A Hybrid Approach Variable Selection Algorithm Based on Mutual Information for Data-Driven Industrial Soft-Sensor Applications


 The development of virtual sensors predicting the desired output requires a careful selection input variables for model construction. In an industrial environment, datasets contain many instrumentation system measures; however, these are often non-relevant or excessive information. This paper proposes variable algorithm based on mutual information examination, redundancy analysis, and reduction soft-sensor modeling. A relevance calculation is performed in first stage to select important using criterion. Then, detection exclusion redundant carried out, penalizing undesired variables. Finally, most relevant subset determined through wrapper method Mallowssans' Cp metric assess fitting prediction performance. approach was successfully applied estimate ethanol concentration distillation column process adaptive network-based fuzzy inference architecture as non-linear dynamic regression model. comparative study considering application correlation analysis proposed this study. Simulation results show effectiveness providing search suitable models that achieve faster developing soft oriented applications.