This paper presents a soft sensor model for a high-pressure digestion process to control the raw material proportioning for the bauxite slurry blending process in the alumina metallurgical industry. By dividing the sample data set into several clusters with an improved rival penalized competitive learning clustering algorithm, a distributed support vector machine-based soft sensor is presented to measure the quality of the digested slurry online. Based on expert knowledge and the mechanism of the blending and digestion process, a hybrid expert control system for supervisory control of the blending process is developed to optimize the raw material proportioning. Both the experiments and the industrial applications demonstrate the feasibility and effectiveness of the soft sensor and the developed expert control system. (c) 2013 Elsevier Ltd. All rights reserved.