Author ORCID Identifier
Semester
Fall
Date of Graduation
2022
Document Type
Dissertation
Degree Type
PhD
College
Statler College of Engineering and Mineral Resources
Department
Mechanical and Aerospace Engineering
Committee Chair
Piyush M. Mehta
Committee Co-Chair
Jason Gross
Committee Member
Jason Gross
Committee Member
Yu Gu
Committee Member
Snehalata Huzurbazar
Committee Member
Nasser Nasrabadi
Committee Member
Natalia Schmid
Abstract
In the modern space age, private companies are crowding the already-congested low Earth orbit (LEO) regime with small satellite mega constellations. With over 25,000 objects larger than 10 cm already in LEO, this rapid expansion is forcing us towards the enterprise on Space Traffic Management (STM). STM is an operational effort that focuses on conjunction assessment and collision avoidance between objects. While the equations of motion for objects in orbit are well-known, there are many uncertain parameters that result in the uncertainty of an object's future position. The force that the atmosphere exerts on satellite - known as drag - is the largest source of uncertainty in LEO. This is largely due to the difficulty in predicting mass density in the thermosphere - the neutral region in Earth's upper atmosphere. Presently, most thermosphere models are deterministic and the treatment of uncertainty in density is highly simplified or nonexistent in operations.
In this work, four probabilistic thermospheric mass density models are developed using machine learning (ML) to enable the investigation of the impact of model uncertainty on satellite position for the first time. Of these four models, two (HASDM-ML and TIE-GCM ROPE) are reduced order models based on outputs from existing thermosphere models while the other two (CHAMP-ML and MSIS-UQ) are based on in-situ thermosphere measurements. The data and model development are described, and the models' capabilities, including the robustness of their uncertainty quantification (UQ) capabilities, are thoroughly assessed.
Existing thermosphere models, and the ones developed here, use different space weather drivers to estimate density. In a forecasting environment, there are algorithms and models that forecast the drivers for a given period in order for a density model to make a forecast. The driver forecast models used by the United States Space Force for the HASDM system are assessed to benchmark our current capabilities. Using the error statistics for each driver, we can perturb the deterministic forecasts. This provides an avenue to use the ML thermosphere models to study the effect of driver uncertainty on satellite position, in addition to model uncertainty, for any period with available driver forecasts. Seven periods are considered with diverse space weather conditions to study the isolated effects of the two density uncertainty sources on a 72-hour satellite orbit. This provides insight into the relative importance of density uncertainty on satellite position for various space weather scenarios. This study also functions as a motivation to reconsider our current methods for STM in order to improve our capabilities and prevent future satellite collisions with increased confidence.
Recommended Citation
Licata, Richard J. III, "Probabilistic Space Weather Modeling and Forecasting for the Challenge of Orbital Drag in Space Traffic Management" (2022). Graduate Theses, Dissertations, and Problem Reports. 11600.
https://researchrepository.wvu.edu/etd/11600