Modern process logging systems for electric arc furnaces have the capability of storing large quantities of data, both in terms of variables and number of heats. The objective of this research is to evaluate how this data can be used for improvement of operating practices and optimisation of energy and scrap utilisation. In previous research projects, a process visualisation and monitoring system was developed and installed on four meltshops in Scandinavia. The system consists of a real-time database, a historical database, operator screens and a report generation system. By showing relevant information to the operators during the melting operation and generating reports based on the performance of the different operating teams, reduction of energy consumption, increase of productivity and reduced production costs could be achieved. Further analysis of data from the historical database led to the development of process optimisation tools for post combustion, hot heel practice, slag foaming and charging of scrap baskets. With the use of these tools, energy consumption and power-on time could be decreased further. However, steel scrap is the most important raw-material in electric steelmaking, contributing between 60% and 80% of the total production costs. Today the degree of which the scrap mix can be optimised, and also the degree of which the melting operation can be controlled and automated, is limited by the knowledge of the properties of the scrap and other raw- materials in the charge mix. In this thesis, multivariate regression methods have been used to develop prediction models for final chemical analysis of the steel, total electric energy consumption and metallic yield. The predictions are based on composition of the raw-material mix, total consumption of injected materials (like oxygen, oil, coal, slagformers) and initial condition of the furnace (hot heel composition). A prediction model for total energy consumption for melting of the first scrap basket based on continuous measurements of electrical parameters and wall panel temperatures was also developed. The models have been used to estimate some scrap properties (chemical composition, specific electrical energy consumption, yield coefficients), evaluate the efficiency of post-combustion and to estimate the optimal time to charge the second scrap basket. The results show that it is possible to provide estimates of the levels of tramp elements (Cu, Sn, As, etc.) and alloy elements (Cr, Ni, Mo, etc.) in commonly used scrap grades based on evaluation of commonly logged process data. To ensure that the estimates remain consistent with scrap quality, it is suggested that they be updated on a regular basis. It is also discussed how the estimates of scrap properties can be used for improved process control and monitoring of scrap quality.
Luleå: Luleå tekniska universitet, 2005. , 65 p.