TVA, a combined approach: classic and experimental

Author: Mauro Lucotti

[email protected]
August 2019

Abstract transmission + drive shafts + tire. Well set on the vehicle,

not tuned at testing bed.
This paper deals with the Torsion Vibration Analysis for
reciprocating engines in engine testbeds. Another interesting case is represented by the clutch
skidding that happened when a motorbike powertrain
The aim of this study is to combine the classic mechanic was tested at dynamometer bench, connected with a
calculation of the modal analysis of the set specimen, driveshaft. This phenomenon didn’t happen on the
shaft and brake, solving the homogeneous system of vehicle on the road: the own frequencies of the
linear ODE, with the solution of the complete equations powertrain with the transmission chain and the tire were
in which forcings are represented by the inertia modal completely different from those at the engine test bed, so
components of a crank mechanism and the variable gas the torsional oscillations gave picks of torque,
pressure in the cylinder chamber generated by the fuel responsible of making the clutch skid.
combustion. This for giving a theoretic and practical
method for a fast but deep analysis. Unlike to other types of vibration, shafting vibrations are
not visible, however they may cause serious damages
under certain circumstances.

Introduction These simple case studies tell us that the torsion vibration
analysis shows whether a mechanical system is
The torsion vibration analysis is a well-known representative or not of a complex system like a true
methodology for studying the behaviour of rotors and vehicle on a road with all the possible variables of a Real
shafts in order to have the critical speeds in not harmful Driving.
ranges. Many studies and investigations of the torsion
vibration have been done since the beginning of last The shaft vibration problem may happen at low and high-
century, now we apply the theory and a model of analysis speed rotation and, as we told, it may not occur in real
to engine test cells. In engine testing it is possible to have driving, but it may happen in testing for the different
different conditions when an engine is tested at the bench global conditions.
rather that it runs in a vehicle on the road. This is due to
the different own frequencies of the whole system in each Some experimenters or test bed operators use to find the
application. first and the higher critical speeds operating a natural
deceleration of the system
It happened that, due to torsional oscillations, intake or
outlet valves broke down with an engine in testing at the
bench even if it was perfectly running on the vehicle. In
this case, relevant oscillations of the crank shaft produced
a no longer phase trim for the valve opening; the same
engine on the vehicle didn’t show the same issue. The
torsional oscillation depended on the different own
frequencies of the system engine + shaft + dynamometer
in comparison with the system engine + gearbox +
Figure 1 ‐ Mechanic system

1
in order to practically discover where these speeds are
located, from the highest revolution per minute up to the
engine stop

Figure 4 ‐ Double mass flywheel engine equivalent system

and the two natural frequencies become:

Figure 2 ‐ Critical speeds

Sudden or relevant variations of the torque mean that the

rotation is close to a critical speed. This procedure is very
realistic but, unfortunately, needs the true mechanical
system with all the components installed, so it gives an
information a posteriori; we are looking for an a priori
analysis.
where the C constant is:
In the following paragraphs a step by step approach is
developed in order to reach a quite complete model for
the forced vibration analysis result. Flexional or axial
analysis are subjects not treated in this paper.
With these simple models it is possible to find the first
own frequencies in order to define the inertia and the
Simple models stiffness of the coupling shaft for having the critical
speeds out of the range of working of the engine. We
A very simple approach consists in dealing with the normally try to put them below idle that should be at least
model as at the picture Figure 1 which gives the following √2 times greater than them:
system:

Figure 3 ‐ Two masses equivalent system

where the engine with a simple flywheel is connected to Figure 5 ‐ Amplification or damping
the brake and the natural frequency is
In this way, the greater the ratio 𝜔⁄𝜔 the higher the
damping. Operating in an overcritical zone the damping
ratio ζ of the system appears less important. It mainly
should be chosen for the transient phase when cranking
the engine and crossing over the critical speed up to idle,
Modern engines with a double mass flywheel can be so we try not to have big elongations during this phase.
represented according the following model: This is normally done using specific elastomers that
avoid relevant oscillations and, at the same time, filter the
periodic irregularity, that is relevant for 2, 3, 4 cylinders
engines or with small flywheels.

2
 n degrees of freedom models 𝐽𝜗 𝑘 𝜗 𝜗 0
The calculation model is based on the modal analysis 𝐽 𝜗 𝑘 𝜗 𝜗 𝑘 𝜗 𝜗 0
evaluating eigenvalues and eigenvectors, that represent
the own modes of vibration. Moreover, the analysis of 𝐽 𝜗 𝑘 𝜗 𝜗 0
the response of the system to forced vibrations, produced
by the engine, gives the whole behaviour in running and consequently, the simplest equation is:
conditions. 𝐽 0 0
The calculation is based on the harmonic analysis and on 𝐽 0 𝐽 0 matrix of inertia
the response in the frequency domain of the equivalent 0 0 𝐽
system. This is done with the reduction of the real system
𝑘 𝑘 0
made by the crank shaft and the elements in alternate
moving. The equivalent system with concentrate 𝐾 𝑘 𝑘 𝑘 𝑘 stiffness matrix
parameters, with only rotational degrees of freedom, has 0 𝑘 𝑘
all the parts connected to the crank shaft also reduced to 𝜗
disks in order to let have a complete system to be treated
𝜃 𝜗 ⇒ 𝐽𝜃 𝐾𝜃 0 (1)
in a matrix approach.
𝜗
In reciprocating systems, all the masses with a purely
The extension of the equation (1), with the introduction
translation motion (alternative) are concentrated at the
of the viscous damping leads to a more general equation
crankpin and the system can be replaced, from an energy
in the following form, the differential system equation of
point of view, with a disk, a flywheel, with an inertia that
motion, a system done with n second order linear
varies with the crank angle variation.
differential equations corresponding to n degrees of
This model leads to a differential system made by n freedom
equations in which there are all the components to
analyze (pulley, cylinders, flywheel, clutch, coupling 𝑑 𝑋 𝑑𝑋
𝑀 𝐶 𝐾𝑋 0 2
shaft, brake). 𝑑𝑡 𝑑𝑡
where:
The crankshaft stiffness was, in the past, defined by Ker
Wilson’s formula or Carter’s formula before introducing M = diagonal inertia matrix
a FEA approach. Ker Wilson himself suggested to use
C = damping matrix,
an average value between the outcome of his definition
symmetrical and tridiagonal
and of Carter’s; today we can see that this average value
is quite close to the results of a finite element analysis. K = force or stiffness matrix
Due to a practical approach, we don’t perform the mass symmetrical and tridiagonal
reduction with a rigorous solution of the motion equation. X = vector of variables
This should give a second order differential equation with
non-linear variable coefficients. So, we follow a t = time
traditional approach simplifying the motion equation
using average and constant equivalent moments of The solution is simpler when the system is reduced to a
inertia, no longer function of the angle of the conrod at new system of 2n ODE of the first order where the
the crankpin. Moreover, we treat as negligible the inertia variables are the state variables, the variables with their
due to the small oscillations of the con rod itself. own first derivatives.

So, starting from a simple model to extend it to a quite The solution of the homogeneous equation leads to find
complex one that considers the stiffens and the damping out the critical speeds, the own frequencies, i.e. the
of the engine and the shaft, we obtain: eigenvalues, and to define the modes of vibration, i.e. the
eigenvectors, the eigenvector represents the mode shape
of the vibration corresponding to the eigenvalue or
natural frequency.

The mechanical model for a 4-cylinder reciprocating

engine ideally starts from a configuration like this:

Figure 6 ‐ Three flywheels model

3
 Figure 7 ‐ 4‐cylinder crankshaft with pulley and flywheel

where the mass and the stiffness of the part of the crank
shaft, as well as the front-end and the mass of the
Houdaille damper and flywheel come from the engine
engineering. Internal damping coefficients may come
from literature figures, at first, and by calculation for the
tuned or untuned viscous torsional damper. k and c
coefficient will fill the three diagonals of K and C matrix.

Implementing the system above with the inertia coming

from the reduction of the piston and the con rod, the
inertia of the brake, of the coupling shaft with its
parameters for stiffness and damping, we get the
following torsional model, where the variables are the
angular position of the flywheels (the rotational degree of
freedom), in this case the last eight variables, and their
derivatives, the first eight ones.

and the natural frequencies are given by the imaginary

part of the eigenvalues of the matrix

𝑀 𝐾

𝜆 𝑒𝑖𝑔𝑒𝑛𝑣𝑎𝑙𝑠 𝑀 𝐾
In deeper words, they are calculated bringing C to 0 and
𝜆 represents the diagonalized matrix of eigenvalues.
The result of the equation

Figure 8 ‐ 4‐cylinder model and variables 𝑀 𝐾 𝜆 𝑋 0 3

gives the modes of vibration, mode shapes, associated to
Before developing this 2n system, let us analyze once
more the equation (2) that becomes: each eigenvalue. The eigenvector 𝑋 is a nonzero solution
of the equation (3) and it physically represents the mode
shape of the vibration corresponding to its own pulsation,
i.e. its natural frequency.

Then we have:

rad
𝜔 𝐼𝑚 𝜆
s

4
and the natural or own frequencies are:
𝜔
𝑓 Hz
2𝜋
So, thanks to the eigenvectors of the same matrix
calculated for each eigenvalue we get all the mode shape
of vibration of the mechanical system.

The nonzero angular displacement 𝑋, the vector

associated to each 𝜆 , is:

𝑋 𝑒𝑖𝑔𝑒𝑛𝑣𝑒𝑐 𝑀 𝐾, 𝜆
With reference to a 4 cylinders 4 strokes 2 liters Diesel
engine of the market, the engine that from now ahead will
be used for this analysis, and an asynchronous IGBT Figure 11 – First 7 modes of vibration
driven brake and a typical coupling shaft, we get, for
instance: The picture above shows the relative displacements of
each part of the torsional system at each natural
frequency, these are normalized values, not absolute.
The higher the relative displacement, the higher the stress
and the strain in the portion of the element.

The problem of finding the resonance conditions can be

easily approached with the use of the Campbell’s
diagram. This diagram shows on the abscissa the
frequency range to investigate and on the ordinates the
natural frequencies of the system, the bundle of lines are
the harmonics. Putting horizontal lines corresponding to
Figure 9 – Engine test bed the most relevant frequency modes, the possible
resonance conditions are given when there is an
intersection of one of the horizontal lines with the bundle.
The corresponding abscissa gives the possible resonance
with the n.th harmonic.

Figure 10 – Coupling shaft

Figure 12 – Campbell’s diagram

5
The picture above shows that in the operating range of The driving and inertia torque, as per equation (4), will
the engine there are not harmful resonances for this be applied in the model to the variables representing the
system. rotational degree of freedom of the cylinders, i.e. the four
variables 𝑥 ⋯ 𝑥 that stand for the cylinders, all the
other elements will be 0.
Forced vibrations With steady state forced vibration in the form:
The equation (2) is the basis for the analysis of the forced 𝑋 𝑡 𝐼𝑚 𝜒 𝑡
vibrations. This homogeneous equation gives us the free
vibrations of the system, the natural frequencies, the equation (5) becomes:
mode shape of vibration and the possible resonances.
The complete equation, with the second term no longer
𝑑𝑋 /
𝐹𝑋 𝑓𝑒
zero, gives us the behaviour of the system under a 𝑑𝑡
harmonic or complex force. On each disk or flywheel of and looking for solution in the form:
the equivalent system representing the crank of the real
system there is a torque that varies with time and that can /
𝜒 𝑡 𝑋 𝑒
be defined as sum of an “engine moment” or “driving
torque” and of a “inertia torque”. The total torque acting we get:
on each crankshaft throw / / /
𝑋 𝑖𝜔𝑒 𝐹𝑋 𝑒 𝑓𝑒
𝑀 𝜔 𝑀 𝜔 4
𝑖𝑘𝜔
𝐼 𝐹 𝑋 𝑓
can be written as Fourier series. The calculation of the 2
modal force should be developed according the stellar
diagram of the engine. The vector sum of the harmonics 𝑖𝑘𝜔
𝑋 𝐼 𝐹 𝑓 6
of driving torques and moments of inertia produces the 2
resulting stress on the system. The calculation of the
From this point ahead all the efforts are concentrated in
harmonics of the driving torque and the modal forces for
the definition of 𝑓 .
4-stroke engines is normally calculated using the Ker
Wilson formulas regardless of the phase. The driving torque comes from the pressures in the
cylinders and the map of these pressures vs. the crank
In the next paragraphs there will be a combination of the
angle and the rotational speed has the shape:
classic and practical theory for the definition of the stress
on the system with the results of the investigation on the
real combustion and pressure into the cylinder. The
complete equation will maintain the historical
formulation for the inertia while the driving moment will
come from the pressure into the cylinder.

For an easier analysis the equation (2) becomes now:

𝑑𝑋
𝐹𝑋 𝛷 5
𝑑𝑡
where the system is reduced to a new system of 2n ODE
of the first order according to the picture Figure 8 and the
matrix 𝐹 is:

𝑂 𝐼
𝐹
𝑀 𝐾 𝑀 𝐶 Figure 13 – Pressures in the cylinder

and 𝑂 is the null matrix, 𝐼 the diagonal identity matrix, and for all the cylinders, considering the firing angle and
the vector 𝑋 has the rotational degrees of freedom and the firing order, we have:
their derivatives

𝑥 … 𝑥

𝑥 𝜗 … 𝑥 𝜗
and the vector 𝛷 stands for the forcing.

6
 p(α)

Figure 15 – Fourier series at max pressure speed

This exciting pressure will be tuned, in a first

Figure 14 – Pressures in the cylinders
approximation, according to the behavior of the nmep for
We may do, from these pressures, a good approximation having a harmonic exiting function vs. rotational velocity
using a Fourier series based on the crank angle α and ω.
tuned on the base ω in order to define the first addendum
The amplitude and the phase of the k-th harmonic is:
for the equation (4). The calculation should be done for
two revolutions, so between -2π and +2π, and for the first
n harmonics, the greater n, the better it is. The data
represented in the picture Figure 14 should be “splined”
in order to get a function for getting Fourier coefficients,
let’s call this spline function at a certain revolution ω:

𝑔 𝛼
Considering the first n harmonics and Modal forces, for the same equation (4) can be calculated
with Ker Wilson formulas. The so-called inertia torque
𝑘 ∈ 0, 𝑛 is not negligible for the second, fourth, sixth and eighth
harmonic, the amplitudes of these harmonics in sine and
the arrays for cosine and sine coefficients (A, B) for each cosine components are:
harmonic are:
𝑎 2𝜋 𝑏 2𝜋 𝑛 20

2 𝛼 𝑎
𝐴 𝑔 𝛼 𝑐𝑜𝑠 2𝑘𝜋 𝑑𝛼
𝑏 𝑎 𝑏 𝑎

2 𝛼 𝑎
𝐵 𝑔 𝛼 𝑠𝑖𝑛 2𝑘𝜋 𝑑𝛼
𝑏 𝑎 𝑏 𝑎

At the chosen speed ω the pressure vs. the crank angle

becomes:
(7)
1 𝛼 𝑎 𝛼 𝑎
𝑝 𝛼 𝐴 𝐴 𝑐𝑜𝑠 2𝑘𝜋 𝐵 𝑠𝑖𝑛 2𝑘𝜋
2 𝑏 𝑎 𝑏 𝑎

and this last, the equation (7), is the modal moment that
comes from the kinematic of the crank mechanism, r and
Area are the crank radius and the section of the piston,
where:

7
 cAm  0.330 if k 2 Further developments
k
0.220 if k 4
Starting from the elements of the vector 𝑋
it is
0.110 if k 6 possible to not only verify the oscillation of the
0.051 if k 8 whole system but the stress at the shaft, at the shaft
0 otherwise dampers, as well as the vibrations at the engine mounting
and, from here, the vibration to the base plate can be
cBm  0.140 if k 2 calculated combining the elements of the vector in all
k
0.040 if k 4 the following mechanical or structural analysis.
0.048 if k 6 In the following, some examples are shown for the
0.051 if k 8 harmonics and for some other results got from the
analysis of the 4 cylinders 4 strokes Diesel engine as at
0 otherwise
pictures Figure 9, Figure 10.
b  0.0588 if k 2
k
0.5 if k 4
0.176 if k 6
0.0138 if k 8
0 otherwise

are the coefficients according to Ker Wilson studies, the

sum of the amplitudes of the harmonics should be done
considering the corresponding phases.

At last, the sum (4) gives the torque to apply to the

variables 𝑥 ⋯ 𝑥 (the rotational displacements that
represent the cylinders at the crank shaft) and it becomes
the sum of each torque, driving and modal time the phase
and time the component of the eigenvector. For cylinder
j we will have:
Figure 16 – Torsional oscillation of the flywheel
𝑇, 𝜔 𝑋 𝑀 , 𝜔 cos 𝜑 , 𝑀 , 𝜔 cos 𝜑

and the equation (6) will have the term 𝑓

0
0
⎛ 0 ⎞
⎜ 0 ⎟
⎜ 0 ⎟
⎜ 0 ⎟
⎜ 0 ⎟
⎜ 0 ⎟
𝑓 𝜔 ⎜ 0 ⎟
⎜𝑇 , 𝜔 ⎟
⎜ ⎟
𝑇 𝜔
⎜ , ⎟
⎜𝑇 , 𝜔 ⎟
⎜𝑇 , 𝜔 ⎟
⎜ 0 ⎟
0 Figure 17 – Torsional oscillation elastic coupling at the
⎝ 0 ⎠ flywheel

Now the equation (6) is complete and it can give all the Considering the parameters of the elastic coupling as at
information for investigating the forced torsional picture Figure 10 it is possible to check it for all the range
vibrations at the crankshaft, at the coupling shaft, at the of use and verifying whether the fatigue limits are
brake and so on with the results in the vector 𝑋 . exceeded or not in resonance conditions.

8
 Further analysis of the phenomenon can be done,
nowadays, with dedicated software that implements the
classic equation of motion considering all the parameters
of the crankshaft mechanism, geometry, material,
kinematics lubrication, etc. The torsional behavior can
be analyzed not only by analytical but by FEA too with
the well-known ANSYS or Comsol environment. The
results of these are quite close to our simple analysis even
if more precise when the geometry comes from the
engineering department.

When investigating on existing test beds or existing

systems this approach can be more powerful thanks to the
possibility of doing a reasonable simplification of the
Figure 18 – Torque at flexible coupling ‐ flywheel side model considering, by experience, the most influential
parameters.
The picture Figure 18 shows a strong resonance at a
rotational velocity below idle and must be crossed over
quite fast in cranking the engine. In all the possible range
of running the engine the admissible torque is never References
exceeded.
[1] Harris & Crede “Shock and Vibration Handbook” -
Following the same concepts, it is possible to investigate 2009
the reactions at the engine mounting due to the torsional [2] J.P. Den Hartog “Mechanical Vibrations” – 1st
vibration, regardless the dynamic performance of a edition 1934
certain test or duty cycle: [3] G. Genta “Principi e metodologie della
progettazione meccanica” - 1992
[4] D. Giacosa “Motori endotermici” - 2000
[5] A.J. Martyr & M.A. Plint “Engine testing” - 2012
[6] AVL “Excite Designer Theory” – 2007
[7] W. Ker Wilson “Practical solution of torsional
vibration problems” - 1946
[8] Amitpal Singh Punewale, Mr. Tushar Khobragade,
Mr. Amit Chaudhari, Prof. N. M. Khandre
“Analysis & Optimization of Torsional Vibrations
in a Four-Stroke Single Cylinder Diesel Engine
Crankshaft – IJETR 2015

Figure 19 – Load at engine mounting for torsional vibrations

Conclusions
With this approach a classic investigation of torsional
vibration in order to check natural frequencies and
resonances is developed. Moreover, thanks to
experimental data acquisition of the pressure in the
cylinders the forced vibrations influence on the whole
system is studied and verified for compliance. This study
tries to put together the engineering information and the
experimental acquisitions. This was done with
mathematical worksheets or environment, e.g. Matlab or
Mathcad, that can manage arrays, matrix and complex
functions.