Dynamic Data-Driven Prediction, Measurement Adaptation
Abstract
Scientific advancements in the development of fast dynamic data-driven tools for forecasting and detection of thermo-acoustic instabilities as well as of fatigue damage in aircraft structures: The DDDAS paradigm has been used with streaming sensor data for forecasting and detection, and classification of emerging anomalies. The algorithmic advancements are achieved using differential-geometric concepts of deep neural networks and statistical learning with hidden Markov models for sequential pattern classification, as an alternative to symbolic time series analysis. The proposed methods have been validated on time-series data, generated from a laboratory-scale combustor apparatuses, operating under different protocols at varying air-fuel premixing levels. Applicability of the proposed method has been demonstrated with respect to anomaly detection and regime identification with limited data requirements, making it a potential candidate for near-real-time monitoring and active control of thermo-acoustic instabilities in commercial-scale combustors. Experimental Research and Validation: A computer-instrumented and computer-controlled laboratory apparatus has been designed.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 13, 2021
- Accession Number
- AD1155231
Entities
People
- Asok Ray
Organizations
- Pennsylvania State University