optimization & analytical opportunities in the …
TRANSCRIPT
OPTIMIZATION & ANALYTICAL
OPPORTUNITIES IN THE STEEL
INDUSTRY
BISWAJIT ROYCHOWDHURY
TATA STEEL
IIT D
elh
i –
IBM
Researc
h
Opera
tional R
esearc
h W
eekend
April 9
th2006
Agenda
Conte
xt
-Tata
Ste
el
-In
tegra
ted S
teel P
lant O
pera
tions B
ack G
round
Are
as for O
ptim
ization is S
teel In
dustry
-Pro
cess
-Logis
tics
-C
apacity
An u
nsolv
ed p
roble
m
Challe
nges
45 M
inute
s
Tata Steel Limited
Tata Steel Limited
5 MT fully integrated steel
plant
Key S
egm
ent:Automotive
and Construction
40,000 work force
No#1 Steel Company in the
world-World Steel Dynamics
2006
Fro
m d
irt to
door panels
Inbound
Iron
Coal
Limestone
MRO
Energy
Scrap
Outbound
Distribution
Center
Service
Center
Tier 1
Automotive
Assembly
Appliance
OEM
Torpedo
COAL
Iron Ore
Lime Stone
Imported Coal
Blending Yard
Blending Yard
Calcining
Coke Ovens
Sinter
Other fluxes
Sponge iron
CHARGE
Reheating F
cs Descale
rR
oughin
g M
illC
oil
Box
Descale
rC
ontinuous M
illLam
ella
r
Coolin
g
Coile
rH
R C
oil
Un-c
oile
rLevele
rTrim
mer
Pre
cis
ion
Levele
r
Fly
ing
Shear Bundle
s
HR
Coil
yard P
icklin
gC
old
Rolling M
ill
BAF
Galv
aniz
ing 1
Galv
aniz
ing 2
Skin
pass
Recoiling
Packagin
gC
R c
oil
Yard
Hot M
eta
l
Handlin
g
Desulp
hurization
BOF
RH
LF
Caste
r
Sla
bIn
spection
RsRsRsRs/ ///
− −−−R
oad T
ransport
Rail
Tra
nsport
Ete
rnal Pro
cessin
g
Sto
ck yard
Outb
ound T
ransport
Oth
er transfe
r s
eg
Tube,
Tin
pla
te e
tc
An integrated steel plant is a complex
industrial system in which LIMITED
materials flow through VARIOUS routes to
generate NUMEROUS products through
DIFFERENT series of production units.
Areas of Optimization
OR
orororor
Optimization
!!
Pro
cess
Modelin
g
Capacity
bala
ncin
g
Logis
tics
optim
ization
•Load b
ala
ncin
g o
n
each w
ork
cente
r
•Modelin
g d
em
and
•Pro
duct M
ix
optim
ization
•Invento
ry
optim
ization
•Desig
n b
uffer
•Schedulin
g e
ach
work
cente
r
Etc
..
•Desig
n
transportation
netw
ork
to reduce
freig
ht cost.
•Raw
Mate
rial
movem
ent and
sto
rage.
•WIP
flo
w
bala
ncin
g
Etc
…
Illustrative Scope of Optimization
•Ble
nd o
ptim
ization
of coal m
ix, sin
ter
charg
e, Bla
st
Fces.
•Heat and m
ass
bala
nce for ste
el
make o
pera
tion
•Modelin
g b
low
pattern
in B
OF
•Modelin
g o
f
decarb
urization in
RH
-degasser
•BAF fill
ing facto
r
Etc
…
Process Modeling
Optim
al allo
cation o
f Ele
ctric
al energ
y
•Pro
ble
m–
Energ
y in s
teel pla
nt is
a s
carc
e a
nd e
xpensiv
e resourc
e.
–Typic
ally
easte
rn p
art o
f In
dia
the g
ap b
etw
een the s
upply
and d
em
and
is v
ery
hig
h.
–In
ste
el pla
nt som
e o
f th
e u
nits a
re c
onsid
ere
d a
s e
ssential pow
er
consum
ing u
nits ( s
uch a
s B
F C
O e
tc..) re
quires fix
ed u
nits o
f pow
er.
In case of shortages in the power supply, how the consumption to
be distributed most profitably.
•M
odel O
bje
ctive
–M
axim
izin
g P
rofit contrib
ution
–M
inim
ize the c
ost of pro
duction
–M
axim
ize the T
hro
ughput.
•M
odel used
–Lin
ear Pro
gra
mm
ing w
ith P
ow
er as a
lim
itin
g c
onstrain
t.
Edelman Recognized
Online Slab Temperature Calculation and Control
Problem:
A m
odel to
pre
dic
t the h
eat transfe
r from
the reheating furn
ace
atm
osphere
to
the s
urface o
f th
e s
lab a
nd fro
m the s
lab s
urfaces to the c
ore
of th
e s
lab.
Heating p
rofile
of th
e reheating furn
ace d
epends o
n
•Targ
et dro
p o
ut te
mpera
ture
•Entry tem
pera
ture
•D
iffe
rent th
erm
al pro
pertie
s for diffe
rent gra
des o
f ste
el
•O
pera
tional fluctu
ation in the d
ow
n s
team
pro
cesses
Model Objective:
•U
nders
tand the therm
al m
odel
•U
nders
tand the o
pera
tional re
quirem
ents
•C
alib
rate
the m
odel para
mete
rs
Model Used:
•M
ath
em
atical fu
nctions u
sin
g b
asic
heat transfe
r equations.
Assumptions:
�H
eat transfe
r m
ode is p
rim
arily
radia
tive.
�In
sula
ting e
ffect on h
eat conduction d
ue to s
cale
form
ation is n
eglig
ible
�H
eat transfe
r in
the s
lab is o
nly
fro
m the top a
nd b
ottom
surfaces, m
ode b
ein
g
conduction
�Shadow
ing e
ffect due to s
kid
layout is
sm
all
�Tem
pera
ture
thro
ughout th
e h
orizonta
l pla
n in the s
lab is u
niform
.
Contd..
Online Slab Temperature Calculation and Control
Basic parameters:
Reheating furn
ace o
f w
alk
ing b
eam
type o
f effective therm
al le
ngth
39 m
ete
rs
with m
ax o
pera
ting tem
p 1
350
0C
.
The furn
ace d
ivid
ed into
12 o
pera
ting z
one
-Recupera
tive (to
p a
nd b
ottom
) -P
reheating ( top a
nd b
ottom
)
-Heating (to
p a
nd b
ottom
)-P
resoakin
g (to
p a
nd b
ottom
)
-Soakin
g rig
ht (top a
nd b
ottom
)-S
oakin
g left (to
p a
nd b
ottom
)
Calc
ula
ted n
odal poin
t te
mpera
ture
was e
xtrapola
ted thro
ughout th
e furn
ace
length
to p
rovid
e tw
o c
ontinuous radia
tion c
urv
es
-Alo
ng the c
eili
ng
-Alo
ng the flo
or
This
continuous tem
pera
ture
pro
file
used a
s h
igher te
mpera
ture
and the s
lab
surface a
s a
low
er te
mpera
ture
in the S
tefa
n-Boltzm
an’s
radia
tion e
quation.
This
calc
ula
tes the h
eat flux d
ensity o
f th
e s
lab.
Online Slab Temperature Calculation and Control
Radiativeheat Transfer from the furnace to the slab surface
�N
odal poin
t at each Z
one ( m
axim
um
tem
pera
ture
) w
ere
identified
�M
easure
d tem
pera
ture
by the T
Cs
neare
st to
the n
odal poin
ts u
sed a
s p
rim
ary
ra
dia
tion d
ata
.
�An o
ffset ( fixed a
nd v
ariable
) w
as a
dded to the m
easure
d tem
pera
ture
for
absorb
ing the e
ffect of th
e convective h
eat transfe
r
•Fix
ed p
art-Additio
nal heat transfe
r fo
r th
e s
tandard
fuel flow
at each z
one
•Variable
part-D
evia
tion in c
onvective h
eat transfe
r due to c
hanges in fuel
flow
w.r.t. sta
ndard
.
PI control for achieving required slab temperature
�C
alc
ula
ted s
lab tem
pera
ture
fro
m the 7
layers
then p
rocessed a
savera
ge u
pper
half a
nd a
vera
ge low
er half tem
pera
ture
and c
om
pare
d w
ith the respective s
lab
targ
et te
mpera
ture
.
�The d
evia
tions w
ere
avera
ged for all
the s
labs in a
zone.
�This
was u
sed a
s feed b
ack to P
I control lo
op w
hic
h c
alc
ula
tes the tem
pera
ture
set poin
t fo
r all
Level 1 s
yste
m therm
ocouple
s
�O
ut put dow
n loaded to L
evel 1 s
yste
m
�This
dedic
ate
dly
tries to m
ain
tain
the s
et poin
t by regula
ting the fuel flow
rate
.
Online Slab Temperature Calculation and Control
Conductive heat Transfer from the slab surface to the core of the slab
�The m
odel used o
ne d
imensio
nal Fourier’s h
eat conduction e
quation.
�The tem
pera
ture
and h
eat flux e
quations w
ere
solv
ed u
sin
g fin
ite
ele
ment
techniq
ue
�Sla
b thic
kness w
as d
ivid
ed into
7 layers
each a
t 35m
m a
part.
�Top a
nd b
ottom
surface tem
pera
ture
calc
ula
ted fro
m the radia
tion
equations
was u
sed a
s b
oundary
valu
e.
�Physic
al pro
pertie
s ( s
uch a
s therm
al conductivity, specific
heat etc
..) of th
e
pla
in c
arb
on s
teel w
ere
taken fro
m s
tandard
table
.
�In
terv
al betw
een s
uccessiv
e c
alc
ula
tions w
as taken o
ne m
inute
s.
Logistics Optimization
Cost optimized executable dispatch plan
•Problem
–Tra
nsportation o
f th
e fin
ished g
ood in s
teel in
dustry is
a c
om
ple
x a
nd e
xpensiv
e
opera
tion.
–Typic
ally
easte
rn p
art o
f In
dia
the infrastructu
re is n
ot fa
voura
ble
for th
e
industrie
s.
–R
esourc
es s
uch a
s tra
nsportation m
odes a
re n
ot w
ell
org
anis
ed.
–M
ost of our m
ovem
ents
are
aw
ay fro
m E
aste
rn z
one.
How to use the resources optimally at the lowest cost
•Model Objective
–M
inim
ize the c
ost of finis
hed g
oods m
ovem
ent per to
n.
–M
inim
ize the c
ongestions o
f finis
hed g
oods a
t each n
odes.
–M
axim
ize the m
ate
rial m
ovem
ent.
–Pre
dic
t budget allo
cation requirem
ent.
•Model used
–Lin
ear Pro
gra
mm
ing.
Mother Mill
C u s t o m e r
St.Yd/Hub
EPA
Rail
Road
Container
Products
HR
/ C
R c
oils
/ Sheets
& P
late
s /
Sla
bs
WR
coils
, R
ebars
, Billets
,
Transport
Mode &
Costs
Routes
& Costs
Rail siding
PARAMETERS :
Cost optimized executable dispatch plan
Annual Sales Plan
Product group/ volume/
destination Detailed Annual Business plan
Product group/ Volume / Destination
Total Delivery System Cost
Product group / Route /
Destination / Mode / Cost
data
Transport Capacity
Constraints
Annual Shutdown Plan
Each major resources
L P model
for Cost Optimization
Rework with
Shipment
execution
constraints
Matrix
develo
ped o
n
XL
spre
adsheet.
Cost-optimized & executable Dispatch Plan
Product group / route / mode / destination / volume / cost
Capacity Balancing
•Limited abilityto
proactively fulfill
rapidly-changing
customer demand.
•Infeasible supply plans
due to inability to model
all relevant supply chain
bottlenecks
•Low asset utilization
due to inability to get a
single global view into
entire supply chain
•Sub-optimal costs and
lost salesdue to limited
analytics and
optimization
•Inability to allocate
limited product supply
through sales channels
Problem
•Maximize customer
responsivenessby
integrating supply planning
to strategic planning and
forecasting processes
•Gain global visibilityinto
extended supply chain by
modeling all relevant
production and distribution
bottlenecks.
•Increase feasibilityof
detailed plan by
representing them in global
supply plans.
•Simultaneously optimize
entire supply chain in a real
world fashion by
considering real costs and
multiple goals
•Allocate limited supply as
ATP reservations
Solutions
•Reduced planning
cycle time
•Generation of feasible
plan
•Higher service level
for key customers
•Reduced Inventory
costs
•Reduced
obsolescence cost
•Improved profitability
by allocating most
profitable customer
Benefits
What does the Capacity balancing do to the business ?
Order Planner -Purpose
•The p
rim
ary
goal of O
rder Pla
nner is
to develop a feasible production plan
for
satisfy
ing d
em
and requirem
ents
pla
ced o
n the O
rder (i.e
., c
usto
mer / fo
recast
ord
ers
) respecting the due date promised
•The p
roduction p
lan is d
evelo
ped w
ith the s
imultaneous c
onsid
era
tion o
f th
e
follo
win
g c
onstrain
t ty
pes:
–demand constraints
: due d
ate
s a
nd their tole
rances
–material constraints
: m
ate
rials
availa
bility
(quantities a
nd tim
ing) –
used for
outs
ourc
ed R
M
–capacity constraints
: re
sourc
es a
vaila
bility
(quantities a
nd tim
ing)
•O
rder Pla
nner is
an inte
lligent decision support system
for short-range a
nd
mediu
m-r
ange p
roduction p
lannin
g.
An unsolved problem
Already Existing Body of Work
1.
1.
Modelin
g a
nd S
olu
tion A
ppro
aches to P
ractical Schedulin
g P
roble
mM
odelin
g a
nd S
olu
tion A
ppro
aches to P
ractical Schedulin
g P
roble
ms in
s in
Iron &
Ste
el In
dustry b
y
Iron &
Ste
el In
dustry b
y Dr. Lixin Tang
Dr. Lixin Tang, , Dr. Jiyin Liu
Dr. Jiyin Liu
and
and Professor Zihou
Professor Zihou
Yang
Yang
2.
2.
Ste
el
Ste
el --m
akin
g p
rocess s
chedulin
g u
sin
g
makin
g p
rocess s
chedulin
g u
sin
g L
agra
ngia
nLagra
ngia
nre
laxation b
y
rela
xation b
y Lixin
Lixin
Tang
Tang, Peter B.
Peter B. Luh
Luh, , Jiyin
JiyinLiu and Lei Fang
Liu and Lei Fang
3.
3.
A S
urv
ey o
fA S
urv
ey o
fM
ath
em
atical Pro
gra
mm
ing A
pplic
ations
Math
em
atical Pro
gra
mm
ing A
pplic
ations
in Inte
gra
ted S
teel
in Inte
gra
ted S
teel
Pla
nts
Pla
nts
by
by Goutam
GoutamDutta
Dutta
and
and Robert
Robert Fourer
Fourer
4.
4.
Utilit
y a
nd S
tability
Measure
s for Agent
Utilit
y a
nd S
tability
Measure
s for Agent --Based D
ynam
ic S
chedulin
g o
f Based D
ynam
ic S
chedulin
g o
f
Ste
el C
ontinuous C
asting b
y
Ste
el C
ontinuous C
asting b
y D.
D. Ouelhadj
Ouelhadj , , P. I. Cowling
P. I. Cowling, , S.
S. Petrovic
Petrovic
And there
are
many m
ore
math
em
atical m
odels
develo
ped o
n this
su
And there
are
many m
ore
math
em
atical m
odels
develo
ped o
n this
subje
ct
bje
ct
Already Existing Software Solutions From…
1.
1.
Sie
mens
Sie
mens
2.
2.
Rockw
ell
Auto
mation
Rockw
ell
Auto
mation
3.
3.
IBM
IBM
4.
4.
Bro
ner
Bro
ner
5.
5.
i2i2
6.
6.
PO
SD
ATA
PO
SD
ATA
7.
7.
OM
Partners
OM
Partners
Leavin
g a
sid
e m
any h
om
egro
wn s
yste
ms
Leavin
g a
sid
e m
any h
om
egro
wn s
yste
ms
…….B
ut, th
e d
epth
and b
readth
to c
over fo
r th
e v
ariability
of ste
.But, th
e d
epth
and b
readth
to c
over fo
r th
e v
ariability
of ste
el pro
duction
el pro
duction
as w
ell
as a
ll th
e p
roduction s
teps rem
ain
ed a
question
as w
ell
as a
ll th
e p
roduction s
teps rem
ain
ed a
question ……
....
For in
sta
nce, if c
aste
r and H
SM
are
synchro
niz
ed, th
en the
impact on L
H a
nd R
F a
re ignore
d
How
should
the s
chedulin
g logic
change intra-d
ay if th
ere
is a
change in h
ot m
eta
l qualit
y
Re-ladle
Re-ladle
Vessel
#1
Vessel
#2
Vessel
#3
OLP
OLP
OLP
RH
LF #1
LF #2
Caster#1
Caster#2
Caster#3
Scarfing
Grinding
Inspection
Testing
Direct
For Hot Rolling
Melt Shop Scheduling
Edging
t +
Melt Shop Scheduling
Custo
mers
Ord
ers
Desig
n c
oils
D
esig
n S
labs
Rolling S
ch
Cre
ation
Day R
olling p
lan
Melt s
hop
Capacity
Logis
tics
constrain
ts
Melt s
hop
Constrain
ts
Yes
No
Day S
teel pla
n
Melt s
hop
schedule
LF/R
H a
vaila
bility
Tundis
hLoad
No o
f w
idth
Caste
r health
etc
Level -1
Flo
w
Melt Shop Scheduling
Ste
el pla
n d
epends o
n:
a)D
ow
n s
tream
dem
and b
ased o
n O
rder due
date
b)P
roduction c
onstrain
ts
a)
Technic
al constrain
ts
b)
Shop C
onstrain
ts
c)D
ispatc
h c
onstrain
ts
d)M
axim
ize c
apacity u
tiliz
ation a
t each
resourc
es
e)W
IP reduction
Melt Shop Scheduling
Challenges
Critical Success Factors
••A large % of such initiatives don
A large % of such initiatives don’’ t see the
t see the
light at the end of the tunnel
light at the end of the tunnel ……
……..
….. Due to non-technical reasons
Because
“Technical problems are the easier ones to solve... The soft
issues are always the hardest.”
What Kind of Project is this???
•IT
?
•R
& D
?
•Pro
cess c
hanges ?
•Perform
ance Im
pro
vem
ent?
•Busin
ess S
olu
tion that w
ill s
hape the futu
re ?
Who’s Project is this???
Source: Dr Brian Eck
Low
Hig
h
LowHigh
Understanding of Business
Depth in thinking
Academia
Business
Practitioner
Colla
bora
tion
Business Vision
•C
lear unders
tandin
g o
f
pro
ble
m d
efinitio
n
•Alig
nm
ent w
ith b
usin
ess
strate
gy
•Id
entify
and a
gre
e o
n
direction
•D
efine c
ritical success
facto
rs for th
e b
usin
ess
and p
roje
ct
•U
nify indiv
idual vis
ions
Organizational Issues
•Busin
ess v
isio
n, strate
gy
•Top M
anagem
ent
com
mitm
ent
•C
hange M
anagem
ent
•Scope m
anagem
ent
•U
ser expecta
tions
•R
eportin
g issues
•Tra
inin
g
•Im
ple
menta
tion tim
ing
Manage Expectations
•Expecta
tions, w
hile
necessarily
am
bitio
us,
must be g
rounded in realit
y.
Perform
ance
Perform
ance
Expectations
Expectations
ROI
ROI
Flexibility
Flexibility
Compliance
“I haveto do it this
new way”
Reaction
“I will react to this change if I must”
Avoidance
“I must avoid this change”
Negative perception
“I feel threatened by this
change”
Positive perception
“I see the opportunity in this
change”
Experimenting and Testing
“I will experiment with this change -and
check that it really works”
Action
“I will act to achieve this change”
Commitment
“I wantto do it this
new way”
Awareness
“I am being told about
something”
Understanding
“I know what will change and why”
Engagement
“I see the implications for me/us”
Change can be achieved through
commitment or compliance