Authors:
Samuel Latham
and
Cinzia Giannetti
Affiliation:
Faculty of Science and Engineering, Swansea University, Fabian Way, Swansea, Wales
Keyword(s):
Root Cause Analysis, Machine Learning, Classification, Data Analytics, Knowledge Integration, Hot Strip Mill, Steel Industry.
Abstract:
Data is one of the most valuable assets a manufacturing company can possess. Historical data in particular has much potential for use in automated data-driven decision-making which can result in more efficient and sustainable processes. Although the technology and research behind data-driven systems for Root Cause Analysis has developed vastly over decades, their use for real time automated detection of root causes within steel manufacturing has been limited. Typically, root cause analysis still involves a lot of human interaction both in the pre-processing and data analysis phases, which can lead to variability in results and cause delay when devising corrective actions. In this paper, an application for automated Root Cause Analysis in an Hot Strip Mill is proposed for the purpose of demonstrating the effectiveness of such an approach against a manual approach. The proposed approach classifies temperature defects of steel strip Width Pull using a variety of machine learning algorit
hms in conjunction with k-fold cross validation.
(More)