USP Electronic Research Repository

Estimation of the insulation deterioration of metallurgical ladle by use of RBFNN models

Christova, Nikolinka and Vatchkov, Gantcho L. (2012) Estimation of the insulation deterioration of metallurgical ladle by use of RBFNN models. [Conference Proceedings]

[thumbnail of Estimation of the Insulation Deterioration of Metallurgical Ladle by Use of RBFNN Models]
Preview
PDF (Estimation of the Insulation Deterioration of Metallurgical Ladle by Use of RBFNN Models) - Published Version
Download (1MB) | Preview

Abstract

A model-based approach for estimation and diagnosis of the deterioration in the metallurgical ladle insulation is proposed in this paper. It is based on using the diverse information that comes from the so called thermovision analysis (thermographic images), which show the temperature profile on the surface of the ladle. A group of Radial Basis Function Neural Network(RBFNN) models with different structures is developed and used for such estimation. Each model has different number of input parameters and a different output, in order to estimate the respective parameters of the insulation deterioration (the defect), such as its depth, width and shape. The created RBFNN models are a kind of diagnostic models because they solve the inverse problem, namely: finding the parameters of the defect, taking into account the available measured symptoms (the selected parameters from the thermographic images).
The estimation results from all proposed diagnostic models
are shown and discussed in the paper, by using simulated input-output data sets. Respective suggestions and procedure for selection of the best diagnostic model are also given in the paper.

Item Type: Conference Proceedings
Additional Information: DOI 10.1109/EMS.2012.94
Subjects: T Technology > TN Mining engineering. Metallurgy
Divisions: Faculty of Science, Technology and Environment (FSTE) > School of Engineering and Physics
Depositing User: Gancho Vatchkov
Date Deposited: 01 May 2013 04:23
Last Modified: 20 Jul 2016 02:59
URI: https://repository.usp.ac.fj/id/eprint/5737

Actions (login required)

View Item View Item