Process Control & Automation
Global industry, including steel, faces parallel crises of worsening shortage of skilled workers with accelerated retirement rates of skilled trades holding critical knowledge. Add to these challenges the evolution of communication preferences from written words to digital imagery. State-of-the-art augmented reality (AR) systems are proven to ease these challenges. This paper will provide data of the challenges along with a practical overview of current AR technology, evidence of how AR has and can ease challenges of skills, available time, retaining knowledge, safety and bridging generational challenges, amongst others.
Digital twin solutions are set to play a very active role in many digital transformation and Industry 4.0 plans. Existing implementations of digital twins have a wide breadth of scope and usage, from basic data viewing, to large, centralized predictive control setups. Exploring the use of widely available tools in steelmaking to create a combination of modular and specialized interfaces for a pilot process and improving on some existing development may help create designs that help with analyzing data, improve operator understanding of obscure data, and look toward other new tools and possible process implementations.
This project team was the recipient of the 2021 AIST Digital Technologies for Steel Manufacturing Grant.
In this paper, the application bottleneck of big data in blast furnace (BF) ironmaking will be revealed, characteristics and problems of original data in an actual commercial BF will be sorted out, and the optimization technical measures will be put forward. The results show that a standard process should be carried out from the collection of original data to data application, and the centralized control platform was established to support the application of data.
The model training pool and ensemble learning were employed to adjust the big data algorithm.
The application effect after optimization was evaluated, and the improved accuracy proved that the optimization measures were effective.
The vacuum degassing (VD) process decreases the hydrogen content in molten steel, improving steel quality. One problem is that no tool allows real-time monitoring of hydrogen content in molten steel. The application of data analytics methods can help predict hydrogen content, reducing the time and energy consumption of the VD. For this reason, a machine-learning model was developed to estimate hydrogen content in molten steel, achieving a mean absolute error of 0.1298, 0.2400, and 0.2179 ppm for training, test, and production data, respectively. Furthermore, statistics parameters display acceptable results, showing a robust model capable of generalizing the results for new data entry. This real-time prediction allowed for the decrease of hydrogen content in the final product by helping the operator to make better decisions through the visualization of live operational variables, thus generating recommendations of process variables, increasing the quality of the manufactured steel, increasing the productivity of the vacuum degassing process, and reducing energy consumption in the steelmaking process.
Stirring generated by low gas flowrates helps improve steel quality and minimize oxidation. The soft bubbling phenomenon has been studied using a combination of multiple microphones and image sensors in both physical modeling and industrial trials. In the cold model, stirring was monitored under a range of conditions including closed-eye situations. In addition, ladle eye sizes measured at thicker oil heights indicated different trends than previous studies. Overall, the study suggested that an acoustic approach can provide a more robust signal for closed-ladle-eye stirring at low gas flowrates compared to image sensors. In particular, the closing of an eye is easily detected from a sound signal. Further testing in an industrial environment is required.
The Gerdau Special Steel North America Fort Smith plant has expanded use of the ibaPDA system from the initial data acquisition for troubleshooting functionality to a central component of a data analytics platform. With the upgrade to a historical data server, the ibaPDA system provides the data lake foundation as well as the first level of analytics for data cleaning, filtering and process variable aggregation. Process events triggered upon data ingestion are used to automatically execute statistical analyses or machine-learning models implemented using open-source libraries for automatic notifications, anomaly detection and process diagnostics. The integration with video cameras further expanded troubleshooting capabilities.