Generating Anomalous Regenerable CO2 Removal System Data for Environmental Control and Life Support System Self-Awareness

Human spaceflight beyond Earth orbit will require autonomous deep space habitats that can keep the crew alive when present and keep the habitat "alive" when not. To achieve this goal, the autonomous agent must be both self-aware and self-sufficient. A self-aware Environmental Control and Life Support System (ECLSS) that can perform diagnostics and failure prognostics will be especially crucial towards enabling autonomy. A machine learning-based autonomous agent requires time-dependent data to train, test, and evolve the algorithm. Unfortunately, such data are not available during nominal or anomalous ECLSS operations. The Simulation Testbed for Exploration Vehicle ECLSS (STEVE), a 13X zeolite sorbent bed with CO2-laden simulated cabin atmosphere flow, was developed along with a Simulink and Aspen Adsorption-based computational model of STEVE to produce data of a regenerable CO2 removal system. Experiments and simulations can be conducted at nominal operating conditions and with faults to rapidly generate a diverse set of data. This paper describes the design and development of STEVE and the corresponding computational models. We recommend guidelines for generating data to develop machine learning algorithms for ECLSS self-awareness.