Document Type: Original Article
Department of Chemical Engineering, Petroleum University of Technology, Abadan, P.O. Box 63187, Iran
Catalytic pyrolysis of plastic wastes is recently deemed as an efficacious way to tackle the severe issue of plastic waste accumulation in the globe. In order to extend the pyrolysis process to industrial scale and optimize its operational conditions, development of suitable models seems crucial. In current study, time series modeling has been investigated for a semi industrial pyrolysis employing two distinctive multivariate modeling methods namely, Least Square Support Vector Machine (LS-SVM) and Artificial Neural Network (ANN). Sufficient experimental data sets were utilized to form both LS-SVM and ANN models effectively. After proper training of both models, two unseen test data sets were applied to measure the predicting performance of developed models. In both training and prediction sections, accuracy of both models was analyzed using two statistical error measuring methods namely, the Mean Squared Error (MSE) and the coefficient of determination (R2). Both methods achieved the R2 of more than 0.99 and the MSE of less than 0.1 in training part. In prediction part, the resulted R2 values for both models were analogous to values obtained in training part while the MSE values were slightly more than those values were achieved in training part. Finally, obtained outcomes from training and prediction parts revealed that both ANN and LS-SVM methods are quite reliable and precise tools for time series modeling of pyrolysis process.