Abstract: Based on the idea approach based on multi-model with a stronger robustness and precision, combining with strong coupling and the highly nonlinear characteristics in multi-effect evaporation process, a soft sensor modeling method of fuzzy kernel clustering was proposed based on differential evolution and multi-squares support vector machines. The method made clustering samples under different conditions by use of DE-KFCM, each sub-models were constructed using LSSVM using fuzzy clustering algorithm to cluster the input data under different conditions, and finally get the system output by weighting. Taking alumina evaporation process as an example, for the soft sensor modeling of export of feed concentration, the better results have been obtained.

Key words: DE algorithm, least squares support vector machine, evaporation process, KFCM