30.07.2013 Views

WAVES AND VIBRATIONS IN INHOMOGENEOUS STRUCTURES ...

WAVES AND VIBRATIONS IN INHOMOGENEOUS STRUCTURES ...

WAVES AND VIBRATIONS IN INHOMOGENEOUS STRUCTURES ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

R<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

q1=1<br />

q2=1<br />

q1=1<br />

q2=3<br />

q1=3<br />

q2=3<br />

0<br />

0 0.1 0.2 0.3 0.4 0.5<br />

design variable<br />

0.6 0.7 0.8 0.9 1<br />

R<br />

1<br />

0.9<br />

0.8<br />

0.7<br />

0.6<br />

0.5<br />

0.4<br />

0.3<br />

0.2<br />

0.1<br />

q1=0.25<br />

q2=0.25<br />

q1=1<br />

q2=0.25<br />

q1=1<br />

q2=1<br />

0<br />

0 0.1 0.2 0.3 0.4 0.5<br />

design variable<br />

0.6 0.7 0.8 0.9 1<br />

Figure 2 - Reflection, R, versus the design variable ̺, left: polynomial interpolation (Eq. (6)),<br />

right: inverse polynomial interpolation (Eq. (7)), for different values of the parameters q1 and<br />

q2. Parameters: ρ2/ρ1 = κ2/κ1 = 500.<br />

Fig. 2(right) shows reflection versus ̺ for different parameter values. The curves are<br />

clearly more well-behaved with a non-vanishing slope for ̺ → 1. The different parameter<br />

combinations have been used in the optimization algorithm and the simplest<br />

choice of q1 = q2 = 1 yields good results with well-defined air and solid regions in the<br />

optimized designs. Naturally, a simple 1D model cannot fully account for all necessary<br />

properties, and more work should be put into fully understanding the effect of the<br />

choice of interpolation functions.<br />

DISCRETIZATION <strong>AND</strong> SENSITIVITY ANALYSIS<br />

To solve the model equation (with the appropriate boundary conditions) and the optimization<br />

problem, we discretize the complex amplitude field p and the design field ̺<br />

using finite elements:<br />

N<br />

Nd <br />

p = φipi, ̺ = ψi̺i, (8)<br />

i=1<br />

where φi and ψi are basis functions, and the vectors p = {p1 p2 ...pN} T and ̺ =<br />

{̺1 ̺2 ...̺Nd }T contain the nodal values of the two fields. We use the commercial<br />

FEM package FEMLAB on top of MATLAB for assembling and solving the discretized<br />

equation:<br />

S(̺)p = f(̺). (9)<br />

Throughout this work we use a triangular element mesh, and a quadratic approximation<br />

for the pressure field and a linear approximation for the design field.<br />

We now derive the sensitivities of the objective functional Φ(̺,p) with respect<br />

to a single design variable. We use the adjoint method and obtain the adjoint equation<br />

as:<br />

S T λ = −1 ∂Φ<br />

( − i 2 ∂pR<br />

∂Φ<br />

)<br />

∂pI<br />

T , (10)<br />

4<br />

i=1

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!