Abstract: Shaft furnace roasting in iron mineral processing is a synthetic complex process, whose product quality index, i.e. magnetic tube recovery rate (MTRR), is hard to measure online, and the optimizing control of such process is very difficult. Based on analysis of its dynamic characteristic and manual operation mode, a multivariable intelligent optimizing control approach is developed, combining intelligent method, forward-feed control andfeedback control. The target is the optimization of integrated production indices including MTRR, hour's yield and consumed gas. The setpoints of control loops of the process are adjusted online automatically according to the targeted integrated production indices and variations of boundary conditions, so the optimizing control is achieved. The proposed approach has been applied to a roasting process of a mineral processing factory of China,and optimization of the integrated production indices is realized and remarkable benefits are obtained.