A genetic algorithm evolving charging programs in the ironmaking blast furnace

In the ironmaking blast furnace, the distribution of the charged burden plays an important role because it influences the gas distribution in the shaft and the shape and the position of the cohesive zone. Because of enormous mechanical wear and high temperatures and pressure, the possibilities to reliably measure the distribution in real time are severely limited. Even though devices that provide information about the burden surface level have been developed, the high investment and maintenance costs make them economically infeasible in small or medium-size blast furnaces. A simplified first-principles model of the burden distribution forms the basis of the work presented in this article. A method is proposed by which a desired radial ore-to-coke distribution can be achieved by developing charging programs by a genetic algorithm, which was found to be a technique that can tackle this complex and nondifferentiable optimization problem. The algorithm evolves different charging programs subject to practical constraints of the charging (such as maximum skip size and movable armor spans), with the goal to find a charging program that minimizes the differences between the desired and calculated burden distribution. The article describes the method and presents a few illustrative examples on charging programs evolved by it.