Modeling and predicting of the flash point of chemical compounds

Flash point is one of the most important flammability characteristics of chemical compounds. In the present study, we developed a neural network model for accurate prediction of the flash point of chemical compounds, using the number of hydrogen and carbon atoms, critical temperature, normal boiling point, acentric factor and enthalpy of formation as model inputs. Using a robust strategy to efficiently assign neural network parameters and evaluate the authentic performance of the neural networks, we could achieve an accurate model which yielded average absolute relative errors of 0. 97, 0. 96, 0.99 and 1.0% and correlation coefficients of 0.9984, 0.9985, 0.9981 and 0.9979 for the overall, training, validation and test sets, respectively. These results are among the most accurate ever reported ones, to date.in this article method for selection the best learning algorithm and transfer function are clearly presented and relative error for these parameters are represented in detail .