Creating Situational Awareness with Spacecraft Data Trending and Monitoring

Data trending and monitoring are crucial in the spacecraft operations for maintaining the spacecraft health and safety and for evaluating the system performance and accuracy. The existing trend and data monitoring approaches are insufficient for time dependent datasets. The monitoring of these datasets would be very difficult if not possible without a time dependent trend being established, and the determination if a time dependent dataset at a given time is normal or in an error state requires significant engineering analysis efforts. This talk presents a Satellite Data Trending and Monitoring Toolkit(SDTMT), which implements a machine learning system for an automated and integrated trending and monitoring of time dependent datasets exhibiting the diurnal characteristics. Satellite data trending and monitoring are a natural fit to the operational concepts of a machine learning system. The data training in the machine learning system obtains the time dependent trend for datasets represented by the time function and standard deviation. The real-time or near real-time data monitoring determines if a data point is consistent with its time dependent trend. The potential anomalies can be detected in real time, which creates the enhanced situational awareness for autonomous spacecraft operations. The adaptive trending and limit monitoring algorithm and the neural networks are implemented in SDTMT as the machine learning algorithms. The machine learning approach is systematic, autonomous and adaptive. The application of the machine learning system to Geostationary Environment Operational Satellite(GOES) Imager data processing process is presented. It shows that the machine learning system enables the real-time monitoring of the instrument data calibration process that would have been impossible with the standard statistical trending approach. SDTMT can have many potential applications from the spacecraft health and safety to the science instrument data processing process, and it represents a significant advance toward an autonomous spacecraft operations.