A Health State Assessment Method Based on Autoencoder and Analytic Hierarchy Process for Spacecraft Components

Health assessment of complex systems mainly relies on expert experience or a data-driven health assessment model trained by a large volume of historical condition monitoring data. However, only relying on expert experience to handle the health information contained in a large amount of telemetry data is very time-consuming and labor-intensive for spacecraft, and it does not make full use of a large amount of historical data. Moreover, due to limitations such as space-ground data transmission resources, it is also difficult to obtain all the condition monitoring parameters at each system level for different health states. In response to these challenges, this paper proposes a health assessment framework combining expert experience and data-driven strategy to achieve a multi-level quantitative assessment for spacecraft components. Specifically, a deep autoencoder is trained to construct the state change curves of the key condition monitoring parameters, and the quantitative score map is constructed for each parameter. Then, by analyzing the matrix of pairwise importance for parameters based on expert experience, the health status change curve at a second level of the component is constructed. Finally, a quantitative score map of the component status is provided using the health status change curve to realize the multi-level quantitative health assessment. A case study on health assessment for an attitude control actuator is conducted to show the overall procedure and effectiveness of the health assessment framework. The results demonstrate that the proposed method can effectively integrate expert knowledge and data characteristics, providing a feasible framework for the multi-level health assessment of spacecraft components.