Aeroelasticity Analysis
Aeroelasticity Analysis: Theoretical Foundations
What is Aeroelasticity?
Professor, aeroelasticity is a coupled problem of aerodynamics and structures, right? Why is it considered particularly important?
Aeroelasticity is explained by Collar's triangle. It's a problem where three forces combine: Aerodynamic Forces, Elastic Forces, and Inertial Forces. The interaction of these three forces causes dangerous phenomena like flutter, divergence, and buffeting.
Especially flutter is a self-excited vibration caused by the coupling of aerodynamic and structural elastic forces, where the amplitude grows exponentially leading to structural failure. In aircraft design, proving that flutter does not occur within the flight envelope is a mandatory requirement for type certification.
2-Degree-of-Freedom Flutter Model
Please explain the basic mechanism of flutter.
A typical airfoil flutter is explained by a 2-degree-of-freedom model of bending (plunge, $h$) and torsion (pitch, $\alpha$). The equations of motion are:
Here $m$ is mass, $S_\alpha$ is the static unbalance moment, $I_\alpha$ is the moment of inertia, $K_h$, $K_\alpha$ are spring constants, $L$ is lift, and $M_{ea}$ is the moment about the elastic axis.
How is the condition for flutter onset determined?
Since unsteady aerodynamic forces are a function of velocity, as the velocity increases, the frequencies of the bending and torsion modes approach each other (frequency coalescence), and at a certain velocity, the net energy transfer becomes positive, causing the vibration to diverge. This critical velocity is the flutter speed $V_F$.
Here $\sigma$ is the eigenvalue of the system.
Theodorsen's Unsteady Aerodynamic Theory
How are unsteady aerodynamic forces calculated?
Classically, the Theodorsen function $C(k)$ is used. The unsteady lift for a harmonically oscillating airfoil is:
Here $k = \omega b / U$ is the reduced frequency, $b$ is the semi-chord length, and $a$ is the elastic axis position. $C(k)$ is expressed using Bessel functions, converging to $C \to 1$ for $k \to 0$ (quasi-steady) and $C \to 0.5$ for $k \to \infty$.
So if we use CFD, we don't need to rely on the Theodorsen function, right?
Exactly. In CFD-based aeroelastic analysis, unsteady aerodynamic forces are calculated directly from CFD, enabling predictions beyond the applicability limits of theoretical models (linear, potential flow), such as transonic flutter and large-amplitude oscillations.
The Lockheed L-188 Disaster Changed Aeroelasticity
In 1960, a series of Lockheed L-188 Electra aircraft disintegrated in mid-air due to propeller resonance flutter. The investigation revealed that a design flaw in the propeller's vibration damper changed the natural frequency, inducing flutter under certain flight conditions. Following this accident, the submission of aeroelastic flutter analysis became mandatory for aircraft type certification in the United States. This is a classic example where an accident forced the implementation of theory, and flutter theory is now ingrained as a fundamental rule in aircraft design.
Computational Methods for Aeroelasticity Analysis
Classification of CFD-Based Aeroelastic Analysis
What approaches are there for aeroelastic analysis using CFD?
The most commonly used in practice is CFD + Modal Analysis. The structure is represented by an expansion in its natural modes, and the time evolution of the generalized coordinates $q_i(t)$ for each mode is solved coupled with unsteady aerodynamic forces from CFD.
Here $Q_i$ is the generalized aerodynamic force calculated from CFD. A standard workflow is to obtain structural modes from MSC Nastran SOL 146 (Flutter Analysis) and pass those mode shapes to the CFD solver.
How is data transferred between Nastran and the CFD solver?
Using Ansys System Coupling, mesh displacement and surface pressure can be automatically mapped between Fluent (fluid) and Ansys Mechanical (structure). For coupling with Nastran, you can build a custom workflow via Fluent UDF to pass modal coordinates, or use dedicated tools (e.g., MSC FlightLoads, Zona ZAERO).
V-g Method and p-k Method
What is the standard method for determining flutter speed?
The v-g method (velocity-damping method) and p-k method for linear flutter analysis are the basics.
V-g Method: Assumes harmonic oscillation at each velocity and determines the required structural damping $g$. The velocity where $g = 0$ is crossed is the flutter speed. Nastran's SOL 145 uses this method.
p-k Method: Determines eigenvalues in the time domain and plots damping ratio and mode frequency as functions of velocity. The p-k method gives physically more correct damping information and can also be used for damping estimation in the subcritical region. Nastran's SOL 145 PK option.
In CFD-based flutter analysis, can't these classical methods be used?
For CFD, the most direct method is the "time-marching method," which calculates the unsteady time history directly and observes whether the response damps or diverges. The flutter boundary is identified by incrementally increasing the velocity parameter. However, since computational cost is high, it's efficient to first narrow down the approximate range using linear theory (DLM: Doublet Lattice Method + Nastran SOL 145/146) and then refine with CFD.
DLM and CFD Correction
Is DLM (Doublet Lattice Method) still used?
DLM is still the mainstay for aircraft flutter certification. It can efficiently calculate unsteady aerodynamic forces in the frequency domain under potential flow assumptions. However, it cannot handle shock wave effects in the transonic regime, so "CFD-corrected DLM," which corrects DLM aerodynamic forces using steady pressure distributions calculated by CFD, is widely used.
"Method Selection" for Aeroelastic Analysis Has Always Been a Headache
Before CFD-based aeroelastic analysis became widespread, designers used flutter calculations combining linear panel methods and structural modes. Accuracy was rough but computation was fast. Using CFD improves accuracy but skyrockets cost. Records remain of an Airbus development team seriously debating in the late 1990s "how much CFD should be used," concluding that "nonlinear CFD is unnecessary if the required accuracy is within ±5% of the flutter speed." The importance of having criteria for method selection remains unchanged today.