The purpose of this work was to investigate the utility of electronic
aroma detection technologies for the detection and identification of i
gnitable liquid accelerants and their residues in suspected arson debr
is. Through the analysis of ''known'' accelerants and residues, a trai
ned neural network was developed for classifying fire debris samples.
Three ''unknown'' items taken from actual fire debris that had contain
ed the fuels, gasoline, kerosene, and diesel fuel, were classified usi
ng this neural network. One item, taken from the area known to have co
ntained diesel fuel, was correctly identified as diesel fuel residue e
very time. For the other two ''unknown'' items, variations in sample c
omposition, possibly due to the effects of weathering or increased sam
ple humidities, were shown to influence the sensor response. This mani
fested itself in inconsistent fingerprint patterns and incorrect class
ifications by the neural network. Sorbent sampling prior to aroma dete
ction was demonstrated to reduce these problems and allowed improved n
eural network classification of the remaining items which were identif
ied as kerosene and gasoline residues.