The burden distribution plays a key role in the daily control of the blast furnace due to its effect on the gas distribution and the chemical and thermal efficiency of the dry part of the shaft, as well as on the shape and level of the cohesive zone. Because of difficulties to measure the burden distribution on-line, operators base their control measures on know-how and experience from cold model trials, and indirect on-line information, such as gas temperatures and composition measurements from above-burden or in-burden probes. An abundance of possible charging programs makes the decisions of control actions difficult. This paper describes a method where a genetic algorithm is applied to automatically evolve novel charging programs. The programs are interpreted by a mathematical model of the burden distribution, and the differences between the desired and calculated radial distributions of ore (sinter and pellets) and coke are minimized subject to constraints of the charging (movable armor settings, skip size, etc.). The method is used on a set of examples where the charging programs evolved by it are illustrated and analyzed.
|Tidskrift||Transactions of the Indian Institute of Metals|
|Status||Publicerad - 2006|