Dynamic Analysis of Elastic

D. A. Turcic1 Mechanism Systems. Part I:

Assistant Professor,
Department of Mechanical Engineering
and Mechanics,
Applications
Drexel University,
Philadelphia, Pa., 19104 In a previously reported work, generalized equations of motion were developed for
the dynamic analysis of elastic mechanism systems. Emphasis was placed on the
need for including vibration effects; these, until recently, had generally been
Ashok Midha neglected, primarily for reason of the complexity of the mathematical analysis of
such systems. Finite element theory was used to facilitate the modeling of elastic
Associate Professor, mechanisms. In this work, Part I of two companion papers, an approach to the
School of Mechanical Engineering,
Purdue University, solution of the equations of motion is offered. Some related considerations are also
West Lafayette, Ind. 47907 discussed. Applications to the analysis and solution method are presented. To
demonstrate generality, an example is selected wherein the geometrically complex
follower link of a four-bar mechanism is modeled with quadrilateral finite
elements. Part II of this paper presents an experimental investigation of an elastic
four-bar mechanism.

Introduction
In general, machine designers have traditionally relied upon systems. The mathematical model utilizes the displacement
rigid-body dynamic analysis methods, wherein all mechanism finite elements. The reader is referred to [1] for detailed
members are treated as rigid-bodies. Subsequent to the theoretical developments. The generality of the approach is
dynamic analysis, a force and stress analysis of the rigid exemplified here by its application to a high-speed, four-bar,
members is undertaken. This approach tends to be reasonably punching machine, whose geometrically complex follower
accurate as long as the speed of machine operation remains link is modeled with linear quadrilateral finite elements.
very much lower than the natural frequencies of the
mechanism system. Solution of the Equations of Motion
As pointed out earlier [1], mechanisms are now being
required to run at higher speeds while maintaining greater The equations of motion for the entire mechanism system
positioning accuracy, and exerting a greater force per unit have been expressed in [1] as follows:
mass of the mechanism. A vibration analysis is necessitated to [M]{u}+[c]{u} + [K]{u} = {Qex}
predict the mechanism response and its stability during
operation. Unstable parametric resonant conditions would -[M\{U0}-2[{Md\ + [MyA]]{u]
not only result in large elastic deformations, but would also -[[A«cfl+2[Mtfvel] + [Macc]](«) (1)
shorten the life of the mechanism due to early fatigue failure.
The latter results from coupling of high stress levels and their where [M\, [C], and [K\ represent the system mass, viscous
cyclic occurrence. damping and stiffness matrices, respectively; {u}, [ii], («),
It is, therefore, not surprising that much research has been and [ U0) contain the displacement, velocity, acceleration,
reported in recent years addressing the above issues. A and rigid-body acceleration, respectively, of the degrees-of-
representative set of these works are reported in [1]. By and freedom describing the mechanism system. The generalized
large these works have been divided into two categories, the forces are contained in [Qex}. The matrices [Md], [Mvel],
first being works of those researchers who treated the elastic [Mdd], [MdwA], and [Macc] involve mass distribution and
links of four-bar or slider-crank mechanisms as continuous rigid-body orientation, velocity and acceleration. For details
systems. A second group treated the elastic links as discrete of derivation, the reader is referred to [ 1 ].
systems, wherein a finite number of degrees-of-freedom are The coefficient matrices of equation (1) are periodic
utilized to describe the motion of the discretized system. functions of the mechanism position. This dependence on
It was in this latter vein that [1] presented a derivation of position comes about because of the transformation matrix
the generalized equations of motion for elastic mechanism [TR], described in [1], which is necessary to assure com-
patibility of the displacement, velocity, and acceleration of
the degrees-of-freedom common to two or more links. For a
Formerly at Department of Mechanical Engineering, The Pennsylvania mechanism operating at a constant speed, the matrices will be
State University, University Park, Pa., 16802
Contributed by the Dynamic Systems and Control Division for publication in
position- as well as time-periodic. The approach utilized for
the JOURNAL OF DYNAMIC SYSTEMS, MEASUREMENT AND CONTROL. Manuscript solution of the equations of motion (1) is one developed by
received by the Dynamic Systems and Control Division, May 17,1983. Midha [2]. This method is capable of calculating the transient

Journal of Dynamic Systems, Measurement, and Control DECEMBER 1984, Vol. 106/249

Copyright © 1984 by ASME

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 Table 1 Mechanism parameters for the four-bar linkage with a
complex follower link
Parameter Crank Coupler Follower

Area .835 in ,2
.835 in 2 .2 in (thickness)
(5.39 a / ) 15.39 cnT) (.508 cm)

Length 2.5 in 7.0 in 6.7 in

(6.35 cm) (17.78 cm) (17.02 cm)

Bending moment
of inertia 1.94x10"?, in? 1.94X10"3, in?
(8.1x10 cm ) (8.1x10"^ cm )

Distance between ground pivots - 6.4 in (16.25 cm)

Modulus of elasticity, E = 10.3x106 psi (7.10xl07 kPa)

Weight density = .098 lb/in3 (2.66X10-2 N/cm3)

300

-300

-600

id - 9 0 0
Fig. 1 Four-bar mechanism with a geometrically complex follower link

as well as the steady-state response of a linear second-order -1200,

72 144 216 288 360
differential equation with time-periodic coefficients, without
costly numerical integration of many cycles. The numerical Crank Angle (Degrees)
algorithm is based upon dividing the fundamental time period Fig. 2 Punching force acting on the follower link versus crank angle
into an appropriate number of sub-intervals; during each sub-
interval the time-dependent coefficients of the differential rigid-body and elastic motion coupling terms, shown in the
equations are assumed to be invariant. Constraint equations, equations of motion (1).
which require both displacement and velocity continuity, are The equations of motion are expressed in a recursive form
applied at the boundaries of each discrete time subinterval. as:
The resulting algebraic set of constraint equations is then
solved, yielding either the steady-state or transient system [M]lu)ii + [Qiu)li + lKHu},i = {QeX\-[M][U0}
response. With additional improvements discussed in [3], this
method provides an efficient means for computing the steady- - 2[[Md] + [Mvel]] (it) ,,„ „ - [[Mdd]
state solution to the equations of motion. The equations of + 2[Mc?vel] + [Macc]] [«)<„_„ (3)
motion (1), however, must be modified slightly in order to where subscript ii denotes the iteration number. The iteration
implement the steady-state solution technique.
proceeds as follows. In the first iteration, ii = 1 and (H)(,,_I)
Equation (1) represents a coupled set of differential = w ii) o = 0 , and the following equation
equations. The equations must be first decoupled. This = " (/;-1)
is solved for (u} x and {«), in each time subinterval.
procedure is complicated by the presence of the nonsymmetric
[Md], [Mdd], [Mvel], [Mdvel], and [Macc] matrices in the M [ M } , + [ q { « ) , + [ A ] l H ) , = {Q w }-[Afl(tt 0 } (4)
equations. These matrices represent the coupling terms In the second iteration, the values of [u)i and («), are
between the rigid-body velocities and accelerations and the substituted into the right-hand side of equation (3) and the
elastic displacements {u} and velocities (u}. When con- improved solution ([u}2, [u}2) is evaluated. This procedure
sidering small displacements, these terms are small when is repeated until the changes in the {u),, and (it},, vectors, for
compared with the other terms in the equation and have each point in the time period, meet some predetermined small
generally been neglected by most investigators. However, value.
these terms may become significant with high operating
speeds of the mechanism. The recursive scheme proved to be efficient and converged
in very few (less than 5) iterations for the cases examined. It
For convenience at this point in the development of the should be noted that each of these cases produced only small
solution method, these rigid-body and elastic motion coupling changes from the original or base solution. Inaccuracy
terms will temporarily be neglected. Thus, equation (1) problems could occur if the rigid-body and elastic motion
becomes, coupling terms were to cause large variations from the base
[M][ti)+[c]lu}+[K]lu] = lQex}-[Mi{U0} (2) solution. Presently, in the iterative procedure, the eigenvalues
and eigenvectors of equation (3) are assumed to remain
The foregoing equation can now be decoupled by using unaffected by the rigid-body and elastic motion coupling
modal analysis techniques [2, 4-6]. Midha's steady-state terms. Therefore, if the coupling terms significantly change
solution algorithm [2] can then be applied to the decoupled the eigenvalues and eigenvectors of the original system, a
equations of motion yielding the steady-state solution of these variation of the above iteration procedure would be necessary
equations directly. which takes the large changes into account. At this point, the
recursive scheme is used only as a means for determining
An Iterative Method for Inclusion of Coupling Terms whether the coupling terms are significant. For the cases
An iterative method is presented here for including the examined thus far the coupling terms proved to be negligible.

250/Vol. 106, DECEMBER 1984 Transactions of the ASME

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 I900r
Punching
Force
1640

1380

1120

Element
Number 860
Eight
600, 72 144 216 288 360
Crank Angle (Degrees)
Fig. 4 First mode natural frequency versus crank angle

Fig. 3 Follower link illustrating the position of the punching force and An
element number eight 9le
Quasi-Static Analysis
Research is continuing into the stability and accuracy
properties of the iterative method for larger excursions of the
final solution from the base solution.
A
Geometrically Complex Four-Bar Mechanism
Figure 1 shows the four-bar mechanism with a
geometrically complex follower link, modeled with linear f) n
quadrilateral finite elements [5-7]. The finite element model L ocafi°
L
contains 241 elastic degrees-of-freedom. This particular onthe
mechanism constitutes part of a high-speed punching Fo\ lower

machine. Table 1 contains the various parameters associated

with the mechanism. Two separate approaches are taken to
determine stresses in the punching mechanism, as it operates
at 5000 rpm. The first is a quasi-static analysis, that assumes
the velocity (it ] and acceleration (it} corresponding to the Vibration Analysis
generalized coordinates [1] to be negligible. The quasi-static
Fig. 5 Maximum principal stress as a function of location on the
method is traditionally the method of design for industrial follower and the crank angle
mechanisms. The second method is a vibration analysis based
on the methods developed in [1], The stresses determined by
the two methods are then compared. principal stresses versus the crank angle of the mechanism, for
Figure 2 shows the punching force (force of resistance to element number eight only (see Fig. 3), and for the entire
shearing), modeled as a one-half sine wave that acts on the follower link, respectively. The discontinuities in the plots
follower as shown in Fig. 3. The variation of the first mode occur whenever the maximum principal stress shifts in
natural frequency of the system with position of the direction, or when the element, in which the maximum stress
mechanism is shown in Fig. 4. As seen from the figure, the occurs, changes. The figures show much higher maximum
natural frequency variation for this system is quite large (300 principal stresses for the vibratory response.
percent). Figure 5 shows a comparison of the maximum These results indicate that a realistic mechanism of this type
principal stress obtained using the two methods of analysis. could be undersigned if the vibratory response were not taken
The x-axis represents the location of the point on the follower into account.
with the origin corresponding to the top of the follower (see
Fig. 3). Thej-axis represents the maximum principal stress at Analytical Model
a point and the z-axis the crank angle of the mechanism for Part II of this work will present an extensive experimental
one cycle. A qualitative comparison of the two plots shows study of an elastic four-bar mechanism, for verification of the
much higher principal stresses corresponding to the vibratory analytically determined results. The analytical model used in
response than to the quasi-static one. Figures 6 and 7 show a the verification is developed below. Figure 8 illustrates the
comparison of the quasi-static and vibratory maximum basic configuration of the four-bar crank-rocker mechanism.

Journal of Dynamic Systems, Measurement, and Control DECEMBER 1984, Vol. 106/251

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 The parameters defining the mechanism are listed in Table 2. where 6 is the angular orientation of the link. From [1], the
For the link dimensions chosen (Table 2), each of the [mevd] and [meacc] matrices are shown to be as follows:
members may be treated as a slender beam and modeled with
beam finite elements. The mass and stiffness matrices for an [«evei] = \p[N]T[Tm]T[tm][N\dV
(10)
Euler beam element can be shown [5-8] to be [mU = MN]T[Tm]T[fm][N]dV

"1/3 0 0 1/6 0 0

0 13/35 11L/210 0 9/70 -13L/420

0 11ZV210 LV105 0 13L/420 -L2/140

[me] = pAL (5)
1/6 0 0 1/3 0 0

0 9/70 13L/420 0 13/35 -11L/210

0 - 13L/420 -LV140 0 -11L/210 L 2 /105

' EAX/L 0 0 -EAX/L 0 0

0 12E//L 3 6EI/L 2 0 - 12E//Z,3 6E//L 2

0 6EI/L 2 4EI/L 0 - 6E//L 2 -2EI/L

(6)
w\= -EAX/L 0 0 EAX/L 0 0

0 -12E//Z, 3 -6EI/L2 0 12E//Z,3 -6E//L2

0 6EI/L2 2EI/L 0 -6E//L2 4E//L

where The shape function matrix [N] for a beam element is described
p = mass density in [7]. Substituting for transformation matrix [Tm] and its
A = cross-sectional area derivatives, from equations (7)-(9), and performing the in-
L = element length dicated integrations yields the following expressions for [meyA]
E = Young's modulus and [/<„.]:
/ = area moment of inertia about the bending axis. 80000
mum Prin ;ipal Stress (lb/in )

For two-dimensional problems, the transformation matrix

[T,„] in [1] is defined as
60000 •Vibratory Response

cos0 — sinfl
[TJ = (7)
sini9 cosf? 40000

and its derivatives as 20000

" - sint9 -COS0 "
[Tm] = e (8)
cost9 -sin0

" - sin0 - cosd "cos0 --sin0 -20000^ 72 144 216 288 360
[Tm] = e l (9)
cosS - sinfl -e sint9 cos©
Crank Angle (Degrees)
Fig. 7 Follower maximum principal stress for the entire follower
versus crank angle
20000

Vibrolory Response
12000

4000
Fol lower
Q.
E -4000
E

-12000

IE - 2 0 0 0 0 ,
72 144 216 288 360
Crank Angle (Degrees)
Fig. 6 Follower maximum principal stress for element number eight
versus crank angle Fig. 8 Schematic drawing of the four-bar mechanism

252/Vol. 106, DECEMBER 1984 Transactions of the ASME

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Table 2 Four-bar mechanism parameters 350
Parameter Crank Coupler Follower
Area .167 i n 2 , .063 i n 2 , .063 i n 2 ,
(1.07 cm ) (.406 c m ) (.406 cm ) 330

Length 4.25 i n 11.0 i n 10.65 i n

(10.795 cm) (27.94 cm) (27.051 cm)
Bending moment 3.88xl0~4,in4. 2.08xl0~5.in4. 2.08xl0~5„in4.
of i n e r t i a (1.62x10 cm ) (8.67x10 cm ) (8.67x10 cm )

D i s t a n c e between ground p i v o t s = 1 0 . 0 i n (25.4 cm) 290

Lumped mass of t h e b e a r i n g assembly = .0976 l b (.443 N)
at the crank-coupler connection
Lumped mass of t h e b e a r i n g assembly = .08112 l b (.368 N) 270
at t h e coupler-follower connection
Modulus of e l a s t i c i t y , E - 1 0 . 3 x l 0 6 p s i (7.10x10 kPa)
250,
Weight d e n s i t y = .098 l b / i n 3 ( 2 . 6 6 x l 0 ~ 2 N/crn3) 360

Crank Angle (Degrees)

Fig. 10 First mode natural frequency versus crank angle

400

380

360

^ 340

320-

300,
0 72 I44 2I6 288 360
Crank Angle (Degrees)
Fig. 11 Second mode natural frequency versus crank angle

Fig. 9 Elastic mechanism model three

800

[»J,ei] = %m%\ (11) 700

2 e
[mlc \ = 6[m<
]-6 [m ] (12)
where 600

o -r
[mec]=\P[N]T [N]dV 500
1 0
or, 400
0 -42 -6L 0 18 4L -Models Two, Three, and Four

42 0 0 18 0 0 300,
0 72 I44 2I6 288 360

pAL 6L 0 0 4L 0 0 Crank Angle (Degrees)

[m'c (13) Fig. 12 Third mode natural frequency versus crank angle
120 0 -18 -4L 0 42 6L
18 0 0 42 0 0 elements, and twenty-seven degrees-of-freedom. Figures 10,
11, and 12 show the variations in the first, second and third
-4L 0 0 --6L 0 0 mode natural frequencies with crank angle for the nine
element model (model three) of the four-bar mechanism as
for this two-dimensional beam element. The above matrix is well as a three (model one), six (model two) and fifteen
antisymmetric. (model four) element model of the mechanism. As is
The preceding equations give expressions for each of the manifested in the figures, the natural frequencies converge to
matrices that appear in the differential equations of motion a common functional variation, for each of the modes, as the
(equation (21) of [1]) for a single finite element. The link and complexity of the model increases. A similar convergence in
the mechanism system matrices are subsequently assembled the steady-state, coupler and follower midpoint strains has
from the element matrices. Described in a previous section are also been found as the complexity of the finite element model
techniques for solving the equations of motion of the is increased. These results indicate that a nine-element model
mechanism system. of the four-bar mechanism, in the absence of any applied
Figure 9 illustrates the finite element model of the elastic forcing, is sufficiently accurate.
four-bar mechanism. The model is comprised of nine beam When deriving the stiffness properties of a beam finite

Journal of Dynamic Systems, Measurement, and Control DECEMBER 1984, Vol. 106/253

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 element, small strain theory assumes that the transverse directly without costly numerical integration over many
displacements are independent of the axial displacements or cycles. This technique, with additional improvements
forces. In reality, however, a compressive axial force would discussed in [3] has been found to be a very efficient method
tend to increase any transverse displacement of the beam, thus for determining the steady-state solution of the differential
effectively decreasing the transverse stiffness of the beam. A equations of motion (of the Hill type) possessing time-
tensile axial force would have the opposite effect. This effect, periodic coefficients. An iterative approach for including the
called geometric stiffening or stress stiffening, could become rigid-body and elastic coupling terms in the equations of
important in mechanism analysis where large axial forces are motion has also been discussed.
known to occur. One approximate method of including this To demonstrate the generality of the analysis methods an
effect is to calculate a geometric stiffness matrix [kea] based example has been presented wherein the geometrically
on large strain theory that would represent the coupling complex follower link of a four-bar crank-rocker mechanism
between the axial and transverse displacements. The has been modeled with linear quadrilateral elements. An
"geometric stiffness matrix" for a beam element is developed inherent characteristic of such mechanisms is verified in that
in [8] and is shown below: they manifest very large variations in their natural (modal)
frequencies with position. This example illustrates the im-
0 0 0 0 0 0 portance of including vibration effects in the analysis of high-
speed mechanisms. It has been shown that it is possible to
0 6/5 L/10 0 -6/5 L/10 obtain much lower stresses and strains when using the quasi-
static method of analysis. The traditional quasi-static ap-
F 0 L/10 2LV15 0 -L/10 -L 2 /30 proach, therefore, carries with it a risk of underdesigning, or
Wc incorrectly designing, high-speed elastic mechanisms.
L 0 0 0 0 0 0
Also presented herein has been the development of the
0 -6/5 -L/10 0 6/5 -L/10 analytical model that will be used in Part II of this paper. It
will serve to verify the analytical modeling and solution
0 L/10 2
-L /30 0 -L/10 2L2/15 techniques, discussed in an earlier and the present works, via
comparison of analytical and experimental strains in an
(14) elastic four-bar mechanism.
where F is equal to the axial force in the element. The matrix Acknowledgments
[keG] represents the change in transverse stiffness due to an
axial force in the element. To include this coupling effect in The authors would like to acknowledge the support of
the equations of motion, the geometric stiffness matrix is National Science Foundation through Grant No. MEA-
simply added to the element stiffness matrix [ke]. The axial 8104703, and AMP Incorporated of Harrisburg, Penn-
force (F) in the member is approximated by a quasi-static sylvania. The Engineering Computer Laboratory is gratefully
force analysis, wherein the displacements and internal forces acknowledged for the use of its facilities.
are calculated due to inertial and external forces on the Reference
mechanism system. This proved to be a reasonable assump-
tion and as will be verified in Part II, the geometric stiffening 1 Turcic, D. A., andMidha, A., "Generalized Equations of Motion for the
effect does contribute significantly to the solution. Dynamic Analysis of Elastic Mechanism Systems," ASME JOURNAL OF
DYNAMIC SYSTEMS, MEASUREMENTS, AND CONTROL, Vol. 106, No. 4, Dec. 1984,
pp.243-248.
2 Midha, A., Erdman, A. G., and Frohrib, D. A., "A Closed-Form
Observation and Conclusions Numerical Algorithm for the Periodic Response of High-Speed Elastic
Linkages," ASME Journal of Mechanical Design, Vol. 101, No. 1, Jan. 1979,
The generalized equations of motion for an elastic pp. 154-162.
3 Turcic, D. A., "A General Approach to the Dynamic Analysis of Elastic
mechanism system were derived in an earlier work using Mechanism Systems," Doctoral Dissertation, The Pennsylvania State
displacement finite element theory. The equations were first University, November 1982.
developed for a single finite element, then for a single link, 4 Midha, A. Erdman, A. G., and Frohrib, D. A., "Finite Element Ap-
and finally for the entire mechanism system. The equations proach to Mathematical Modeling of High-Speed Elastic Linkages,"
were general in that a three-dimensional elastic mechanism Mechanism and Machine Theory, Vol. 13, 1978, pp. 603-618.
5 Zienkiewicz, O. C , The Finite Element Method in Engineering Science,
could be modeled, allowing the use of any finite element type McGraw-Hill, 1971.
for its description. 6 Bathe, K. J. and Wilson, E. L., Numerical Methods in Finite Element
For the work presented here, the primary method of Analysis, Prentice-Hall, 1976.
7 Cook, R. D., Concepts and Applications of Finite Element Analysis,
solution of the equations of motion has been a steady-state Wiley, 1974.
solution method initially developed by Midha [2], This 8 Przemieniecki, J. S., Theory of Matrix Structural Analysis, McGraw-
method allows the steady-state solution to be calculated Hill, 1968.

254/Vol. 106, DECEMBER 1984 Transactions of the ASME

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 02/01/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use