Comparative evaluation of the predictability of neural network methods on the flammability characteristics of extruded polystyrene from microscale combustion calorimetry
Predictions of both combustible material flammability and heat release parameters have been long goals in fire safety research, for its complex heat, mass transfer and chemical reaction process in gas phase. In this study, neural network method is employed to predict materials flammability considering its wide application in predicting key properties of engineering problems. The use of group method of data handling (GMDH) and feed forward back-propagation (FFBP) neural networks in predicting the heat of combustion and heat release capacity (HRC) from microscale combustion calorimetry has been examined. The study presented models with excellent predictability though GMDH out-performed FFBP. The deviation of the predicted and measured HRC data from this study was compared with the results of other predictive modelling techniques used in flammability studies. The GMDH neural network results presented the least mean deviation of 4.01 signifying accurate predictions. Hence, this study proposed the use of GMDH in predicting flammability characteristics of materials.
KeywordsHeat of combustion Heat release capacity Microscale combustion calorimetry Group method of data handling Feed forward back-propagation neural network
This research is supported by the National Natural Science Fund of China, No. 51776098, and the Fundamental Research Funds for the Central Universities, No. 30918015101.
- 2.Shi Y, Xing W, Wang B, Hong N, Zhu Y, Wang C, Hu Y. Synergistic effect of graphitic carbon nitride and ammonium polyphosphate for enhanced thermal and flame-retardant properties of polystyrene. Mater Chem Phys. 2016;177:283–92. https://doi.org/10.1016/j.matchemphys.2016.04.029.CrossRefGoogle Scholar
- 4.Standard Test Method for Determining Flammability Characteristics of Plastic’s and Other Solid Materials Using Microscale Combustion Calorimetry. ASTM D7309-13.Google Scholar
- 5.Walters RN, Lyon RE. Microscale combustion calorimeter for determining flammability parameters of materials. Evol Technol Compet Edge. 1997;42:1335–44.Google Scholar
- 13.Arbib MA. The handbook of brain theory and neural networks, A Bradford Book. 2nd ed. Cambridge: The MIT Press; 2003.Google Scholar
- 14.Hecht-Nielsen R. Theory of backpropagation neural networks. Presented at IEEE proceedings, international conference on neural network, Washington, DC; 1989.Google Scholar
- 26.Li RYM, Fong S, Chong KWS. Forecasting the REITs and stock indices: group method of data handling neural network approach. Pac Rim Prop Res J. 2017;23(2):123–60.Google Scholar
- 27.Mueller JA. Automatic model generation based on GMDH. In: Kerckhoffs EJH, Snorek M, editors. Modelling and simulation 2001 (Proceedings of the 15th European Simulation Multiconference, Prague, 2001), SCS Europe 2001; 2001, p. 661–8.Google Scholar
- 29.Najafzadeh M, Barani GA, Hessami-Kermani MR. Group method of data handling to predict scour depth around vertical piles under regular waves. Sci Iran. 2013;20(3):406–13.Google Scholar
- 31.Schrijver MC, Kerckhoffs EJ, Water PJ, Saman KD. Model tuning with the use of heuristic-free GMDH (group method of data handling) networks. In: Proceeding 14th European Simulation Symposium, (c) SCS Europe BVBA, 2002.Google Scholar
- 37.May R, Dandy G, Maier. Review of input variable selection methods for artificial neural networks. In: Artificial neural networks-methodological advances and biomedical applications, vol. 10. Intech; 2011. p. 16004.Google Scholar
- 40.Mensah RA, Xu Q, Asante-Okyere S, Jin C, Bentum-Micah G. Correlation analysis of cone calorimetry and microscale combustion calorimetry experiments. J Therm Anal Calorim. 2018;10:1–11.Google Scholar