Sei sulla pagina 1di 35

Installation Eects on Propeller Aeroelasticity

Literature Review and Preliminary Scheme of Work Survey


Harry Smith May 3, 2012

This document is the result of the preliminary literature survey for a Ph.D. CASE Studentship with Dowty Propellers (DP), the Aircraft Research Association (ARA) and the University of Glasgow (UoG). The papers, books and other sources referenced herein encompass a fundamental grounding in the history of propeller aerodynamics. Additionally, the zeitgeist of rotary wing unsteady aerodynamics, aeroelasticity and structural dynamic modelling has been researched and presented as relevant to this project. The key issues relevant to this project, likely methodologies and analytical techniques that shall be used will be presented herein and a preliminary outline of the sequence of research and how work is expected to proceed is presented at the end of the document. To fully appreciate the relevance of these source and why they have been chosen, the outline of the project is explained along with an extract from the project brief project in section 2 on page 4.

Contents
1 Summary 2 The Project 3 Propeller aerodynamics 3.1 History of Propeller Aerodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Unsteady Rotary Aerodynamics 4.1 Dynamic Inow/Finite-State Induced Flow Models . . . . . . . . . . . . 4.1.1 Historical Development - Nonuniform Inow/Air Mass Dynamics 4.1.2 The First Modern Dynamic Inow Model - Pitt and Peters . . 4.1.3 Beyond Pitt and Peters Model . . . . . . . . . . . . . . . . . . . 4.1.4 Dynamic Inow for propeller applications . . . . . . . . . . . . . 4.1.5 Dynamic Inow Summary . . . . . . . . . . . . . . . . . . . . . . 4.2 Nonlinear Aerodynamics - Dynamic Stall . . . . . . . . . . . . . . . . . 3 4 5 5 8 11 12 15 17 19 22 23 24 24 25 25 27 27 28 29 30 35

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

5 Propeller Structural Dynamic Model 5.1 Lower order models: rotating beams . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Beyond beam models: Finite Element Analysis . . . . . . . . . . . . . . . . . . . . . . 5.3 Structural Dynamic Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Proposed Initial Scheme of Work/Model Construction 6.1 Initial Model Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Initial Aerodynamic and Dynamic Inow Models . . . . . . . . . . . . . . . . . . . . . 6.3 Structural Dynamic Model addition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Testing and Validation A Proposed Timeline

Summary

The exploration of the various relevant literature is fully detailed in this report. A brief summary of the key ndings is included here for clarity and brevity. Propeller aerodynamic theory for design and analysis is found to be geared towards performance calculations and little else - no major theoretical work has been completed for propeller calculations since that of Theodorsen. Vortex theory has been used to model, to some degree, the propeller wake - the ideal propeller can be designed by extension of wake theory to predict minimum-loss propeller loading. No major work has been done to calculate unsteady propeller loading or o-design cases - this is the area upon which this project aims to expand the analytical tools available by taking advantage of developments in the rotorcraft eld for unsteady aerodynamic analysis. Unsteady rotary aerodynamics have been to date largely unexplored for propellers. This is perhaps due to the mindset that if a propeller works then detailed analysis of vibratory loads, secondary ows etc. are unnecessary as a propeller provides propulsion and not lift. Conversely, a helicopter main rotor provides propulsion and lift and its continued safe performance is more mission-critical. Accordingly, the bulk of unsteady aerodynamic theory for this project has been taken from rotorcraftfocussed literature. Dynamic inow appears to be the most amenable analytical tool for this project - it is a proven unsteady aerodynamic tool in frequent use in rotorcraft analysis, easily compatible with structuraldynamicanalysis. It has arisen as a coalescence of nonuniform inow theory with investigations into unsteady ow for helicopters. It has not been proven to work for propeller analysis and a large part of this project will be proving the feasibility of such an application via working through the mathematical theory from the ground-up. The suitability of a Dynamic Inow model for the purposes of propeller analysis is, as mentioned, untested. On a propeller the method should provide no real issues - the problems of compressible ow and advance ratio are foreseen as being negligible when looking at 1p-loading. That is, if we were attempting to ascertain the detailed lift distribution on each blade at each azimuthal position, span and chordwise ordinates, then we may need to turn to a more detailed analysis e.g., free wake analysis. For the purposes of determining the discontinuities in loading as the propeller completes each revolution, dynamic inow should prove a suitable method to determine unsteady eects. Structural dynamics for this project are less groundbreaking than the unsteady aerodynamic theory. As suggested in the project brief, and discussed in Section 5, the best approach shall be to start with a simple beam model and increase the delity step by step. For the rst phase of this project, a Rayleigh-Ritz approach for a simple beam capable of twisting and bending is likely to be sucient. It shall hopefully become clear whether this will provide enough detail for the intended nal code, or if extension to a full structural FE model is necessary. Section 6 and Appendix A provide the provisional scheme of work.

The Project
The aim of this research project is to develop a mathematical model of propeller aerodynamics and aeroelastics that is capable of accurate predictions of propeller 1P loading, i.e. azimuthal variation of propeller blade loading caused by asymmetrical propeller inow. Dowty Project Brief, 2011

As laid out in the extract from the project brief from Dowty, partly quoted above, the project requires the formulation of a mathematical capable of determining the eect of nonuniform inow on 1P-loading (once per revolution loading). No one single model will be capable of such analysis alone. Instead, the nal model will be formed from separate models of propeller steady and unsteady aerodynamics and propeller structural dynamics/aeroelasticity. As shall be discussed further, the use of a Dynamic Inow model will likely be amenable to this task. Accordingly, Figure 2.1, below, is a re-working of similar diagrams presented in rotorcraft dynamic inow literature, adapted for the propeller problem. By looking at the dierent areas required, the work may be split down into subtasks, research areas and subroutines within the nal model.
Induced Flow Mode/Unsteady Aerodynamic Model (s, ) (s, ) CT , C M , C N
mc ms or n , n

a/c Flight Environment & Prop. Conditions

, , V ac , ac

Propeller Aerodynamic Environment

Vb (s) b (s)

Blade Aerodynamic Model

dL, dD BE Forces

h(s, ) h(s, )

(s, ) (s, ) Structural Dynamic Model L(s), D(s) Inertial Loads

Figure 2.1: Basic Model Structure Required

Propeller aerodynamics

Before developing an unsteady aerodynamic model, one needs to have an appreciation of the steady aerodynamics in play on a propeller. Propellers have been in use since the birth of powered aviation the Wright Brothers, although experimenters and not theorists, took the step of changing to a pair of thin, highly twisted blades turning at high speed as opposed to the cumbersome, windmill-like blades of other designers of the time. With little power to play with from the engines of the day, this step was key in securing their role as the founders of heaver-than-air ight. Although, as stated above, they were experimenters and not theorists, there is evidence that they may have been the rst to combine BladeElement and Momentum theory to determine the AoA distribution along the blade span (Wald, 2001). Since this simple modication in blade shape, much work has been completed on propeller theory and design to bring us to the present day where we see scimitar-type blades capable of producing excellent propulsive eciency, even at high-subsonic ight regimes, comparable to the performance of a turbofan. The history of propeller theory and design for optimum eciency is laid out chronologically in great detail in a 2006 review of propeller aerodynamics (Wald, 2006). This forty-page work provides an overview of not only the key developments, but goes into detail on the theory of propellers and shows the dierent means via which propeller calculations have been performed through dierent ight regimes; propeller, vortex-ring and windmill states. This document will likely be of use for the duration of this project. It should be noted that some of the works and citations discussed in the following section are taken directly from Wald (2006), as the works themselves are not written in English and no translations readily available.

3.1

History of Propeller Aerodynamics

The earliest applicable theoretical work on propellers was performed by Froude and Rankine (1865), working on marine propellers. The basic theory of the propulsive uid momentum equation was formulated, and thus created actuator-disc theory. Although this crude approximation allows designers to set performance and sizing requirements, it involves no real detail on actual propeller aerodynamics, instead proposing an innitesimal imaginary disc through which there is a discontinuity in pressure and momentum, resulting in thrust. In 1892 (and later in 1901), Drzewiecki modelled propeller blade elements moving through a uid medium on a helical path, but did not account for the propeller-induced velocity due to the shed wake. It was not until Prandtls lifting line theory with free and bound vorticity that the modern theory of propeller aerodynamics would be possible. Using this theory, the propeller may be considered to be a lifting surface about which there is a circulation associated with bound vorticity and a vortex sheet shed continuously from the trailing edge - propeller wake theory gives a better appreciation of the actual physical ow eld than the crude approximations of actuator-disc theory. In 1919, Betz showed that the loading distribution for lightly-loaded 1 propellers is such that the
1 A lightly loaded propeller is described as one where the induced velocities are small compared to the propeller velocity (Makinen, 2005) - this was never clearly dened by Betz in the original literature.

shed vorticity forms regular helicoidal sheets moving aft uniformly from the propeller. This work was reprinted by Prandtl and Betz (1927) with an appendix showing a closed-form approximation to this ow for small advance ratios. Goldstein (1929) solved the problem of the potential eld and circulation distribution based on a helicoidal vortex system as outlined by Betz. This circulation distribution was presented in tabulated form for two and four-bladed propellers as the ideal distribution. Ideal design via Goldsteins function is thereby a design procedure to t this ideal circulation distribution . Although the work by Betz, Prandtl and Goldstein marked the development of a better understanding of propeller aerodynamics via vortex theory, work continued on the renement of Blade Element Momentum Theory (BEMT), notably by Glauert. This development as in (Glauert, 1926b, 1943) is still applied in practical calculations today, despite the dependence on an independence of blade elements, which was shown to be without physical justication2 . The long out-of-print Theory of Propellers (Theodorsen, 1948) is essentially a collection of NACA TNs 775-778 (Theodorsen, 1944a,b,c,d) - Theodorsen followed from Goldsteins ideal circulation distribution and studied the shed wake far downstream, rather than at the propeller itself and as such was able to negate the need for the lightly-loaded condition that had previously been invoked. Tibery and Wrench Jr (1964), applied mathematicians, used modern computational and mathematical techniques to tabulate accurately functions related to the Goldstein function over a wider range of propeller parameters than had previously been attempted. A design procedure developed by Larrabee (1979), capable of determining the geometry of minimum induced loss propellers based upon specic conditions (e.g., ; loading, , n etc.) has the benet of convenience, quoted as ecient...adaptable to pocket calculators (Wald, 2006). However, the theory behind the work is based on the original vortex assumptions by Prandtl and Betz (1927) and does not utilise the more rigorous and general concepts introduced by Theodosen. The preceding section gives a comprehensive overview of the key developments in steady propeller aerodynamics. It should be noted that the most recent signicant work dates to 1979, with the most major theoretical developments being that of Theodorsen in 1948. Despite the main principles of propeller design being over sixty years old, they have enabled designers and blade manufacturers to produce ecient and powerful propellers. Where there has been a lack of propeller-focussed research, and where there are gains to be made in terms of reduced fatigue etc., is in the propeller unsteady aerodynamics. By analysing how the blade forces and moments change with respect to time-varying or non-uniform inow, correlation between 1P-loading and design features may be determined and blades designed/adapted for maximum life between service intervals. Although there has been a lack of propeller-focussed unsteady aerodynamic theory, there is a wealth
2

Taken from Wald (2006)

of information on rotorcraft-based investigation. This is down to a combination of the fact that a helicopter rotor is generally subject to more unsteady ow (e.g.; more manoeuvring ight, edgewise ow and transition between propeller, vortex-ring and windmill-brake states) and the fact that a helicopter rotor provides the main lifting surface in addition to propulsion. Therefore unsteady aeroelastic phenomena such as utter etc. are clearly not only more likely without careful and safe design procedure, but also more catastrophic than they would be on a propeller (although one would still hope to avoid blade utter on a propeller). It should be noted that work on nonuniform pressure distributions was completed by Glauert (1926a) during development of autogyro theory, but not applied to propeller aerodynamics - likely because of the reduced eect of inow variation on the propeller problem, and the aforementioned critical aspect of a rotor vs. a propeller as lifting surface vs. pure propulsion.

Unsteady Rotary Aerodynamics

As mentioned in the previous section, there is little information available on propeller unsteady theory, and the key sources found that deal with rotary unsteady aerodynamics have come from helicopterfocussed literature. An excellent starting point for rotorcraft unsteady aerodynamics is in Principles of Helicopter Dynamics (Leishman, 2000). This book gives an overview of the dierent unsteady theories and how they compare under dierent ight environments/analytical procedure. Before going into detail on the model that is anticipated shall be most suitable for propeller analysis, this section shall briey cover the key developments in unsteady aerodynamics as laid out by Leishman (2000, Chapter 8 - Unsteady Aerodynamics), and why they are suited or unsuited to propeller analysis. To explain the relevance of unsteady aerodynamics, Leishman presents a breakdown of the sources of unsteady aerodynamic loading that may be found on helicopter rotor. Figure 4.2, below, is an adaptation of Leishmans breakdown of the sources of unsteady aerodynamic loading, put into context here on a propeller. The relative frequencies and amplitudes of dierent eects are given; the frequency dictates the level to which unsteady eects dominate the ow via reduced frequency k = b V

Blade Motion

Pitch

Flap

Torsion (high freq., high amp.)

Control (low freq., high amp.)

Flap* (low freq., high amp.)

Bending (high freq., low amp.)

(a) Unsteady Aerodynamics due to Blade Motion *(only applicable if including the proposed pin-jointed hinge)
Floweld Structure

Periodic

Aperiodic

Velocity (low freq., low amp.)

Downwash (low freq., low amp.)

Sweep (low freq., low amp.)

Fuselage oweld (low amp.)

Wake distortion (low amp.)

Discrete vortices (high amp.)

(b) Unsteady Aerodynamics due to Floweld Structure

Figure 4.2: Unsteady Aerodynamic Sources on a Propeller; adapted from Leishman (2000, taken from Beddoes (1980))

Before going into the detail of dierent unsteady theories, Leishman lays out the requisites of any useful unsteady aerodynamic model, for use in practical rotary aerodynamics. The following are largely common sense, but fundamental to choosing any aerodynamic model to be used as part of a larger calculation procedure and hence are included here. 1. The assumptions and limitations of any model need to be fully assessed, understood and justied if invoked. e.g., incompressibility requires not only local M 1 but M k 13 . 2. The model must be written in a form that is easily coupled with structural dynamic model. e.g., the model may be in terms of ODEs at radial blade elements, or written in state-space form at the disc level. 3. If choosing an integral approach, i.e., a BE model, the concentric radial and azimuthal integration loops place a strict limit on the computational cost of any unsteady model4 . Leishman discusses reduced frequency to be used as a measure of the degree to which unsteady t eects dominate the oweld, and the reduced time, s = 1 0 V dt, as a non-dimensional ordinate b representing distance that the blade travels through the oweld in semi-chords during a given time interval, t. An preliminary analysis of the eect of unsteady ow in the propeller problem will be benecial before constructing the model, and by calculating k over the disc it is anticipated that a better appreciation of when and where in the oweld unsteady eects dominate. This shall assist with preliminary qualitative analysis of initial results. The book goes through the development of unsteady attached theory, quasi-steady thin-aerofoil theory, Theodorsens theory/function with returning-wake additions by Loewy and Jones and through other frequency-domain theories. As shall be discussed below, such frequency domain based models are not foreseen to be used in the proposed model as it shall be preferable to have a time domain based solution. However, it is important that they have been researched as they may prove useful at a later date. Indeed, as shall be discussed in section 4.1.4, the proposed extension of a Dynamic Inow model to the propeller problem and validation thereof is not guaranteed and this work may have to rely on a Theodorsen method or similar to model unsteady aerodynamics. This will be determined at a decision gate after the theoretical work to extend dynamic inow to the propeller case is complete. Leishman (2000, 8.14.2: State-Space Solution) highlights the advantage of state-space form for an unsteady aerodynamic model that is to be used as part of a larger aeroelastic model - in such a form, the governing dierential equations may be used in conjunction with the time domain based structural-dynamic model. Another fundamental benet of using a system that is in state-space form is that it is possible to calculate the response of a propeller/rotor system at a given time (i.e., for a particular state), rather than a models validity being limited to the stability boundary. Eorts to formulate a state-space model of 2D unsteady aerofoil theory via Laplace transform methods of the indicial response have been performed by many using approximations to Theodorsen and Wag3 Incompressibility may not be assumed in a blade-level model as tip speeds will be high subsonic/transonic. The validity of assuming the wake compressibility may be negated is discussed in Section 4.1.4 4 As shall be discussed in Section 6, use of vectorisation in code construction negates the concentric loops for some calculations, although a numerical integration scheme shall require them still.

ners functions - however, the approximations essentially amount to curve-tting and have no physical basis. For a problem with little previous theoretical work and without a wealth of experimental data, it is dicult to justify such a method and as such it will be preferable to use a more mathematically rigourous formulation.

10

4.1

Dynamic Inow/Finite-State Induced Flow Models

The theories described in the preceding section are an attempt to account for unsteady aerodynamics at the blade level. Described herein are Dynamic Inow models - also referred to as Finite-State Induced Flow models 5 in the relevant literature. It should be noted that Dynamic Inow has been developed as a helicopter analysis tool and as such the nomenclature used in this section is rotorcraftfocussed. That is, disc shall be synonymous with rotor disc until the discussion of extension to the propeller problem in section 4.1.4, where the distinction shall be made clear. Dierent, isolated work on theories related to but not aimed at producing a dynamic inow model have arisen over the last century, and the latest incarnation(s) of dynamic inow is are amalgams of these previous theories, reworked with modern and rigorous mathematical formulations. As such, the basic principles fundamental to dynamic inow are given here: 1. The model utilises some distribution of pressure/induced velocity variation over the disc; e.g., radial, azimuthal (or both) expansion, say, Fourier series with nite coecients i.e., the nite states. 2. It contains some accounting for transient ow characteristics such that second-order ow characteristics are modelled e.g., using unsteady actuator disc theory. Hence dynamic inow. 3. A coupling of rotor loads and induced ow such that the model may be linked with a basic aerodynamic model e.g., BEMT to form a closed-loop model amenable to eigenvalue stability analysis. Such that the basic format of a dynamic inow model is, in matrix form: [M ]u + [L]1 u = F (4.1)

Where u and F are state vectors of induced ow and aerodynamics forces, respectively. [M ] and [L] are the mass matrix and gain matrix, respectively. Their product is dened [ ] [L][M ] as the time constant matrix. In researching this project, the development of the method has been taken from A New Appreciation of Inow Modelling for Autorotative Rotors (Murakami, 2008) as this thesis sheds light on some of the interim mathematics than Peters et al. leave out in publications. It may be observed that eqn. 4.1 is loosely in the same form as Eulers equation of motion for an incompressible uid: u + (u ) u = P (4.2) t Which itself is an extension of Newtons second law: Fext = m aabs (4.3)

Hence it may be seen the modern dynamic inow model, although reached by complex mathematics, has simply the fundamental laws of motion at its roots. The latest incarnations of dynamic inow have been reached via years of investigation into what shall be termed herein nonuniform inow and air
nite, here, refers to the relatively small number of states as compared to other computation methods e.g., CFD which, whilst still not comprising an innite number of states (or nodes in CFD) may have several million whereas Dynamic Inow will only have single-digit numbers of states.
5

11

mass dynamics, the meaning of the former hopefully being clear, and the latter denoting research into the dynamic response of the uid medium surrounding a rotor to changes in the forces acting upon it and vice versa. Hence now, at least, the etymology of dynamic inow should be clear to the reader. The two of these have been investigated separately over the last century for various rotorcraft purposes and combined together in the latter half for incarnations of dynamic inow models. Historical development of both methods shall be presented in the next section, although it should be clear that not all of the following work was performed with a nite state/dynamic inow methodology in mind, but dynamic inow has at its roots the fundamentals proposed herein. 4.1.1 Historical Development - Nonuniform Inow/Air Mass Dynamics

From the earliest rotorcraft design, the inow over the the rotor was presumed to be uniform, enabling easy calculations by momentum theory providing reasonable insight into hover performance, despite being a large simplication of the actual ow situation (Bramwell et al., 2000). During forward ight (or, infact, any translational ight), there is a radial and azimuthal variation in inow due to dierent factors: A one-dimensional variation in the axis of ight direction due to translational lift causing rotorinduced upwash at the leading edge, much like an aerofoil of equivalent size - this eect is much, much smaller on a propeller at an angle of attack than on a rotor blade. A radial variation of local tangential velocity and therefore inow angle along the blade axis. A lateral (azumuthal) variation of inow due to asymmetric summation of forward speed and local rotational velocity on advancing and retreating blades at AoA = 0. The autogyro is clearly not capable of axial ight, and as part of his work on autogyro theory, the variation in inow on the rotor was rst accounted for by Glauert (Glauert, 1926b) by modelling the disc as an elliptically-loaded circular wing with associated downwash distribution. He proposed a mean and longitudinal variation in inow of the form: vi0 = vi T 2AV = vi0 (1 + Kc x cos ) (4.4) (4.5)

r Where the nondimensional radial ordinate is x = R and Kc is a coecient slightly larger than one, chosen so that the equation 4.5 produces an upwash at the leading edge and a positive linear gradient of vi towards the trailing edge i.e., along the line ( = 0 or ).

Much work was done over the following years to rene the denition of and values for Kc , through both experimental and theoretical means. Most of the work for this shall not be included in this document - for a full review of the development of Kc for autogyro and helicopter ight, the Ph.D. thesis by Murakami (2008) provides a good perspective. Arguably the rst innovative addition to Glauerts work was provided by Coleman and Feingold (1945) where they dened Kc to be a function of , the wake skew angle. 12

Kinner (1937)6 introduced the use of elliptical co-ordinates to the problem and proposed the use of pressure distributions as a series of the associated Legendre functions - satisfying the Laplace equation and providing a discontinuity in pressure (i.e., lift) across the disc. This step was key to gaining the dynamic inow models that are used today, but it should be noted that Kinner was not exploring unsteady induced ow, rather the distribution of the steady induced ow. Mangler and Squire (1950) extended the work of Kinner by associating his lift distribution with the induced velocity eld - using Eulers equations and a variant of actuator disc theory, they were able to calculate the acceleration of the induced ow distribution so as to satisfy the requisite hub loading. This coupling, as Murakami (2008) notes, may be regarded as the theoretical forebearer of modern dynamic inow theory - more recent dynamic inow models couple pressure distribution to inow, but the concept is the same of coupling the loads. Joglekar and Loewy (1970) extended Mangler and Squires work to incorporate the eects of the shed wake geometry, but the technique is less sophisticated than in modern methods, as noted in (Murakami, 2008) - additionally various sources show that the models based on Mangler and Squires work, although innovative, fail to show good correlation with experimental data; (Chen, 1989). Around the same time as Mangler and Squires work, Sissingh (1951) attempted to account for a cross-coupling in roll and pitch damping of a helicopter rotor, rst reported by Amer (1950), by assuming a nonuniform distribution of induced velocity based on the rst harmonic variation of the lift coecient, or: v = v0 + v1 cos (4.6) His results for rotor damping cross-couping matched predictions and the results were well received, nding use in commercial simulation programmes (Gaonkar and Peters, 1988). However, the model above utilises a Fourier distribution based only on the azimuthal variation, neglecting the radial variation. Additionally, Sissinghs work presumes that the reaction of inow to the change in rotor loads is instantaneous, and does not account for the aerodynamic hysteresis due to dynamic ow eects and is termed the quasi-steady model. Nonetheless, the step of explaining the cross-coupling by linking with nonuniform inow lays the foundations for dynamic inow theory. The relationship between rotor load and induced ow had been noted much earlier than the above analyses in Wheatleys 1935 analysis of autogyro rotor blade vibration, concluding that Glauert and Locks previous work on autogyro theory was quantitatively usable except for the blade motion...[which] cannot be calculated rigourously without the accurate determination of the induced ow. Wheatleys report proved the link between induced ow and rotor loads, but it was not until the advent of the hingeless rotor nearly thirty years later that inclusion of this eect gained momentum as a necessary step in a rotor modelling. This is due to the ability of the hingeless rotor to apply moments directly to the hub noted in by Ormiston and Peters (1972)The nonuniform inow is a direct result of the
6

German language paper, but theory described by Joglekar and Loewy (1970)

13

large hub moments developed by hingeless rotors...neither the induced inow nor elastic blade bending are of comparable importance for articulated rotor response. This is a key point for the proposed propeller analysis as current propeller design is essentially a hingeless design, rather than articulated7 and as such, the fundamental principles of work based on the correlation between rotor loads and inow may be carried forward. Carpenter and Fridovich (1953) explored time delay in control response of rotorcraft - observing a lag in the build up of rotor thrust in response to a change of collective pitch whilst exploring the jump take-o of heavily-loaded helicopters. For this, a rapid increase in collective is applied to a pre-rotated rotor, eecting a vertical take-o. The observed time delay or lift hysteresis was attributed to Newtons second law and the force required to eect a momentum change in the uid surrounding the disc. Following the theory of NACA TN 197 (Munk, 1924), they added an apparent addition mass to the 1D momentum equation of 63.7% of the mass of a sphere of uid with the same diameter as the rotor disc, and their method is now referred to as unsteady momentum theory8 . In their work on nonuniform rotor loads and the relationship with pitch and roll moments Curtiss and Shupe (1971) developed a model for axial ight in a quasi-steady formulation. Rather than incorporating dynamic ow eects using a rigorous physical formulation, they developed the reduced Lock number, eectively the same as modelling a heavier blade. The two methodologies for modelling the lift hysteresis have been explored by Banerjee and Crews (1979) and found that the dynamic inow model works better for low advance ratio regimes (Murakami, 2008). The models developed up until this point had been proved valid only in axial ight, and showed poor correlation with forward ight. Although our propeller case is closer in oweld situation to the axial case, the aircraft will be cruising at an angle of attack and it is this skewed inow/wake that we want to determine the eects of. Work from the 1960s developed techniques for experimental measurement of the induced velocity distribution, and the results showed that unsteady momentum theory of Carpenter and Fridovich matched the result poorly (Gaonkar and Peters, 1986). Dynamic inow needed developing in order to cover forward ight conditions, and much work was done to amend the theory. Methods for developing new methodologies to cover the forward ight regime included local momentum theory and a variety of methods combining blade-element and momentum theory, but few showed success. Development of these methods is explained in (Murakami, 2008) and shall be disregarded here. Ormiston and Peters (1972) developed a quasi-steady model in which nonuniform inow distribution was included in a hingeless rotor model. Their model was based on Kutta-Joukowski circulation theory and developed in a matrix formulation relating induced ow and rotor thrust and moments.
Aside from the proposed pin-jointed hinge proposed as conceptual design, the eect of which on 1p-loading is to be determined. 8 In (Carpenter and Fridovich, 1953) they do not follow the method of TN 197 fully, rather they adapt a two-dimensonal method to match with experimental data. This is not documented in the paper, but the reliability of their method is noted in other sources and by prominent researchers e.g., Peters and colleagues.
7

14

Peters (1974) extended this model using unsteady momentum theory to develop the basic formulation of dynamic inow in its whole form. The model was limited to hovering ight (o-diagonal gain matrix elements are not included). Murakami (2008) also notes that the elements of the matrices were not mathematically rigorously determined. 4.1.2 The First Modern Dynamic Inow Model - Pitt and Peters

The preceding section describes the eorts undertaken to develop an accurate model for nonuniform inow and dierent means of accounting for unsteady eects. These works have been integral to developing the fundamentals for dynamic inow , but it is has been the work of Professor David A. Peters and colleagues that has built on these foundations and ne-tuned the mathematics to make it into the modern analytical tool that it is today, nding accurate and reliable use on helicopter, autogyro and wind turbine analysis. Described in the following section is the work that has been done on modern dynamic inow models9 . Gaonkar and Peters 1988 paper, Review of Dynamic Inow Modelling for Rotorcraft Flight Dynamics, contains an excellent overview of the key work that has been done. Ormiston (1976) provided the rst dynamic inow model in which o-diagonal elements of the component matrices are evaluated. The model uses a rst-harmonic expansion of inow components, apping and pitch angles and is capable, to some extent, of accounting for the cross-coupling of rotor behaviour. Peters and Gaonkar (1980) investigated ap-lag stability and extended the quasi-steady reduced Lock number approach (Banerjee and Crews, 1979) to the advanced dynamic inow model suitable for analysing ap-lag stability. Banerjee and Crews (1979) attempted to populate the gain matrix via experimental methods and found that the model based on unsteady momentum theory was more reliable the the reduced Lock number approach over the low advance ratio ( < 0.4) regime. The most crucial development in dynamic inow theory is the Pitt-Peters model (Pitt and Peters, 1981) - this model is essentially an extension of the theory of the original model (Ormiston and Peters, 1972) using linearised unsteady actuator disc theory, entirely neglecting the reduced Lock number approach and dening the gain matrix as a function of the wake skew angle, thereby returning to the concept rst introduced by Coleman and Feingold in 1945. The Pitt-Peters model was the rst threestate, rst harmonic, perturbation dynamic inow model to nd practical usage in rotorcraft ight dynamics (Chen, 1989). In two NASA Ames Research Center papers by Chen (1989); Chen and Hindson (1986) comparison is made between the dierent dynamic inow models for use in rotorcraft ight dynamics. The former paper compares various rst-harmonic inow models (including the Pitt-Peters mode) against both computational free-wake models and experimental wind-tunnel data. Chen nds that the PittPeters model performs the most favourably of the rst-harmonic methods which all perform well - in his words all the rst-harmonic inow models predict the induced velocity as well (or as poorly) as the free-wake methods when compared to a set of new data at advance rations of 0.15, 0.23 and 0.3, whilst noting previously in the paper that computational limits place a constraint on the feasibility of free-wake CFD methods and this is still the case nearly thirty years later, certainly for the proposed application where computational cost is to minimised and speed and simplicity of model optimised.
modern, here, refers to dynamic inow models of the form of, or theory integral to the dynamic inow model(s) proposed to be used in the propeller problem, despite some of the work being 35 years old.
9

15

The number of states in the Pitt-Peters model refers to the order of the Fourier/Legendre expansion and includes the mean (uniform) inow component. Hence we see odd numbers of states in dierent versions of the Pitt-Peters model e.g., uniform inow plus n sine and cosine terms, where n is the order of the expansion. The gain and mass matrices of the models will be scaled accordingly i.e., a 3 3 matrix for a three-state model and a 5 5 for a ve-state model. Nagabhushanam and Gaonkar (1984) compared three and ve-state variants of the Pitt-Peters model in a forward-ight air resonance stability investigation. They determined that for simulation of two, three and four-bladed rotors, the three-state model was more reliable than the ve-state model, concluding that the ve-state model contaminates the damping of dierent modes depending on the number of blades;
For instance for N = 3, it [the ve-state model] gives extraneous terms which contaminate the damping level of all the modes except the lag collective mode. For N = 4, these extraneous terms contaminate only the lag dierential collective mode. Concluding remarks, Nagabhushanam and Gaonkar (1984)

The correlation of the performance of dierent models with the number of blades is a curious point. Since dynamic inow is a disc-level model, the relationship is counterintuitive. This result does not seem to warrant further discussion in their paper, nor in any associated literature. It shall be interesting to see if similar correlations may be found with recent developments of dynamic inow models and, if so, possible physical justication for it. As shall be discussed further in this document, validation of a dynamic inow model for the propeller case will most likely require a full working from the ground up, ensuring that assumptions made are fully valid for a propeller as opposed to a rotor e.g., use of Legendre/Bessel functions in the distribution implies a certain loading distribution that varies between planforms. Pitt and Peters model is widely used and has demonstrably good performance when used in analysis. It is not, however, without its aws in implementation and formulation, and there have been a series of developments made to improve upon performance and make models with more rigorous formulations. It should be noted that the publications by Peters et al. do not provide a clear and intuitive formulation of the Pitt and Peters model (nor, in fact, the later Peters and He model). The theory behind and limitations thereof are explained well, but the mathematical formulation is provided with large steps missing in some fairly involved mathematical work. Dr. Yoh Murakamis 2008 thesis A new appreciation of inow modelling for autorotative rotors provides sections which, in his own words, improve the lucidity of the explanation of the mathematical derivation of this mode by complementing the details of some dicult rearrangements of equations and correcting some typographical errors...10 . The work has been cited earlier, but will be mentioned here again. As an engineer and not a mathematician, this author is greatly thankful for Murakamis work. It has been useful thus far as a good source for references on the development of dynamic inow theory, but it shall prove even more useful in this elucidation of the mathematics behind the theory, when it comes to validating a model for a new planform.
10

here, referring to the Peters-He model, but he provides similar instructive steps for the Pitt-Peters model.

16

4.1.3

Beyond Pitt and Peters Model

As mentioned previously, the Pitt-Peters Dynamic Inow model is not without its shortcomings. The following noted shortcomings are taken from (Murakami, 2008): 1. The formulation of the model is such that it is unable to dierentiate between variations in loading that provide the same net forces. That is, the model relies on hub loads rather than distributed loads. Since we require blade structural analysis and aeroelasticity for the propeller problem, a model that can distinguish a specic lift distribution will be advantageous. 2. Despite involving complicated inow distributions and mathematical formulations, the underlying theory is reliant on the actuator disc assumption. Murakami notes that this may reect a limitation of the model in accounting for phenomena related to the blade geometry - this may more pronounced for the propeller problem due to increased twist and solidity, and reduced aspect ratio. The implications of this shall become clearer during rederivation of the model for propeller analysis. 3. The Pitt and Peters model as presented in (Pitt and Peters, 1981) is expressed in linearised, perturbation form. This is not only a simplication that is not required with the computational techniques available - for our purposes it shall be more useful to have variables dened in net terms. 4. The Pitt and Peters model denes quantities in the rotating frame, but for some applications it is more useful to dene the model in wind-axes. The relevance of this nal point on our problem is questionable since were likely to stay in the rotating frame for aeroelastic analysis, but the modications made to the model to accommodate it shall be mentioned herein. Pitt and Peters (1983) noted these shortcomings and reevaluated their model against other inow models. Gaonkar et al. (1983) present a hierarchical list of dierent dynamic inow models (thirteen in total) and the authors note that prior to the Pitt and Peters model, there was no suitable replacement for momentum theory. Their conclusions on the varying reliability of three and ve-state models supports the previous ndings of Nagabhushanam and Gaonkar (1984) and the correlation thereof with the number of blades, N . Peters and HaQuang (1988) demonstrated a fully nonlinear version of the Pitt and Peters model, along with a more instructive derivation than that given in (Pitt and Peters, 1981). Again, Gaonkar and Peters (1988) provide an excellent summary of work that has been done and a comparison between linear and nonlinear models. Murakami argues that the most important development in dynamic inow modelling since the PittPeters model is the Peters-He model, explained in concise form in Peters and He (1995) and rst presented six years earlier (Peters et al., 1989). The Peters-He model is praised in relevant literature as being more rigorous from from rst principles and in its mathematical foundations. Murakamis thesis once again provides a clearer and more instructive narration of the model formulation than the concise one provided by Peters and He. The Peters-He model allows for the calculation of the axial component of induced velocity, using two families of functions for the inow expansion; Legendre functions of the rst and second kind,

17

m m Pn () and Pn ()/, respectively. Both sets included m + n O11 Developments on the Peters-He model aord the ability to account for wake curvature in forward ight by modifying the gain matrix. Further extensions to the model introduced in the augmented Pitt and peters model allow the wake skew, wake curvature and wake spacing to be included as extra states in transition from hover to forward ight. These extensions are not so relevant for the propeller case, and will be disregarded in this document. The Peters-He model is a notable improvement on the Pitt-Peters model in its formulation; the dierence noted in (Murakami, 2008, Sec 2.3). Murakami notes, and it has also been found by this author, that little comparison is made between the Pitt-Peters and Peters-He model in the literature aside from a couple of small examples which refer to the performance of the two rather than their formulation. Rather, this appears to be the case with a lot of the work of Peters et al., that the models developed show promising correlation, but details on the theory behind are provided only loosely, and little insight given into their respective formulations. The Peters-He model, whilst providing good correlation with forward ight data for rotorcraft, is not without its limitations or problems. The limitations of the Peters-He model are laid out by Morillo (2001) in his doctoral thesis, A Fully Three-Dimensional Unsteady Rotor Inow Model from a Galerkin Approach. The limitations as listed are: 1. The ow in uid volume above the rotor disc is calculated in the axial direction only. The ability to calculated all three components of the ow would be benecial in many modelling situations; e.g., ground eect or wing-propeller interference. 2. The model is unable to simulated mass-injection at the rotor. This point is insignicant for the propeller problem and refers to additions such as tip-jets or blade slats capable of blowing/sucking. 3. The model relies fundamentally on the superposition of pressure principle. This is a form of linearisation in which the total pressure eld is assumed to be formed of a superposition of the static and unsteady pressure of a eld. Morillos work forms the Morillo-Peters model, and remedies the above aws in the Peters-He model. By starting with the conservation equations of mass and momentum, Morillo utilises a Galerkin approach to form a more generalised expansion series for the velocity and pressure elds. This allows all three components of the velocity modes to be known in the innite hemisphere above the rotor disc (i.e., on-disc and o-disc ow using n + m Even Legendre coecients - the theory will be expanded later if used, but may be found in details in his thesis.), as opposed to solely the axial direction as with the Peters-He model. By expanding the velocity potential function, , the three-dimensional velocity vector is then given by its gradient, . It should be note that although the nal form of the Peters-He and Peters-Morillo model are the same, the derivation is quite dierent (Makinen, 2005; Murakami, 2008).
11

( 1, 2Z + 1), E

( 2Z), i.e., the sets of odd and even numbers, respectively

18

The Peters-Morillo model overcomes the aforementioned limitations of the Peters-He model, successfully extending the validity of dynamic inow simulations into situations involving mass source terms (e.g., tip-thrusters), and including the capability to calculate all three components of velocity above the rotor disc. 4.1.4 Dynamic Inow for propeller applications

The preceding development of dynamic inow models has been rotorcraft-focussed, but the methodology has found usage in wind-turbine analysis - see (Suzuki, 2000), for example. The extension to wind turbines is fairly simple, and at rst glance, the use of dynamic inow on the propeller problem seems logical. The possible problems that are foreseen and for which solutions must be found/assumptions validated and extended are: The use of the Euler equations in the derivation of the dierent incarnations of dynamic inow requires the compressible ow assumption to be invoked. On a helicopter rotor, although the local mach number at the blades may be 1, dynamic inow is a disc-level model, so it is only the compressibility in the wake that matters. Since typical rotorcraft ight is M < 0.3, compressibility may be safely invoked. For the propeller problem, the freestream Mach number may be 0.4 < M < 0.5 in a given ight, so justifying the incompressibility assumption is more dicult. Discussion of this point shall be elaborated in the following section Lift distribution - the lift distribution utilised in the Peters-He model (and others) assumes a simple planform and a radial lift distribution of the associated Legendre functions. The shape of propeller blades is very dierent from the typical helicopter blade, and the eect of this on the formulation of the model may need to be considered - possibly rederivation of the component matrices of the model. Blade twist - the average helicopter rotor blade has much less twist than the typical propeller blade. As such, the assumption that the lift acts in purely normal direction to the disc may be warranted for a rotor blade. This is less easily justiable for the propeller case - Makinen (2005) attempts to include dierent weighting functions to account for this on a dynamic inow model for tilt-rotor aircraft in propeller mode. Discussion of his methodology and results shall be included later in this section. Swirl velocity - the dierent dynamic inow models neglect to account for losses due to imparting a swift velocity to the wake, as the thrust is assumed to act purely axially. Again, Makinens thesis attempts to compensate for this factor by changing the denition of the equivalent air mass accelerated from an axial column to an cylindrical annulus - this point is a corollary, really, of the previous problem of purely axial lift vs. canted lift from a twisted blade. The relative eects of the preceding on this problem have been largely unexplored in the literature. What follows herein is an attempt to lay out the key research areas that will need to be tackled before condence may be gained in using dynamic inow for the propeller problem. Wake compresibility: As mentioned above, the freestream velocity for a turboprop aircraft will be surpassing the upper Mach number where compressibility may be typically neglected. Since the use of the Euler equations 19

is integral to the development of the dierent incarnations of dynamic inow, due to the decoupling of the uid equations of motion, rederivation using fully compressible uid equations is beyond the scope of this project. This leaves us with two options; either adapt the current dynamic inow models to incorporate compressibility eects OR justify using the model without compressibility eects, and include in the lift model instead. Of the two alternatives, the latter is preferable since adaptation of the dynamic inow model to include compressibility eects will require changing the fundamental equations on which it is based. To justify this approach, we need to look at what eect the dynamic inow model will have on the overall lifting model. As shown in gure 2.1, the dynamic inow model feeds back the induced ow to the aerodynamics model. More specically, the blade aerodynamic model (likely to be a blade-element model) takes in both the freestream and induced velocities and hence the relative angle of attack. The induced velocity component is likely to be < 10% of the total inow velocity for a helicopter in edgewise ow and respectively much less for a propeller at small AoA, and hence any changes due to compressibility in the inow will be small. Additionally, we want to nd the eect of dynamic inow on the 1p-loading, and any changes due to compressibility are likely to be relatively uniform over the disc. The qualitative azimuthal (and, less crucially, radial) variation in ow is likely to show the same trends with and without the incompressibility assumption. Further exploration of the above will need to be performed, but it is preliminarily proposed that the dynamic inow model may be used without adaptation to incorporate compressible ow eects. Validation of this assumption will need to be performed, and means of this need to be determined. Fundamentally, the performance of the model needs to be compared against Mach number, and hence the suitable range of the model calculated. The steady aerodynamic model is best suited to a blade-element model with sectional properties utilised as functions of angle of attack, density and velocity - if compressibility eects are included in this part of the model, then neglecting them in the inow model is hoped will be of negligible deteriment. Lift distribution: The Peters-He/Peters-Morillo models need to be worked-through from a propeller standpoint. Assumptions made in the shape of the blades and how they relate to the lift distribution need to be re-evaluated for the propeller case. Twisted, scimitar-type blades of a propeller are very dierent from the rectangular planform assumed in the dynamic inow literature available. Since dynamic inow is a disc-level model, the shape of the blade itself is unlikely to be of issue, but the radial lift distribution needs to be reassessed for the propeller case. Blade twist and swirl velocity: As highlighted by Makinen (2005), the spanwise distribution of twist for a propeller aects the direction of the thrust vectors used in the dynamic inow model. It is presumed in the various Peters models that the thrust acts purely axially and hence the equivalent mass upon which the thrust acts at

20

each radial station12 may be visualised as a vertical column of air. Since the propeller is highly twisted at the inboard sections, this approximation is not valid - Makinen proposes that the mass matrix is modied to account for this non-axial component, and the swirl velocity it imparts by modelling the thrust acting on a swirling cylinder of air as shown in Figure 4.3 on page 21.

(a) Vertical column of apparent uid mass for axial velocity component

(b) Cylindrical annulus of apparent uid mass for swirl velocity component

Figure 4.3: Comparison of original apparent uid mass and propeller modication to a cylindrical annulus, as proposed by Makinen (2005)

Makinens work attempts to modify the Peters-He and the Peters-Morillo model by adding weighting functions to account for the swirl velocity. He attempts to match the ideal solutions of Goldstein and Prandtl, which is quite a dierent aim to the problem at hand. For accurate predictions of 1p-loading, we are not so interested in the circulation distribution and how it tallies with the ideal case, rather the salient point is the azimuthal variation of inow as this is what will lead to the largest bending stresses and thereby the most signicant 1p-loads, and the radial distribution is less of an issue. The key ndings of his work are that it is the presence of physical phenomena at the inboard blade region that causes the dynamic inow model to dier from Goldstein and Prandts optimal solutions. Inclusion of weighting factors to account for the inertia of the inboard swirl velocity component allows, in his view, the applicable range for the dynamic wake/inow model to be expanded whilst retaining the same form. As previously stated, though, his aims are dierent to that of this project - the ideal distribution or, indeed, complete calculation of the radial distribution is less important for 1p-loading. Additionally, Makinen nds that the largest discrepancies from the exact theoretical distribution are at the blade root - where aerodynamic forces have a much smaller moment arm than those at the blade median and outboard sections, where his work shows that the dynamic inow solutions show much better agreement with exact solutions.
12 nb: the dynamic inow model does not integrate along the blade radius, but for the purposes of this discussion it is easier to talk about variation of thrust direction/apparent mass along the blade radius

21

Interestingly, Makinen doesnt mention the relevance of the incompressibility assumption in his thesis - it doesnt even come into discussion during the derivation of the original Peters-He model. In this authors view, this should be taken as an indication of how much of an undertaking it would be to include compressible ow in the dynamic inow model, rather than an acceptance of the incompressibility assumption for the propeller case. In summary - there is much work to be done in proving the suitability of dynamic inow for the propeller problem. Owing to the specications from Dowty; that the model be simple, platformindependent and computationally inexpensive, there is little alternative. Alternative state-space form models, (e.g., blade-element unsteady models (Leishman and Nguyen, 1990)) are unfavourable as they are based on numerical approximations to data and amount to curve-tting rather than a reliance on actual ow physics. 4.1.5 Dynamic Inow Summary

The discussion in the preceding section gives a grounding in dynamic inow theory and a brief insight into its application for propellers as an extension from its original rotorcraft origins. What is clear is that there is a lack of research into the application of dynamic inow to the propeller planform, and the method needs to be worked out from rst principles to ensure that it is valid and applicable to our problem. The timeline of where this ts into the project is shown in Appendix A - it is proposed that upon returning from placements at ARA and DP, eort is concentrated on theoretical work and a full rederivation of the Peters-He focussed towards the propeller planform and the aforementioned problems is performed. Murakamis work shall be integral to performing this bookwork.

22

4.2

Nonlinear Aerodynamics - Dynamic Stall

The preceding discussion of propeller aerodynamics is based around an attached-ow assumption such theory will be capable of modelling transient behaviour, but will be unable to model the large load that will occur if and when the ow separates. Dynamic stall occurs on an aerofoil section undergoing rapid changes of angle of attack; the rapid change causes a strong vortex to be shed from the leading edge and travels rearward, increasing the lift produced on the aerofoil. When the vortex is shed, lift is lost and the aerofoil is stalled. Such rapid oscillations are commonplace on rotary wings owing to the aerodynamic environment - particularly on rotorcraft as sections of reversed ow are present on the rotor disc at any signicant forward ight speed. Dynamic stall is an inboard blade phenomena owing to the lower tangential velocity component, and the outboard section13 will continue to produce lift - this leads to a (torsional) shear stress in the blade. Loads due to dynamic stall are likely to be some of the highest encountered on the blade and accurate prediction of when and where dynamic stall occurs and the resultant change to lift distribution will be necessary for the proposed model. However, since dynamic stall is less likely to occur on a propeller owing to reduced chance for reversed ow, it will be neglected entirely in the initial model and work will continue with a fully-attached assumption. Research into the various models for predicting dynamic stall will continue, though, and a survey of the pertinent rotorcraft literature provides many sources that shall prove useful for this task such as the NASA Rotorcraft Report by Bousman (2000), which lists an array of dynamic stall models and the relevant source literature.

inboard/outboard is an arbitrary distinction here; the region of stall will grow larger and include include more outboard blade as rotational speed is reduced, forward speed is increased (for nonzero angle of attack/sideslip for xedwing).

13

23

Propeller Structural Dynamic Model

This section outlines the research that has been undertaken for the propeller structural-dynamic model, and the proposed scheme of work/progression of model delity that has been formulated. The work to produce a structural-dynamic model for a propeller is less groundbreaking than implementing a dynamic inow model for a propeller, but is of equal importance and has been treated accordingly. It is as yet unknown what structural-dynamic models are available at ARA/Dowty, their capabilities and techniques. It has been presumed that some Finite-Element model is in use, but whether this is going to be feasible to couple with a dynamic inow model is unknown - these issues shall become clearer during the placement work in July. Nonetheless, it remains prudent to build a structural dynamic model from the ground up, ensuring that theory is fully understood and correctly implemented with the remainder of the model. As with all models, it is common sense to start building with the simplest model and add capability step by step - this ensures that problems and compatibility issues with interoperating theories can be isolated with ease if necessary. An excellent background text for this problem is written by Bielawa (2006); Rotary Wing Structural Dynamics and Aeroelasticity (Second Edition). This text whilst is, again, more rotorcraft-focussed, has chapters of interest e.g., instabilities of rotor pylon systems which is highly relevant for propellers, and the general concepts of a structural dynamic model rotor may be applied to the propeller problem without any signicant modication foreseen except, hopefully, some simplication due to the fewer degrees of freedom that a propeller blade has compared to a rotor blade. Based on general background reading and knowledge from undergraduate courses, the sensible way to create the structural dynamic model(s) is in the following order:

5.1

Lower order models: rotating beams

Single DoF Model () - This is be a simple model of blade rotary dynamics, without incorporating bending/torsion/lead-lag DoFs. This model allows simply to determine the variation of blade element tangential and normal velocities and therefore the elemental angle of attack and how they all vary with forward speed, a/c and a/c . 2 2DoF beam models - Using simple beam theory, models shall be developed separately to account for blade apwise-bending and torsion about the blade span. Lead-lag bending is not foreseen to have a signicant eect and will not be considered at this stage of analysis. The aeroelastic deformation of the blade shall be modelled using the Rayleigh-Ritz method of assumed modes - pertinent literature used/to be used will be listed later. 3DoF beam model - The full equations of motion for a beam subject to bending and torsion may easily be derived from the previous work. 3DoF beam model with pre-twist - The models prior to this assume a uniform, isotropic and homogenous beam - the rst assumption to be removed is the uniformity of the blade.

24

Development of these models to include a pre-twist rate along the blade span is not largely involved. There are numerous works on the development of such rotating beam models, many focussing on Timoshenkos theory of beam-bending (Weaver Jr and Timoshenko, 1974). A good narrative of the development of blade equations of motion is presented by Zhu (2010) and will be of use during this project. Development of beam-based models up to this point does not allow modelling of complex geometry - indeed, the scimitar-type blades that the model shall be used for incorporate high sweep, highly nonuniform chord and anisotropic structural properties owing to composite construction. Although the theory presented up to this point does not allow for modelling of such characteristics, it should be accurate enough to determine the eect of using a dynamic inow theory as opposed to a fullyuniform inow method. As shall be discussed in the following section, the eect of using dierent inow models will need to be comparable at all incarnations of the model, and it will be advantageous to determine the dierences with reduced-order structural dynamic models to eciently determine any compatibility issues between the theories involved.

5.2

Beyond beam models: Finite Element Analysis

As mentioned previously, is is expected that FE models are available for the various propellers, but the complexity of using these models in conjunction with a dynamic inow and aerodynamic model created can not be ascertained until both more work has been done on producing a workable preliminary model of aerodynamics and also when more information is available from ARA/Dowty. Nonetheless, FE theories have been researched for propeller application - plenty of applicable work has been performed and documented in the relevant literature. series of publications from the Israel Institute of Technology (TECHNION) details a nite-element methodology for aeroelastic analysis of propellers (Yadykin et al., 2006, 2004, 2005), starting with simple beam methods and following through into using a nite-element model based on the theory of Kosmatka and Friedmann (1987). Kosmatka and Friedmann (1987) provide the theory for FE analysis of advanced composite propellers with the aforementioned non-uniformities and anisotropic structural relationships. This work builds upon rotorcraft theory to develop a propeller model, in what appears to be a common theme for recent works concerning advanced propeller analysis.

5.3

Structural Dynamic Summary

The basic outline of propeller structural dynamic theory required has been laid out in the preceding section, and the dierent theories that it is proposed are used are listed. What remains is both application of the work and further investigation into FE models and the feasibility of such techniques to be included in the larger overall propeller model.

25

Owing to the timescales involved and the dependence of gaining condence in simpler modelling techniques before moving on to FE modelling, the following tasks may be performed concurrently: Successive formulation, validation and analysis of the beam models mentioned in section 5.1. Ongoing survey of propeller/rotor FE literature - research on this is more widespread than dynamic inow research, which is largely conned to Peters colleagues. It shall be perspicacious to remain up-to-date with any new developments, as formulation of the full model is some time away. Review of ARA/Dowtys current structural-dynamic modelling techniques.

26

Proposed Initial Scheme of Work/Model Construction

Based on the preliminary survey into the relevant literature summarised in the preceding sections, it is proposed that a sequence of models increasing in complexity and delity are constructed - a brief outline of the progression of this development is outlined below.

6.1

Initial Model Formulation

The preliminary models will be very simple - work has already begun on their formulation in MATLAB. In order to gain familiarity with the propeller operating environment, a single degree of freedom model has been developed. This model is able to calculate the azimuthal and radial variation of tangential and normal velocity as a function of aircraft angle of attack () and sideslip () - hence, by coupling with a suitable aerodynamic strip-theory model, the variation of CL and CD may be calculated and hence F (, r) can be determined for a given angle of attack and sideslip, assuming a uniform inow. Although simplistic and not analytically valuable, this model gives a good indication of the qualitative eect upon what the propeller disc sees with variation of freestream velocity direction. Before going into detail on the progression of individual models it is appropriate, here, to discuss the general code philosophy; how it shall be construction - which programming languages shall be used, data inputs and outputs (for subcomponents and for the total model). It is proposed that for the initial development of the mathematical model, the pre-industrialisation phase, the programming is kept in MATLAB. This is due to personal familiarity with the software, simplicity and capabilities for data presentation and exibility of data import/export to and from other software. Owing to the nature of this problem i.e., repeat calculations through azimuthal and radial discretisation (and, later, time) it will be prudent to utilise vectorisation of calculations rather than concentric for loops - this is simple to perform in MATLAB, but needs further research of the relevant literature for FORTRAN. Each subcomponent of the model shall be formulated as a fully-functioning function in MATLAB. That is, no global variables shall be shared between the Dynamic Inow model, the Aerodynamic Model or the Structural-Dynamic model. Starting with a given aircraft angle of attack, sideslip and the propeller operating conditions, simple momentum theory shall calculate the steady induced velocity and the steady aerodynamic model shall determine the lift/pressure distribution as an output. This is then used by the dynamic inow and structural dynamic models to determine the unsteady induced velocity and this eective change to angle of attack as a function of blade span, respectively as outputs. These are then input back into the steady model at each time step. By keeping model subroutines separate, this shall hopefully allow for simple modication to incorporate dierent model versions - e.g., if at a later date it is deemed suitable to add in a nite-element model, this may be incorporated without modifying the dynamic inow model.

27

6.2

Initial Aerodynamic and Dynamic Inow Models

It is proposed that the model may be adapted to incorporate 3/4 dierent dynamic inow variants: Pitt-Peters original model from Section 4.1.2 Peters-He model from Section 4.1.3 Peters-Morillo model from Section 4.1.3 shall not be included, as the benets it holds over the Peters-He model are not applicable to propellers. Variants of He/Morillo models incorporating Makinens weighting functions from Section 4.1.4 The nal model must be able to determine the eect of incorporating dynamic inow at all levels, and hence this functionality must be built into the model from the outset - i.e., the ability to vary only the inow model and keep all other elements of the model functionality the same must be possible. The aerodynamic model is to be kept simple and easily adaptable - a simple strip-theory model based on a variable-width but spanwise-constant commonplace aerofoil prole e.g., Clark-Y will be ideal. This will allow for calculation of lift and drag as a function of Mach number and angle of attack. Use of a simple public-domain prole means that the amount of data readily available is greater than that for an ARA-D prole but also, perhaps more importantly for both my own academic aims but also for the University of Glasgow, it means that the data produced shall be easier to publish in relevant journals14 This model and variants thereof will be termed the baseline models and dierentiated in code and literature by baseline pitt, baseline he etc. . These model(s) will allow for comparison of dierent dynamic inow models and fully uniform inow on lift distribution in radial and azimuthal directions without the incorporation of: Structural dynamic eects e.g., (s), (s) - these shall be incorporated by adding bending and torsion degrees of freedom initially, as shall be explained below. Aerodynamic nonlinearities e.g., dynamic stall - before incorporating the eect of these, a fullyattached model needs to be validated. Inclusion of a dynamic stall model such as the ONERA model will be considered at a later date once condence in the model reliability is determined. Wing-propeller interaction - this is (hopefully) less complicated than the preceding two eects. Ideally, variation of upwash/downwash due to the wing as the aircraft changes angle of attack and sideslip will be incorporated. A good start for the eect of wing-propeller interference is available in (Witkowski and Lee, 1989). Raw data for this should be readily available from sources such as (Doring, 2008) investigating the ight environment of propellers, although CFD simulation may be suitable to determine continuous variation with ac and ac .

Although any publication resulting from the work in this project shall be subject to the review procedure as outlined by ARA/Dowty.

14

28

6.3

Structural Dynamic Model addition

Although the theory behind the propeller structural dynamics and formulation of a mathematical model thereof is not technically overwhelming, it is hoped that the addition of blade structural dynamics can wait until the dynamic inow model is successfully formulated. It makes no sense to formulate the entire model at the outset, and it will be sensible to determine the eect of dynamic inow without incorporating blade dynamics initially. With this in mind, the following progression of models is suggested. Similarly to the dynamic inow model(s), it shall be prudent to create and test this model in isolation; validating a structural dynamic model for the uniform inow case before incorporating into the nal overall model. It is proposed that the initial structural-dynamic model is adapted by the inclusion of simple, low-order structural-dynamic eects. As mentioned previously, the rst incarnations of the model shall be utilising simple beam-theory incorporating the Rayleigh-Ritz method of assumed modes/trial functions for spanwise bending and torsion e.g., 2/3 polynomial bending modes and a single linear torsion mode. Models for bending and torsion shall be developed independently and combined by linear superposition to create a model for combined bending and torsion. Theory for this is that as taught in (Gillies, 2009), and the Rayleigh-Ritz method has been successfully applied to rotating Timoshenko beams by Zhu (2010) to develop a validated model for a homogenous, isotropic beam with linearly varying spanwise twist. Timoshenko theory of beam bending is intended to be employed for the standard form of beam equations of motion - theory taken from Vibration Problems in Engineering (Weaver Jr and Timoshenko, 1974).

29

Testing and Validation

Although the model is far from being at the operational stage, the testing and validation of the individual mathematical models, as well as the overall propeller model needs to be considered. In the early stages, it will be desirable to have means to qualitatively evaluate the performance of the code i.e., are the maxima/minima/discontinuities of variation of loading in a sensible azimuthal position? Are the general trends correct with changes to and ? To be able to answer these questions, access to relevant literature is necessary. Although this has not been covered largely in the literature survey, it is hoped that data at ARA and DP may help for this - if not, more intensive search for pertinent papers needs to be performed over the coming months (and for the remainder of this project). In terms of proving the potential of dynamic inow for this purpose, the following needs to be considered. Fundamentally - how can the functionality of dynamic inow to predict 1p-loading be compared both to validated data and to other methods? The following questions/points need to be answered/considered over the coming months. This activity may be researched concurrently with model formulation. What data is available? (e.g., 1p FT/Wind Tunnel data) This shall hopefully become clearer once access to documents at ARA and DP is available. What other data is available that is publishable? What restrictions will be placed on publication of results comparing data from ARA-D aerofoil banks? Possibility of comparing dynamic inow with CFD methods (e.g., Brown et al.s Vorticity Transport Method (VTM)-based code - currently being utilised in UoG. - alternatively compare with NUMECA?) Comparison of the dynamic inow model to both CFD and experimental data will be advantageous. It will be ideal to show the trade-o (if any) of accuracy vs. computational cost of dynamic inow compared to free-wake methods.

30

References
Amer, K. . (1950). Theory of Helicopter Damping in Pitch or Roll and Comparison with Flight Measurements. NACA. Banerjee, D. and Crews, S. (1979). Parameter identication applied to analytic hingeless rotor modeling. Journal of the American . . . . Beddoes, T. S. (1980). Unsteady Flows Associated With Helicopter Rotors. Agard Report 679. Bielawa, R. L. (2006). Rotary wing structural dynamics and aeroelasticity. AIAA Education Series. Bousman, W. (2000). Airfoil Dynamic Stall and Rotorcraft Maneuverability. Technical report, Moett Field, California. Bramwell, A. R. S., Done, G., and Balmford, D. (2000). Helicopter Dynamics. Elsevier Ltd., second edition edition. Carpenter, P. and Fridovich, B. (1953). Eect of a rapid blade-pitch increase on the thrust and induced-velocity response of a full-scale helicopter rotor. Chen, R. T. N. (1989). A survey of nonuniform inow models for rotorcraft ight dynamics and control applications. Technical report, Ames Research Center, Moett Field, California, Ames Research Center, Moett Field, California. Chen, R. T. N. and Hindson, W. S. (1986). Inuence of Dynamic Inow on the Helicopter Vertical Response. Technical report, Ames Research Center, Moett Field, Caliornia, Ames Research Center, Moett Field, Caliornia. Coleman, R. P. and Feingold, A. M. (1945). Evaluation of the Induced-Velocity Field of an Idealized Helicoptor Rotor. Technical report, National Advisory Committee for Aeronautics, National Advisory Committee for Aeronautics. Curtiss, H. K. and Shupe, N. K. (1971). A Stability and Control Theory for Hingless Rotors. In Annual Forum of the American Helicopter Society, Washington, D.C. Doring, J. (2008). Flight environment of the propellers on commuter aircraft. soar.wichita.edu. dg. Gaonkar, G. H. and Peters, D. A. (1986). Eectiveness of Current Dynamic Inow Models in Hover and Forward Flight. Journal of the American Helicopter Society. Gaonkar, G. H. and Peters, D. A. (1988). Review of Dynamic Inow Modeling for Rotorcraft Flight Dynamics. Vertica, 12(3):213242. Gaonkar, G. H., Sastry, V., Reddy, T., Nagabhushanam, J., and Peters, D. A. (1983). The Use of ActuatorDisc Dynamic Inow for Helicopter FlapLag Stability. Journal of the American Helicopter Society. Gillies, E. A. (2009). Aircraft Vibration and Aeroelasticity 5. In Course Notes: Fifth Year Aeronautical Engineering, pages 114, University of Glasgow, Department of Aerospace Engineering. 31

Glauert, H. (1926a). A General Theory of the Autogyro. Technical report, Scientic Research Air Ministry, Scientic Research Air Ministry. Glauert, H. (1926b). The Elements of Aerofoil and Airscrew Theory. Cambridge University Press. Glauert, H. (1943). Chapter VII, Div. L: Airplane Propellers. In Durand, W. F., editor, Aerodynamic Theory: A General Review of Progress. California Institute of Technology, Guggenheim Fund for the Promotion of Aeronautics. Goldstein, S. (1929). On the vortex theory of screw propellers. In Proceedings of the Royal Society of London., pages 440465. Joglekar, M. and Loewy, R. (1970). An Actuator-Disc Analysis of Helicopter Wake Geometry and the Corresponding Blade Response. Technical report, US Army Air Mobility Research and Development Lab (USAAVLABS), Fort Eustis No. 69-66, US Army Air Mobility Research and Development Lab (USAAVLABS), Fort Eustis No. 69-66. Kinner, W. (1937). Die kreisfrmige Tragche auf potentialtheoretischer Grundlage. Archive of o a Applied Mechanics. Kosmatka, J. B. and Friedmann, P. P. (1987). Structural Dynamic Modeling of Advanced Composite Propellers by the Finite Element Method. In . . . , pages 112124, Monterrey, Canada. AIAA. Larrabee, E. E. (1979). Practical Design of Minimum Induced Loss Propellers. Technical report, Warrendale, PA. Leishman, J. G. (2000). Principles of Helicopter Aerodynamics. Cambridge Aerospace Series. Leishman, J. G. and Nguyen, K. Q. (1990). State-Space Representation of Unsteady Airfoil Behavior. AIAA, 28(5):836844. Makinen, S. (2005). Applying dynamic wake models to large swirl velocities for optimal propellers. PhD thesis, Washington University Sever Institute of Technology Department of Mechanical Engineering. Mangler, K. W. and Squire, H. B. (1950). The induced velocity eld of a rotor. Technical report. Morillo, J. (2001). A Fully Three-Dimensional Unsteady Rotor Inow Model from a Galerkin Approach. PhD. Thesis, pages 1241. Munk, M. (1924). Some tables of the factor of apparent additional mass. Murakami, Y. (2008). A New Appreciation of Inow for Autorotative Rotors. PhD thesis, University of Glasgow Dept. of Aerospace Engineering. Nagabhushanam, J. and Gaonkar, G. H. (1984). Rotorcraft Air Resonance in Forward Flight with Various Dynamic Inow Models and Aeroelastic Couplings. Vertica, 8(4):373394. Ormiston, R. A. (1976). Application of Simplied Inow Models to Rotorcraft Dynamic Analysis. Journal of the American Helicopter Society, 21(3):3437.

32

Ormiston, R. A. and Peters, D. A. (1972). Hingeless Helicopter Rotor Response with Non-Uniform Inow and Elastic Blade Bending. Journal of Aircraft, 9. Peters, D. A. (1974). Hingeless rotor frequency response with unsteady inow. NASA Ames Research Center Rotorcraft Dynamics: SP 352, pages 112. Peters, D. A., BOYD, D., and HE, C. (1989). Finite-State Induced-Flow Model for Rotors in Hover and Forward Flight. Journal of the American Helicopter Society, 34(4):517. Peters, D. A. and Gaonkar, G. H. (1980). Theoretical FlapLag Damping with Various Dynamic Inow Models. Journal of the American Helicopter Society. Peters, D. A. and HaQuang, N. (1988). Dynamic Inow for Practical Applications. Journal of the American Helicopter Society. Peters, D. A. and He, C. J. (1995). Finite State Induced Flow Models Part II: Three-Dimensional Rotor Disk. Journal of Aircraft, 32:111. Pitt, D. M. and Peters, D. A. (1981). Theoretical Prediction of Dynamic-Inow Derivatives. Vertica, 5(1):2134. Pitt, D. M. and Peters, D. A. (1983). Rotor dynamic inow derivatives and time constants from various inow models, 9th European Rotorcraft Forum. In 9th European Rotorcraft Forum, page 216, Stresa, Italy. Stresa. Prandtl, L. and Betz, A. (1927). Schraubenpropeller mit Gerinstem Energieverlust, Gttinger o Nachrichten, Gttinggen. Vier Abhandlungen uber Hydro- und Aerodynamik, pages 193197. o Rankine, W. J. M. (1865). On the mechanical principles of the action of propellers. Transactions of the Institute of Naval Architects. Sissingh, G. J. (1951). The Eect of Induced Velocity Variation on Helicopter Rotor Damping in Pitch or Roll. Aeronautical Research Council Technical Report - CP No. 101. Suzuki, A. (2000). Application of dynamic inow theory to wind turbine rotors. PhD thesis, Dept. of Mechanical Engineering, The University of Utah. Theodorsen, T. (1944a). The Theory of Propellers I: Determination of the Circulation Function and the Mass Coecient for Dual-rotating Propellers. NACA Technical Reports. Theodorsen, T. (1944b). The Theory of Propellers II: Method for Calculating the Axial Interference. NACA Technical Reports. Theodorsen, T. (1944c). The Theory of Propellers III: The Slipstream Contraction with Numerical Values for Two-blade and Four-blade Propellers. NACA Technical Reports. Theodorsen, T. (1944d). The Theory of Propellers IV: Thrust, Energy, and Eciency Formulas for Single and Dual Rotating Propellers with Ideal Circulation Distribution. NACA Technical Reports. Theodorsen, T. (1948). Theory of propellers. 33

Tibery, C. L. and Wrench Jr, J. W. (1964). Tables of the Goldstein factor. David Taylor Model Basin, Report 1524, Applied Mathematics Laboratory, Washington, DC, page 69. Wald, Q. R. (2001). The Wright Brothers propeller theory and design. Proceedings AIAA Joint Propulsion Conference, Salt Lake City. Wald, Q. R. (2006). The Aerodynamics of Propellers. Progress in Aerospace Sciences, 42(2006):88128. Weaver Jr, W. and Timoshenko, S. P. (1974). Vibration problems in engineering, 1990. John Wiley. Wheatley, J. B. (1935). An Aerodynamic analysis of the autogiro rotor with a comparison between calculated and experimental results. Technical report, NACA, NACA. Witkowski, D. and Lee, A. (1989). Aerodynamic interaction between propellers and wings. Journal of Aircraft. Yadykin, Y., Ognev, V., and Rosen, A. (2006). Aeroelastic Analysis of Propellers. Technical report, Faculty of Aerospace Engineering: Technion - Israel Instritute of Technology, Faculty of Aerospace Engineering: Technion - Israel Instritute of Technology. Yadykin, Y., Tenetov, V., Weissberg, I., and Rosen, A. (2004). Aeroelastic Analysis of Propellers. Technical report, Department of Aerospace Engineering - Technion, Israel Institute of Technology, Israel. Yadykin, Y., Tenetov, V., Weissberg, I., and Rosen, A. (2005). Aeroelastic Analysis of Propellers. Technical report, Faculty of Aerospace Engineering, Technion - Israel Institute of Technology, Faculty of Aerospace Engineering, Technion - Israel Institute of Technology. Zhu, T. L. (2010). The vibrations of pre-twisted rotating Timoshenko beams by the RayleighRitz method. Computational Mechanics, 47(4):395408.

34

Proposed Timeline

Figure A.4: Proposed Timeline

35

Potrebbero piacerti anche