基于ISOMAP-ELM的软测量建模及化工应用

针对化工生产反应过程复杂,难以直接建立对应的机理模型,提出了一种基于ISOMAP-ELM的软测量模型。将等距离映射与极限学习机结合,通过等距离映射对输入数据进行降维,消除输入数据的共线性,在低维空间提取更具代表性的特征分量,最后对特征分量采用极限学习机训练,以此建立软测量模型。验证结果表明,所提出算法与传统的基于ELM以及基于欧式距离降维的MDS-ELM模型相比,具有更高的预测精度,模型的均方误差仅为0.28,软测量模型的命中率达到94%,对化工生产具有一定的指导意义。

Soft Measurement ModeIing and ChemicaI AppIication Based on ISOMAP-ELM NeuraI Network

In view of the chemical process,it is difficult to build a corresponding mechanism model,a soft sensor model based on ISOMAP-ELM is proposed. It combines ISOMAP with ELM,through the ISOMAP,the input data can reduce its dimensionality,eliminate the colinearity among each other,and extract more representative characteristics. Finally,the extracted feature components are sent to ELM to build a soft sensor model. Verification results show that the proposed algorithm has a high prediction precision,model squared error is only 0. 28,and the hit rate of the soft sensor model is 94%,it can give guidance for chemical process.