| Anomaly monitoring for Satellite is one of the core technologies that ensure satellite operating with in-orbit-safety and reliability. Because of the complex structure of satellite and the complex space environment it works with, the factors that affect the satellite status can be numerous. These effects make it difficult to establish an accurate analysis of model. Consequently the traditional model-based method is not suitable.Moreover many of the key components of satellite are small, which makes it difficult to obtain large amount of historical data and enough sufficient knowledge on abnormal components and rules. In this dissertation, taking the anomaly monitoring problems of lithium-ion battery as the background, based on the study of the Multivariate State Estimation Techniques (MSET), an anomaly monitoring methodology of the key components of satellite is proposed. It can be a new way to address the anomaly monitoring of key components of satellite through the actual in-orbit telemetry obtained data to achieve the anomaly monitoring of key components of satellite. The main research production of this dissertation includes the following aspects:(1) The physics of failure and selction of anomaly monitoring parameters with satellite battery are studied. Taking lithium-ion battery as a case, the basic principles, failure mechanisms and performance parameters are studied. And then the anomaly monitoring method of lithium-ion battery is proposed by monitoring the changes of electrolyte resistance and charge transfer resistance.(2) The state estimation and life prediction methods based on MSET are studied, which can solve the state estimation problem when absence of historical data. A MSET estimation method is presented based on the actual in-orbit state telemetry obtained data. And the data standardization process, training data selection, memory matrix structure and the nonlinear operator selection and optimization of MSET implementation technologies are given respectively. Furthermore a life prediction method based on MSET is proposed in accordance with the relationship between the actual residuals based on MSET and the degradation of the products.(3) The anomaly detection is studied based on SPRT, which can solve the anomaly detection problem when actual residuals distribution is known or unknown. A non-parametric test method, sequential rank-sum probability ratio test (SRPRT), is presented. According to the actual residuals of different distribution, the sequential probability ratio test (SPRT) and the sequential rank-sum probability ratio test (SRPRT) are compared in different situations.(4) Taking the lithium-ion battery electrolyte resistance and charge transfer resistance as the monitoring parameters, the anomaly monitor problem of lithium-ion battery is studied with the above methods. Comparing with the traditional threshold monitoring method, the results show that the method presented is more suitable for solving the anomaly monitoring of the key components of Satellite. |
