This dissertation is dedicated to the development of data-driven soft sensors for monitoring process variables in an electric arc furnace (EAF), a critical component in the steel scrap recycling process. Due to the nature of the process, continuous measurement of key process variables such as steel bath tapping temperature and dissolved oxygen content is challenging. These variables have a significant impact on the efficiency, energy consumption and overall productivity of steel recycling processes. Conventional methods for measuring these variables are invasive, disruptive to production and associated with high operational costs and energy inefficiencies.
The dissertation presents a novel fuzzy model soft sensor approach that utilizes fuzzy clustering and Takagi-Sugeno (TS) modelling in combination with particle swarm optimization (PSO) to estimate the bath temperature of the EAF. Achieving the prescribed temperature is crucial to achieve the desired steel quality and to ensure the appropriate steel properties for further processing. The presented method is based on initial temperature measurements and subsequent EAF inputs, enabling continuous estimation of the bath temperature throughout the refining process. The proposed model is characterized by a high prediction accuracy and suggests that it can significantly reduce the number of temperature measurements required, reduce energy losses and shorten tapping times.
A soft sensor for dissolved oxygen content in a steel bath, based on a TS model and prediction intervals (PIs), is proposed to predict unobserved output under difficult measurement conditions. Based on the amount of dissolved oxygen, operators can estimate the dissolved carbon content, which is a key parameter to ensure the quality of the steel and its mechanical properties. The inclusion of PIs provides a systematic way to account for system variability and allows the derivation of the narrowest possible band containing a prescribed percentage of the data despite the presence of heteroscedastic noise. A comparative analysis with actual EAF operational data demonstrates the effectiveness of these models in overcoming the challenges of scarce output measurements.
Finally, a novel key performance indicator for electric arcs, called the arc quality index (AQI), is proposed. A three-phase equivalent circuit integrated with the Cassie-Mayr (CM) arc model captures the nonlinear and dynamic characteristics of arcs, including the processes of arc breakage and ignition. A PSO technique is applied to real EAF data with current and voltage measurements to estimate the parameters of the CM model. The AQI is determined based on deviations from optimal arc conditions and provides a qualitative assessment of arc quality that includes aspects of arc stability and arc coverage. Consequently, operators can take targeted measures to improve the AQI, including adjustments to the power level or arc length and initiatives to increase the slag height, thereby optimizing operating conditions and energy utilization.
|
