Modeling In-Orbit Radiation for Fault-Injection

Supervisor: Andrew Price / Mathieu Salzmann
Type of Project: Masters thesis, 1 student
Duration: Dependent on masters thesis program
Image courtesy of: “Everything you wanted to know about space radiation but were afraid to ask”, Jeff Chancellor et al.

 Deep neural networks can be applied to many in-orbit applications such as in-orbit servicing, debris removal and construction. However, the deployment of these algorithms in-orbit raises the question of radiation tolerance. Exposure to high energy particles can induce single event upsets and other errors. An algorithm’s sensitivity to such errors is a function of many factors including: the algorithm architecture, the hardware, the orbit, Earth’s magnetic field, and the solar cycle.

Project Scope

The student will review space environment models and implement one in combination with an orbit integrator. The student will explore the effects of different orbits, satellite attitudes, and hardware footprint size on expected radiation fluxes. These fluxes will be used as input to a neural network fault injection tool developed by Andrew. These modeled fluxes will be validated against literature and, timeline permitting, validated with radiation experiments.

Tasks

  • Reviewal of space environment models
  • Implementation of an environment model into an orbit integrator
  • Exploration of the impact of several factors on radiation dose (e.g., satellite attitude, hardware footprint etc).
  • Validation of expected radiation dose against literature
  • [Schedule dependent] Validation of expected radiation dose with in-lab experiments

Prerequisites

  • This project is intended for a masters thesis student interested in mechanistic modeling and the space environment.
  • Python programming skills are mandatory.

Contact

References

  • Standard ECSS-E-ST-10-04C Space Environment
  • Flavio Ponzina, Rubén Rodríguez Álvarez, José Miranda Calero, Mathieu Salzmann, Jacques Viertl, Tajana Rosing, Miguel Peón-Quirós, David Atienza. “Using ensemble learning to improve radiation tolerance of CNNs in space applications.” Proceedings of the 1st SPAICE Conference on AI in and for Space.
  • Fernando Fernandes dos Santos, Paolo Rech, Angeliki Kritikakou, and Olivier Sentieys. “Neutron-Induced Error Rate of Vision Transformer Models on GPUs.” 2023 23rd European Conference on Radiation and Its Effects on Components and Systems (RADECS)