Sintering optimization of manganese ore fines: A mixed multilinear regression and linear programming model

"Due to the randomness of all factors influencing the sintering operation, a two-stepped procedure is proposed to find the best way of using stocked fines at a plant producing Fe-Mn alloys for the steelmaking industry. The approach is based on the following: (1) predict the result of the sintering operation by establishing the best multilinear regression model using some of the characteristics of the loaded material; and (2) find the best solution under a minimizing cost criteria by solving a linear optimization problem using the regression equation plus constraints on the feed and product. IntroductionSome tests carried out using 20% to 30% of sinter in the material loaded into a 60 000 KVA Tysland-Hole electric furnace producing Fe-Mn alloys for the steelmaking industry indicated: (1) a meaningful decrease in the off-gas temperature and, consequently, a better energy use due to the material higher porosity; (2) more uniform heat distribution as a consequence of the electrode reaction zone enlargement; (3) an improved furnace control resulting in a smoother operation; and (4) a decrease in gas production facilitating the off-gas treatment. For some years, this plant used only the coarse fraction of the Mn ores, thus stocking all the received and own-generated fine particles. In view of the mentioned results, a strategy for increasing the use of stocked fines was formulated, including a step where it was necessary to decide which components and respective quantities should be used in the blends to be sintered (Almeida, 1984).It is well known that several factors affect productivity and energy consumption in sinter production, also influencing the product quality. Raw materials characteristics, namely chemical and size composition, affect permeability to gases and may or may not allow a deeper bed. In general, finer ores decrease permeability, increasing electricity consumption by fans, but due to their higher specific surface area, react more completely with other components in the mixture. For coke granulometry, a compromise must be obtained in order not to burn it too quickly nor to widen the flame front, therefore decreasing productivity. Decreasing coke in the blend also acts in sinter bonding, decreasing its consistency and increasing crushing and return fines. Moisture addition also influences components mixing, pre-agglomeration, bed permeability and, consequently, sintering productivity. Therefore, optimum conditions must be established for both a given sintering plant and the components in the feed (Dawson, 1993)."