Prediction of polypropylene business strategy for a petrochemical plant using a technique for order preference by similarity to an ideal solution‐based artificial neural network

Abstract

AbstractThe quality of polypropylene is one of the major components in the plastic industry. The quality of the polypropylene depends upon the melt flow index and the xylene solubility under the given condition such as hydrogen flow, donor flow, pressure, temperature, etc. This study investigates the use of artificial neural network (ANN) modeling for the prediction of the quality of polypropylene for petrochemical plants. This study proposes an ANN model to predict the “melt flow index (MFI) and xylene solubility” as the quality of polypropylene depends upon these two factors. Hydrogen (H2) flow, pressure, temperature, and donor flow are the controlling parameters for the MFI and xylene solubility. The study proposes a new approach for the selection of the best topology using the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) multicriterion decision‐making (MCDM) process for the ANN technique. Experimental data are trained with three training algorithms each with four combinations of training functions and each with three combinations of numbers of neurons. In this study, the best model for ANN is found by the Levenberg–Marquardt backpropagation training algorithm with logarithmic sigmoid and hyperbolic tangent sigmoid transfer functions. The best topology for the ANN model for prediction is selected by using the TOPSIS method. Some sensitivity analyses are provided graphically to show the physical nature of the problem.