Permeability prediction model for imperial smelting furnace based on improved case-based reasoning

Abstract

An intelligent prediction modeling approach integrating case based reasoning (CBR) with adaptive particle swarm optimization (PSO) is proposed for the permeability index prediction of smelting process in the imperial smelting furnace (ISF), to deal with the difficulties in describing the process with accurate mathematical models and information uncertainty. The case base is constructed directly from the production data. The cases most similar to the target case are retrieved from the case base, whose similarity measure is larger than a pre-specified threshold value. The result of the prediction model is obtained by reusing the solutions of the retrieved cases in weighted averaging. The weighted k-nearest neighbor algorithm (k-NN) is used in case retrieval, where the number of nearest neighbors and the weighted vector of features are optimized online using adaptive PSO to improve the retrieval accuracy of CBR. The experimental results of the industrial field production data show that the improved CBR model is better than the standard CBR and the model accuracy can satisfy the technological requirements.