When aircraft begin to vibrate in flight, things get uncomfortable – or even dangerous in the worst case. Flow-induced oscillations on wings or tailplanes can have catastrophic consequences. These are exactly the phenomena we study at the Institute of Aeroelasticity. Our mission: to understand when and why an aircraft begins to oscillate in the air – and how to prevent it. Two of the most well-known types of such oscillations are flutter and buffeting. Traditionally, they have always been viewed as two distinct phenomena. But our current research shows that this strict separation doesn’t hold up under closer scrutiny.
So, what is flutter? In the classical sense, flutter refers to a condition in which an elastic structural component – such as a wing – becomes unstable as the flow speed increases. Two of its natural modes, such as a bending and a torsional mode, couple via unsteady aerodynamic forces and begin to amplify each other. Once a certain speed is exceeded – the so-called flutter boundary – the system no longer settles down: the oscillations grow exponentially. This is extremely dangerous – and ensuring safety against it, both computationally and experimentally, is a major part of every aircraft certification process.
Buffeting, on the other hand, has traditionally been understood differently. Here, the assumption is that the flow itself becomes unstable – for example, at transonic speeds with significant flow separation. This unstable flow generates oscillating aerodynamic forces, which then set the aircraft structure in motion. The structure responds – but it is not considered to be causally involved. The critical region begins here with the so-called buffet onset, which can already be identified using a rigid wing geometry in flow simulations or wind tunnel tests – at least, according to the traditional theory.
However, in our numerical simulations we were able to show: Buffeting, too, can be understood as a stability problem – just like flutter. The difference is this: it’s not two structural modes that couple, but a structural mode and a low-frequency fluid-dynamic mode. The latter typically arises in transonic flows with mild boundary layer separation. The flow then exhibits damped natural oscillations on its own. When an elastic structure is added to the system, the result is again a coupled system that can become unstable.
At first glance, these resulting oscillations might be mistaken for a pure structural response. However, their frequency is only close to – but not exactly equal to – the original frequency of the pure fluid-mechanical buffet instability. In addition, it becomes evident that the supposed buffet onset seems to depend on the elasticity of the responding structure – which means it is no longer purely a feature of the flow. That shouldn’t be possible under the old way of thinking. The only explanation: we are not observing a passive structural response but rather flutter “with limited amplitude” – so-called limit cycle oscillations. The traditional distinction between "response" (buffeting) and "instability" (flutter) begins to blur.
In CFD simulations, we were able to demonstrate the transition from classical flutter in attached flow to the so-called fluid-mode flutter in separated flow for various configurations. Particularly striking: in the separated flow regime – precisely where flutter was long thought to be irrelevant – these new instabilities appear. And in most cases, only one structural degree of freedom is required. The second degree of freedom is effectively provided by the fluid. Similar phenomena had been observed experimentally in the past – but the connection to buffeting was simply not recognized.
These findings have significant implications for flutter assessment in aircraft design. We may need to perform flutter analyses in parts of the flight envelope previously considered uncritical – for example, at high local angles of attack, such as during certain maneuvers or when encountering gusts. The good news: all the necessary simulation and analysis tools are essentially already available; we just need to slightly adjust our approach.
So far, our results come only from computer simulations. The next step is experimental validation: If we could demonstrate that the supposed buffet onset is in fact a generalized flutter boundary – one that depends on the elasticity of the wind tunnel model – it would be the proverbial “smoking gun.” In a joint research effort with scientists from Japan’s JAXA and France’s ONERA, we are currently working on this. In one or two years, we’ll know more.
Jens Nitzsche, Department of Aeroelastic Simulation, DLR Institute of Aeroelasticity
