A new ELM based on interval-value for modeling in industry systems

In actual industry systems, some input/output variables of the control and optimization models may be not crisp instead of interval-valued. This paper proposes a new Extreme Learning Machine based on interval-value(I-ELM). The model is composed of two parts, one is an ordinary ELM for modelling the midpoint of the interval-valued data, the other is a modified ELM for modelling the width of the interval-valued data. In the modified ELM model, the constrained least-squares estimation method is used to obtain the output weights. Also, Marzullo sensor fusion algorithm is introduced into the ELM model to improve its prediction accuracy. Results of numerical comparison based on data from an actual continuous casting process show the usefulness of the proposed ELM model based on interval-value.