Fuzzy mechanistic model with neural compensation for Estimation of shaft furnace’s product quality

Usually grey-box modeling has better accuracy than black-box identification. The grey-box can be regarded as combination of mechanistic modelling and intelligent identification. But in many cases, mechanistic models are not available, for example the production quality of the shaft furnace which will be discussed in this paper, we can use fuzzy technique to obtain the mechanistic models that can be checked by the physical meanings. In this paper, we propose a novel modeling approach for complex nonlinear systems, it has a fuzzy mechanistic model and a neural compensator. For the training of the neural network, we propose a new fast and stable algorithm. Finally, the above method is successfully applied on estimation of the production quality of the shaft furnace.