Automated Contingency Management for Water Recycling System

NASA’s exploration program envisions the utilization of a Deep Space Habitat (DSH) for human exploration of the space environment in the vicinity of Mars and beyond. Communication latency and extreme limitations of power and life-supporting resources make it imperative to operate the DSH systems in a highly autonomous fashion. One such system is the Environmental Control and Life Support System (ECLSS) which needs to be monitored and optimized to support its designated missions. Integrated System Health Management (ISHM) technologies have been developed in the past decades to provide a real-time assessment of system health and use this information to improve system safety and reduce operation cost. The goal of this study is to apply the ISHM principle to ECLSS, and bring in innovative solutions to address several key elements that NASA has been soliciting in its recent research announcements. To achieve this goal, Water Recycling System (WRS) deployed at NASA Ames Research Center’s Sustainability Base is selected as a testbed. Efforts are then made to develop a physics-based model with high fidelity fault simulations and data-driven model enhancements. With the system in place, an Automatic Contingency Management (ACM) system concept is introduced to embrace the fault diagnosis, prognosis, and automated fault accommodation through control reconfiguration and optimization, thus contingencies in the system are managed automatically without human in the loop. The ACM system architecture and key enabling techniques are developed. The fault diagnosis and prognosis task are accomplished by a novel method combining Lebesgue sampling (LS) and unscented Kalman filter (UKF) that allows computing resources to be allocated efficiently based on system health conditions without sacrificing performance. Multi-Stage PID and Time-Varying Weight Model Predictive Control (MPC) based fault mitigation strategies are designed for automatic fault accommodation which allows the integration of diagnostics and prognostics in the control strategy. Special consideration is given to distinguish sensor faults from system component faults since faulty sensor information could confuse the ACM system if not addressed sufficiently. Overall, the development and enhancements of WRS modeling, the ACM system architecture for the WRS, as well as the novel fault diagnosis and control algorithms constitute the major contributions from this study