Session: DAC-09-01-Design for Resilience and Failure Recovery

Paper Number: 89132

89132 - A Computational Framework for the Evaluation of Resilience in Deep Space Habitat Systems 

Resilience is a vital consideration for designing and operating a deep space habitat system. The numerous hazards that may affect a deep space habitat and its crew during its lifecycle need to be considered early in the design. Trade-off studies are the typical method used to assess the cost and value of different design choices. Here we develop a modular dynamic computational framework intended for rapid simulation and evaluation of the resilience of different system configurations. The framework uses a system-level phenomenological Markov model of the habitat systems, enabling us to assess multiple habitat configurations and evaluate their performance in the presence of several hazards and user-defined control policies. System fault detection and repairs are modeled. External disturbances, including meteorites impact, temperature fluctuations, and dust, are modeled based on the lunar environment, envisioning a deep space habitat design. We use a reflective health management subsystem that prioritizes recovery actions based on the robotic agent's availability to close the loop. In addition to performance, a resilience metric is included to quantify the system's resilience  over the design lifecycle. We illustrate the use of the framework for supporting early-stage design decisions of a habitat system. Our case study focuses on designing the power generation system considering cost and energy efficiency.

Presenting Author: Amir Behjat Purdue University

Presenting Author Biography: Amir Behjat is a Post-Doctoral researcher at Purdue university. He received his Ph.D. in Mechanical Engineering from University at Buffalo in 2021 He graduated with his BS and MS degrees in Mechanical engineering from Sharif University of Technology. His research focuses on neuroevolution, physics-aware machine learning, co-Design, and surrogate/physics based design and optimization.

Authors:

Amir Behjat Purdue University
Roman Ibrahimov Purdue University
Ali Lenjani Purdue University
Aaron Behrkat Purdue University
Kathleen Martinus Purdue University
Amin Maghareh Purdue Universtiy
Dawn Whitaker Purdue University
Illias Bilionis Purdue University
Shirley Dyke Purdue University