Nv Taylor

A structure-coupled CFD method for time-marching flutter analysis

Aeroelastic analysis is a critical area of the aircraft design process, as a good understanding of the dynamic behaviour of the wing structure is essential to safe operation of the vehicle. The inevitable inaccuracies present in the modelling of such phenomena impose mass penalties, as large safety margins are necessitated, which in turn lead to overly stiff designs. In an effort to reduce the uncertainty in analysis methods, fully coupled CFD and structural models are under widespread development. We describe the results produced by such a system for a series of test cases based on the AGARD445.6 and MDO wings

Aeroelastic analysis through linear and non-linear methods: a summary of flutter prediction in the PUMA DARP

This paper presents a comparison of linear and non-linear methods for the analysis of aeroelastic behaviour and flutter boundary prediction. The methods in question include NASTRAN and ZAERO, based on linear aerodynamics, and the non-linear coupled CFD-CSD methods RANSMB and PMB, developed at the Universities of Bristol and Glasgow respectively. The test cases used for this comparison are the MDO and AGARD 445.6 weakened wing. In general, it was found that the non-linear methods demonstrate excellent agreement with respect to pressure distributions, deflections, dynamic behaviour, and flutter boundary locations for both cases. This is in contrast to previous studies involving similar methods, where notable differences across the MDO span were found, and is taken to imply good performance of the CFD-CSD interpolation schemes employed here. While the linear methods produce similar flutter boundaries to the coupled codes for the aerodynamically simple AGARD 445.6 wing, results for the transonic ‘rooftop’ MDO wing design did not agree as well.

Investigation of Structural Modelling Methods for Aeroelastic Calculations

Flutter clearance and aeroelastic analysis of aircraft is of vital importance, as avoidance of the non-linear phenomena involved is critical to the safe operation of an aircraft. However, the complexity of the behaviour involved makes detailed analysis extremely dicult and, hence, in practise simplied linear methods have been used for all previous and current designs. However, the non-linear nature of the actual responses means that such methods have inherent inaccuracies, which requires the use of large safety margins, in turn imposing signican t weight penalties. Increasing computational power has allowed more complicated forms of analysis and, hence, fully coupled CFD and CSD solution methods are currently under development, allowing more accurate results to be generated. This paper considers a coupled full non-linear CFD solver and structural dynamic solver, and compares dieren t approaches to modelling the structural behaviour in this method. Specically mode extraction and mass and stiness matrix analysis are considered. The former is more commonly used, the latter more complex and computationally intensive. However, it is demonstrated here that for small to moderate structural grid sizes, the latter has signican t advantages in generality of approach and response accuracy and consistency, at negligible computational expense. The prediction and analysis of the aeroelastic behaviour of an aircraft is of critical importance, as any unstable response to time varying or impulsive loads (gusts, etc.) may rapidly lead to disastrous structural failure. This danger requires signican t safety margins, proportionate to the level of inaccuracy and uncertainty inherent in the prediction method used. This increased safety margin results in unnecessary stiness, and hence a higher structural weight. Methods currently common for aeroelastic analysis in industry involve the use of linear techniques for both utter and aeroservoelastic problems, allowing uncoupling of the aerodynamic and structural equations. This reduces the accuracy of the methods, particularly in the transonic igh t regime where the aerofoil thickness (usually neglected) plays a crucial role in the development and movement of shock waves. They are also inapplicable to non-linear instabilities such as control surface buzz and limit cycle oscillation. A new approach is therefore under widespread development, consisting of simulation by non-linear CFD analysis coupled to a dynamic structural representation of the body under investigation (e.g. references 1{14). This allows time-accurate non-linear analysis of dynamic behaviour, leading to much more accurate methods for utter investigation. The majority of such methods rely for prediction of structural motion on a summation of the displacement produced by a limited number of modes derived from an analysis of the structure by a commercial FE solver. Whilst this provides a simple method for producing a structural model within the main CFD algorithm that may be rapidly solved, it does mean that the delit y of the model is compromised, particularly at high frequencies. As an alternative, the full mass and stiness matrices may be extracted from the FE analysis, and once inverted, used to evolve the structure forward in time (this method is henceforth referred to as ‘MK’ for brevity). This increases the complexity of the analysis, but retains the higher frequency properties.

Moving Mesh CFD-CSD Aeroservoelastic Modelling of BACT Wing with Autonomous Flap Control

Initial results produced by aeroservoelastic simulations of the BACT (Benchmark Active Control Technology) wing are presented. This is a low aspect ratio wing with large trailing edge ap. A fully time-synchronised aeroelastic simulation tool has previously been developed at the University of Bristol, by coupling a 3D central-dierence, nite-v olume, muliblock structured, unsteady CFD code with a linear computational structural dynamics code, in the time-domain. This has recently been extended to account for control surface representation and igh t control system integration, and various aspects of aeroservoelastic simulation have been analysed. All control surface motion is driven through control laws acting on the structure, allowing both commanded and un-commanded deections and distortions in the body modelled. The simulated response of the BACT wing and mount to aerodynamic perturbations and ap angle demands are presented, and a number of key issues relating to the modelling of surface motion are identied, the most signican t of which being the need for consideration of structural and aerodynamic grid behaviour at hinge lines. I. Introduction The accurate prediction of various types of non-linear aeroelastic behaviour is an important area of aircraft design. Whilst utter is the most spectacular example of the dangers posed by unpredicted and uncontrolled oscillations, other phenomena, whilst posing a less immediate threat to vehicle integrity, can have serious eects in the long term due to structural fatigue. Such phenomena would include, for example, Limit Cycle Oscillations (LCO’s) of control surfaces. Methods currently common in industry for aeroelastic analysis involve the use of linear techniques, allowing uncoupling of the aerodynamic and structural equations. However, these methods are not applicable in the most demanding design areas, i.e. non-linear regimes such as transonic utter, control surface LCO’s, and buzz. This can lead to the need for signican t safety margins, creating over-sti and hence high mass designs. More signican tly, it is likely that any errors in design will only be picked up during either ground vibration, or more likely igh t testing. This extends the length of the igh t test program (typically at a cost of thousands of dollars an hour), as a lack of condence in the prediction means that the initial tests must be carried out well below the expected onset of the non-linear behaviour, and speed increased only gradually. Further, the cost of xing any errors encountered at this stage is considerably greater than would be the case if they were to be identied before prototype manufacture through accurate modelling. Prediction methods proven to be of greater accuracy will therefore lead directly to signican t cost savings, without requiring any advances in the underlying aerodynamic or structural design methodologies. For this reason, non-linear techniques are under widespread development (e.g. references 1{22) consisting of time-accurate CFD (Computational Fluid Dynamics) analysis of the o w, coupled to a Computational Structural Dynamics (CSD) structural model. This oers the potential for modelling both the structure and aerodynamics in a non-linear fashion, although admittedly at a far higher cost in terms of CPU time. These methods can also be used to obtain high delit y o w data in regions of particular interest. At the

A comparison of linear and moving mesh CFD-CSD aeroelastic modelling of the BACT wing

Consideration of the aeroelastic and aeroservoelastic behaviour of an aircraft is important throughout design, as catastrophic failure may occur in extreme cases, and more generally aeroelasticity can impact upon safety critical factors such as fatigue, control and vibration, and hence lead to limits upon the safe operating envelope. At low speeds, linear representations of the structural and aerodynamic interactions may be employed with reasonable success. In the transonic region, however, the non-linear nature of the o weld means that a linearised model no longer accurately represents the physical behaviour, and more complex forms of analysis become necessary, such as fully coupled non-linear computational schemes. Such methods are validated against experimental test cases, and in turn the linear methods compared with both. This usually results in broad agreement between all methods up to the transonic range, whereupon the linear methods diverge from both experiment and the more advanced non-linear schemes. The BACT wing, however, is somewhat unusual in this regard in that the most accurate prediction of the utter speed in the transonic dip has been produced by a (highly localised) linear representation, whilst coupled codes have struggled to predict even lower speed utter points, and completely missed the dip itself. The reasons for this are investigated in this paper through a comparison of a model based on linear aerodynamics and an unsteady, fully coupled CFD solver. It is found that using the reported data, the linear method signican tly under predicts the utter speed. Using data derived from unsteady CFD simulation, a close correspondance is found with the full simulation, despite an over prediciton of utter speed of about 10% at Mach 0.77. Although compared to previous CFD simulations, USCRANSMB performed well, the boundary produced was found in places to vary signican tly from the experimental result. A theory for this discrepency advanced elsewhere relating to the centre of gravity location could not be reproduced in this case. Although inconclusive, some evidence was produced suggesting that the diering utter boundary is caused by the eects of the wind tunnel itself (not modelled in the CFD), and/or other errors relating to shock position and strength.