@article{MuhrKlapoetkeHoll2022,
  author    = {Matthias Muhr and Thomas M. Klap{\"o}tke and Gerhard Holl},
  title     = {Sensory Monitoring of Drop Hammer Experiments with Multivariate Statistics},
  series   = {Propellants, Explosives, Pyrotechnics},
  volume    = {47},
  number    = {11},
  publisher = {Wiley-VCH},
  address   = {Weinheim},
  issn      = {0721-3115},
  doi       = {10.1002/prep.202200025},
  url       = {https://nbn-resolving.org/urn:nbn:de:hbz:1044-opus-64201},
  year      = {2022},
  abstract  = {A precise characterization of substances is essential for the safe handling of explosives. One parameter regularly characterized is the impact sensitivity. This is typically determined using a drop hammer. However, the results can vary depending on the test method and even the operator, and it is not possible to distinguish the type of decomposition such as detonation and deflagration. This study monitors the reaction progress by constructing a drop hammer to measure the decomposition reaction of four different primary explosives (tetrazene, silver azide, lead azide, lead styphnate) in order to determine the reproducibility of this method. Additionally, further possible evaluation methods are explored to improve on the current binary statistical analysis. To determine whether classification was possible based on extracted features, the responses of equipped sensor arrays, which measure and monitor the reactions, were studied and evaluated. Features were extracted from this data and were evaluated using multivariate methods such as principal component analysis (PCA) and linear discriminant analysis (LDA). The results indicate that although the measurements show substance specific trends, they also show a large scatter for each substance. By reducing the dimensions of the extracted features, different sample clusters can be represented and the calculated loadings allow significant parameters to be determined for classification. The results also suggest that differentiation of different reaction mechanisms is feasible. Testing of the regressor function shows reliable results considering the comparatively small amount of data.},
  language  = {en}
}