An Integrated CBR Model for Predicting Endpoint Temperature of Molten Steel in AOD

Abstract

An integrated Case Based Reasoning model is proposed to predict the endpoint temperature of molten steel in AOD. Case's problem part is represented by a set of feature attributes and a set of state attributes. The boundary value of state attributes can be obtained by M5P. Case similarity is computed based on Grey Relational Degree with different weights of attributes in order to solve the problem of obtaining the accurate results with incomplete information. Entropy Weight Method is adopted to determine the weights of attributes. A two-step case retrieval, composed of rough search and delicate search, is provided to decrease the search time greatly. Seven methods, Multiple Linear Regression, M5P, Artificial Neural Network, Cased Based Reasoning with case similarity based on Euclidean Distance and Equal Weights of attributes (CBR_ED_EW), Cased Based Reasoning with case similarity based on Euclidean Distance and Different Weights of attributes (CBR_ED_DW), Cased Based Reasoning with case similarity based on Grey Relational Degree and Equal Weights of attributes (CBR_GRD_EW) and Cased Based Reasoning with case similarity based on Grey Relational Degree and Different Weights of attributes (CBR_GRD_DW), are employed for a comparison. The results show that CBR_GRD_DW outperforms the other methods, and the integrated Case Based Reasoning model is effective in predicting the endpoint temperature of molten steel in AOD.