
High Quality Scrap
Task 19 Team
-
Prof. Zushu Li
University of Warwick
-
Prof. Giovanni Montana
University of Warwick
-
Dr Yijun Quan
University of Warwick
Introduction
Increasing scrap usage is a strategy to help decarbonise the UK steel industry and maintain its sustainability considering the excessive scrap supply in the UK and significantly reduced CO2 emissions of using recycled scrap in steelmaking. However, there are various challenges associated with this. The first and most well-known challenge is the scrap quality – its measurement and control. The scrap supplied to steelmakers in general is a mixed material without chemistry measurement. The main measures are “trust”, visual inspection, and occasional spot check by hand-held XRF (which requires direct contact between XRF analyser and the sample surface). This limits the accuracy, consistency, and assurance of compatibility with product chemistry and value-in-use (VIU) that the materials (recycled scraps and associated raw materials) can offer.
The research in Task 19 aims to develop the industry highly sought-after, artificial intelligence (AI)-based tool for quantifying the scrap quality, and subsequently improve the scrap value-in-use (VIU) model for optimising its usage in steelmaking. The project will build on existing work in SUSTAIN, as well as the latest developments from the wider community such as Viability, RAP Prosperity Partnership, CircularMetal, DEFRA projects.
Initially, this Task will quantify the scrap quality by applying the computer vision-based system for scrap quality monitoring. This will provide assurance of scrap quality compatible with steel products. Work will then quantify the true value (contribution and penalty) associated with the use of a particular scrap, from metallurgical understanding of the effects of scrap quality (e.g. combined residuals and alloying elements, sterile) on steelmaking process (e.g. productivity, energy/materials efficiency, costs, CO2 emissions), downstream process, and steel properties via theoretical simulation, laboratory experiments and industry data. Finally, the scrap usage in producing high quality steel grades can be improved/optimised.
This project is looking to develop a computer vision-based steel scrap quality monitoring system. We aim to develop a computer vision-based system, with images of steel scraps as an input, so that it can output a quality assessment of the scraps, including the chemical composition of the batch. This quality assessment tool will be easy to operate to allow both metal recyclers and steelmakers to use them for their daily operation and trading. Currently, we have completed image-based tramp element copper percentage classification in the lab-environment for fragmentised steel scraps.
Progress To Date
Task 19 have conducted lab-environment image-based copper percentage estimation for fragmentised steel scraps. Fragmented steel scraps with copper contaminants were collected from Recyclers’ scrap yard. The origins of the scraps are end-of-life vehicles and domestic wastes, which makes scraps representative for commercial 3A/3B fragmentised scraps. The steel scraps were mixed with different amount of copper contaminants from the process and a labelled image dataset was constructed with both weight information for the tramp material (free copper) and image segmentation map for the objects. Using segmentation as a guide, Task 19 show steel scrap tramp material composition classification can be done using images only for test data with high accuracy in differentiating rich copper mixtures from low copper mixtures.
To extend the work from characterising tramp material only to performing overall composition estimation, Task 19 are also looking to develop an image-based steel scrap intrinsic composition prediction tool by working closely with the Rectifi Project based at Swansea University. They are looking to use an automated characterisation process to collect composition data of steel scraps and link them with images. To allow an automated steel scrap characterisation pipeline, we collected data and built an object detector for steel scraps on a conveyor belt for the Rectifi project.
Potential for Industrial Impact
The research would provide a computer vision-based tool for steel scrap quality assessment. The impact would be manifold. It would allow the metal recycler to quickly assess the cleanliness of their products and apply re-sorting if necessary. It would provide a standardised quality assessment framework between the metal recycler and steelmaker to evaluate and agree the value of the steel scraps by quality. It can also serve as a monitoring tool for daily operation for steelmakers to monitor the scraps charged into a furnace. Overall, this would allow easier use of steel scrap in steelmaking and eventually would encourage more steel scraps to be reused for steelmaking, moving the whole society to be more sustainable and getting closer to the goal of net-zero.