A novel 3D variational aeroelastic framework for flexible multibody dynamics: Application to bat-like flapping dynamics

Guojun Li; Yun Zhi Law; Rajeev K. Jaiman

doi:10.1016/j.compfluid.2018.11.013

A novel 3D variational aeroelastic framework for flexible multibody dynamics: Application to bat-like flapping dynamics

Li, Guojun;Law, Yun Zhi;Jaiman, Rajeev K. 2018-07-12 00:00:00 We present a novel three-dimensional (3D) variational aeroelastic framework for apping wing with a exible multibody system subjected to an external incom- pressible turbulent ow. The proposed aeroelastic framework consists of a three- dimensional uid solver with a hybrid RANS/LES model based on the delayed detached eddy simulation (DDES) treatment and a nonlinear monolithic elastic structural solver for the exible multibody system with constraints. Radial basis function (RBF) is applied in this framework to transfer the aerodynamic forces and structural displacements across the discrete non-matching interface meshes while satisfying a global energy conservation. For the consistency of the interface data transfer process, the mesh motion of the uid domain with large elastic defor- mation due to high-amplitude apping motion is also performed via the standard radial basis functions. The uid equations are discretized using a stabilized Petrov- Galerkin method in space and the generalized- approach is employed to integrate the solution in time. The exible multibody system is solved by using geometrically exact co-rotational nite element method and an energy decaying scheme is used to achieve numerical stability of the multibody solver with constraints. A nonlinear iterative force correction (NIFC) scheme is applied in a staggered partitioned itera- tive manner to maintain the numerical stability of aeroelastic coupling with strong Preprint submitted to Elsevier Science 13 July 2018 arXiv:1807.04411v1 [physics.flu-dyn] 12 Jul 2018 added mass eect. An isotropic aluminum wing with apping motion is simulated via the proposed aeroelastic framework and the accuracy of the coupled solution is validated with the available experimental data. We next study the robustness and reliability of the 3D exible multibody aeroelastic framework for an anisotropic apping wing ight involving battens and membranes with composite material and compare against the experimental results. Finally, we demonstrate the aeroelastic framework for a bat-like wing and examine the eects of exibility on the apping wing dynamics. Key words: 3D exible multibody aeroelasticity, Radial basis functions, Flapping wing, Large elastic deformation, Partitioned iterative. 1 Introduction 1.1 Background on biologically-inspired apping ight Biologically-inspired apping ight has intrigued human-kind over the past several centuries. Recently, there is a growing trend in aeronautical engineer- ing applications to incorporate the understanding of apping ight dynamics of birds, insects and bats. A lot of remarkable research work in this area has been carried out during the past decades, including experiments [1,2], compu- tational uid dynamics (CFD) simulations [3,4] and real apping robots [5]. A number of recent reviews on several aspects of apping ight have been doc- umented [6,7,8,9,10]. Compared with a xed-wing ight vehicle, the apping ight of birds and bats involve active morphing and adaptive exible wing con guration, which may oer some unique bene ts with regard to eciency, noise and manoeuvrability [11]. Therefore, the understanding of apping wing Corresponding author Email address: mperkj@nus.edu.sg (R. K. Jaiman). 2 mechanisms can be useful to incorporate in the designs of micro air vehi- cles (MAVs) and unmanned aerial vehicles (UAVs). An unsteady turbulent aerodynamics interacting with nonlinear exible multibody dynamics poses a serious challenge in the study of such apping ight dynamics of animals. The emerging engineering requirements and the fundamental understanding of bio-inspired apping ight have motivated the present computational devel- opment focusing on the exible multibody solver interacting with a separated turbulent ow. During the past decades, numerous researchers [9,12,13] have explored the biologically-inspired apping ight for dierent ying species (e.g., insects, birds and bats). The aerodynamic characteristics around rigid and exible apping wings as well as the coupled nonlinear structural responses were in- vestigated in detail. However, several challenges are still remained for a more comprehensive and complete understanding of apping mechanism and the role of exibility during ight [9]. For example, previous researches are lim- ited to some speci c airfoils and simpli ed insect-like wings, which lack the generalization to a real apping ight. Biological structures are quite dierent for various species, which lead to their unique apping patterns with speci c mechanisms. Most of the insects have single or several pairs of wings consisted of the vein and membrane components with widely varying distributions of exibility [12]. Speci cally, a wing of the bird is made of bones and muscles to control its ight attitude and the surface is covered by dierent types of feathers. In particular, the unique skeletal anatomical structure of a bat has more degrees of freedom, compared with insects and birds and the membrane- like wing skin has high exibility with anisotropic material properties [14]. It is noticed that a physical structure of dierent wings can be considered as a generic exible multibody system with kinematic constraints. Recently, several work focus on the study of uid- exible multibody interaction, and a realistic apping wing including the vein and membrane components is incorporated 3 into the simulation. For instance, Gogulapati et al. [15] developed an approx- imate aerodynamic model coupled with shell elements to simulate a exible anisotropic apping wing. Farhat et al. [16] built an ALE-embedded compu- tational framework to deal with aeroelastic problems with large structural de- formation and demonstrated this framework for the anisotropic apping wing. The underlying structural models in these two work are discretized only with shell elements. In Cho et al. [17], the co-rotational beam elements are used for veins and the co-rotational shell elements are employed for the anisotropic apping wing simulation. All the numerical methods in the above mentioned research work were developed to analyze and to provide a physical understand- ing of the aeroelastic phenomena around insect-like or bird-mimicking wings. Only a handful of publications on 3D fully coupled aeroelastic analysis on a bat apping dynamics can be found in the literature, compared with a large number of work on the ight of birds and insects [18]. The main objectives of the present work are (i) to develop a fully-coupled aeroelastic framework for exible multibody analysis of apping ight dynamics, and (ii) to demonstrate the proposed aeroelastic framework for a bat-like wing. 1.2 Aspects of computational modeling Generally, a typical numerical simulation of the aeroelastic problem involves the coupling of the governing equations of uid and structural dynamical sys- tems in two dierent domains. Currently, two main aeroelastic schemes are considered to couple the uid equations and the structural equations namely, monolithic [19,20] and partitioned [21,22]. A monolithic approach can achieve good numerical stability for the aeroelastic problem with strong added mass eect, which assembles the uid and structural equations into a single block then solves the coupled equations in a uni ed manner. However, it is dicult to take a full advantage of the existing stable and advanced uid and struc- 4 tural solvers, which restricts the scalability and exibility of an aeroelastic framework [20,23,24]. Considering the drawbacks of a monolithic scheme, a partitioned approach is developed to employ the existing suitable uid and structural solvers to solve the complex and generic aeroelastic problem. The uid and structural equations are solved in a sequential manner, and the trac- tion and velocity continuities are satis ed along the uid-structure interface to achieve numerical stability and accuracy [21,25,26]. However, the added mass eect associated with acceleration of the apping wing during ight and geometry of the wing with thin structures may result to numerical instability in a partitioned approach [27,28,29]. For the purpose of generality of exi- ble multibody aeroelastic analysis, we adopt a partitioned approach in the present study. In what follows, some backgrounds and relevant literature are associated with the partitioned aeroelastic framework for exible multibody simulations. In a typical partitioned-based aeroelastic scheme, the surface boundary data must be transferred along the interface between the uid and structural do- mains to satisfy a Dirichlet-type interface conditon (displacements or velocity) and a Neumann-type ( uid momentum ux or traction) interface condition. A proper care during interpolation and projection process is required to transfer the physical data accurately across non-matching meshes between the parti- tioned uid and structural domains [26,30,31]. Global and local energy conser- vation should be satis ed during the aeroelastic coupling while maintaining the accuracy of data transfer along the interface via Dirichlet-Neumann coupling. Recently, a high accuracy interface projection scheme with the global and local energy conservation is successfully applied to 3D uid-structure problems [31]. In addition to the interface projection problem, a large structural deformation due to high-amplitude apping motion poses a challenge for mesh motion to maintain good mesh quality in a transient aeroelastic computation. In the recent years, the radial basis function (RBF) method has been demonstrated 5 as a simpli ed approach to interpolate scattered data [32,33] while satisfying the property of the conservation of global energy transfer. For several aeroe- lastic problems, the RBF-based interpolation has been employed for the data transfer across non-matching meshes [34,35] and the mesh motion with large deformation [36,37]. The marked advantage of the RBF method is that the connectivity of the nodes is not required, which provides an eective and con- venient way to implement RBF in any existing framework. For the ease of implementation during the coupling of uid domain with exible multibody surfaces, we consider the radial basis functions for the interface coupling and the mesh motion in the present study. In the context of the present aeroelastic work, the structure of a exible wing consists of various components with dierent material properties and the rel- ative inertia, which can be considered as a nonlinear elastic multibody sys- tem. The elastic behavior of such a system is inherently nonlinear, hence the elastic displacements and rotations cannot be assumed to be small [38]. Geo- metrically exact elements are able to deal with arbitrarily large displacement and rotation components in a multibody system exactly and those compo- nents are de ned in a common inertial frame. Energy preserving (EP) time integrators are generally employed to ensure the discrete conservation of the total mechanical energy while achieving nonlinear unconditional stability for multibody systems. However, the undesired high frequency oscillations lead to numerical divergence for the solution of the nonlinear equations of motion when solving some complex problems with large number of degrees of freedom [39]. The nonlinearities of the system induce the energy to transfer from low to the high frequency modes. The energy decaying (ED) scheme [40,41,42] is designed based on energy preserving (EP) scheme to deal with the high fre- quencies in a exible multibody system, which can achieve nonlinear uncon- ditional stability. The multibody interaction is solved by a time discontinuous Galerkin scheme based on the ED approach. The constraints in the system are 6 typically enforced by the Lagrange multiplier technique. Such a exible multi- body formulation has been successfully applied for the interaction of turbulent ow with a oater-mooring system comprising a rigid- oater body and a long exible riser and mooring lines [43]. We adopt a similar exible multibody formulation for the present aeroelastic framework. In nature, the ight condition of ying animals varies from low Reynolds numbers to high Reynolds numbers, depending on the real size, ying speed and ight environment of these animals. Such a highly variable ight condition involves rich aerodynamic physics during ight, including leading edge vortex (LEV) generation, massively separated ow, laminar-turbulent transition and trailing edge vortex (TEV) shedding [12]. Therefore, a practical turbulent model is desired to capture the complex ow dynamics around a apping wing. Compared to direct numerical simulation (DNS) and large eddy simulation (LES) methods, the delayed detached eddy simulation (DDES) method has signi cant advantages to save the computational resources while keeping the reasonable accuracy for the separated turbulent ow, as demonstrated by [44] for large-scale uid-structure simulations. The DDES model oers an eective and reasonable way to simulate the vortex structures played during a apping ight, hence we employ this model in our computational framework. 1.3 Contributions and organization In this study, a novel 3D variational aeroelastic framework is developed to simulate the apping ight with exible multibody system (e.g., insect-like and bat-like wing) in a turbulent ow. While the Petrov-Galerkin nite ele- ment method is used to solve Navier-Stokes equations for the external uid ow, geometrically nonlinear co-rotational nite element method is applied to the exible multibody system in the structural domain. The DDES method is employed to simulate the turbulent separated ow during apping motion. 7 The ALE uid-turbulent solver and the exible multibody solver are cou- pled via a partitioned iterative scheme combined with the nonlinear iterative force correction (NIFC) approach, which achieve numerical stabilization for the coupled aeroelastic framework. The RBF-based data interpolation ap- proach is implemented in our framework to transfer the aerodynamic forces and the structural displacements across the uid-structure interface while sat- isfying the property of conservation of energy transfer. The interface force is corrected at the end of each uid sub-iteration by means of the NIFC method. The RBF approach with a compact support is utilized to handle the mesh motion with a large deformation condition and keep the initial mesh quality in simulation. For the initial validation, an isotropic aluminum wing and an anisotropic wing with composite material in apping ight condition are sim- ulated by using the proposed novel 3D variational aeroelastic framework with exible multibody dynamics. Results are compared against results obtained from experiments and literature. Finally, we demonstrate the proposed aeroe- lastic framework to simulate a bat-like exible wing with supported skeletons and covered membranes. In the present paper, we address two important challenges associated with the variationally coupled aeroelastic framework with exible multibody sys- tem for the exible apping wing simulation: (i) coupling of an incompressible turbulent ow and a exible multibody system with geometrically nonlin- ear co-rotational nite elements, (ii) the interpolation of aerodynamic forces and structural displacements between uid surface elements and structural nite elements in a exible multibody environment. For the rst challenge, the proposed partitioned iterative scheme referred above is used to achieve the coupling between the uid solver and the multibody structural solver in a robust and generic manner. It is worth noting that two main problems should be considered when dealing with the second challenge. Firstly, the information on the multiple surfaces belonging to dierent components of an entire struc- 8 tural model need to be collected then exchanged with the information from the uid domain in an energy conservation manner. While the aerodynamic tractions at the nodes of each uid element are interpolated to the target multiple structural elements by using the RBF method, the collected nodal structural displacements and the velocity vectors of each structural element are interpolated to the corresponding uid mesh nodes. To address the numer- ical instability caused by the added mass eect, the strongly-coupled NIFC implementation [45] is employed. The force equilibrium and the velocity con- tinuity condition conservation on the interface are satis ed by evaluating the approximate interface force corrections in the nonlinear sequence transforma- tion. The generalization of Aitken's extrapolation technique is applied to the iterative sequence coupling, which achieves a stable and convergent force updating process. The remainder of this manuscript is organized as follows. In Section 2, the variational formulations for the uid and the exible multibody system with constraints are reviewed. The detailed procedures for aeroelastic interface in- terpolation and the NIFC-based uid- exible multibody coupling by using RBF method are presented. The implementation of RBF method in mesh motion with large deformation condition is then introduced. An isotropic alu- minum wing and an anisotropic wing with composite material in apping ight condition are simulated with the proposed aeroelastic framework and compared with the available simulation and experimental data for validation purpose in Section 3. Section 4 presents an application on a bat-like exible wing with skeletons and membrane and explores the eects of exibility and aerodynamic load. The key conclusions of the present work are summarized in Section 5. 9 2 Partitioned aeroelastic framework for exible multibody system The governing equations and the formulations for the present 3D variational aeroelastic framework are similar to those of [43]. For the sake of complete- ness, we review the variational formulations for the moving uid and exible multibody solvers, whereas the Navier-Stokes equations are solved using the Petrov-Galerkin nite element method in the ALE coordinates and the exible multibody system is solved via nonlinear co-rotational nite element method. 2.1 Petrov-Galerkin nite element for turbulent ow Consistent with the work of [43], the Navier-Stokes equations are discretized using a stabilized Petrov-Galerkin formulation. The gerneralized- method is implemented to integrate the ALE ow solution in time domain, which can achieve unconditionally stable as well as second-order accuracy for linear problem. Furthermore, user-controlled high frequency damping desired for a coarser discretization in space and time is enabled by this scheme. The solution updates for the ow variables with the generalized- scheme can be written as f;n+1 f;n f;n f;n+1 f f u = u + t((1 )@ u + @ u ) (1) t t h h h h f;n+ f;n+1 f;n f f u = u + (1 )u (2) h h h f;n+ f;n+1 f;n m f f @ u = @ u + (1 )@ u (3) t t t h m h m h m;n+ m;n+1 m;n f f u = u + (1 )u (4) h h h f;n+1 m;n where u and u represent the uid and mesh velocities de ned for h h f f f each spatial uid point x 2 (t), respectively, whereas x and t are the f f f spatial and temporal coordinates. Here, , and represent the standard 10 integration parameters given as 1 1 3 1 f f 1 f f f = ; = ; = + (5) m m f f 2 2 1 + 1 + 1 1 The uid spatial domain can be discretized into n number of non-overlapping el f e f;h el nite elements and . While S represents the space of the trial e=1 f;h solutions, V denotes the space of test function. The variational formulation of the uid equations within the Petrov-Galerkin framework can be written f;n+ f;n+1 f f;h f;h as: nd [u ; p ] 2 S such that 8[ ; q ] 2 V h h h f f f f;n+ f;n+ m;n+ f;n+ f (@ u + (u u )ru ) d h h h h h Z Z f;n+ f;n+ f f des + : r d + : r d h h h h e e f;n+ rq u d el f f f;n+ m;n+ f f f e + ( (u u )r +rq )R (u ; p)d m h m h h h e=1 el f (6) f f;n+ + r r u d h h e=1 el f f;n+ f e (R (u ; p)ru )d m m h h e=1 el f f e r : ( R (u ; p) R (u ; p))d m m m m e=1 Z Z f f f n+ f = b (t ) d + h d h h f f where is the density of the uid and b is the body force applied on the f;n+ f;n+ des uid, represents the turbulent stress term, is the Cauchy h h stress tensor for a Newtonian uid. Here, and q denote the test functions of the uid velocity u and pressure p. In Eq. (6), the Galerkin terms for the ow equations are shown in the rst, second and third lines. The Petrov- Galerkin stabilization term for the momentum equation is shown in the fourth line and the term for the continuity is in the fth line. The remaining terms 11 represent the approximation of the ne scale velocity on the element interiors. The stabilization parameters and are added to the fully discretized m c formulation and R (u ; p) denotes the residual of the momentum equation [46]. A hybrid RANS/LES model based on the DDES treatment is applied to model the ow turbulence for high Reynolds numbers. The DDES model behaves as RANS model in the near wall region and switches to LES-mode in the separated turbulent region. Further details can be referred to [47]. 2.2 Flexible multibody solver with constraints The motion equations for a exible structure are discretized using nite ele- ment method and can be written into a weak variational form using the virtual work principle Z Z Z n+1 2 s @ d s h s s s s + (E(d )) : r d dt = h h h s s @t i i (7) Z Z Z n+1 s s s s b d + t d dt h h s s s where denotes the multibody domain, and represent the test function i h for the structural displacements and the structural density, respectively. Here, s s d is the structural displacement, is de ned as the rst Piola-Kirchho s s s stress tensor, b denotes the body force on the multibody and E(d ) is the i h simpli ed Cauchy-Green Lagrangian strain tensor. The external force caused by ow on the interface is de ned as t . The kinematic joints, like revolution joint and sphere joint, are used to connect dierent components in the exible multibody system and can be considered as multibody constrains with the following expression: c(d ) = 0: (8) 12 Generally, the discretized motion equations with constraints for exible multi- body system can be written into a matrix form Z n+1 Z n+1 t t s s s s s s (M d (t) + C d (t) + K d (t))dt = (F (t))dt (9) n n t t s s s where M , C and K represents the mass, constrain and stiness matrices for the exible multibody system respectively. All the body forces and external forces caused by ow acting on the multibody system can be combined into a whole force matrix F . The forces from constraints will not produce any work for the multibody system at the discrete solution level, which are not considered in the force matrix. The structural variables are updated via an unconditionally stable energy decaying scheme temporally, which is obtained by using a linear time dis- continuous Galerkin approximation to Eq. (9). We brie y present the general discretized motion equations for exible multibody system with dierent geo- metrically nonlinear co-rotational nite element models, including beam, ca- ble, shell and membrane. The detailed description of geometrically exact shell model in multibody dynamics can be found in [48,49]. 2.3 Aeroelastic coupling and mesh motion via radial basis functions In this section, we brie y introduce the RBF method and its applications to the multibody aeroelastic coupling and the mesh motion interpolation in our ALE formulation. In the proposed partitioned iterative coupling scheme, the interface data is exchanged between the uid domain and the multibody structural domain. Across the non-matching meshes along the aeroelastic in- terface, the surface data need to be transferred in a conservative manner with reasonable accuracy. 13 2.3.1 Review of RBF interpolation process Assume that data are interpolated from a set of control points to a set of c c c c target points. Let x = (x ; y ; z ) be the coordinate of the i-th control point, i i i i t t t t x = (x ; y ; z ) be the coordinate of the j-th target point, N and N be the c t j j j j number of control points and target points. The global interpolation function de ned by a radial basis interpolation is given by i=N g(x) = + p(x) i i i=1 i=N = (kx x k) + p(x) (10) i=1 where g(x) is the global interpolation function and x is the coordinate of an arbitrary point in space and kk denotes the Euclidean norm. The coecient represents the weights related to the i-th basis function and p(x) is a linear polynomial to recover translation and rotation motion, where its expression is p(x) = + x + y + z. 0 1 2 3 As shown in [32,33], the interpolation relationship between the control vector c t G and the target vector G can be written in the matrix form as t I I 1 c I c G = A (M ) G = H G (11) where 2 3 2 3 0 0 6 7 6 7 6 7 6 7 6 7 6 7 6 0 7 6 0 7 6 7 6 7 6 7 6 7 6 7 6 7 0 0 6 7 6 7 6 7 6 7 6 7 6 7 t c G = 6 7 ; G = 6 7 (12) 0 0 6 7 6 7 6 7 6 7 6 7 6 7 t c g(x ) g(x ) 6 7 6 7 1 1 6 7 6 7 6 7 6 7 . . . . 6 7 6 7 . . 6 7 6 7 4 5 4 5 t c g(x ) g(x ) N N t c 14 I where H denotes the nal interpolation matrix between the control vector I I and target vector. A and M are the interpolation matrices given by 2 3 0 0 0 0 1 1 1 6 7 6 7 6 7 c c c 60 0 0 0 x x x 7 1 2 N 6 7 6 7 c c c 6 7 0 0 0 0 y y y 6 7 1 2 N 6 7 6 7 c c c M = 6 7 (13) 0 0 0 0 z z z 1 2 N 6 7 6 7 c;c c;c c;c 6 7 c c c 1 x y z 6 7 1 1 1 11 12 1N 6 7 6 7 . . . . . . . . . . . . . . . 6 7 . . . . . . . 6 7 4 5 c;c c;c c;c c c c 1 x y z N N N N 1 N 2 N N c c c c c c c 2 3 t;c t;c t;c t t t 1 x y z 6 1 1 1 11 12 1N 7 6 7 . . . . . . . 6 . 7 . . . . . . . . A = 6 . 7 (14) . . . . . . . 6 7 4 5 t;c t;c t;c t t t 1 x y z N N N N 1 N 2 N N t t t t t t c c;c t;c c c t c with = ( x x ) and = ( x x ), respectively. A compactly ij i j ij i j supported Wendland's C function [32,34] has been proved as an eective ba- sis function for the interpolation with improved accuracy, which is considered in the framework. The basis function with a compact support is de ned as (kxk) = (1kxk =r) (4kxk =r + 1), where r denotes the support radius. 2.3.2 Multibody aeroelastic coupling The aeroelastic coupling involves two data transfers across the interface: (i) forces from the uid domain to the structural domain and (ii) displacements from the structural domain to the uid domain. A schematic diagram of in- terface data interpolation via RBF method is depicted in Fig. 1. The energy conservation should be satis ed in the interpolation of forces and displace- ments along the aeroelastic interface. Generally, the work W done by forces in structural domain on the interface equals to the work W done by the aerodynamic forces according to the de nition of energy conservation, which 15 Rigid Beam Shell Fluid element Multibody Structural node structure Structural Fluid node displacements Data transfer via RBF Fluid forces Fluid Fig. 1. Schematic of interface data transfer via RBF method across non-matching meshes between uid domain and multibody structure system involving rigid link, exible beam and shell members. is given as Z Z s s s s s f f f f f W (d ) = ( n ) d d = ( n ) d d = W (d ) (15) fs fs s f where d and d are the displacements along interface in the structural and uid domains, respectively. For the aeroelastic interface coupling process, the interpolation equation of displacements along interface can be written as f;I I s;I D = H D (16) ss f;I s;I where D and D are the displacement vectors along interface for the uid domain and the structural domain, respectively. Both the vectors have the same vector form as shown in Eq. (12). The matrix H interpolates the dis- ss placements from the structural surface nodes to the uid surface nodes. As the balance of energy transfer at the interface must be achieved during the interpolation process, according to Eq. (15) and Eq. (16), the interpolation 16 between the uid forces and the structural forces can be written as s;I I T f;I F = (H ) F (17) ss f;I s;I where F and F are the force vectors along the interface for the uid do- main and the structural domain, respectively. We numerically evaluate the mo- mentum ux through the boundary faces and provide the interpolated forces to the structural solver. Such an interpolation process across aeroelastic inter- face is performed at each time step in the partitioned iterative scheme with the conservation of energy transfer at the uid-structure interface. 2.3.3 Mesh motion interpolation The aeroelastic coupling involve a deformation of the uid mesh in response to a movement of the structure. The interpolation of the mesh motion using radial basis function can handle mesh motion with a relatively moderate deformation without distorting the mesh quality. Moreover, the mapping matrix used for the interface displacement interpolation can be easily extended to a three- dimensional uid domain. In our framework, the mesh motion of the uid nodes in the domain is interpolated from the uid nodes on the interface. The relationship between the displacement vector of uid nodes in the domain D f;I and the displacement vector of the uid nodes on the interface D is given as f I f;I D = H D (18) v vs s where H denotes the interpolation matrix for the uid space nodes. Note vs that, H includes the polynomial term p(x) shown in Eq. (10), where its vs value increases linearly with x. While these terms can preserve good mesh quality in the near wall region, an undesired translation and rotation motion will be interpolated to the uid nodes at the boundary of the computational 17 domain. Hence, a smooth cut-o function [33] is implemented to adjust the deformation far away from the interface. 2.4 Nonlinear interface force correction scheme Nonlinear interface force correction scheme [45] can provide a numerical stabil- ity for the partitioned aeroelastic coupling when signi cant added mass eect is encountered in an aeroelastic simulation. This scheme has been extended to the exible multibody aeroelastic coupling simulation to achieve numeri- cal stability and the iterative force correction procedure is brie y summarized herein. The formulation for a coupled linear system between the uid domain and the structure domain which is discretized by the nite element method can be written into the matrix form AU = R, where U denotes the vector of the unknowns for the coupled system and R represents the right-hand side. The abstract matrix form can be written as 8 9 8 9 2 3 > > > > > > > > A 0 0 A d R > > > > 11 14 1 6 7 > > > > > > > > 6 7 > > > > > > > > 6 7 > > > > < = < = 6A A 0 0 7 d R 21 22 2 6 7 = (19) 6 7 > > > > 6 7 > > > > 0 A A 0 >q > >R > 32 33 3 6 7 > > > > > > > > 4 5 > > > > > > > > > > > > : I; : ; 0 0 A A f R 43 44 4 where A denotes the mass and stiness matrices of exible multibody system and A is the load vector on the solid surface. A represents the 14 21 transformation matrix which maps the structural displacements to aeroelastic interface and A denotes the force calculation and transform to the interface. A and A are the identity matrices. A gives the relationship between the 22 44 32 displacement on interface and the ALE mapping in the uid domain. A is the coupled uid velocity and pressure linear system. d denotes the structural displacement for the exible multibody system and d represents the displace- ment along the coupling interface. q is the unknown variables in uid domain 18 I and f is the force on interface. The right-hand side vector R shows the resid- ual of this linear system for the dierent equations. In this matrix form, the rst equation is the exible multibody system and the third equation is re- lated to the uid and turbulence equations. The second and fourth equations correspond to the displacement and traction transformation between the uid domain and structure domain, respectively. The o-diagonal term A can be eliminated through the static condensation and Eq. (19) can be rewritten as 2 3 8 9 8 9 > s> > > > > > > A 0 0 0 d R 11 > > > 1> 6 7 > > > > > > > > 6 7 > > > > > > > > 6 7 > > > > < = < = A A 0 0 d R 6 7 21 22 2 6 7 = (20) 6 7 > > > > 6 7 > > > > 0 A A 0 >q > >R > 32 33 3 6 7 > > > > > > > > > > > > 4 5 > > > > > > > > : I; : ; ~ ~ 0 0 0 A f R 44 4 1 1 1 1 ~ ~ where A = A A A A A A A A and R = R A A (R 44 44 43 33 32 22 21 11 14 4 4 43 33 3 1 1 A A (R A A R )). An iterative scheme is used to correct the forces 32 2 21 1 22 11 between the uid domain and structure domain with a feedback process. The formulation of nonlinear iterative force correction is given as I I 1 ~ ~ f = f + A R (21) 4(k) (k+1) (k) 44 where k denotes the nonlinear sub-iteration with a single time step. The cur- rent nonlinear iterative force correction process can be considered as a gen- eralization of Aitken's extrapolation while updating a dynamic stabilization parameter, which could transform a divergent xed-point iteration to a con- vergent and stable force correction [45]. Finally, a numerical stability can be achieved with the aid of the NIFC scheme for the partitioned uid- exible multibody system coupling. 19 2.5 Fluid- exible multibody coupling procedure The partitioned iterative coupling procedure of the uid solver with the ex- ible multibody structural solver is brie y summarized in this section. The schematic of aeroelastic coupling procedure based on a predictor-corrector al- gorithm is shown in Fig. 2. The predictor displacement is calculated from the exible multibody equations and the ALE uid equations can be solved to provide the forces on the uid-structure interface as a corrector step. The structural displacement caused by the aerodynamic forces at time t is de- s s n ned as d (x ; t ). Firstly, the structural displacements of each component in the exible multibody system is solved via the time discontinuous Galerkin approximation under the loads from the uid domain at time t . Then the structural displacements obtained from the previous step are transferred to the uid solver and it is satis ed the ALE compatibility and the velocity con- fs tinuity on the interface . To satisfy the consistency of the non-matching uid and structural mesh con gurations, the mesh displacements are set equal to the structural displacements along the interface as m;n+1 s fs d = d on (22) m;n+1 n+1 where d represents the mesh displacement at time t . Meanwhile, the fs velocity continuity on interface is satis ed when the uid velocity equals to the mesh velocity. m;n+1 m;n f f d d f;n+ m;n+ fs u = u = on (23) In the third step, the Navier-Stokes equations under the ALE framework with the turbulence model equations are solved and compute the uid forces. Fi- nally, the obtained uid forces are corrected based on NIFC scheme to achieve the numerical stability then transferred to the exible multibody solver. One 20 aeroelastic sub-iteration is nished and the sub-iteration mentioned above will be executed continuously until the convergence criterion has achieved. Sub- sequently, all the variables in this aeroelastic solver are updated for the next n+1 time step t . The Generalized Minimal RESidual (GMRES) algorithm is applied in the cur- rent framework to compute the velocity, pressure and mesh displacement in uid equations discretized with nite element method. This algorithm relies on the Krylov subspace iteration and the modi ed Gram-Schmidt orthog- onalization. In order to minimize the linearization error at each time step, the Newton-Raphson scheme is considered for the aeroelastic framework. Fur- thermore, the discretized algebraic equations based on the exible multibody system with the co-rotational framework are solved via a classical skyline solver. 3 Mesh convergence and validation Before we demonstrate our multibody aeroelastic framework for a bat-like wing, a mesh convergence study is rst conducted to ensure a sucient mesh resolution employed on both the uid and structure domains. We validate our multibody aeroelastic solver against the experiment work of [50] by simulating a exible apping wing with an anisotropic material, where the structural responses are compared to the experimental data, and the vortex patterns are analyzed. 21 f f n+1 u ¯ (x , t ), f f n+1 s n+1 p ¯ (x , t ), d (x , t ) f f n+1 ν ˜ (x , t ) k = nIter k = 2 I I I f = f + δf k+1 k (4) (3) (2) (1) k = 1 f f n u ¯ (x , t ), s n f f n p ¯ (x , t ), d (x , t ) f f n ν ˜ (x , t ) f fs Ω Γ Fig. 2. A schematic of predictor-corrector procedure for ALE uid solver and exible multibody solver coupling via nonlinear iterative force correction. (1) Solve exible multibody system with constraints, (2) Transfer predicted structural displacement, (3) Solve ALE uid and turbulence equations, (4) Correct forces via NIFC lter. nIter represents the maximum number of nonlinear iterations to achieve a desired n n+1 convergence tolerance in a time step at t 2 [t ; t ]. 3.1 Mesh convergence The mesh convergence study is conducted by simulating a wing with isotropic material using three dierent mesh resolutions on the uid and structure do- mains. The wing simulated is adopted from an experimental study [50], and its properties are summarized in Table 1. It is an isotropic aluminum wing with Zimmerman shape, which is designed to investigate the aerodynamic charac- teristics under apping ight condition. A schematic diagram of the geometry of the wing and its surrounding aeroelastic computational domain are shown in f s u ¯ = u ALE ﬂuid & turbulence Flexible multibody wing 75mm Leading edge Fixed square 25mm Wing tip Trailing edge (a) wing 60C in out 80C 60C (b) Fig. 3. Flow past a apping wing: (a) geometry of isotropic wing, (b) schematic of computational setup. Fig. 3. The root chord length C is 25 mm and the span length is 75 mm, which results in an aspect ratio of 7.65. The apping condition is hovering motion with 10 Hz apping frequency and 21 apping amplitude. The reference ow velocity is the velocity measured at the wing tip and the freestream velocity is given as zero. The distances of the inlet ( ) and outlet ( ) boundaries in out from the leading root of the wing are both 30C . The distance between the side walls ( ) on top and bottom is 60C and the distance increases to 80C for slip the side walls on both sides. The components of the ow velocity are de ned f f f f as u = (u ; v ; w ). The freestream velocity along the X -axis at the inlet 23 f boundary is given as u = U . Slip boundary condition is applied on the in @u top and bottom boundaries ( ), where = 0 and w = 0. Slip bound- slip @z @u ary condition is applied on both side boundaries ( ), where = 0 and slip @y v = 0. A traction-free boundary condition is de ned at the outlet boundary , where = = = 0. The boundary condition on apping wing out xx yx zx surfaces is no-slip boundary condition. The three-dimensional uid computational domain is discretized by unstruc- tured eight-node brick nite element mesh and the structural computational domain is discretized by structured four-node rectangular nite element mesh. A summary of three dierent mesh resolutions is shown in Table 2. A boundary layer with y 0:5 in the wall-normal direction is maintained, with a stretch- ing ratio, y =y of 1.2. The discretization along chord direction, span-wise i+1 i direction and outside the boundary layer is varied. The non-dimensional time step size, tU =C is chosen as 0.01. Details of mesh statistics in space and ref on surface for the uid domain with M2 are shown in Fig. 4a and 4b, respec- tively. A nite element mesh on structural surface with M2 is shown in Fig. 4c. For the purpose of mesh convergence investigation, the amplitude of lift coef- cients for three dierent meshes are calculated and compared with literature data, which are shown in Table 3. The results indicate that the gap between M1 and M3 is within 1% and it reduces to 0.15% for M2 and M3. It is con- cluded that M2 has achieved mesh convergence and it can be used as the reference case to compare with experiment data. The comparison of lift coe- cient histories in one cycle is displayed in Fig. 5a and the simulated result for M2 shows a good comparison with the result of [51]. The normalized displace- ment at wing tip is compared in Fig. 5b with that of the experiment [50]. It can be seen that the normalized displacement at wing tip in one non-dimensional cycle has a good match with the experimental measurements. 24 (a) Fluid space mesh (b) Fluid surface mesh (c) Structural surface mesh Fig. 4. Schematic of mesh characteristics: (a) in space for apping wing in the uid domain (M2), (b) on surface for apping wing in the uid domain (M2), (c) on surface for apping wing in the structural domain (M2). The velocity magnitude contour, X -vorticity contour for a slice at the quarter position along chord-wise and Y -vorticity contour for a slice at the middle position along span-wise at both t=T = 0:3 and t=T = 0:48 are shown in Fig. 6, respectively. It can be seen from the Y -vorticity contours that a pair of main vortices is generated at the leading edge and trailing edge during the hovering motion. The normalized velocity magnitude distributions along vertical direction of the apping wing at wing tip and mid-span for both time instants t=T = 0:3 and t=T = 0:48 are extracted from the ow eld and compared with those obtained from [50] and [51], which are shown in Fig. 7. The experiment result is displayed with 95% errorbars of the instantaneous values. The normalized velocity magnitude distributions simulated via our aeroelastic framework show similar trends with those in literature and the experiment. 25 Table 1 Aeroelastic parameters for an isotropic aluminum wing under hovering motion Parameters Value Semi-span at quarter chord 0.075 m Chord length at wing root 0.025 m Structural thickness 0.0004 m Poissons ratio 0.3 Material density 2700 kg=m Youngs modulus of material 70.0 GPa Reference ow velocity (hovering) 1.0995 m/s Air density 1.209 kg=m Mean chord-based Reynolds number 2605 Flapping frequency 10 Hz Flapping amplitude 21 Aspect ratio 7.65 Table 2 Mesh statistics for an isotropic wing under hovering motion Mesh Fluid nodes Fluid elements Structural nodes Structural elements M1 509,082 492,630 114 90 M2 816,312 795,614 182 150 M3 1,311,120 1,284,226 506 450 The instantaneous turbulent wake elds of the complex three-dimensional ow around the apping wing at four various time instances with t=T = 0, 0.25, 0.5 and 0.75 are displayed in Fig. 8. The non-dimensional Q-criterion at isosur- face of 1 is used to depict the detailed and complex vortex structures around apping wing and it is colored by the normalized velocity magnitude during hovering motion. 26 Table 3 Mesh convergence study for the lift coecient C . The percentage dierences are computed based on M3 result. Results Amplitude of lift coecient C Present (M1) 6.65 (0.61%) Present (M2) 6.6 (-0.15%) Present (M3) 6.61 Aono et al. 6.24 1.5 Present (M2) Present (M2) Aono et al. (2010) Experiment 0.5 -0.5 -2 -1 -4 -1.5 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 t/T t/T (a) (b) Fig. 5. Comparisons of time traces of the isotropic apping wing: (a) lift coecient: (4) Present simulation, () Aono et al. [51], (b) normalized displacement at wing tip: (4) Present simulation, () Experiment [50]. 3.2 Validation of multibody aeroelastic framework A multibody exible wing adopted from the experiment [50] with composite material apping in a vacuum is rst simulated to validate the exible multi- body structural solver. The wing apping in air ow is considered for our validation purpose of the proposed aeroelastic solver. This wing is designed to mimic a real hummingbird wing based on Zimmerman planform and the membrane with Capran material is supported by several skeletons made of unidirectional carbon bers. The schematic of this anisotropic wing for a de- tailed geometric information is depicted in Fig. 9a. A rigid triangle made of δ /C z,tip (a) t=T = 0:3 (b) t=T = 0:48 (c) t=T = 0:3 (d) t=T = 0:48 (e) t=T = 0:3 (f ) t=T = 0:48 Fig. 6. Flow elds around the isotropic apping wing in uniform ow: velocity magnitude at (a) t=T = 0:3, (b) t=T = 0:48, X -vorticity at (c) t=T = 0:3, (d) t=T = 0:48, and Y -vorticity contours at (e) t=T = 0:3, (f ) t=T = 0:48. 28 4 2.5 Present (M2) Present (M2) Aono et al. (2010) Aono et al. (2010) 2 Experiment with 95% errorbars Experiment with 95% errorbars 1.5 0.5 -1 -0.5 -3 -2 -1 0 1 -3 -2 -1 0 1 Z/C Z/C (a) t=T = 0:3 at mid-span (b) t=T = 0:3 at wing tip 2.5 5 Present (M2) Present (M2) Aono et al. (2010) Aono et al. (2010) Experiment with 95% errorbars Experiment with 95% errorbars 1.5 3 1 2 0.5 1 0 0 -1 -0.5 -3 -2 -1 0 1 -3 -2 -1 0 1 Z/C Z/C (c) t=T = 0:48 at mid-span (d) t=T = 0:48 at wing tip Fig. 7. Flow past an isotropic appig wing: Comparison of instantaneous velocity magnitude for isotropic wing at various time instants: (a) t=T = 0:3 at mid-span, (b) t=T = 0:3 at wing tip, (c) t=T = 0:48 at mid-span, (d) t=T = 0:48 at wing tip. (4) Present simulation, () Aono et al. [51], () Experiment with 95% errorbars [50]. three layers of bidirectional carbon ber is used to mount the wing during the apping deformation and it is inserted at the corner of the wing root and the leading edge. A schematic of the topological layout is presented in Fig. 9c. This anisotropic wing is reinforced at the leading edge, the wing root and the surface of membrane with dierent layers of carbon ber strips of 0.8 mm width. Dierent wings with speci c layout schemes are termed LiBj, where L represents the leading edge, B stands for the batten and i, j denotes the kUk /U kUk /U ref ref kUk /U ref kUk /U ref (a) t=T = 0 (b) t=T = 0:25 (c) t=T = 0:5 (d) t=T = 0:75 kUk /U : 0 0.2 0.4 0.6 0.8 1 1.2 1.4 ref Fig. 8. Flow past an isotropic apping wing: Wake structures based on the instan- @u @u 1 j taneous iso-surfaces of Q(= ) value at (a) t=T = 0, (b) t=T = 0:25, (c) 2 @x @x j i + 2 t=T = 0:5, (d) t=T = 0:75. Iso-surfaces of non-dimensional Q Q(C=U ) = 1 ref are colored by the normalized velocity magnitude kUk =U . ref number of layers in the leading edge and the batten, respectively. The wing root is reinforced with 2 carbon ber layers for all wings. L2B1 and L3B1 models are considered to validate the high- delity exible multibody solver in our proposed aeroelastic framework. For L2B1 and L3B1 models, the leading edge is reinforced with two and three layers, respectively. 30 Each batten is reinforced with 1 layer for both models. Details of mate- rial properties of the composite skeleton and the wing membrane for the anisotropic wing with L2B1 and L3B1 models are summarized in Table 4. Some structural material parameters are adjusted to satisfy the structural mode frequencies and shapes obtained from experiments according to the cor- rection work in [52]. Geometrically exact co-rotational shell elements with anisotropic material are employed for skeletons and shell elements with ho- mogeneous isotropic material are used for the wing membrane. Dierent nite elements are connected with their adjacent elements via kinematic constraints. The structural model is discretized by 315 structured four-node rectangular nite elements and the detailed mesh characteristic is given in Fig. 9b. The rst six order natural frequencies are analyzed for L2B1 and L3B1 models with speci c material properties. The comparison with frequencies obtained from the experiment [50] and literature [52] shows that the present results have good agreement, which is given in Table 5. In the experiment, this anisotropic wing is given a prescribed rotation apping motion along wing root direction at the mounting rigid triangle in a vacuum. The apping amplitude is 35 and the apping frequency is set to 25 Hz. The time transient structural responses at wing tip is measured and compared with those obtained from experiment [50]. Here denotes the actual location in the vertical direction of wing z;tip tip at computational coordinate and represents the distance between w;tip deformed actual wing tip and the undeformed reference plane. The comparison of the initial reference plane, the undeformed reference plane and the deformed actual plane is depicted in Fig. 9d. Comparisons of time histories for the normalized location in the vertical di- rection and the normalized displacement at the wing tip for L2B1 and L3B1 models undergoing prescribed rotational motion in a vacuum are shown in Fig. 10, respectively. The results indicate that the high- delity exible multi- body solver can simulate the nonlinear structural responses of the anisotropic 31 Table 4 Material properties of the composite skeleton and wing membrane for the anisotropic wing (L2B1 and L3B1 models) [52] Component Material properties E =233 GPa E =23.7 GPa G =10.5 GPa G = G =1.7 GPa 23 31 Carbon ber prepreg (Properties of one v =0.05 layer) v = v =0.32 23 31 =1740 kg=m Thickness=0.1 mm E=2.76 GPa v =0.489 Capran membrane 12 (From experiments) 3 =1384 kg=m Thickness=0.015 mm wing with multiple components and shows reasonable agreement with the date obtained from the experiment. In our simulation results, the wing tip defor- mation is varied in dierent cycles, which is similar to the conclusion obtained in [52]. The structural responses are averaged over several cycles, according to the periodic assumption in experimental measurements. One possible reason causing discrepancies in the comparison is some uncertain factors connected with the actuation mechanism [50]. The instantaneous structural displacement contours of L2B1 model for four time instants in a whole apping cycle are given in Fig. 11. The anisotropic wing shows a relatively large elastic defor- mation and the wing twist due to the high exibility of the wing material. In the experiment, the anisotropic wing apping in air condition is also per- formed to investigate the aeroelastic eect on the apping motion. Therefore, an anisotropic wing with L3B1 model is simulated by using the proposed 32 Table 5 Comparison of the natural frequencies for anisotropic wing Natural frequencies (Hz) Model Result st nd rd th th th 1 2 3 4 5 6 Present 45.03 73.16 76.39 94.45 106.35 116.31 L2B1 Gogulapati A. 47.00 72.00 76.50 88.00 109.00 118.80 Experiment 42.00 84.00 126.00 Present 62.71 76.13 79.55 106.31 111.00 119.39 L3B1 Gogulapati A. 65.00 75.50 76.80 107.00 109.00 120.00 Experiment 59.00 104.00 138.00 Rigid triangle Leadning edge Batten 25mm Root Trailing edge 9.375mm 18.75mm 18.75mm 18.75mm 9.375mm (a) (b) Number of layers for leading edge layup L3B1 0.12g L2B1 0.10g L1B1 0.08g Number of layers for batten layup L3B2 0.13g L2B2 0.11g L1B2 0.09g (c) (d) Fig. 9. Problem set-up for anisotropic wing con guration: (a) geometry information, (b) nite element representation, (c) the topological layout [50], (d) initial reference plane, undeformed reference plane and deformed actual plane. 33 3 Present Present Experiment 2 Experiment -1 -1 -2 -3 -2 -50 -25 0 25 50 -50 -25 0 25 50 α ( ) α ( ) (a) L2B1 (b) L2B1 Present Present Experiment Experiment 0.5 -1 -0.5 -2 -1 -3 -50 -25 0 25 50 -50 -25 0 25 50 α ( ) α ( ) (c) L3B1 (d) L3B1 Fig. 10. Comparison of structural responses for anisotropic wing: (a) normalized location in vertical direction of L2B1 model at wing tip, (b) normalized displace- ment of L2B1 model at wing tip, (c) normalized location of L3B1 model in vertical direction at wing tip, (d) normalized displacement of L3B1 model at wing tip. (4) Present simulation, () Experiment [50]. aeroelastic framework for validation purpose. In the experiment, the material properties of the anisotropic wing are summarized in Table 4 and a hov- ering motion is employed with zero freestream velocity. Detailed simulation parameters for the aeroelastic simulation are given in Table 6. Considering the similar simulation condition as the isotropic apping wing, a similar aeroelas- tic computational domain is shown in Fig. 3b. Identical boundary conditions for the computational domain of anisotropic wing are applied as those of the δ /C δ /C z,tip z,tip δ /C w,tip δ /C w,tip (a) (b) (c) (d) δ : -0.03 -0.021 -0.012 -0.003 0.006 0.015 0.024 Fig. 11. Structural displacement contours of L2B1 model at various time instants: (a) t=T = 0:19 with apping angle 32:54 , (b) t=T = 0:34 with apping angle 29:55 , (c) t=T = 0:55 with apping angle 10:50 , (d) t=T = 0:69 with apping angle 32:82 . isotropic wing con guration. A similar mesh distribution is adopted for the three-dimensional uid computational domain with 492,630 unstructured - nite elements. The structural mesh is same as shown in Fig. 9b with 315 structured nite elements. The distance between the rst grid point in the 4 + boundary layer and wing surface is set as 9.2810 C with y 0:5 and the number of divisions along the vertical direction of the wing surface is 25 and stretching ratio, y =y is 1.2. The non-dimensional time step size, i+1 i tU =C is set to 0.0018. ref For the purpose of validation, the comparisons of normalized location in ver- tical direction and normalized displacement at the wing tip are given in Fig. 12, respectively. Results indicate that the overall trends of the apping wing are well predicted, compared with the experimental data. Fig. 13 graphically 35 Table 6 Aeroelastic parameters for an anisotropic wing under hovering motion Parameters Value Reference velocity (hovering) 4.58 m/s Air density 1.206 kg=m Reynolds number 7304 Flapping frequency 25 Hz Flapping amplitude 35 3 1.5 Present Present Experiment Experiment 2 1 1 0.5 0 0 -0.5 -1 -2 -1 -3 -1.5 -50 -25 0 25 50 -50 -25 0 25 50 ◦ ◦ α ( ) α ( ) (a) (b) Fig. 12. Comparison of coupled aeroelastic responses for anisotropic apping wing: (a) normalized location of L3B1 model in vertical direction at wing tip, (b) normal- ized displacement of L3B1 model at wing tip. (4) Present simulation, () Experi- ment [50]. shows the iso-surface of the non-dimensional Q-criterion of 1 for L3B1 model colored by the normalized velocity magnitude during a apping motion pe- riod. Detailed turbulent wake structures and their evolution process can be ob- served. Overall, the proposed uid- exible multibody solver is able to simulate exible apping wing with multibody components. The detailed aerodynamic characteristics around the apping wing and nonlinear structural responses are captured accurately, and compared well with the available experimental data. δ /C z,tip δ /C w,tip (a) t=T = 0 (b) t=T = 0:25 (c) t=T = 0:5 (d) t=T = 0:75 kUk /U : 0 0.2 0.4 0.6 0.8 1 1.2 1.4 ref Fig. 13. Flow past an anisotropic apping wing: Wake structures of L3B1 model @u @u 1 j based on the instantaneous iso-surfaces of Q(= ) value at (a) t=T = 0, 2 @x @x j i (b) t=T = 0:25, (c) t=T = 0:5, (d) t=T = 0:75. Iso-surfaces of non-dimensional + 2 Q Q(C=U ) = 1 are colored by the normalized velocity magnitude kUk =U . ref ref 4 Application to bat-like apping dynamics As mentioned earlier, the ight patterns and the mechanism of bats are quite dierent from birds and insects. The lack of understanding about how bats y eciently prompts researchers to investigate via experiments and CFD 37 simulations. However, the complex ight kinematics, the special wing struc- tures comprising skeletons with multi-degree of freedom and highly exible wing-membrane skin as well as the signi cant aeroelastic phenomena during ight prevent the research progresses for bat ight. The presently available results about the bat apping dynamics are quite limited, especially from fully-coupled computational modeling. For the rst time, we demonstrate the capability of our fully-coupled multibody aeroelastic framework to simulate a bat-like exible wing. The eects of exibility and aerodynamic load are investigated to explore their impact on the dynamics of the wing. 4.1 Structural model setup A bat-like wing made of several wing skin-like membranes and reinforced skele- tons is constructed to investigate the aeroelastic responses with dierent ight patterns. We consider the bat-like wing but ignore its body and other compo- nents. A schematic of the constructed bat-like wing is shown in Fig. 14a. The basic, simpli ed shape of the bat-like wing is referred to a apping bat robot made in Brown University [5]. In our simulation, the chord at the wing root is 0.27 m and the span is 0.69 m, resulting in a wing area of 0.124 m . A rigid strip is applied at the wing root to x the wing like the bat body. Two strips near to the wing root represent the humerus bone and radius bone, respectively. Several strips in the outer region stand for bat ngers. All these components are used to support the covered wing membranes. The detailed geometric sizes of these bones and ngers are designed to approach those of a real bat, based on the research work of P. Watts [53]. The widths and thicknesses of these bones and ngers are varied along their axial direction in order to adjust and achieve reasonable stiness distributions to support the wing membranes. Meanwhile, the thickness of the wing membrane becomes thinner in the outer region than that near to the wing root, which allows 38 Table 7 Material properties of the skeletons and wing membrane for the bat-like wing (! represents the variation of thickness along axial direction) Components Material properties M1/M2/M5 M3/M4 B1/B2 B3/B4 B5 B6 Young's modulus (MPa) 4 3 5000 3000 2500 2000 Poisson ratio 0.3 0.3 0.3 0.3 0.3 0.3 Density (kg=m ) 1100 1100 2200 2200 2200 2200 Thickness (mm) 0.8/0.6/0.4 0.2 5/4 2!1.5 3!2 2 0.69m M1 M3 Fingers Humerus bone B5 B2 B1 B3 B4 M4 0.27m Radius bone B6 M2 M5 Rigid wing root Wing membrane (a) (b) Fig. 14. Schematic of bat-like wing: (a) geometric information, (b) indication of dierent components. a relative larger elastic deformation in the outer region. The indication of six bones and ve membranes in this wing is provided in Fig. 14b. All the components in this bat-like wing are assumed to be made of homogeneous, isotropic material. Material properties of the skeletons and wing membranes in dierent wing regions are summarized in Table 7. 4.2 Eect of exibility To investigate the eect of exibility on the dynamics of a bat-like exible wing, a rigid wing with identical geometry is simulated and compared to its exible counterpart. To eliminate the eect of apping motion, the bat-like wing in a gliding ight is considered. The gliding ight is simulated by applying a uniform ow to the wing while constraining its root in all six degree-of- 39 freedom. Structural deformation due to aerodynamic load is allowed for the exible wing, but not allowed for the rigid wing. The detailed parameters for the simulation are summarized in Table 8. A schematic diagram of the bat-like wing setup is depicted in Fig. 15a. The length of chord at the wing root is de ned as C . The distances between the inlet boundary ( ) and in the outlet boundary ( ), the top and bottom boundaries ( ) and the out slip boundaries on both sides are the same of 200C . The no-slip Dirichlet boundary condition is applied on the surface of bat-like wing. The freestream velocity f f along the X-axis at the inlet boundary is de ned as u = U , where u in f f f f is the X-component of the ow velocity u = (u ; v ; w ). The slip boundary condition is applied on the top and bottom as well as the side boundaries. A traction-free boundary condition is de ned at the outlet boundary , where out = = = 0. xx yx zx The three-dimensional uid computational domain is discretized by 454,258 unstructured eight-node brick nite element meshes. A boundary layer is main- tained around the wing such that y < 1:0 in the wall-normal direction. Mesh distribution slice in the uid domain at mid-span is shown in Fig. 15b. Geo- metrically exact co-rotational shell elements are employed for wing membrane and bones. The structural model is discretized by 496 structured four-node rectangular nite elements. Mesh characteristics on the wing surface in both the uid domain and the structural domain are compared and given in Fig. 16a and 16b, respectively. The non-dimensional time step size is selected as tU=C = 0:022. Comparison of the deformed exible wing and rigid wing colored by non- dimensional displacement in the vertical direction is shown in Fig. 17a. The maximum displacement is observed at the wing tip and the covered mem- branes have some small wrinkles on the surface. Comparison of the mean lift coecient (C ) and the mean drag coecient (C ) for the rigid and exi- L D ble wing is presented in Table 9. C decreases from 0.6234 to 0.5746 and C L D 40 Table 8 Simulation parameters for a bat-like wing under gliding motion Parameters Value Semi-span length 0.69 m Chord length at wing root 0.27 m Wing area 0.124 m Freestream velocity 2.0 m/s Angle of attack 20 Air density 1.225 kg=m Mean chord-based Reynolds number 24609 reduces by 13.17% due to the passive deformation of exible wing and its compliant membrane components. As a result, the lift-to-drag ratio of exible wing (C /C ) increases by 6.13%, compared with the rigid wing. L D For further investigation, wake structures for both rigid and exible wings based on instantaneous iso-surfaces of non-dimensional Q-criterion of 0.25 col- ored by the normalized velocity magnitude are given in Fig. 17b. The shedding vortex structures behind the bat-like wing are changed by the wing deforma- tion via aeroelastic coupling. Fig. 18 provides streamlines around the rigid and exible wings colored by the normalized velocity magnitude. A massive sepa- ration ow on the upper surface is observed and the ow patterns are altered by the wing deformation, compared with the rigid wing counterpart. In sum- mary, the exibility of the wing signi cantly aects the dynamics of a bat-like wing. Further investigation can be conducted by simulating a range of exibil- ities to quantitatively investigate its eect on the aerodynamic performance of the wing. 41 wing 200C in Γ out 200C 200C (a) (b) Fig. 15. Flow past a bat-like apping wing: (a) schematic of computational setup, (b) mesh distribution slice in the uid domain at mid-span. Table 9 Comparison of mean lift coecient (C ), mean drag coecient (C ) and L D lift-to-drag ratio (C /C ) for rigid and exible wing L D Results C C C /C L D L D Rigid wing 0.6234 0.2391 2.61 Flexible wing 0.5746 (7.83 % #) 0.2076 (13.17% #) 2.77 (6.13 % ") 42 (a) (b) Fig. 16. Mesh characteristics on bat-like wing surface in (a) the uid domain, (b) the structural domain. (a) (b) Fig. 17. (a) Comparison of non-dimensional displacement contours in the vertical direction for both rigid and exible wings. (b) Wake structures of bat-like wing @u @u 1 j based on the instantaneous iso-surfaces of Q(= ) value for both rigid and 2 @x @x j i + 2 exible wings. Iso-surfaces of non-dimensional Q Q(C=U ) = 0:25 are colored ref by the normalized velocity magnitude kUk =U . ref 4.3 Eect of aerodynamic load In this section, we investigate the eect of aerodynamic load on the structural responses of the apping wing. Both uncoupled and coupled aeroelastic cases are simulated herein. The uncoupled case does not consider aerodynamic load in the apping motion, while the coupled aeroelastic case includes the cou- pling with surrounding air ow. The angle of attack is assumed as zero and the freestream velocity is set as 1.0 m/s for the apping ight. The apping amplitude is given as 20 and the apping frequency is set as 1 Hz. The iden- 43 (a) Rigid wing (b) Flexible wing kUk /U : 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 ref Fig. 18. Streamlines colored by the normalized velocity magnitude kUk =U around ref the bat-like wing for (a) rigid wing, (b) exible wing. tical mesh distributions in Section 4.2 for both uid and structural models are adopted for the apping wing simulation. The non-dimensional time step size is selected as tU=C = 0:00926. The comparisons of the normalized location and the displacement in the verti- cal direction at wing tip for both cases are shown in Fig. 19, respectively. The nonlinear dynamic responses for the exible bat-like wing have been aected by the aeroelastic coupling eect signi cantly and the lagging eect of coupled case becomes stronger than the uncoupled one. The iso-surfaces of the non- dimensional Q-criterion of 0.25 for the bat-like wing colored by the normalized velocity magnitude during a whole apping motion period are given in Fig. 20. The evolution of vortex generated from the leading edge, the trailing edge and the wing tip during apping ight can be observed. For a detailed analysis, streamlines around this wing at t=T = 0:25 and t=T = 0:75 are shown in Fig. 21, respectively. A massive separation ow on upper surface near to the wing tip is viewed at the downstroke motion. Such complex, nonlinear aerodynamic phenomena, including leading edge vortex (LEV) generation and trailing edge vortex (TEV) shedding, in uence the structural responses via the aeroelastic 44 1 0.4 Uncoupled Uncoupled Coupled Coupled 0.5 0.2 -0.5 -0.2 -1 -0.4 -20 -10 0 10 20 -20 -10 0 10 20 α ( ) α ( ) (a) (b) Fig. 19. Comparison of uncoupled and coupled aeroelastic structural responses of the bat-like wing for (a) normalized location in the vertical direction at wing tip, (b) normalized displacement at wing tip. coupling process. The eect of aerodynamic load on the bat-like apping wing is quite substantial. According to the discussion presented above, the proposed multibody aeroelas- tic solver is able to capture the physics of aeroelastic phenomena of the exible multibody apping wing and it can be extended to a more general bat-like wing. Meanwhile, the implementation of RBF method in current framework and combination with NIFC approach have been proved as an eective scheme to simulate exible apping wing. Finally, a bat-like wing with dierent exi- bilities for the bones and the membranes and the apping frequencies should be explored for further mechanism investigation based on the proposed aeroe- lastic framework. 5 Conclusion In this paper, a three-dimensional multibody aeroelastic framework is devel- oped by coupling an incompressible turbulent ow solver with DDES method and a exible multibody structural solver discretized by geometrically exact δ /C z,tip δ /C w,tip (a) t=T = 0 (b) t=T = 0:25 (c) t=T = 0:5 (d) t=T = 0:75 kUk /U : 0 0.2 0.4 0.6 0.8 1 1.2 1.4 ref Fig. 20. Flow past a bat-like wing: Wake structures of bat-like wing based on the in- @u @u 1 j stantaneous iso-surfaces of Q(= ) value at (a) t=T = 0, (b) t=T = 0:25, (c) 2 @x @x j i + 2 t=T = 0:5, (d) t=T = 0:75. Iso-surfaces of non-dimensional Q Q(C=U ) = 0:25 ref are colored by the normalized velocity magnitude kUk =U . ref co-rotational nite elements via partitioned domain decomposition strategy. The nonlinear iterative force correction approach has been implemented and coupled with the RBF interpolation method during the integration of incom- pressible turbulent ow and a exible multibody system to achieve a stable and robust partitioned coupling process. The proposed aeroelastic framework provides a feasible and high- delity tool to explore the design and optimization of a wide range of exible apping wings, such as bionic MAVs and UAVs. We validated the accuracy of our framework by simulating an anisotropic wing made of multiple reinforced battens and membranes. A good agreement to 46 (a) t=T = 0:25 (b) t=T = 0:75 kUk /U : 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 ref Fig. 21. Streamlines colored by the normalized velocity magnitude kUk =U around ref the bat-like wing at (a) t=T = 0:25, (b) t=T = 0:75. an experiment has been achieved, where the characteristic responses of the apping wing compared quite well with the experimental data. We further demonstrated the applicability of our framework by investigating the aeroe- lastic responses of a bat-like wing. The rigid and exible wings under gliding ight simulations were carried out and the eect of exibility on the dynamics of the bat-like wing was quanti ed. The lift-to-drag ratio increases by 6.13% for the exible wing due to the passive deformation of the wing and its com- pliant membranes, compared to its rigid counterpart. Subsequently, the eect of aerodynamic load has been studied by simulating a coupled and an uncou- pled aeroelastic cases. It is observed that the aerodynamic load enhances the lagging eect of the nonlinear structural responses signi cantly. In future, var- ious aspects of a bat ight can be investigated via the developed framework, which may include structural optimization, acoustic reduction, ight control, and among others. 47 Acknowledgements The authors wish to acknowledge supports from the National University of Singapore and the Ministry of Education, Singapore. References [1] K. Jones, M. Platzer, An experimental and numerical investigation of apping- wing propulsion, in: 37th Aerospace Sciences Meeting and Exhibit, 1999, p. [2] B. Singh, I. Chopra, Insect-based hover-capable apping wings for micro air vehicles: Experiments and analysis, Aiaa Journal 46 (46) (2008) 2115{2135. [3] S. Heathcote, D. Martin, I. Gursul, Flexible apping airfoil propulsion at zero freestream velocity, AIAA journal 42 (11) (2004) 2196{2204. [4] M. Hamamoto, Y. Ohta, K. Hara, T. Hisada, Application of uid{structure interaction analysis to apping ight of insects with deformable wings, Advanced Robotics 21 (1-2) (2007) 1{21. [5] J. W. Bahlman, S. M. Swartz, K. S. Breuer, Design and characterization of a multi-articulated robotic bat wing, Bioinspiration & biomimetics 8 (1) (2013) [6] S. Wei, M. Berg, D. Ljungqvist, Flapping and exible wings for biological and micro air vehicles, Progress in Aerospace Sciences 35 (5) (1999) 455{505. [7] K. V. Rozhdestvensky, V. A. Ryzhov, Aerohydrodynamics of apping-wing propulsors, Progress in Aerospace Sciences 39 (8) (2003) 585{633. [8] M. S. Triantafyllou, A. H. Techet, F. S. Hover, Review of experimental work in biomimetic foils, IEEE Journal of Oceanic Engineering 29 (3) (2004) 585{594. [9] M. F. Platzer, K. D. Jones, J. Young, J. S. Lai, Flapping wing aerodynamics: progress and challenges, AIAA journal 46 (9) (2008) 2136{2149. 48 [10] W. Shyy, H. Aono, S. K. Chimakurthi, P. Trizila, C. K. Kang, C. E. S. Cesnik, H. Liu, Recent progress in apping wing aerodynamics and aeroelasticity, Progress in Aerospace Sciences 46 (7) (2010) 284{327. [11] N. S. Proctor, M. S. Lynch, Patrick J., Manual of ornithology: Avian structure and function, Quarterly Review of Biology. [12] W. Shyy, Y. Lian, J. Tang, D. Viieru, H. Liu, Aerodynamics of low Reynolds number yers, Vol. 22, Cambridge University Press, 2007. [13] H. Dai, Computational modeling of uid-structure interaction in biological ying and swimming, Ph.D. thesis, Vanderbilt University (2013). [14] S. M. Swartz, Skin and bones: the mechanical properties of bat wing tissues, Bats: Phylogeny, Morphology, Echolocation, and Conservation Biology (1998) 109{126. [15] A. Gogulapati, P. P. Friedmann, E. Kheng, W. Shyy, Approximate aeroelastic modeling of apping wings in hover, AIAA journal 51 (3) (2013) 567{583. [16] C. Farhat, V. K. Lakshminarayan, An ale formulation of embedded boundary methods for tracking boundary layers in turbulent uid{structure interaction problems, Journal of Computational Physics 263 (2014) 53{70. [17] H. Cho, N. Lee, S. J. Shin, S. Lee, S. Kim, Improved computational approach for 3-d realistic insect-like apping wing using co-rotational nite elements, in: 55th AIAA Aerospace Sciences Meeting, 2017, p. 1417. [18] S. Wang, X. Zhang, G. He, T. Liu, Lift enhancement by bats' dynamically changing wingspan, Journal of the Royal Society Interface 12 (113) (2015) [19] F. J. Blom, A monolithical uid-structure interaction algorithm applied to the piston problem, Computer methods in applied mechanics and engineering 167 (3-4) (1998) 369{391. [20] J. Liu, R. K. Jaiman, P. S. Gurugubelli, A stable second-order scheme for uid{ structure interaction with strong added-mass eects, Journal of Computational Physics 270 (2014) 687{710. 49 [21] C. A. Felippa, K. Park, C. Farhat, Partitioned analysis of coupled mechanical systems, Computer methods in applied mechanics and engineering 190 (24-25) (2001) 3247{3270. [22] A. Yenduri, R. Ghoshal, R. Jaiman, A new partitioned staggered scheme for exible multibody interactions with strong inertial eects, Computer Methods in Applied Mechanics and Engineering 315 (2017) 316{347. [23] J. Hron, S. Turek, A monolithic fem/multigrid solver for an ale formulation of uid-structure interaction with applications in biomechanics, in: Fluid-structure interaction, Springer, 2006, pp. 146{170. [24] R. K. Jaiman, S. Sen, P. S. Gurugubelli, A fully implicit combined eld scheme for freely vibrating square cylinders with sharp and rounded corners, Computers & Fluids 112 (2015) 1{18. [25] R. Jaiman, P. Geubelle, E. Loth, X. Jiao, Combined interface boundary condition method for unsteady uid{structure interaction, Computer Methods in Applied Mechanics and Engineering 200 (1-4) (2011) 27{39. [26] R. Jaiman, P. Geubelle, E. Loth, X. Jiao, Transient uid{structure interaction with non-matching spatial and temporal discretizations, Computers & Fluids 50 (1) (2011) 120{135. [27] D. D. Chin, D. Lentink, Flapping wing aerodynamics: from insects to vertebrates, Journal of Experimental Biology 219 (7) (2016) 920{932. [28] P. Causin, J.-F. Gerbeau, F. Nobile, Added-mass eect in the design of partitioned algorithms for uid{structure problems, Computer methods in applied mechanics and engineering 194 (42-44) (2005) 4506{4527. [29] M. Olivier, A uid-structure interaction partitioned algorithm applied to exible apping wing propulsion, Ph.D. thesis, Universit Laval (2014). [30] R. Jaiman, X. Jiao, P. Geubelle, E. Loth, Assessment of conservative load transfer for uid{solid interface with non-matching meshes, International Journal for Numerical Methods in Engineering 64 (15) (2005) 2014{2038. 50 [31] Y. Li, Y. Law, V. Joshi, R. Jaiman, A 3d common-re nement method for non- matching meshes in partitioned variational uid-structure analysis, Journal of Computational Physics. [32] T. Rendall, C. Allen, Uni ed uid{structure interpolation and mesh motion using radial basis functions, International Journal for Numerical Methods in Engineering 74 (10) (2008) 1519{1559. [33] M. Lombardi, N. Parolini, A. Quarteroni, Radial basis functions for inter-grid interpolation and mesh motion in fsi problems, Computer Methods in Applied Mechanics and Engineering 256 (2013) 117{131. [34] A. Beckert, H. Wendland, Multivariate interpolation for uid-structure- interaction problems using radial basis functions, Aerospace Science and Technology 5 (2) (2001) 125{134. [35] A. de Boer, A. Van Zuijlen, H. Bijl, Review of coupling methods for non- matching meshes, Computer methods in applied mechanics and engineering 196 (8) (2007) 1515{1525. [36] A. De Boer, M. Van der Schoot, H. Bijl, Mesh deformation based on radial basis function interpolation, Computers & structures 85 (11-14) (2007) 784{795. [37] F. M. Bos, Numerical simulations of apping foil and wing aerodynamics: Mesh deformation using radial basis functions, Ph.D. thesis, Delft University of Technology (2010). [38] O. A. Bauchau, Flexible multibody dynamics, Vol. 176, Springer Science & Business Media, 2010. [39] O. Bauchau, G. Damilano, N. Theron, Numerical integration of non-linear elastic multi-body systems, International Journal for Numerical Methods in Engineering 38 (16) (1995) 2727{2751. [40] O. Bauchau, N. J. Theron, Energy decaying scheme for non-linear beam models, Computer Methods in Applied Mechanics and Engineering 134 (1-2) (1996) 37{ 51 [41] C. L. Bottasso, M. Borri, Energy preserving/decaying schemes for non-linear beam dynamics using the helicoidal approximation, Computer Methods in Applied Mechanics and Engineering 143 (3-4) (1997) 393{415. [42] O. A. Bauchau, C. L. Bottasso, On the design of energy preserving and decaying schemes for exible, nonlinear multi-body systems, Computer Methods in Applied Mechanics and Engineering 169 (1-2) (1999) 61{79. [43] P. S. Gurugubelli, R. Ghoshal, V. Joshi, R. K. Jaiman, A variational projection scheme for nonmatching surface-to-line coupling between 3d exible multibody system and incompressible turbulent ow, Computers & Fluids 165 (2018) 160{ [44] V. Joshi, R. K. Jaiman, A variationally bounded scheme for delayed detached eddy simulation: Application to vortex-induced vibration of oshore riser, Computers & Fluids 157 (2017) 84{111. [45] R. Jaiman, N. Pillalamarri, M. Guan, A stable second-order partitioned iterative scheme for freely vibrating low-mass blu bodies in a uniform ow, Computer Methods in Applied Mechanics and Engineering 301 (2016) 187{215. [46] A. N. Brooks, T. J. Hughes, Streamline upwind/petrov-galerkin formulations for convection dominated ows with particular emphasis on the incompressible navier-stokes equations, Computer methods in applied mechanics and engineering 32 (1-3) (1982) 199{259. [47] P. R. Spalart, Detached-eddy simulation, Annual review of uid mechanics 41 (2009) 181{202. [48] J. C. Simo, D. D. Fox, On a stress resultant geometrically exact shell model. part i: Formulation and optimal parametrization, Computer Methods in Applied Mechanics and Engineering 72 (3) (1989) 267{304. [49] O. A. Bauchau, J.-Y. Choi, C. L. Bottasso, Time integrators for shells in multibody dynamics, Computers & structures 80 (9-10) (2002) 871{889. [50] P. Wu, Experimental characterization, design, analysis and optimization of 52 exible apping wings for micro air vehicles, Ph.D. thesis, University of Florida (2010). [51] H. Aono, S. K. Chimakurthi, P. Wu, E. S allstr om, B. K. Stanford, C. E. Cesnik, P. Ifju, L. Ukeiley, W. Shyy, A computational and experimental study of exible apping wing aerodynamics, in: 48th AIAA aerospace sciences meeting including the new horizons forum and aerospace exposition, 2010, pp. 4{7. [52] A. Gogulapati, Nonlinear approximate aeroelastic analysis of apping wings in hover and forward ight., Ph.D. thesis, University of Michigan (2011). [53] P. Watts, E. J. Mitchell, S. M. Swartz, A computational model for estimating the mechanics of horizontal apping ight in bats: model description and validation, Journal of Experimental Biology 204 (16) (2001) 2873{2898. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Physics arXiv (Cornell University) http://www.deepdyve.com/lp/arxiv-cornell-university/a-novel-3d-variational-aeroelastic-framework-for-flexible-multibody-L5HyG3unsy

Loading next page...

References (62)

Yulong Li, Y. Law, V. Joshi, R. Jaiman (2017)
A 3D common-refinement method for non-matching meshes in partitioned variational fluid-structure analysis
J. Comput. Phys., 374
H. Ahn, Y. Kallinderis (2006)
Strongly coupled flow/structure interactions with a geometrically conservative ALE scheme on general hybrid meshes
J. Comput. Phys., 219
F. Blom (1998)
A monolithical fluid-structure interaction algorithm applied to the piston problem
Computer Methods in Applied Mechanics and Engineering, 167
O. Bauchau, G. Damilano, N. Theron (1995)
Numerical integration of non‐linear elastic multi‐body systems
International Journal for Numerical Methods in Engineering, 38
Dd Wang, Js Chen, 王东东 (2004)
Computer Methods in Applied Mechanics and Engineering
O. Bauchau, Jou-Young Choi, C. Bottasso (2002)
Time integrators for shells in multibody dynamics
Computers & Structures, 80
S. Heathcote, D. Martin, I. Gursul (2003)
Flexible Flapping Airfoil Propulsion at Zero Freestream Velocity
AIAA Journal, 42
Jintai Chung, G. Hulbert (1993)
A Time Integration Algorithm for Structural Dynamics With Improved Numerical Dissipation: The Generalized-α Method
Journal of Applied Mechanics, 60
A. Brooks, T. Hughes (1990)
Streamline upwind/Petrov-Galerkin formulations for convection dominated flows with particular emphasis on the incompressible Navier-Stokes equations
Computer Methods in Applied Mechanics and Engineering, 32
A. Boer, M. Schoot, H. Bijl (2007)
Mesh deformation based on radial basis function interpolation
Computers & Structures, 85
C. Bottasso, M. Borri (1997)
Energy preserving/decaying schemes for non-linear beam dynamics using the helicoidal approximation
Computer Methods in Applied Mechanics and Engineering, 143
Armin Beckert, H. Wendland (2001)
Multivariate interpolation for fluid-structure-interaction problems using radial basis functions
Aerospace Science and Technology, 5
C. Felippa, K. Park, C. Farhat (2001)
Partitioned analysis of coupled mechanical systems
Computer Methods in Applied Mechanics and Engineering, 190
A. Yenduri, R. Ghoshal, R. Jaiman (2017)
A new partitioned staggered scheme for flexible multibody interactions with strong inertial effects
Computer Methods in Applied Mechanics and Engineering, 315
Diana Chin, D. Lentink (2016)
Flapping wing aerodynamics: from insects to vertebrates
Journal of Experimental Biology, 219
Abhijit Gogulapati, P. Friedmann, Eugene Kheng, W. Shyy (2013)
Approximate Aeroelastic Modeling of Flapping Wings in Hover
AIAA Journal, 51
Beerinder Singh, I. Chopra (2008)
Insect-Based Hover-Capable Flapping Wings for Micro Air Vehicles: Experiments and Analysis
AIAA Journal, 46
U. Küttler, W. Wall (2008)
Fixed-point fluid–structure interaction solvers with dynamic relaxation
Computational Mechanics, 43
M. Fernández, Jean-Frédéric Gerbeau, C. Grandmont (2007)
A projection semi‐implicit scheme for the coupling of an elastic structure with an incompressible fluid
International Journal for Numerical Methods in Engineering, 69
J. Simo, D. Fox, M. Rifai (1990)
On a stress resultant geometrically exact shell model
Hu Dai (2013)
Computational modeling of fluid–structure interaction in biological flying and swimming
P. Gurugubelli, R. Ghoshal, V. Joshi, R. Jaiman (2017)
A Variational Projection Scheme for Nonmatching Surface-to-Line Coupling between 3D Flexible Multibody System and Incompressible Turbulent Flow
Computers & Fluids, 165
R. Jaiman, Subhankar Sen, P. Gurugubelli (2015)
A fully implicit combined field scheme for freely vibrating square cylinders with sharp and rounded corners
Computers & Fluids, 112
O. Bauchau, C. Bottasso (1999)
On the design of energy preserving and decaying schemes for flexible, nonlinear multi-body systems
Computer Methods in Applied Mechanics and Engineering, 169
Mathieu Olivier (2014)
A fluid-structure interaction partitioned algorithm applied to flexible flapping wing propulsion
R. Jaiman, P. Geubelle, E. Loth, X. Jiao (2011)
Combined interface boundary condition method for unsteady fluid–structure interaction
Computer Methods in Applied Mechanics and Engineering, 200
P. Watts, E. Mitchell, S. Swartz (2001)
A computational model for estimating the mechanics of horizontal flapping flight in bats: model description and validation.
The Journal of experimental biology, 204 Pt 16
N. Proctor, P. Lynch (1993)
Manual of Ornithology: Avian Structure and Function
P. Spalart (2009)
Detached-Eddy Simulation
Annual Review of Fluid Mechanics, 41
M. Platzer, K. Jones, J. Young, J. Lai (2008)
Flapping Wing Aerodynamics: Progress and Challenges
AIAA Journal, 46
Jaroslav Hron, S. Turek (2006)
A Monolithic FEM/Multigrid Solver for an ALE Formulation of Fluid-Structure Interaction with Applications in Biomechanics
(2010)
Experimental characterization
W. Shyy, M. Berg, Daniel Ljungqvist (1999)
Flapping and flexible wings for biological and micro air vehicles
Progress in Aerospace Sciences, 35
P. Causin, Jean-Frédéric Gerbeau, F. Nobile (2005)
Added-mass effect in the design of partitioned algorithms for fluid-structure problems
Computer Methods in Applied Mechanics and Engineering, 194
H. Matthies, R. Niekamp, J. Steindorf (2006)
Algorithms for strong coupling procedures
Computer Methods in Applied Mechanics and Engineering, 195
M. Lombardi, N. Parolini, A. Quarteroni (2013)
Radial basis functions for inter-grid interpolation and mesh motion in FSI problems
Computer Methods in Applied Mechanics and Engineering, 256
V. Joshi, R. Jaiman (2017)
A variationally bounded scheme for delayed detached eddy simulation: Application to vortex-induced vibration of offshore riser
Computers & Fluids, 157
F. Bos (2010)
Numerical simulations of flapping foil and wing aerodynamics: Mesh deformation using radial basis functions
Shizhao Wang, Xing Zhang, G. He, Tianshu Liu (2015)
Lift enhancement by bats' dynamically changing wingspan
Journal of The Royal Society Interface, 12
W. Shyy, Y. Lian, Jian Tang, Dragos Viieru, Hao Liu (2007)
Aerodynamics of Low Reynolds Number Flyers
A. Boer, A. Zuijlen, H. Bijl (2007)
Review of coupling methods for non-matching meshes
Computer Methods in Applied Mechanics and Engineering, 196
J. Simo, D. Fox (1989)
On stress resultant geometrically exact shell model. Part I: formulation and optimal parametrization
Applied Mechanics and Engineering, 72
J. Bahlman, S. Swartz, K. Breuer (2013)
Design and characterization of a multi-articulated robotic bat wing
Bioinspiration & Biomimetics, 8
Jie Liu, R. Jaiman, P. Gurugubelli (2014)
A stable second-order scheme for fluid-structure interaction with strong added-mass effects
J. Comput. Phys., 270
M. Triantafyllou, A. Techet, F. Hover (2004)
Review of experimental work in biomimetic foils
IEEE Journal of Oceanic Engineering, 29
(1998)
Skin and bones: the mechanical properties of bat wing tissues
Abhijit Gogulapati (2011)
Nonlinear approximate aeroelastic analysis of flapping wings in hover and forward flight
Haeseong Cho, N. Lee, Sang-Joon Shin, Seungsoo Lee, Sangyong Kim (2017)
Improved Computational Approach for 3-D Realistic Insect-like Flapping Wing using Co-rotational Finite Elements
K. Rozhdestvensky, V. Ryzhov (2003)
Aerohydrodynamics of flapping-wing propulsors
Progress in Aerospace Sciences, 39
(1999)
An experimental and numerical investigation of flappingwing propulsion
O. Bauchau, N. Theron (1996)
Energy decaying scheme for non-linear beam models
Computer Methods in Applied Mechanics and Engineering, 134
T. He, Jian Yang, C. Baniotopoulos (2018)
Improving the CBS‐based partitioned semi‐implicit coupling algorithm for fluid‐structure interaction
International Journal for Numerical Methods in Fluids, 87
R. Jaiman, X. Jiao, P. Geubelle, E. Loth (2005)
Assessment of conservative load transfer for fluid–solid interface with non‐matching meshes
International Journal for Numerical Methods in Engineering, 64
W. Shyy, H. Aono, S. Chimakurthi, Pat Trizila, Chang-kwon Kang, C. Cesnik, Hao Liu (2010)
Recent progress in flapping wing aerodynamics and aeroelasticity
Progress in Aerospace Sciences, 46
Pin Wu (2010)
Experimental characterization, design, analysis and optimization of flexible flapping wings for micro air vehicles
Masaki Hamamoto, Yoshiji Ohta, Keita Hara, T. Hisada (2007)
Application of fluid–structure interaction analysis to flapping flight of insects with deformable wings
Advanced Robotics, 21
H. Aono, S. Chimakurthi, Pin Wu, Erik Sällström, B. Stanford, C. Cesnik, P. Ifju, L. Ukeiley, W. Shyy (2010)
A computational and experimental study of flexible flapping wing aerodynamics
R. Jaiman, N. Pillalamarri, M. Guan (2016)
A stable second-order partitioned iterative scheme for freely vibrating low-mass bluff bodies in a uniform flow
Computer Methods in Applied Mechanics and Engineering, 301
C. Farhat, V. Lakshminarayan (2014)
An ALE formulation of embedded boundary methods for tracking boundary layers in turbulent fluid-structure interaction problems
J. Comput. Phys., 263
O. Bauchau (2010)
Flexible multibody dynamics
R. Jaiman, P. Geubelle, E. Loth, X. Jiao (2011)
Transient fluid–structure interaction with non-matching spatial and temporal discretizations
Computers & Fluids, 50
T. Rendall, C. Allen (2008)
Unified fluid–structure interpolation and mesh motion using radial basis functions
International Journal for Numerical Methods in Engineering, 74

ISSN: 0045-7930
eISSN: ARCH-3341
DOI: 10.1016/j.compfluid.2018.11.013
Publisher site: See Article on Publisher Site

Abstract

We present a novel three-dimensional (3D) variational aeroelastic framework for apping wing with a exible multibody system subjected to an external incom- pressible turbulent ow. The proposed aeroelastic framework consists of a three- dimensional uid solver with a hybrid RANS/LES model based on the delayed detached eddy simulation (DDES) treatment and a nonlinear monolithic elastic structural solver for the exible multibody system with constraints. Radial basis function (RBF) is applied in this framework to transfer the aerodynamic forces and structural displacements across the discrete non-matching interface meshes while satisfying a global energy conservation. For the consistency of the interface data transfer process, the mesh motion of the uid domain with large elastic defor- mation due to high-amplitude apping motion is also performed via the standard radial basis functions. The uid equations are discretized using a stabilized Petrov- Galerkin method in space and the generalized- approach is employed to integrate the solution in time. The exible multibody system is solved by using geometrically exact co-rotational nite element method and an energy decaying scheme is used to achieve numerical stability of the multibody solver with constraints. A nonlinear iterative force correction (NIFC) scheme is applied in a staggered partitioned itera- tive manner to maintain the numerical stability of aeroelastic coupling with strong Preprint submitted to Elsevier Science 13 July 2018 arXiv:1807.04411v1 [physics.flu-dyn] 12 Jul 2018 added mass eect. An isotropic aluminum wing with apping motion is simulated via the proposed aeroelastic framework and the accuracy of the coupled solution is validated with the available experimental data. We next study the robustness and reliability of the 3D exible multibody aeroelastic framework for an anisotropic apping wing ight involving battens and membranes with composite material and compare against the experimental results. Finally, we demonstrate the aeroelastic framework for a bat-like wing and examine the eects of exibility on the apping wing dynamics. Key words: 3D exible multibody aeroelasticity, Radial basis functions, Flapping wing, Large elastic deformation, Partitioned iterative. 1 Introduction 1.1 Background on biologically-inspired apping ight Biologically-inspired apping ight has intrigued human-kind over the past several centuries. Recently, there is a growing trend in aeronautical engineer- ing applications to incorporate the understanding of apping ight dynamics of birds, insects and bats. A lot of remarkable research work in this area has been carried out during the past decades, including experiments [1,2], compu- tational uid dynamics (CFD) simulations [3,4] and real apping robots [5]. A number of recent reviews on several aspects of apping ight have been doc- umented [6,7,8,9,10]. Compared with a xed-wing ight vehicle, the apping ight of birds and bats involve active morphing and adaptive exible wing con guration, which may oer some unique bene ts with regard to eciency, noise and manoeuvrability [11]. Therefore, the understanding of apping wing Corresponding author Email address: mperkj@nus.edu.sg (R. K. Jaiman). 2 mechanisms can be useful to incorporate in the designs of micro air vehi- cles (MAVs) and unmanned aerial vehicles (UAVs). An unsteady turbulent aerodynamics interacting with nonlinear exible multibody dynamics poses a serious challenge in the study of such apping ight dynamics of animals. The emerging engineering requirements and the fundamental understanding of bio-inspired apping ight have motivated the present computational devel- opment focusing on the exible multibody solver interacting with a separated turbulent ow. During the past decades, numerous researchers [9,12,13] have explored the biologically-inspired apping ight for dierent ying species (e.g., insects, birds and bats). The aerodynamic characteristics around rigid and exible apping wings as well as the coupled nonlinear structural responses were in- vestigated in detail. However, several challenges are still remained for a more comprehensive and complete understanding of apping mechanism and the role of exibility during ight [9]. For example, previous researches are lim- ited to some speci c airfoils and simpli ed insect-like wings, which lack the generalization to a real apping ight. Biological structures are quite dierent for various species, which lead to their unique apping patterns with speci c mechanisms. Most of the insects have single or several pairs of wings consisted of the vein and membrane components with widely varying distributions of exibility [12]. Speci cally, a wing of the bird is made of bones and muscles to control its ight attitude and the surface is covered by dierent types of feathers. In particular, the unique skeletal anatomical structure of a bat has more degrees of freedom, compared with insects and birds and the membrane- like wing skin has high exibility with anisotropic material properties [14]. It is noticed that a physical structure of dierent wings can be considered as a generic exible multibody system with kinematic constraints. Recently, several work focus on the study of uid- exible multibody interaction, and a realistic apping wing including the vein and membrane components is incorporated 3 into the simulation. For instance, Gogulapati et al. [15] developed an approx- imate aerodynamic model coupled with shell elements to simulate a exible anisotropic apping wing. Farhat et al. [16] built an ALE-embedded compu- tational framework to deal with aeroelastic problems with large structural de- formation and demonstrated this framework for the anisotropic apping wing. The underlying structural models in these two work are discretized only with shell elements. In Cho et al. [17], the co-rotational beam elements are used for veins and the co-rotational shell elements are employed for the anisotropic apping wing simulation. All the numerical methods in the above mentioned research work were developed to analyze and to provide a physical understand- ing of the aeroelastic phenomena around insect-like or bird-mimicking wings. Only a handful of publications on 3D fully coupled aeroelastic analysis on a bat apping dynamics can be found in the literature, compared with a large number of work on the ight of birds and insects [18]. The main objectives of the present work are (i) to develop a fully-coupled aeroelastic framework for exible multibody analysis of apping ight dynamics, and (ii) to demonstrate the proposed aeroelastic framework for a bat-like wing. 1.2 Aspects of computational modeling Generally, a typical numerical simulation of the aeroelastic problem involves the coupling of the governing equations of uid and structural dynamical sys- tems in two dierent domains. Currently, two main aeroelastic schemes are considered to couple the uid equations and the structural equations namely, monolithic [19,20] and partitioned [21,22]. A monolithic approach can achieve good numerical stability for the aeroelastic problem with strong added mass eect, which assembles the uid and structural equations into a single block then solves the coupled equations in a uni ed manner. However, it is dicult to take a full advantage of the existing stable and advanced uid and struc- 4 tural solvers, which restricts the scalability and exibility of an aeroelastic framework [20,23,24]. Considering the drawbacks of a monolithic scheme, a partitioned approach is developed to employ the existing suitable uid and structural solvers to solve the complex and generic aeroelastic problem. The uid and structural equations are solved in a sequential manner, and the trac- tion and velocity continuities are satis ed along the uid-structure interface to achieve numerical stability and accuracy [21,25,26]. However, the added mass eect associated with acceleration of the apping wing during ight and geometry of the wing with thin structures may result to numerical instability in a partitioned approach [27,28,29]. For the purpose of generality of exi- ble multibody aeroelastic analysis, we adopt a partitioned approach in the present study. In what follows, some backgrounds and relevant literature are associated with the partitioned aeroelastic framework for exible multibody simulations. In a typical partitioned-based aeroelastic scheme, the surface boundary data must be transferred along the interface between the uid and structural do- mains to satisfy a Dirichlet-type interface conditon (displacements or velocity) and a Neumann-type ( uid momentum ux or traction) interface condition. A proper care during interpolation and projection process is required to transfer the physical data accurately across non-matching meshes between the parti- tioned uid and structural domains [26,30,31]. Global and local energy conser- vation should be satis ed during the aeroelastic coupling while maintaining the accuracy of data transfer along the interface via Dirichlet-Neumann coupling. Recently, a high accuracy interface projection scheme with the global and local energy conservation is successfully applied to 3D uid-structure problems [31]. In addition to the interface projection problem, a large structural deformation due to high-amplitude apping motion poses a challenge for mesh motion to maintain good mesh quality in a transient aeroelastic computation. In the recent years, the radial basis function (RBF) method has been demonstrated 5 as a simpli ed approach to interpolate scattered data [32,33] while satisfying the property of the conservation of global energy transfer. For several aeroe- lastic problems, the RBF-based interpolation has been employed for the data transfer across non-matching meshes [34,35] and the mesh motion with large deformation [36,37]. The marked advantage of the RBF method is that the connectivity of the nodes is not required, which provides an eective and con- venient way to implement RBF in any existing framework. For the ease of implementation during the coupling of uid domain with exible multibody surfaces, we consider the radial basis functions for the interface coupling and the mesh motion in the present study. In the context of the present aeroelastic work, the structure of a exible wing consists of various components with dierent material properties and the rel- ative inertia, which can be considered as a nonlinear elastic multibody sys- tem. The elastic behavior of such a system is inherently nonlinear, hence the elastic displacements and rotations cannot be assumed to be small [38]. Geo- metrically exact elements are able to deal with arbitrarily large displacement and rotation components in a multibody system exactly and those compo- nents are de ned in a common inertial frame. Energy preserving (EP) time integrators are generally employed to ensure the discrete conservation of the total mechanical energy while achieving nonlinear unconditional stability for multibody systems. However, the undesired high frequency oscillations lead to numerical divergence for the solution of the nonlinear equations of motion when solving some complex problems with large number of degrees of freedom [39]. The nonlinearities of the system induce the energy to transfer from low to the high frequency modes. The energy decaying (ED) scheme [40,41,42] is designed based on energy preserving (EP) scheme to deal with the high fre- quencies in a exible multibody system, which can achieve nonlinear uncon- ditional stability. The multibody interaction is solved by a time discontinuous Galerkin scheme based on the ED approach. The constraints in the system are 6 typically enforced by the Lagrange multiplier technique. Such a exible multi- body formulation has been successfully applied for the interaction of turbulent ow with a oater-mooring system comprising a rigid- oater body and a long exible riser and mooring lines [43]. We adopt a similar exible multibody formulation for the present aeroelastic framework. In nature, the ight condition of ying animals varies from low Reynolds numbers to high Reynolds numbers, depending on the real size, ying speed and ight environment of these animals. Such a highly variable ight condition involves rich aerodynamic physics during ight, including leading edge vortex (LEV) generation, massively separated ow, laminar-turbulent transition and trailing edge vortex (TEV) shedding [12]. Therefore, a practical turbulent model is desired to capture the complex ow dynamics around a apping wing. Compared to direct numerical simulation (DNS) and large eddy simulation (LES) methods, the delayed detached eddy simulation (DDES) method has signi cant advantages to save the computational resources while keeping the reasonable accuracy for the separated turbulent ow, as demonstrated by [44] for large-scale uid-structure simulations. The DDES model oers an eective and reasonable way to simulate the vortex structures played during a apping ight, hence we employ this model in our computational framework. 1.3 Contributions and organization In this study, a novel 3D variational aeroelastic framework is developed to simulate the apping ight with exible multibody system (e.g., insect-like and bat-like wing) in a turbulent ow. While the Petrov-Galerkin nite ele- ment method is used to solve Navier-Stokes equations for the external uid ow, geometrically nonlinear co-rotational nite element method is applied to the exible multibody system in the structural domain. The DDES method is employed to simulate the turbulent separated ow during apping motion. 7 The ALE uid-turbulent solver and the exible multibody solver are cou- pled via a partitioned iterative scheme combined with the nonlinear iterative force correction (NIFC) approach, which achieve numerical stabilization for the coupled aeroelastic framework. The RBF-based data interpolation ap- proach is implemented in our framework to transfer the aerodynamic forces and the structural displacements across the uid-structure interface while sat- isfying the property of conservation of energy transfer. The interface force is corrected at the end of each uid sub-iteration by means of the NIFC method. The RBF approach with a compact support is utilized to handle the mesh motion with a large deformation condition and keep the initial mesh quality in simulation. For the initial validation, an isotropic aluminum wing and an anisotropic wing with composite material in apping ight condition are sim- ulated by using the proposed novel 3D variational aeroelastic framework with exible multibody dynamics. Results are compared against results obtained from experiments and literature. Finally, we demonstrate the proposed aeroe- lastic framework to simulate a bat-like exible wing with supported skeletons and covered membranes. In the present paper, we address two important challenges associated with the variationally coupled aeroelastic framework with exible multibody sys- tem for the exible apping wing simulation: (i) coupling of an incompressible turbulent ow and a exible multibody system with geometrically nonlin- ear co-rotational nite elements, (ii) the interpolation of aerodynamic forces and structural displacements between uid surface elements and structural nite elements in a exible multibody environment. For the rst challenge, the proposed partitioned iterative scheme referred above is used to achieve the coupling between the uid solver and the multibody structural solver in a robust and generic manner. It is worth noting that two main problems should be considered when dealing with the second challenge. Firstly, the information on the multiple surfaces belonging to dierent components of an entire struc- 8 tural model need to be collected then exchanged with the information from the uid domain in an energy conservation manner. While the aerodynamic tractions at the nodes of each uid element are interpolated to the target multiple structural elements by using the RBF method, the collected nodal structural displacements and the velocity vectors of each structural element are interpolated to the corresponding uid mesh nodes. To address the numer- ical instability caused by the added mass eect, the strongly-coupled NIFC implementation [45] is employed. The force equilibrium and the velocity con- tinuity condition conservation on the interface are satis ed by evaluating the approximate interface force corrections in the nonlinear sequence transforma- tion. The generalization of Aitken's extrapolation technique is applied to the iterative sequence coupling, which achieves a stable and convergent force updating process. The remainder of this manuscript is organized as follows. In Section 2, the variational formulations for the uid and the exible multibody system with constraints are reviewed. The detailed procedures for aeroelastic interface in- terpolation and the NIFC-based uid- exible multibody coupling by using RBF method are presented. The implementation of RBF method in mesh motion with large deformation condition is then introduced. An isotropic alu- minum wing and an anisotropic wing with composite material in apping ight condition are simulated with the proposed aeroelastic framework and compared with the available simulation and experimental data for validation purpose in Section 3. Section 4 presents an application on a bat-like exible wing with skeletons and membrane and explores the eects of exibility and aerodynamic load. The key conclusions of the present work are summarized in Section 5. 9 2 Partitioned aeroelastic framework for exible multibody system The governing equations and the formulations for the present 3D variational aeroelastic framework are similar to those of [43]. For the sake of complete- ness, we review the variational formulations for the moving uid and exible multibody solvers, whereas the Navier-Stokes equations are solved using the Petrov-Galerkin nite element method in the ALE coordinates and the exible multibody system is solved via nonlinear co-rotational nite element method. 2.1 Petrov-Galerkin nite element for turbulent ow Consistent with the work of [43], the Navier-Stokes equations are discretized using a stabilized Petrov-Galerkin formulation. The gerneralized- method is implemented to integrate the ALE ow solution in time domain, which can achieve unconditionally stable as well as second-order accuracy for linear problem. Furthermore, user-controlled high frequency damping desired for a coarser discretization in space and time is enabled by this scheme. The solution updates for the ow variables with the generalized- scheme can be written as f;n+1 f;n f;n f;n+1 f f u = u + t((1 )@ u + @ u ) (1) t t h h h h f;n+ f;n+1 f;n f f u = u + (1 )u (2) h h h f;n+ f;n+1 f;n m f f @ u = @ u + (1 )@ u (3) t t t h m h m h m;n+ m;n+1 m;n f f u = u + (1 )u (4) h h h f;n+1 m;n where u and u represent the uid and mesh velocities de ned for h h f f f each spatial uid point x 2 (t), respectively, whereas x and t are the f f f spatial and temporal coordinates. Here, , and represent the standard 10 integration parameters given as 1 1 3 1 f f 1 f f f = ; = ; = + (5) m m f f 2 2 1 + 1 + 1 1 The uid spatial domain can be discretized into n number of non-overlapping el f e f;h el nite elements and . While S represents the space of the trial e=1 f;h solutions, V denotes the space of test function. The variational formulation of the uid equations within the Petrov-Galerkin framework can be written f;n+ f;n+1 f f;h f;h as: nd [u ; p ] 2 S such that 8[ ; q ] 2 V h h h f f f f;n+ f;n+ m;n+ f;n+ f (@ u + (u u )ru ) d h h h h h Z Z f;n+ f;n+ f f des + : r d + : r d h h h h e e f;n+ rq u d el f f f;n+ m;n+ f f f e + ( (u u )r +rq )R (u ; p)d m h m h h h e=1 el f (6) f f;n+ + r r u d h h e=1 el f f;n+ f e (R (u ; p)ru )d m m h h e=1 el f f e r : ( R (u ; p) R (u ; p))d m m m m e=1 Z Z f f f n+ f = b (t ) d + h d h h f f where is the density of the uid and b is the body force applied on the f;n+ f;n+ des uid, represents the turbulent stress term, is the Cauchy h h stress tensor for a Newtonian uid. Here, and q denote the test functions of the uid velocity u and pressure p. In Eq. (6), the Galerkin terms for the ow equations are shown in the rst, second and third lines. The Petrov- Galerkin stabilization term for the momentum equation is shown in the fourth line and the term for the continuity is in the fth line. The remaining terms 11 represent the approximation of the ne scale velocity on the element interiors. The stabilization parameters and are added to the fully discretized m c formulation and R (u ; p) denotes the residual of the momentum equation [46]. A hybrid RANS/LES model based on the DDES treatment is applied to model the ow turbulence for high Reynolds numbers. The DDES model behaves as RANS model in the near wall region and switches to LES-mode in the separated turbulent region. Further details can be referred to [47]. 2.2 Flexible multibody solver with constraints The motion equations for a exible structure are discretized using nite ele- ment method and can be written into a weak variational form using the virtual work principle Z Z Z n+1 2 s @ d s h s s s s + (E(d )) : r d dt = h h h s s @t i i (7) Z Z Z n+1 s s s s b d + t d dt h h s s s where denotes the multibody domain, and represent the test function i h for the structural displacements and the structural density, respectively. Here, s s d is the structural displacement, is de ned as the rst Piola-Kirchho s s s stress tensor, b denotes the body force on the multibody and E(d ) is the i h simpli ed Cauchy-Green Lagrangian strain tensor. The external force caused by ow on the interface is de ned as t . The kinematic joints, like revolution joint and sphere joint, are used to connect dierent components in the exible multibody system and can be considered as multibody constrains with the following expression: c(d ) = 0: (8) 12 Generally, the discretized motion equations with constraints for exible multi- body system can be written into a matrix form Z n+1 Z n+1 t t s s s s s s (M d (t) + C d (t) + K d (t))dt = (F (t))dt (9) n n t t s s s where M , C and K represents the mass, constrain and stiness matrices for the exible multibody system respectively. All the body forces and external forces caused by ow acting on the multibody system can be combined into a whole force matrix F . The forces from constraints will not produce any work for the multibody system at the discrete solution level, which are not considered in the force matrix. The structural variables are updated via an unconditionally stable energy decaying scheme temporally, which is obtained by using a linear time dis- continuous Galerkin approximation to Eq. (9). We brie y present the general discretized motion equations for exible multibody system with dierent geo- metrically nonlinear co-rotational nite element models, including beam, ca- ble, shell and membrane. The detailed description of geometrically exact shell model in multibody dynamics can be found in [48,49]. 2.3 Aeroelastic coupling and mesh motion via radial basis functions In this section, we brie y introduce the RBF method and its applications to the multibody aeroelastic coupling and the mesh motion interpolation in our ALE formulation. In the proposed partitioned iterative coupling scheme, the interface data is exchanged between the uid domain and the multibody structural domain. Across the non-matching meshes along the aeroelastic in- terface, the surface data need to be transferred in a conservative manner with reasonable accuracy. 13 2.3.1 Review of RBF interpolation process Assume that data are interpolated from a set of control points to a set of c c c c target points. Let x = (x ; y ; z ) be the coordinate of the i-th control point, i i i i t t t t x = (x ; y ; z ) be the coordinate of the j-th target point, N and N be the c t j j j j number of control points and target points. The global interpolation function de ned by a radial basis interpolation is given by i=N g(x) = + p(x) i i i=1 i=N = (kx x k) + p(x) (10) i=1 where g(x) is the global interpolation function and x is the coordinate of an arbitrary point in space and kk denotes the Euclidean norm. The coecient represents the weights related to the i-th basis function and p(x) is a linear polynomial to recover translation and rotation motion, where its expression is p(x) = + x + y + z. 0 1 2 3 As shown in [32,33], the interpolation relationship between the control vector c t G and the target vector G can be written in the matrix form as t I I 1 c I c G = A (M ) G = H G (11) where 2 3 2 3 0 0 6 7 6 7 6 7 6 7 6 7 6 7 6 0 7 6 0 7 6 7 6 7 6 7 6 7 6 7 6 7 0 0 6 7 6 7 6 7 6 7 6 7 6 7 t c G = 6 7 ; G = 6 7 (12) 0 0 6 7 6 7 6 7 6 7 6 7 6 7 t c g(x ) g(x ) 6 7 6 7 1 1 6 7 6 7 6 7 6 7 . . . . 6 7 6 7 . . 6 7 6 7 4 5 4 5 t c g(x ) g(x ) N N t c 14 I where H denotes the nal interpolation matrix between the control vector I I and target vector. A and M are the interpolation matrices given by 2 3 0 0 0 0 1 1 1 6 7 6 7 6 7 c c c 60 0 0 0 x x x 7 1 2 N 6 7 6 7 c c c 6 7 0 0 0 0 y y y 6 7 1 2 N 6 7 6 7 c c c M = 6 7 (13) 0 0 0 0 z z z 1 2 N 6 7 6 7 c;c c;c c;c 6 7 c c c 1 x y z 6 7 1 1 1 11 12 1N 6 7 6 7 . . . . . . . . . . . . . . . 6 7 . . . . . . . 6 7 4 5 c;c c;c c;c c c c 1 x y z N N N N 1 N 2 N N c c c c c c c 2 3 t;c t;c t;c t t t 1 x y z 6 1 1 1 11 12 1N 7 6 7 . . . . . . . 6 . 7 . . . . . . . . A = 6 . 7 (14) . . . . . . . 6 7 4 5 t;c t;c t;c t t t 1 x y z N N N N 1 N 2 N N t t t t t t c c;c t;c c c t c with = ( x x ) and = ( x x ), respectively. A compactly ij i j ij i j supported Wendland's C function [32,34] has been proved as an eective ba- sis function for the interpolation with improved accuracy, which is considered in the framework. The basis function with a compact support is de ned as (kxk) = (1kxk =r) (4kxk =r + 1), where r denotes the support radius. 2.3.2 Multibody aeroelastic coupling The aeroelastic coupling involves two data transfers across the interface: (i) forces from the uid domain to the structural domain and (ii) displacements from the structural domain to the uid domain. A schematic diagram of in- terface data interpolation via RBF method is depicted in Fig. 1. The energy conservation should be satis ed in the interpolation of forces and displace- ments along the aeroelastic interface. Generally, the work W done by forces in structural domain on the interface equals to the work W done by the aerodynamic forces according to the de nition of energy conservation, which 15 Rigid Beam Shell Fluid element Multibody Structural node structure Structural Fluid node displacements Data transfer via RBF Fluid forces Fluid Fig. 1. Schematic of interface data transfer via RBF method across non-matching meshes between uid domain and multibody structure system involving rigid link, exible beam and shell members. is given as Z Z s s s s s f f f f f W (d ) = ( n ) d d = ( n ) d d = W (d ) (15) fs fs s f where d and d are the displacements along interface in the structural and uid domains, respectively. For the aeroelastic interface coupling process, the interpolation equation of displacements along interface can be written as f;I I s;I D = H D (16) ss f;I s;I where D and D are the displacement vectors along interface for the uid domain and the structural domain, respectively. Both the vectors have the same vector form as shown in Eq. (12). The matrix H interpolates the dis- ss placements from the structural surface nodes to the uid surface nodes. As the balance of energy transfer at the interface must be achieved during the interpolation process, according to Eq. (15) and Eq. (16), the interpolation 16 between the uid forces and the structural forces can be written as s;I I T f;I F = (H ) F (17) ss f;I s;I where F and F are the force vectors along the interface for the uid do- main and the structural domain, respectively. We numerically evaluate the mo- mentum ux through the boundary faces and provide the interpolated forces to the structural solver. Such an interpolation process across aeroelastic inter- face is performed at each time step in the partitioned iterative scheme with the conservation of energy transfer at the uid-structure interface. 2.3.3 Mesh motion interpolation The aeroelastic coupling involve a deformation of the uid mesh in response to a movement of the structure. The interpolation of the mesh motion using radial basis function can handle mesh motion with a relatively moderate deformation without distorting the mesh quality. Moreover, the mapping matrix used for the interface displacement interpolation can be easily extended to a three- dimensional uid domain. In our framework, the mesh motion of the uid nodes in the domain is interpolated from the uid nodes on the interface. The relationship between the displacement vector of uid nodes in the domain D f;I and the displacement vector of the uid nodes on the interface D is given as f I f;I D = H D (18) v vs s where H denotes the interpolation matrix for the uid space nodes. Note vs that, H includes the polynomial term p(x) shown in Eq. (10), where its vs value increases linearly with x. While these terms can preserve good mesh quality in the near wall region, an undesired translation and rotation motion will be interpolated to the uid nodes at the boundary of the computational 17 domain. Hence, a smooth cut-o function [33] is implemented to adjust the deformation far away from the interface. 2.4 Nonlinear interface force correction scheme Nonlinear interface force correction scheme [45] can provide a numerical stabil- ity for the partitioned aeroelastic coupling when signi cant added mass eect is encountered in an aeroelastic simulation. This scheme has been extended to the exible multibody aeroelastic coupling simulation to achieve numeri- cal stability and the iterative force correction procedure is brie y summarized herein. The formulation for a coupled linear system between the uid domain and the structure domain which is discretized by the nite element method can be written into the matrix form AU = R, where U denotes the vector of the unknowns for the coupled system and R represents the right-hand side. The abstract matrix form can be written as 8 9 8 9 2 3 > > > > > > > > A 0 0 A d R > > > > 11 14 1 6 7 > > > > > > > > 6 7 > > > > > > > > 6 7 > > > > < = < = 6A A 0 0 7 d R 21 22 2 6 7 = (19) 6 7 > > > > 6 7 > > > > 0 A A 0 >q > >R > 32 33 3 6 7 > > > > > > > > 4 5 > > > > > > > > > > > > : I; : ; 0 0 A A f R 43 44 4 where A denotes the mass and stiness matrices of exible multibody system and A is the load vector on the solid surface. A represents the 14 21 transformation matrix which maps the structural displacements to aeroelastic interface and A denotes the force calculation and transform to the interface. A and A are the identity matrices. A gives the relationship between the 22 44 32 displacement on interface and the ALE mapping in the uid domain. A is the coupled uid velocity and pressure linear system. d denotes the structural displacement for the exible multibody system and d represents the displace- ment along the coupling interface. q is the unknown variables in uid domain 18 I and f is the force on interface. The right-hand side vector R shows the resid- ual of this linear system for the dierent equations. In this matrix form, the rst equation is the exible multibody system and the third equation is re- lated to the uid and turbulence equations. The second and fourth equations correspond to the displacement and traction transformation between the uid domain and structure domain, respectively. The o-diagonal term A can be eliminated through the static condensation and Eq. (19) can be rewritten as 2 3 8 9 8 9 > s> > > > > > > A 0 0 0 d R 11 > > > 1> 6 7 > > > > > > > > 6 7 > > > > > > > > 6 7 > > > > < = < = A A 0 0 d R 6 7 21 22 2 6 7 = (20) 6 7 > > > > 6 7 > > > > 0 A A 0 >q > >R > 32 33 3 6 7 > > > > > > > > > > > > 4 5 > > > > > > > > : I; : ; ~ ~ 0 0 0 A f R 44 4 1 1 1 1 ~ ~ where A = A A A A A A A A and R = R A A (R 44 44 43 33 32 22 21 11 14 4 4 43 33 3 1 1 A A (R A A R )). An iterative scheme is used to correct the forces 32 2 21 1 22 11 between the uid domain and structure domain with a feedback process. The formulation of nonlinear iterative force correction is given as I I 1 ~ ~ f = f + A R (21) 4(k) (k+1) (k) 44 where k denotes the nonlinear sub-iteration with a single time step. The cur- rent nonlinear iterative force correction process can be considered as a gen- eralization of Aitken's extrapolation while updating a dynamic stabilization parameter, which could transform a divergent xed-point iteration to a con- vergent and stable force correction [45]. Finally, a numerical stability can be achieved with the aid of the NIFC scheme for the partitioned uid- exible multibody system coupling. 19 2.5 Fluid- exible multibody coupling procedure The partitioned iterative coupling procedure of the uid solver with the ex- ible multibody structural solver is brie y summarized in this section. The schematic of aeroelastic coupling procedure based on a predictor-corrector al- gorithm is shown in Fig. 2. The predictor displacement is calculated from the exible multibody equations and the ALE uid equations can be solved to provide the forces on the uid-structure interface as a corrector step. The structural displacement caused by the aerodynamic forces at time t is de- s s n ned as d (x ; t ). Firstly, the structural displacements of each component in the exible multibody system is solved via the time discontinuous Galerkin approximation under the loads from the uid domain at time t . Then the structural displacements obtained from the previous step are transferred to the uid solver and it is satis ed the ALE compatibility and the velocity con- fs tinuity on the interface . To satisfy the consistency of the non-matching uid and structural mesh con gurations, the mesh displacements are set equal to the structural displacements along the interface as m;n+1 s fs d = d on (22) m;n+1 n+1 where d represents the mesh displacement at time t . Meanwhile, the fs velocity continuity on interface is satis ed when the uid velocity equals to the mesh velocity. m;n+1 m;n f f d d f;n+ m;n+ fs u = u = on (23) In the third step, the Navier-Stokes equations under the ALE framework with the turbulence model equations are solved and compute the uid forces. Fi- nally, the obtained uid forces are corrected based on NIFC scheme to achieve the numerical stability then transferred to the exible multibody solver. One 20 aeroelastic sub-iteration is nished and the sub-iteration mentioned above will be executed continuously until the convergence criterion has achieved. Sub- sequently, all the variables in this aeroelastic solver are updated for the next n+1 time step t . The Generalized Minimal RESidual (GMRES) algorithm is applied in the cur- rent framework to compute the velocity, pressure and mesh displacement in uid equations discretized with nite element method. This algorithm relies on the Krylov subspace iteration and the modi ed Gram-Schmidt orthog- onalization. In order to minimize the linearization error at each time step, the Newton-Raphson scheme is considered for the aeroelastic framework. Fur- thermore, the discretized algebraic equations based on the exible multibody system with the co-rotational framework are solved via a classical skyline solver. 3 Mesh convergence and validation Before we demonstrate our multibody aeroelastic framework for a bat-like wing, a mesh convergence study is rst conducted to ensure a sucient mesh resolution employed on both the uid and structure domains. We validate our multibody aeroelastic solver against the experiment work of [50] by simulating a exible apping wing with an anisotropic material, where the structural responses are compared to the experimental data, and the vortex patterns are analyzed. 21 f f n+1 u ¯ (x , t ), f f n+1 s n+1 p ¯ (x , t ), d (x , t ) f f n+1 ν ˜ (x , t ) k = nIter k = 2 I I I f = f + δf k+1 k (4) (3) (2) (1) k = 1 f f n u ¯ (x , t ), s n f f n p ¯ (x , t ), d (x , t ) f f n ν ˜ (x , t ) f fs Ω Γ Fig. 2. A schematic of predictor-corrector procedure for ALE uid solver and exible multibody solver coupling via nonlinear iterative force correction. (1) Solve exible multibody system with constraints, (2) Transfer predicted structural displacement, (3) Solve ALE uid and turbulence equations, (4) Correct forces via NIFC lter. nIter represents the maximum number of nonlinear iterations to achieve a desired n n+1 convergence tolerance in a time step at t 2 [t ; t ]. 3.1 Mesh convergence The mesh convergence study is conducted by simulating a wing with isotropic material using three dierent mesh resolutions on the uid and structure do- mains. The wing simulated is adopted from an experimental study [50], and its properties are summarized in Table 1. It is an isotropic aluminum wing with Zimmerman shape, which is designed to investigate the aerodynamic charac- teristics under apping ight condition. A schematic diagram of the geometry of the wing and its surrounding aeroelastic computational domain are shown in f s u ¯ = u ALE ﬂuid & turbulence Flexible multibody wing 75mm Leading edge Fixed square 25mm Wing tip Trailing edge (a) wing 60C in out 80C 60C (b) Fig. 3. Flow past a apping wing: (a) geometry of isotropic wing, (b) schematic of computational setup. Fig. 3. The root chord length C is 25 mm and the span length is 75 mm, which results in an aspect ratio of 7.65. The apping condition is hovering motion with 10 Hz apping frequency and 21 apping amplitude. The reference ow velocity is the velocity measured at the wing tip and the freestream velocity is given as zero. The distances of the inlet ( ) and outlet ( ) boundaries in out from the leading root of the wing are both 30C . The distance between the side walls ( ) on top and bottom is 60C and the distance increases to 80C for slip the side walls on both sides. The components of the ow velocity are de ned f f f f as u = (u ; v ; w ). The freestream velocity along the X -axis at the inlet 23 f boundary is given as u = U . Slip boundary condition is applied on the in @u top and bottom boundaries ( ), where = 0 and w = 0. Slip bound- slip @z @u ary condition is applied on both side boundaries ( ), where = 0 and slip @y v = 0. A traction-free boundary condition is de ned at the outlet boundary , where = = = 0. The boundary condition on apping wing out xx yx zx surfaces is no-slip boundary condition. The three-dimensional uid computational domain is discretized by unstruc- tured eight-node brick nite element mesh and the structural computational domain is discretized by structured four-node rectangular nite element mesh. A summary of three dierent mesh resolutions is shown in Table 2. A boundary layer with y 0:5 in the wall-normal direction is maintained, with a stretch- ing ratio, y =y of 1.2. The discretization along chord direction, span-wise i+1 i direction and outside the boundary layer is varied. The non-dimensional time step size, tU =C is chosen as 0.01. Details of mesh statistics in space and ref on surface for the uid domain with M2 are shown in Fig. 4a and 4b, respec- tively. A nite element mesh on structural surface with M2 is shown in Fig. 4c. For the purpose of mesh convergence investigation, the amplitude of lift coef- cients for three dierent meshes are calculated and compared with literature data, which are shown in Table 3. The results indicate that the gap between M1 and M3 is within 1% and it reduces to 0.15% for M2 and M3. It is con- cluded that M2 has achieved mesh convergence and it can be used as the reference case to compare with experiment data. The comparison of lift coe- cient histories in one cycle is displayed in Fig. 5a and the simulated result for M2 shows a good comparison with the result of [51]. The normalized displace- ment at wing tip is compared in Fig. 5b with that of the experiment [50]. It can be seen that the normalized displacement at wing tip in one non-dimensional cycle has a good match with the experimental measurements. 24 (a) Fluid space mesh (b) Fluid surface mesh (c) Structural surface mesh Fig. 4. Schematic of mesh characteristics: (a) in space for apping wing in the uid domain (M2), (b) on surface for apping wing in the uid domain (M2), (c) on surface for apping wing in the structural domain (M2). The velocity magnitude contour, X -vorticity contour for a slice at the quarter position along chord-wise and Y -vorticity contour for a slice at the middle position along span-wise at both t=T = 0:3 and t=T = 0:48 are shown in Fig. 6, respectively. It can be seen from the Y -vorticity contours that a pair of main vortices is generated at the leading edge and trailing edge during the hovering motion. The normalized velocity magnitude distributions along vertical direction of the apping wing at wing tip and mid-span for both time instants t=T = 0:3 and t=T = 0:48 are extracted from the ow eld and compared with those obtained from [50] and [51], which are shown in Fig. 7. The experiment result is displayed with 95% errorbars of the instantaneous values. The normalized velocity magnitude distributions simulated via our aeroelastic framework show similar trends with those in literature and the experiment. 25 Table 1 Aeroelastic parameters for an isotropic aluminum wing under hovering motion Parameters Value Semi-span at quarter chord 0.075 m Chord length at wing root 0.025 m Structural thickness 0.0004 m Poissons ratio 0.3 Material density 2700 kg=m Youngs modulus of material 70.0 GPa Reference ow velocity (hovering) 1.0995 m/s Air density 1.209 kg=m Mean chord-based Reynolds number 2605 Flapping frequency 10 Hz Flapping amplitude 21 Aspect ratio 7.65 Table 2 Mesh statistics for an isotropic wing under hovering motion Mesh Fluid nodes Fluid elements Structural nodes Structural elements M1 509,082 492,630 114 90 M2 816,312 795,614 182 150 M3 1,311,120 1,284,226 506 450 The instantaneous turbulent wake elds of the complex three-dimensional ow around the apping wing at four various time instances with t=T = 0, 0.25, 0.5 and 0.75 are displayed in Fig. 8. The non-dimensional Q-criterion at isosur- face of 1 is used to depict the detailed and complex vortex structures around apping wing and it is colored by the normalized velocity magnitude during hovering motion. 26 Table 3 Mesh convergence study for the lift coecient C . The percentage dierences are computed based on M3 result. Results Amplitude of lift coecient C Present (M1) 6.65 (0.61%) Present (M2) 6.6 (-0.15%) Present (M3) 6.61 Aono et al. 6.24 1.5 Present (M2) Present (M2) Aono et al. (2010) Experiment 0.5 -0.5 -2 -1 -4 -1.5 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 t/T t/T (a) (b) Fig. 5. Comparisons of time traces of the isotropic apping wing: (a) lift coecient: (4) Present simulation, () Aono et al. [51], (b) normalized displacement at wing tip: (4) Present simulation, () Experiment [50]. 3.2 Validation of multibody aeroelastic framework A multibody exible wing adopted from the experiment [50] with composite material apping in a vacuum is rst simulated to validate the exible multi- body structural solver. The wing apping in air ow is considered for our validation purpose of the proposed aeroelastic solver. This wing is designed to mimic a real hummingbird wing based on Zimmerman planform and the membrane with Capran material is supported by several skeletons made of unidirectional carbon bers. The schematic of this anisotropic wing for a de- tailed geometric information is depicted in Fig. 9a. A rigid triangle made of δ /C z,tip (a) t=T = 0:3 (b) t=T = 0:48 (c) t=T = 0:3 (d) t=T = 0:48 (e) t=T = 0:3 (f ) t=T = 0:48 Fig. 6. Flow elds around the isotropic apping wing in uniform ow: velocity magnitude at (a) t=T = 0:3, (b) t=T = 0:48, X -vorticity at (c) t=T = 0:3, (d) t=T = 0:48, and Y -vorticity contours at (e) t=T = 0:3, (f ) t=T = 0:48. 28 4 2.5 Present (M2) Present (M2) Aono et al. (2010) Aono et al. (2010) 2 Experiment with 95% errorbars Experiment with 95% errorbars 1.5 0.5 -1 -0.5 -3 -2 -1 0 1 -3 -2 -1 0 1 Z/C Z/C (a) t=T = 0:3 at mid-span (b) t=T = 0:3 at wing tip 2.5 5 Present (M2) Present (M2) Aono et al. (2010) Aono et al. (2010) Experiment with 95% errorbars Experiment with 95% errorbars 1.5 3 1 2 0.5 1 0 0 -1 -0.5 -3 -2 -1 0 1 -3 -2 -1 0 1 Z/C Z/C (c) t=T = 0:48 at mid-span (d) t=T = 0:48 at wing tip Fig. 7. Flow past an isotropic appig wing: Comparison of instantaneous velocity magnitude for isotropic wing at various time instants: (a) t=T = 0:3 at mid-span, (b) t=T = 0:3 at wing tip, (c) t=T = 0:48 at mid-span, (d) t=T = 0:48 at wing tip. (4) Present simulation, () Aono et al. [51], () Experiment with 95% errorbars [50]. three layers of bidirectional carbon ber is used to mount the wing during the apping deformation and it is inserted at the corner of the wing root and the leading edge. A schematic of the topological layout is presented in Fig. 9c. This anisotropic wing is reinforced at the leading edge, the wing root and the surface of membrane with dierent layers of carbon ber strips of 0.8 mm width. Dierent wings with speci c layout schemes are termed LiBj, where L represents the leading edge, B stands for the batten and i, j denotes the kUk /U kUk /U ref ref kUk /U ref kUk /U ref (a) t=T = 0 (b) t=T = 0:25 (c) t=T = 0:5 (d) t=T = 0:75 kUk /U : 0 0.2 0.4 0.6 0.8 1 1.2 1.4 ref Fig. 8. Flow past an isotropic apping wing: Wake structures based on the instan- @u @u 1 j taneous iso-surfaces of Q(= ) value at (a) t=T = 0, (b) t=T = 0:25, (c) 2 @x @x j i + 2 t=T = 0:5, (d) t=T = 0:75. Iso-surfaces of non-dimensional Q Q(C=U ) = 1 ref are colored by the normalized velocity magnitude kUk =U . ref number of layers in the leading edge and the batten, respectively. The wing root is reinforced with 2 carbon ber layers for all wings. L2B1 and L3B1 models are considered to validate the high- delity exible multibody solver in our proposed aeroelastic framework. For L2B1 and L3B1 models, the leading edge is reinforced with two and three layers, respectively. 30 Each batten is reinforced with 1 layer for both models. Details of mate- rial properties of the composite skeleton and the wing membrane for the anisotropic wing with L2B1 and L3B1 models are summarized in Table 4. Some structural material parameters are adjusted to satisfy the structural mode frequencies and shapes obtained from experiments according to the cor- rection work in [52]. Geometrically exact co-rotational shell elements with anisotropic material are employed for skeletons and shell elements with ho- mogeneous isotropic material are used for the wing membrane. Dierent nite elements are connected with their adjacent elements via kinematic constraints. The structural model is discretized by 315 structured four-node rectangular nite elements and the detailed mesh characteristic is given in Fig. 9b. The rst six order natural frequencies are analyzed for L2B1 and L3B1 models with speci c material properties. The comparison with frequencies obtained from the experiment [50] and literature [52] shows that the present results have good agreement, which is given in Table 5. In the experiment, this anisotropic wing is given a prescribed rotation apping motion along wing root direction at the mounting rigid triangle in a vacuum. The apping amplitude is 35 and the apping frequency is set to 25 Hz. The time transient structural responses at wing tip is measured and compared with those obtained from experiment [50]. Here denotes the actual location in the vertical direction of wing z;tip tip at computational coordinate and represents the distance between w;tip deformed actual wing tip and the undeformed reference plane. The comparison of the initial reference plane, the undeformed reference plane and the deformed actual plane is depicted in Fig. 9d. Comparisons of time histories for the normalized location in the vertical di- rection and the normalized displacement at the wing tip for L2B1 and L3B1 models undergoing prescribed rotational motion in a vacuum are shown in Fig. 10, respectively. The results indicate that the high- delity exible multi- body solver can simulate the nonlinear structural responses of the anisotropic 31 Table 4 Material properties of the composite skeleton and wing membrane for the anisotropic wing (L2B1 and L3B1 models) [52] Component Material properties E =233 GPa E =23.7 GPa G =10.5 GPa G = G =1.7 GPa 23 31 Carbon ber prepreg (Properties of one v =0.05 layer) v = v =0.32 23 31 =1740 kg=m Thickness=0.1 mm E=2.76 GPa v =0.489 Capran membrane 12 (From experiments) 3 =1384 kg=m Thickness=0.015 mm wing with multiple components and shows reasonable agreement with the date obtained from the experiment. In our simulation results, the wing tip defor- mation is varied in dierent cycles, which is similar to the conclusion obtained in [52]. The structural responses are averaged over several cycles, according to the periodic assumption in experimental measurements. One possible reason causing discrepancies in the comparison is some uncertain factors connected with the actuation mechanism [50]. The instantaneous structural displacement contours of L2B1 model for four time instants in a whole apping cycle are given in Fig. 11. The anisotropic wing shows a relatively large elastic defor- mation and the wing twist due to the high exibility of the wing material. In the experiment, the anisotropic wing apping in air condition is also per- formed to investigate the aeroelastic eect on the apping motion. Therefore, an anisotropic wing with L3B1 model is simulated by using the proposed 32 Table 5 Comparison of the natural frequencies for anisotropic wing Natural frequencies (Hz) Model Result st nd rd th th th 1 2 3 4 5 6 Present 45.03 73.16 76.39 94.45 106.35 116.31 L2B1 Gogulapati A. 47.00 72.00 76.50 88.00 109.00 118.80 Experiment 42.00 84.00 126.00 Present 62.71 76.13 79.55 106.31 111.00 119.39 L3B1 Gogulapati A. 65.00 75.50 76.80 107.00 109.00 120.00 Experiment 59.00 104.00 138.00 Rigid triangle Leadning edge Batten 25mm Root Trailing edge 9.375mm 18.75mm 18.75mm 18.75mm 9.375mm (a) (b) Number of layers for leading edge layup L3B1 0.12g L2B1 0.10g L1B1 0.08g Number of layers for batten layup L3B2 0.13g L2B2 0.11g L1B2 0.09g (c) (d) Fig. 9. Problem set-up for anisotropic wing con guration: (a) geometry information, (b) nite element representation, (c) the topological layout [50], (d) initial reference plane, undeformed reference plane and deformed actual plane. 33 3 Present Present Experiment 2 Experiment -1 -1 -2 -3 -2 -50 -25 0 25 50 -50 -25 0 25 50 α ( ) α ( ) (a) L2B1 (b) L2B1 Present Present Experiment Experiment 0.5 -1 -0.5 -2 -1 -3 -50 -25 0 25 50 -50 -25 0 25 50 α ( ) α ( ) (c) L3B1 (d) L3B1 Fig. 10. Comparison of structural responses for anisotropic wing: (a) normalized location in vertical direction of L2B1 model at wing tip, (b) normalized displace- ment of L2B1 model at wing tip, (c) normalized location of L3B1 model in vertical direction at wing tip, (d) normalized displacement of L3B1 model at wing tip. (4) Present simulation, () Experiment [50]. aeroelastic framework for validation purpose. In the experiment, the material properties of the anisotropic wing are summarized in Table 4 and a hov- ering motion is employed with zero freestream velocity. Detailed simulation parameters for the aeroelastic simulation are given in Table 6. Considering the similar simulation condition as the isotropic apping wing, a similar aeroelas- tic computational domain is shown in Fig. 3b. Identical boundary conditions for the computational domain of anisotropic wing are applied as those of the δ /C δ /C z,tip z,tip δ /C w,tip δ /C w,tip (a) (b) (c) (d) δ : -0.03 -0.021 -0.012 -0.003 0.006 0.015 0.024 Fig. 11. Structural displacement contours of L2B1 model at various time instants: (a) t=T = 0:19 with apping angle 32:54 , (b) t=T = 0:34 with apping angle 29:55 , (c) t=T = 0:55 with apping angle 10:50 , (d) t=T = 0:69 with apping angle 32:82 . isotropic wing con guration. A similar mesh distribution is adopted for the three-dimensional uid computational domain with 492,630 unstructured - nite elements. The structural mesh is same as shown in Fig. 9b with 315 structured nite elements. The distance between the rst grid point in the 4 + boundary layer and wing surface is set as 9.2810 C with y 0:5 and the number of divisions along the vertical direction of the wing surface is 25 and stretching ratio, y =y is 1.2. The non-dimensional time step size, i+1 i tU =C is set to 0.0018. ref For the purpose of validation, the comparisons of normalized location in ver- tical direction and normalized displacement at the wing tip are given in Fig. 12, respectively. Results indicate that the overall trends of the apping wing are well predicted, compared with the experimental data. Fig. 13 graphically 35 Table 6 Aeroelastic parameters for an anisotropic wing under hovering motion Parameters Value Reference velocity (hovering) 4.58 m/s Air density 1.206 kg=m Reynolds number 7304 Flapping frequency 25 Hz Flapping amplitude 35 3 1.5 Present Present Experiment Experiment 2 1 1 0.5 0 0 -0.5 -1 -2 -1 -3 -1.5 -50 -25 0 25 50 -50 -25 0 25 50 ◦ ◦ α ( ) α ( ) (a) (b) Fig. 12. Comparison of coupled aeroelastic responses for anisotropic apping wing: (a) normalized location of L3B1 model in vertical direction at wing tip, (b) normal- ized displacement of L3B1 model at wing tip. (4) Present simulation, () Experi- ment [50]. shows the iso-surface of the non-dimensional Q-criterion of 1 for L3B1 model colored by the normalized velocity magnitude during a apping motion pe- riod. Detailed turbulent wake structures and their evolution process can be ob- served. Overall, the proposed uid- exible multibody solver is able to simulate exible apping wing with multibody components. The detailed aerodynamic characteristics around the apping wing and nonlinear structural responses are captured accurately, and compared well with the available experimental data. δ /C z,tip δ /C w,tip (a) t=T = 0 (b) t=T = 0:25 (c) t=T = 0:5 (d) t=T = 0:75 kUk /U : 0 0.2 0.4 0.6 0.8 1 1.2 1.4 ref Fig. 13. Flow past an anisotropic apping wing: Wake structures of L3B1 model @u @u 1 j based on the instantaneous iso-surfaces of Q(= ) value at (a) t=T = 0, 2 @x @x j i (b) t=T = 0:25, (c) t=T = 0:5, (d) t=T = 0:75. Iso-surfaces of non-dimensional + 2 Q Q(C=U ) = 1 are colored by the normalized velocity magnitude kUk =U . ref ref 4 Application to bat-like apping dynamics As mentioned earlier, the ight patterns and the mechanism of bats are quite dierent from birds and insects. The lack of understanding about how bats y eciently prompts researchers to investigate via experiments and CFD 37 simulations. However, the complex ight kinematics, the special wing struc- tures comprising skeletons with multi-degree of freedom and highly exible wing-membrane skin as well as the signi cant aeroelastic phenomena during ight prevent the research progresses for bat ight. The presently available results about the bat apping dynamics are quite limited, especially from fully-coupled computational modeling. For the rst time, we demonstrate the capability of our fully-coupled multibody aeroelastic framework to simulate a bat-like exible wing. The eects of exibility and aerodynamic load are investigated to explore their impact on the dynamics of the wing. 4.1 Structural model setup A bat-like wing made of several wing skin-like membranes and reinforced skele- tons is constructed to investigate the aeroelastic responses with dierent ight patterns. We consider the bat-like wing but ignore its body and other compo- nents. A schematic of the constructed bat-like wing is shown in Fig. 14a. The basic, simpli ed shape of the bat-like wing is referred to a apping bat robot made in Brown University [5]. In our simulation, the chord at the wing root is 0.27 m and the span is 0.69 m, resulting in a wing area of 0.124 m . A rigid strip is applied at the wing root to x the wing like the bat body. Two strips near to the wing root represent the humerus bone and radius bone, respectively. Several strips in the outer region stand for bat ngers. All these components are used to support the covered wing membranes. The detailed geometric sizes of these bones and ngers are designed to approach those of a real bat, based on the research work of P. Watts [53]. The widths and thicknesses of these bones and ngers are varied along their axial direction in order to adjust and achieve reasonable stiness distributions to support the wing membranes. Meanwhile, the thickness of the wing membrane becomes thinner in the outer region than that near to the wing root, which allows 38 Table 7 Material properties of the skeletons and wing membrane for the bat-like wing (! represents the variation of thickness along axial direction) Components Material properties M1/M2/M5 M3/M4 B1/B2 B3/B4 B5 B6 Young's modulus (MPa) 4 3 5000 3000 2500 2000 Poisson ratio 0.3 0.3 0.3 0.3 0.3 0.3 Density (kg=m ) 1100 1100 2200 2200 2200 2200 Thickness (mm) 0.8/0.6/0.4 0.2 5/4 2!1.5 3!2 2 0.69m M1 M3 Fingers Humerus bone B5 B2 B1 B3 B4 M4 0.27m Radius bone B6 M2 M5 Rigid wing root Wing membrane (a) (b) Fig. 14. Schematic of bat-like wing: (a) geometric information, (b) indication of dierent components. a relative larger elastic deformation in the outer region. The indication of six bones and ve membranes in this wing is provided in Fig. 14b. All the components in this bat-like wing are assumed to be made of homogeneous, isotropic material. Material properties of the skeletons and wing membranes in dierent wing regions are summarized in Table 7. 4.2 Eect of exibility To investigate the eect of exibility on the dynamics of a bat-like exible wing, a rigid wing with identical geometry is simulated and compared to its exible counterpart. To eliminate the eect of apping motion, the bat-like wing in a gliding ight is considered. The gliding ight is simulated by applying a uniform ow to the wing while constraining its root in all six degree-of- 39 freedom. Structural deformation due to aerodynamic load is allowed for the exible wing, but not allowed for the rigid wing. The detailed parameters for the simulation are summarized in Table 8. A schematic diagram of the bat-like wing setup is depicted in Fig. 15a. The length of chord at the wing root is de ned as C . The distances between the inlet boundary ( ) and in the outlet boundary ( ), the top and bottom boundaries ( ) and the out slip boundaries on both sides are the same of 200C . The no-slip Dirichlet boundary condition is applied on the surface of bat-like wing. The freestream velocity f f along the X-axis at the inlet boundary is de ned as u = U , where u in f f f f is the X-component of the ow velocity u = (u ; v ; w ). The slip boundary condition is applied on the top and bottom as well as the side boundaries. A traction-free boundary condition is de ned at the outlet boundary , where out = = = 0. xx yx zx The three-dimensional uid computational domain is discretized by 454,258 unstructured eight-node brick nite element meshes. A boundary layer is main- tained around the wing such that y < 1:0 in the wall-normal direction. Mesh distribution slice in the uid domain at mid-span is shown in Fig. 15b. Geo- metrically exact co-rotational shell elements are employed for wing membrane and bones. The structural model is discretized by 496 structured four-node rectangular nite elements. Mesh characteristics on the wing surface in both the uid domain and the structural domain are compared and given in Fig. 16a and 16b, respectively. The non-dimensional time step size is selected as tU=C = 0:022. Comparison of the deformed exible wing and rigid wing colored by non- dimensional displacement in the vertical direction is shown in Fig. 17a. The maximum displacement is observed at the wing tip and the covered mem- branes have some small wrinkles on the surface. Comparison of the mean lift coecient (C ) and the mean drag coecient (C ) for the rigid and exi- L D ble wing is presented in Table 9. C decreases from 0.6234 to 0.5746 and C L D 40 Table 8 Simulation parameters for a bat-like wing under gliding motion Parameters Value Semi-span length 0.69 m Chord length at wing root 0.27 m Wing area 0.124 m Freestream velocity 2.0 m/s Angle of attack 20 Air density 1.225 kg=m Mean chord-based Reynolds number 24609 reduces by 13.17% due to the passive deformation of exible wing and its compliant membrane components. As a result, the lift-to-drag ratio of exible wing (C /C ) increases by 6.13%, compared with the rigid wing. L D For further investigation, wake structures for both rigid and exible wings based on instantaneous iso-surfaces of non-dimensional Q-criterion of 0.25 col- ored by the normalized velocity magnitude are given in Fig. 17b. The shedding vortex structures behind the bat-like wing are changed by the wing deforma- tion via aeroelastic coupling. Fig. 18 provides streamlines around the rigid and exible wings colored by the normalized velocity magnitude. A massive sepa- ration ow on the upper surface is observed and the ow patterns are altered by the wing deformation, compared with the rigid wing counterpart. In sum- mary, the exibility of the wing signi cantly aects the dynamics of a bat-like wing. Further investigation can be conducted by simulating a range of exibil- ities to quantitatively investigate its eect on the aerodynamic performance of the wing. 41 wing 200C in Γ out 200C 200C (a) (b) Fig. 15. Flow past a bat-like apping wing: (a) schematic of computational setup, (b) mesh distribution slice in the uid domain at mid-span. Table 9 Comparison of mean lift coecient (C ), mean drag coecient (C ) and L D lift-to-drag ratio (C /C ) for rigid and exible wing L D Results C C C /C L D L D Rigid wing 0.6234 0.2391 2.61 Flexible wing 0.5746 (7.83 % #) 0.2076 (13.17% #) 2.77 (6.13 % ") 42 (a) (b) Fig. 16. Mesh characteristics on bat-like wing surface in (a) the uid domain, (b) the structural domain. (a) (b) Fig. 17. (a) Comparison of non-dimensional displacement contours in the vertical direction for both rigid and exible wings. (b) Wake structures of bat-like wing @u @u 1 j based on the instantaneous iso-surfaces of Q(= ) value for both rigid and 2 @x @x j i + 2 exible wings. Iso-surfaces of non-dimensional Q Q(C=U ) = 0:25 are colored ref by the normalized velocity magnitude kUk =U . ref 4.3 Eect of aerodynamic load In this section, we investigate the eect of aerodynamic load on the structural responses of the apping wing. Both uncoupled and coupled aeroelastic cases are simulated herein. The uncoupled case does not consider aerodynamic load in the apping motion, while the coupled aeroelastic case includes the cou- pling with surrounding air ow. The angle of attack is assumed as zero and the freestream velocity is set as 1.0 m/s for the apping ight. The apping amplitude is given as 20 and the apping frequency is set as 1 Hz. The iden- 43 (a) Rigid wing (b) Flexible wing kUk /U : 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 ref Fig. 18. Streamlines colored by the normalized velocity magnitude kUk =U around ref the bat-like wing for (a) rigid wing, (b) exible wing. tical mesh distributions in Section 4.2 for both uid and structural models are adopted for the apping wing simulation. The non-dimensional time step size is selected as tU=C = 0:00926. The comparisons of the normalized location and the displacement in the verti- cal direction at wing tip for both cases are shown in Fig. 19, respectively. The nonlinear dynamic responses for the exible bat-like wing have been aected by the aeroelastic coupling eect signi cantly and the lagging eect of coupled case becomes stronger than the uncoupled one. The iso-surfaces of the non- dimensional Q-criterion of 0.25 for the bat-like wing colored by the normalized velocity magnitude during a whole apping motion period are given in Fig. 20. The evolution of vortex generated from the leading edge, the trailing edge and the wing tip during apping ight can be observed. For a detailed analysis, streamlines around this wing at t=T = 0:25 and t=T = 0:75 are shown in Fig. 21, respectively. A massive separation ow on upper surface near to the wing tip is viewed at the downstroke motion. Such complex, nonlinear aerodynamic phenomena, including leading edge vortex (LEV) generation and trailing edge vortex (TEV) shedding, in uence the structural responses via the aeroelastic 44 1 0.4 Uncoupled Uncoupled Coupled Coupled 0.5 0.2 -0.5 -0.2 -1 -0.4 -20 -10 0 10 20 -20 -10 0 10 20 α ( ) α ( ) (a) (b) Fig. 19. Comparison of uncoupled and coupled aeroelastic structural responses of the bat-like wing for (a) normalized location in the vertical direction at wing tip, (b) normalized displacement at wing tip. coupling process. The eect of aerodynamic load on the bat-like apping wing is quite substantial. According to the discussion presented above, the proposed multibody aeroelas- tic solver is able to capture the physics of aeroelastic phenomena of the exible multibody apping wing and it can be extended to a more general bat-like wing. Meanwhile, the implementation of RBF method in current framework and combination with NIFC approach have been proved as an eective scheme to simulate exible apping wing. Finally, a bat-like wing with dierent exi- bilities for the bones and the membranes and the apping frequencies should be explored for further mechanism investigation based on the proposed aeroe- lastic framework. 5 Conclusion In this paper, a three-dimensional multibody aeroelastic framework is devel- oped by coupling an incompressible turbulent ow solver with DDES method and a exible multibody structural solver discretized by geometrically exact δ /C z,tip δ /C w,tip (a) t=T = 0 (b) t=T = 0:25 (c) t=T = 0:5 (d) t=T = 0:75 kUk /U : 0 0.2 0.4 0.6 0.8 1 1.2 1.4 ref Fig. 20. Flow past a bat-like wing: Wake structures of bat-like wing based on the in- @u @u 1 j stantaneous iso-surfaces of Q(= ) value at (a) t=T = 0, (b) t=T = 0:25, (c) 2 @x @x j i + 2 t=T = 0:5, (d) t=T = 0:75. Iso-surfaces of non-dimensional Q Q(C=U ) = 0:25 ref are colored by the normalized velocity magnitude kUk =U . ref co-rotational nite elements via partitioned domain decomposition strategy. The nonlinear iterative force correction approach has been implemented and coupled with the RBF interpolation method during the integration of incom- pressible turbulent ow and a exible multibody system to achieve a stable and robust partitioned coupling process. The proposed aeroelastic framework provides a feasible and high- delity tool to explore the design and optimization of a wide range of exible apping wings, such as bionic MAVs and UAVs. We validated the accuracy of our framework by simulating an anisotropic wing made of multiple reinforced battens and membranes. A good agreement to 46 (a) t=T = 0:25 (b) t=T = 0:75 kUk /U : 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 ref Fig. 21. Streamlines colored by the normalized velocity magnitude kUk =U around ref the bat-like wing at (a) t=T = 0:25, (b) t=T = 0:75. an experiment has been achieved, where the characteristic responses of the apping wing compared quite well with the experimental data. We further demonstrated the applicability of our framework by investigating the aeroe- lastic responses of a bat-like wing. The rigid and exible wings under gliding ight simulations were carried out and the eect of exibility on the dynamics of the bat-like wing was quanti ed. The lift-to-drag ratio increases by 6.13% for the exible wing due to the passive deformation of the wing and its com- pliant membranes, compared to its rigid counterpart. Subsequently, the eect of aerodynamic load has been studied by simulating a coupled and an uncou- pled aeroelastic cases. It is observed that the aerodynamic load enhances the lagging eect of the nonlinear structural responses signi cantly. In future, var- ious aspects of a bat ight can be investigated via the developed framework, which may include structural optimization, acoustic reduction, ight control, and among others. 47 Acknowledgements The authors wish to acknowledge supports from the National University of Singapore and the Ministry of Education, Singapore. References [1] K. Jones, M. Platzer, An experimental and numerical investigation of apping- wing propulsion, in: 37th Aerospace Sciences Meeting and Exhibit, 1999, p. [2] B. Singh, I. Chopra, Insect-based hover-capable apping wings for micro air vehicles: Experiments and analysis, Aiaa Journal 46 (46) (2008) 2115{2135. [3] S. Heathcote, D. Martin, I. Gursul, Flexible apping airfoil propulsion at zero freestream velocity, AIAA journal 42 (11) (2004) 2196{2204. [4] M. Hamamoto, Y. Ohta, K. Hara, T. Hisada, Application of uid{structure interaction analysis to apping ight of insects with deformable wings, Advanced Robotics 21 (1-2) (2007) 1{21. [5] J. W. Bahlman, S. M. Swartz, K. S. Breuer, Design and characterization of a multi-articulated robotic bat wing, Bioinspiration & biomimetics 8 (1) (2013) [6] S. Wei, M. Berg, D. Ljungqvist, Flapping and exible wings for biological and micro air vehicles, Progress in Aerospace Sciences 35 (5) (1999) 455{505. [7] K. V. Rozhdestvensky, V. A. Ryzhov, Aerohydrodynamics of apping-wing propulsors, Progress in Aerospace Sciences 39 (8) (2003) 585{633. [8] M. S. Triantafyllou, A. H. Techet, F. S. Hover, Review of experimental work in biomimetic foils, IEEE Journal of Oceanic Engineering 29 (3) (2004) 585{594. [9] M. F. Platzer, K. D. Jones, J. Young, J. S. Lai, Flapping wing aerodynamics: progress and challenges, AIAA journal 46 (9) (2008) 2136{2149. 48 [10] W. Shyy, H. Aono, S. K. Chimakurthi, P. Trizila, C. K. Kang, C. E. S. Cesnik, H. Liu, Recent progress in apping wing aerodynamics and aeroelasticity, Progress in Aerospace Sciences 46 (7) (2010) 284{327. [11] N. S. Proctor, M. S. Lynch, Patrick J., Manual of ornithology: Avian structure and function, Quarterly Review of Biology. [12] W. Shyy, Y. Lian, J. Tang, D. Viieru, H. Liu, Aerodynamics of low Reynolds number yers, Vol. 22, Cambridge University Press, 2007. [13] H. Dai, Computational modeling of uid-structure interaction in biological ying and swimming, Ph.D. thesis, Vanderbilt University (2013). [14] S. M. Swartz, Skin and bones: the mechanical properties of bat wing tissues, Bats: Phylogeny, Morphology, Echolocation, and Conservation Biology (1998) 109{126. [15] A. Gogulapati, P. P. Friedmann, E. Kheng, W. Shyy, Approximate aeroelastic modeling of apping wings in hover, AIAA journal 51 (3) (2013) 567{583. [16] C. Farhat, V. K. Lakshminarayan, An ale formulation of embedded boundary methods for tracking boundary layers in turbulent uid{structure interaction problems, Journal of Computational Physics 263 (2014) 53{70. [17] H. Cho, N. Lee, S. J. Shin, S. Lee, S. Kim, Improved computational approach for 3-d realistic insect-like apping wing using co-rotational nite elements, in: 55th AIAA Aerospace Sciences Meeting, 2017, p. 1417. [18] S. Wang, X. Zhang, G. He, T. Liu, Lift enhancement by bats' dynamically changing wingspan, Journal of the Royal Society Interface 12 (113) (2015) [19] F. J. Blom, A monolithical uid-structure interaction algorithm applied to the piston problem, Computer methods in applied mechanics and engineering 167 (3-4) (1998) 369{391. [20] J. Liu, R. K. Jaiman, P. S. Gurugubelli, A stable second-order scheme for uid{ structure interaction with strong added-mass eects, Journal of Computational Physics 270 (2014) 687{710. 49 [21] C. A. Felippa, K. Park, C. Farhat, Partitioned analysis of coupled mechanical systems, Computer methods in applied mechanics and engineering 190 (24-25) (2001) 3247{3270. [22] A. Yenduri, R. Ghoshal, R. Jaiman, A new partitioned staggered scheme for exible multibody interactions with strong inertial eects, Computer Methods in Applied Mechanics and Engineering 315 (2017) 316{347. [23] J. Hron, S. Turek, A monolithic fem/multigrid solver for an ale formulation of uid-structure interaction with applications in biomechanics, in: Fluid-structure interaction, Springer, 2006, pp. 146{170. [24] R. K. Jaiman, S. Sen, P. S. Gurugubelli, A fully implicit combined eld scheme for freely vibrating square cylinders with sharp and rounded corners, Computers & Fluids 112 (2015) 1{18. [25] R. Jaiman, P. Geubelle, E. Loth, X. Jiao, Combined interface boundary condition method for unsteady uid{structure interaction, Computer Methods in Applied Mechanics and Engineering 200 (1-4) (2011) 27{39. [26] R. Jaiman, P. Geubelle, E. Loth, X. Jiao, Transient uid{structure interaction with non-matching spatial and temporal discretizations, Computers & Fluids 50 (1) (2011) 120{135. [27] D. D. Chin, D. Lentink, Flapping wing aerodynamics: from insects to vertebrates, Journal of Experimental Biology 219 (7) (2016) 920{932. [28] P. Causin, J.-F. Gerbeau, F. Nobile, Added-mass eect in the design of partitioned algorithms for uid{structure problems, Computer methods in applied mechanics and engineering 194 (42-44) (2005) 4506{4527. [29] M. Olivier, A uid-structure interaction partitioned algorithm applied to exible apping wing propulsion, Ph.D. thesis, Universit Laval (2014). [30] R. Jaiman, X. Jiao, P. Geubelle, E. Loth, Assessment of conservative load transfer for uid{solid interface with non-matching meshes, International Journal for Numerical Methods in Engineering 64 (15) (2005) 2014{2038. 50 [31] Y. Li, Y. Law, V. Joshi, R. Jaiman, A 3d common-re nement method for non- matching meshes in partitioned variational uid-structure analysis, Journal of Computational Physics. [32] T. Rendall, C. Allen, Uni ed uid{structure interpolation and mesh motion using radial basis functions, International Journal for Numerical Methods in Engineering 74 (10) (2008) 1519{1559. [33] M. Lombardi, N. Parolini, A. Quarteroni, Radial basis functions for inter-grid interpolation and mesh motion in fsi problems, Computer Methods in Applied Mechanics and Engineering 256 (2013) 117{131. [34] A. Beckert, H. Wendland, Multivariate interpolation for uid-structure- interaction problems using radial basis functions, Aerospace Science and Technology 5 (2) (2001) 125{134. [35] A. de Boer, A. Van Zuijlen, H. Bijl, Review of coupling methods for non- matching meshes, Computer methods in applied mechanics and engineering 196 (8) (2007) 1515{1525. [36] A. De Boer, M. Van der Schoot, H. Bijl, Mesh deformation based on radial basis function interpolation, Computers & structures 85 (11-14) (2007) 784{795. [37] F. M. Bos, Numerical simulations of apping foil and wing aerodynamics: Mesh deformation using radial basis functions, Ph.D. thesis, Delft University of Technology (2010). [38] O. A. Bauchau, Flexible multibody dynamics, Vol. 176, Springer Science & Business Media, 2010. [39] O. Bauchau, G. Damilano, N. Theron, Numerical integration of non-linear elastic multi-body systems, International Journal for Numerical Methods in Engineering 38 (16) (1995) 2727{2751. [40] O. Bauchau, N. J. Theron, Energy decaying scheme for non-linear beam models, Computer Methods in Applied Mechanics and Engineering 134 (1-2) (1996) 37{ 51 [41] C. L. Bottasso, M. Borri, Energy preserving/decaying schemes for non-linear beam dynamics using the helicoidal approximation, Computer Methods in Applied Mechanics and Engineering 143 (3-4) (1997) 393{415. [42] O. A. Bauchau, C. L. Bottasso, On the design of energy preserving and decaying schemes for exible, nonlinear multi-body systems, Computer Methods in Applied Mechanics and Engineering 169 (1-2) (1999) 61{79. [43] P. S. Gurugubelli, R. Ghoshal, V. Joshi, R. K. Jaiman, A variational projection scheme for nonmatching surface-to-line coupling between 3d exible multibody system and incompressible turbulent ow, Computers & Fluids 165 (2018) 160{ [44] V. Joshi, R. K. Jaiman, A variationally bounded scheme for delayed detached eddy simulation: Application to vortex-induced vibration of oshore riser, Computers & Fluids 157 (2017) 84{111. [45] R. Jaiman, N. Pillalamarri, M. Guan, A stable second-order partitioned iterative scheme for freely vibrating low-mass blu bodies in a uniform ow, Computer Methods in Applied Mechanics and Engineering 301 (2016) 187{215. [46] A. N. Brooks, T. J. Hughes, Streamline upwind/petrov-galerkin formulations for convection dominated ows with particular emphasis on the incompressible navier-stokes equations, Computer methods in applied mechanics and engineering 32 (1-3) (1982) 199{259. [47] P. R. Spalart, Detached-eddy simulation, Annual review of uid mechanics 41 (2009) 181{202. [48] J. C. Simo, D. D. Fox, On a stress resultant geometrically exact shell model. part i: Formulation and optimal parametrization, Computer Methods in Applied Mechanics and Engineering 72 (3) (1989) 267{304. [49] O. A. Bauchau, J.-Y. Choi, C. L. Bottasso, Time integrators for shells in multibody dynamics, Computers & structures 80 (9-10) (2002) 871{889. [50] P. Wu, Experimental characterization, design, analysis and optimization of 52 exible apping wings for micro air vehicles, Ph.D. thesis, University of Florida (2010). [51] H. Aono, S. K. Chimakurthi, P. Wu, E. S allstr om, B. K. Stanford, C. E. Cesnik, P. Ifju, L. Ukeiley, W. Shyy, A computational and experimental study of exible apping wing aerodynamics, in: 48th AIAA aerospace sciences meeting including the new horizons forum and aerospace exposition, 2010, pp. 4{7. [52] A. Gogulapati, Nonlinear approximate aeroelastic analysis of apping wings in hover and forward ight., Ph.D. thesis, University of Michigan (2011). [53] P. Watts, E. J. Mitchell, S. M. Swartz, A computational model for estimating the mechanics of horizontal apping ight in bats: model description and validation, Journal of Experimental Biology 204 (16) (2001) 2873{2898.

Journal

Physics – arXiv (Cornell University)

Published: Jul 12, 2018

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

A novel 3D variational aeroelastic framework for flexible multibody dynamics: Application to bat-like flapping dynamics

A novel 3D variational aeroelastic framework for flexible multibody dynamics: Application to bat-like flapping dynamics

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

A novel 3D variational aeroelastic framework for flexible multibody dynamics: Application to bat-like flapping dynamics

A novel 3D variational aeroelastic framework for flexible multibody dynamics: Application to bat-like flapping dynamics

References (62)

Abstract

Journal

Recommended Articles

There are no references for this article.

Our policy towards the use of cookies