Soft sensor modeling method based on k-nearest neighbor and RBF neural network

Author

Weijun Zhang ; Hongbo Gao

Author_Institution

Sch. of Mater. & Metall., Northeastern Univ., Shenyang, China

fYear

2012

fDate

29-31 May 2012

Firstpage

11

Lastpage

15

Abstract

A soft sensor modeling method based on k-nearest neighbor and RBF neural network is presented to diminish the effects of outliers on the developed soft sensor model. Firstly, the anomaly degree of each modeling data pairs is calculated by using the k-nearest neighbor algorithm. Then, the weight of each modeling data pairs is determined according to the calculated anomaly degrees. Lastly, a soft sensor model is developed by using RBF neural network with weighted training error. Simulation is performed using functional data and production data from Nosiheptide fermentation process, and the simulation results show the effectiveness of the presented approach.

Keywords

fermentation; pattern clustering; radial basis function networks; Nosiheptide fermentation process; RBF neural network; anomaly degree calculation; functional data; k-nearest neighbor; modeling data pair weight; outliers; production data; radial basis function networks; simulation results; soft sensor modeling method; weighted training error; Analytical models; Biological system modeling; Biomass; Data models; Estimation; Neural networks; Training; RBF neural network; k-nearest neighbor; modeling; outlier; soft sensor;

fLanguage

English

Publisher

ieee

Conference_Titel

Natural Computation (ICNC), 2012 Eighth International Conference on

Conference_Location

Chongqing

ISSN

2157-9555

Print_ISBN

978-1-4577-2130-4

Type

conf

DOI

10.1109/ICNC.2012.6234583

Filename

6234583