Optimization and Computational Fluid Dynamics - Department of ...
Optimization and Computational Fluid Dynamics - Department of ...
Optimization and Computational Fluid Dynamics - Department of ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
3 Mathematical Aspects <strong>of</strong> CFD-based <strong>Optimization</strong> 71<br />
Consider the following (locally) convex optimization problem<br />
min gk(pk):=fk(yk(pk),pk) (3.35)<br />
pk<br />
where k =1, 2,...,L is the resolution or discretization parameter, where<br />
L denotes the finest resolution, <strong>and</strong> pk is the (unconstrained) optimization<br />
variable in the space Vk. For the resolution k, one chooses an appropriate<br />
discretized state variable yk <strong>and</strong> an accordingly resolved objective evaluation<br />
fk. For variables defined on Vk, we introduce the inner product (·, ·)k with<br />
associated norm �y�k =(y,y) 1/2<br />
k . Between spaces Vk, restriction operators<br />
I k−1<br />
k<br />
: Vk → Vk−1 <strong>and</strong> prolongation operators I k k−1 : Vk−1 → Vk are defined.<br />
We require that (I k−1<br />
k y,v)k−1 =(y,I k k−1 v)k for all y ∈ Vk <strong>and</strong> v ∈ Vk−1.<br />
On each space, denote with Sk an optimization algorithm, e.g., a gradient<br />
<strong>of</strong> the solution to (3.35),<br />
based technique. Given an initial approximation p0 k<br />
the application <strong>of</strong> Sk results in gk(Sk(p0 k )) 1 :<br />
1. Pre-optimization. Define p1 k = Sk(p0 k ).<br />
2. Set up <strong>and</strong> solve a coarse-grid minimization problem. Define p1 k−1 =<br />
I k−1<br />
k p1 k <strong>and</strong> σk−1 = ∇gk−1(y1 k−1 ) − Ik−1<br />
k ∇gk(y1 k ). The coarse-grid minimization<br />
problem is given by<br />
�<br />
min gk−1(pk−1) − σ<br />
pk−1<br />
T �<br />
k−1 pk−1 . (3.36)<br />
Apply one cycle <strong>of</strong> MG/OPT(k-1) to Eq. (3.36) to obtain p 2 k−1 .<br />
3. Line-search <strong>and</strong> coarse-grid correction. Perform a line search in the I k k−1 (p2 k−1 −<br />
p 1 k−1 ) direction to obtain τk. The coarse-grid correction is given by<br />
p 2 k = p 1 k + τk I k k−1(p 2 k−1 − I k−1<br />
k p 1 k) .<br />
4. Post-optimization. Define p 3 k = Sk(p 2 k ).<br />
Roughly speaking, the essential guideline for constructing gk on coarse<br />
levels is that it must sufficiently well approximate the convexity properties<br />
<strong>of</strong> the functional f(y(·), ·) at finest resolution. In addition, we have that the<br />
gradient <strong>of</strong> the coarse-grid functional at p1 k−1 = Ik−1<br />
k p1 k equals the restriction<br />
<strong>of</strong> the gradient <strong>of</strong> the fine-grid functional at p1 k . In fact, by adding the term<br />
−σT k−1 pk−1 in Step 2, we have that<br />
∇ � gk−1(pk−1) − σ T k−1 pk−1<br />
�<br />
|p1 = I<br />
k−1 k−1<br />
k ∇gk(p 1 k ) .