QSPR and Artificial Neural Network Predictions of Hypergolic Ignition Delays for Energetic Ionic Liquids

Abstract

Due to their negligible volatility, energetic ionic liquids are being considered as next generation hypergolic fuels for replacing toxic monomethylhydrazine. One design challenge for energetic ionic liquids is to maintain their ignition delays as close to that of monomethylhydrazine. The ignition process of ionic liquids with an oxidizer, such as nitric acid, is a complex process and, to date, there are no theoretical methods for predicting the ignition delay. The present work examines two correlation methods, Quantitative Structure Property Relationship (QSPR) and Artificial Neural Networks (ANNs), for their ability to predict this quantity. A set of five descriptors were chosen from a pool of more than 160 to establish these correlations. A good QSPR correlation was obtained using these descriptors. We then trained an artificial neural network and examined the predictive ability of the network using an extensive 5-fold cross validation process with the same set of descriptors. A number of data normalization techniques were examined for network training and validation. The results show that ANNs exhibit excellent prediction capabilities for this application.