Melt index(MI)is the most important parameter in determining polypropylene’s specifications and also the key of quality control.Considering the continuity of the polypropylene production process and the characteristics of gray prediction model

a new GMC(1

N)melt index prediction model is proposed in this paper to infer the MI of polypropylene from other process variables

which shows good applicability. Furthermore

a back propagation network is built to correct the residual of the model and the GMC(1

N)-BP model is improved with better precision.Standard SVM(support vector machines)

LS-SVM and weighted LS-SVM model are taken for comparison.A practical polypropylene industrial process is taken as case study.The results show that the proposed improved model achieves good results in practical melt index prediction.