Soft Sensing Of Polypropylene Melt Index

Melt index is a quality index of polypropylene product. Nevertheless, it is difficult to measure on line. It will affect the quality of production and stability of the system. Using Spheripol II polypropylene unit as a research background, several soft sensors were developed by combining first principle modeling and data-driven modeling methods. Systematic and in-depth study was carried on a number of important aspects of the soft-sensing technique for the Spheripol II polymerization process.Soft-sensing technique of polypropylene melt index was briefly described on the theoretical research and application prospects. Spheripol II polymerization process and propylene polymerization reaction mechanism were introduced. Mechanism modeling for loop reactor was proposed. And in accordance with the reactor model, the response of the system was analyzed after the changes in operating conditions. Through the preliminary prediction on field density and temperature, the correctness of the model was verified.Two mechanism models for dual-loop reactor were established. According to the mechanism of propylene polymerization, the framework of the mechanism model for single-loop reactor was derived. Based on this model and blend equations, two mechanism models for dual-loop reactor were derived. The recursive least square method with forgetting factor (FFRLS) and particle swarm optimization (PSO) algorithm were used to identify unknown model parameters respectively. And the validity of the models was verified through on-site collective data.Two hybrid soft sensors combining the established mechanism model and intelligent method were presented. The mechanism model was combined to strengthen Fuzzy C-Means (FCM) clustering algorithm and Fuzzy Inference System (FIS). Considering traditional FCM clustering algorithm requires determining the clustering number firstly and is sensitive to the initial value, PBMF indicator was used to determine the optimal clustering number, besides using Cluster Center Initialization Algorithm (CCIA) to determine the initial clustering center. In order to optimize the structures of FIS, subtractive clustering was used to simplify fuzzy rules. Besides, Particle Swarm Optimization (PSO) algorithm was used to optimize its parameters. Validations of industrial data show that the soft-sensors achieve good predictions.Finally, a conclusion of the full text was made, besides summarizing the solved problem and the harvest. Some future research areas were highlighted.