Real-Time Anomaly Detection Method for Space Imager Streaming Data Based on HTM Algorithm
Lei Song, Haoran Liang, Taisheng Zheng · 2019
Healthy operational status of space imager is important for the successful completion of space exploration tasks, and any abnormal event may lead to serious faults or disasters. Real-time anomaly detection is an important technique for finding abnormal parameters and potential faults of space equipment. The downlink data of space imager is streaming data that presents technique challenges and opportunities. The fundamental capability of anomaly detection techniques for space imager streaming data is to model each stream in an unsupervised fashion and detect unusual. Under the influence of operating instructions, environmental conditions and equipment performance, the streaming times series fluctuates acutely and indicates obvious concept drifting. Besides, application constraints require the method to process data in real-time, not batches. However, most anomaly detection methods need offline training of amount of historical data, and it is difficult to realize the online learning and detecting continuously. In this paper, a novel method is proposed that meets the constraints. The method is based on online time series memory and learning algorithm called Hierarchical Temporal Memory (HTM). We also present the comparison results of the proposed method and other algorithms, and the experiments show the proposed method could realize the real-time anomaly detection effectively.