VKI Seminar Series
Optical techniques for multi-phase systems: a support to space propellant behavior (open to public)
Dr. Alessia Simonini, Research Engineer, von Karman Institute
Deep space exploration, robotic and humans, need to be supported by efficient space engines. Several two-phase flows problematics need to be investigated in such systems to increase their reliability and reduce safety margins. Numerical codes and reduced models, used for the design of space systems, need a strong support for validation by means of experimental databases. Optical techniques might serve to this scope: they can provide a new insight on two-phase phenomena happening at low TRL providing a unique database of test cases.
The two-phase flows addressed in this presentation includes the flashing injection in the combustion chamber, the film cooling of small rocket engines and the sloshing investigation. Different techniques will be presented aiming quantitative measurements of liquid temperature, velocity and thickness and free surface characterization in the aforementioned two-phase flows.
The conditions for such techniques to be applied in flight demonstrators will be considered for an open discussion!
Moment Methods for Non-Equilibrium Low-Temperature Plasmas with Application to Electric Propulsion (open to public)
PhD Defense of Stefano Boccelli, Collaborative PhD VKI/Politecnico di Milano,
Electric space propulsion devices are able to outperform the efficiency of classical chemical rockets, by showing a much higher specific impulse. This has profound effects on the spacecraft design, as significantly more mass can be destined to the payload with respect to the previous technologies. The Hall effect thruster is one of the most employed electric propulsion devices and is based on low-temperature plasma technology. While Hall thrusters have been successfully employed for decades, the design of up- or down-scaled Hall thrusters for future manned missions and for micro-satellites requires renewed modeling efforts. Due to the low pressure of the space environment, together with the presence of electric and magnetic fields, these devices show strong deviations from thermodynamic equilibrium. An accurate computer modeling requires such effects to be carefully considered.
This work discusses the application of the Maximum-Entropy moment methods to such non-equilibrium conditions. In particular, the order-4 Maximum-Entropy methods will be considered, resulting in a set of 14 moment equations that extend the validity of the classical Euler/Navier-Stokes-Fourier fluid formulations towards strong non-equilibrium conditions. After briefly discussing the structure and properties of the 14-moment method, it will be shown that such formulation allows to describe accurately both the low-collisional ion dynamics inside the thruster channel and the strongly anisotropic and asymmetric electron distribution function in presence of crossed electric and magnetic fields. The computational cost of the method appears larger than the simpler fluid methods, but still advantageous with respect to more expensive fully-kinetic simulations.
The work was carried at Politecnico di Milano, in collaboration with the von Karman Institute for Fluid Dynamics. The work have also benefited from visits at the Laboratoire de Physique des Plasmas (LPP, Paris) and at the University of Ottawa.
Turning Data into Value - Sustainability through Probabilistic Analytics (open to public)
Dr. Georg Rollmann , Siemens Energy, Germany
Georg Rollmann has been working on Data Science methodologies ad applications to Gas Turbines R&D and Service business and other areas of the power generation industry for more than 14 years. He is responsible for the Technology Field “Data Analytics & AI” across Siemens Energy leading a cross-organizational team of experts.
Abstract: Across all areas in the energy business, sustainability and reliability aspects are becoming more and more important. For example, power plant operators are looking for condition-based maintenance concepts and optimized products and solutions to fit their specific needs in the heterogeneous energy landscape. For this, a good understanding of the potential failure modes and associated risks is inevitable. A key challenge here is given by the large number of uncertainties from different sources that influence system behavior, such as manufacturing variations, material imperfections, and power plant operating conditions. The Probabilistic approach, that combines domain-specific, e.g. physics-based, modelling with data science and statistical methods, aims at modelling these uncertainties in an explicit way. This allows for more accurate risk assessments and can be leveraged to enable a more sustainable use of natural resources while maintaining, or even improving, system reliability.
Recent advances in variational data assimilation at CMRE (open to public)
Dr. Paolo Oddo, CMRE, Centre for Maritime Research and Experimentation, Italy
Data assimilation is here introduced in its variational formulation as used at CMRE. The classical stationary formulation of the variational cost function is then augmented by including the information retrieved from ensemble simulations to improve the adherence of the background error covariance matrix to the time-evolving regime-dependent errors. The hybrid scheme is used to both correct the systematic error and to improve the representation of small-scale errors in the background error covariance matrix. Optimal exploitation of remotely sensed data is also investigated. To this end, the “hybrid” variational scheme is modified relaxing the usual assumption of uncorrelated observation errors, and physics or statistical based bias procedures are proposed and assessed. Furthermore, the “adjoint free” version of selected observation operators is tested using a new data-driven approach, based on canonical correlation analysis or neural network, to assimilate remotely observed quantities. Real or synthetic experiments data are used to illustrate the benefit deriving from the recent developments.
ARGO: a high order multiphysics solver (open to public)
Dr. Pierre Schrooyen, Senior Research engineer at Cenaero (VKI graduates, PhD UCL)
Pierre's talk will present the Argo platform, which is a multi-physics code based on Discontinous Galerkin discretization. Discontinuous Galerkin methods provides high order accuracy on unstructured meshes while ensuring local conservation of physical quantities. This type of method handles a wide variety of elements types and are well suited for local adaptation in mesh size and interpolation order. The method will be briefly presented and the current capabilities and limitations of the Argo software will be discussed.
Since his graduation, Pierre has been working as a research engineer at CENAERO and is also an invited professor at UCL, giving the Aerospace dynamics course. He is collaborating with VKI since his PhD work on several research projects, further developing ARGO.
Online Seminar on MLOps – the data science solution to efficiently manage the full lifecycle of computational software in the Cloud (open to public)
Dr. David Vanden Abeele, partner and senior quantitative modeller at Credo Software and Radovan Parrák, Data Scientist and YQ product owner
Developing, testing, deploying, running, monitoring and maintaining a computational code on your own PC is easy.
Doing it with a team is harder.
Doing it with a team on a cluster is even harder.
Doing it with a team on a cluster for tens or hundreds of codes simultaneously becomes a complete mess unless you are organized for it and supported by adequate tooling.
The good news is that you are not alone! This problem also appears in Data Science, where things are even worse. ‘Machine Learning Operations’ (MLOps, the Data Science equivalent of DevOps) offers a promising solution to these challenges in the Cloud. This talk discusses the benefits and pitfalls of building and deploying computational software in an MLOps fashion, based on practical experiences in the financial sector, on AWS and Azure Cloud platforms.
Online Seminar on shock-layer radiative heating for the Mars 2020 mission, which just delivered the Perseverance Rover on 18 February this year (open to public)
Dr. Christopher Johnston from NASA Langley Research Center
The impact of shock-layer radiative heating to the backshell of capsules entering Earth or Mars was assumed negligible until less than 10 years ago. Since then, modeling improvements and new experimental data have shown that radiative heating is actually the dominant heating contributor, relative to convective, over much of the backshell surface. This talk presents the story of how this backshell radiative heating component was ignored for so long for Mars entry, how it was discovered, and how its modeling was refined and utilized for Mars 2020.
Chris Johnston has worked at NASA Langley since 2006, with the main focus of modeling shock-layer radiative heating. He is the primary author of the HARA radiation code, which is one NASA’s standard aerothermal tools. Has worked on the Mars 2020 and Orion programs, and is currently working on the Mars Sample Return Earth Entry vehicle.
Online Seminar on SU2-NEMO: an open-source framework for nonequilibrium flows (open to public)
Prof. Marco Fossati from the University of Strathclyde
Prof. Fossati is associate professor in computational aerodynamics and the head of the Future Air-Space Transportation Technologies Laboratory. His research interests are in the area of multiphysics computational aerodynamics, and his expertise is in the field of aircraft aerodynamics and non-equilibrium flows, modal-based Reduced Order Modeling for aircraft aerodynamics, mesh optimisation and generation.
The talk will give an overview on the developments of an open-source code to address high-enthalpy non-equilibrium flows called SU2-NEMO. NEMO is the outcome of a collaborative effort between the University of Strathclyde, the VKI, Stanford University, and the University of Arizona. The rationale behind NEMO is to develop open-source simulation capabilities to address nonequilibrium physics such as finite-rate chemistry and nonequilibrium energy transfer that characterise the aerothermodynamic interaction of objects flying at high Mach regimes.
Online Seminar on Machine Learning Moment Closures for Accurate and Efficient Simulation of Polydisperse Evaporating Sprays (open to public)
Dr. James B.Scoggins, Postdoctoral researcher at the von Karman Institute
Dr. James B. Scoggins to the AR seminar, where he will talk about his paper presented at this year's AIAA SciTech conference:Machine Learning Moment Closures for Accurate and Efficient Simulation of Polydisperse Evaporating Sprays
He will present a novel machine learning moment method for the closure of the moment transport equations associated with the solution of the Williams-Boltzmann equation for polydisperse, evaporating sprays. The method utilizes neural networks to learn optimal closures approximating the dynamics of the kinetic equation using a supervised learning approach. The neural network closure is compared to reference solutions obtained using a Lagrangian random particle method as well as two other state-of-the-art closure models, based on the maximum entropy assumption. Results on 0D and 1D test cases demonstrate that the closures obtained using the machine learning approach is significantly more accurate than the maximum entropy closures with comparable CPU performance. This suggests that such models can be used to replace expensive Lagrangian techniques with similar accuracy at far less cost.
Online Seminar on Characterization of Supersaturated Nitrogen In Hypersonic Wind Tunnels & The Design Of A Fast-Acting Valve In ANDLM6QT (restricted to VKI member)
Erik Hoberg, PhD at University of Notre Dame, USA
Erik is PhD candidate at the University of Notre Dame and he received the 2020 BAEF fellowship to study for 6 months at VKI
Erik received his bachelors degree in aerospace engineering in 2017 at New Mexico State University and his masters in aerospace engineering from the University of Notre Dame January of this year. At Notre Dame, he has worked on flow characterization and design in the arc heated hypersonic wind tunnel and the large hypersonic quiet tunnel.
Seminar on Bayesian aerothermal assessments
Pranay Seshadri from Alan Turing Institute, United Kingdom
Pranay Seshadri is the Group Leader in Aeronautics at the Alan Turing Institute — the UK’s national institute for data science artificial intelligence. He is concurrently a Research Fellow in the Department of Mathematics at Imperial College London. He obtained his PhD in 2016 from the University of Cambridge in robust turbomachinery design. After his approx. 30min talk, Pranay will continue with a mini-workshop on 'equadratures': an open-source code for uncertainty quantification, data-driven dimension reduction, surrogate-based design optimisation and numerical integration. The workshop will finish around 15:00.
Abstract of the talk and mini-workshop:
Abstract (for talk):
In this talk, I will give you an overview of my group’s research that is broadly focused on Bayesian aerothermal measurements — the science and statistics of inferring aerothermal quantities. The “Bayesian” perspective arises from both uncertainties in the measurements and uncertainties in regions where we do not have measurements. I will present two case studies of some of the underlying research that cuts across statistics and aerodynamics. The first case study is the measurement of jet-engine sub-system (fans, compressors, turbines, etc.) efficiencies, where the measurements are sparsely placed pressure and temperature rakes. The second case study is the quantification of density from old Schlieren images, where the sensors are densely sampled pixels that measure the refractive index gradient field.
Abstract (for mini-workshop):
This mini-workshop is focused on ‘equadratures’: an open-source code for uncertainty quantification, data-driven dimension reduction, surrogate-based design optimisation and numerical integration (see: https://equadratures.org/). Although we will not have time for tandem coding, I will aim to show code snippets for some of the aforementioned capabilities, so you can try running the code on your own data-sets. There will be quite a few turbomachinery examples presented, including your very own LS-89.