Your browser doesn't support javascript. This means that the content or functionality of our website will be limited or unavailable. If you need more information about Vinnova, please contact us.

Find patterns in new data flows

Reference number
Coordinator Jernkontoret
Funding from Vinnova SEK 1 500 000
Project duration March 2017 - July 2018
Status Completed

Purpose and goal

The huge amounts of data in the Swedish metal industry are not fully utilized but new methods of analysis makes it possible to extract new knowledge from them. However, there have been no or few examples of the benefits of the new uses of data. Costly data integration has been prioritized while the value has been unclear. This project breaks the circle by showing concrete values of investing in data integration and analysis. We show that machine learning can be used to identify reasons for quality deficiencies.

Expected results and effects

The total production data from hot-rolling rolling at Outokumpu has been shown to, with machine learning algorithms, be able to predict telescoping of hot-rolling rolls, and where in production the risks for this appear to occur. Together with production analysts, this knowledge has been translated into concrete proposals for actions. A demonstrator developed in the project shows how the new knowledge can be used by operators and analysts to better understand the manufacturing process.

Planned approach and implementation

Around 2600 production parameters from the entire production line, with all details from 3 months of production were used to train a dozen machine learning algorithms and evaluate them for their ability to predict telescoping. As there is a high level of prediction early in the line it is profitable to invest there to achieve a more controlled production. With Deep Learning, a computer model was trained to predict telescoping based solely on the band´s side deviation.

The project description has been provided by the project members themselves and the text has not been looked at by our editors.

Last updated 8 January 2019

Reference number 2017-01531

Page statistics