Modeling Process Safety Scenarios
Global and local quantitative structure–property relationship models to predict the impact sensitivity of nitro compounds
Version of Record online: 5 JUN 2012
Copyright © 2012 American Institute of Chemical Engineers (AIChE)
Process Safety Progress
Volume 31, Issue 3, pages 291–303, September 2012
How to Cite
Fayet, G., Rotureau, P., Prana, V. and Adamo, C. (2012), Global and local quantitative structure–property relationship models to predict the impact sensitivity of nitro compounds. Proc. Safety Prog., 31: 291–303. doi: 10.1002/prs.11499
- Issue online: 14 AUG 2012
- Version of Record online: 5 JUN 2012
- quantitative structure–property relationships;
- density functional theory;
- applicability domain;
- impact sensitivity;
- nitro compounds
New quantitative structure–property relationships were developed to predict accurately the impact sensitivity of nitro compounds from their molecular structures. Such predictive approaches represent good alternative to complete experimental testing in development process or for regulatory issues (e.g., within the European REACH regulation).
To achieve highly predictive models, two approaches were used to explore the whole diversity of nitro compounds included in a dataset of 161 molecules. In the first step, local models, dedicated to nitramines, nitroaliphatics, and nitroaromatics, were proposed. After that, a global model was developed to be applicable for the whole range of the nitro compounds of the dataset.
In both cases, large series of molecular descriptors were calculated from quantum chemically calculated molecular structures, and multilinear regressions were computed to correlate them with experimental impact sensitivities. Both the global and local models could predict nitramines and nitroaliphatics in high accuracy whereas nitroaromatics were more difficult to be predicted due to their complex decomposition mechanisms.
The proposed models were validated in the perspective of potential regulatory use according to the OECD principles, including internal, external validation, and the definition of their applicability domain. So, they could then be used for prediction either separately or in a consensus approach. © 2012 American Institute of Chemical Engineers Process Saf Prog, 2012