Predicting the impact sensitivity of explosive molecules using neuromimetic networks

Area of Science:

• Computational chemistry
• Materials science
• Chemical engineering

Background:

• Predicting the impact sensitivity of explosive molecules is crucial for safety and material design.
• Traditional methods often rely on empirical data and may lack predictive accuracy for diverse molecular structures.

Purpose of the Study:

• To introduce a new computational method for predicting the impact sensitivity of explosive molecules.
• To evaluate the performance of a neural network approach using specific molecular descriptors.

Main Methods:

• A neural network model was developed using formal neurons.
• The model incorporated molecular descriptors such as oxygen balance and enumeration of specific chemical groups.
• A dataset of 124 explosive molecules from various chemical families was used for training and validation.

Main Results:

• The neural network model achieved a satisfactory prediction accuracy, correctly classifying 80% of the molecules.
• The classification was performed on a scale of four distinct sensitivity levels.
• Comparison with multivariate linear regression analysis indicated a slight performance advantage for the neural network method.

Conclusions:

• The developed neural network method offers a promising approach for predicting explosive molecule impact sensitivity.
• The model's ability to correctly classify a high percentage of molecules suggests its utility in safety assessments and new energetic material development.
• The findings highlight the potential of artificial intelligence in advancing the field of energetic materials research.