Академический Документы
Профессиональный Документы
Культура Документы
T=0
Optimal Topology
Optimal Temperature
Topology Distribution
{[ ]}
affecting the dominance of one physical process
ρ (u ⋅ ∇u ) = −∇P + ∇ ⋅ η ∇u + (∇u )T − αu . (5) relative to another. For simplicity, these
weighting values were selected manually.
Eq. (4) represents the fluid incompressibility
constraint, and Eq. (5) describes Navier-Stokes To determine the optimal steady state fluid flow
fluid flow. In these expressions ρ and η are the and channel layout the thermal conductivity and
fluid density and dynamic viscosity, inverse permeability of the porous medium were,
respectively. The inverse permeability of the respectively, interpolated using the penalty
porous medium, α, is assumed to be method from [1] and the convex interpolation
approximately valid for an actual porous scheme from [4]. These effective properties (i.e.
medium, per [4]. The state variables include the k and α) were interpolated via a main design
fluid pressure, P, and velocity field terms in the parameter, γ, which varied from 0 (minimally-
vector, u. porous, non-conductive solid) to 1 (conductive
fluid).
Additionally, the governing equation for steady
state convection-diffusion heat transfer is, 5.2 Computational Model Description
ρC (u ⋅ ∇T ) = ∇ ⋅ (k∇T ) + Q , (6)
To illustrate the multiphysics optimization
process a 2-D three-terminal structure, separated
into four subdomains, was considered; see
where C represents the heat capacity and k is the Figure 5. A domain roughly square in size and
thermal conductivity of the fluid. having a height a little larger than its width was
selected to provide slightly greater distance
A dual objective function, A2, was implemented between the two fluid outlet terminals. The
to optimize for both heat transfer and fluid flow. design domain was meshed with approximately
Specifically, the objective was specified to 6,200 quadrilateral elements.
minimize the mean temperature and total fluid
power dissipated in the system, Fixed temperature, parabolic normal fluid flow
was assumed at the single device inlet.
A2 = w1 B + w2 C , where (7) Convective flux, zero pressure normal flow was
assumed at both outlets. No-slip adiabatic
{ }
B = ∫ k (γ )(∇T )2 + ρC [T (u ⋅ ∇T )] dΩ ; (8) boundary conditions were enforced on all
external walls of the device. Uniform heat
Ω
generation, Q, and the main design parameter, γ,
were specified only on the primary design
domain.
Adiabatic, Zero Pressure,
No-Slip Walls Convective
Fixed Flux, Fluid
Temperature, Outlet
Parabolic
Flow,
Fluid Inlet Uniform Heat
Generation, Q
Zero Pressure,
Primary Convective
Adiabatic,
Design Domain Flux, Fluid
No-Slip Walls
Outlet
Figure 5: 2-D topology optimization design domain Figure 8: Pressure contours for optimal heat transfer
and boundary conditions for fluid flow and heat and fluid flow.
transfer.
The inlet fluid Reynolds number was set
arbitrarily to Re ≈ 50. A minimum Darcy number
of 10-4 was selected to represent a relatively
impermeable solid [4]. Fluid density, heat
capacity, and dynamic viscosity were all set to
unity. The solid volume fraction constraint was
set 0.6 for designs exhibiting a lesser amount of
solid material.
8. Conclusions
Figure 9: Optimal ‘Y-branch’ topology with pressure
The application of gradient based topology
contours and fluid flow vectors for minimum flow
power dissipation only (i.e. obtained with w1 = 0 and optimization within COMSOL Multiphysics via
w1 = 1 in Eq. (7)). a MMA optimizer was presented. An initial
single physics pure heat conduction problem was
Note that setting w1 = 0 and w2 = 1 in Eq. (7) selected to evaluate a custom MATLAB script.
leads to a simple ‘Y-branch’ topology and a Self-similar branching structures were obtained
reduced pressure drop, Figure 9, for minimum for this benchmark problem both in 2-D and 3-D.
fluid power dissipation. Thus, the relatively large The method was then extended to multiple
pressure drop observed in Figure 8 is an artifact physical processes including convection-
of a larger weighting given to the thermal portion diffusion and Navier-Stokes flow with
of the objective function, B, in Eq. (7). Thus, the corresponding objectives. The approach was then
topology in Figure 6, obtained for dual heat applied to optimize a 2-D three-terminal device
transfer and fluid flow objectives, is a logical having characteristics attributable to the various
superposition of the dendritic structural physics involved.
characteristics seen in Figure 3 and the
straightforward ‘Y-branch’ structure in Figure 9. Despite additional computational time needed for
The conclusion is that the relative values of w1 larger 3-D structures, this approach provides
and w2 in Eq. (7) have an important effect in interesting starting points for synthesizing
‘tuning’ the final result obtained through the effective thermal / fluid structures. Future
optimization process. research should focus on automating the
weighting strategy for multiple objectives to
better interrogate a prospective design space. The
7. Discussion application of this computational method to the
design of various vehicle systems is also a
The multiphysics optimization process described logical focus for future work.
in this paper may be applied to a variety of heat
transfer and fluid flow problems. Additional 9. References
physics including, for example, both static and
dynamic structural loading may be incorporated 1. M.P. Bendsoe and O. Sigmund, Topology
into the general process. Moreover, the optimization, 2nd Edn., pp. 71-272, Springer-
advantage of using an MMA optimizer is that it Verlag, Berlin (2004).
has been shown to be effective in handling 2. O. Sigmund, A 99 line topology optimization
optimization problems with multiple constraints. code written in MATLAB, Structural and
Multidisciplinary Optimization, 21, pp. 120-127
The primary limitation in applying this method (2001).
to a broader range of three dimensional problems 3. K. Svanberg, The method of moving
continues to be the computational time required asymptotes – a new method for structural
optimization, International Journal for
Numerical Methods in Engineering, 24, pp. 359-
373 (1987).
4. L.H. Olesen, F. Okkels, and H. Bruus, A high-
level programming-language implementation of
topology optimization applied to steady-state
Navier-Stokes flow, International Journal for
Numerical Method in Engineering, 65, pp. 975-
1001 (2006).
5. B. Lemke, Z. Liu, and G. Korvink,
Implementation of structural topology
optimization in COMSOL, COMSOL Users
Conference, Frankfurt, Germany (2006).
6. A. Bejan and S. Lorente, Natural design with
constructal theory, Mechanical Engineering, 13,
pp. 44-48 (2009).
7. J.N. Reddy and D.K. Gartling, The finite
element method in heat transfer and fluid
dynamics, 2nd Edn., pp. 7-10, CRC Press, Boca
Raton (2000).