Application of fuzzy inference system to polypropylene business policy in a petrochemical plant in India

Polypropylene is a versatile thermoplastic resin available in a wide range of formulations for engineering applications. This paper presents a new approach to predict the quality of polypropylene in petrochemical plants. A model is constructed based on a large number of data collected from a renowned petrochemical plant in India and used to predict the polypropylene quality through the proposed approach. The quality of polypropylene depends on the indices like melt flow index and the xylene solubility of the product. The parameters controlling these two indices are hydrogen flow, donor flow, pressure and temperature of polymerization reactors. Using these four input and two, output parameters, four Mamdani fuzzy inference systems are formed depending on the different membership functions of the variables. The model outcomes are then compared with the collected plant data and a sequence of statistical data analyses selects the most suitable model among them. Some sensitivity analyses with respect to some parameters are also performed to validate the proposed models. The raw materials for producing polypropylene are very much costly specially the catalyst teal. So if the desired grade of PP is not achieved by the trial and error run then the production cost becomes uncontrollable. With the help of our proposed approach by controlling some parameters during the production phase, the quality of polypropylene can be improved. Thus it will save both time and cost by reducing the yield of non-prime, off-grade products obtained by the conventional procedure. (C) 2015 Elsevier Ltd. All rights reserved.