Вы находитесь на странице: 1из 7

FINISHING PROCESSES

Sequencing and speed optimisation in


strip processing
New developments in strip
sequencing and speed control during
the operation of linked pickling and
tandem mill lines have proved
extremely efficient for increasing
productivity. The potential of VAIs
LinkMaster system for the control of
production has been demonstrated in
a way that is beyond the scope of
manual control. It is possible to

Figure 1 The sections of a linked pickling process and tandem mill line

increase the quality of the strip by


running the line more consistently and within
tighter operational limits.
The development process is not yet
complete and further advances are
confidently expected.

millennium steel2 k 2

Gerhard Finstermann, Helmuth Aflenzer,


Franz Mitter and Roman Winkler
VOEST-ALPINE Industrieanlagenbau
GmbH & Co

224

Process linkage and continuous operation are key


elements in efficient, high-quality cold strip
production. Important technologies have been
developed over the past decade to equip plants with
such capabilities. Some continuous working
procedures are already well-established, but the
combination into one line of the pickling process and
tandem mill cold rolling, and possibly coating, is now
the subject of intense activity. Strong competition in
flat-product markets and increasing demands for
improved quality make the prospect of line linkage
particularly attractive as a means of saving costs and
improving product consistency.
The conversion of a line from de-coupled to
coupled operation can be a difficult process. The
underlying operational philosophy of coupled
operation differs substantially from the level of

experience of many flat-product producers. Deciding


factors for achieving the full savings potential and
quality consistency of a linked line are the availability
and efficiency of the line. While the former is largely
a matter of well-established maintenance practice,
the latter depends on a level of production planning
and process control that, during the first years of
operation of a linked line, may be beyond the scope
of the plant personnel. Even linked operations that
are well-established may exhibit a degree of
inefficiency. Some sources of inefficiency are related
to the speed factor and are perhaps inevitable, while
others are operator-dependent and can be
minimised by sophisticated control strategies. Tools to
reduce linked-line inefficiencies are fairly new in the
industry and certainly offer considerable potential for
further improvements.
VAI is well-known as a developer of process
technology and a builder of turnkey plants for linked
lines. The company began extensive work in this
field some years ago and has recently developed a
complete package of software and control tools,
enabling lines to be run at a high level of efficiency.
In this article we shall outline the key elements of this
performance package, which is designed to
optimise production scheduling, and which includes
the LinkMaster, a Level 2 automation tool for
controlling the entire production process. Examples
are presented from simulation studies and from
practical applications to demonstrate the potential
for improved performance, and hence the cost
saving and product quality benefits offered by these
control tools.

FINISHING PROCESSES

Figure 2 Criteria and objectives of strip sequencing optimisation

There is in an inevitable loss of efficiency during


production where strips from different products are
joined together to meet a particular order. The
processing of these different products can result in
bottlenecks in different sections and hence different
coil cycle times for the sectionloopersection
combinations. This will infringe on mass flow
conditions and there is need for a method of
adaptive, on-line adjustment of section speeds. To
meet the design capacity of the process line, this effect
is taken into account by introducing a speed (or
efficiency loss) factor, which can be up to 10 per cent.
In other words, the loss of efficiency is compensated
by a pre-defined overcapacity.
As will be explained later, the optimisation of strip
sequencing under various boundary conditions
represents a significant influencing variable with
regard to the natural efficiency of linked process lines.
The more smoothly the products can be varied from
strip to strip and the strips joined into a continuous
stream, the more successful is the simplified
assumption of throughput calculation. This
assumption approaches the optimum condition
asymptotically where there is a continuous sequence
of the same products in the linked process line. In
reality, this condition is only seldom achieved.
Apart from the speed factor, the availability of the
line must be taken into account when determining the
ideal overcapacity required for meeting the rated
capacity under actual operating conditions. If strip
sequencing can be implemented in an optimised
manner then the speed factor-dependent design
overcapacity can be utilised to some extent during
operations as a potential means of reducing specific
production costs.
Two other factors offer a similar potential for the
optimisation of linked lines: disturbance-driven
inefficiency and operator-driven inefficiency. Clearly,
apart from the rate of operational disturbance that
leads to lower line speeds, or even causes stoppages,

millennium steel2 k 2

Linked pickling-tandem mill lines consist of four


sections: entry, pickling, side trimming, and the
tandem mill with exit coilers (see Figure 1). Each
section is usually operated at different speeds and in a
differential mode, bearing in mind, for example, that
the welding of strips on entry requires a line stoppage,
while a change in strip width between adjacent coils
requires stoppage of the side trimmer. Since the critical
sections the pickling process and the tandem mill
should operate at fairly constant speeds, a method of
speed compensation is necessary; this is done by the
use of strip loopers or accumulators for storage of the
strip. The mass flow from entry through to exit of the
strip must always be constant to avoid line overfill or
emptying. This means that the integral of speed over a
specific time interval for each section must comply with
the requirement of constant mass flow. The differential
speed variation of sections within a particular time
frame either fills or empties the respective loopers, that
is, depending on the speed differential between the
entry and exit side of the looper.
The key element in the design of process lines is
throughput calculation, which serves not only to
determine the rate of production, or the annual
capacity of the line, but also to determine the size of
the loopers. The main input parameters required are
the permissible strip speeds in the main process
sections, which are dependent on the dimensions
and composition of the strip, the degree of reduction,
as well as strip pre-treatment parameters. Speed
capabilities in the other line sections are also used as
input parameters in the throughput calculation.
Individual throughput data are calculated for a
representative selection of products, differing in
dimensions, cold reduction and type, taken from the
product mix. This simplified approach basically
determines the achievable coil cycle or coil-to-coil
time for each product for the various linkages in the
line: entry sectionentry looperpickling section;
pickling sectionintermediate looperside trimmer
section, and side trimmer sectionexit loopertandem
mill. Depending on product properties, any one of
these linkages may be a bottleneck for a particular
product. The speeds of sections where there is no
likelihood of a bottleneck can be reduced to comply
with the mass flow principle.
Since this simplified procedure is applied on a
specific product basis, the line can be regarded as
producing the equivalent of a continuous stream of
strips from the same product, with the different
products being treated independently of one another.
At the planning stage this is the best possible estimate
that can be made, because no information is available
from actual strip sequencing based on combinations
of different products. Obviously, under production
conditions, strip sequencing will change according to
order intake, market requirements and so on.

225

FINISHING PROCESSES

for the entry section, side trimmer and tandem mill, is


an extremely critical factor in reaching an acceptable
level of efficiency. Usually the trimmer section
operator is the prime mover of this coordination
process. His task is to control the line-disturbance and
product-sequence-driven interaction of speed
deviations between the pickling section and the
tandem mill in such a way that the speed in both
sections is kept as constant as possible, and that
neither of these sections has to be stopped because
looper limits are exceeded. Since coil cycle times in
the line sections can vary to a certain extent and line
disturbances occur in a stochastic manner, this
coordination task can be quite demanding.
The foregoing considerations provided the
motivation for VAIs development of the LinkMaster, a
computer-controlled method of speed supervision
designed to replace the need for manual
coordination. Apart from relieving the operators of
the burden of speed coordination, there are
additional benefits such as improvement in line
efficiency and product quality due the facility of
maintaining process speeds within set limits. Efficiency
is improved through the use of sophisticated methods
of optimisation that are based on actual line
conditions and the expected cycle times in the various
sections. The system reacts extremely quickly to all
kinds of process disturbances by selective speed
adaptation in the line sections, thus ensuring a fast
speed recovery. The new system also has the
advantage of compensating for inefficiency effects
introduced by inexperienced operators and this is
particularly helpful during the start-up of a linked
process line, as will be described later.

millennium steel2 k 2

Strip sequencing
Strip sequencing is the procedure whereby a batch of
coils is lined up to provide a chain of strips for
continuous processing through the line. A strip
sequence is subject to restrictions imposed by
variations in hot band and finished gauges, and in
strip width and material types permissible between
adjacent strips. These restrictions are, for example,
associated with:

226

Figure 3 Optimised strip sequencing results for


width, entry and exit thicknesses, and yield stress
for 162 strips

the time needed to recover from such disturbances


results in loss of efficiency.
Timely intervention by an operator, in particular in
the coordination of inter-sectional speed control
changes between the different operators responsible

Welding machine: joining different materials,


differences in strip thickness and width
Side trimmer: differences in strip width
Rolling mill: differences in strip width and entry/exit
thickness; differences in materials.
From the economic viewpoint, the essential task
of sequencing is to minimise speed factor-related
inefficiency. Safety aspects can also be exemplified
by the need to reduce the incidence of strip and
weld breaks particularly during the rolling of weld
joints. The criteria for strip sequence optimisation

FINISHING PROCESSES

Optimisation of strip sequencing


We have seen that the optimisation of strip sequencing
can maximise throughput while achieving specified
product quality and consistency parameters. The first
step is, of course, to maximise process speeds in all
line sections; for example, by defining maximum
pickling speeds for individual products, minimising
downtimes required for shearing and welding
operations, and maximising pass schedule rolling
speeds, bearing in mind any technological restrictions.
This first step optimises the processing of single
products only in the different sections of a continuous
line; a second step is necessary to optimise the
interaction of all sections as well as the strip sequence.
VAIs solution to this problem is to apply the
TabuSearch concept proposed by Hinterleitner and
Zeman of the Johannes Kepler University in Linz.
The basic idea is to take any starting value and to
proceed along a path with a decreasing objective
function while avoiding inverse steps in the search
path. The objective function of this approach is the
sum of all restriction weights for the entire product mix.
This solution is not necessarily unique, and the
objective function value of the global solution
generally is not zero because of wide variation in the
product mix. Therefore, a set of solutions is searched
at an initial stage; the respective production times are
then calculated for each solution, with the final solution
being a compromise between the two premises.

Figure 4 Throughput optimisation by speed coordination

The computational complexity of TabuSearch can


be reduced to the order of O(n3) (where n is the
number of strips); this has a considerable effect
when dealing with several hundreds or even
thousands of strips, since the solution set in its
original form has a size of n!
Additional procedures can be optimised, such as:
Handling priority orders
Thermal crown build-up on strips after roll change,
width distribution
Work-roll changes, and so on.
The example in Figure 3 concerns an optimised coil
sequence, based on a product mix of 162 coils, and
shows the variation of the main strip parameters:
width, entry thickness, exit thickness and yield stress.
In Figure 3a, the width distribution was ranked as the
most important parameter (using penalty techniques)
so that a typical coffin shape resulted; in effect this
means a monotonic decrease of widths during the
sequence.
The optimiser arranges the coils in the sequence so
that apart from meeting this governing rule of
monotonic decrease in widths, there are as few as
possible violations of the permissible variations in
thickness and yield stress, from coil to coil. The aim
is, of course, to avoid any such violations but this
does depend on the potential for rearrangement of
the particular product mix. The prioritising of a coffinshaped width distribution is the customary practice
but it is not the only solution. The optimiser can be
forced to adopt any other type of strategy; for
instance, the governing rule can be to permit no
change in thickness or in yield stress, which then
leads to the type of distributions shown in Figures 3
b, c and d.

millennium steel2 k 2

and the objectives of the process are summarised in


Figure 2.
Product-mix cycle times usually vary from one to two
weeks in duration, depending mainly on the
complexity of the mix. The number of coils available
for sequencing ranges from 1,000 to 2,500 coils for
an annual line capacity of 1.3 Mt and from 2,000 to
4,000 for a capacity of 2.0 Mt. The production
planning time window is not necessarily coincident
with this mix cycle time and this reduces the actual
number of coils available for sequencing operations.
The width of this window determines the degrees of
freedom for strip sequencing under the various
restrictions to be observed. The narrower the window,
the greater is the chance of violation of joining
restrictions, which reduces the stability and efficiency
of production. In principle, a planning window can be
as narrow as one day, or even one shift, which for
complex product mixes makes strip sequencing
difficult but not necessarily impossible. Our
experience is that, for most product mixes, it is usually
possible to sequence hot coil inventories in an
optimised manner for planning windows of three days
or more. This provides sufficient production flexibility
even for medium-size capacity lines where the
producer must take into account the volatility of
market conditions.

227

FINISHING PROCESSES

with the requirement of achieving constant speed


over the length of the strip.
There are some restrictions in the solution of the
optimisation problem as a consequence of the factors
stated below:

millennium steel2 k 2

Figure 5 Measured and optimised values for the entry section-entry


looperprocess section speeds, and for the looper fill level

228

The application of LinkMaster


The essential objectives of LinkMaster, as illustrated in
Figure 4, are constant speeds in both the pickling
process section and the tandem mill, that is, except
where special speed profiles are required for weld
passing and cutting. The result is a process that is
more homogeneous in operation, with products of a
higher standard of quality, as well as the attainment
of a maximum line throughput.
Strip speeds for the pickling section, trimming
section and tandem mill must be optimised within the
technological ranges while keeping the content of the
loopers within the operational range.
Conventional approaches to the coordination of the
sections of a linked line (manual control together with
state-of-the-art automation) usually take into account
only the plant situation and the strips in the line. But
the line should have the capability for optimum
running not just for a short time but for long periods.
The calculation and optimisation executed by the
LinkMaster is therefore aimed at a period of 15 or 20
minutes. Coils in position awaiting entry to the line are
taken into consideration according to the planned
production sequence.
The central feature of VAIs solution is the
consistent application of a mathematical optimisation theory. The objective function, that is, the
function to be minimised or maximised, essentially
represents the total entry length into the tandem mill
within a fixed period of time. The vector of variables
to be optimised comprises the speeds of the relevant
strips in each section of the line, that is, except in the
case of entry section speed which is directly
controlled by Level 1 automation using a
straightforward strategy. The action of setting up only
one aim-speed per strip and section then complies

Lower and upper limits for the strip speed in the


various sections The feasible ranges for the pickling
and tandem mill speeds are the result of preliminary
calculations made using the corresponding set-up
models; the trimming speed is limited by plant
restrictions. In all sections additional limits can be set by
the operating personnel.
Lower and upper limits for looper filling The
checking of these limits is only necessary in case of
speed changes in the sections immediately before or
after the looper. For practical purposes, it is sufficient
to replace acceleration/deceleration ramps in the
speed profile by speed jumps in the middle of the
acceleration or deceleration interval; this yields a
finite number of time points to be checked.
Upper limit for the speed difference between
two sections linked by a looper These limits are
design parameters determined by the number of
looper strands and the maximum permissible speed
of the looper cars.
For the solution of these non-linear, restricted
optimisation problems, the VAI LinkMaster uses
Schittkowskis sequential quadratic programming
method. The evaluation of the restrictions and the
objective functions within the iterations of the algorithm
requires the construction of profiles for speed versus
time and looper content versus time over the entire 15
or 20 minute period that was mentioned earlier.
Starting with the actual plant condition as described
by the looper filling rates, the speed in each section,
the position of materials in the line, and the duration
of stoppages in the entry or trimming sections, the
calculation must take account of the following:
Procedures at reduced speeds for weld passing in
the trimming section and tandem mill, and for
cutting at the tandem mill exit side
Procedures involving planned stoppages for notching
and side trimmer adjustment, for example, because
of a change in width between adjacent strips
Speed control in the entry section when the tail end
of the strip approaches the shear; stoppages for
tail-end cutting and welding; and acceleration to
the maximum possible speed until the entry looper
is full or the next tail-end sequence starts, and
synchronisation with the pickling speed if the
sequence has not yet started
Although the LinkMaster does not control the speed
in the entry section, the speed profile must be
modelled so as to evaluate the restriction concerning
the lower limit of the entry looper, and also that

FINISHING PROCESSES

concerning the absolute difference between the


entry-section speed and the pickling speed.

Figure 6 Measured and optimised values for processed strip length in the
various sections

Figure 7 Performance monitoring display. (Note that the time intervals of


the chart abscissa run from right to left)

and tandem mill. Such predictions depend on the


product and might vary from strip to strip. One
important side effect is that the more homogeneous
the speed, the greater the efficiency; in other words,
the longer the strip length processed in a given time
frame. By comparing the predicted and measured
speed profiles, it is possible to locate any bottlenecks
over a specified observation period. This comparison
is not required in on-line applications because
optimum section speeds are determined in a cyclic
mode and imposed on the line speed as they are
generated.
Figure 5 was generated off-line using measured
line data and represents several successive speed
predictions over a 30 min observation period. The
measured values for speed and looper content are
shown in blue and the predicted, optimised values in
red. The diagrams show data for the entry

millennium steel2 k 2

With regard to the duration of the planned


stoppages in the entry and trimming sections, only
estimates can be used in the calculation. In fact,
several influences that are difficult to predict may
affect stoppage times, such as the possibility of
manual intervention by operating personnel or
variation in the quality of coils entering the line. The
incoming hot band may, for example, be damaged or
subject to severe dimensional deviations in the head
and/or tail, resulting in line delays.
To ensure a rapid reaction in the event of line
disturbances, the LinkMaster executes optimisation
calculations in a cyclic mode (cycle time ~ 30 sec). If,
for example, the LinkMaster predicts that a looper
limit will be exceeded before the next cyclic
optimisation calculation is completed, it will
immediately start the actual optimisation procedure.
Results from a previous calculation are used to
estimate the variable strip speeds in the next
calculation and to help minimise the iteration needed
to find the optimum solution.
The optimised speeds for following strips that enter
the process section, trimming section and tandem mill,
and, if necessary, also the data for the strips already in
these sections are passed to the Level 1 automation
system. The transmission of set-up information only for
the following strips is designed to comply with the
requirement for constant speed over the strip length.
Off-line applications Apart from the normal online application, the VAI LinkMaster can be used offline to analyse the production performance of linked
lines. In fact, this was the original motivation for the
development because the detection of low efficiency
operation in linked lines, based on measured speed
profiles of the various sections, is a tedious process.
The visible impact of low efficiency on the section
speed does not necessarily coincide with the source
because inefficiency factors usually propagate
through the line in a waveform, and the effects can be
evident at either the entry or exit end of the line. Spot
checks of speeds are difficult to interpret and thus it is
likely that any attempt made to identify a bottleneck
will fail due to the subjective nature of operators
observations. LinkMaster is able to attack the problem
in the off-line mode by reversing the strategy used online in order to compare section speeds.
From the measured production data, any suitable
point of time is used as a starting point to initialise the
program using the relevant line data, such as section
speeds or looper content. The program then predicts,
in a manner analogous to the on-line application, the
changes necessary in section speeds, for all the strips
in the line and those following, to obtain a fairly
homogeneous speed profile in the pickling process

229

FINISHING PROCESSES

millennium steel2 k 2

Figure 8 Comparison of manual and LinkMaster supervision of speed


control of a linked line during the start-up stage

230

sectionentry looper-pickling section grouping.


Similar plots were obtained for other groupings in the
line. The speed distributions clearly indicate the
beneficial optimisation effect of LinkMaster compared
with as-run values. The LinkMaster cannot, of
course, remove the strip-sequence-driven inefficiency
effects but it does effectively balance the section
speeds, which considerably improves the quality
consistence of products. Mention should also be
made of the differences in the utilisation of the loopers
in the as-run and optimised conditions which is due
mainly to the elimination of operator-driven, safety
cushioning systems provided to avoid exceeding the
fill limit.
There is an increase in the rate of production, as
shown in Figure 6, where production is expressed as
accumulated strip length processed in the different
sections. In this example, the increase is almost 20 per
cent on entry into the tandem mill but this is a rather
extreme value. For a well-established production
operation with manual speed coordination, the
maximum increase is likely to be about 10 per cent on
average. But even if the increase in productivity is only
to 5 per cent, the benefits of using LinkMaster are
considerable; in addition, there is the advantage of
greater consistency in production conditions.
On-line applications Figure 7 is a visual display
screen of VAIs performance-monitoring tool that
supplements the LinkMaster. This tool monitors line
performance with various features provided to indicate
line bottleneck status (in increments of one minute), to
show speed targets and actual speeds of each line
section (lower chart), and to provide statistical data.
The process efficiency can be shown for individual
sections or for the entire line. The histograms show the
efficiency distribution over a particular time period in
the pickling process and the tandem mill.
The histogram bars at the left of Figure 7, in both
the process section and tandem mill charts, within the

efficiency range of zero to 10 per cent, indicate the


presence of bottlenecks on line entry; this has resulted
in speed reductions in both the sections.
The bars in the process section chart, within the
efficiency range of 10-20 per cent, also indicate line
entry delays, but to a much lesser extent; these delays
have progressed through to the tandem mill as a
consequence of the damping effect of the loopers in
relation to disturbance propagation.
The observation period in this example, as
indicated in the lower chart, is roughly 11/2 hours.
During this period, most of the products were run with
the process section operating at 50-70 per cent
efficiency and the tandem mill at 70-100 per cent
efficiency. This indicates that the natural bottleneck in
the line (disregarding random delays) was the mill
(due to the rolling of thin, wide and hard products)
with the process section consequently running at
speeds that were lower than those permissible in
order to comply with the mass flow principle. Here,
variation of product dimensions is not the key issue.
The thinner the exit thickness (high rolling reductions)
in conjunction with a large strip width and high yield
stress, the greater the need to run the mill at the drive
power limits. This obviously limits the speed of the mill
and consequently the speed of upstream sections
(pickle line) has to reduced otherwise the looper will
overflow. Thus the utilisation (or efficiency) of the
pickle line is reduced (in Figure 7 to between 50 and
70 per cent). Actual mill utilisation ranges from 70 to
100 per cent. (Note, that the histograms show the
percentage of coils run at various levels of utilisation
over the observation period of 11/2 hours .)
Start-up applications The final example, shown in
Figure 8, is concerned with the capabilities of
LinkMaster during the start-up stage of a linked
installation. The histograms show overall line efficiency
and the superimposed lines the weekly average speed
of the process section. The yellow bars are data
obtained during manual supervision of a linked line
and the purple bars for LinkMaster supervision. The
average speed in the process section is an additional
indicator of running efficiency. In this particular
example, the line efficiency and average pickling speed
were significantly increased when the LinkMaster was
applied, thus minimising the need for an inexperienced
operator to intervene in the control of line speed.
LinkMaster is a registered trade name of VAI.
Gerhard Finstermann, is Vice President, Cold Band;
Aflenzer Helmuth and Mitter Franz are both Project
Engineers, Automation and Drive Systems for
Rolling and Processing, Process Optimisation; and
Winkler Roman is R&D Project Manager, all at
VOEST-ALPINE Industrieanlagenbau GmbH & Co
in Linz, Austria.

Вам также может понравиться