Strength of Materials, Vol. 34, No.

6, 2002

ON TRANSVERSE VIBRATION OF AIRCRAFT LANDING GEAR

N. P. Plakhtienkoa and B. M. Shifrinb UDC 534

Transverse elastic-frictional vibrations of landing gear with respect to a hull of infinite mass during
an airplane high-speed motion on the runway are studied theoretically within the framework of the
nonlinear shift hypothesis. The authors have derived approximate amplitude-phase equations that
describe vibration in one-dimensional mechanical systems with arbitrary analytical velocity-
dependent friction characteristics.

Keywords: airplane, landing gear, vibration, shift, yawing, self-excited vibrations.

In most cases the failure of components of the landing-gear compartment is mainly due to the action of
unanticipated transverse, i.e., across the hull longitudinal axis, cyclic loads. The works [1, 2] present the results of
analysis of causes of failure of the main landing-gear strut and break-away of the splined-joint attachment block of a
long-life airplane F100. The combined transverse and torsional vibrations of the gear strut, as arising during the
airplane motion on a runway, were accepted as possible causes of the failure. In the work [3] we proposed a nonlinear
model for the analysis of transverse elastic-frictional vibration of gear struts and an airplane as a whole. When
constructing this model, we made the following main assumptions:
an airplane is represented by a mechanical system of two bodies (a hull and main gear struts) whose coupling
is compliant only during transverse relative displacement of the bodies of this mechanical system;
the airplane longitudinal velocity and the angle of yaw of the hull longitudinal axis are constant;
the transverse component of friction forces on rolling wheels is described within the framework of the
nonlinear-shift hypothesis [4].
In addition to [3], we assume here that the airplane hull is vibration-free and, using the mathematical model,
we analyze transverse elastic-frictional vibrations of landing-gear struts in the near-critical range of the shift angles.
The assumption of the vibration-free hull is made also during the investigation of the wheel shimmy phenomenon, it is
equivalent to a requirement that the hull should have an infinitely large mass in comparison to that of the landing-gear
struts. While studying this technical problem we obtained certain results regarding arbitrary one-dimensional
mechanical systems with nonlinear friction.
Approximation of a Friction Characteristic. Nonlinearity of the model is due to nonlinearity of the
dependence of the transverse component of friction force for a rolling wheel on the shift angle ϕ:

µ * (ϕ ) = µ max f (ϕ ), (1)

where µ * and µ max are the effective friction coefficient and its maximum value, respectively, f (ϕ ) is the odd
nonlinear function of the effective friction coefficient, f ∈[ −1, 1]. The function increases monotonically up to a
critical value ϕ * and then either remains constant or decreases. Let us study vibrations at positive near-critical shift
angles and pass on to a variable x = ϕ ϕ * in expression (1). We set x ∈[ x1 = 0. 8, x 2 = 1. 2] and consider five friction
characteristics f ( x ) (curves 1–5 in Fig. 1) in the above mentioned range of the relative shift angle. The first of these
characteristics is described by the quadratic function
a
Timoshenko Institute of Mechanics, National Academy of Sciences of Ukraine, Kiev, Ukraine. bState Flight
Academy of Ukraine, Kirovograd, Ukraine. Translated from Problemy Prochnosti, No. 6, pp. 79 – 88, November –
December, 2002. Original article submitted December 26, 2000.

584 0039–2316/02/3406–0584$27.00 © 2002 Plenum Publishing Corporation

 TABLE 1. Coefficients of the Friction Characteristics
Characteristic No. µ0 µ1 µ2 µ3 µ4 µ5 C0
in Fig. 1
2 0.5071 0.1417 1.3237 − 0.8632 − 0.3600 0.2509 0.9883
3 0.3662 0.5430 1.0705 − 1.0252 − 0.1873 0.2330 0.9899
4 0.3433 0.5912 1.0490 − 0.9552 − 0.3514 0.3235 0.9916
5 0.2954 0.6863 1.0877 − 1.0931 − 0.3427 0.3669 0.9932

Fig. 1. Friction characteristics.

f ( x ) = 2x − x 2 , (2)

while the others by the quintic polynomials

5
f (x ) = ∑ µ i x i . (3)
i=0

Table 1 gives the coefficients µ i and the average value of C 0 of the f ( x ) function within the above
segment,
x2
−1
C 0 = ( x 2 − x1 ) ∫ f (x )dx.
x1

For friction characteristic 1 (Fig. 1) we have C 0 = C 01 = 0.9867. The friction characteristics are shown in
Fig. 1 and are consistent with theoretical and experimental findings in [5–8]. (Numbers of the curves in Figs. 3–5
correspond to the numbers of the friction characteristics.) Note that polynomials (3) are approximations of four curves
of the one-parameter set of piecewise-parabolic curves:

f = κ + (1 − κ )(2x − x 2 ), (4)

where κ is a constant parameter,

κ =0 for x ∈[ x1 , 1] and 0 ≤ κ ≤ 1 for x ∈[1, x 2 ].

For characteristics 2, 3, 4, and 5 the parameter κ is 0.25, 0.5, 0.75, and 1.0, respectively. It should be
mentioned also that setting κ = 0 in expression (4) yields formula (2). Thus, in the subcritical range of the shift angle
all five curves are close to parabola. When κ =1 the f ( x ) characteristic described by equation (4) degenerates its
second section to become a straight line parallel to the axis of relative angles. Characteristic 1 consists of two
symmetrical branches, while the others consist of asymmetrical branches.

585
 Fig. 2. Mechanical model of vibration of landing-gear struts.

Mechanical Model of Vibrations of an Airplane Landing Gear. Let us consider an airplane as a system of
two bodies which are different in size. The larger body (the hull) executes a uniform straight-line progressive motion
with the angle of yaw of the longitudinal axis ψ and angle of yaw of velocity ψ a being constant and equal (Fig. 2).
Axes OXYZ are the fixed ground axes. A plane OXZ is the plane of the runway surface. The ψ angle is such that the
OX and OZ axes are parallel to the longitudinal and transverse axes of the airplane hull, respectively, and the ground
speed of the hull coincides with a selected course line, i.e., the runway axis which is shown as dashed line in Fig. 2,
over the whole motion section.
An equation of transverse motion of the smaller body, i.e., the landing-gear strut, is written as

M 2 Z&&2 + C ( Z 2 − Z1 ) = −R sgn( Z&2 ). (5)

Here, M 2 is the strut mass, Z1 and Z 2 are the coordinates of the hull and the strut, respectively, C is the rigidity of
the internal coupling between the hull and the strut, the dot indicates differentiation with respect to time t, and R is
the transverse component of friction between wheels and the runway surface,

R = µ max f (ϕ ) N ,

where N is the normal response of the runway surface. The Z1 ( t ) function is known, Z1 = Z&1 t, Z&1 = const. We
rewrite equation (5) as a set of two first-order differential equations and introduce the dimensionless phase variables

dϕ dτ = z − Ef (ϕ ), dz dτ = W − ϕ. (6)

The phase variables are the wheel shift angle ϕ and the dimensionless transverse strain of the struts z:

ϕ = Z&2 V , z = ω ( Z1 − Z 2 ) V ,

V = const is the run speed or the ground speed of the airplane (Fig. 2) and ω = (C M 2 )1/ 2 is the natural frequency of
vibration of the landing-gear strut. In (6): E and W are the constant positive parameters of the problem and τ = ωt is
the dimensionless time. The E parameter is directly proportional to µ max and tends to zero as V approaches the
take-off speed:
E = (µ max N ) ( M 2 ωV ). (7)

The parameter W = arctan (Z&1 V ) ≈ Z&1 V = ψ.

586
 Let us assume that the airplane runs for ∆t = 5–10 s at a speed V close to the take-off speed. (In [1] it was
found that the failure occurred upon vibration lasting 6 s.) For the sake of convenience of estimation of the order of
the E parameter, formula (7) is rewritten as

E = (µ max Ng ) ( mGωV ),

where g = 9.8 m/s 2 , G is the airplane weight, and m is the strut-to-airplane mass ratio. When m = 0.03, ω = 10 Hz,
V = 55.6 m/s, we obtain
E = 0.1µ max N G. (8)

The maximum value of the effective friction coefficient depends on runway surface condition and is of the
order ≤1. At speeds close to the take-off speed (second half of take-off roll or first half of landing roll) the ratio
N G <<1, hence it follows that E <<1. This conclusion allows using the averaging method [9] to obtain analytical
solutions to set (6). The calculations shows that for each type of airplane one can specify the run speed close to the
take-off speed, V* , whereby the dimensionless strain is numerically equal to the dimensional one expressed in meters.
This enables us to judge the extent of transverse loading of the landing-gear struts from the values z.
Equation (6) supplemented with relationship (2) or (3) is a nonlinear mathematical model of elastic-frictional
transverse vibration of landing-gear struts with respect to a hull of infinite mass. In the present work, using this model
we study vibrations in the near-critical range of the shift angle in order to clarify the causes for shortening the service
life of aircraft components.
Amplitude-Phase Equations. The f ( x ) function on a segment x ∈[ x1 , x 2 ] is represented by the Fourier
series
n
f ( x ) = C 0 + ∑ [C k cos( bk x ) + S k sin( bk x )], (9)
k =1

where C 0 , C k , S k , and bk are the known constants. The values C 0 are given in Table 1, the other constants can be
found from the known formulas of the Fourier series theory, but what matters for further calculations is only the
possibility of representing (9) in principle. The initial equations (6) are rewritten as

dx dτ = y − ε∆( x ), dy dτ = w − x, (10)

where ε = E ϕ * , w = W ϕ * , y = z ϕ * − εC 0 , and ∆( x ) = f ( x ) − C 0 . Note that w is a parameter of the hull

transverse velocity Z&1 .
We set ε = 0 and find a generating solution to the set of equations (10):

x = a sin θ + w, y = a cos θ, θ = τ + ϑ, (11)

where a and ϑ are certain constants. Note that the amplitude value of strain z is equal to A = aϕ * . A solution to the
set of equations (10) will be sought for in the form of (11), and in doing so, the symbols a and ϑ will be taken to
mean the sought-for functions of time. Substitution of (11) and (9) into (10) and the known elementary
transformations yield
n
da
= − ε sin θ ∑ {C k cos[ bk ( w + a sin θ )] + S k sin[ bk ( w + a sin θ )]},
dτ k =1
(12)
adϑ n
= − ε cos θ ∑ {C k cos[ bk ( w + a sin θ )] + S k sin[ bk ( w + a sin θ )]}.
dτ k =1

587
The expression in braces is represented as

{...} = C * k cos( bk a sin θ ) + S * k sin( bk a sin θ ), (13)

where
C * k = C k cos( bk w) + S k sin(bk w), S * k = − C k sin(bk w) + S k cos( bk w).

As to completeness of the mathematical model, the set of equations (10) is equivalent to (12). Let us average
the right-hand members of equations (12) over the total phase angle θ. In doing so, we take into account (13) and the
fact that the following expansions are valid for arbitrary a and θ:

cos( a sin θ ) = J 0 + 2 J 2 ( a )cos(2θ ) + 2 J 4 ( a )cos(4θ ) + ...,

(14)
sin(a sin θ ) = 2 J1 sin(θ ) + 2 J 3 ( a )sin(3θ ) + 2 J 5 (a )sin(5θ ) + ... .

Hereinafter, J k is the first-rank Bessel function of order “k.”

The averaging gives the approximate amplitude-phase equations

da n
dϑ
= − ε ∑ S * k J1 ( bk a ), = 0. (15)
dτ k =1 dτ

The second equation of (15) indicates that the ϑ angle is constant. Let us dwell on the first one. The Bessel
function is written in the form of the series [10]

J1 ( bk a ) = bk a 2 − ( bk a 2) 3 (12 ⋅ 2) + ( bk a 2) 5 (12 ⋅ 2 3 ⋅ 3) − ... . (16)

Substitution of (16) into the equation under consideration gives

da dτ = −εΦ( a ), Φ( a ) = µ 1* ( a 2) + µ 3* (a 3 16) + µ *5 ( a 5 384) − ... , (17)

where
n n n
µ 1* = ∑ S * k bk , µ 3* = − ∑ S * k bk3 , µ 5* = ∑ S * k bk5 , ... . (18)
k =1 k =1 k =1

A comparison between (9) and (18) shows that

µ *i = ∂ i f ∂x i , i = 1, 3, 5, ... . (19)
x= w

Thus, with ε <<1 and an arbitrary analytical function f ( x ) the equation for amplitudes is of the form (17),
where the µ *i coefficients are determined by expressions (19).
Determination of the Self-Excited Vibration Parameters of a Landing-Gear Strut. Let us study vibration
of a landing-gear strut using equations (17). Self-excited vibration is possible if the equation

Φ( a ) = 0 (20)

has nonzero real roots. Figure 3 shows the Φ( a ) curves plotted for w = 0.95 (dashed lines) and w = 1.05 (solid lines).
We assess the stability of the limiting cycle by the sign of the dΦ da derivative with a = a * , where a * is the
nonzero real root of Eq. (20). If the derivative is positive the limiting cycle is stable and a * is the amplitude of
self-excited vibration [11]. Thus, it can inferred that self-excited vibrations occur if the f ( x ) characteristic is not a
parabola and w >1. In view of the ratio of the maximum permissible speed of strictly cross wind to the take-off speed
we come to a conclusion that the hull transverse velocity parameter w can exceed unity.

588
 Fig. 3. Curves of the Φ( a ) functions.

Fig. 4. Amplitude of self-excited vibrations of landing gear a * as a function

of the hull transverse speed parameter w.

Let us dwell on the determination of the a * roots. Equation (20) is rewritten as

Φ( a ) = aF ( a 2 ) 2, F = pa 4 + qa 2 + c, (21)

where p = µ *5 192 , q = µ *3 8 , and c = µ 1* . To find the self-excited vibration amplitudes a * it is necessary to solve
the biquadratic equation F = 0. If the f ( x ) characteristic is a parabola, then p = q = 0. Figure 4 shows the a * ( w)
curves whose plotting required fulfillment of the set of inequalities

a * + w ≤ 1. 2, − a * + w ≥ 0. 8.

Transient Processes. Since an airplane’s high-speed run during standard take-off (landing) roll does not last
long, the duration of transient processes takes on special importance. To study such processes we now turn to equation
(17). In view of (21), the separation of variables gives

d ( a 2 ) [ a 2 F ( a 2 )] = −εdτ. (22)

589
 Fig. 5. On finding a preset amplitude of self-excited vibration of landing gear.

Then, upon integration for a parabola (characteristic 1 in Fig. 1) we obtain

a1 (T ) = a 0 exp( − cT 2), T1 = (2 c) ln ( a 0 a k ). (23)

Hereinafter, a i is the amplitude for the respective characteristic (Fig. 1), a 0 and a k are respectively the initial and
final amplitudes of vibration, T = ετ is the time parameter, and Ti is the time of reaching the amplitude a k . It is
evident from relationships (23) that the vibration is damped out when c > 0 (the ascending branch of the parabola),
the amplitudes remain unchanged when c = 0, and the vibration builds up when c < 0 (the descending branch). For
characteristics 2–5 (Fig. 4) we obtain [12]

Ti = Ti1 + Ti 2 , i = 2, 3, 4, 5,

Ti1 = (0. 5 c) ln[(a 0 a k ) Fk F0 ],
4

Ti 2 = − [ p (2cD )]ln [ ( F10 F1k )(F2 k F20 ) ],


 (24)
Fk = F ( a = a k ), F0 = F ( a = a 0 ), D = [( q − 4 pc)] ,
2 2 2 2 2 1/ 2

F = 2 pa 2 + q − D, F = 2 pa 2 + q + D,
 1k k 2k k
F = 2 pa 2 + q − D, F = 2 pa 2 + q + D.
 10 0 20 0

Note that Ti1 = T1 when p = q = 0.

Using formulas (23) and (24) we find the values of wi = wi * (i = 2, 3, 4, 5) whereby the self-excited vibration
amplitudes a i * will be equal to a preset value. (The sought-for parameters can be found, to a first approximation,
from Fig. 4.) Assume that a i * = 0.1. Then, we have w2* = 1.00866, w3* = 1.01856, w4* = 1.03357, w5* = 1.06461.
Using formula (24) we construct diagrams of the transient processes for self-excited vibrations with a steady-state
amplitude a i * = 0.1 (Fig. 5). For comparison, Fig. 5 shows also the curves of building-up of vibrations of the system
in the case of the friction characteristic being a parabola, which were plotted by formula (23).
In view of formula (8), let us estimate the interval of the dimensionless time parameter ∆T, which
corresponds to that of real time ∆t. When expanded,

 N 
∆T =  0.1µ max ω  ∆t.
 Gϕ* 

590
The ϕ * angle is of order 0.1–0.2 rad, the µ max coefficient is 0.1 to 1.0 depending on the presence of moisture on the
runway. Assume that ∆t =1 s, ϕ * = 1.5 rad, µ max = 0.5, ω =10 Hz. Then, we find that the realizable interval of the
dimensionless time parameter ∆T will be approximately 7π N G, where N G <<1.
Using numerical integration of equations (10) by the Runge-Kutta method, we estimate the order of ε,
whereby the amplitude equation (17) yields acceptable results. The initial values of variable x and y are denoted by
x 0 and y0 , respectively. We set w = w* , x 0 = w, and y0 = 1. 2 − w. A comparison between the numerical integration
results and the data found by formulas (24) has demonstrated that the results obtained by Eq. (17) are correct even
when ε =1.
Thus, in the present work we studied the elastic-frictional vibration of a landing-gear strut with respect to a
hull of infinitely large mass, which executes a high-speed uniform motion. It has been found (Fig. 4) that self-excited
vibrations with strain amplitudes up to 4 cm are likely to occur. Such vibrations are dangerous and can cause fatigue
failure of the gear-compartment components. Furthermore, in view of the finite hull mass the above mentioned
vibrations are responsible for intensive manifestation of lateral oscillations of those aircraft components whose partial
frequency is close to the natural frequency of the landing-gear struts. In case the initial value a is somewhat different
from the a * amplitude, a shorter segment of the transient process, which is similar to that illustrated in Fig. 5, is
realized.
It has been found that during the first half of landing roll or the second half of take-off roll there arise
conditions whereby the transverse elastic-friction vibrations of landing-gear struts occur in a similar manner to
vibration in a linear system without damping. We have studied theoretically the behavior of one-dimensional
mechanical system during speed-dependent nonlinear nonmonotonic friction. Amplitude-phase equations have been
derived for an arbitrary analytic function of friction. A difference has been shown in the system behavior in the
near-critical speed range in symmetrical and nonsymmetrical segments of the friction characteristic prior to and after
the critical speed value.

REFERENCES

1. R. Den Hertog, “Problem solved,” Aircraft Eng., 62, No. 12, 2–4 (1990).
2. R. Van der Valk and H. B. Pacejka, “An analysis of a civil main gear shimmy failure,” Vehicle System
Dynamics, 22, 97–121 (1993).
3. N. P. Plakhtienko and B. M. Shifrin, “Transverse elastic-frictional vibrations of an airplane moving on the
runway,” Prikl. Mekh., 37, No. 5, 136–143 (2001).
4. M. A. Levin and N. A. Fufaev, Theory of Rolling Motion of a Deformable Wheel [in Russian], Nauka,
Moscow (1989).
5. T. I. Ligum, S. Yu. Skripchenko, and A. V. Shishmarev, Aerodynamics of Tu-154B Airplane [in Russian],
Transport, Moscow (1985).
6. Yu. L. Bychkov, “Determination of traction coefficient between aircraft tire and artificial runway,” Trudy
GosNIIGA, Issue 233, 34–38 (1985).
7. V. A. Sannikov, “Determination of lateral skid resistance of wheels by flight test results,” Trudy GosNIIGA,
Issue 192, 111–119 (1980).
8. P. A. Davis, V. J. Martinson, T. J. Yager, and S. M. Stubbs, “26 × 6.6 radial-belted aircraft tire performance,”
SAE Techn. Pap. Ser., No. 912157 (1991).
9. N. N. Bogolyubov and Yu. A. Mitropol’skii, Asymptotic Methods in the Theory of Nonlinear Vibrations [in
Russian], Fizmatgiz, Moscow (1963).
10. H. B. Dwight, Tables of Integrals and Other Mathematical Data, The Macmillan Company, New York
(1961).
11. A. N. Obmorshev, Introduction to the Theory of Vibrations [in Russian], Nauka, Moscow (1965).
12. Yu. A. Brychkov, O. I. Marichev, and A. P. Prudnikov, Tables of Indefinite Integrals [in Russian], Nauka,
Moscow (1986).

591