Research Internship Opportunity - 6 months
Institut Pprime - CNRS
Location: Institut Pprime, Poitiers, France
Team: Curiosity Research Group
Supervisors:
Philippe Traoré philippe.traore@univ-poitiers.fr
Lionel Agostini lionel.agostini@cnrs.fr
Duration: 6 months
Starting Date: Flexible (2024)
Phd funding available
I. Context & Objectives
An investigative program is proposed, focusing on the efficacy of controlling
near-wall turbulence through the use of optimised active devices. These
devices are designed to generate substantial, meandering streamwise vortices
by capitalising on secondary flow instabilities that manifest over a concave
wall. The features of these vortices will be regulated by plasma actuators with
the aim of maximising the capacity for heat transfer. Improving the performance
of heat exchangers is a critical technological challenge essential for
enhancing efficiency and cost-effectiveness in engineering systems where heat
transfer is a fundamental process. With the swift global increase in energy
consumption in recent decades, the demand for innovative concepts, methodolo-
gies,
and designs has significantly risen. Regardless of the application,
whether in heating or cooling processes, enhancing the heat transfer capacity
of exchangers is a crucial step towards achieving superior efficiency. This
research is intended to fulfil this requirement by managing the generation of
streamwise vortices to increase heat transfer while keeping the pressure loss
to a minimum.
A specific application of this approach can be found in the heat exchangers,
such as the ones installed in the engine flow paths . A significant challenge
that arises during engine systems design is the requirement for the heat
exchanger to encompass a broad spectrum of conditions within the operational
range, including the accommodation of rare events like extreme hot day
conditions. They refer to ambient conditions with a ground temperature of 54 °C.
As this condition is within the envelope of the engine specification,
the heat exchangers must be designed to accommodate it, even though it
is extremely rare. Typically, it is estimated that such conditions will be
experienced
less than 5 times in the life of an aircraft engines. However, the
extreme hot day condition ends up sizing the heat exchanger, resulting with
a large thermal margin on typical days, with no real added operational value
and the associated air-side pressure loss, resulting in a specific consumption
and fuel burn penalty.
The study focuses on flow over a concave wall, illustrated in Fig.1. The
wall’s curvature induces centrifugal forces, leading to secondary flow
instabilities,
-i.e. the formation of coherent streamwise vortices, known as Görtler
vortices. The primary objective is to understand the impact of these structures
on heat transfer in turbulent wall-bound flow. Subsequently, a control
law will be designed using plasma actuators to provide an optimal control
strategy that maximises heat transfer at minimum drag penalty.
Many techniques or actuators have emerged depending on the flow configurations
and desired effects. Among all these methods, a new technology
based on the use of cold plasmas is currently expanding rapidly. These types
of plasma actuators, called Dielectric Barrier Discharge (DBD), rely on the
optimal use of the ionic wind induced by the plasma discharge. They generally
consist of a system of electrodes installed on one of the walls of the flow
domain to be controlled. By applying a sufficient potential difference between
these electrodes and according to their respective positions, a plasma
discharge is generated. This discharge allows a transfer of energy to the fluid
which results in the creation of an ionic wind which is in fact a direct
conversion
of electrical energy into kinetic energy. Thanks to this process, plasma
discharges can be used to create a tangential flow on the wall in order to
accelerate the flow and especially modify the velocity profile in the boundary
layer. They can also be used to induce a flow perpendicular to the wall this
time in order to increase the turbulent intensity of the flow. The aim is to
trigger and control instabilities promoting large-scale mixing.
The efficiency of DBD plasma actuators greatly depends on their positioning
on the wall as well as on numerous control parameters :
- Number of électrodes,
- distance between electrodes and electrode length,
- amplitude of the electrical potential difference or electrical power,
- shape of the alternating electrical signal (square, triangular,
sinusoidal),
- frequency of the signal.
This wide variety of possible configurations makes these devices highly
modular and adaptable. Moreover, the interest in plasma actuators lies in
particular in their non-intrusive character in the flow and their low energy
implementation cost, as well as their very short response times compared to
traditional mechanical actuators. However, their main disadvantage is that
the amplitude of the perturbations directly imposed on the fluid remains
limited. This is why we wish to exploit a different approach using plasma
actuators,
not to directly force the flow, but to drive and amplify instabilities
naturally present in the boundary layer. The objective is thus to take advantage
of secondary instabilities, such as longitudinal vortices, which have
their own dynamics. The plasma control then aims to optimally excite these
coherent structures in order to maximise the desired effects, in particular the
increase of near-wall heat transfers.
The combinations offered by these devices are numerous. The use of numerical
simulation and Artificial Intelligence in this context is an essential
asset in order to optimize the plasma discharge process according to the
targeted application, thus avoiding too many experimental tests which can some-
times be expensive to carry out. The numerical tool also makes it possible to
better understand the physics of this energy exchange by simulating the behavior
of dif-ferent theoretical models. Artificial Intelligence makes it possible to
optimize the op-eration of the actuator by adjusting the different parameters
that determine its opera-tion.
II. Research tasks
In the context of our numerical simulation activities in EHD (Electro-
Hydro-Dynamics), we have developed a computational code : Oracle3D [3],
for solving the Navier-Stokes equations coupled with the Maxwell equations.
This code written in Fortran 95 is based on the finite volume method. It
is a block-structured code capable of handling non-orthogonal hexahedra to
simulate flows in complex geometries. The code is parallelized using the MPI
library.
The missions of the PhD student will consist in performing direct numerical
simulations (DNS) to analyze the heat transfer enhancement by controlling
secondary instabilities such as Görtler vortices. First, a parametric study
will be conducted by testing different plasma control parameters to identify
their effects on the flow dynamics, and more specifically on the heat transfer.
These parameters will be defined manually based on observations of the
effects of plasma actuators on the flow dynamics. Although the probability
of identifying the optimal control law through this approach is limited, it is
not impossible (see work on drag reduction [2]).
The parametric study aims at facilitating and accelerating the subsequent
implementation of Deep Reinforcement Learning, by first exploring the parameter
domain and acquiring knowledge that will make it possible to better
guide and constrain the learning algorithm. For this purpose, the objectives
are :
- To perform a broad but coarse exploration of the control parameter
domain in order to restrict it for the DRL. Thus, with the search
domain being more limited, it will allow the DRL to converge more
quickly towards a robust solution.
- To improve the understanding of the physics of controlled flows, this
will facilitate the design of relevant cost and reward functions, by
incorporating
physical knowledge. With cost and reward functions based
on physics, the DRL will be more effective at finding the optimal control law.
III. Candidate profile and prerequisites
This thesis involves digital development. The candidate should have programming
skills and an interest in numerical simulation. Knowledge of one
of the following languages - Fortran, C/C++ or Python - would be appreciated.
Knowledge of parallelizing calculations using libraries such as MPI and
PETSc would be an asset but is not essential.
In terms of theoretical knowledge, the candidate will need to have a good
foundation in fluid mechanics as well as, ideally, an understanding of machine
learning and/or control theory.
The ideal profile is someone with a Master’s degree in Fluid Mechanics
or Applied Mathematics, with some initial experience in programming and
numerical simulation.
More generally, the desired profile is a motivated, rigorous and autonomous
student, with strong skills in computer programming as well as a strong
interest in numerical simulation in fluid mechanics and artificial intelligence.
IV. Application process
Send by email to lionel.agostini@cnrs.fr & philippe.traore@univ-poitiers.fr
with subject “internship_DBD_application_#yourname” :
- Resume
- Transcripts from Master 1 & 2
- Contact information for two possible references
|
