Nonlinear soft sensing modeling by combining multiple RBFN-based models

A new method (named MRBF algorithm) for nonlinear soft sensing modeling is proposed. Multiple RBFN-based models are combined by the fuzzy c-means clustering (FCM) algorithm for nonlinear modeling. The MRBF algorithm has been evaluated by a nonlinear function example and a practical problem of modeling the product quality in hydrocracking fractionator. The detailed algorithm of MRBF method is presented and the comparison results between the MRBF and RBFN are also provided. It shows that MRBF method achieves better generalization result and costs shorter training time than pure RBFN