# Embedded real-time analysis of continuous casting for machine-supported quality optimisation

## TL;DR

This paper introduces a real-time digital system to improve continuous casting quality by predicting and preventing defects during steel production.

## Contribution

The novel contribution is a real-time support system using a 3D digital twin and Big Data to predict top-freezing events in continuous casting.

## Key findings

- Defect-promoting scenarios were identified through statistical analysis and surface defect detection.
- A 3D digital twin was developed to simulate thermal and fluid-mechanical conditions in the casting mould.
- Real-time prediction of top-freezing events significantly improves plant safety and inspection efficiency.

## Abstract

Thermal and fluid-mechanical conditions in continuous casting (CC) moulds are only roughly known although highly relevant for the product quality. Manual process control is difficult due to the big number of influencing factors. During continuous casting, manual top-freezing controls must be carried out. Every manual performed mould control can affect the strand quality and even increase the risk of failure. Therefore, regular top-freezing controls are performed after a certain casting duration. However, top-freezing events between the regular controls cannot be detected and are a major risk for plant safety.

In the RFCS project RealTimeCastSupport, the aim of the research was the digitalisation and optimised control of continuous casting machines. A real-time support system was developed to predict quality-relevant top-freezing events and thus achieve improved control. This was reached by offline material tracking, synchronisation of data streams and statistical analysis by application of Big Data technologies, the development of a digital twin and the exploitation of various CC data and surface inspection to predict reliability of steel production.

The following results were achieved:

Identification of defect promoting scenarios by correlation of statistical results and surface defect detection.

Realisation of an offline 3D digital twin of the mould considering heat transfer, inert gas feeding and solidification.

Offline reproduction of the identified defect promoting scenarios with the 3D digital twin to find thermal and fluid mechanical reasons for the detected behaviour.

The application of a real-time support system enables the prediction of top-freezing events during the whole casting process. Subsequently, this significantly increases the plant safety and offers to carry out top-freezing inspections in a more targeted manner in the future.

This publication is part of a series of papers in the frame of the dissemination project METACAST.