12/22/2009
1
Optimisation for Thermo-Fluids
Engineering
Dr. R.J.M. (Rob) Bastiaans
Combustion Technology
Mechanical Engineering
4M020 Design Tools
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation
Eggert 2000: Engineering design
� Engineering design is the set of decision-making
processes and activities used to determine the form of
an object given the functions desired by the customer.
� During the parametric design phase we determine values
for the controllable parameters, called design variables,
identified as unknown during the configuration phase.
� CAE refers to computer software and hardware systems
used in the analysis of engineering designs to validate
functional performance.
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Thermo-Fluids Engineering
What is Thermo-Fluids Engineering
� Covered by
� Energy Technology
� Process Technology
� Combustion Technology
� Common factor: Fluid flow
� Often multi-scale multi-physics problems
� Much research less optimal design
� Implication on how to use computer-capacity
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Fluid Flow
Many problems in many areas
� Meteorology
� Astrophysics
� Biology
� Agriculture
� Process technology
Common factor: Navier Stokes Equations
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Multi-Scale flows
Examples
� Turbulence
� Atmospheric dispersion
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Multi-Physics
Often flow is not the problem but interactions are
� Buoyancy induced flows
� Mixing of different fluids
� Dispersion of pollutants
� Flows with heat transfer
� Reactive flows; combustion
� Compressible flows
� Acoustics
� Shock waves
� MHD (Magneto Hydro-Dynamics)
� Flow structure interaction
� Combinations of the above
12/22/2009
2
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Multi-Scale Multi-Physics flows research
Examples
� Turbulent combustion:
� Compressible flow
� Heat transfer
� Many chemical species and reactions
� Acoustics, stability
� Flame-thickness independent length scale
� Application: Gas-turbines for aeroplanes and el. power generation
� Very important for society: Emissions, Climate, Energy
� Optimisation??
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Multi-Scale Multi-Physics flows research
Gas turbines:
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Modelling of reactive flows
Turbulent combustion
� Not only interesting from an industrial point of
view, but also from an academic point of view
� Large range of time and length scales makes
numerical simulation of turbulent combustion
far from easy and very expensive
� Development of accurate and efficient models
for turbulent combustion is one of the most
challenging tasks facing the combustion
community today
DL
R, G
erm
an
y
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Approach
From small to large scale, from fundamentals to application
� One-dimensional flame modelling with detailed description of
chemistry and transport
� Flamelet-based reduction (FGM) to simplify chemistry model
� Direct numerical simulation (DNS) of turbulent flame to unravel
chemistry-turbulence interaction
� Model for turbulence-chemistry interaction (e.g. a sub-grid scale
model for large-eddy simulations)
� Large-eddy simulation (LES) of lab-scale flames
� Reynolds-averaged Navier-Stokes (RANS) simulations of
industrial applications
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
DNS of turbulent flame kernels
DNS of spherically expanding premixed
turbulent flames
� Validation of FGM vs detailed chemistry
� Analyse turbulence/chemistry interaction
� Practical relevance is found in IC engines
Leeds, UK
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Validation of FGM
Mass fraction of OH radical. FGM 100 times faster than detailed chemistry!
This enables paramteric studies.
12/22/2009
3
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
From DNS to LES of reactive flows
� Averaging DNS results enables a-priori testing of LES sub-grid scale
models.
� Application of LES-FGM in premixed turbulent Bunsen flame:
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
RANS of gas turbine combustor
NOx formation in gas turbine combustor
� Fired in lean premixed mode:
� Fired in diffusion mode (start-up): NOmax is ~100x larger
� Investigate the influence of hydrogen addition
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Application to biomass conversion
� Application of fundamental knowledge of reactive flows to biomass
conversion
� Multi-scale, multi-physics approach:
� Small scale: single particle
kinetics, pyrolysis, heat/mass transfer
� Intermediate scale: fixed/fluidized bed
two-phase flow, heat/mass transfer
� Large scale: reactor, furnace
flow pattern, radiation, control
Info
rma
tion
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Multi-scale multi-physics
� Multi-scale multiphysics problems require multi-scale multi-physics approach
� Because physics at the smallest scales can have a large impact on large
scale design constraints
� NOx formation in turbulent combustion has its origin in very thin oxidation layers
� typically 100 µm
� combustor: typically 1 m
� for accuracy you need 10 points for a relevant gradient to resolve: 10 µm grid
� 3D problems require (105)3 =1015 calculation points
� Time dependent problems require 105 timesteps
� Problem size: n x 1020 operations.
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Large computer power for solving physics
Turbulent combustion in gas turbines
http://www.cerfacs.fr/cfd/
Helicopter engine, using 2000 procs:
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Large computer power for solving physics
Big machines required:
Worldwide:
http://www.top500.org/list/2008/06/100
Netherlands:
http://www.sara.nl/
(Huygens)
12/22/2009
4
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation of design
Calculating for design
� BOEING (conservation)
� Bernoulli
� Potential flow models
� Potential flow with viscous layers
� Euler
� Navier-Stokes
� Incompressible
� Boussinesq
� Variable density
� Compressible
Analytical
Partial differential
equations
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Flow models and optimisation
General: Trade off
� Analytical/Exact but crude physics
� Accurate physics but crude approximations; numerics
� Accuracy very important in optimisation!!
� You cannot optimise in % if your calculation/prediction
accuracy is not on the same level
� Low Reynolds laminar steady flows (nano-technology,
micro compact, heat exchangers, lab-on-a-chip)
� Control problems (e.g. suppressing vortex shedding)
� Future: Optimisation for more and more complex problems
General: Trade off/limitations
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation in CFD
Approach
� Good strategy requires detailed knowledge of fluid dynamics and
computational methods.
� Use of Computational Fluid Dynamics (CFD),
� Solving sets of partial differential equations (PDE’s);
� started in the 80s;
� philosophy: calculate and analyse a certain design
� Recent: B. Mohammadi & O. Pironneau, Shape Optimization in Fluid
Mechanics, Annual Rev. Fluid Mech., 2004, 36, pp 255-279.
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation in CFD
Approach
� Highly educated designer determines the parameters to optimize
� Changing parameters will often lead to edge (predefined) optima
� Pushing the boundaries will often result in changing physics
� e.g. becomes unsteady, no Stokes flow anymore etc.
� CFD learns designers what parameters might be important
� Parameter study learns designers to really look at new concepts
� Automatic optimization in multi-parameter/multi-physics still far away
� Multi-parameter is domain of mathematicians, genetic algorithms etc.
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation in CFD
Certainly a great future for
Opimisation in CFD!!� Relatively unexplored
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation in CFD
Examples:
12/22/2009
5
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation in CFD
Examples:
� Q. Li et al., Evolutionary Structural Optimization, Int.J.Heat Mass Transfer, 42, (1999)
� A. Gersborg-Hansen et al., Topology optimization of channel flow problems, Struct.
Multidisc. Optim., 30, (2005)
� D.N. Srinath, S. Mittal, A stabilized finite element method for shape optimization in
low Reynolds number flows, Int. J. Num. Meth. Fluids, 54, (2007)
� A. Gersborg-Hansen et al., Topology optimization of heat conduction problems
using the finite volume method, Struct. Multidisc. Optim., 31, (2006)
� D.E. Hertzog et al., Optimization of a microfluidic mixer for studying protein folding
kinetics, Anal. Chem., 78, (2006)
� H. Antil et al., Optimal design of stationary flow problems by path-following interior
point methods, Struct. Multidisc. Optim., submitted (2007)
� L.Debiane et al., Temperature and pollution control in flames, Proc. Summer Progr.,
Center for Turbulence Research, (2004).
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation in CFD
Examples:
� Limited application
� Mathematical methods
� Limited parameter space
� Hertzog et al.:
� Navier Stokes
� Convection-Diffusion
� Comsol
� 40 % reduction in mixing time
� Debiane et al., Center for Turbulence Research 2004:
� Application in flames
� I was there but I found this article only yesterday!!
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation in CFD
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation in CFD
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation in Comsol
Conclusion:
Let us just start ourselves with an experiment in Comsol:
Double glazing:
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation in Comsol
� Optimise double glazing design
� Optimisation parameters:
� Minimization of heat flux
� Maximization of acoustic isolation
� Maximization of mechanical strength, resitance to impact
� Minimization of costs
12/22/2009
6
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation in Comsol
� Hypothesis 1
� The thicker the air layer the more isolating
� But the air is not stagnant, so
� Hypothesis 2
� At larger distance, L, the Ra number becomes higher
– Third power:
– Flow becomes more vigorous
– Eventually instationary
– Heat transfer by convection increases
– More heat losses
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation in Comsol
� Physical problem:
� Conduction
� Natural convection
� Partial differential equations:
� Convection and Conduction (CC)
� Navier-Stokes equations (NS)
� Mutual influence
• Buoyancy force as function of T solved by CC in NS
• Velocities for convection of heat, solved from NS in CC
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation in Comsol
� Setup and boundary conditions
� All (other) walls: no slip, adiabatic
� H=0.1 m, L(initial)=0.01 m, d=0.002 m
T=320 K T=280 K
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Equations in Comsol
� Equations
T=320 K T=280 K
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation in Comsol
� Parameters:
� Distance between glazing
� Thickness of the glass
� Height of the glass/how to simulate full height
– Variation of height
– Inflow/outflow
� Relevant temperature difference
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation in Comsol
� Parameters:
� Physical properties of the glass
– Conductivity
– Density
– Heat capacity
� Physical properties of the medium (argon, water)
� Pressure: What isolates better
– Low pressure (low density, capacity)
– High pressure (higher force needed for momentum)
� Instationary behaviour?
12/22/2009
7
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation in Comsol
� Setup the model in Comsol; save in Matlab
� What do we do with the pressure?
� We are going to change the geometry, what does
this mean for the gridding?
� Constants:
� Air: density 1.2, k=0.025, Cp=1006, eta=1.7 10-5
� Glass: density 2500, k=1.1, Cp=840
� Units?
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation in Comsol
Heat flux analysis:
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation in Comsol
Flow analysis:
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Optimisation in Comsol
Flow analysis:
Instationary
behaviour?
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Assignment
3 possibilities:
� Double glazing: more parameter variations
– Research possible unsteady behaviour
– Influence of glass thickness
– Use argon and water (determine changes in
Ra and Pr in advance)
� New: Cilinder in a box, disturbing convection
– Box is a lid driven cavity
– Scalar flux (temperature, species) at the top
– Fixed value at the bottom
– Research influence of position and size of a
cilinder, with no slip walls.
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Assignment
Cilinder in a box:
12/22/2009
8
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Assignment
Cilinder in a box:
– Determine base flow
– Add the scalar problem
– Put cilinder in
– Vary, determine cost and analyse
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Assignment
3rd possibility:
Your own optimisation problem
(in this case you need to know and discuss with
me today)
4M020 Design Tools; Optimisation in Thermo-Fluids Engineering
Further information:
Dr. R.J.M. Bastiaans (Rob)
Combustion Technology
Mechanical Engineering, WH 3.141Eindhoven University of Technology
P.O. Box 513, 5600 MB Eindhoven, The Netherlands
E: [email protected]: +31 40 2474836
F: +31 40 2433445
www.combustion.tue.nl