Forecasting Spacecraft Telemetry Using Modified Physical Predictions

Ryan Mackey and Igor Kulikov
Submission Type: 
Full Paper
Supporting Agencies (optional): 
phmc_10_022.pdf455.14 KBAugust 30, 2010 - 9:51pm

Among systems that provide sensor data of their performance, one approach to prognostic estimation is forecasting, i.e. prediction of measurable parameters and comparison of predicted values against established operational limits. Forecasting can be attempted statistically, or can be based on rigorous physical simulation. However, combining these approaches is difficult where system mode behavior or timing of system activities is uncertain, limiting the accuracy or applicability of a forecast.

In this paper we describe a method to modify simulation outputs to better match current telemetry. We begin with a familiar autoregressive approach to model residuals between predicted spacecraft performance, provided by physics-based modeling tools, and up-to-date spacecraft telemetry. This result is then improved by transforming the simulation result to better fit recent data, and the transformation is applied to generate more accurate future predictions. The method is suitable for real-time signal prediction. We will motivate this approach and characterize its performance using example telemetry from the NASA Mars Exploration Rover spacecraft.

Publication Control Number: 
Submission Keywords: 
sensor fusion
physical modeling
Submitted by: 

follow us

PHM Society on Facebook Follow PHM Society on Twitter PHM Society on LinkedIn PHM Society RSS News Feed