Data-driven Dynamical Systems Analysis in Fluid Mechanics

Please notice: The Focus Group is organising a conference.
SoTiC 2020 - Symposium on Thermoacoustics in Combustion: Industry meets Academia
14-17 September 2020, TUM


This Focus Group is co-led by Rudolf Diesel Industry Fellow Dr.-Ing. Mirko R. Bothien (Ansaldo Energia Switzerland) and Hans Fischer Fellow Dr. Luca Magri (University of Cambridge), who are hosted by Prof. Dr.-Ing. Thomas F. Sattelmayer (Thermodynamics, TUM) and Prof. Dr. Wolfgang Polifke (Thermo-Fluid Dynamics, TUM) respectively. Dr. Nguyen Anh Khoa Doan is the IAS-TUM postdoctoral fellow working with Luca Magri. This Focus Group partly continues the work in flow instability of the Alumni Focus Group Advanced Stability Analysis.

We study one of the most intractable and persistent challenges in the development of propulsion and power-generation engineering with unconventional approaches: Prediction, understanding of reacting flow dynamics’ mechanisms, and control of extreme events in fluids with artificial-intelligence algorithms and advanced clean power-generation technologies.

Projections indicate that combustion-based energy conversion systems will continue to be a predominant approach for the majority of our energy usage for the coming decades. In aviation, emission regulations for aero-engines have been steadily tightened over the past decades (ACARE Flightpath 2050 - Europe's Vision for Aviation, www.acare4europe.org). At the Paris Climate Conference in December 2015, 195 countries adopted the first-ever universal global climate deal. The climate and energy framework sets three stringent key targets to be achieved by 2030: (i) 40% or more cuts in greenhouse emissions; (ii) 27% or more share for renewable energy; and (iii) 27% or more improvement in energy efficiency.

The multi-physical fluid mechanics of aeronautical propulsion

Air transportation is anticipated to double over the next couple of decades, which calls for novel methods to cut back on up to 80% in oxides of nitrogen (NOx) and up to 50% in noise, as set by the Advisory Council for Aerospace Research. To develop new clean aircraft engines, gas turbines are designed to burn in a lean regime to reduce NOx emissions. The downside is that lean flames burn very unsteadily because they are sensitive to the turbulent environment of the combustion chamber. In this complex multi-physical environment, three main physical subsystems can be identified: (i) acoustics, (ii) aerodynamics, and (iii) flame dynamics (chemical reaction). These subsystems interact with each other contributing differently, but fundamentally to engine noise; thermo-acoustic instabilities; and rare and extreme events.

On the one hand, all of these phenomena (here referred to as extreme events for brevity) are unwanted and have to be eliminated during the design or controlled if they occur. On the other hand, these phenomena are bound to increase as turbines become cleaner. These two contrasting situations make the design of low-emission aircraft engines particularly challenging. This project will propose a new framework to predict and control extreme events with artificial intelligence and adjoint methods. With a better design, the new aircraft engine will be cleaner, healthier and quieter, keeping the design, repair and replacement costs low.

The multi-physical fluid mechanics of clean power generation

The world of power generation is changing rapidly. Analyses of future electricity generation predict renewables and natural gas to have the highest growth of all fuels in the coming decades. According to the reference scenario in the International Energy Outlook 2017, their combined share is increasing to 57% in 2040, 26% being covered by natural gas, i.e., gas turbines. Playing such a major role for the future energy mix and the change-over to renewables, it is of utmost importance to further enhance the efficiency and reduce the emissions of gas turbine power plants. Additionally, gas turbines need to exhibit outstanding fuel flexibility. This does not only include strongly varying compositions of natural gas. In order to reduce the environmental impact, excess power from renewables will be used for energy storage, for example, by producing hydrogen enriched fuels. However, fuels with very low calorific values from waste processes or biomass gasification complement the wide range of alternative fuels, for which a reliable combustion in gas turbines has to be ensured. Gas turbine efficiency is constantly increasing, especially in the recent past, and is projected to further increase. At the same time, detrimental emissions have to be reduced. Not only will gas turbine power plants considerably contribute to the overall world’s energy mix, but their major role will be also to compensate the fluctuations of renewables, especially from wind energy. For this purpose, a gas turbine has to be able to respond quickly and reliably to balance the electricity production with the demand. Accordingly, fast load ramp rates but also optimum efficiency under all load conditions of the power plant are a crucial requirement.

Efficient, ultra-low emission gas turbines are key to meet the global challenges of reliable energy production and minimal environmental impact of power generation. The need for efficiency and power increase calls for increasing firing temperatures. This in turn requires extremely short combustion chambers and rapid mixing so that the post-flame residence times are sufficiently low keeping detrimental NOx emissions below the limits. However, part-load CO emission burn-out has to be guaranteed as well, which is diametrically opposite of having short combustors. To accomplish this goal, the sequential combustion principle is best suited. A sequential combustor consists of two combustion chambers arranged in series, which allows the machine to switch off the second stage and hence to park the engine at extremely low loads. During the times the renewables satisfy the energy demand this is an unbeatable advantage: The gas turbine can be operated at very low loads and, when required, is able to deliver power fast because it does not need to be ignited first. In contrast to this, single combustor gas turbines cannot be turned down to similarly low levels and eventually they would have to be turned off with the disadvantage of requiring longer time to get to full load and penalties in engine lifetime due to start-stop cycles.

Alstom Power was the pioneer of reheat gas turbines and, after acquiring the latest sequential combustion technology from Alstom, Ansaldo Energia, is now continuing to be at the forefront of reheat engines. A paradigm shift from gas turbines relying on single combustors based on lean premixed propagation stabilized flames to sequential combustion systems is in full swing.

Methods

Prediction of rare and extreme events with data-driven algorithms

The common and paramount feature of the extreme events is the exceedingly high sensitivity to the design parameters and the operating conditions. First, there are elusive physical mechanisms, which cannot be easily detected, for example, the intertwined interactions between the acoustic, aerodynamics and chemical reaction. These not yet fully understood physical interactions appear in our models as hidden (or latent) variables, which increase the uncertainty of the predictions. Second, the parameters of the models are uncertain because they are difficult to measure and quantify. Because of their large sensitivity to the configuration and operating point, they are bound to have large uncertainties. For these reasons, traditional engineering methods, such as computational fluid dynamics, in particular large-eddy simulations (LES), and aero-thermo-acoustic models, in particular Helmholtz solvers and wave-based low-order models, have proven to be limited at accurately predicting and controlling extreme phenomena, both at design and testing stage. From an industrial point of view, finding the presence of extreme events results in a high financial risk owing to the high costs in re-designing a combustion chamber in a late stage of development.

A variety of novel computational techniques based on artificial intelligence, adjoint methods, and weather forecasting techniques, will be developed and applied for accurate physical description, prediction and control of extreme events. In lack of a full physical description, existing database and experimental data will be used to develop hybrid predictive tools, which will be physics-based and data-driven. In more detail, the four main objectives are

  • Modelling fluid mechanics systems with Artificial Intelligence;
  • Predicting and preventing extreme events with Data Assimilation;
  • Controlling in real time with Adjoint Methods;
  • Calculating robust uncertainty quantification.

A clean power-generation technology: Sequential combustion

The physical driving mechanisms causing an interaction of flow, acoustics and flame are sought to be experimentally unraveled. The results are compared to flame transfer matrices retrieved from large-eddy simulation (LES) and system identification techniques. Based on these analytical models both for low and high frequencies are set up. For the experimental investigations an existing flat combustor (HTRC: High-Frequency Transverse Mode Reheat Combustor) for lean, premixed combustion is available, which is partly stabilized by deflagration and by auto-ignition as in systems of technical interest. In this context, the term “high frequency” refers to a frequency regime, which is beyond the cut-on frequency of the combustor. This leads to a relation between flame and modal length scales on the same order of magnitude. Hence, local interactions between flame and acoustic mode require consideration, rendering the situation as thermoacoustically non-compact.

Although mission-oriented, this focus group also aims at developing new mathematical and computational methods, which will be applied in other fields involving multiple-scales and multi-physics, such as aeroelasticity and bio-engineering.

Publications of the Focus Group

2020

  • Nguyen Anh Khoa Doan, Wolfgang Polifke, Luca Magri: Learning Hidden States in a Chaotic System: A Physics-Informed Echo State Network Approach. 2020 mehr…

2019

  • Bothien, Mirko R.; Ciani, Andrea; Wood, John P.; Fruechtel, Gerhard: Toward Decarbonized Power Generation With Gas Turbines by Using Sequential Combustion for Burning Hydrogen. Journal of Engineering for Gas Turbines and Power 141 (12), 2019 mehr…
  • Ciani, Andrea; Bothien, Mirko; Bunkute, Birute; Wood, John; Früchtel, Gerhard: Superior fuel and operational flexibility of sequential combustion in Ansaldo Energia gas turbines. Journal of the Global Power and Propulsion Society 3, 2019, 1-16 mehr…
  • Doan, Nguyen Anh Khoa; Polifke, Wolfgang; Magri, Luca: Physics-Informed Echo State Networks for Chaotic Systems Forecasting. In: Lecture Notes in Computer Science. Springer International Publishing, 2019 mehr…
  • Gant, Francesco; Scarpato, Alessandro; Bothien, Mirko R.: Occurrence of multiple flame fronts in reheat combustors. Combustion and Flame 205, 2019, 220-230 mehr…
  • H. Yu, M. P. Juniper, and L. Magri: Interpretability within a level-set data assimilation framework. SIAM Workshop on Frontiers of Uncertainty Quantification in Fluid Dynamics, 2019 mehr…
  • Magri, Luca; Doan, Nguyen Anh Khoa: Physics-informed data-driven prediction of turbulent reacting flows with Lyapunov analysis and sequential data assimilation. In: Apollo - University of Cambridge Repository, 2019 mehr…
  • N. A. K. Doan, W. Polifke, and L. Magri: Physics-Informed Echo State Networks for the Prediction of Extreme Events in Turbulent Shear Flows. 72nd Annual Meeting of the APS Division of Fluid Dynamics, 2019 mehr…
  • Nguyen Anh Khoa Doan, Wolfgang Polifke, Luca Magri: A physics-aware machine to predict extreme events in turbulence. 2019 mehr…
  • Nilsson, Thommie; Yu, Rixin; Doan, Nguyen Anh Khoa; Langella, Ivan; Swaminathan, Nedunchezhian; Bai, Xue-Song: Filtered Reaction Rate Modelling in Moderate and High Karlovitz Number Flames: an a Priori Analysis. Flow, Turbulence and Combustion 103 (3), 2019, 643-665 mehr…
  • Traverso, Tullio; Magri, Luca: Data Assimilation in a Nonlinear Time-Delayed Dynamical System with Lagrangian Optimization. In: Lecture Notes in Computer Science. Springer International Publishing, 2019 mehr…
  • Yu, Hans; Jaravel, Thomas; Ihme, Matthias; Juniper, Matthew P.; Magri, Luca: Data Assimilation and Optimal Calibration in Nonlinear Models of Flame Dynamics. Journal of Engineering for Gas Turbines and Power 141 (12), 2019 mehr…
  • Yu, Hans; Jaravel, Thomas; Ihme, Matthias; Juniper, Matthew P.; Magri, Luca: Data Assimilation and Optimal Calibration in Nonlinear Models of Flame Dynamics. Volume 4B: Combustion, Fuels, and Emissions, American Society of Mechanical Engineers, 2019 mehr…
  • Yu, Hans; Juniper, Matthew P.; Magri, Luca: Combined state and parameter estimation in level-set methods. Journal of Computational Physics 399, 2019, 108950 mehr…
  • Yu, Hans; Juniper, Matthew; Magri, Luca: Physics-informed data-driven prediction of premixed flame dynamics with data assimilation. , 2019 mehr…

2018

  • Acharya, Vishal S.; Bothien, Mirko R.; Lieuwen, Timothy C.: Non-linear dynamics of thermoacoustic eigen-mode interactions. Combustion and Flame 194, 2018, 309-321 mehr…
  • Aditya, Konduri; Gruber, Andrea; Xu, Chao; Lu, Tianfeng; Krisman, Alex; Bothien, Mirko R.; Chen, Jacqueline H.: Direct numerical simulation of flame stabilization assisted by autoignition in a reheat gas turbine combustor. Proceedings of the Combustion Institute, 2018 mehr…
  • Berger, Frederik M.; Hummel, Tobias; Romero Vega, Pedro; Schuermans, Bruno; Sattelmayer, Thomas: A Novel Reheat Combustor Experiment for the Analysis of High-Frequency Flame Dynamics: Concept and Experimental Validation. Volume 4B: Combustion, Fuels, and Emissions, American Society of Mechanical Engineers, 2018 mehr…
  • Bothien, Mirko R.; Lauper, Demian; Yang, Yang; Scarpato, Alessandro: Reconstruction and Analysis of the Acoustic Transfer Matrix of a Reheat Flame from Large-Eddy Simulations. Journal of Engineering for Gas Turbines and Power, 2018 mehr…
  • F. Garita, H. Yu, L. Magri, and M. Juniper: A Bayesian Approach for Predicting Thermoacoustic Oscillations in an Electrically-Heated Rijke Tube. 71st Annual Meeting of the APS Division of Fluid Dynamics, 2018 mehr…
  • Ghirardo, G.; Boudy, F.; Bothien, M. R.: Amplitude statistics prediction in thermoacoustics. Journal of Fluid Mechanics 844, 2018, 216-246 mehr…
  • Ghirardo, G.; Di Giovine, C.; Moeck, J. P.; Bothien, M. R.: Thermoacoustics of Can-Annular Combustors. Journal of Engineering for Gas Turbines and Power 141 (1), 2018, 011007 mehr…
  • Ghirardo, Giulio; Juniper, Matthew P.; Bothien, Mirko R.: The effect of the flame phase on thermoacoustic instabilities. Combustion and Flame 187, 2018, 165-184 mehr…
  • H. Yu, T. Jaravel, M. Ihme, F. Garita, M. P. Juniper, and L. Magri: Data assimilation and parameter estimation of thermoacoustic instabilities in a ducted premixed flame. 71st Annual Meeting of the APS Division of Fluid Dynamics, 2018 mehr…
  • T. Traverso, A. Bottaro, and L. Magri: Data assimilation in thermoacoustic instability with Lagrangian optimization. EuroMech Vienna, 2018 mehr…

2016

  • García-Cáceres, Cristina; Quarta, Carmelo; Varela, Luis; Gao, Yuanqing; Gruber, Tim; Legutko, Beata; Jastroch, Martin; Johansson, Pia; Ninkovic, Jovica; Yi, Chun-Xia; Le Thuc, Ophelia; Szigeti-Buck, Klara; Cai, Weikang; Meyer, Carola W.; Pfluger, Paul T.; Fernandez, Ana M.; Luquet, Serge; Woods, Stephen C.; Torres-Alemán, Ignacio; Kahn, C. Ronald; Götz, Magdalena; Horvath, Tamas L.; Tschöp, Matthias H.: Astrocytic Insulin Signaling Couples Brain Glucose Uptake with Nutrient Availability. Cell 166 (4), 2016, 867-880 mehr…