Industrial case study of causal modeling of continuous casting and lamination of steel tubes

Black-box models have shown high flexibility and accuracy in prediciting what values certain variables involved in industrial processes will assume in the future, given the values of certain other variables. These models, however, are frequently too complex to be interpreted by a human operator, and...

Full description

Saved in:

Bibliographic Details
Published in:	2021 IEEE Latin American Conference on Computational Intelligence (LA-CCI) pp. 1 - 6
Main Authors:	Caldeira Silva, Danilo A., Salis, Turibio Tanus, Braga, Antonio Padua
Format:	Conference Proceeding
Language:	English
Published:	IEEE 02-11-2021
Subjects:	Casting Causal modelling Computational modeling counterfactual queries Data models Electron tubes Lamination Mathematical models Predictive models rule-extraction seamless steel tubes
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	Black-box models have shown high flexibility and accuracy in prediciting what values certain variables involved in industrial processes will assume in the future, given the values of certain other variables. These models, however, are frequently too complex to be interpreted by a human operator, and are frequently unable to furnish adequate answers to queries regarding interventions in a given system, or to answer counterfactual queries. Causal models, however, frequently can. In this work we explore the causal modeling of two stages in the production of seamless steel tubes, extracting directed acyclic graphs, which can then be used for rule extraction, as well as for predictive, intervention and counterfactual queries.
DOI:	10.1109/LA-CCI48322.2021.9769827