Autonomous Identification and Quantification of Chemical Species with VCAM for use Onboard the ISS

Abstract

The vehicle cabin atmosphere monitor (VCAM) instrument is designed to autonomously detect and identify trace organic species in the international space station (ISS) cabin air and monitor changes in species concentrations over time after chemical events. The physical instrument is comprised of two subsystems. The first subsystem is a preconcentrator gas chromatograph (PCGC) which separates chemical analytes in time, based on compound specific properties such as molecular weight. The second subsystem is a Mass Spectrometer (MS) which measures the abundance of ionized analytes, separated in the GC phase, at specific mass-to-charge ratios. The VCAM PCGC/MS produces a time-series of mass fractionation patterns, indicative of the chemical compounds present, which is used for subsequent compound detection, identification, and quantification. In order to autonomously identify and quantify chemical species from the PGGC/MS data, VCAM employs a variant of the de-facto industry standard automated mass spectral deconvolution and identification system (AMDIS) algorithm developed by the National Institute of Standards and Technology (NIST). AMDIS was chosen first for its superior performance, when compared to a neural network classifier developed in-house and a proprietary, third-party, commercial algorithm, and second for its reputation within the mass spectrometry community. In this paper we provide an overview of AMDIS, including GC peak identification and spectral matching, as well our variations and additions to the core algorithm for performing mass calibration beforehand and species quantification afterward. We also discuss some of the challenges faced creating an independent implementation of AMDIS for delivery to VCAM flight software. Testing our algorithm, both individual components and in its entirety, was a particularly challenging, as the VCAM instrument was still in development and only periodically able to produce validation datasets.