Identification of ARX Model with Multi-Gaussian Noises

In most industrial environments, the real process data points are usually subject to contamination from a variety of noises. Most of the traditional methods assume that the process noise is Gaussian white noise, which can lead to poor robustness of the model to abnormal data. In this paper, we consi...

Full description

Saved in:
Bibliographic Details
Main Authors: Bai, Wentao, Guo, Fan, Chen, Lei, Hao, Kuangrong, Huang, Biao
Format: Conference Proceeding
Language:English
Subjects:
ARX
continuous stirred tank reactor
Expectation-Maximization algorithm
Markov chain
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In most industrial environments, the real process data points are usually subject to contamination from a variety of noises. Most of the traditional methods assume that the process noise is Gaussian white noise, which can lead to poor robustness of the model to abnormal data. In this paper, we consider an ARX model with multi-Gaussian white noises, assuming that the collected data is affected by different noise models and switching of noise model follows Markov chain probability. The parameters of the model are estimated by Expectation-Maximization (EM) algorithm. A numerical example and a continuous stirred tank reactor are employed to verify the effectiveness of the proposed algorithm.
ISSN:2642-3901
DOI:10.23919/ICCAS47443.2019.8971542