BEGIN:VCALENDAR
VERSION:2.0
METHOD:PUBLISH
BEGIN:VEVENT
ORGANIZER;CN=ESTAD 2023:mailto:info@metec-estad.com
LOCATION:Room 2
SUMMARY:"Hands-Free" Plant-Scale, Anomaly Detection AI
DESCRIPTION:Steel manufacturing processes are heavily instrumented, generating large volumes of automation data in the form of time series to the tune of over 5 million data points per second per plant. The time series data contain adequate information to represent the state of a physical system and production process at any given point in time. However, prevalent data-driven fault detection methods require significant setup efforts and expert inputs for modeling every known state of the system. 

This paper presents a novel self-supervised AI approach that does not require any setup effort and is capable of monitoring every existing process parameter and asset metric at high speed. Our approach utilizes a deep learning architecture based on Convolutional Variational Autoencoders (CVAE) that can start learning from small amounts of data to identify excursions, can automatically and incrementally learn as the underlying behavior of the asset changes and can process millions of measurements a second across thousands of time series. The automated time series AI informs plant operations of conditions that require human attention and provides diagnostics of underlying issues - leading to informed production and maintenance decision-making. Self-supervised AI overcomes the challenge the conventional machine learning method faces scaling to the needs of steel manufacturing by accommodating the challenges of constant equipment, environment, and product changes that hinder classical supervised learning methods. This paper will show this new AI in commercial steel manufacturing operations today.

CLASS:PUBLIC
DTSTART:20230614T114000
DTEND:20230614T120000
END:VEVENT
END:VCALENDAR