Generalized regression and feed forward back propagation neural networks in modelling flammability characteristics of polymethyl methacrylate (PMMA)

The capability of artificial neural networks in predicting microscale combustion calorimeter (MCC) parameters of polymethyl methacrylate (PMMA) was carried out in this study. Using values of sample mass and corresponding heating rate, feed forward back propagation (FFBP) and generalized regression neural network (GRNN) models were developed to predict MCC parameters. On the whole, GRNN outperformed FFBP in predicting HRC data while FFBP model saw an improvement over GRNN when estimating pTime. It was also discovered that GRNN obtained better THR, pTemp and pHRR predictions during training but generated a relatively poor correlation when estimating the testing data. Sensitivity analysis on the ANN models revealed that heating rate had a more significant effect on the models' outcome. Also, the ANN models observed the least error deviation when compared with HRC results for PMMA from structure-property models. Hence, ANN presents a reliable method for predicting flammability characteristics of PMMA from MCC test.