� DIGITALIZATION � tapping temperature. These parameters are critical to ensuring stable production and high-quality output. Traditional models struggle to accurately predict these conditions due to the complex chemical and physical dynamics of blast furnaces. However, Primetals Technologies has developed a machine learning model that uses existing instrumentation to predict key performance indicators with remarkable accuracy over a two-hour forecast period. By visualizing a probability funnel for predicting hot metal quality parameters, the solution enables operators to trust AI-driven insights, fostering confidence in the new technology and changing the way humans interact with systems in the optimization phase.
Another example where data analytics methods are key to achieving better results is determining the optimal EAF operating strategy in terms of scrap cost, conversion cost, or tap-to-tap time, depending on the current market situation. Operating an EAF without exploiting the potential of large production data will result in suboptimal performance. Attempting to analyze exported large production data in Excel will cause a loss of focus on properly identifying variances, and information from time series data will be missing. Primetals Technologies offers EAF best practice analysis services based on a selfdeveloped big data filtering and big data sorting engine, which assists EAF operators to achieve their individual goals by finding the operational best practice based on actual heats. First, scrap qualities are clustered, then a reference heat is defined for each cluster. Similar heats in historical data are identified, Level 1 data from these similar historical heats are analyzed, and finally a new EAF best practice is derived. The database consists of thousands of heats over a period of years and a time series with millions of rows.
A third application of AI methods is quality control throughout the entire production chain of a steel
plant. Fast identification of quality deviations and fast definition of corrective and compensatory actions are key to achieving a high yield of products that precisely meet quality targets. Primetals Technologies has developed the digital know-how-based quality control solution through-process quality control( TPQC). TPQC collects seamless quality data along the entire production process(“ through process”) by connecting to various data sources in a steel plant. These data points create a product genealogy that provides insight into signal details, heat maps, and product and signal comparisons visualized in various dashboards. Using artificial intelligence to analyze this vast amount of collected data, TPQC predicts and analyzes mechanical properties and identifies root causes in a timely manner.
The transition phase: Embracing new routes
In the transition phase, the steel industry faces the challenge of operating in
Prediction results of hot metal quality parameters powered by AI and using quality indication( circled in red).
Green Steel World | Issue 17 | April 2025 35