[Effects of microbial degradation effects based on generalized regression neural network on the detection of accelerants].

To explore the effects of microbial degradation at fire scenes on the identification of accelerants, samples of general and nutrient soil were injected into an accelerant, considering sealed storage time as a variable. The accelerant residues in the samples were identified through gas chromatography-mass spectrometry (GC-MS) with passive headspace pretreatment. The result indicated that the composition of the accelerants was changed due to microbial degradation, with varying degrees of degradation degree in different soil samples. Degradation in the general soil samples was superior to that in nutrient soil samples. C9 to C12 n-alkanes and monosubstituted aromatic hydrocarbons were more easily degraded, compared to other components. With the increase of substituents, the degradation of polysubstituted aromatic hydrocarbons became increasingly difficult. Principal component analysis (PCA) was used for data dimension reduction based on the type of soil samples. Generalized regression neural network (GRNN) analysis classified the soil samples with 100% accuracy.