AffiliationPeraton Labs, Test Resources Management Center
MetadataShow full item record
CitationLau, R., Bagchi, A., Shen, J., Triolo, T., Sanchez, K., Yao, L., Kovarskiy, J., & Castro, R. (2022). Advanced Multi-Variate Time Series Analytic Techniques (Attends). International Telemetering Conference Proceedings, 57.
AbstractWe describe an advanced architecture supporting fast decisions by using multi-variate time series analytic techniques on voluminous datasets that were previously inaccessible. The system, Advanced Multi-Variate Time Series Analytic Techniques (ATTENDS) automates data ingestion, knowledge extraction, and Artificial Intelligence/Machine Learning (AI/ML) algorithm configuration for anomaly detection, failure prediction, causal analysis, and diagnosis. To enable reusability, ATTENDS presents a set of Application Programming Interfaces (API) to support user configurability and remote invocation. The APIs implement state-of-the art AI/ML algorithms for predictive maintenance, sensor component correlation for problem diagnosis, and unsupervised learning of sensor measurement anomaly for support of automated testing and evaluation. We will present two use cases including prediction of Remaining Useful Life (RUL) of Turbofan  and sensor diagnosis and recommendation for maintenance actions, as well as detection and quantification of target location error in an airborne platform.