Concurrent Static and Dynamic State Monitoring Using a Global-Local Model for the Fluid Loop in Space Stations

Abstract

The fluid loop is the core of thermal control system in the space station. Owing to the complex and changeable operating conditions and long time in orbit, the fault rate and performance degradation rate increase. Therefore, it is of practical significance to timely and accurately monitor the operating state of the fluid loop. This article proposed a combined strategy of canonical variate analysis and slow feature analysis to concurrently monitor the static deviation and process dynamics for the operating state. First, multiple sensors are used to measure the data in all condition of the space station, and typical faults are simulated by fault seeding to obtain fault data. Following that, the serially correlated canonical subspaces for each normal operating condition are modeled, and they are further explored to extract slow features to establish the multiple local static monitoring models. Besides, a global model is built using differential slow features in all condition to detect the dynamic anomalies by monitoring the temporal variations. To illustrate the feasibility and efficacy, the proposed monitoring strategy is applied to monitor the test data consisting of normal samples and fault samples.