Electronic Nose for detecting Acetone as a Potential Precursor in Triacetone Triperoxide (TATP)-based Improvised Explosive Devices (IEDs)

Abstract

The paper aims to fabricate an electronic nose for detecting volatile organic compounds (VOCs) emitted by acetone as a potential precursor of Triacetone Triperoxide (TATP) using Artificial Neural Network (ANN) and Principal Component Analysis (PCA). Specifically, it aims to build an array of sensors aided with Raspberry Pi and Python programming language to implement the necessary algorithms, together with the documentation of gathered information to conclude the effectiveness of the system. ANN is used for pattern recognition and classification, while PCA is for feature reduction and extraction. Sensors are selected by PCA through dimension reduction of signals obtained, which resulted to sensors MQ8, MQ136, MQ7 having the highest retained information of 83.07%, 8.79% and 7.27%. The classifier was trained by feeding acetone with varying concentrations from 5% to 100% with their corresponding expected classifications that resulted to 0.02% of training error. Using a Confusion matrix, accuracy was determined which assesses the classification between the E-nose and expected classification of commercial products. This study concludes that the fabrication of E-nose with PCA and ANN algorithms produces promising results of classifying acetone as a potential TATP precursor, with an overall accuracy of 87.88%.