A novel signal processing approach enabled by machine learning for the detection and identification of chemical warfare agent simulants using a GC-QEPAS system.

Abstract

The detection and identification of chemical warfare agents (CWAs) present challenges in emergency response scenarios and for safety and security applications. This study presents the development and validation of an innovative analytical method using a gas chromatography (GC) and quartz-enhanced photoacoustic spectroscopy (QEPAS) sensor for the detection of stimulants for six CWAs. Following the guidelines of the European Network of Forensic Science Institute (ENFSI) and the Commission Implementing Regulation (EU) 2021/808, the analytical method was validated. The validation results demonstrated the robustness and reliability of both the GC and QEPAS modules. Moreover, with regard to the toxicological threshold levels, this study highlights the efficacy of a prototype of a portable device for real security and safety applications. Furthermore, a machine learning (ML) approach was developed to automate the detection and identification of CWAs' stimulants. The workflow involved two interconnected stages: detection based on chromatographic retention times (RTs), and identification using infrared (IR) spectra through the one-class support vector machines classifier. The classifier was activated only after obtaining a positive detection based on RTs. The results highlight the ML model's effectiveness in CWA detection and identification, combining RT analysis and IR spectrum classification, achieving 97% accuracy at a 95.5% confidence interval and 99% accuracy at a 99.7% confidence interval; this result demonstrates the model's utility for real-world security and safety applications for CWAs.