A Fuzzy Associative Memory for the Classification of Chemical Warfare Agent Simulants

Abstract

This paper presents the development and testing of a fuzzy associative memory (FAM) architecture for use in the classification of chemical warfare agent simulants. A hybrid ion mobility spectrometry time-of-flight mass spectrometry (IMS(tof)MS) instrument was used to collect data for two chemical warfare agent simulants: dimethyl methyl phosphonate (DMMP) and tributyl phosphate (TBP). A preprocessor was developed to convert the raw IMS(tof)MS data file into a set of triplets containing the values for the mass, K0 (reduced mobility), and intensity for each point in the original 2-dimensional data set. Due to the small amount of available real data, synthetic data sets were also created. A classification system was constructed consisting of a FAM trained by either DMMP data or TBP data. Repeated experiments were run using different sample set configurations for training and testing. Experiment scenarios included instances where real data sets were used for training, and cases where synthetic data were used for training; the test sets contained a mixture of both real and synthetic data each time. Training was done with training sets as small as only a single sample. The results were excellent: the system was able to correctly classify the DMMP and TBP data, both real and simulated, 100% of the time.