Satellite Fault Identification

Spacecraft operators need advanced capabilities that rapidly and effectively identify the cause of detected satellite anomalies based on flight telemetry data. In this Phase II, the USSF/USAF requests that industry design, implement, integrate, and test the most promising and effective algorithm(s) with ground software to classify detected satellite faults in near real time. The need for such a fault classification capability is because the space environment is harsh and can lead to many kinds of onboard faults. Identifying these faults in a timely fashion is difficult since communication with the ground is either intermittent or may involve long latencies to deliver to the ground for analysis. Furthermore, the amount of data to be processed by a ground operator is cognitively overwhelming. The problem is made even more difficult given the complex interdependence between the various satellite components and subsystems and between satellite fault behaviors. Present-day fault diagnosis is performed manually with some degree of automated, but not autonomous, fault protection systems. Such fault protection systems perform low-level, coarse fault detection and classification coupled with rule-based remediation strategies designed for rapid recovery from common anomalous events. The most common remediation rule is to reset the bus, which temporarily places the bus in a safe mode while the system restarts. However, such hard resets create service outage leading to operational blind spots and reduced return on investment. This current state of practice is limited, lacks accuracy, is time consuming, and can result in failure to deliver critical services. In this effort, Trusted Space will continue its work from Phase I to design, implement, integrate, and test a prototype solution with ground software to classify detected satellite faults in near real time. Our Phase II prototype software will be based on our recommended Phase I Time Series Classification and Dynamic Bayesian Network based Fault Diagnostics Engine (FDE) solution for enhanced and scaled autonomous spacecraft fault identification. Specifically, Trusted Space team will use its deep expertise in spacecraft systems, Machine Learning and Artificial Intelligence, and strong software development capabilities to: expand the limited Attitude Determination and Control System (ADCS) models and fault behaviors from Phase I to a more comprehensive set of behaviors, including fault sources located in other satellite subsystems that cascade to the ADCS subsystem, develop a Hierarchical architecture of the FDE (HFDE) to enhance the scalability of the solution, enhance the algorithm implementation of the HFDE for computational speed, efficiency, and scalability, generate a high-quality HFDE software prototype that meets software development and operations best practices and that can be integrated with existing ground software, and test, evaluate and demonstrate the HFDE software prototype.