Melt index prediction by RBF neural network optimized with an MPSO-SA hybrid algorithm

Abstract

Melt index (MI) is considered as an important quality variable which determines the product specifications, so reliable estimation of MI is crucial in the quality control of practical polypropylene (PP) polymerization processes. A novel MPSO-SA-RNN (modified PSO-SA algorithm and RBF neural network) MI prediction model based on radial basis function (RBF) neural network and artificial intelligent algorithms particle swarm optimization (PSO), and simulated annealing (SA) is presented, where the traditional PSO is modified first and then combined with SA to overcome the inherent defects in PSO and SA, and to achieve better optimization performance. The proposed optimization algorithm, MPSO-SA algorithm, is then used to optimize the parameters of the RBF neural network. Then the network is employed to build the MI prediction model, and the MPSO-SA-RNN model is thereby developed. Based on the data from a real plant, the approach presented above is evaluated and the research results confirm the validity of the proposed model, as well as the advantage of MPSO-SA algorithm to the traditional PSO and SA algorithms in handling optimization problems.