Proceedings of The th Intemational Conference on Stel Rolling (STEEL ROLLING '98),
1998, Chiba, Japan, The Iron and Steel Institue of Japan
FA
Rolling Mills
Modelling and Simulation of Vibrational Phenomena in
Paul A. MEEHAN, W. John EDWARDS and Glen A. WALLACE
Industria) Automation Services Ply. Ld, Teralba, NSW, 2284, Australia
ABSTRACT
‘The vibratinal phenomenon known as chatter is investigated using mathematical models to analyse and simulate its
‘ccutrence in roting mils. Models for torsional, third ara fith octave chatter are presented and discussed, & stabity erteron
{or the erticalroling speed is developed for third octave chatter in tems of known roling parameters, This cretion provides a
Useful practical design tool forthe avoidance of third octave chatter. An investigaion ofthe effct of various mil parameters on
the occurrence of third octave chatter is performed. tn general, the results presented inthis paper, coreate wall with practical
‘observation anc provide significant new ineights into methogs of avoidance of chater instability
KEY WORDS: chatter vibration: numerical simulation; tied octave: ith octave,
INTRODUCTION
Over the past century, significant advances in
technology have allowed rolling mills to operate at
relatively high speeds to increase productivity and
efficiency. However, these speed increases have often
been accompanied by mechanical vibrations, generally
referred to a5 chater, which cause periodic, ransverse,
bands of light and dark appearance across the stip. In
some eases, a matching thickness variation is associated
with the “chatter bands”. This phenomenon is highly
undesirable. Not only must the affected product be
rejected, but strip beaks may lead to damage inthe mill
equipment. Usuell ithas been found that as mill speeds
ate increased the vibrations become more severe. Thus,
at present, the only reliable remedy for chatter is to
reduce the operating speed of the rolling mill which
adversely affects mill productivity,
Generally, three distinet types of mill chater, have
been identified, namely:
* Torsional chatter, which causes significant chatter
bands across the strip and small thickness
fluctuations, occurs in the 5-25 Hz range. This is
‘often referred to as rumble or shudder, reflecting the
Jow frequency range in relation to the audible range.
‘The small variations in thickness may cause
uctuations in surface reflectivity which are
aesthetically unacceptable.
‘Third octave mode chatter, which produces large
thickness variations and strip rupture lies inthe 125-
240 Hz range. This form of chatter is found to occur
mostly in high speed tandem cold mills rolling thin
strip products. It is characterised by a sudden
‘occurrence (usually <5 seconds) and thus appears to
be selexcited as opposed to externally excited
+ Fifth octave mode chatter occurs in the range 500-
800 Hz and results in transverse banding of the
backup and work rolls and matching transverse
surface marking ofthe strip. To cure the problem, an
unscheduled backup roll change is often required
166
There is experimental evidence to suggest that
although there is negligible thickness variation, the
strip is not flat and has a periodic corrugated
waveform, These strip marks (or comugations) are
sull visible after the strip has been temper rolled and
painted.
Of the three types of chatter, third octave made
chatter is the most destructive and has the most
detrimental effects on mill productivity due tothe lower
rolling speeds required to avoid the phenomena,
However fifth octave chatter seems to be more
prevalent ia rolling mills and is of increasing concern as
customers are demanding better surface quality
It seems likely that with better knowledge of the
chatter mechanism, more effective strategies may be
formulated. If accurate chatter simulation models can be
developed they will provide an invaluable tool for the
investigation of alternative methods for suppressing the
phenomena. The present paper describes recent research
into the modelling, analysis and simulation of chatter.
Results pertaining to torsional, third and fifth octave
chatter will be dealt with in tum. The mathematical
symbols and notation used are defined in the
nomenclature section
2 TORSIONAL CHATTER
Torsional vibration involves the mill components,
shown in Fig. 1. The rolls undergo periodic angular
speed variations, while the drive spindles act as
torsional springs. The strip also acts as a spring in this
system, but under most conditions its influence is not
significant compared to that of the spindles.
‘When the top and bottom work rolls are out of phase
the response of the rolling torque to an asymmetrical
rolling condition needs to be considered in the analysis,
In the past, this aspect has usually been neglected.
Recent research") has shown that differential roll speeds
lead to a reduced roll force and increased torque.Fig. 1 Components ofa 4-High Mill Involved in
Torsional Vibration
Early experimental work by Moller and Hoggart®!
identified the main parameters involved in the torsional
vibration of @ mill rolling aluminium strip. A relatively
simple calculation of the natural torsional vibration
frequency of an experimental 2-high mill agreed closely
with their observations. For a typical mill, the natural
frequency 0, is givenby:
fai 283 ~)
Sad (On +”°Q,) 2
where
(Q,. = torsionalstifines ofthe drive shaft
Op. = torsionalstiffess ofthe spindls/couplings
Jj =momentofineriaof the rolls and spindles
J, =momentof inertiaof the motor
armatureand diveshat
and n= gearbox ratio (motor:roll speed).
Fora 4-high mill, the moments of inertia of both work
and backup rolls need to be included in the term J. tis
‘often the case that the torsional stiffness of the motor
‘gearbox shaft is much higher than that of the roll drive
spindles, Furthermore, if the referred motor inertia
n° J', is significantly greater than the effective rotl
inertia, the expression collapses to:
2 2)
5 @
Some occurrences of sustained torsional vibration
have been attributed to speed control systems. Badly
tuned control loops or noisy speed feedback signals
from tachometers have been reported to cause
instability. Gears in the drive train have also been
identified as the source of vibration in some cases.
3. THIRD OCTAVE CHATTER
It is generally accepted that third octave chatter is a
self-excited phenomena. Insight into the dynamics of
this mode of chatter have been provided by several
authors!” From these analyses it is clear that the
phenomenon has a relationship with interstand tension
‘STEEL ROLLING °98, Chibe, ISIS
fluctuations in the strip and the inherent vertical
dynamics of the roll stack. In particular, Paton and
Critchley”? confirmed experimentally that there is @ 90
degree phase shift between the workroll positions and
the stand eniry tension oscillations at the chatter
frequency. —Thusty" and Tamiya showed that
instability could arise if the tension was out of phase
with the roll deflections but neglected forward slip
variations, Using a different approach, Roberts derives
instability criteria from simplified roll force equations
that predict chatter due to excessive entry tension and
inadequate roll-bite lubricity, For a more complete
model it appears that the incorporation of tension
effects, slip interaction and roll stack dynamics is
required with minimal assumptions about the dynamics
of each mechanism. Chefneux et al! amongst others,
have developed such models to numerically simulate
chatter, however, no major insights were gained
pertaining to the exact condition for instability due to
the complexity of the models. More recently, Jepsen
and Kneppe" have used simplified models to obtain
qualitative conditions for third octave chatter in a
tandem mill however the significant interaction due to
slip was not included,
31 Dynamic Modelling
A key aim of this research is to develop an explicit
criterion for predicting the rolling speed at which
chatter instability will occur in terms of the rolling
model parameters. For the purpose of investigating third
octave chatter, the vertical roll stack dynamics may be
modelled by utilising the discrete spring-mass-damper
system depicted in Fig, 2. Note that the work roll - strip
flattening stiffness k,.is included explicitly,
By assuming that the roll stack is vertically
symmetrical about the stand passline, the relative
movement between the backup chocks and rolls is small
and the damping is low, the system shown in Fig. 2
reduces to a 4 degree of freedom system as described in
Fig. 3. This figure illustrates
each of the modal roll
displacements and natural
frequencies of the system.
I may be seen from Fig. 3
that only modes 2 and 4 will
be excited by roll force
variations in the strip and
cause significant thickness
variations. Only mode 2 has a
natural frequency within the
range of observed third
‘octave chatter. Modes 3 and 4
could be in the fifth octave
chatter range.
For chatter to occur, the
vertical oscillations of the roll
stack must be excited. The
‘most likely mechanism for
Fig. 2 Dynamic Model
of Four High Roll
167 =STEEL ROLLING "98, Chiba, ISI
Fig. 3 Mode Shapes of Vertical Motion of Four
High Roll Stack
this in third octave chatter is via strip deformation
interactions which incorporate interstand dynamics.
Practical observations!” suggest that tension oscillations
in the strip become large during chatter and have a
relationship with the vertical motion of the roll stack
‘due to roll force induced variations. Tlusty et al") and
Pawelski et al.” also indicate that the roll bite geometry
may vary ftom conventional models when the ratio of
the chatter vibration period is of the same order as the
rollgap transit time.
To understand the relationship between the roll force,
vertical roll vibrations and interstand tension dynamics
it is instructive to inspect the block diagram shown in
Fig. 4 and to consider the consequence of a perturbation
in the gap between the work rolls of stand i.
Hee
Fig. 4 Simplified Block Diagram Illustrating Rollstack
Vibration and Interstand Tension Interaction
Firstly, the exit thickness will increase. To ensure
continuity of mass flow, the strip entry speed will have
to inerease by the same fraction. This produces a speed
‘mismatch with the strip exiting the previous stand, and
the interstand tension force will start to increase as
would be the case if the interstand strip was a static
elastie spring with a relative movement between its
ends. An increasing tension will reduce the roll force
and lead to a reduction in strip exit thickness by moving
the work and backup rolls, partially cancelling the effect
of the initial gap perturbation. The tension increase also
induces slip changes of opposite sign in the stands at
each end of the interstand gap. The tension response
will stabilise at a finite value which reflects the
combined influence of the thickness and slip
interactions. The response of the tension to a change in
roll gap position may be shown to be equivalent to a
first order linear system with a time constant s, given by
the equation:
L
ye ey
vwen|(r'e,)_ laren)
‘The crucial issue here is that the time constant +, is
inversely proportional to mill speed. This means that at
ow speed the time constant r, is large enough to inhibit
significant tension fluctuations at the natural frequency
of the roll stack. As the mill speed increases, the time
constant reduces allowing tension interaction to occur,
This speed dependent interaction will contribute a
component of negative damping to the system. Under
certain conditions, this may be sufficient to cancel the
inherent positive damping in the roll stack, and
sustained vibration, or chatter, will then ensue.
‘The chatter models and mechanism described above
were investigated via stability analysis and numerical
simulation. In the following, a brief description of these
methods and results is provided,
32 Stability Criterion
The equations of motion for the mill dynamic models
described may be solved algebraically to derive a
relationship between the primary variables involved in
the chatter mechanism. Classical stability analysis
techniques can then be employed to obtain an explicit
expression for the critical rolling speed. Details of the