Abstract
Condition monitoring of the blast furnace system plays an important role in the safe and efficient production of high quality hot metal. This article proposes to use a state space model for monitoring the dynamic blast furnace system. The blast furnace data is assumed to be generated by some independent non-Gaussian source signals and a state space model is used to extract the source signals. An optimization problem with the objective function of minimum Kullback-Leibler divergence, i.e., maximum independence between the source signals is constructed. The system matrix and non-Gaussian signals are obtained by solving the optimization problem. Based on the extracted signals, the support vector data description (SVDD) is used for constructing monitoring statistics. Operational data collected from a real blast furnace containing both normal and faulty data are analyzed and used to test the proposed monitoring strategy. The proposed method is then compared with the dynamic independent component analysis (DICA) based monitoring strategy. It is shown that the state space model based monitoring strategy is more appropriate for monitoring of blast furnace faults.
