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

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

1624 J. S. JENSEN<br />

h<br />

2h<br />

fcos Ωt<br />

Figure 11. A short cantilever subjected to a harmonic tip-load with frequency .<br />

is to be distributed so that a chosen objective function is minimized. The objective function is<br />

c3 = 1<br />

N<br />

utipūtip<br />

N k=1<br />

so that the response of the tip is chosen as a measure of the dynamic compliance.<br />

A plane stress model with unit thickness is considered and the cantilever is discretized using<br />

80 × 40 bi-linear quadratic elements. The load is applied in three nodes in middle of the right<br />

boundary. Damping coefficients are = 0.05 and = 0.002. The Matlab implementation from<br />

Section 4.1 is used, and a sensitivity filter (filter-size: 1.5×element-size) is included to avoid<br />

mesh-dependent solutions and checkerboard formation [23]. The initial design has 50% material<br />

uniformly distributed in the domain.<br />

Figure 12 shows the cantilever optimized for four different frequency intervals. If the frequency<br />

interval is in the low range ( = 0–0.02) well below the first resonance frequency of the initial<br />

design, then the optimized cantilever (Figure 12(a)) resembles the structure that is obtained for a<br />

static tip-load. The increased static and low-frequency dynamic stiffness is obtained by moving<br />

the first resonance frequency up in the frequency range. Figure 12(b) shows that if the frequency<br />

range is increased to = 0–0.04, the material is slightly redistributed so that the first resonance<br />

frequency is further increased and the high stiffness frequency range broadened. If the optimization<br />

interval is increased to higher frequency ranges, the optimized structures (Figure 12(c), (d)) change<br />

qualitatively.<br />

Figure 13 shows the corresponding responses for the four optimized structures as well as for<br />

the initial design. As mentioned before, the structures in Figure 12(a) and (b) have responses in<br />

which the first resonance frequency is moved up in the frequency range. Similar observations were<br />

made in [22]. However, the structures in Figure 12(c) and (d) obtain a low average response by<br />

a different mechanism: the static and low-frequency stiffness is lower (higher response) but an<br />

anti-resonance is introduced in the optimization interval. At the anti-resonance the tip response is<br />

small whereas the response of the rest of the structure is high. Thus, if a different measure of the<br />

dynamic compliance had been used, e.g. the total energy, these designs would not be optimal.<br />

It should be mentioned that the choice of N (number of expansion terms) is critical for this<br />

example. The initial problem in which the structure is homogeneous is very ill-conditioned and if<br />

N>2 the errors introduced in the PA computation and the corresponding sensitivities become too<br />

large with a double-precision implementation. Therefore, in the shown examples the first iterations<br />

are done with N = 2 and up to eight PA expansions. After a few iterations, say 15–20, N is<br />

increased (up to N = 5) and the number of expansions reduced to one or two.<br />

Copyright q 2007 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2007; 72:1605–1630<br />

DOI: 10.1002/nme<br />

(74)

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

Saved successfully!

Ooh no, something went wrong!