REVISTA DYNA NEW TECHNOLOGIES

Key words:

sistemas difusos tipo-2, modelado de caja gris, aprendizaje híbrido, estimación de temperatura, laminado en caliente, type-2 fuzzy inference systems, grey-box modeling, hybrid learning, temperature estimation, hot rolling

Article type:

ARTICULO DE INVESTIGACION / RESEARCH ARTICLE

Section:

RESEARCH ARTICLES

ABSTRACT:
In hot rolling, on-line estimation of the bar head-end temperature at the scale breaker entry is of crucial importance for the finishing mill set-up, and thus fulfilling the bar head-end quality requirements. In most mills, a physical model performs the estimation from measured roughing mill exit temperature and travelling time. However, estimation is greatly affected by measurement uncertainties, variations in the incoming bar conditions, and product changes. To overcome these problems several fuzzy inference systems and fuzzy based grey-box models have been developed and tested previously. In this work, hybrid-learning type-1 and type-2 Mamdani fuzzy grey-box models will be developed, such systems have not been presented earlier for this application. Data from a real-life mill were collected and two different sets were randomly drawn. The first set is used for training the systems, while the second one is used for validation. The performance of the systems is evaluated by five performance measures applied on the prediction error with the validation set and is compared with that of the fuzzy systems presented earlier and the physical model used in plant. The results show that hybrid-learning Mamdani fuzzy grey-box modeling improves estimation.

Keywords: type-2 fuzzy inference systems, grey-box modeling, hybrid learning, temperature estimation, hot rolling.