Jtam-A4.dvi


JOURNAL OF THEORETICAL

AND APPLIED MECHANICS

52, 1, pp. 107-115, Warsaw 2014

DYNAMIC ANALYSIS OF TORSIONAL DISCRETE-CONTINUOUS

SYSTEMS WITH POSITION-DEPENDENT VARIABLE INERTIA

Amalia Pielorz

Kielce University of Technology, Faculty of Management and Computer Modelling, Kielce, Poland

e-mail: apielorz@tu.kielce.pl

In the paper, dynamic investigations of multi-mass discrete-continuous systems torsionally
deformed are performed taking into account the position-dependentmassmoment of inertia
of one of rigid bodies. It is assumed that the rigid body representing a motor moves with
a variable velocity. An originally nonlinear problem is linearized, and the wave approach
leading to solving equations with a retarded argument having variable coefficients is used
for the solution. Exemplary numerical calculations are done for a three-mass system.

Key words: variable inertia, discrete-continuous systems, wave approach

1. Introduction

Variable inertia problems play an important role in dynamics of various mechanisms and ma-
chines undergoing torsional deformations. It appears that in rotating and vibrating systems, the
mass moment of inertia of rigid bodies can depend on the angular displacement. For example,
in the crankshaft of a multi-cylinder engine the total moment of inertia of the rotating parts
reduced to the axis of the crankshaft is not constant but it depends on the location of the
cranks, see Hesterman and Stone (1995), Koser and Pasin (1995, 1997), Pasricha and Carnegie
(1979), Sinha andButcher (1997), Turhan andKoser (2004), Zajączkowski (1987). Systemswith
position-dependent inertia in the technical literature are analyzed using discrete models, cf Pa-
sricha andCarnegie (1979), Sinha andButcher (1997), Zajączkowski (1987), discrete-continuous
models, cf Koser and Pasin (1995, 1997), Pielorz and Skóra (2006, 2009a) or a finite element
method, cf Turhan and Koser (2004). Two-mass discrete-continuous systems were studied by
Koser and Pasin (1995, 1997) while multi-mass systems were discussed by Pielorz and Skóra
(2006, 2009a). In the papers by Koser and Pasin (1995), Pielorz and Skóra (2006, 2009a) it is
assumed that the rigid body representing a motor has a constant velocity.
The aim of the present paper is to generalize the considerations performed by Koser and

Pasin (1995, 1997), Pielorz and Skóra (2006, 2009a). This is done for multi-mass systems with
the assumptions that the rigid body representing the motor moves with a variable velocity. In
the discussion, a wave approach is applied similarly to studies of discrete-continuous systems,
see e.g. Nadolski and Pielorz (2001), Pielorz (1999, 2003, 2007, 2010), Pielorz and Skóra (2006,
2009b). In the numerical calculations, the effect of variable inertia on the behavior of a three-
mass system is considered, including parameters describing the motor.

2. Governing equations for a multi-mass system

Consider the multi-mass discrete-continuous system torsionally deformed as shown in Fig. 1. It
consists of rigid bodies connected by means of the shafts. The central axis of shafts, together
with elements settled on them, coincide with themain axis of the system. The x-axis is parallel
to themain axis of the system, and its origin coincides with the position of the left-hand end of



108 A. Pielorz

the first shaft at time instant t = 0. The equations of motion for the shafts are classical wave
equations.

Fig. 1. Discrete-continuousmulti-mass torsional system

The i-th shaft, i=1,2, . . . ,N, is characterized by the length li, density ρ, shearmodulus G
and polar moment of inertia I0i. The i-th rigid body of themodel is characterized by themass
moment of inertia Ji. Themassmoment of inertia of the last rigid body depends on its angular
displacement, i.e., JN+1 = JN+1(θN). The considered system can represent shafts connected to
mechanisms with position-dependent inertia.
The first rigid body, having the constantmassmoment of inertia J1, represents an electrical

motor. It is loaded by themotor torque M1 =M0+K(Ω0−∂θ1/∂t), where M0 is the nominal
torque, Ω0 is the nominal angular velocity and K is the slope of the motor characteristic. The
last rigid body is loaded by an equivalent resistance torque equal to M0. Damping in the shafts
is neglected. The study is concentrated in a forced response resulting from the variable inertia
of (N+1)-th rigid body.
On the above assumptions, the determination of displacements θi of the elastic elements of

the considered system is reduced to solving N equations

∂2θi
∂t2
− c2
∂2θi
∂x2
=0 i=1,2, . . . ,N (2.1)

with the following boundary conditions

J1
∂2θ1
∂t2
−GI01

∂θ1
∂x
=M0+K

(

Ω0−
∂θ1
∂t

)

for x=0

θi(x,t)= θi+1(x,t) for x=
i
∑

k=1

lk i=1,2, . . . ,N −1

−Ji+1
∂2θi
∂t2
−GI0i

∂θi
∂x
+GI0,i+1

∂θi+1
∂x
=0 for x=

i
∑

k=1

lk i=1,2, . . . ,N−1

JN+1
∂2θN
∂t2
+
1

2

dJN+1
dθN

(∂θN
∂t

)2

+GI0N
∂θN
∂x
=−M0 for x=

N
∑

k=1

lk

(2.2)

where c2 =G/ρ is the wave speed. The condition in the cross-section x= l1+ l2+ . . .+ lN can
have various forms. Here it is assumed in the form suggested by Koser and Pasin (1995, 1997).
Boundary conditions (2.2) aremore general than those in thepapers byPielorz andSkóra (2006,
2009a). They differ in the condition for x=0.
The above problem is nonlinear. The paper is a generalization of the studies presented by

Koser and Pasin (1995, 1997), Pielorz and Skóra (2006, 2009a), so in analogy to these papers,
it is linearized by introducing the following new unknown functions

θi(x,t)=Ω0t+αi(x,t)−
M0
GI0i

(

x−
i
∑

k=1

lk

)

i=1,2, . . . ,N (2.3)



Dynamic analysis of torsional discrete-continuous systems... 109

the variable mass moment of inertia JN+1(θN)= JN+1(Ω0t+αN) with its first derivative

dJN+1
dθN

=
JN+1
dθN
(Ω0t+αN)

are expanded in the Taylor series around Ω0t, and the second order together with higher order
terms are neglected.

Then the determination of functions αi is reduced to solving N wave equations

∂2αi
∂t2
−c2
∂2αi
∂x2
=0 i=1,2, . . . ,N (2.4)

with linear boundary conditions

J1
∂2α1
∂t2
+K
∂α1
∂t
−GI01

∂α1
∂x
=0 for x=0

αi(x,t)=αi+1(x,t)+
M0
GI0,i+1

li+1 for x=
i
∑

k=1

lk i=1,2, . . . ,N−1

−Ji+1
∂2αi
∂t2
−GI0i

∂αi
∂x
+GI0,i+1

∂αi+1
∂x
=0 for x=

i
∑

k=1

lk i=1,2, . . . ,N −1

JN+1
∂2αN
∂t2
+
dJN+1
dt

∂αN
∂t
+GI0N

∂αN
∂x
+
1

2

d2JN+1
dt2

αN (2.5)

=−
1

2
Ω0
dJN+1
dt

for x=
N
∑

k=1

lk

The last boundary condition has now such a form that the direct effect of variable inertia
can be investigated. It should be pointed out that after transformation (2.3), the nominal to-
rque moment M0 appears only in the boundary condition for displacements in cross-sections
x = l1 + l2 + . . .+ li, i = 1,2, . . . ,N −1. However, when the two-mass system with N = 1 is
considered, no effect of M0 is observed, cf Koser and Pasin (1995, 1997).

Boundary conditions (2.5) differ from the corresponding boundary conditions in the papers
byPielorz andSkóra (2006, 2009a) by the condition in the cross-section x=0.This is connected
with the fact that the velocity of the first rigid body of the multi-mass system in those papers
was constant while in the present paper it is variable.

Upon the introduction of the following dimensionless quantities

x=
x

l0
t=
ct

l0
αi =

αi
α0

Kr =
I01ρl0
J0

Ei =
J0
Ji

Ω0 =
Ω0l0
α0c

K =
Kl0
J0c

M0 =
M0l

2
0

J0c2α0

JN+1(t)=
JN+1(t)

J0
li =
li
l0

Bi =
I0i
I01

(2.6)

the determination of displacements αi(x,t) is reduced to solving N equations

∂2αi
∂t2
−

∂2αi
∂x2
=0 i=1,2, . . . ,N (2.7)

with the following boundary conditions



110 A. Pielorz

∂2α1
∂t2
+KE1

∂α1
∂t
−KrE1

∂α1
∂x
=0 for x=0

αi(x,t)=αi+1(x,t)+Ci+1 for x=
i
∑

k=1

lk i=1,2, . . . ,N−1

−

∂2αi
∂t2
−KrBiEi+1

∂αi
∂x
+KrBi+1Ei+1

∂αi+1
∂x
=0

for x=
i
∑

k=1

lk i=1,2, . . . ,N−1

JN+1(t)
∂2αN
∂t2
+
dJN+1
dt

∂αN
∂t
+KrBN

∂αN
∂x
+
1

2

d2JN+1
dt2

αN

=−
1

2
Ω0
dJN+1
dt

for x=
N
∑

k=1

lk

(2.8)

where Ci = M0(KrBi)
−1li, α0 is a fixed angular displacement, J0 is a fixed mass moment

of inertia and l0 is a fixed length, correspondingly. Moreover, the bars denoting dimensionless
quantities are omitted for convenience.
We are interested in torsional vibrations, so for simplicity, we assume zero initial conditions

for αi, i.e.

αi(x,t)=
∂αi
∂t
(x,t) =0 for t=0 (2.9)

The solutions to equations (2.7), similarly to problemsdiscussedbyPielorz andSkóra (2006),
are sought in the form

αi(x,t)= fi

(

t−x+2
i
∑

k=1

lk−
N
∑

k=1

lk

)

+gi

(

t+x−
N
∑

k=1

lk

)

i=1,2, . . . ,N (2.10)

The functions fi and gi in (2.10) represent the waves caused by variable inertia, propagating
in the i-th shaft in the positive and negative senses of the x-axis, respectively. These functions
are continuous and equal to zero for negative arguments. In (2.10), it is taken into account that
the first disturbance appears in the i-th element in the cross-section x0i = l1+ l2+ . . .+ li at
the time instant t0i = li+1+ li+2+ . . .+ lN.
Substituting postulated solutions (2.10) into boundary conditions (2.8) and denoting the

largest argument in each boundary condition separately by z, we obtain the following set of
ordinary differential equations for the unknown functions fi and gi

f ′′1(z)+r01f
′

1(z)=−g
′′

1(z−2l1)+r02g
′

1(z−2l1)

fi(z)= fi−1(z−2li)+gi−1(z−2li)−gi(z−2li)−Ci i=2,3, . . . ,N

rN1g
′′

N(z)+rN2g
′

N(z)+rN3gN(z)=F(z)+rN4f
′′

N(z)+rN5f
′

N(z)+rN6fN(z)

g′′i (z)+ri1g
′

i(z)=−f
′′

i (z)+ri2f
′

i(z)+ri3g
′

i+1(z) i=N−1,N −2, . . . ,1

(2.11)

where

r01 =E1(Kr +K) r02 =E1(Kr−K) rN1(z)= JN+1(z)

rN2(z)=KrBN +J
′

N+1(z) rN3(z)=
1

2
J′′N+1(z) rN4(z)=−rN1(z)

rN5(z)=KrBN −J
′

N+1(z) rN6(z)=−rN3(z) ri1 =KrEi+1(Bi+Bi+1)

ri2 =KrEi+1(Bi−Bi+1) ri3 =2KrEi+1Bi+1 F(z) =−
1

2
J′N+1(z)Ω0

i=1,2, . . . ,N−1

(2.12)



Dynamic analysis of torsional discrete-continuous systems... 111

Equations (2.11) are differential equations with a retarded argument. Coefficients in (2.11)
are variable. Equations (2.11) are solved numerically by means of the Runge-Kutta method. It
should be pointed out that functional equations having shifted arguments are investigated in
the literature, see e.g. Cherepennikov (2008), Cherepennikov and Ermolaeva (2006), Hale and
Verduyn Lunen (1993).

Themass moment of inertia of the last rigid body, after expanding in the Taylor series aro-
und Ω0t, depends on Ω0t and now it can be described by an arbitrary function containing Ω0t.
This function may also have the form of the Fourier series

JN+1(t)= a0+
∞
∑

m=1

(amcosmΩ0t+ bm sinmΩ0t) (2.13)

where ai and bi are constants.

3. Numerical results

Exemplary numerical results are performed for the three-mass system. The variable moment of
inertia of the last rigid bodies is described by the formula

JN+1(Ω0t)= a0+a2cos(2Ω0t) N =2 (3.1)

by analogy to the papers byKoser andPasin (1995, 1997). Comparing relations (2.13) and (3.1),
it is seen that function (3.1) is the Fourier series with m = 2 and with constants a0 and a2
being the only nonzero constants. Such a form of the mass moment of inertia was proposed by
Koser and Pasin (1995, 1997) for the analysis of a two-mass discrete-continuous system using
an analytical approach.

Diagrams presented below concernmainly solutions in the steady state of motion. They are
plotted out after solving equations (2.11) with

N =2 Kr =0.05 B1 =B2 =1 E2 =0.8

l1 = l2 =1 a0 =1 a2 =0.05 C2 =0
(3.2)

Theparameters E1 and K, characterizing themotor, can vary.Thefirst fournatural frequencies
of the linear system with E1 =1.0 are ω1 =0.222, ω2 =0.359, ω3 =3.158, ω4 =3.183. We are
interested in vibrations caused by the variable inertia, so the constant C2 is neglected in the
numerical analysis.

The wave approach enables us to determine the required quantities in various cross-sections
of the considered discrete-continuous systems in transient and in steady states of motion.

In Figs. 2 and 3, numerical results for angular displacements in the cross-sections x = 0,
0.5, 1.0, 1.5, 2.0 are presented for E1 =1, K =0.1 with Ω0 =0.2 and Ω0 =0.5, respectively.
The diagrams in Fig. 2, show the results in the transient state of motion while the diagrams
in Figs. 3a and 3b show the angular displacements in the steady state. In the steady state, the
solution behaves as a harmonic vibration with the period equal to the period of function (3.1)
describing the variable mass moment of inertia J3, i.e., T = π/Ω0. One can notice that the
amplitude is steady for t> 8000when Ω0 =0.2 and for t> 3500when Ω0 =0.5. In the case of
Ω0 =0.2, the highest displacement amplitude occurs in the cross-section x=2and the smallest
one in x = 0.5. For Ω0 = 0.5, the displacement amplitudes increase with the increase of x.
From Figs. 2 and 3, it follows that the steady state is gained earlier for a larger value of Ω0.

The displacement amplitudes αA versus Ω0 shown in Fig. 4, plotted for E1 = 1, K =0.1,
contain four resonant regions (ω1 =0.222, ω2 =0.359, ω3 =3.158, ω4 =3.183). The resonances



112 A. Pielorz

Fig. 2. Angular displacements in the transient state with Ω0 =0.2

Fig. 3. Angular displacements in the steady state and with Ω0 =0.2 (a) and with Ω0 =0.5 (b)

correspond to Ω0 = ωi/2. In the third and fourth resonances, the displacement amplitudes
are much higher than in the first and second resonant regions. In the first resonant region, the
highest amplitude occurs in the cross-section x=2and the smallest one in x=1. In the second
resonant region the highest amplitude occurs in the cross-section x = 1 and the smallest one
in x = 0.5. From Fig. 4, it follows that for the assumed parameters (3.2), the displacement
amplitudes are small except for resonant regions. Besides, near p=0.78, the amplitude of the
displacement in the cross-section x=1.5 is equal to 0.0419 while the displacement amplitudes
in the remaining cross-sections are smaller than 0.013.

Fig. 4. Displacement amplitudes versus Ω0



Dynamic analysis of torsional discrete-continuous systems... 113

The first rigid body representing themotor in themodel shown in Fig. 1 is characterized by
the parameter E1 (in the dimensional quantities by themassmoment of inertia J1) and by the
parameter K (the slope of characteristics). The effect of these parameters on the amplitudes αA
in the cross-section x=2 for 0.003<Ω0 < 0.245 is shown in the next twofigures.Thediagrams
contain the first two resonant regions (ω1 = 0.222, ω2 = 0.359). For bigger values of Ω0, the
displacement amplitudes are small except for resonances where they become high. In these
figures, apart from numerical results obtained by using equations (2.11) with (3.2), the curves
marked by broken lines are plotted using the analytical solution for the three-mass system
derived in the paper by Pielorz and Skóra (2009a) when K→∞ and E1 =1.

The diagrams for displacement amplitudes αA with E1 = 0.2, 0.5, 1.0, 1.5 are given in
Fig. 5a. From these diagrams, it follows that the natural frequencies increase, the resonant
regions become wider and the maximal amplitudes decrease with the increase of E1. Besides,
for small values of E1, i.e., for large J1, the numerical results approach the analytical solution.

Fig. 5. (a) Effect of E1 in x=2 with K =0.1; (b) effect of K in x=2with E1 =1.0

The effect of the parameter K is discussed for K = 0.1, 0.15, 0.25, 0.35, 0.5, 1.0, 2.0. In
Fig. 5b, four resonant regions are shown: two concern the analytical solution derived in the
papers by Pielorz and Skóra (2006, 2009a) with the natural frequencies ω1 = 2Ω0 = 0.132,
ω2 = 2Ω0 = 0.332 and two regions for the three-mass system with a variable velocity of the
motor with ω1 =0.222, ω2 =0.359.

The behavior of the system is different for small and larger values of K. Namely, the para-
meter K with smaller values (up to 0.35) plays the role of a damping coefficient which decreases
the amplitudes while for K > 0.35 the maximal amplitudes increase with the increase of K
approaching the analytical solution. Besides, it is seen that the numerical solutions with smaller
values of K are in the resonant regions corresponding to the system under considerations while
numerical solutions with K > 0.35 lay in the resonant regions determined for the system with
the motor working with a constant velocity as studied by Pielorz and Skóra (2009a).

From Figs. 4 and 5, it follows that an additional resonant region may occur for Ω0 close
to 0.05. It concerns E1 = 1.0, 1.5 or K < 0.35. For Ω0 < 4.0, no other additional resonant
regions, except for those connected with the linear system, were found assuming ∆Ω0 =0.001.

It is interesting to compare the results obtained by means of the wave approach with the
results obtained using another method. It should be pointed out that the wave method enables
us to determine solutions in transient as well as in steady states of motion.

Some comparisons could bedone for the discrete-continuous torsional system shown inFig. 1
with the first rigid body representing a motor working with a constant velocity as investigated



114 A. Pielorz

by Pielorz and Skóra (2006, 2009a). Then, the boundary condition for x=0 in (2.8) is simpler
and has the form

∂α1
∂t
=0 (3.3)

This case corresponds to K →∞. For the problemwith this boundary condition, itwas possible
to derive analytical solutions. These solutions are valid in steady states of motion.

Fig. 6. Comparison of numerical (dashed line) and analytical solutions (continuous line) for
displacements of the three-mass system in x=2 in a steady state with boundary condition (3.3)

and Ω0 =0.05

Comparative results concerning angular displacements are plotted out in Fig. 6 with
Ω0 = 0.05 using appropriate analytical solutions for the three-mass system derived by Pielorz
and Skóra (2009a).

4. Final remarks

Problems of discrete-continuous systems torsionally deformed with variable inertia, taking into
account variable velocity of the rigid body representing an electrical motor, can be described by
classical wave equations with boundary conditions having variable coefficients. These equations
can be solved applying themethod using the dAlembert approch to the solution of equations of
motion which leads to equations with a retarded argument.
The presented numerical calculations show that the variablemotor velocity has a significant

influenceonthebehavior of theconsidered systems.For a largemassmomentof themotor J1 and
a large slope of its characteristics K, the numerical solutions approach the analytical solutions
derived in the paper byPielorz and Skóra (2009a) for a simpler case of boundary conditions, i.e.
for the systemwith the motor moving with a constant velocity.
From the comparison of the results given in the paper by Pielorz and Skóra (2006) and in

the present paper, it follows that the wave approach, i.e., applying the solution of dAlembert
type, is more effective in the case of the motor with a variable velocity. This is connected with
the fact that the parameter K plays partly the role of a damping coefficient.

References

1. Cherepennikov V.B., 2008, Polynomial quasisolutions of a boundary value problem for linear
differential-difference equations,Functional Differential Equations, 15, 4, 143-157

2. Cherepennikov V.B., Ermolaeva P.G., 2006, Polynomial quasisolutions of linear differential-
difference equations,Opuscula Mathematica, 26, 3, 431-443



Dynamic analysis of torsional discrete-continuous systems... 115

3. Hale J.K., Verduyn Lunen S.M., 1993, Introduction to Functional Differential Equations,
Springer-Verlag, NewYork

4. Hesterman D.C., Stone B.J., 1995, Secondary inertia effects in the torsional vibration of re-
ciprocating engines – a literature review, Proceedings of the Institution of Mechanical Engineers,
Part C: Journal of Mechanical Engineering Science, 209, 1, 11-15

5. Koser K., Pasin F., 1995, Continuous modeling of the torsional vibrations of the drive shaft of
mechanisms, Journal of Sound and Vibration, 188, 1, 17-24

6. Koser K., Pasin F., 1997, Torsional vibrations of the drive shafts of mechanisms, Journal of
Sound and Vibration, 199, 4, 559-565

7. Nadolski W., Pielorz A., 2001, Theoretical investigations of nonlinear loads on gear teeth in
single gear transmission, International Journal of Mechanical Sciences, 43, 299-311

8. Pasricha M.S., Carnegie W.D., 1979, Formulation of the equations of dynamic motion inc-
luding the effects of variable inertia on the torsional vibrations in reciprocating engines, Part I,
Journal of Sound and Vibration, 66, 181-186

9. Pielorz A., 1999, Non-linear vibrations of a discrete-continuous torsional system with non-
linearities having characteristic of a soft type, Journal of Sound and Vibration, 225, 2, 375-389

10. Pielorz A., 2003,Vibration problems in selected torsionalmechanical systems,Meccanica, 38, 6,
731-738

11. Pielorz A., 2007, Nonlinear equations with a retarded argument in discrete-continuous systems,
Mathematical Problems in Engineering, 2007, doi: 10.1155/2007/28430

12. PielorzA., 2010,Vibrations of rod-rigid element systemswith a local non-linearity, International
Journal of Non-Linear Mechanics, 45, 888-894, doi: 10.1016/j.ijnonlinmec.2009.11.016

13. Pielorz A., Skóra M., 2006,Modeling ofmultimass systems torsionally deformedwith variable
inertia,Differential Equations and Nonlinear Mechanics, 2006, ID 20758, 1-11

14. Pielorz A., Skóra M., 2009a, Analytical approach in torsional multi-mass discrete-continuous
systems with variable inertia,Meccanica, 44, 121-131

15. Pielorz A., Skóra M., 2009b, Torsional vibrations of discrete-continuous systems with lo-
cal nonlinearities having soft characteristics, Journal of Theoretical and Applied Mechanics, 47,
211-227

16. Sinha S.C., Butcher E.A., 1997, Symbolic computation of fundamental solution matrices for
linear time-periodic dynamical systems, Journal of Sound and Vibration, 206, 1, 61-85

17. TurhanO., KoserK., 2004,Parametric stability of continuous shafts, connected tomechanisms
with position-dependent inertia, Journal of Sound and Vibration, 277, 223-238

18. Zajączkowski J., 1987, Torsional vibration of shafts coupled by mechanisms, Journal of Sound
and Vibration, 116, 221-237

Manuscript received January 29, 2013; accepted for print June 17, 2013