Time Domain Aeroelastic Analysis of an Aerofoil T. Varghese Mathew1 and Sharanappa V. Sajjan2 1Student, Dept. of Aerospace Engg., MVJ College of Engineering, Bangalore – 560067, India 2Computational and Theoretical Fluid Dynamics Division, CSIR-National Aerospace Laboratories, Bangalore – 560017, India ABSTRACT Fully implicit time-marching scheme is used to determine the flutter boundary of the NACA64A010, typical wing section of Isogai wing. Unsteady RANS equations are solved to predict the flow parameters required for the flutter estimation of the wing section. The implicit RANS solver is coupled with the governing structural equations motion. Dual time stepping approach is used to achieve the complete convergence of the flow in each time step. The response of the wing section to the flow fluctuations are used to determine the critical flutter index at each Mach number. Further the coupled tool has been used in prediction of flutter boundary of the aerofoil. The results are obtained in the form of stable, unstable and neutral oscillations and transonic dip for the aerofoil. The comparison of the computed results with the available data is good. Key words : Aero-elasticity, RANS solver, Transonic Flutter, Transonic dip NOMENCLATURE b Airfoil semi chord c Airfoil chord Cl Coefficient of lift Cm Coefficient of moment about elastic axis h Plunging displacement of the elastic axis ,positive down m Airfoil mass per unit span Mae Total moment about the elastic axis , positive nose up M∞ Free stream mach no. t Real time t* Pseudo time Uf Flutter speed α Angle of attack (degrees) Δα Pitching motion forcing amplitude Δt Implicit real time step μ Airfoil mass ratio, μ = m/πρb2 ρ Air density ωf Frequency of forced oscillation ωh,ωα Uncoupled natural frequency of section in plunge and pitch. τ Non-dimensional structural time , (τ= ωα*t) {F} Force matrix [K] Stiffness matrix [M] Mass matrix [C] Damping matrix {q} Displacement vector π’ β∩ Grid velocity π£ Velocity vector Introduction With the requirement for higher speed and maneuverability of aircraft, many high speed operating aircraft configuration were designed and brought to service. In-order to reduce the weight and drag in transonic regimes and to operate above the sonic speed the aircraft wing section thickness were reduced as it was found out that wave drag produced in the structure were directly proportional to the square of the thickness resulting in thin airfoils section. Due to this, the effect of inertia force at this flight regimes were of greater significance and its interaction with the elastic properties of the airfoil section resulted in emergence of the study of dynamic aero-elasticity especially flutter estimation. Transonic regime poses a challenge in predicting the flutter speed due to the flow non-linearity's occurring in this regime due to large non linear forces and moments being setup with small change in the aerodynamic configuration. The interaction of the moving shock and the boundary layer further complicates the problem for applying a linear approach. Isogai [1] showed the existence of a sharp decrease in the flutter speed of a swept wing in transonic flight (M∞ ≈ 0.7 − 0.9). This “transonic dip” phenomenon can only be predicted by including the flow equation with non-linearity in the model. Frane et al. [2] in his report stated that the in-capabilities of linear code in predicting the accurate flutter boundary. The report further stated that the use of Euler code predicted significantly lower flutter code compared to experimental result due to not including the viscous effect. His work states that in order to accurately predict the flutter index, viscous effects should be considered such as the boundary layer and the shock-boundary layer interaction. The computational work carried by Frane using the RANS model showed good agreement with the experimental results. Francisco Palacios et al. [3] Describes the various capabilities of su2 .It describes the various numerical (space and time integration schemes) methods available for solving physical problem in the domain of CFD in SU2 RANS model. Udbhav Sharma [4] His work mainly dealt with altering the flutter index graph by using the variation provided in the baseline geometry and structural parameters to redesign the airfoil to fit the required flutter index graph envelope. He too stated that linear models are incapable of predicting the flutter boundaries as the flow is highly unsteady in the transonic regime. Jameson [5] described in his report the use of multi-grid in solving fully implicit time stepping approach for wing and airfoil pitching. Edward et al. [6] used Hytran 2 code to study the accuracy and stability of various numerical techniques which may be used to transient time marching equations. His computational work stated out that the nonlinear curve of flutter is the direct effect of inclusion of angle of attack for NACA64A010 airfoil. The above works states that Euler and linear model are incapable in capturing the exact flow phenomenon occurring during Transonic flight regimes and the use of RANS model provides a accurate tool for predicting the flutter index which is utilized by using the RANS solver provided in su2 [3]. This paper describes the use of su2 code to predict the flutter boundary in the transonic flight regime using fully coupled RANS model with the structural modal equation for NACA64A010 section for Isogai swept wing. AEROELASTIC SOLVER Typical wing section module of su2 is the two dimensional analogy of the three-dimensional wing. It consists of a point mass-spring system with two degree of freedom analysis, pitch and plunge. AERODYNAMIC SOLVER Reynolds -Average Navier-stroke equations (Time marching implicit viscous flow) is implemented to determine the flow parameters in the computation domain. RANS equations : ππ’ + ∇πΉπ΄πΏπΈ πΆ − ∇. πΉ π£ − π = 0 ..............................(for t(time ) > 0)) ππ‘ where , π’ πΉπ΄πΏπΈ πΉπ£ π = πΆ {π, ππ£ , ππΈ } ...............................................................[Conservative variables] π(π£ − π’ β∩) = { π π£ (π£ − π’ β ∩) + πΌ ΜΏ π } π πΈ (π£ − π’ β∩)+ππ£ . πΜΏ = { } π. ΜΏ π£ + π ∗ π‘ππ‘ ππ ∇π = ........................................[ Convective Flux] ..............................................[Viscous Flux] ππ { πππ£β }.....................................................................................[Source Terms] πππΈ For a given problem the RANS model is used with no turbulence kind model resulting in the use of laminar Navier-strokes equation. Following steps are performed by the solver to set the flow for internal calculations for viscous flow conditions ο· The solver uses the free stream temperature and the gas constant to determine the speed of sound ο· Velocity vectors are determined from the above computed speed of sound and the free stream Mach number and angle of attack (Deg.) provided by the user . ο· ο· ο· computes the free stream viscosity from the viscosity model specified or it takes the default values hardcoded in the solver free stream density is computed using the definition of Reynolds number and the other parameters computed from above steps The free stream pressure is determined using the perfect gas law. Spatial integration is performed using the finite volume method while the time integration in unsteady flow is done using the several implicit and explicit schemes. In this particular analysis gradients of the flow variables are calculated using the weighted least square method at all grid nodes and then averaged and used to represent the gradient value at the cell faces. An Euler implicit scheme is implemented for time integration and the linear solver used to solve the equation of time integration is Bi-conjugate Gradient stabilized method. STRUCTURAL DYNAMICS SOLVER Two set of equations are as stated below π βΜ + ππΌ πΌΜ + πΎβ β = −πΏ ππΌ βΜ + πΌπΌ πΌΜ + πΎπΌ πΌ = πππ Where L and Mae are the lift and the moment about elastic axis, Kh and Kα are the bending and the torsion stiffness. The time is non-dimensionalized by multiplying it by the pitching frequency which can be used to rewrite the equation in the standard governing form as [π]{πΜ } + (πΎ){π} = {πΉ} Where, 1 [π] = [ π₯πΌ π π₯πΌ [π β ] [πΎ] ] , = [ πΌ ππΌ 2 0 2 0 ππΌ ] 2 are the non-dimensional mass and stiffness matrices and, β −πΆπ {πΉ} = [ ] , {π} = { πΌπ } π πΆπ are the load and displacement vectors. ππ 2 The problem is solved in su2 tool as follows, ο· For time step n, the flow parameter in the computational domain is solved using the implicit RANS solver module to obtain the value of Cl and Cm. ο· The corresponding values of the flow coefficient contributes to the load vectors in the standard governing equation on the R.H.S ο· The structural modal equation is solved to obtain the values of pitch and plunge ο· ο· The values of pitch and plunge are used by the grid movement code to move the grid to a new position and the above procedure is repeated again for non-dimensional time n+1. As the solver exit at Time = total time interval provided by the user the response of the system can be observed in the post processing tool for neutral oscillation which indicates the flutter speed at the corresponding Mach number. GEOMETRY AND COMPUTIONAL DOMAIN NACA64010 airfoil is used as the geometry for typical wing section simulation with chord length c=1m. C-grid is generated around the airfoil using the point-wise software. The domain is extended to 15*c from the trailing of the airfoil in the downstream direction and 15*c in the direction normal to the axis of the airfoil. The grid generated consists of 354928 structured Quadrilaterals. The spacing of 0.0001 is provided towards the airfoil from the top and bottom of the domain whereas a spacing of 0.001 is provided towards the trailing edge from the downstream domain edge. The computational domain along with the generated grid is given below figures 1-2. Fig. 1: Computational grid Fig. 2: Near airfoil grid BOUNDARY CONDITIONS The computational grid as shown in figures 1-2, totally, two types of boundary conditions applied in su2 for flow initialization in flow domain in time marching scheme. The outermost region was given the far field boundary condition with the values of free stream pressure of 101325 N/m2 and free stream Mach number for different transonic flow regimes and the surface of the airfoil being provided the condition of wall (no slip). RESULTS AND DISCUSSION The flutter boundary of the Naca64A010 airfoil is predicted using the "Typical Wing Section" module of the su2 software using time marching Euler implicit scheme with RANS model coupled to the structural modal equation with following structural parameters hard coded in the modal equations. xα = 1.8, r2α= 3.48, a = −2.0, ωh = 100rad/sec, ωα = 100 rad/sec, μ = 60. Dual time step of 2nd order is used for unsteady simulation with total physical time of 1.4 sec with number of internal iteration set to 20. As the pivot point lies ahead of the elastic axis, the x coordinate for the moment origin is provided as -0.5m. The flow is maintained at Reynolds no of 12.56e06 and the grid movement kind chosen for grid deformation is "AEROELASTIC", which means it uses the spring analogy to deform the grids as per the body moves. The simulation is run at various speeds keeping the Mach number constant in order to determine the response of the airfoil with respect to the forces acting upon it. Fig. 3: Pitch v/s Time plot at Mach no-0.75 At low speeds index (Vf), the disturbances caused due to interaction between the aerodynamic and elastic forces setup on and in the body tends to dampen out quickly which leads to the conclusion that the elastic forces are high enough to dampen the effect of disturbances resulting in damped oscillation. Fig. 4: Pitch v/s Time plot at Mach no-0.8 Fig. 5: Pitch v/s Time plot at Mach no-0.85 As the Vf reaches a particular value it exhibits a self sustaining neutral oscillation where the inertial force, aerodynamic force and the elastic force are in exact equilibrium and denotes the threshold value of flutter speed index denoted by critical flutter index. Above this critical flutter index the elastic forces are not sufficient to dampen the oscillation caused and the time history plot obtained is a diverging. The different responses are shown in figure 3, figure 4 and figure 5 respectively for particular operating Mach number. Fig. 6: Pitch v/s plunge plot at Mach no=0.85 (neutral response) Fig. 7: Pitch v/s plunge plot at Mach no=0.85 (Diverging and converging response) The pitch v/s plunge plots for three response case at Mach number 0.85 are provided in figures 6 and 7. The figure 7 shows that the area under the hysteresis curve increases as the response of the system increases from converging to diverging. The diverging hysteresis loop encloses the converging case as the pitch and plunge variation increases along the response. The present work on flutter simulation predict the transonic dip occurring at Mach number 0.85 at the flutter index value of Vf = 0.5388. The result is compared with the available flutter index v/s Mach number curves as in figure 8. Alonso and Jameson's Euler code closely resembles the flutter graph predicted by su2, but deviations in the critical flutter speed are observed at Mach number 0.8 and 0.85 due to effect of viscosity. The critical flutter speed predicted by su2 is above the values of those predicted by Alonso and Jameson code due to consideration of viscous effects in the present calculations at Mach number 0.8 and 0.85 respectively. The su2 code also simulates the effect of angle of attack on flutter index up to Mach no 0.9 when plotted against the Edward [6] simulated data. In general the su2 typical wing simulation takes into account the effect of varying angle of attack and viscous effect on flutter boundary estimation of NACA64A010 airfoil. The multiple flutter index obtained through non viscous calculation disappears by considering RANS viscous model providing distinct critical flutter index values at each Mach number as in figure 8. Fig. 8: Flutter index v/s Mach no. plot (transonic dip curve) CONCLUSION Su2 solver predicts the transonic dip at Mach number 0.85 with the critical flutter index at 0.5388. The flow is accompanied by the presence of moving shock at the top and bottom surface which interacts with the boundary layer and greatly alters the pressure distribution on airfoil surface, causing changes in the local values of pressure coefficient resulting in changing lift and moment about z-axis which in turn results in varying and time dependent highly non linear aerodynamic forces acting on the airfoil contour and affects the structural dynamics solutions. The data obtained from su2 seems to be in good agreement with the earlier simulated reference data. The simulation confirms the role of moving shock in making the flutter phenomenon highly non- linear. REFERENCES [1] Isogai, K., “On the Transonic-Dip Mechanism of Flutter of a Sweptback Wing,” AIAA Journal, Vol. 17, No. 7, pp. 793-795, July 1979. [2] Frane MajiΔ, Ralph Voss2, Zdravko Virag, “boundary layer method for unsteady transonic flow", Journal of Mechanical Engineering 58(2012)7-8, 470-481, 2012-04-03. [3] Stanford University Unstructured (Su2): Open-Source Analysis And Design Technology For Turbulent Flows, AiAA SciTech, AIAA 2014 -0243. [4] Udbhav Sharma, "Effects of Airfoil Geometry and Mechanical Characteristics on the Onset of Flutter", School of Aerospace Engineering Georgia Institute of Technology December 10, 2004. [5] Jameson, A., “Time Dependent Calculations Using Multigrid, with Applications to Unsteady Flows Past Airfoils and Wings,” AIAA Paper 91-1596, June 1991. [6] Edwards, J. W., et al., “Time-Marching Transonic Flutter Solutions Including Angle of Attack Effects,” Journal of Aircraft, Vol. 20, No. 11, November 1983, pp. 899-906. [7] Kousen, K. A., “Non-Linear Phenomena in Computational Transonic Aeroelasticity,” Ph.D. Dissertation, Dept. of Mechanical and Aerospace Engineering, Princeton University, January 1989.