Professional Documents
Culture Documents
by
compare with reasonably good accuracy. It was found that the Descnbing Function
down.
I would Iike to thank Peter Ba.rrîngton, of Carleton Universis., my Master's thesis
, has provided me with guidance and assistance through t h i s project. 1
s u p e ~ s o rwho
would also like to thank my CO-supervisorBen Lee of the Institute for Aerospace
ResearchoNational Research Council for his help and insight into this topic, and for the
use of the NRC facilities. I would also like to thank my third CO-supewisorDick Kind,of
Carleton University. for going to great lengths ensure this project drew to cornpletion and
not forgetting me. Thanks aiso to the people at Davis Engineering for giving me tirne.
facilities and support to complete the project. The support of d l these people is greatly
appreciated-
...
ABSTRACT .......................................................... i~i
1.0LNTRODUCTION ................................................... 1
1-1 Problem Definition .............................................. 1
1-2 Scope of Present Work .......................................... - 3
2.0BACKGROUND .................................................... 5
2-1 Previous Work ................................................ - 5
3.0THEORY ......................................................... 12
~.lEquationsofr\noti~n ........................................... 13
3 .1.1 Quasi-Steady Aerodynamics .............................. 18
3.1.2 Unsteady Aerodynamics ................................. 33
3.3SolutionMethods .............................................. 23
,
-
,
4OANGLYSIS ........................................................ 36
4.1 System Def-ition ............................................ - 3 6
4.3 Time Marching Simulation Solutions .............................- 3 7
3.3 Describing Function Solutions ................................... - 4 2
Figure Description
vii
5-1 Sinusoidal Type Pitch Response for Hysteresis Nodine-
Power Spectral Density of Pitch Response for Hysteresis
Nonlinearity
Pitch Response with Harmonic Structure for Hysteresis
Noniinearity
Power Spectral Densi- of Pitch Response for ETysteresis
Nonlinearïty
Results Plots for Hysteresis Nonlinearïty R u 06 fiom Table 1.1
Res1dts Plots for Hysteresis Nodinearity Run 10 fiom Table 4.1
Results Plots for Hysteresis Nonlinearity Run 14 fiom Table 4.1
Results Plots for Hysteresis Nonlinearity Run 15 fiom Table 4.1
Resdts PIots for Hysteresis Nonlinearity Run 16 fiom Table 4.1
Resdts Plots for Hysteresis Nonlinearity Run 100 fiom Table 4.1
Results Plots for Hysteresis Nonlinearity Run 2 0 1 fiom Table 4.1
Results Plots for Hysteresis NonIinearity Run 102 fiom Table 4.1
Results Plots for Hysteresis NodineanS. Run 103 fiom Table 4.1
Results Plots for Hysteresis Nonlinearity Run 1O4 fiom Table 4.1
Results Plots for Hysteresis Nonlinearity Run 105 fiom Table 4-1
Results Plots for Hysteresis Nonlinearity Run 110 fiom TabIe 4-1
Results Plots for Hysteresis Nonlinearity Run I l 1 from Table 4-1
Rcsults PIots for Hysteresis Nonlinearity Run 25 fiom Table 4.1
Results Plots for Hysteresis Nonlineariq Run 27 fiom Table 4.1
Results Plots for Hysteresis Nonlinearity Run 28 fiom Table 4.1
Results Plots for Hysteresis Nonlinearïty Run 29 fiom Table 4.1
Results Plots for Hysteresis Nonlinearity Run 30 from Table 3.1
Results Plots for Hysteresis Nonlinearity Run 3 1 fiom Table 4.1
Results Plots for Hysteresis Nonlinearity Run 32 fiom Table 4.1
Results Plots for Hysteresis Nonlinearity Run 33 Eom Table 4.1
Resuits Plots for Hysteresis Nonlinearity Run 120 fiom Table 4.1
Results Plots for Hysteresis Nonlinearity Run 121 from Table 4.1
viii
5.28 azb,c Results Plots for Hysteresis Noniinearity Run 130 £iom Table 4.1
5.29 ab,c Results PLos for Hyaeresis Nonlinearity Run 131 from Table 4.1
5-30 a,blc Results Plots for Hysteresis Nonlinearity Run 132 from T b l z 4.1
5.3 1 ab$ Results Plots for Hysteresis Nodine* Run 133 from Table 3.1
amplitude of input s i p a l to Describing Function a(t) = B + Asin(wt+e)
semi-chord
coefficient of lifi
Linear heave smictural damping coefficient per unit span, see Equation 3.2
linear pitch structurai damping coefficient per unit spul. see Equation 3.2
heave position
radius of 2 ~ t i o about
n elastic axis
pitch angle
d m p i n g term
phase mgle
ot+8
The work descnbed in thïs thesis was performed for the purpose of investigating
and to compare results between cornputer simulations and a semî-analytic solution. The
goal was to evaluate the effectiveness of using the descrÏbed semi-analy-tical solution
rnethod to predict the flutter behaviour of this nonlinear system and determine where the
rnethod's accuracy breaks down for the hysteresis nonlinearity and why.
The studied system represented a generic airfoil section that had been sirnplified
When subjected to large aerodynamic forces?i.e. airspeed, and an initial pitch angle the
dynarnic system can be shown to enter into a state of oscillation that afier the removal of
transient effects results in a system that is either stable and decays to zero, or is unstable
grow out of control is known asfitter and is a dangerous condition that c m cause violent
damage to aircrafi muctures. The airspeed at which flutter is induced in a linear
By introducing structural nonlùiearities into the elastic response of the airfoil' the
dparnics can be reevaluated with respect to the stability of the system and may
subsequenrly include steady state oscillations that are sustainable and neither decay nor
grow beyond the bounds of the system. Typicd nonlinearities that have been anaiysed in
theoretical work include thefieeplay nonlineanty where a dead band of zero change in
restoring force occurs over a range of displaceme. cubic nodinearity where the reacaon
nonlinearity where the reaction force response c u v e is dependent upon the cment state
of the system. It is uiis hysteresis type noniinearis which was snidied in the work being
forces and the reaction forces of an elastic structure. Aerodynamic forces deform an
elastic structure and the deformation of the structure in tum changes the nature of the
aerodynamic forces acting on it. In a dynamic system these forces can feedback ont0
each other causing reactions that are sometimes cornplex and dinicult to predict. The
feedback nature of the dynamic system can, under certain conditions, lead to instabilities
combine large aerodynamic forces with low structural stifbess. Specifically wing
sections and tail sections are susceptible to damage ffom aeroelastic innabilities. The
aerodynamic lifting forces can cause these relatively low stifiess components to
dynamically oscillate under loading. If the airspeed is great enough and certain stability
conditions are met the oscillations can increase to the point where failure of the part
occun. For the purposes of design, it is important to be abIe to predict when these
unstable conditions are met such that the structure can be designed to fly within the safe
operating envelope.
are performed through the use of finite elemenr type numencal problems that use linear
models to simulate the response of the structure to the aerodynamic loading. The f i t e
behaviour for the model generally models the response of the structure well but even
smdl nonlinearities. that are present in any real world smcture, can have a large
solution for the behaviour of the defined dynamic system. Simulations were performed
for a large rnatrix of airfoi1 configurations and initial conditions and the results were
The
amassed to characterize the response of each airfoi1 ~ o d ï ~ p a t i o n . semi-analfical
describing function method was also utilized to characterize the same airfoi1
configurations and initial conditions.
This thesis will present the results of the cornparisons betwen the hvo methods
for the hysteresis nonlinearity present in the pitch degree of fkeedom of the airfoil. Their
similarities are discussed and reasons for the differences and relative shortcomings are
explained The areas where the accuracy of the semi-andy-tical method breaks down are
examined and methods are proposed for predicting these conditions. Real-world
applications for the methods are discussed as well as topics for friture work
The importance of aeroelastic analysis has been realized since the early days of
aviation. This chapter presents an overview of the work that preceded the work described
in this thesis. Previous work that lead up to the presented thesis material included
experimenral and analyticai work with both linear aerosystems and structural
nonlinearïties-
were laid by Theodorsen and Fung who developed methods for characterizhg flutter
throiigh experiment and theor).. Theodorsen and Garrick (1 940) performed numerous
experiments using scale mode1 wing sections in the hi& speed wind tunnels of NACA.
His experiments cIearly showed the dangers involved with uncontrolled flutter. Both
Theodorsen and Garrick and Fung (1955) used linearizing assumptions for simpli@ing
the dynamic systems they studied. W l e these assumptions provided a basis for solution
of the systems it was evident that nonlinearities were still quite important for the resulting
behaviour of the airfoil. Woolston et. al. (1957) performed analyses using an analog
cornputer for a number of structural nonlineârities including a fkee play spring, hysteresis,
6
and cubic nonlinear restoring forces. They were able to correlate their results with wind
tunnel experiments. They made note of how the stabilis of the system was highly
dependent upon the initial displacement from equilibrium. Subsequeritly work was
presented by Shen and Hsu (1957) showing a cornparison and explanation of W-oolston*~
of aircrafi components. Nonluiear aerodynamics usually refer to transonic flows and are
not considered within this thesis. The importance of nonlinearities in the performance of
aircraft structures was broadly- outlined by Breitbach (1977) who gives an overview of the
types of nonlinear behaviours exhibited by aircrafi structures. His analysis was denved
partly from theory and p d y from observations and experiments. He classified the
noniinear effects as either being distributed noniineuriries, the nonlinear response of the
aircrafi structure due to the many srnall nonlinearities dispersed throughout the aircraft, or
concennated nonlinearities, which are locally acting on one particular degree of fieedom
of the structure. It is the behaviour of these concentrated nonlineanties that analytic and
numerical methods such as those presented in this thesis attempt to model. Concentrated
displacement can be in such forms as a dead band of response in the centre of the
displacement range (fieeplay) such as might be due to a loose control cable or a preload
materials as well as by solid fkiction in control cables and hïnge bearings. A dynamic
system with one of these nonlinearities in the restoring force or moment has the
change of the restoring force with displacement. would normally be a constant for a Iinear
system. The resulting system cm, piven a certain set of initial conditions, reach a state of
oscillation where the system response neither decays IO rest nor grows b e o n d the bounds
of the system. These Iimit cycle osciUations &CO) can occur at speeds si-onificantly
below the flutter speed predicted by the linear approximation, and the ampiinide can be
such as to affect the flight dynamics of the aircrafr or cause damage. Breitbach stressed
the need for nonlinear analysis methods in aeroelastic investigations as he found the
these problems and nonlinear systems in general. Dowell paid particular attention to the
aeroeIastic analysis of heIicopter rotor biades and for solution methods for nodinear
vibrations.
surfaces with structural nonlinearitïes. They found that the effective stifkess of the
resdting system was less than that for the corresponding linear system. This result leads
to the conclusion that nonlinearïties are likely to induce flutter at a lower speed than is
systems where the output of the system is first assumed to be harmonie, then an efecfive
stifiess for the systern is determïned. The resulting linear systern approximates the
nonlinear system such that the resulting magnitude of linear system hannonic response
for a prescribed set of conditions approaches that of the nonlinear system being modelled.
Gelb and Vander Velde (1968) gave a good description of the method and outlined how it
Many uses for the describing function technique for the purposes of modelling
nonlinear systems for linear anaiysis have been developed. C. L.Lee (1985) descnbed an
iterative procedure for making use of the describing function and how it was compared to
experiments and simulations for flutter analysis of large dynarnic systems. He provided a
good validation of the method for a range of nonlinearities of increasing complexity and
showed good agreement with the experimental and simulation data. mur^ (1995) used a
describing function method to solve systems with nonlineanty in the torsional degree of
freedom. Mast and Pierce (1995) made use of the technique for the analysis of nonlinear
fighter aircrafi. They used a describing function method to predict the flutter
characteristics of a CF-18 wing fold hinge. They showed that the nonlinearities in the
hinge could be modelled using springs with bilinear stiffnesses and fieeplay zones. They
9
used the descnbing function m e h d to predict the speed at which flutter would ensue and
solve the equations of motion using numerical rime rnarching methods. Modem
cornpuhg advances have made feasible the use of algorithms utilising millions of
and ~onfi~mations.
Jones and Lee (1985) showed how a single degree of fieedom
dynamic system known as D u f i g ' s equabon could be solved using a tirne marching
of a step by step solution for the aircrafi structural response. He utilised a recurrence
m a h k solurion to solve for the stepwise tirne based simulatioo of the aircraft response.
Jones and Lee were able to show that solving the Dufkg's equation system through a
sirnilar numerical means was both feasible and in some areas more accurate than previous
analytical work. A similar numerical technique was later used by Lee and Leblanc (1986)
nonlineanties. They used Houbolt's scherne to investigate the flutter response of the
airfoi1 to varying initial conditions. They f o n d that different initiai pitch angles could
cause die airfoil to achieve Iimit cycle flutter at speeds below the linear flutter speed. Lee
and Desrochers (1987) used the same numencal method to investigate airfoils with
restoring forces that contained bilinear and fieeplay nodinearïties. They numerïcally
determined the speed at which different airfoi1 configurations would achieve either limit
cycle or divergent flutter for differing initial pitch angles of the airfoil. The analysis was
IO
performed by simulating the motion of the system for each confi,guration and set of initial
conditions through a sufficient number of steps to ensure the system had reached ~eaciy
characterized the response of the airioil. Lee and Desrochers also invesrigated the
relationships between the amplitude of the limit cycle flutter and the simulated airspeed
for various confi_grations of the airfoil and found the amplitude to be independent of the
initial conditions,
They found that the response of the system depended greatiy upon the initiai conditions
and could range fÎom damped decay to unstable oscillations (flutter) to Limit cycle
response or chaotic motion. They drew a conclusion that the system would not
Price, Alighanbari, and Lee (1994) performed work extending the results of Lee
and Desrochers on two-dimensional airfoil systems with bilinear and cubic restoring
forces. They solved the systems using both the numerical tirne marching method and a
semi-analytical describing fimction method. They found the describing function method
was able to predict reasonably well both the speed of the onset of flutter and the
magnitude to which a b i t cycle flutter condition would grow Pnce e t al. were also
able to show that, c o n w q to the results of Hauenstein et. al., smdl regions of chaotic
motion could be obtained using single cubic or bilinear nonlinearitïes in the aeronirface
greaat nurnber of possibiIities to observe some of the more intriguing aspects of nonlinear
motion such as chaos. Stockard et- al. (1967) showed, using a nonlinear oscillator: how
could be observed. Yang and Zhao (1988) used a harmonic baiance method compared to
could be observed for the same airfoi1 confi_muation subjected to the sarne airspeed
condition. Subsequent work by Zhao and Yang (1990) detailed their obsewations of
chaotic responses in airfoil systems with cubic nonlinearities in the pitch degree of
fkeedorn.
The anaiysis and characterization of chaotic responses within the context of the
problem being exarnined herein is beyond the scope of the present thesis. It does
vibrating system with an aerody-nunic forcing term that is dependent upon the value of
the pitch degree of fieedom. The two degrees of motion as shown ui Fi*me 3-1 are
rotation about the elastic axis bitch: a) and vertical motion (heave, h). The system is
elastically constrained about each of the degrees of freedon where the restoring force for
the pitch degree of fieedorn is torsion and the restoring force for the heave degree of
fteedom represents what w-ould be rhe bending of a wing in a Ehree dimensional aircrafi.
The "motion induced" forcing terms result fiom aerodynamics forces evaluated according
to simple unsteady thin airfoil theory using the assumptions of incompressible inviscid
airflow. The aerodynamic forces on the system are in turn dependent upon the value of
the pitch angie a. the heave rate ahBr, and the prescribed airspeed. Classical solution
methods exist for both a coupled two degree of fieedom linear system and for the
presented U1this thesis to outline how the sûxctural and aerodynamic portions of the
The two degrees of freedom in the system are egectively coupled both due to the
s ~ ~ c t uofr sthe airfoil system and the aerodynamics. The aerodynamic response to the
pitch of the airfoi1 is lift. The change in lift causes both a change in heave and a
subsequent change in pitch due to inertiai effects. The changes in pitch and heave feed
back to a change in lifi and the resuiting system effectively couples the displacements,
forces and moments. The result is a two degree o f f eedom dyn-c system where the
This chapter will present the ecpaùons of motion that govern the described
dynamic system and develop them into a form that is suitable for solution both by
To derive the equations of motion that govem the two degree of freedom system
we begin with the equations of motion for a 2 degrees of fieedom dymarnic syslem which
where cc represents the airfoi1 pitch angle and h is the heave displacement at the elastic
axi-S. Note that the negative s i s on the Lifi term arises because h has been defined as
positive for downward motion. The structural properties of the airfoi1 are depicted in
Figure 3.1 showing the relative locations of the centre of rn- elastic a i s , and
aerodyrzarnic centre of the airfiil. The single and double dots over the variables indicate
the first and second derivative respectively with respect to time. S represents the static
(3-lc)
and m is the airfoil mass per unit span. Ch and C, zre the structural dampin, coefficients
elastic mis, and K,, is the heave spring stEness per unit span. Variables b, ah?and x, are
depicted in Figure 3.1 and represent the position of the midchord, and positions of the
elastic axis and centre of mass of the airfoi1 measured from the midchord respectively.
F(a)represents the nodinear function by which the pitch degree of freedom restoring
force is determined. This function acts as a nonlinear pitch degree of fieedom stifiess
detem-ning the restoring torque according to the value of the airfoi1 pitch. L(t) and M(t)
are the motion induced lifting force and moment arising f?om the aerodynamic forces
the pitch degree of fieedom represented by the spnng constant K, and zero initial
becomes:
In order to non-dimensiondize the te-, the foilowkg defuiitions were used:
where o,and ocrepresent the natural fiequencies of the pirch and heavc degrees of
freedom when the static offset S from Equation 3. lc is O. Additional definitions for the
velocity:
where V is the reference velociq and a, is the nanirai fiequency of the system in the
ratio of the natural fiequemies of the two degrees of fieedom according to:
Using this system of equations as the basis, solution methods can now be utilised
that approach the problem either in the time domain or in the fiequency domain. The
aendynamic side of the system also requires more refmement as it is not only dependent
upon time i but d s o upon the pitch angle a. The aerodynamics for the system c m be
approached in one of two ways. For the purposes of this thesis, an assumption of quasi-
neady aerodynamics was initially used to develop the solution method in a marner that
was easier to implement. Because of the time based nanire of the simulations, a solution
incorporating unsteady aerodynamics was then later implemented to provide for a more
accurate sohtion.
For the purposes of illustration only. the above systems (Equations 3 -9 and 3.10)
Because the mathematics involved in using this simplined definition of the aerodynamics
are more basic, the solution method will be outiined in this section using the quasi-steady
19
aerodymamics assumption. Details on defining the system using more accurate unsteady
aerodynamics tenns mil1 be oudined in the next section and c m then be substituted into
that the aerodynamic forces acting upon the aerosurface change instantaneously with
changes in airfoi1 pitch. For the purposes of illustrating the development of the mode1 of
the system, the quasi-steady assumption is cmde but valid. For the purposes of numencai
accuracy for subsequent simulations and analytical solutions the unsteady aerodynamics
Let the aerodynamic relationship of lifi (L) force and moment (M) torque to the
pitch and heave positions of the airfoil be represented by the aerodynarnic transfer rnatrïx:
niin airfoi1 theory shows that for an inviscid incompressible steady 80- the Lfi and
span and is equal to Zb. It should be noted that equations 3.13 and 3-14 assume that the
pitch angle makes the only contributions to lift and moment of the airfoil. In reality: the
heave rate term Ii and pitch rare term 4 wodd alter the effective angle of attack and
contribute to lifi and moment. Since the steady assumptions for this method are for
illustrative purposes, the Ioss of accuracy by neglecting the heave contribution is not
important Thus for quasi-steady aerodynamics. the t e m s of the [A] mat& in equation
3-12 become:
assuming that the lifi c w e dope for the airfoii is CLa=Zn. If the stnicniral darnping
terms are neglected, when this equation is substituted back into equation 3.4 we then
obtain:
Non-dimensiondizing this matrix as before. it simplifies to:
Accordkg to the well known theory of dynamic systems (Steidel 1971 Thomson
1993), the roots of the characteristic equarion, det PG)] = 0' are the eigenvalues of the
system and they reveal whether the system is stable or unstable. Instability in chis case
indicates that the system would achieve flutter for the given configuration and initial
conditions. If det [B@J = O is Wntten out as a polynornial in and the real parts cf dl
roots (eigenvalues) of the equation are negative, then the system is considered stable and
d l vibrations will evennially damp out to zero. A solution with positive real roots and no
imaginary roots indicates a divergent solution, while roots that are complex with a
positive real component indicate an oscillatory instability, that is flutter in this case.
37
Using this method, the stability of a systern with Iinev sprùig constants for both
aerodynamics. This solution allows one to determine the linear flutter speed for various
configurations of the airfoil as well as to examine the relationship between the stiffness of
method of the previous section were replaced by unsteady aerodpamics terms. A more
where fiom Fung (1955), f=l, a=0.165' b=0.0455, c=0.335, d=O.jOO for inviscid
incompressible flow- Wagner showed that the Laplace Transforrn of the expressions for
Equation 3.2 1 is therefore once again in the form of equation 3-12 and can be substituted
into equation 3.9 in order to solve directly for the stabïiity conditions for a given airfail
confi5guration. Note that the only term lefi that c m prevent a direct solution for the
problern is the nonlinear term, F(a) in equation 3.2 representing the pitch degree of
Fung (1955) d s o gives the following expressions for Iifk and moment in the time domain:
freedom, the pitch s t m e s s K, in equations 3.9 and 3.1 0 \vas replaced witb a h c t i o n
F(a)relating the pitch angle to the restoring moment. Solutions for soivïng the system
Note that the aeroelastic system has now been described in a non-dimensional
form in both the tirne domain (equations 3.9,3 .Z,j-24) and the Laplace domain
35
(equations 3.1 0,3.2 1.3-22). Two diEerent approaches. numencal and anaiytical for
The tirne domain equations will be solved using the Houbolt h i t e difference
method [Houbolt 19501 to give a time stepping simulation solu6on for the response of
the system to initial stimuli. The Laplace domain approach wï1I be used to provide an
analytical solution to the synem. The only factor preventing solution ofboth systems is
the nonlinear restoring moment F(a) which now replaces the linear pitch stiffbess a.
Within the Houbolr method simulation, the parameter F(a) can be evaluated at each time
step according to the instantaneous value of a as outlined in section 32.1. For the
Laplace domain equatio~sthe nonlinear term must fnt be linearized through the use of a
With the aeroelastic systen characteszed in both the tirne and Laplace domains.
the nonlinear restorïng moment F(a) in the pitch degree of freedom is subnituted into
equations 3 -9 and 3.10 in place of the assumed linear pitch stifiess a.Because of the
non-constant nature of the nodinear restorïng moment F(a), directly solving for the
stability of the system as shown above?using the Laplace domain, is not feasible. Two
approaches were used to evaluate the response of the airfoil with a structural nonlinearity
and determine the stability of the system. n i e Houbolt finite difference scheme solves
for the state of the dynarnic system at successive time steps. By simulating the system for
a sufncient number of t h e steps to reach a steady state condition where all mimient
26
effects in the system have disâppeared, the stability of the system may be explicitly
anaiytical rnethod by which the flutter speed and amplitude of the limit cycle may be
The use of a backward lookïngf;nire diBrence solution allows the user to solve
the nonlinear differential equation that results fkom eq-uation3.1 0. The Houbolt finite
difference method uses a fourth order explicit finite diEerence method to evaluate the
pitch and heave parameters of the system using a time marchine scheme that effectively
tracks through Ume the response of the airfoi1 to a given stimulus. As in previous work
(Lee and Desrochers 1987, Price et. al. 1993) a fourth order Houbolt scheme was utilized.
While thïs is a relatively low order method of solution, it has been shown that higher
order methods such as eighth order Houbolt or Runge Kutta schemes provide a greater
degree of accuraccy but can be more cornplex to use as weIl as being more
computationally intensive p e e and Leblanc, 19851. Using the fouah order scheme, the
where n represents the curent time step for which ai1 conditions are known. Thus n+l
represents the subsequent r h e s e p for which the system is being solved, and n-1 and n-2
represent solutions ai the two previous steps. The AT represents the size of the time step.
where expressions for H and X can be found in Appendix A. The H matrix represents the
time domain differential equation terms in equation 3.10 and X represents the time
domain unsteady aerodynamics ternis fkom equations 3 .Z and 3.24. This equation h a .
been modified such that F[a(s)]has been replaced by F,(a;Ü) + F,(a), where a is an
represented as:
Solving for the unknowns by inverting the P ma& gives the solution for the system at
requires ~ avalues
t at time r = O and two t h e steps previous be lmown before the system
c m be stmed. Since the initial conditions for the systern (oi(O), &(O),a(O), j(0), :(O)-
c(0)) are required to be known in order to defme the problem, only the bvo previou tïme
steps need to be deterrnined to begin the simulation. By using a Taylor series around the
1-0 point, values for n-1 and n+l for both the first and second derivatives can be
estimated and substituted into the M matrix (See Appendix A for details). Conditions
for three consecutive time steps are then known and Houbolt's scheme can be used to
AT
I ( -A) = e Il (5) + - {9A(5
24
+ Ar) -
(3 -3O)
19L(~)e - 51(~ - 4 ~ ) e + k(r - AT)^ -3bAr}
See Appendix A for details. Iz(Ar) is sïmilarly expressed by substituthg d in the place of
b in equation (3 -30).
systems by replacing the nonlinear portion of the system with a linear replacement where
the gain, or stifkess in the case of a spring, of the linearized spring is dependent upon the
amplitude of the harmonic motion. As it is Unplemented here, the nonlinear pitch spring
sinusoidal input to either the nonlinear or the hearized spring will produce the sarne
output response,
The describing funetion method requires the assumption that the resulting motion
of the system wiIl be h m o n i c with an amplitude that may Vary only slowly. The
linearized spring is also sometimes known as an equivalent snijj%ess and allows for a
solution that in the descnbed case approximates die amplitude versus airspeed
relationship of the nonlinear system, as w-el1as the airspeed at which limit cycle flutter
will ensue. In the system described in Section 3.1 the nodinear operator is the pitch
The descnbing function method seeks to replace the nonlinear spring in the
system with the linearized one by fint assuming a hamonic input to the spring in terms
F(t) = N, B + n A sin(ot
P
+ 0) + nq A cos(or + 8)
where NB is the relationship beween the mean input B, and the mean output and
NA= n, +jn, : where j denotes the imaginary component of the complex variable, is the
relationship between the sinusoidal input and the fundamental harmonic output. In the
wefl known harmonic equation 3.3 1, parameter A represents the amplitude of the input
signal and B represents the DC offset or bias of the signal. In the case of the symrnetric
hysteresis loops studied here, there is no bias of the signal and both NBand B can be
relations form the describing function portion of the system depicted in Figure 3.3.
For the system studied here, the describing fùnction seeks to linearize the
nodinearity in the restoring moment of the pitch degree of freedorn. The input to the
32
Applying these results for the purposes of illustration to the well knovvn, simpler
given by:
The solution c m be expressed [Gelb and Vander Velde, 19671 in the form:
where o = a, + 8 and the describing function terms n, and n, have been defined by:
Thus the nonlinearïty has in effect been replaced by a linear proportional plus derivarive
network whose coefficients are functions of the amplitude and fiequency of the -stem.
The linearization is essentÏaiIy derived h m an averaging of the of the amplitude over
one complete cycle of the oscillation. From Equation 3.38 it c m be seen that for this type
of dynamic system the parameter n, acts in place of the spring stiffness coefficient and
the n, term comprises a darnping term. The cornplex linear gain NAdetermined fiom the
descnbing fünction therefore incorporates both stifkess and darnping tems n, and n,
respectively.
For a hysteresis nonluiearity, Gelb and Vander Velde (1968) l i a the parameters
which is valid only where A > 6 + D/m,meanhg that the describing fûnction
approximation can o d y be used where the amplitude of the pitch oscillations lies outside
the bounds of the hysteresis loop. The parameter m represents the siope of the linear
portion of the hysteresis loop fiom the relation between pitch angle a and the restoring
moment F(a) as illustrated in Figure 3.2. In al1 cases presented here the slope was
assumed to be unity, or 1 non-dimensional moment unit per radian of pitch angle. The
parameter F is e q d to hW2 and D is equai to VI as illustrated in Figure 3.2. The
fiinctions f() is the saturation function that describes the deadband axa where:
f(t/> = -1 :J < -1
= (2/x)(sin-'y + y fiz) /if-< 1
=1 y> I
where y is a d m y variable. For illustration the values for the describing fùnction
parameters N,, I+,, and n, for the airfoil confiUrnationused in Run 14 (~=100,G=0&
The feedback nature of the system is illustrated in Fi-we 3 -3 where the describing
function is the linear operator N that replaces the nonlinear operator of the hysteresis
spring F(a) in the otherwise linear dynamic airfoil system represented in Fiame 3-3 as
L(S). The method determines what sinusoidal input is required for the linearized operator
to emulate the nonlinear outpur. Thus the properties of the operator are dependent upon
the amplitude of the input signal. An iteratïve approach ro solving the system of
equations is therefore required that first guesses at the amplitude of the pitch oscillation
of a(t)?then compares the output of the system to the guess. How this rnethod was
A similar solution method was also applied to the bilinear type nonlinearity
illusnated in Figure 3.4 for comparison with results by Lee and Desrochers (1986) and
component to the moment about the two-dimensional airfoil and NBrepresents the bias of
35
the input signal to llinear spring (see equation 3-37): then NBmust equal zero. In the case
of the hysteresis loops that are s p m e t r i c about the ongin of the restoring moment versus
pitch angle space B is also equal to zero and the Sinusoidal Input Describing-Function
(SIDF) formulation may be used. The system can therefore be soived in an iterative
rnanner by k s t assuming a value for pitch amplitude A and solving the equations 3.40
and 3 -41for n, and n, as a function of A. A value for NA-, +jq, whîch represents the
equivalent stifiess of the nonlinear spring, can then be substituted into the equations of
motion for the airfoil in place of F(a) and a Iinear solution detemiined. The resulting
pitch amplitude of the system is then compared to the initial estirnate, A, and revised.
The iterations are repeated untd the solution resolves to a single a m p h d e solution.
The malysis of the studied systems consisted of cornparisons between the
numerical solution method involving ùme marching simulations, and the semi-andytical
approximation using the describing function approach. This section describes how the
the solution rnethods were applied. The parameters which were examined and the
in a free Stream of air with restoring forces in both the pitch and heave degrees of
lifting forces were derermined fiom thin airfoil theory where the unsteady aerodynamic
Sû-ucturally, the restoring force in the heave direction is modelled as a linear spring and
for validation with previous work as iIlustrated in Figure 3.4, and a hysteresis type
37
The airfi il used for the analysis was confipured such that the elasac axis was
placed at the % chord position and the centre of mass placed at the Y8 chord position
downs~eamof the elastic auis. The resdting parameters used for the a o i l structure as
a+,= -0.5
r, = 0.5
x,= 0.25
where r, represents the radius of , ~ a t i o nof the airfoi1 about the elastic axis. These
values were kept constant throughout the analysis and are consistent wiîh those used in
previous works. Thin airfoil theory specifies that the aerodparnic centre is located at the
% chord position. For the sake of cl- and cornparison it was decided to keep these
characteristics of the airfoi1 consistent. The damping coefficients for the system were
neglected throughout the analysis and o d y the fiequency ratio a,the mass ratio p and
non-dimensional aùspeed U were varied. For each airfoil confi,guration the airspeed was
expressed as a fiaction of the linear flutter speed U*, the airspeed at which divergence of
the system occurs if the pitch spring is linear, and was denoted as UN*.
Borh the simulation method and the describing function method were first used to
reproduce die results of Lee and Desrochers (1986) by hnplementing the mode1 with a
38
bilinear spring in the pitch degree of freedom. The Houbolt finite difference method was
used to produce a time based "recording" of the airfoil response in pitch and heave in a
time marching manner. The solution method was subsequently used for Hysteresis s p ~ g
The response o f the systern was simulated subject to a set of given initial
conditions, in this case maidy the initial pitch angie and airspeed were varîed' to produce
a response of the system with time in each of the degrees of fkeedom. The simulation was
computed until the sysrern response either damped out to zero, diverged beyond
acceptable bounds, or had compieted enough time steps to be certaio transient effects had
-
parametrïc b a i s for each airfoil configuration Iisted in Table 4.1 by varying the airspeed
(WU*) and initial pitch angle @,)as the set of initial conditions. The results of each
simulation were then evaluated to determine which of the three conditions resuited:
For cases where LCOYsresulted, the amplitude of the steady state oscillation was
recorded. Frorn these results, diagrams showing the Butter boundaries, regions of
stability and LCO's within the airspeed UN* versus initial pitch angle a, parameter
39
space, as well as the airspeed U/U+ verszrs LCO amplitude relationship w-ere evduated.
For the bilinear spring simulations, close cornparison of results with those fiom
the previous work c o b e d the correct implementation of the model. Al1 r e d t s agreed
w-el1wi-th the results of Lee and Desrochers with some srnail differences explicable by
Chapter 5.
The hysteresis spring was subsequently implemented into the mode1 ming the
nomenclature outiined in Fiame 3-2 where the spring3 parameters have been
characterized by the AV, AJ3, V, and Ho,, variables. The hysteresis spring was
varied depending upon the direction of change of the pitch ( Oa / 3 c ). The spring was
moment unit per radian as shown in the diagiam except when on the deadband portions of
the hysteresis at the top and bonom of the figure. Exception handling was also
response. The code was implemented such that a reversal in direction of pitch between
H' and H- maintains the pitch-moment dope of 1 between the V, and Vz marks. The
parameters AV and AH From 1.0 to 0.1 and with vertical and horizontai offsets up to 0.5
with various combinations of d l four pararneters. The pararneters for the simulated
systems are listed in Table 4.1 - Results consisted of time based traces of the airfoi1
motions in both the pitch and heave degrees of fieedorn which were then interpreted to
provide flutter boundary maps indicating the regions of stability as well as the limit cycle
The f i t e dserence method was used for each parameter set to simulate the
response of the system to an initial perturbation, the initial pitch angle ao,to the point in
time where the transient effects were no longer affecting the system. The three stability
1. Darnped (stable) - the oscillations of the system died out over time such
bounds of the simulation and the system was deerned to have become
1imït cycle response. The criterion implemented was "rime steps" >
50000.
The results of particular interest are diose where the oscillations either enter a steady state
cause damage or failure. This flutter type behaviour c m not be predicted by linear
For each airfoil configuration of interest, simulations were performed for a range
of non-dimensional airspeed ratios of 0.1 to 1.O. For each simulation run the steady state
response of the system was recorded with respect to the stability of the system as
specified above. Also recorded was the amplitude of the limit cycle response if the Lunit
cycle existed-
4.3 Describing Function Solutions
a means of emulating the effect of the nonlinear elernent by relating a Luiearized spring
stifiess to the output amplitude. In effect?the noniinear spring modelled in the pitch
approximates the same amplitude of oscillations as the nonlinear spring it replaces. The
method allows for the development of a relationship between the airspeed and the
The Describing Function @F) analysis technique was implemented in this project
O
To provide for a means by which to compare the results of the simulation
with freeplay, illustrated in Figure 3-4, as evaluated numerically by Lee and Desrochers
(1 986) and both numencaily and using DF's by Price, et. al. (1994). M e r successful
cornparison with the simulations for the bilinear spring and the DF work by Price, et. al.
43
(1994) the method was modified to use the d e s c r i h g function parameters corresponding
with unsteady aerodynamics represented in Equation 3-21 can be solved directly for
stability if the structural elements are linear. The descnbing h c t i o n method \vas
inplemented by first choosing the value of the limit cycle amplitude A. -4value for N=n,
+jn, fkom equations 3.33 and 3.34 was found and represented the Iinear pitch stifkess
K, replacing nodïnear spring F(a), in equation 3.17. When the system is simplified and
represented in the form of Equation 3.19, the system is in the form of a fiee vibration
response problem. As described in Chapter 3: the roots of the system are the solutions to
The expression for [BI was determined fkom the equations for unsteady
aerodynamics by substituting Equation 3-22 into Equation 3-21and then substituting the
expressions for lift and moment into equation 3.9. The polynomial was explicitly
determined using the symbolic math processing software Mapte. The resulting
expression was solved usine the software analysis tool Matlab. Using Matiab the explicit
roots for the characteristic equation were found for the syaem using the linearized pitch
stifbess N-
The roots of the p] matrix could then plotted in a root locus plane and the speed
at which the first real root crossed the real=O line revealed the Butter speed for that pitch
stiffness. Figure 4.10 is an exarnple of a root locus result plotting the values of the roots
44
of the characteristic equation for the given aeroelastic system as the airspeed U* is varïed.
In the case of Fi,we 4-10 the roots for the linear system where o = 0.2 are plotted -4th a
pitch stifiess term of 1.0 substituted for the nonlinear F(a) functiori. The result shows
that the system will become unstable for a non-dimensional airspeed of U* = 6.285. This
value represents the linear flutter speed for this airfoi1 ~onfi~wation.
This solution
method allows for a relationship to be established between the pitch stiffness and the
flutter speed. Fi_mire 4.1 1 is a plot of how the flutter speed varies uith pitch stiffriess for
The describing function approach essentially provides a means for relating the
pitch amplitude to the pitch stiffness, calculating linearized pitch stifhess N fiom A'
while the flutter analysis essentially provides for a relation between the pitch stifniess and
the flutter speed. By assuming a harmonic output for the dynamic system the nonlinear
pitch relation F(a) was replaced by the describing fûnction equivalent stifkess effectively
creating a solution method that relates the pitch amplitude to the flutter speed.
The flutter speed is solved in an iterative manner, varying the speed U for which
the dynamic system is analysed until the speed for which the system goes unstable is
found. The result is an approximate solution for the limit cycle amplitude A for the speed
The results described in the previous section were compiled for the various airfoi1
confi,wations and initial conditions. This section presents an analysis of those results as
well as cornparisons between the two solution methods. Stren,ghs and weaknesses of the
methods are discussed and improvements to the analysis suggested. The results are put
The results of the pararnetric study of various airfoil configurations are plotted in
Fiagres 5.5 through 5.3 1. The 'a' figures (Fi=gre 5.5% 5.6q 5.7% ...,5.3 la) plot the
flutter map results £kom the finite difference method numerical simulations. The plots
show the parameter space of airspeed ratio, U / U f ,versw initial pitch angle, a,,. Regions
within the parameter space are indicated where for the specific airfoi1 confi~ouz-ation
unstable oscillations were induced. The x's represent a change in stability fkom damped
to LCO or vice versa. The 'b' fipures (Fiagre 55b, 56b, 57b, ... 5.2%) plot the pitch
?
restoring moment F(a) against the pitch angle a(r) indicating the geometry of the
hysteresis loop for each of the indicated cases. The 'c' figures (Figure 5.5c,5.6c, 5 . 7 ~...'~
5-3Ic) indicated the amplitude of Iirnit cycle oscillations, where they w-ere observed, as
46
induced by the varjing airspeeds U n * . Where the describing function \vas used. the
5.1.1 Validation
As stated previously. the results of the numerical simularions were in the form of
time based histones of the pitch and heave motions of the airfoil system as illustrated for
example in Figure 4.1. Results were first obtaùled for airfoi1 ~onfi~gurations
using a
bilinear spring with freeplay and preload as explored by Lee and Desrochers (1987). A
large number of simulation plots were accessible for comparîson with their work and
present the results compared very well with theirs, with only a veq- few minor
discrepancies. For individual cases compared, Le. the same airfoi1 configuration' initial
conditions and bilinear nonlinear sprïng, in dl cases the same stability condition was mer
for a given simulation. SimilarIy, cornparisons of boundaries of lirnit cycle and divergent
flutter for given airfoil properties matched exactly. It was found that for a few
configurations for which the system was just barely stable the simulations did not reach
stability after the exact same amount of tirne. This discrepancy was easily explainable by
the sensitivity of the nonlinearities to very small changes in initiai conditions. A slightly
differences in the numerical accuracy of the methods al1 contributed to small differences
that were amplified by the nonlinearity into larger differences in the results of the
simulation m. It was not found that these differences cawed discrepancies in the overall
47
characterizatiori of the airfoi1 system. It was rather found that there were noticeable
could be determined. By utilinng a bilinear sprhg and comparing with previous w-ork.
the irnplementation of the Houbolt scheme was validated. Subsequenùy validating the
behavïour of the hysteresis spring and substituthg it into the simulation code dlow-ed the
motions of the spring and its response to various stimuli. The response of the spring was
analysed graphically to ensure that the pitch angle versus restoring moment response of
the operator followed the specified curve. Figure 3.2 shows the general structure of the
c w e and the difference in the path of the c u v e depending upon the direction of the
change in pitch. Exceptions in the path of the c w e were noted during certain portions of
direction of pitch change wirhùi the "loop" portion of the hysteresis curve. It was
determined that the loop was best described by dowing the restoring moment to r e m
dong the sloped path of the cuwe when reversal occurred as depicted in Fiawe 4.8.
When reversa1 occurred on the flat portion of the cuve the response was modelled such
that the dope of m=l was maintained as depicted in Figure 4.9. The spring response was
Figures 5-5b to 5.9b and Figures 5.1 Sb to 5.2% trace the path of the hysteresis
loop restoring moment for an example of each of the corresponding runs. A small
48
arnount of under-nui c m be observed in some of the plots. This resdt is in fact just a
;ru&g of the cornero' by path of the restoring moment. This particularly occurs for
cases where the amplitude of the pitch response is large and the pitch angle is therefore
changing veq- rapidly as it moves through the area of the hysteresis loop. The finite
difference source code includes 2 smaU aigorithm to look ahead and test each s e p for
ovemin and ensure that the moment follows the curve explicitly.
Simulation results for the airfoil system were subsequently obtained for a
hysteresis sp&g with a variew of spring parameters as outlined in Chapter 4. The results
were first analysed in the f o m of plots of the stabili- charactenstics of the system for
m.rying airspeeds as shown for example in Fiame 5% This data dlowed for a
"mappingooof üie flutter boundaries indicating at what airspeeds the system achieved
stability, limit cycle oscillations, or instabiw. Results were also plotted for the
The flutter maps in Figures 5 5 a to 5.3 la show thaf as predicted, there lies a zone
of lirnit cycle flutter at airspeeds lower than the linear flutier speeds. The linear flutter
speed for the airfoi1 confi_pratïon was determïned using both the simulation and the
analytical solution methods. The linear flutter speed was detennined using the classical
analytical solution method by detennining the speed U* at which the system became
unstable. As described in the previous chapter, this was performed by assuming a linear
sprïng in the pitch degree of fieedom then solving the equations of motion fkom the
49
Laplace domain for the eigenvdues of the sysrern. The linear flutter speed was found by
v q i n g the airspeed co fïnd the speed at which the system just goes unstable. indicated by
negative reaI portions of the roots. The numericd simulation method was later used to
confimi the linear flutter speed by setting the spring to be a linear spring and perfomlln;
iterative simuiations to find the speed at which the systen; became unstable. The results
The flutter boundaries show the extTeme sensitivity of the nonlinear system to the
initial conditions. For example, Figures 3.2 to 4.7 show how the sarne initial conditions
will result in different stability conditions for the system depending upon the airspeed.
The pitch angle time naces fiom Run 14, depicted in Fiegres 1.2 to 4.7: demonstrate how
for the same initial conditions the stability of the system varies depending on the
the speeds and initial pitch mgles for which there is LCO motion. For the same initial
pitch angle of a = 6" the system would achieve a lunit cycle condition if the non-
dimensional flutter speed waç Setween 0.864 and 0-910 and between 0.943 and 1.0. The
system was stable below a speed of 0.864, but a zone of stability also existed between
0.91 0 and 0.942. From these results it can be seen that it becomes v e c difficult to
predict the stabilip of a system frein a given set of initial conditions without prefonnïng
It was also observed that the flutter boundaries were a continuous line for most of
the ii = 0.8 nuis with a clear boundary showing the stable and LCO regions of the
parameter space. Most of the o = 0.2 runs however showed the boundary of flutter as a
50
large area of conditions were the airfoil exhibired ~ a b i l i t yor LCO depending upon small
differences in speed U/U*' as s h o w for example in Fiagure 5.12a. This fi=ure shows the
flurter map for the given airfoil c o n f i a ~ t i o nand conditions where the x's mark each
airspeed U N * where for the same initial pitch angle the stabiliq- condition has switched
fiom stable to LCO motion or vice versa compared to the next lowest ainpeed simulated.
In ail cases the runs diverged to instability at the point Um*=l .O regardless of
configuration or initial conditions. This result is as expected since the nonlineaity tends
to reduce the stifiess of the degree of fkeedom in the region of the hysteresis loop.
Therefore there is no tendency for the system to diverge when the pi~champlitude is
much larger than the size of the hysteresis loop unless the linear flutter speed has been
achieved,
In addition to compiling results with respect to the stability of the dynamic system
at various speeds, for speeds at urhïcha limit cycle stability existed the amplitude of the
limit cycle was recorded. These results were then plotted to show the relationship
between the limit cycle amplitude and the Butter speed for various airfoi1 configurations
and initial conditions, as shown for exarnple in Figure 5%.The fist charactenstic for
these relationships that was noted was that for d l the airfoil confi_gurations tested in this
work?the limit cycle amplitudes were for the most part invariant with initial conditions.
This result is consistent with the results published by Lee and Desrochers (1987) and
Price et. al. (1994) for many bilinear and cubic springs. The exceptions to this case are
51
the bifurcation diagrams in the Pnce et. ai. (1994) report where specific confi_mations of
the nonlinear spring lead to diffcrent LCO anplinides. Le. different limit cycle
amp!itudes could be achieved depending upon the initial conditions. From the resdts it
c m be seen that these Ends of amplitude jumps were present mainl- in the results of the
-
o = 0.2 farnily of results. This resuit would seem t o indicate that the results for the w=
0.8 group were the more neady sinusoidai cases without the harmonic structures observed
in the o = 0 2 cases.
Observations were also made of the type of response achieved depending upon the
shape of the hysteresis cuve. As would be expected, the response of the system becomes
the sarne as for a linear system as AH approaches zero- and the same as for a system with
5.17~ that as the ratio AWAV approaches infinity fiom zero the difference between the
two solution rnethods for the speeds at which LCO may be induced and the linear flutter
speed becomes greater. Table 5.1 shows thm for the o = 0.8 cases the speeds at which
CC0 motion is initiated varies from U/U* = 0-816 for the M A V = I O case to UA-..* =
0.954 for the M A V = 0.1 case. For the 65 = 0.2 cases the confiCpration for which the
speed at which LCO motion was initiated was the case of AV = O, a bilinear type operator
sirnilar to that shown in Fiame 3.4, for whicti LCO was initiated at U N * = 0.236. From
the table it can be seen that there is a distinct correlation berneen the speeds of initiation
introduce certain errors into the solution because of the nature of their numencal
approximation. One example of this -pe of error is the introduction of whar is termed in
would in this case be termed an czmj?cial damping component. This error is introducecl
because the f i t e measure of thne, Ar can not exactly descnbe the motion of the airfoil
when changes occur very rapidly. This approximation to the airfoil behaviour has the
From the validation nins performed to determine the Iinear flutter speed for the
airfoi1 ~onfi~gurations,
the known solution of a direct analfical analysis of the linear
flutter speed was compared to the finite difference solution for the same parameter. For
the airfoi1 confi,guratïon where p = 100 and 6 = 0.8, the analytical solution to the flutter
U* = 4.1 128. The error of the finite dserence solution was therefore -0.04% and was
therefore not determined to be important for the purposes of the present analysis.
It should however be noted that the finite difference method is not an exact
solution to the problem and should not be considered absolutely correct when compared
to the describing function method. It becomes quite difficult to evaluate the error
inherent in the finite difference solution, especially for cases such as the studied nonlinear
spring where no exact solution for cornparison exists. The two studied solution methods
53
were essentially being compared IO each other and a determination of one or the other
As with the numerical simulations, the describing function method w-as first
applied to systems \vit& the bilinear spring with freeplay nonlinearities as reported by
Price et. al (1994). Sample systems were solved as described in the theory section and
the results were compared with both the resdts of the bilinear spring simulations and the
results published by Pnce et. a1 (1994). The solutions were found to agree quite weil
between the describing function method and the finite difference simulation method.
spring this solution method was used to rnodel the systems with hysteresis nonlinearïties.
The syçtems with hysteresis springs that were modelled were symmeûic springs with AV
and AH varying from 0.1 to 1.0 and cases completed for both o = 0.8 and o = 0.2.
Fi-=es 5 3 to 5.3 l c plot the values of pitch angle a vs flutter speed U N * and where
solution. It c m be seen from these plots that the agreement between the nvo methods is
quite good considering the approximations and assumptions that have been made in order
observed between the shape of the hysteresis curve, descnbed by the W A V ratio, and
the accuracy of the describing function method compared to the finite ciifference results.
54
Frorn Table 5.1 ,the largest discrepancies between prediction of the speed at which LCO
motion is initiared were found in the cases where the shape ratio H A V was highest.
The largest percent error between the two solution methods was found for the case where
in the prediction of the initiation of LCO motion was for the case where hWhV = 10.
There wodd appear to therefore aiso be a correlation between the accuracy of the
descnbing function method and the shape of the hysteresis loop as described by the ratio
WAV.
The predictions for the airspeeds at which LCO oscillations initiate were found to
be in good agreement with the f i t e difference method and the semi-analytical method.
The agreement fiom Table 5.1 ranges Eorn 0.2 % to 37.5 % with the cases where o = 0.8
agreeing somewhat better than the cases where G = 0.2, It was noted earlier that the
simulation results for the = 0.8 cases appeared to be more sulusoidal, without the
harrnonics observed in the = 0.2 cases so that better agreement is to be expected for G =
0.8.
Quantifying the accuracy of the describing function for predicting the amplitude
versw airspeed relation, as predicted by the finite ciifference solution method, was
difficult due to the multiple values of LCO amplitude for many of the finite difference
results. Observations of results in Fiames 5 . 5 ~to 5.1 5c showed that the describing
h c t i o n method results generally followed the frnite difference results quite closely. The
-
o = 0.8 nui that appeared to have the largest error between the two methods is shown in
Figure 5 . 5 ~corresponding to nin 6- From Table 5.1 it was dso seen that this
55
confiauration also had the largest error in predicting the speed at which LCO motion
uùtiated. In the ii = 0.2 family of results, the figures displayuig the larges discrepancies
between the two solution results were in Figures 5.1 Oco5.12c, and 5. 16c corresponding to
nins 100, 102 and 110 respectively. Of the three confi3~ationsFigure 5.16~appeared ro
have the largest error, although diEcult to quant if.^' due to the mdtiple amplitude
solutions. Table 5.1 indicates that these are the same hysteresis configurations thar had
the larger errors between the describing function method and h i t e difference method
Similady, the diEerences between the descnbing function prediction of the LCO
noted that the accuracy of the describing function method solution was greatly improved
as the amplitude of the limit cycle increased greatly. This relation is expected since, as
before, the accuracy of the describing fûnction method will increase as the response
becomes more sinusoidal. As the amplitude of the limit cycle becomes large with respect
to the size of the hysteresis loop, the effects of the noniineariv become less si_gïficant
function it was decided to restrict the describing function solutions to those hysteresis
loop configurations that were syrnmetric, i-e. no offsets- From the n u e r i c a l simulation
results for the offset hysteresis configurations as shown in Fieires 5.18 to 5.3 1, there
appeared to be little effect in terms of the general behaviour of the system. The same
56
relationships with respect to the fkequency ratio and the shape of hysteresis loops were
observed in the offset results as were described above for the symmetnc results.
5.3 Cornparisons
method followed quite closely the results of the finite difference method in the majority
of the cases. In particular the describing fuaction agreed better for the cases where ii =
0.8. This result is not surprising, considering that the time-based results fiom the
simulations indicated a purely harmonic result. Since the describing fiuiction method has
expected to compare quite closely with the analytical approximation. From Figures 5-52
and 5 . 8 ~it was seen that as the hysteresis sprin; approached a condition of AV=O the
describing fünction method was less accurate than those where AH approached zero
(linear spring). This result c o n f i s the conclusion that the describing function merely
loses accuracy as the nonlinearity has a greater impact on the response of the systeem.
Fi,gures 5.1 to 5.4. The sample run s h o w in Figure 5.1, from run 105, shows little or no
difference from an ordinary sinusoid and the power spectrum of the sample, as shown in
Figure 5.2 shows integer harmonies of the fundamental but of liale power and no
additional structure. Figure 5.3 however. fiom run 102- obviously bas additional
fiequencies in response. The power spectnim in Figure 5.4 shows that the sample was
57
definitely not sinusoidd in nature. Comparing the flutter maps and describing function
method solutions for those two runs shows thar of the two, nin 102 has the "bifurcation"
iype muftiple amplitudes in the LCO amplitude plots as well as a very clutrered flutter
map. We c m infer Iiom the cornparison of these two sets of resdts that the higher
hWAV shaped hysteresis nodinearities are introducing harrnonics into the response of
the system that the describing function is fundarnentally unable to account for.
interesting" results for those cases where AV approaches zero, i.e. the bilinear case, it
could be observed that the bilincar case represents a good conti,vation for examining the
behaviour of the nonlinear system. Previous works has mostly utilized bilinear and cubic
functions for nonlinearities because they are more easily implemented. The quenion
arises whether by sirnpli@inp the anaiysis process they were limiting their potentiai to
find unusual behaviour such as chaotic motion. It would appear that the hysteresis loup
provides what amounts to a continuum fiom the simple sinusoidal response of a linear
system where AH approaches zero, to the much more unpredictable behaviour of the
good candidate for observing musual behaviour in nodinear systems, at least when
The question of why the describing function breaks down as the systern response
behaviour in the solution method. Why the system response becomes less sinusoidal
should also be examined. For the cases where the shape of the hysteresis nodinearity
58
affect the harmonies of the system, it can be seen that as the spring becomes Iess like the
linear case where AV = O?the "parh" of the restorinj moment with pitch angle is
deflected m e r and M e r fiom the linear- It can be sunnised that it is this deflection
that ïntroduces the harmonic structure into the system response. Thus the bilinear s p ~ g
k e d at zero and only AV is acting upon the systern. For the cases where the difference
in the fiequency ratio o S e c t e d the response of the systern, it can be assumed that since
as discussed previously the two degrees of freedom are couple& reducing the ratio of
their fiequencies could make the system more sensitive to the effects of the nonlinearity.
Whiie the accuracy of the simulations has been shown ro be quite good in a
anaiysis methods to "real world" situations. In most cases a real aircraft wing section is
subject to more than simply two degrees of keedom and three dimensional effects are
going to keep the flutter motion of the wïng fiom being exactly as assumed in the present
model. However there are some cases whereby components such as control surfaces do
behave in a very two-dimensional rnanner. For example a flap might pivot about its
hinge and be far enough outboard for the bending of the wing to be ~ e a t e das pure heave.
The pure heaving motion of the wing as a whole is actually more of a bending motion in
a ke-dimensional situation and the pitching of the wing occurs as a twist that varies the
59
angle of the wing section depending on its outboard position. Despite these limitations
there are still some good generalities that can be S e r r e d from the mode1 results. From
the simulations it is quite apparent that Iimit cycle flutter due to nonlinearïties in the
response of the wing is achievable at speeds subsrantially lower than the linear flutter
speed calculated using classical aeroelastic rnethods. Work by people such as WooInon
et. al. (1 957) and Breitbach (1977) have recognized this phenomenon for some time now
but adequate tools for predicting the results of these nonlinearities are still being
While this reduction in the actual flutter speed of the aircrafi component is of
concem, both the simulation results and the descnbing function resdts point to the fact
that the amplitude of the limit cycle oscillation slowly increases with speed, building up
the operation of the aircraft. As the wing flutter amplitude increases with speed, either
sensors installed in the wing or the feedback through the flight controls can be used to
wam the pilot of impending £lutter danger similar to the rnanner in which aircrafi
It has been shown that these types of limit cycle oscillations c m occur at airspeeds
well below the Iinear flutter speed and care m u t therefore be taken when analyshg the
aeroelastic properties of a component to ensure that the flight speeds do not encroach
upon the speeds at which the small aeroelastic vibrations are induced. While die
60
amplitude of the oscillations may not be lase enough to affect the flight dynamics it is
There are some cases where die nonlineanties present in the response of an
airfoiI. Lee and Tron (1 989) were successfuliy able to mode1 the flutter characteristics on
a CF-18 control surface using a describing fuaction type solution method. They showed
how the solution method could be used to correctly predict LCO behaviour at speeds well
The system studied in this thesis makes the assurriprion of inviscid incompressible
air flow. There are nonlinearities associated with compressible flows and shocks that are
is quite good so long as the limitations of the solution method are &&en into account and
caution is exercised to ensure the underlying assumptions are still valid. Validation with
a simulation type solution method could be valuable for the purposes of checking the
become more pronounced and the response less sinusoidal, the describing function
becomes less accurate. Thus applying it to solve problems where nodinearities are
prevalent becomes difficdt to justiS. It's greater value is perhaps for those cases where
Several aspects of the results of this work appear appropnate for M e r study.
The h t area for continuation of this work lies with the analysis of the time based pitch
angle results from the Houbolt finite difference rnethod. Many of the snidies performed
analysing the results for evidence of chaotic behaviour. The anaiysis tools for studying
the chaotic nature of this kind of Urne based response include power spectral densities
(PSDs), phase-plane plots and Poincaré sections. For example Pnce et. al. (1994)
analysed the responses of bilinear and cubic systems and detected chaotic behaviour in
continue and apply chaoric analysis tools to rhese results to determine if regions of
chaotic motion are also present. A cornparison of the chaotic motion for the hysteresis
and contrast with the response of the bilinear and cubic nonlinearities. By studying and
characterizing the chaotic response of this type of nonlinear system, researchers hope to
be able to in tum characterise the system itseifand fmd new anaiytical methods of
predicting the response of the system without the need for time consuming simulations.
describing fimction rnethod to systems whose responses are not purely sinusoidd. It has
been shown here and in other works that it is the nature of the describing function to
62
break down when predicting higher order system responses. While ùiis is a fundamental
limitation of this solution method, a technique could be detemined to evaluate the system
befoe the describing function solution method is applied. TooIs such as PSDs and Bode
Plots codd be used to determine the sensitivip of the system to higher frequencies. The
system could then be evaluated with respect to its suitability for solution by descnbing
h c t i o n rnethod-
investigation of the eEects of adding structural damping to the system might be studied to
approach the problem ~ o ammore redistic perspective. Since it is the goal of this type
of work to provide some usefül real-world application of these theories, adding some
structural damping to the system mi& indicate what some of the ciifferences might be
when applied to a real situation. That is to Say, al1 real-world systems have a certain
amount of darnping and in the case of aircraft components the amount of damping may
not be negligible. Pnce e t al. (1994) investigated several cases where they added
smictural damping to the system and they found that the damping tended to e h ï n a t e the
chaotic nanire of the system response. T t might then be inferred that a red-world
From ùie work summarized and presented in this thesis, several conclusions may
be drawn-
The describing function solution method provided good agreement with the
numencal simulation solutions, except in those cases where the response of the system
There were two parameters that appeared to most a e c t the accuracy of the
Of the niro configuration parameters the hWAV ratio appeared to have the greater effect
The G = 0.2 famiIy of nins exhibited LCO behaviour had much lower airspeeds,
compared to the linear flutter ratio, then did the o = 0.8 set of r u s . Additionally the
Butter maps for the G = 0.2 m s were much less "ordered" than those of the 5 = 0.8
results' showinj a much greater sensitivity to the UN* flou- speed and changing back
Because of the noted effects of the shape of the hysteresis nonlinearity on the
response of the system, it was concluded that the bilinear nonlinearïty that has been used
for many investigations of this type into nodinear system provides good opporNnities. at
leasr equal to that of the hysteresis nonlinearïq, for observing unmual behaviour-
Barrington, P., 1994, 87.462 Introduction to Aeroelasticity course notes, Dept of
Mechanicd and Aerospace Engineering, Carleton Universi~.
Fung, Y.C., 1955, An Iniroduction ro the Theory of Aeroelasticity, John Wiley & Sons?
Inc., New York-
Gelb, A., Vander Velde, W. A., 1968, Multiple-Inpur Describing Funcriom and
hionlinear S'stems Design, McGraw-Hill: New York.
Hauenstein, A- J., Zara, 5. A., Eversman, W., Qumei, 1. K., 1992, "Chaotic and Nonlinear
Dynamic Response of Aerosurfaces with Structural Nonlinearities", AIAA-92-2547-CP,
AIAA/ASME/ASCE/AHS/ASC, 33rd Structures, Structural Dynamics and Materials
Conference
Houbolt, J. C., 1950, ''A Recurrence Matrix Solution for the Dynamic Response of
Elastic Aïrcrafc", Journal of Aeronautical Sciences, Vol. 17, pp. 540-550.
Jones, D.J., Lee, B. K. K., 1985, "Tirne Marchïng Numerical Solution of t h e Dynamic
Response of Nonlinear Systerns, National Research Council of Canada, Aeronauticd
Note Na-AN-25.
Laurenson, R. M., Tm, R. M., 1980, "Flutter Analysis of Missile Control Surfaces
Contauiing Stnictural Nonlinearities", AIR4 Journal, Vol. 18, October, pp. 1245-1251.
Lee, B. H. K.' Leblanc, P., 1986, "Flutter Analysis of a Two-Dimensional Airfoi1 -<th
Cubic Non-Linear Restoring Force"' National Research Council of Canada,
NAE-AN-36, NRC NO. 25338.
Lee, C. L., 1985, "An Iterative Procedure for Nonlinear Flutter Analysis", Proceedings of
the 761hAM-85-0688, AL4r4/ASU/ASCE/AHS Sr>.zictures.Structural Llynamics and
Materials Conference, Apnl 15- 17.
Mchtosh, S. C. Jr., Reed, R E., Rodden, W. P., 1981, "Experimentd and Theoretical
Study- of Nonlinear Flutter?', Journal of Aircraftl Vol. 18; No. 12, pp. 1057-1063.
Shen, S. F., Hsu, C. C., 1957, "halytical Results of Certain Nonlinear Flutter Problemss',
Journal of the Aeronauticaz Sciences, Vol. 23, Feb., pp. 136-137.
Stockard, D. P., Johnson, T. L.,Sean' F. W., 1967, "Snidy of Amplitude Jumps".
American Journal of Physics, Vol. 35, pp. 961-963.
Suikwerda J., C.' 1989, Finite Difference Schemes and Partid Dilferenrial Eq~arions~
Wadsworth and Brooks: Pacific Grove, California-
Thomson, J. M. T., Stewart, H. B., 1986, Nonlineur Dynamics and Chaos, John Wiley:
Chichester.
Thomson, W. T.? 1993, Theory of Vibration with Applications, 4th ed., Prentice Hall.
New Jersey.
Woolston. D. S., Runyan, H. W., Andrews, R. E.; 1957: "An Investigation of Effects of
Certain Types of Structural Nonlineanbes on Wing and Control Surface Fluner", Jourml
of the Beronautical Sciences, Vol. 24, Jan., pp. 57-63.
Yang,2. C., Zhao, L. C., 1988, "Analysis of Limit Cycle Flutter of an Airfoi1 in
Incompressible Flo+'. Journal of Sound and Vibration' No. 123, pp. 1-13.
Zhao, L. C., Yang, 2. C.. 1990, "Chaotic Motions of an *Moi1with Non-linear Stifbess
in incompressible Flow7', Jozlrnal of Sound and Vibration,No. 138,pp. 245-251.
TABLES
Run # Refer to 1 Frequency 1 Mass 1 Hyçteresis 1 Hysteresis
I I
Vertical Horizontal
1
- - - - - - - - -
v'
>'
Describina
:
$1
N
Function
1
1
FU)
Linear
J
Svstem
L ( S )
FIGURE 3.3 Block Diagram of Descnbing Function Method in Limit Cycle System
c(t)
>
- -*
-
- C
- N
- O!
- O
-- 00
w
- n
V
-
a
12
- Tl-
- 00
- C
-
- CV
Oc
- O
c C r e a t e c by: 3 r a ï ï 3 r o ok i n g , C ¢ r L z ï o n University
C
CCCCCCCCCCCCCC~CcC~CCCCCCCcCCcCCcCcCcccccccc~cccccccccccccc~c~~cc~
CCCSC
PROGW-k? EVST
I M P L I C F T LOGICAL (A-Z]
C C a n s t c n ï IBLGTIM: NllnDer o f cime s t e 9 s t o declere srezSy s t c z s
C LPFEEQ: Frcquency co o u t p u c r e s x l t s
C TIMECSFK: I n i t c l t i r n t C O ellow cransienc effeccs ro
dmp out
INTEGER 13IGTIM, IPZREQ, T I Y I C U E K
C C c n s t a n t TH-RESX: Thrêsnold t o c e t e - d n e dm.ped sysien
REPL TESESU
PPAX'XETER (LBIGTIM=50000)
P-LWXETER (TH-XESE=0. O 07 !
PAEWKETXR [ IFFREQ=S)
PFM14ZTER (TIMECXEK=300 )
REAL S I I , F 12,S~DA(4},EIEGLF(131GTIP4).XZCXI(IBIGTIM)
REPL G u rDT;T,XA,.W, EL,A, 3,C, 5, ZETAAL. ZETFXI. o-wa., UST,PZ, XE-
REAI; PLPHO, DALPHO, DDPLPO, TDALPO. QDALPO, XFO, DXIO, DDXIO, TDXIO
E U , D-nJIPEf(4),DD.?LPE(4) ,DXI(4),DDXI(4)
REFJ;*8 A L O t f ( 4 ) .XI(O)
MXL V I , V2, ZlMIN, BI-WX, E2MIN, 321!C!Xr ULIN, SIPZAIK, LOFEAKr COPJ"\I%T
X A L INCR, DV, DB,VOFF, XOTF, X I X I . L O X I , EMIN. CMAX,QDXIO
37 CONVRT=O. 1 7 4 5 3 2 5 2 5 2
38 TEMPFILE="cemp"
3 9 C Find out if inieractive, or date file
40 C PRINT *, ' Zzter che name of the o u t p ~ tf i l e . '
41 C E a n D (=, 1 0 ) OÜTFILZ
42 PXINT *
43 PRINT *, ' Enter: X u t t e r Map=7 (def 1 or SFnqle ru11=2'
44 READ (*, 5 ) TYPE
45 5 FORMAT (13)
46 IF(TYPE.NE.2) TFIEN
47 TYPE=L
48 SNDIF
49 I F (TYPZ.5 0 . 2 ) THEN
50 OUTFILE="pitcnresp .d z t "
51 OPEN(YNIT=2,FILZ=OUTFILE)
52 OPEN(~NIT=IO,FILE='spring.dat')
53 PRINT *,'Enter t h e i n i t i a l A l p h a . '
54 -XAD ( * , 2 0 ) F L I S T ( l )
55 PXINT *,'Enter t h e Linear F l u t t e r Speed - Ulinl
56 RSAD ( + , 2 0 ) U L I N
112 CALL GR S E T U P ( O U T F I L E r G T I T L E , G ç U B T r X ~ B E L r Y ~ E L r T Y Z E )
113 cal1 TI& (char-the)
114 C X L DATE (TODAY)
115 PRINT *, ' Starting: ' ,CHAR-TIME, ' ' ,TODAY
116 C Assign const variables f o r this case
-
Byster3 for
117 C VO ET-O, 25
118 C HOFF=O - 5
119 V I = (DV/2+VOFFI -COWRT
120 V2= (-DV/~LVOEZ' j -COWRT
Z2L DV=DV-COWRT
122 DE=DSTCOYJRT
123 VOFF=VOFFtCONV3T
124 KOFF=EOX*COMVX'I
125 S-SOFF=90FF
126 C Oril>- -AL330 Fs vzrled, D-ALPEO, XIO, 9x13 a r e c o r s t = O
127 FFYPFILE="picch- d z t "
125 OEZN (UNIT=LI,FILZ=F-MPFILZ 1
L29 X I F I L = = " heaye,&c "
13 0 OPZN(UNIT=LS,FILZ=XïTILE]
131 C Kasrer Loop: Loo- once Eor each i n i t i s l âzçle
132 00 90 COUNT=L, NUX.NG
133 &LPEO=ALIST {COüNT) 'CONVEIT
134 F R I T Z (*, 21) ' I ~ i t i ê lPicch Angle: ' ,1"2LHO/CC>NTqRT
135 21 FOW-%-T ( A 2 2 , ?8 - 2 )
136 STATE=I
137 C Loop for each Speed R a t i o
138 DO 90 I N G Z X = l , NUM
139 C Variable I n i t i a l i t a t i o n
1 4O SMFE-a,F(S=O
l4 1 HOFF=F' HOFF
142 H~MIN=([DV+DE! / 2 + H O F I )
143 F EïMFX=( (DV-DH) /S+HOFE')
1-4 H~X=FJIMF~X
1-43 S 2 M I N = ( - (DV+DH) /2+HOFF)
140 E'-E2kLW= ( - (DV-DH) /2fHOFF)
147 X2MPX= F-E2MAX
148 USTA3= (UKIN+-INCR* (INDEX-1 1 ) * U L I N
149 HI?ELU=- 7-
150 LO?EAEY=I.
-Ï 51 DALP-JO=O.
152 XIO=O .
153 DXIO=O.
L54 C
155 PX=-O. 5
156 W=3, 5
157 XE-= O - 25
158 DT=O - I
155 EL=I*
160 $-=O. 163
161 B=0.0455
162 C=O - 335
163
164 --D=G .3
L-r)T-2
165 ZSTPFL,=O -0
166 ZZ'rXI=O, O
167 C W R I E (=, j ' VI= ' ,Vi/CONVRT
+
I T Y L E , INCREASE)
C
C C ~ l c u l a t eT-11 and F-12
F-I1=ZXP (-B*DT) 'F -II+DT/2T. '(9- * ( 3 D X I ( 4 ) i (O. 5-PX) +3DALE'E (4)-
2 63 ENDIF
264
-- .GT, (100- )
(EIP3A?<-LOPE-A.i<)
L;LSEIF(-QAS ) T4ZN
265 GOTO 70
266 ENDZF
267 ENDI?
268 C
2 6 9 C Xotation of a r r z ÿ s
270 DO 30 I=L,3
271 p..Lp~(I) = _ i ? L p ~ fil)
272 RECPLL (LOG-;=>!FE ( I + I )
273 DALPE (Z]=E-AL?X (1+1)
274 DDALPE (1)=DCGLPH (1-1)
273 XI (1)=XZ (I+I)
275 RECXI (LOOP) =XI (Irl)
277 D X I (Ij =DU1 ( T + I )
278 DDXL (1)=DDXI (I+1]
275 LESrl3DA(I) =L,A?Y3D.L (1-1)
280 30 CONTINUE
2 8 1 C To reach h e r e Fs cc achieve LCO
282 IF(TYPZ.NE.1) T%EN
283 30 4 0 T=3,100P-I,I?fEEQ
284 C W X T Z ( I , * ) (1-2)*9TrRZC-?LP(I)ICONVRT
285 CALI,
GR WITZ ( I 1-21 -DT,REC-ALP(11 / c o ~ ~R-CXI
T, !I / C O ~ V R T )
2 8 6 40 CONTINUE
287 1 R I N T *, ' LCO o c h i e v e c aï: Ur=', UST,2\3/ULIN, ' P i t c h znp=',
288 1 (A3.S (HIPEPL~S-LCJE-PLK)/ 2 ) /CONVRT
283 ELSZ
290 iF(STATE.NZ.2) THEN
m
IF(STATE-Sï3.lj LELNY-
313 ENDIF
314 ST,o-TE=L
3L.5 COTO 90
31670 CONTINUE
317 C U n s t a b l e oscillziion.
318 I? (TYOS - N5.1) TEEN
319 DO 50 I=3,L00P-I,IPFXQ
320 CiLL
GOTO 400
ELSE
INCwnSE=O
HRITE ( *, * ) INC--?SE, T
BOZF=F-HOFF
Z2M-X=F-X2MJX
IIIY!%X=F -ZIFGX
ENDIF
ENDIF
Eysïer3. f o r
520 EMZTF
521 C
522 RET'J'XN
523 ENS
524 C
5 2 5 C C a l c u l ê c e cnc recurn che F n i t i â l alc ces or- zze
. k d ,
&rics
526 C of a l g k c & xi
527 SCi32CGTIN3
V2, K I I I h ' , EIMMX<,I I Z X I K , E2Ei1;, ZHÜ, DT, DV, 95, V O X , <.EOFS, COh-v~T,
2 EOFF, F-;fLM?iX, F-ii2KnX
- -
ELSf
EY-FA=V2
ENDIF
ENDIF
EETURN
END
-.
*
Z Function to c a l c u l a t e the response f o r c e s l o p e for
jiven
for a
C gioen p i c c h cnqle z i p h a - N o t e c h z t here c h e r e is
curvature,
DDF=O .
-ETURN
END
RSAD (1,310)TISLE
2EAD (1,3101 SUBTITLE
E A D (1,320)DV, DE,VOFF, HOFF,L'LIN
READ ( 1 , 3 3 0 ) NUKLWG, UMINr UM.'X/ INCR
X A D ( I . 3 4 0 ) HMU,OMEGA
KVM= (UMAX-~1x1 /INCR+I
DO 3 0 0 I=I,EiUMANG
W . D ( l , 3 5 0 ) E L I S T (1)
CONTINUE
FOWAT (A)
FORMAT (OFS.2, F8.4)
FORmT(15,F7c4,F7.4,F774)
'XRITE (2,5101 ' B ACE/gr pzramerer fils '
WRITE(2,510) ' g '
NRfTZf2,510)' g '
WRLTE(2,5lO)'@page 5'
WRITZ(2,510)'@poqe Fnout 5 '
WEIITE (2,510)' @ l i n kpzgs c f:'
W m T E ( 2 , 5 1G ) ' Gdefaulï l i n e s r y l e i '
'NRITE ( 2 , s 10) ' @ d e f a u l = linewicïh L '
'?i'XITE(2,510~'@aefault c o l o r 1'
WRITE (2,5101 ' @default c h a r size 1- 000000'
WRITE(Sr510l'gdefault Zonï 4 '
M3ITE (2,5101 ' @ d e f a u l t font source O ' - - -
SNDIF
'@ view m i n 0.150000 '
-WRITE(2,51C) ~ ~',
~ K- nL- , L L51C) @ view m a x 0-850000 '
W K T Z (2,510)'@ n i e w p i n 0.150000 '
WRITE(2,510) ' @ view wax 0-850000 '
WRLTE(Zr5OO)' @ t i c l e ', GTITLE, ' ' II
WRTTS(2,5i0)' @ t i t l e font 4 1
r m r ~ = ( 510)
2 , '(3 sO e r r o r b a r riser on 1
hXITZ(2,510) 'C sO e r r o r b a r r i s e r l i z e w i d t h I f
WRITE (2,510)' @ sO e r r o r b a r riser l i n e s ï y l e 1'
MRITE (2,510)' @ s o x y z O. 000000, 0~000000 '
WRITE (2,500;' @ sO comtent " ' , O U T F I L E , ' " '
KRITE(2,510} ' @ x e x i s tick o n r
WRITZ(2,510)'@ xaxis t i c k najor 0.1 I
WRITE(2,SIO) ' @ x a x i s r i c k l a b e l on l
864
865 -- ---
FTEIITZ~2,510) '@
VIRA IL!^, 5L0) '@
- - '@
xaxis
xaxis
ticklzbel color I
t i c k l a b e l linewicch I '
s
8 67 XXTt(2.510) '@ X Z X ~ S Y L C X m l n o r on I
8 68 X R I T E ( 2 , 5 L O ) '@xaxis zick d o f a l ï 6 7
WRITE(2,SIC)' @
WRITE (2,510)' @ _
z e r o x a x i s t i c k l o b d t y p e auto '
z e r o x z x i s L l c k l z ~ e )l r e c 2
+ '
WRISE(2,510) ' @ z e r o x a x i s ï i c k l a b e l fo,mtaï d e c i n a l '
WRITS (2,510)' @ z e r o x a x i s t i c k l a b e l append " " '
XEIITZ (2,510) ' @ zeroxaxis ticklcbel prtptnd " " '
XRITE (2,510)' @ zeroxaxis z i c k l a b e l layout horizonzcl'
WZITE (2,5L0)' E zeroxzxis i i c k l a b e l SKIP O t
KRITE (2,510)' @ z e r o x a x i s t i c k in I
WRITE(2,310)' @ z e r o x a x i s t i c k rninor l i n e w i b t n I r
WRITE(2,SlO) ' @ z e r o x a x i s t i c k m i n o r l i n e s t y l e 1'
for
zeroxaxis - .
-LLCK log off 1
zeroxaxis ~ i c ks i z o 1-000000 s
ztroxzxis
zeroxaxis 5 t r oz=
--
r i c k minor s i z e 0 .5030001
I
zsroxàxis bz-r c o l o r 1 I
zeroxzxis bâr I l n e s ï y l s I r
~ I T (2.510)
E z e r oy a x i s bar l i n e w i d t h 1 '
KRITE(2,SlO) ' @ zeroyaxis t i c k najor g r i d o n '
WRITE(2,510) ' @ zeroyaxis t i c k minor g r i d o n '
WRITE (2,510)' @ zeroyaxis t i c k op both I
1
1sgend locqpe view
Legend laycuc O ?
leqend qa? 2 1
lqend hges I 1
r
legend length 4
1
leqend box o5f
leqend b ~ x f i l 1 off 1
. .
Legenc b c x fil1 wri? c s l c r l
l q e n d D o x fil1 r o l o r 9 '
l q e n d box fil1 p a t c s r n I'
leçend Dox c a l o r I 9
-
legend Svx l i n e w L d ï k I 1
l e g e n c box 11112scy13 -
- 1
l q e n d xl 0 - 8 t
l e g e n d y1 0 . 8 1
l e g e a a font 4 1
l e g e n d chcr s i z e I - G û S 3 0 3 '
legena I l n e s t y l e 1 1
l e g e n d linexidth I 1
l-e g e n c c o l o r L 1
1
crame OR
frame t:ee O 1
--
cyme Linescyle 1 1
1
frome l i n e w i d t h 1
frame c o l o r 1 1
f r a n e background c o l o r O '
IMAGE EVALUATION
TEST TARGET (QA-3)
APPLIED -
il lN14GE. lnc
= 1653 East Main Street
--.-
-- - Rochester. NY 14609 USA
---
-
--
--
-- Phone: 716/482-0300
Fax: 7161288-5989