Abstract: Operating regime based soft sensing model of polypropylene melt flow rate (MFR) was developed.First, different operating regimes were determined according to the concentration of hydrogen in the reactor.Then for each operating regime, a nonlinear PLS (NLPLS) model was built by combining PLS with RBF networks.Finally, all the sub-models were combined to formulate the global model. To deal with the problem that the process changed with time, a recursive nonlinear PLS (RNPLS) algorithm was proposed to update the model to adapt process changes.In addition, an on-line model estimation and updating procedure based on moving window approach was proposed.The proposed method was used in an actual plant, and the results showed that the model achieved quite good performance and the accuracy of the model could satisfy the practical need.