Adaptive System Modeling for Spacecraft Simulation

This invention introduces a methodology and associated software tools for automatically learning spacecraft system models without any assumptions regarding system behavior. Data stream mining techniques were used to learn models for critical portions of the International Space Station (ISS) Electrical Power System (EPS). Evaluation on historical ISS telemetry data shows that adaptive system modeling reduces simulation error anywhere from 50 to 90 percent over existing approaches.

The purpose of the methodology is to outline how someone can create accurate system models from sensor (telemetry) data. The purpose of the software is to support the methodology. The software provides analysis tools to design the adaptive models. The software also provides the algorithms to initially build system models and continuously update them from the latest streaming sensor data. The main strengths are as follows: • Creates accurate spacecraft system models without in-depth system knowledge or any assumptions about system behavior. • Automatically updates/calibrates system models using the latest streaming sensor data. • Creates device specific models that capture the exact behavior of devices of the same type. • Adapts to evolving systems. • Can reduce computational complexity (faster simulations).

This work was done by Justin Thomas of Johnson Space Center. For further information, contact the JSC Innovation Partnerships Office at (281) 483-3809. MSC-24419-1

The U.S. Government does not endorse any commercial product, process, or activity identified on this web site.