The application of neural networks to the assessment of impact testing on explosives

Abstract The BAM Fallhammer test is one of the most widely used impact tests, but the results of the test rely on the subjective assessments of the test operator using the senses of hearing, smell and sight. The paper describes an investigation into the suitability of using a neural network for determining the outcome of BAM Fallhammer tests. The network utilises digital data from instrumentation installed around the BAM apparatus, including a microphone and a gas sensor. Selected examples are given to show that neural networks have the potential to distinguish between the test results of ‘no reaction’, ‘decomposition’, and ‘explosion’ for propellants, plastic and high explosives. The technique removes the possible operator dependence of assessments and the study in general may also lead to clarification of the categories involved in defining the test outcome.