Dynamics of in-line six-cylinder diesel engine

with rubber damper

Vclav Ptk, Pavel Novotn
Brno University of Technology, Institute of Transport Engineering,
Brno, Czech Republic

Summary
The paper deals with issues of the structural dynamics of the in-line six-cylinder tractor diesel engine.
The main components of the driving mechanism have been modelled as flexible bodies using
the finite element method (ANSYS) and then converted into a multi-body system (ADAMS/Engine).
The sliding bearings of the cranktrain have been modelled using a database of reaction forces
in the hydrodynamic oil film. Loading forces have been applied based on the values of combustion
pressures measured in an engine cylinder. A rubber damper has been applied to reduce the engine
vibrations. Experimental verification of the computational model results has been performed
by measuring the values on a testing engine unit using non-contact laser measuring technique.

Keywords
FEM, multibody system, diesel engine, crankshaft, rubber damper

23rd CADFEM Users Meeting 2005

International Congress on FEM Technology
with ANSYS CFX & ICEM CFD Conference
November 9 11, 2005, International Congress Center Bundeshaus Bonn, Germany

1.

Introduction

The internal combustion engine development process requires CAE methods. CAE methods are most
efficient at the beginning of the engine development process when different concepts have to be
compared or no hardware is available yet. This is why they are mainly used early in an idealized
development process. CAE results determine the two primary ones: the concept check (identifying
the best approaches) and the virtual check (transferring the chosen concept to series development).
Another requirement concerns how detailed the computer models are to be. Practice has shown that
the amount of usable information gathered from the computer simulation increases less steeply than
proportionally with the complexity of the model. In a project environment, usable information
froma certain model complexity advances because the results become available too late some
of them later than hardware measurement results or because lengthy response times isolate
the CAE engineers from a project dialogue with the CAD area. Expressed in simpler terms,
the designers cannot wait for the computing side before making their decisions, regardless
of the quality of the simulations.
The next demand for CAE usage is that program platforms should, if possible, not be changed during
the development process so that CAE contributions from different sources (engine developers
including pre-series and series development, suppliers, engineering partners and software firms) can
be combined with little effort.

2.

Simulation methods development

The internal combustion engines are thermo-mechanical systems and this demands a special solution
approach. The cranktrain dynamic model is the main part of a virtual engine. A major problem
is the interaction between a crankshaft and an engine block.
High requirements for modern engines force the development of cranktrain models of a higher level.
The complexity of computational models increases with computer technology development. Complex
cranktrain computation models enable analysis of dynamic behavior, understanding of physical
problems and will be powerful tools for optimal solutions whereas the development time
of the machinery will be reduced. These models incorporate CAD/FEM/MBS platforms and enable full
dynamic analysis, which includes torsional, bending and axial vibrations, stress-strain analysis or
fatigue analysis etc. Fig. 2 presents the cranktrain modeling approach based on CAD/FEM/MBS tools.

Fig. 1: Complex cranktrain dynamic modeling approach

23rd CADFEM Users Meeting 2005
International Congress on FEM Technology
with ANSYS CFX & ICEM CFD Conference
November 9 11, 2005, International Congress Center Bundeshaus Bonn, Germany

The engine development process starts with a kinematic model made up of rigid bodies, specified
with weight and inertia, approximations of which can be derived from the first design sketches
for a new engine. The model is excited with its inertia forces under uniform rotation and
with a provisional cylinder pressure signal. This simulation supplies important specification criteria
for the concept phase (approximate bearing forces, free engine forces and torques, loads on
connecting rods and individual cranks).
The crankshaft, flywheel and torsional vibration damper in the dynamic cranktrain model are
subsequently replaced with elastic bodies represented in the FEM program by the way of their static
deformation and vibration modes and added to the MBS program in this form. The cylinder pressure
load can already be defined more precisely now, for example with indicated pressures of the planned
combustion process from a single-cylinder engine. The benefit from such a method depends
on the specific problem (e.g. strength and acoustics). The same applies to the variety of model stages
for the oil film in the crankshaft bearings. Examinations on a freed individual bearing allow a different
choice of detail than on a complete-engine dynamic model [1].

3.

Cranktrain dynamics solution

3.1

FE Models of main parts

In the case of dynamic structural calculations, the Finite Element Method (FEM) is state of the art.
The calculation of the structural transfer behavior of single components is efficiently possible
in the frequency domain, giving consideration to a large number of degrees of freedom. ANSYS
software is used for FE model creations and calculations. First of all CAD models of diesel in-line
6-cylinder engine is created. CAD models give basic geometric information about engine parts and are
used for a creation of FE models (e.g. crankshafts or engine blocks) or MBS rigid parts (e.g. pistons
or piston pins).
A material model of all flexible parts is linear and isotropic. Youngs modulus, Poissons ratio and
density describe each flexible part.
Fig. 3 shows FE models of main parts created on the basis of CAD models. FE model
of the crankshaft includes all geometric details (radius, notches etc.) and can be used for fatigue
calculations.

Fig. 3: Diesel in-line 6-cylinder engine CAD and FE models

3.2

Summary of flexible body reduction

Modal analyses can be performed in ANSYS and are not time-consuming, but for solution in time
domain, these models are very large and require reduction. The discretization of a flexible component
23rd CADFEM Users Meeting 2005
International Congress on FEM Technology
with ANSYS CFX & ICEM CFD Conference
November 9 11, 2005, International Congress Center Bundeshaus Bonn, Germany

into a finite element model represents the infinite number of DOF with a finite, but very large number
of finite DOF. The linear deformations of the nodes of this finite element mode, u, can be
approximated as a linear combination of a smaller number of shape vectors (see equation 3.1).
M

u = i qi

(3.1)

i =1

where

number of mode shapes,

mode shape,

qi

modal co-ordinates.

For reduction of FE models, the Craig-Bampton method is used. The Craig-Bampton method
separates the system DOF into boundary DOF, uB , and interior DOF, uI. Two sets of mode shapes are
defined as follows:
Constrain modes: these modes are static shapes obtained by giving each boundary DOF a unit
displacement while holding all other boundary DOF fixed.
Fixed-boundary normal modes: these modes are obtained by fixing the boundary DOF and computing
an eigenvalue solution.
The relationship between the physical DOF and Craig-Bampton modes and their modal co-ordinates is
illustrated by the following equation

I
u
u = B=
u I IC

0 qC

IN q N

(3.2)

where
uB
column vector of boundary DOF,
uI
column vector of interior DOF,
I,0
are identity and zero matrices,
IC
physical displacements of interior DOF in the constrain modes,

IN

physical displacements of interior DOF in normal modes,

qC
qN

column vector of the modal co-ordinates of the constrain modes,

column vector of the modal co-ordinates of the fixed-boundary normal modes.

The generalized stiffness and mass matrices corresponding to the Craig-Bampton modal basis are
obtained via a modal transformation. The stiffness transformation is

K = K =
T

IC IN

K BB

K BI I

K IB

K II IC

IN

K CC
0

K NN

(3.3)

(3.4)

and mass transformation is

I
M = T M =

IC IN

M BB

M BI I

M IB

M II IC

IN

M CC
M NC

M CN
M NN

where the subscripts I, B, N and C denote interior DOF, boundary DOF, normal mode and constrain
mode. Equations (3.3) and (3.4) have a few properties:

submatrices M
NM a K NM are diagonal because they are associated with eigenvectors

matrix K is block diagonal. There is no stiffness coupling between the constrain modes and
fixed-boundary normal modes

is not block diagonal because there is an inertia coupling between the constrain
matrix M
modes and fixed-boundary normal modes.
The Craig-Bampton method is a powerful method for tailoring the modal basis to capture both
the desired attachment effects and the desired level of dynamic content. However, the raw CraigBampton modal basis has certain deficiencies that make it unsuitable for use in a dynamic system
simulation. These are:
Embedded in the Craig-Bampton constrain modes are 6 rigid body DOF which must be
eliminated before the ADAMS analysis because ADAMS provides its own large-motion rigid
body DOF.

23rd CADFEM Users Meeting 2005

International Congress on FEM Technology
with ANSYS CFX & ICEM CFD Conference
November 9 11, 2005, International Congress Center Bundeshaus Bonn, Germany

The Craig-Bampton constrain modes are the result of a static condensation. These modes do
not advertise the dynamic frequency content that they must contribute to the flexible body.
The Craig-Bampton constrain modes cannot be disabled because it would not be possible
to apply constrains to the system.
The problems mentioned above can be easily resolved by applying a simple mathematical operation
to the Craig-Bampton modes. The Craig-Bampton constrain modes are not an orthogonal set
of modes, as evidenced by the fact that their generalized mass and stiffness matrices M and K ,
encountered in equations (3.3) and (3.4), are not diagonal.
By solving an eigenvalue problem

K q = M q ,

(3.5)
we obtain eigenvectors that we arrange in a transformation matrix N which transforms the CraigBampton modal basis to an equivalent, orthogonal basis with modal coordinates q*.

Nq* = q

(3.6)

The effect on the superposition formula (3.1) is

M

i =1

i =1

i =1

u = i qi = i Nq* = i * qi *
where

(3.7)

orthogonalized Craig-Bampton modes.

The orthogonalized Craig-Bampton modes are not eigenvectors of the original system. They are
eigenvectors of the Craig-Bampton representation of the system and as such they have natural
frequencies associated with them.
3.3

Plain journal bearing

A slide bearing is described as a sleeve around a pin with a lubricating fluid. The lubricant is supplied
with a suitable slot. Tangential and radial motion in combination with a wedged gap generate
a pressure in oil film in the slide bearing. Bearing loading is periodical and a pin center comes through
a bearing trajectory.
Slide bearings are used in the case of a 6-cylinder in-line diesel engine. Plain journal bearings have
great effect on dynamic behavior of the crankshaft and are a very important part of the virtual engine.
3.3.1 Theoretical background
Assuming an incompressible fluid (Newtons fluid) with a constant viscosity in a laminar lubricating
flow, an equation describing the force equilibrium in a differential fluid element can be established.
This equation is also known as the Navier-Stokes equation and can be written as

w(w) = p + 2 w + g

(3.8)

with w(w) denoting specific inertia forces, p specific pressure forces, w specific viscosity
forces, g specific mass forces and w the velocity vector of the fluid element.
Making several reasonable assumptions and applying appropriate transformations and boundary
conditions, equation (3.8) can be rewritten to the form of:
2

2 D 2
+ a ( , Z ) * = b( , Z ) ,
(3.9)
+ *
2 B Z 2
which is also known as Reynolds differential equation. denotes a transformed pressure distribution,
2

and Z are the coordinates in circumference and width direction, D/B a form factor and a and b are
position-dependent coefficients. Equation (3.9), which is the form of a partial elliptic differential
equation, cannot be solved analytically. However, it can solve equation (3.9) numerically, using an
iterative algorithm as it is done in the EHD subroutine.

23rd CADFEM Users Meeting 2005

International Congress on FEM Technology
with ANSYS CFX & ICEM CFD Conference
November 9 11, 2005, International Congress Center Bundeshaus Bonn, Germany

The implementation of this oil film model into ADAMS is done by a GFORCE User- Written-Subroutine
(GFOSUB). The parameters that are needed by the subroutine are rotational speed of the crank shaft,
position and velocity (vector) of journal against the bearing shell at each bearing side
to consider an inclined position of the crank shaft inside the bearing shell. To keep track of the actual
bearing state, two initially coincident markers (shell and journal) are placed at each bearing side [5].
3.3.2 Elasto-Hydrodynamic Oil Film Model in ADAMS
The numerical solution of the equation of the motion in time domain with integration time step
corresponding to crank angular displacement 0.01 requires effective calculation of the hydrodynamic
reaction of a radial slide bearing. The procedure derives from the Reynolds hydrodynamic equation.
The solution of hydrodynamic radial and axial slide bearings is implemented in ADAMS, where loads
are obtained from a precalculated oil film reaction force database.
When hydrodynamic bearings are modeled, it can choose to include or exclude misalignment effects
by using the following methods:
For the two-dimensional method neglecting misalignment, an empirical analytical equation is used.
This approach, which is similar to the impedance method, is the most efficient way to model
hydrodynamic bearings.
For the three-dimensional method accounting for misalignment, the Reynolds equation must be solved
explicitly. To keep the simulation effort in a reasonable range, you must decouple
the hydrodynamic solution from the dynamic solution of ADAMS/Solver. Therefore, the Reynolds
equation is solved for the slide bearing and the results are stored in hydrodynamic databases
representing the dimensionless bearing reactions (forces and force-attachment coordinates)
to dimensionless states (eccentricity and misalignment value). During the dynamic solution,
ADAMS/Solver subroutines do the database access and the necessary analytical steps (coordinate
transformations, and so on).
3.4

Rubber damper design

A torsional damper is an essential part of most engines. Truck or tractor engines cannot generally
work without the torsional damper because torsional vibrations in resonancies are too dangerous
for engine running. Two types of torsional dampers are applied most frequently: a rubber and a
viscous-type vibration damper. The rubber torsional damper is used for the analyzed in-line sixcylinder tractor diesel engine.
The rubber damper consists of a rubber band interposed between an outer inertia mass and a flange
which is attached to the front end of the crankshaft. The rubber is vulcanized to both members. Fig. 5
presents a rubber damper with 90 degree cut for better presentation.

Fig. 5: Designed rubber damper

Rubber stiffness decreases with high deformation while its damping ability remains constant or even
decreases. Likewise, heat reduces both the stiffness and the damping ability. Rubber has the ability
to store enormous amounts of energy and can release most of this energy on retraction.

23rd CADFEM Users Meeting 2005

International Congress on FEM Technology
with ANSYS CFX & ICEM CFD Conference
November 9 11, 2005, International Congress Center Bundeshaus Bonn, Germany

3.

Dynamic analysis results

Dynamic analysis of the cranktrain gives fundamental data for the cranktrain development process.
Accuracy of computational results is compared with experimental diesel in-line 6-cylinder engine
measurements.
The opening analysis of cranktrain dynamics is a modal analysis. Natural frequencies and mode
shapes of the cranktrain are found. A first torsional natural frequency has the greatest effect on
torsional forced damped vibrations and in the case of analyzed engine occurs at 240 Hz. A second
torsional natural frequency occurs at 640 Hz, has only a smaller effect and is included in the solution
of a forced damped solution. Critical engine speeds can be determined from the first torsional natural
frequency and dominant harmonic orders of combustion pressure.
Forced damped vibrations of the crankshaft are computed for the full range of engine speeds. The
torsional vibrations of the pulley of the diesel in-line engine without a torsional vibration damper are
presented in Fig. 6. Orders 6th, 7.5th, 9th and 12th are most significant in the case of the 6-cylinder
engine without a damper and these are compared with measurement.

Fig. 6: Computed and measured pulley torsional vibrations single order amplitudes without damper
A 6-cylinder crankshaft stress level roughly answers to the torsional vibrations of the pulley and the
highest values of the stress can be expected in critical engine speeds. The front end of the crankshaft
includes gear assembly that drives the valvetrain. High torsional vibrations cause timing problems
of the valvetrain. Generally, a decrease of vibrations can be achieved by applying torsional dampers.
Axial vibrations can be solved by using a cranktrain complex model as well. Fig. 7 shows axial
vibrations of the crankshaft pulley at engine speed 1200 rpm and 1600 rpm.

Fig. 7: Pulley axial vibrations without damper computation and measurement

23rd CADFEM Users Meeting 2005
International Congress on FEM Technology
with ANSYS CFX & ICEM CFD Conference
November 9 11, 2005, International Congress Center Bundeshaus Bonn, Germany

The rubber damper may be considered to be similar to attaching an additional mass to the crankshaft
through a rubber band, the latter being equivalent to a flexible shaft of similar stiffness as the rubber.
The dual stiffness of the crankshaft system produces a second frequency of vibration which results
in a frequency below and above the original frequency when no rubber damper is attached.
Fig. 8 presents computed summs of order amplitudes of pulley torsional vibrations without a damper
and with the rubber damper.
1,60
1,40

Angle []

1,20
without damper
rubber damper

1,00
0,80
0,60
0,40
0,20
0,00
1000

1200

1400

1600

1800

2000

2200

2400

2600

Engine speed [rpm]

Fig. 8: Computed pulley torsional vibrations maximum angular displacement without damper
and with rubber damper
At a critical speed the energy of exciting oscillations within the crankshaft is spent in deforming
the rubber and this unwanted energy is damped by transforming it to the rubber in the form of strain
energy, elastic hysteresis losses and heat rejection to the surroundings. Fig. 9 shows a dissipation
power of the rubber damper for engine speed range.

Dissipation power [W]

300
250
200
150
100
50
0
1000

1200

1400

1600

1800

2000

2200

2400

2600

Engine speed [rpm]

Fig. 9: Dissipation power of rubber damper

It therefore follows that by choosing an appropriate damper mass, stiffness and damping properties,
the crankshaft amplitude of vibration can be restricted to safe working limits.

23rd CADFEM Users Meeting 2005

International Congress on FEM Technology
with ANSYS CFX & ICEM CFD Conference
November 9 11, 2005, International Congress Center Bundeshaus Bonn, Germany

4.

Fatigue analysis of crankshaft

This part of the article deals with fatigue calculations of a dynamically loaded crankshaft of a in-line
6-cylinder diesel engine on the basis of finite element calculation results.
The calculation methods used for the assessment of FE results are based on experimental and
theoretical research results, model trials and operational experience.
The basic procedure is based on the influencing quantity methods described in the TGL and the new
FKM guidelines [4]. On the basis of strength data from unnotched samples, local S/N curves are
calculated at FE nodes which are influenced by local component properties and loads.
The influence of notches on the component strength, for example, requires special solutions in the FE
modeling. Since the classical route using notch form factors is not available because of a lack
of nominal cross-sections, a method based on load gradients and material properties is used.
High cycle fatigue can be expressed by using endurance safety factors. The endurance safety factor
of the crankshaft is shown in Fig. 10. Crankshaft surface distribution of endurance safety factors
in Fig. 10 is presented for crankshaft with the rubber damper and for engine speed 2200 rpm.
A minimum value of endurance safety factor of crankshaft without a damper belongs to resonance of
the 6th harmonic order (2400 rpm) on the sixth crank throw. The local minimum values at engine
speeds 1200, 1600 and 1900 rpm belong to torsional resonances of 12th, 9th and 7.5th orders.
Minimum safety factor less than 1 is at the engine speeds 2350 2450 rpm and a crankshaft failure
would occur during engine life.

.
Fig. 10: Crankshaft endurance safety factors
The rubber vibration damper causes an increase of minimum endurance safety factors values
for all engine speed range, mainly above 2200 rpm, and makes a safe engine operation above
2200 rpm possible.
23rd CADFEM Users Meeting 2005
International Congress on FEM Technology
with ANSYS CFX & ICEM CFD Conference
November 9 11, 2005, International Congress Center Bundeshaus Bonn, Germany

6.

Conclusions

For crankshaft dynamic calculations, traditional methods based on simple calculation of torsional and
bending vibrations are state of art but new calculation methods are necessary to achieve more
accurate results that allow a design which is nearer to the mechanical limit.
A combination of CAD, MBS and FEM methods is the most efficient way of calculating cranktrain
dynamics and fatigue. The results are boundary conditions for further investigations, not only for stress
issues but also for questions concerning friction and wearing out of bearings or pistons.
The new techniques will replace the traditional methods with the increasing power of modern
computers. However, further studies are necessary to optimize the interaction of the CAE programs
and to validate the new technology.

Acknowledgement
Published results were acquired using the subsidization of the Ministry of Education
of Czech Republic, project 1M6840770002 Josef Boek Research Centre for Engine and Vehicle
Technology II.

References
[1]
[2]
[3]
[4]
[5]

Novotny P.: " Cranktrain dynamics simulation central module of virtual engine", Ph.D.-Thesis,
Brno University of Technology, December, 2004
Pistek V., Novotny P.:Multi-Body Analysis of diesel Engine Dynamics, XXXV. Mezinrodn
konference kateder a pracovi spalovacch motor eskch a slovenskch vysokch kol.
MZLU v Brn, 8th -9th September 2004
Pistek V., Novotny P.: Virtual Engine - A Tool For Reliability Increase. Reliability 2004, Military
Academy in Brno, 12th 14th October 2003
Magna Powertrain: FEMFAT 4.5 major release. Users manual, 2004
REBBERT M.: Simulation der Kurbewellendynamik unter Berucksichtigung
der hydrodynamischen Lagerung zur Losung motorakusticher Fragen. Disertation. Fakultt fr
Maschinenwesen der Rhenisch-Westflischen Technischen Hochschule Aachen, January 2000

23rd CADFEM Users Meeting 2005

International Congress on FEM Technology
with ANSYS CFX & ICEM CFD Conference
November 9 11, 2005, International Congress Center Bundeshaus Bonn, Germany

10