing techniques used for estimating product quality have been discussed in this paper.Among them,the PLS and RBFN algorithms based on statistical analysis and ANN respectively,are proved to be effective in practical application.Having considered their merits and defects,a combined method of PLS and RBFN (named PLSRBFN algorithm) is presented and used for modelling of jet fuel endpoint in hydrocracking fractionator.Generalization results show that the PLSRBFN algonithm gives more accurate prediction and more reliable extrapolation than the pure PLS and RBFN algorithms.
出处
《华东理工大学学报(自然科学版)》
CAS
CSCD
北大核心
1999年第4期420-423,共4页
Journal of East China University of Science and Technology
基金
高等学校博士学科点专项科研基金
关键词
软测量技术
产品质量估计
加氢裂化
分馏塔
soft sensors
product quality estimation
principal component analysis (PCA)
partial least square regression (PLSR)
radial basis function network (RBFN)