Escolar Documentos
Profissional Documentos
Cultura Documentos
Gerhard Somieski
Deutsche Forschungsanstalt fiir Luft- und Raumfahrt e. V
Institut fiir Robotik und Systemdynamik, Postfach 1116, D-82230 Wepling, Germany
Manuscript received February 3, 1991; revised version April 24, 1997.
Abstract Shimmy oscillations are still a problem in design and operation of aircraft landing gears, and accurate
and appropriate analysis is required to master the task. Based on a nonlinear model of the mechanics of
the landing gear and tire elasticity according to elastic string theory, some well known linear and nonlinear
mathematical methods are applied to the shimmy analysis of a simple model of a nose gear: Computing
eigenvalues, solving analytically the stability boundaries with a parameter space method, getting limit
cycles by analytical formulae from describing functions, and last but not least numerical simulation of
time histories. It seems that linear or quasi linear methods and analytical solutions are well suited to
obtain extensive insight, respecting limitations of these methods. Numerical simulation on the other hand
is a valuable tool for pointing out specific effects of a nonlinear system in large amplitude regions.
Keywords: Shimmy analysis - Nose landing gear - Limit cycles - Instability - Describing functions -
Analytical solutions.
tire damping moment from tread width [Nm] dynamics of the landing gear and to enable a solution
tire aligning moment [Nm] of the stability boundary and the limit cycle equations
describing function analytically. Extensive use of software is made, e.g.
describing function of tire moment M=/F, for computing eigenvalues and numerical simulation,
describing function of tire side force F,/Fz obtaining analytical solutions for stability boundaries
paraineter or parameter set or limit cycles, and for graphical displaying.
yaw rate of landing gear [rad/s]
Laplace variable II - NONLINEAR MATHEMATICAL MODEL
time [s]
input in dynamic system The nonlinear mathematical shimmy model, com-
wheel forward velocity [m/s] pared to a similar model in [2, 5, lo], consists of the
state of dynamic system torsional dynamics of the landing gear, the forces and
output of dynamic system moments, and of approximations to describe the tire’s
lateral shift of leading tire contact point [m] elastic lateral qualities, see Figure 1.
slip angle or deflection angle of tire [rad]
amplitude of limit cycle of a: [rad]
limiting slip angle for aligning moment
[deal
= A/S =amplitude of steady oscillation
limiting slip angle for tire side force [deg]
constant of tread width tire moment
[Nm’/rad]
eigenvalues of linear differential equations 4aw
relaxation length of tire deflection [m] angleW
time constant of tire elasticity Es] tire side lateral shift y1 of leading
yaw angle of landing gear [rad] force Fy contact point
frequency of a-oscillation [rad/s] Y
frequency of limit cycle of Q [rad/s]
Fig. 1. - Top view of shimmy dynamics model.
Torsional Dynamics
I - INTRODUCTION
The torsional dynamics of the lower parts of the
In the landing gear of a taxiing aircraft, shimmy landing gear - sliding member of the shock absorber
is a state of self-excited oscillations, caused by the and wheel - is described by a 2nd order differential
dynamic reaction forces between elastic tires and the equation for the angle $ about the vertical z-axis with
ground. In shimmy analysis there are different ways external moments Ml to n/r,:
of modeling elasticity of tires: Either the elastic string
theory of v. Schlippe, Dietrich, Pacejka [ 1, 21, or
Moreland’s point contact method [3] are applied. Both Iz$ = Ml ($) + Mz (4, + M3 (a) + M4 (d/V) (1)
methods seem to give adequate results [4, 51, though
they are not easily comparable [6]. Most studies Moment Ml is a linear spring torque provided by
treat the torsional dynamics of the landing gear as a the turning tube and the torque link and iV2 is a
nonlinear multi degree of freedom mechanical system, combined damping moment from viscous friction in
which is simulated [7]. After linearization eigenvalues the bearings of the oil-pneumatic shock absorber and
are computed, from which stability in a small region from shimmy damper:
around nominal conditions can be derived [8].
But there are other efficient linear or quasilinear
methods, e.g. analytically finding stability boundaries
in parameter space and limiting amplitudes by Tire Force and Moments
describing functions, which provide comprehensive Lateral tire deformations due to side slip and yaw
answers. Additionally, for the nonlinear region the rate cause a tire moment n/r, and a tire damping
system can be analyzed by simulation. All these moment n/r,, respectively. Moment MS is composed
methods are displayed here, for details see [9]. Data of aligning torque Mz about the tire’s center and of
used are representative for the nose landing gear of cornering force FY acting with caster e as lever arm:
a civil aircraft of 10 tons of weight. The benefits of
the simple model are to give sufficient insight into the MS = Mz - eFy (3)
Aerospace Science and Technology
Shimmy Analysis of a Simple Aircraft Nose Landing Gear Model Using
Different Mathematical Methods 547
+
Aligning moment Mz/Fz
Instead of complicated functions, like “brush type torsional spring rate c -100000 Nmlrad
tire model” [lo] or “magic formula” [ 111, simpler but
still realistic nonlinear approximations are used, which side force derivative CF~ 20 Urad
only need force and moment derivatives as parameters.
For cornering force a saturation function with linear moment derivative ckfru -2 m/rad
pieces is selected, with a limiting slip angle of 6 = 5
degrees: 0 -50 Nmlradls
torsional damping constant Ic
Fy = cFaaFz for Q < 6
Fy = CF~SF~ sign (cr) for cr 1 6 1 (4) tread width moment constant IF. -270 Nm2/rad
For the coefficient K, one can find in [2, lo]: For a first understanding of the dynamics, the
nonlinear model is linearized and eigenvalues are
6 = -0.15 a2 cFe FZ (7) computed, applying routines from [ 12, 131.
$1
+r YI=V$
+ (e - a) 7j (8) [j] = [; ‘;: y [I]; (11)
From yr an equivalent slip angle is formed, used to
compute cornering force and aligning moment in this where system matrix A consists of the coefficients
instationary model:
a!Zarctana=E (9) C k /E.
fs Cl = -> cz=z+I/I;->
Thus the model is linear with the exceptions of tire I,
force and moment. Parameters (table 1) are: (CM~ - ecFa) F, -V
c3 = , c4=e-a, cj=-.
I*0
p = {v a; e, I,, Fz, c, CFcu, cniior, k; 4 0) (10) 412)
1997; no 8
548 G. Somieski
Characteristic Equation and Eigenvalues The case X = iw = 0 marks the starting point of all
real frequencies and limits the admissible solutions in
From (11) the characteristic equation for the parameter space. At w = 0 the root boundary marks
eigenvalues X follows by forming det (A - M) = 0: the change of a real eigenvalue from stable to unstable
(Real Root Boundary) and for w > 0 the change
of oscillatory eigenvalues (Complex Root Boundary).
x3-(cz+c5) X24-( C2C$-C~-C3Q) A+= 0 (ClCg-VC3)
Conventional stability methods, like Routh or Hurwitz
(13) criteria, see [ 151, can be applied as well.
In evaluation of the roots of this equation, we choose
nominal values for c and k = -10 Nm/rad/s, but vary
velocity V in the range {5.0...85.0 m/s} and caster Analytical Solutions of Stability Boundaries
length e in {-0.5...0.5 m}. Using software for formula manipulation [ 161,
The linear system (11) has a pair of complex roots analytical solutions are found.
X1, 2 and a stable real root X3. For caster lengths
between -0.1 and 0.35 m and for velocities above Stability Boundaries in e-V-Plane
25 m/s the system is unstable due to positive real
parts of the complex roots, Figure 3. In the e-V-parameter space we find a quadratic:
Method
The characteristic equation (13), depending on the
complex eigenvalues X and on parameters pi, ~2, ps,
is solved for parameters pl, p2, such that the stability
condition - poles left of the imaginary axis of the
complex plane - is fulfilled, see [14].
Inserting X = iw, and separating the complex
equation into real and imaginary parts, we arrive at
terms for the desired parameters pl and pz, depending
now on a third parameter p3 and frequency w. If
solvable, we have analytical solutions for pl, ~2, 0 20 40 60 80 100 120 140 160 180
velocity - V [m/s]
which can be evaluated for interesting ranges of w
and pa. Here we select p1 = V, pz = {k, e}, and Fig. 4. - Stability boundaries in e - V-plane, variation of spring
P3 = {c,+F). stiffness c.
bl = V2a
b2 = V31z - Fzclvcr eVg + 2~Vo - u2cV
+ F,cpl,aVa - F,ecF,aVa 0 10 20 39,,W+,,5p,, ,6m4C,70 60 90
+ Fze2cF,Vn
Fig. 6. - Stability boundaries in k - V-plane, variation of vertical
b3 = (IzzcrFzcMcv- IzeFzcMa + d, force Fz.
(19)
- I=gF,ecF, + FzcMcuaIz
- FzecF,aIz + E,e”cF,I,) V2 In principle, we see:
+ Fzchl,l,aoK - F,ecF,aaK lUnder a certain damping value shimmy occurs,
depending on velocity
+ 6/2 - cT2cK. lUnder small velocities (V< 16 m/s) there is
- F,cnt,eglc + Fze2cF,an stability for all values of damping constant k.
Both solutions of eq. (18) are valid, but one branch
is related to positive and thus physically meaningless V - SIMULATION OF NONLINEAR SYSTEM
values of k.
In [9] all parameters are varied, but here the results By numerical simulation [17] the dynamics of the
for varied parameters p3 = {g, F,} are discussed. system (eqs. 1 to 9) can be solved with respect to
Variation of relaxation length a, Figure 5: Small time, even for large amplitudes where linearization no
values of 0 (0 ~0.1 m) cause more instability at longer holds. Changing initial conditions or variation
small velocities and more stability at large velocities. of parameters give detailed insight. On account of long
Large values (Q > 0.1 m) cause more stability at small simulation times, much data output and sometimes
velocities and more instability at large velocities. numerically complicated integration of model, only
Variation of vertical force F,, Figure 6: Values of a limited number of cases can be run to check the
F, change due to weight, acceleration, and braking system. All results confirm those from the linear
at landing and taxiing. Larger values of F, and analysis with additional information for nonlinear
V need larger values of -k for stability. Below regions.
V = 16 m/s there is no instability for negative
damping coefficients. Time Histories and Phase Diagrams
Variation of relaxation 0
-5 For stable parameter cases (damping constant of
2 120 k = -100 Nm/rad/s) the nonlinear results are quite
similar to the linear ones, Figure 7.
g 100
But for unstable cases (weak damping constant of
p.L 80 k = -10 Nm/rad/s), a stable limit cycle occurs based
on nonlinear tire characteristics. For small initial yaw
60 angles (li, = 0.01 rad) the limit cycle is approached
2 from inside (Fig. 8): and for large initial conditions
5 40
($ = 1 rad) from outside amplitudes, Figure 9.
E
20
F
'E. 0 Amplitude of Limit Cycle
E
$ -2ou '//lz ' ' ' ' 1 ' To obtain the amplitudes of the limit cycle,
0 10 20 30 40 50 60 70 80 90 simulation over a time of 4 s is made. Then a stable
velocity - V [m/s]
limit cycle has developed if there is shimmy. From the
Fig. 5. ~ Stability boundaries in k - V-plane, variation of relaxation time histories one can record frequency and amplitude
length CT. as functions of varied parameters V, k.
1997. no 8
550 G. Somieski
; / / ; ; : ;
z 0.4
-E
f=o.3
a,
5
s 0.2
yaw angle - q~ [rad]
N=si=fi+&jw=~+jfz
A
X Aw A A
(23
Torsional dynamics Tire elasticity
For odd symmetric nonlinearities the coefficient ti w pl
fi is zero, and, for dynamic nonlinearities N is a
function of w, too. For tables of describing functions __) L(s) b L2(s)
see [ 191. The method is applicable even for extreme c
nonlinear model elements like ideal relays, where Tire side force
Taylor linearization is not meaningful and therefore a
no eigenvalues or stability boundaries can be found.
As long as the system output is stationary with -eF, t &C”~ t
one dominant frequency, good approximations for
limit cycle oscillations even for large amplitudes are
gained. In case of doubts, the hypothesis for one
basic harmonic must be checked by means of spectral . M, e Nd”) 4
analysis of simulation results. For infinitely small a
amplitudes the results are identical to those from Tire aligning moment
eigenvalue or parameter space stability analysis for
critical stability. Fig. 12. - Block diagram of linearized system.
552 G. Somieski
To solve the system equation (29) analytically, The describing function for the side force F,/F,
approximations of the nonlinearities or of the is a rather complicated function Nz(A), consisting of
describing functions must be used, that lead to arcsin, square root and the argument, see Figure 14,
algebraic terms in N. Especially polynomials of order curve “original”:
n are well suited, because they have polynomial type
describing functions of order n - I. This method is
now applied to the nonlinear tire moment and force
characteristics. (34)
Results of Limit Cycle Computations The resulting formula for amplitude is:
Limit cycles are computed analytically and com-
pared with numerical simulation results. 7r -9k2V” - 274860kV - 30V3k
O!lim = ~
The system equation (29) is solved for frequency 81OOV f315900V2 - 73556100
and amplitude using symbolic manipulations. The (40)
solutions are imported into the simulation software, The amplitudes, Figure 16, increase with velocity
where this static model is evaluated in parameter V and decrease with more damping k. Limit cycle
variation loops for velocity and damping constant. starts at V = 15 m/s for k = 0, reaching a maximum
The formula for frequency shows, contrary to the of 0.21 rad.
simulation result, a weak dependency on velocity and Good coincidence for small and medium amplitudes
damping coefficient: at k = {-20, -30, -40 Nm/rad/s}, compared with simu-
lation results, is shown in Figure 16. For low damping
rnd effects and larger amplitudes (olim 2 0.2 Nm/rad/s)
W]im30270 + kV
= ~ (39)
3 the differences are larger. The structure of the
Frequency values of 3 18 rad/s at k = 0 N&t-ad/s solution with respect to velocity and damping constant
and 328 rad/s for k = -40 Nmlradls and V = 100 approximates the simulated result well.
m/s, see Figure 15, differ by a few percents from the
simulated value of 314.2 rad/s.
VII - MBS-SIMULATIONS
Shimmy limit cycle by describing function
Applying software for multi-body-system (MBS)
3 340 modeling and simulation [20], the simple shimmy
3
E model described is embedded into a more complicated
-E 335 model for the vertical and lateral dynamics of the
z landing gear. Effects of vertical compression of an oil-
3 pneumatic shock absorber, of tire vertical elasticity, of
’ 33c rotating dual wheels, bending of the wheel axle and
6 the leg in fore and side direction are considered. This
s
g 325
E yaw
1
rate - + [rad/s]
p 32C
E
E
* 315
I1 1-1 I I I I1
20 80 100
vePoOcity- $rrn/s]
‘ / /
-0.4 -0.2 angleO
Fig. 15. - Limit cycle frequency w of shimmy oscillations versus yaw - I [ra$
velocity V and damping constant k.
left tire lateral deflection - yl [m]
0.03
Shimmy Limit Cycle
0
-0.03
55 0.24
-0.06
time - t [s]
$ 0.18 yaw angle - W WI
c 0.4
.50.12
2
6v 0.06
.z -0.4
=E 0 Y
0.25 0.75 1.0
time [s]
20 40 60 80 100
velocity - V [m/s] Shimmy oscillations at landing
V=50 m/s, k=-10 Nmlradls, K = 0 Nm’“2/rad
MBS- model consists of 25 states in 12 degrees of in describing functions analysis, a quasi linearization
freedom. Some simulations of landing and taxiing method applicable even for large amplitudes but
are performed, showing gear walk, shock absorber restricted to systems with low pass characteristics.
deflection, and torsional oscillations. By proper selection of algebraic approximations in
Thus MBS-simulation may serve as a completion the interesting range of amplitudes, the equations
and verification of the shimmy system dynamics in are solved analytically. At vanishing amplitudes the
a more realistic model environment with additional results are identical with linear stability analysis.
degrees of freedom, where some of the parameters, Comparisons between the results reveal a close
like vertical force or velocity, are dependent on time. approximation of the limit cycle if the describing
The results from the previous analysis methods, functions fit the amplitude range.
showing stable or unstable shimmy oscillations, are In MBS-simulation an enlarged model of bodies,
verified. In Figure 17 a landing case is simulated. joints and degrees of freedom for landing gear
dynamics is formed, and the simulations confirm the
dynamics of the previous analysis. This kind of
VIII - CONCLUSIONS simulation can be used to verify shimmy results.
Different analysis methods for linear and nonlinear In summary, the methods are characterized in
systems are applied to investigate shimmy of aircraft table 2.
wheels. The main purpose of the work was to In further investigations, modeling of additional tire
demonstrate the applicability of the analysis methods. effects like gyroscopic coupling, or of more degrees
A simple, nonlinear dynamic model of a nose of freedom like rotation of the dual wheels or lateral
gear is described. The parameters chosen influence bending of the nose gear are planned. For validation,
severely the results, and are to be established exactly comparisons of analysis and measurement results are
for a realistic system under investigation. Here they necessary.
are varied in parameter variation loops. For this
simple model one finds, that the tendency for shimmy
increases with growing velocity, lower damping values REFERENCES
and increasing vertical force.
The relation between the different methods can be
characterized as follows: If the nonlinear model can be [l] v. Schlippe B., Dietrich R. - Das Flattem eines bepneu-
linearized, the eigenvalues show the detailed structure ten Rades. Bericht 140 der Lilienthal-Gesellschaft,
of linear dynamics at small amplitudes. The stability 1941 (NACA Report TM 1365, 1941).
boundaries are derived with a parameter space method [2] Pacejka H. B. - The wheel shimmy phenomenon. A
and analytically expressed as functions of parameters. theoretical and experimental investigation with partic-
In judging shimmy tendencies, stability plays a key ular reference to the non-linear problem. Dissertation,
role for the design engineer. De& University of Technology, 1966. ’
[3] Moreland W. J. - The story of shimmy, J. Aeronautical
By numerical simulation, the nonlinear shimmy Sciences, 1954, 21, 793-808.
model is solved in time domain for large amplitudes. [4] Krabacher W. E. - A review of aircraft landing gear
For stable conditions the linear analysis results are dynamics. AGARD R 800, 1995, 2-l to 2-12.
confirmed. For unstable conditions oscillations are [5] Collins R. L. - Theories on the mechanics of tires and
limited in a stable outer limit cycle, shimmy occurs. their applications to shimmy analysis, J. Aircraft, 1971,
If dry friction is added to the model, another unstable 8, 271-277.
inner limit cycle can be observed. [6] KGnig K. - Unsteady tire dynamics and the application
Amplitude and frequency of the outer limit cycle thereof to shimmy and landing load computations.
are evaluated by extensive numerical simulations and AGARD R 800, 1995, 8-l to 8-9.
Aerospace Science and Technology
Shimmv Analvsis of a SimDle Aircraft Nose Landing Gear Model Using.
Differek Malhematical Methods 555
[7] Baumann J., Barker C.R, Koval L. R. - A nonlinear 1141 Ackermann, J. - Robuste Regelung. Springer Verlag,
model for landing gear shimmy. ASME 91-WA -DSC New York, Berlin, 1993.
14, Atlanta, GA, 1991.
[W Follinger, 0. - Regelungstechnik. Elitera Verlag,
[8] van der Valk R., Pacejka H. B. - An analysis of a Berlin, 1972.
civil aircraft main gear shimmy failure, Vehicle System
Dynamics, 1993, 22, 97-121. LlGl Char B. W., Geddes K. O., Gonnet G. H., Leong B.
[9] Somieski G. - Analyse des Flatterns eines einfachen L., Monagan M. B., Watt S. M - Maple V. Language
Modells eines Flugzeugbugfahrwerks. DLR internal Reference Manual. Springer Verlag. New York, Berlin,
report IB 515-96-15. 1996. 1991.
[lo] Pacejka H. B. - Modelling of the pneumatic t171 Somieski G. - MASSIV-Modelling, simulation, and
tire and its impact on vehicle dynamics behavior. evaluation. User manual version 3.0. DLR internal
Lecture Series DR 6.04, Carl-Cranz-Gesellschaft e.V., report IB 515-93-27. 1993.
Oberpfaffenhofen, 1992.
[181 Follinger 0. - Einftihrung in die nichtlinearen Systeme.
[l I] Bakker E., Nyborg L., Pacejka H. B - Tyre modelling Regelungstechnik, 1979, 27, A9-A21.
for use in vehicle dynamics studies. SAE Paper no.
870421, Feb. 1987. u91 Gelb A., Van der Velde W. E. - Multiple-input
[12] Melsa J. L. - Computer programs for computational describing functions and nonlinear system design. MC
assistance in the study of linear control theory. MC Graw Hill Book Company, New York, 1968.
Graw Hill Book Company, New York, 1970. r3.I Korti.im W., Rulka W.; Eichberger A. - Recent
[13] Smith, B. T. et al. - Matrix Eigensystem Routines- Enhancements of SIMPACK and Vehicle Applications.
EISPACK Guide. 2nd ed. Springer Verlag, Berlin, In European Mechanics Colloquium, EUROMECH
1976. 320, Prague, June 6-8, 1994.
1997, no 8