PASSIVE VIBRATION CONTROL VIA UNUSUAL GEOMETRIES - THE APPLICATION OFGENETIC ALGORITHM OPTIMIZATION TO STRUCTURAL DESIGN

In the majority of aerospace structures, vibration transmission proble ms are dealt with by the application of heavy, viscoelastic damping ma terials. More recently, interest has focussed on using active vibratio n control methods to reduce noise transmission. This paper examines a third, and potentially much cheaper method: that of redesigning the lo ad bearing structure so that it has intrinsic, passive noise filtratio n characteristics. It shows that very significant, broadband noise iso lation characteristics (of around 60 dB over a 100 Hz band) can be ach ieved without compromising other aspects of the design. Here, the gene tic algorithm (GA), which is one of a number of recently developed evo lutionary computing methods, is employed to produce the desired design s. The problem is set up as one in multi-dimensional optimization wher e the geometric parameters of the design are the free variables and th e band averaged noise transmission the objective function. The problem is then constrained by the need to maintain structural integrity. Set out in this way, even a simple structural problem has many tens of va riables; a real structure would have many hundreds. Consequently, the optimization domain is very time consuming for traditional methods to deal with. This is where modern evolutionary techniques become so usef ul: their convergence rates are typically less rapidly worsened by inc reases in the number of variables than those of more traditional metho ds. Even so, they must be used with some care to gain the best results . (C) 1995 Academic Press Limited