Research > Funded Research Projects > Machine Learning to Analyse Sensor Data in Steel Plants

Machine Learning to Analyse Sensor Data in Steel Plants

Reducing energy consumption and material loss in steel production using predictive machine learning.

Project Summary

In 2020, global steel production totalled 1,860 million tonnes, leading to the emission of over 3 billion tonnes of CO2, roughly 8% of all man-made greenhouse gas emissions. Steel producers are facing increasing pressure to reduce the CO2 emissions associated with steel production, both from governments and from industrial consumers.

One aspect of this is to apply machine learning (ML) methods to analyse sensor data in steel plants, and optimise the process parameters. A steel plant can have hundreds of thousands of sensors, providing information related to some 5,000+ process parameters. There is an opportunity to employ data-driven approaches to optimise these processes, increasing productivity, reducing energy consumption, and reducing wasted material.

This project is a collaboration between:

• Deep.Meta - Who will focus on the use of ML algorithms to generate real-time insights enabling operators to update control parameters to improve the performance of key processes in steelmaking.
• Materials Processing Institute – Who will lead work on electric arc furnace (EAF) operation, carrying out melts using their pilot-scale plant, enabling Deep.Meta to trial optimisation strategies and measure the results.
• Spartan UK – Who will work with Deep.Meta to analyse process data and create algorithmic schedules for the reheat furnaces and rolling mill that optimise throughput and prevent material loss due to oxidation.
• Grosvenor – Who will lead a life-cycle assessment to determine the impact of the energy savings and resource efficiency improvements in the context of the construction industry, helping them to evidence and quantify the impact of choosing more sustainably produced steels.

This project aims to deliver data-driven process optimisation for the Electric Arc Furnace (EAF), reheat and hot-rolling. The project will establish the data pipelines to extract, transform and load process parameters, and develop and train machine learning algorithms to analyse these data. This will result in real-time actionable insights to optimise Electric Arc Furnace process parameters and reheat/hot-rolling scheduling, enabling savings in energy consumption and material loss.

Partners