Dynamic Data-driven Prediction, Measurement Adaptation, and Active Control of Summary of Efforts q AF Relevance: Development of monitoring & control algorithms and the associated software for propulsion and flight control of both tactical and transport aircraft under nominal operating conditions as well as adverse and emergency situations. Synergistic combinations of dynamic data-driven and model-based information. q Defined AF Mission Needs: Applications to future DoD and commercial programs requiring this advanced technology. Joint Strike Fighter (JSF) Unmanned Compact Air System (UCAS) Versatile Affordable Advanced Turbine Engines (VAATE) Next Generation Product Family commercial engines. q Key Focus of Scientific Research: Theoretical development and experimental validation of a unified dynamic datadriven and model-based methodology for real-time monitoring and active control of combustion instabilities in tactical and transport aircraft. Identification of incipient sources of combustion instabilities Sensing system adaptation and resource allocation for detection of combustion instabilities Online adaptation of the critical measurement system parameters Information fusion of heterogeneous sensor data for fast decision-making on combustion instabilities Analysis and synthesis of active combustion control algorithms for flight and propulsion control systems q Other Performers on the Research Project: In addition to the principal investigator, Prof. Asok Ray, the following are the technical personnel who are\were working on this research project. Senior research personnel: Dr. Shashi Phoha Graduate (doctoral level) students. Current: Sihan Xiong, Pritthi Chattopadhyay, Michael Hauser, Sudeepta Mondal, and Chandrachur Bhattacharya Already graduated: Dr. Nurali Virani and Dr. Devesh K. Jha 1
THEORY and Results New S&T advances for new capabilities through the project: Modeling: Bayesian nonparametric modeling of panel data and high-order time series data; Sequential classification Algorithms: Gibbs sampling algorithm for inference; Conditional tensor factorization and Bayes factor analysis for hypothesis testing and causal inference-making Sensing: Dynamic data of acoustic pressure oscillations and chemiluminescence Systems Software: Python code Time serise1 Time series2 Bayesian Nonparametric Modeling System identification Causal relationship Information fusion Sequential classification Operating condition 2
THEORY and Results New S&T advances for new capabilities through the project: Modeling of the decision & control strategy: Algorithms: Bayesian nonparametric modeling for information fusion and sequential classification; Policy search algorithms for active control; Gaussian process (GP) modeling for obtaining domain knowledge of stability map. Sensing: Distributed pressure sensors and thermocouples for acquisition of high-resolution of spatio-temporal data; and hot-wire anemometers for flow characterization. Systems Software: Python codes. 3
Validation of Theoretical Results New S&T advances for new capabilities through the project: Model of the experimental apparatus: Electrically heated Rijke tube (to be fabricated and instrumented under a forthcoming DURIP grant). Algorithms: Emulation of nominal and off-nominal operating conditions in gas turbine combustors by using a simplified laboratory-scale Rijke tube setup. Developing a real-time active control structure based on different machine-learning algorithms and physics-based modeling. To be supplemented by the experimental data and validated on the test apparatus. Sensing: Sensing devices include: 8 pressure sensors,15 thermocouples, and a hot-wire anemometer for obtaining high-resolution spatio-temporal data. Systems Software: LabView, Flowvision (Alicat), Matlab, Python 4
q Accomplishments Development of theoretical foundations and information-theoretic strategies for real-time monitoring and active control of combustion instabilities. Design of a test apparatus for experimental validation of the theoretical part of the research. The test apparatus design completed in consultation with industry, DoD Labs, and other U.S. and foreign universities. Construction of the test apparatus under a recently funded AFSOR DURIP grant to be completed by August 2018. q Selected archive journal publications (with foreign universities and DoD Laboratory) in 2016 and 2017 (supported by this grant) S. Sarkar, S.R. Chakravarthy, V. Ramanan and A. Ray, Dynamic data-driven prediction of instability in a swirl-stabilized combustor," International Journal of Spray and Combustion Dynamics, vol. 8, no. 4, December 2016, pp. 235-253. doi: DOI: 10.1177/1756827716642091 D.K. Jha, A. Srivastav and A. Ray, Temporal learning in video data using deep learning and Gaussian processes, International Journal of Prognostics and Health Management, ISSN 2153-2648, Vol 7 (Special Issue Big Data and Analytics) 022, pages: 11, 2016. Y. Li and A. Ray, "Unsupervised Symbolization of Signal Time Series for Extraction of the Embedded Information," Entropy, vol. 19, no. 4, April 2017, p. 148 (1-20). doi:10.3390/e19040148. P. Chattopadhyay, S. Mondal, C. Bhattacharya, A. Mukhopadhyay and A. Ray, "Dynamic data-driven design of lean premixed combustors for thermoacoustically stable operations," ASME Journal of Machine Design (Special Issue on Data-Driven Design), in press. doi: 10.1115/1.4037307 S. Xiong, S. Mondal and A. Ray, "Detection of thermoacoustic instabilities via nonparametric Bayesian Markov modeling of time series data," ASME Journal of Dynamic Systems, Measurement, and Control, in press. doi: 10.1115/1.4037288. N. Virani, D.K. Jha, Z. Yuan, I. Sekhawat and A. Ray, "Imitation of demonstrations using Bayesian filtering with Nonparametric Data-Driven Models," ASME Journal of Dynamic Systems, Measuremnt, and Control (Special Issue on Commemorating the life, achievements and impact of Rudolph E. Kalman), in press. Y. Li, D.K. Jha, A. Ray and T.A. Wettergren, "Information Fusion of Passive Sensors for Detection of Moving Targets in Dynamic Environments, " IEEE Transactions on Cybernetics, vol. 47, no. 1, January 2017, pp. 93-104. doi: 10.1109/TCYB.2015.2508024 Y. Li, D.K. Jha, A. Ray and T.A. Wettergren, "Information-theoretic Performance Analysis of Sensor Networks via Markov Modeling of Time Series Data," IEEE Transactions on Cybernetics, in press. doi: 10.1109/TCYB.2017.2717974 UTRC, USA IIT Madras, India Jadavpur University, India General Electric, USA NUWC, U.S. Navy Mitsubishi Electric, USA 5
Dynamic Data-driven Prediction, Measurement Adaptation, and Active Control of q Coordination/Synergy: Other DDDAS PIs: Currently collaborating with Prof. Soumik Sarkar, Iowa State University, Ames, IA. AFRL, DOD Collaboration: Initiating collaboration with AFRL scientists for future research. Data Coordination: Ongoing data coordination with IIT Madras, India and Jadavpur University, India. q Exposure/Use by Other Groups: Future DoD and commercial programs requiring this advanced technology. Demonstrations: Planned demonstration of theoretical results on an apparatus (under construction) within a year. Affiliation with DoD Projects: Collaborated with US Navy on DDDAS research for undersea target detection. Use of Developments in Other Programs: Industry collaboration for extension of DDDAS to Non-DoD research. 6