A combined phenomenological artificial neural network approach for determination of pyrolysis and combustion kinetics of polyvinyl chloride

Abstract

As a widely used plastic material polyvinyl chloride (PVC) accounts for a significant amount of plastic waste but also offers great potential in conversion to chemical feedstock via pyrolysis process. However, development of a sensitive mathematical approach is required for proper process design and monitoring of thermochemical conversion processes. In this work, we attempt to develop an artificial neural network (ANN) model for estimation of mass loss as a function of temperature and heating rate during pyrolysis and combustion of PVC. For this purpose, pyrolysis and combustion characteristics were quantified using thermogravimetric analysis, then non-isothermal kinetics were analysed by iso-conversional models. The results of ANN models show that this method helps predict complex systems with high regression coefficient (R-2) values. The best performed model analysed by ANN for pyrolysis was NN 7 with R-2 = 0.9993, the best performed model for combustion was NN 10 with R-2 = 0.9982. Comparison of experimental results to ANN predictions indicates that ANNs with a quick propagation algorithm can be an effective approach for modelling complex non-linear systems such as thermal degradation of thermoplastics.