Modeling and Optimization Tools for Solving Pd...
60
Modeling and Optimization Tools for Solving P d ti Pl i dS h d li P bl Production Planning and Scheduling Problems Alkis Vazacopoulos CAPD Meeting Carnegie Mellon University Pittsburgh PA March 2011 1
Transcript of Modeling and Optimization Tools for Solving Pd...
Modeling and Optimization Tools for Solving P d ti Pl i d S h d li P blProduction Planning and Scheduling Problems
Alkis Vazacopoulos
CAPD MeetingCarnegie Mellon UniversityPittsburgh PA
March 2011
11
Production Planning and Scheduling Problems
» Strategic» Tactical » Operational
2
Software Companies/Academic Instit/Consulting Companies
» SAP» Oracle» Aspentech» Honeywelly» KBC» Consulting companiesg
» IBM-CPLEX, Gurobi, FICO, GAMS, …..
3
Technologies
» LP » MIP» NLP» CP» Modeling» Platforms » Business Rules
4
LP Performance across releases (FICO Xpress)
LP Performance
25000 800
20000
5000
785
790
795
800
15000
otal
Tim
e(s)
775
780
785
mbe
r Sol
ved
Total TimeNumber Solved
5000
10000To
760
765
770 Num
02003B 2004B 2005B 2006B 2007B 2008A FICO 7.0
Release
750
755
5
Internal test set of 796 public and customer models
LP Performance across releases
LP Performance
25000 800
20000
5000
785
790
795
800
15000
otal
Tim
e(s)
775
780
785
mbe
r Sol
ved
Total TimeNumber Solved
Primal Dual Barrier
5000
10000To
760
765
770 Num
02003B 2004B 2005B 2006B 2007B 2008A FICO 7.0
Release
750
755
6
Internal test set of 796 public and customer models
LP Performance across releases
LP Performance
25000 800
20000
5000
785
790
795
800
Good15000
otal
Tim
e(s)
775
780
785
mbe
r Sol
ved
Total TimeNumber Solved
PrimalGoodDual
Is importantBarrier
5000
10000To
760
765
770 NumIs important
02003B 2004B 2005B 2006B 2007B 2008A FICO 7.0
Release
750
755
7
Internal test set of 796 public and customer models
MIP Performance across releases
250000
300000
270
290
200000
230
250
n Ti
me
lved
100000
150000
210
230
Tota
l Sol
utio
n
Num
ber S
ol
Numbers SolvedTotal Time
50000
100000
170
190
01502003B 2004B 2005B 2006B 2007B 2008A 7.0 7.1
Release
8
Internal test set of 320 public and customer models
MIP Performance across releases
MIP Performance
250000 290
200000
250
270
Heuristics
100000
150000
Tota
l Tim
e(s)
210
230
Num
ber S
olve
d
Total TimeNumber Solved
Branch and Bound
Cuts Classic
problem specific
HeuristicsGeneral and
problem specificparallelize
50000170
190
p p parallelize
02003B 2004B 2005B 2006B 2007B 2008A FICO 7.0
Release
150
9
Feedback Loop
MIP Performance
250000 290
200000
250
270
100000
150000
Tota
l Tim
e(s)
210
230
umbe
r Sol
ved
Total TimeNumber Solved
Reengineer B&B
Decomposition Algorithms –more flexible
Perform Operations -
Node
50000
100000T
170
190
Nu
02003B 2004B 2005B 2006B 2007B 2008A FICO 7.0
Release
150
10
Comparison GUROBI vs CPLEX
» Gurobi: developed to take advantage of multi-core machinesp g» Growing very fast (150 commercial licenses, 25 academic
sales, in addition to the academic program, source: jtoedm comjtoedm.com
» LP 10-30% better than CPLEX (Mittelman website)» MIP 10% - 30% better on optimality» MIP 10% - 30% better on optimality» CPLEX better on Feasibility problems» New MIPLIB data soon» New MIPLIB data soon
11
Open Source
» SKIP – ZIB, Very competitive: used by Dynadec» CBC – used as an entry by many companiesy y y p
» Comparison from Hans Mittelman’s websitep
12
Benchmarking MIP
» Optimality (what about problems won’t solve after 30 minutes, 60 minutes)
» Best feasible solutions» Find k best solutions» Size of the tree» Presolve (reduction of the matrix)
» Develop Sets of Problems » Etc….»
13
Xpress 7 parallel speed ups
Xpress 7 Xpress 7p1 thread
p4 thread Improvement
“Internal” deterministic 25,724 17,447 -47%deterministic“Internal”opportunistic 25,724 13,067 -96%
Coral deterministic 24,484 16,129 -52%
CoralCoralopportunistic 24,484 11,137 -120%
14
mzzv42z: Single/Two Threads Xpress FICO IVE
»1 thread »2 threads
15
»15
Xpress 7 parallel speed ups
Xpress 7 Xpress 7p1 thread
p4 thread Improvement
“Internal” deterministic 25,724 17,447 -47%
Parallel Parallel MIPdeterministic“Internal”opportunistic 25,724 13,067 -96%
Parallel processors
on a Network
Parallel MIP (same
computer)P MIP
Deterministic P-MIP
Coral deterministic 24,484 16,129 -52%
Coral
Network P-MIP
Coralopportunistic 24,484 11,137 -120%
16
Concurrent LP solver
» Efficient LP method parallelization
Primal Dual Barrier Concurrent Overall Speedup
1 thread 19,940 13,459 - -, ,
2 threads 19,940 13,459 6,193 (B+D) 23%
4 threads 19,940 13,459 7,540 7,491 70%
Time(concurrent) ~ min(Time(dual), Time(primal), Time(barrier))
Controls: LPTHREADS and DETERMINISTIC
17
(concurrent is nondeterministic!)
Concurrent LP solver
» Efficient LP method parallelization
Primal Dual Barrier Concurrent Overall Speedup
1 thread 19,940 13,459 - -, ,
2 threads 19,940 13,459 6,193 (B+D) 23%
4 threads 19,940 13,459 7,540 7,491 70%P-Barrier
P-D-BCombinations
Time(concurrent) ~ min(Time(dual), Time(primal), Time(barrier))
Controls: LPTHREADS and DETERMINISTIC
18
(concurrent is nondeterministic!)
Customers want multiple alternative solutionsExample: air04
» Set partitioning problem for airline crew scheduling
» Available from MIPLIB 2003
Default Solver 5-Best Solution Solver
1st S l ti 56137 561371st Solution 56137 56137
2nd Solution 56166 56137
3rd Solution 56307 561373rd Solution 56307 56137
4th Solution 56854 56137
5th Solution 57562 561375th Solution 57562 56137
Total B&B Nodes 141 15 981
Total Time 26 sec 29 sec
19
Total Time 26 sec 29 sec
Customers want multiple alternative solutionsExample: air04
» Set partitioning problem for airline crew scheduling
» Available from MIPLIB 2003
Default Solver 5-Best Solution Solver
1st S l ti 56137 561371st Solution 56137 56137
2nd Solution 56166 56137
3rd Solution 56307 56137
Strategies-Business
RulesRobust -
Strategies
Strategy RefinenementRebalancing
3rd Solution 56307 56137
4th Solution 56854 56137
5th Solution 57562 561375th Solution 57562 56137
Total B&B Nodes 141 15 981
Total Time 26 sec 29 sec
20
Total Time 26 sec 29 sec
Xpress-Tuner: How to Tune (Automatically) an Optimization Problem?
21
Xpress-Tuner:Tuning Process
22
Xpress-Tuner:Detailed Results
23
Xpress-Tuner:Comparing logs
24
Xpress-Tuner:Tuning MIP families
25
Xpress-Tuner:Tuning Methods
26
Benchmarking - Optimality
Problems with less than 60 minutes
Problems with less than 30 seconds
Problems with less than 30
i t
Problems more than 2 hours
Reduce timeParallel MIPOpportunistic MIPminutesSolve in 1/10
of time –examples Google
Reduce timeParallel MIPDeterministic
What is a good solution? Multiple
c MIP
Matrix Gen Deterministic MIP
Multiple SolutionsParallel Opportunistic
Matrix Gen
Modeling
27
Explain Solutions
Languages
MIP Research
» More research on Deterministic Parallel» Predict Size of BB Tree/ Structure
U b l d T /h t B l» Unbalanced Trees/how to Balance
» Solve Smaller problems faster (E Danna – Google)N C t /F ibilit P» New Cuts/Feasibility Pump
» More CP in PresolveH d C t i t / S ft C t i t O ti i BB» Hard Constraints / Soft Constraints – Operations in BB nodes
» Parallelize cuts/Heuristics in root node» How to scale better 4 to 8 cores» Data mining in Parameters – using supervised learning
28
g g p g» **** Adaptive search – change the parameters dynamically
during the Search
Non Linear
» Knitro» CONOPT » Xpress SLP : good for process industry» MINLP CMU-IBM , minlp.orgp g» BONMIN Basic Open Source Nonlinear Mixed Integer
programming
29
Comparisons Mittelman
» Knitro vs IPOPT» 27 better Knitro» IPOPT better in 14 cases
30
Global Optimization other technologies
» Baron» Disjunctive programmingj p g g» Robust» Stochastic
31
Constraint Programming
» ILOG-CP » CHIP» Artelys – Kalis» COMET – Dynadecy» SIMPL (Yunes, Aron, Hooker)
» Benchmarking CP codes????
32
»Example of relaxation for alldifferent
nxxx
j
n,,1
),,(alldiff 1
nx j ,,1»Convex hull relaxation, which is the strongest possible
linear relaxation (JNH, Williams & Yan):
j
n
jj
nJnJJJx
nnx
||with1all)1|(|||
)1(
11
21
Jj
j nJnJJJx ||with ,,1all),1|(|||2
»For n = 4:
1,,,3,3,3,3,3,3
6,6,6,610
4321
434232413121
432431421321
4321
xxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxx
33
4321
»Automatic relaxations
»The following constraints can be relaxed automatically with Xpress‐Kalis :
» linear constraints
alldifferent(X)
occurrence(X,v) opY
distribute(X,A,m,M)
X = Min/Max(Y) X = Min/Max(Y)
absolute value X = | Y |
distance v op |X‐Y|
element(2D) X = A[I] | X = A[I][J]
cycle
34
logical constraints (,,or,and)
An integer knapsack problem with An integer knapsack problem with side constraintside constraints de co st a ts de co st a t
485i 30253subj.to
485min
321
321
xxxxxx
}4,3,2,1{),,(alldiff 321
jxxxx
j
iji
i
ijyx
ll1
all,»»MILP needs additional MILP needs additional constraints and 0constraints and 0--1 1
variables to express the variables to express the j
ij iy all,1variables to express the variables to express the alldiffalldiff::
35
Pure CP model in Xpress-KALISPure CP model in Xpress KALIS
36
HYBRID Model in Xpress-KALISHYBRID Model in Xpress KALIS
37
Modeling Platforms
38
Modeling Languages
» OPL: Integrates MIP and CP» GAMS: All solvers available, NLP, MIP, Stochastic» Mosel : Programming and Modeling together
39
Modeling Languages
» OPL: Integrates MIP and CP» GAMS: All solvers available, NLP, MIP, Stochastic» Mosel : Programming and Modeling together
40
More
» IBM ODM: Optimization Decision Manager» Next generation : Eclipseg p» Simulation » Python linky
41
Eclipse – Platform - Future
42
»CP, MIP and their combination
» Optimization Technologiesp g
» Mixed Integer Programming: MIP» Finite Domain Constraint Programming: CPg g
» Planning & Scheduling »CP » MIP»CP+MIPPlanning & Scheduling
» Long and mid-term planning: MIP» Short-term planning, scheduling: CP
C
S o t te p a g, sc edu g C
Supply chain optimizationRequires MIP, CP, and their combination
» Time Horizon
43
q , ,
Solving Planning and Scheduling Models
»INTEGRATED ENVIRONMENT
C Pl i d S h d li
»INTEGRATED ENVIRONMENT
»Common Planning and Scheduling Model
»LP/MIP »CP»LP/MIP »CP
44
»Combining MIP & CP
»Master problem
MIP
»Sub problem
»MIP
»Sub problem»CP
»Feasible»(OK)
»Infeasible
»Add Cut to MIP
» Mixed logical/Linear Programming framework
»(OK) »Add Cut to MIP
45
ed og ca / ea og a g a e o(Hooker et al)
Project
» LISCOS Project» European Union 5th Framework
BASF Ch i l» BASF - Chemicals» P&G - Food
PSA C» PSA - Cars» Barbot - Paint
46
Production Scheduling Problem
» Schedules Crude-Oil Blend-Shops in Oil-Refineries.» In the Category of “Closed-Shop” Scheduling.g y p g» Involves Quantity, Logic and Quality Decisions.
» Quantity = Flow, Rate, Inventory» Logic = Mixing-Delay, Up-Time, One-Flow-In» Quality = Octane, Sulphur, Yield, Temperature
» Intended to Reduce Quality Variation» Intended to Reduce Quality Variation.Key Planning Proxy
Absolute Maximum Limit
Variability Reduced
Target Set Closer to Limit
TargetLow
47»
TimeManual Scheduling
Crude-Oil Blend Scheduling Installed
Production Scheduling Problem
» Finite-Capacity Scheduling.» Supports Marine and Pipeline Access Blend-Shops.pp p p
Tank1 SourCBH1
PipelineNorth
DayTank1 Pipestill1Tank2 SourCBH2
Tank3SourCBH3
Pipestill2DayTank2
Tank4
Tank5 SweetCBH
PipelineSouth
48
Tank6
Crude-OilBlend-Header
Production Scheduling Technology
» Involves the following OR Problems:» Fixed-Charge Network Flow» Lot-Sizing and Scheduling (Small Time-Bucket)» Facility-Location» Multi-Linear Pooling
» Uses Discrete-Time Formulation.» Allows for all Logic Constraints to be Configurable.» Creates Elastic/Penalty Variables for all Constraints.» Modeled using Xpress-Mosel.
49»
Production Scheduling Technology
» Involves the following Primal Heuristics programmed using Xpress-Mosel:
R l d Fi» Relax-and-Fix» Dive-and-Fix» Fixed-Charge Adjustment» Chronological Decomposition» Smooth-and-Dive» Incumbent Elimination» Disaggregation and Warm-Start» Parallel Diving (Sorted One-Point Crossover)» Local Search Repair of Infeasible Solutions» Local Search Repair of Infeasible Solutions
» Embeds Xpress-Optimizer and Xpress-SLP.
50»
Production Scheduling Approach
» Finds Optimized or Feasible Schedules.» Maximizes Profit and Performance.
» At the core is Bender’s Decomposition.p» Uses Intelligent Problem Solving structure:
» Solve Quantity-Logic (Q-L) first using MILP» Fix Logic, then solve for Quantity-Quality (Q-Q) using SLP» If Q-L infeasible then Q-Q will be infeasible» If Q-L feasible then Q-Q may or may not be feasible» (If Q-Q feasible then Q-L feasible)
51»
Production Scheduler Demo
52
Existing Products with Xpress-MP
» Finished Product Blend Planning (BLEND) uses Xpress-Optimizer.
» Refinery and Petrochemical Planning (RPMS) uses Xpress-Optimizer.
» Supply and Distribution Planning (SAND) to use Xpress» Supply and Distribution Planning (SAND) to use Xpress-Optimizer.
53»
Examples of Production Systems
5454
Petroleum Refining
» Includes many tanks, atmospheric & vacuum fractionators, blenders, reactors, separators, etc. as well as 1100 components and 50 propertiescomponents and 50 properties.» Logistics has 160K rows, 95K columns, 11K binaries: 12-
seconds.» Quality has 250K L rows 25K NL rows 100K columns: 900-» Quality has 250K L rows, 25K NL rows, 100K columns: 900-
seconds.AU I
AU I In
LPG
LABFS
Gas oil
Gas
ATF
AU II
LPG
Gas
ATF
Naphtha
Naphtha
SG
NGOut
SG
NGOut
LS_BH
Out
In
In
In
In
Out
Out
Out
Out
In Out
SG Crude
TK 517
TK 518
TK 519
TK 520
TK 781
Gas
LPG
Naphtha
Gas
LPG
Naphtha
Gas
LPG
LN
Gas
LPG
LN
OutIn
SG Block
Out
Unit
In
AU I RCO
Tank
Blending
RCO
AU I Mix
LS_BH
In Out
SG Group
In Out
NG Group
TK 85
TK 86
TK 90
TK 91
TK 92
ATFRail
In
In
In
In
In
Out
Out
Out
Out
OutATF Blend
AU I ATF
Out
HCU Kero
AU II ATF
TT
ATF
ATF
In
Naphtha
DHDS Diesel
DHDS Feed
AU IV HGO
AU V GO
AU IV LGO
Out
FPU I HD
FPU II HD
LABFS
Gas oil
LABFS
Gas oil
HN
Kero
Gas oil
HN
KAPL
KNPL
KAPL Header
Out
In
In
In
HSD Euro III
SKO
HSD BS II
OutHSD Euro III
SKO
HSD BS II
Naphtha Blend
DHDS Naphtha MTO Blend
AU I ATF
AU II ATF
Out
TK 108,109
In Out
MTOTT
U IV
LG
O
VB GO
VB Nap
InArmy_Navy_FO
IV Kero
In SKO_LABFS
HSD
FO
ATF
I ATF
In
I LABFS
In
I GO
In
II ATF
In
II LABFS
In
II GO
In
III HN
In
III Kero
In
III GO
In
IV HN
In
AU I Streams
AU I Streams
AU II Streams
AU IV Streams
GHC Area
FPU I
In
HDiesel
Unit
VR
Vac Slop
In
HDiesel
VR
Vac Slop
VDU
In
HDiesel
VR
Vac Slop
In Out
TK 821
In Out
TK 822
VBUVR ex 821
VB Tar
VB Naphtha
VB Gas oil
AU II In
LABFS
Gas oil
AU III
AU III In
LPG
HN
Kero
Gas oil
Gas
AU IV
LN
AU II Mix
AU III Mix
NG
HS
Out
_
LS_BH
HS
Out In
In
In
Out
Out
Out
In
In
Out
Out
In Out
NG Crude
TK 782
TK 783
TK 911
TK 703
TK 704
TK 771
Out
VDU feed
AU V RCO
Cold RCO ex GRE
AU IV RCO
Out
FPU II feed
AU IV RCO
Cold RCO ex GHC
AU III RCO
AU II RCO
AU I RCO
Out
FPU I feed
Cold RCO ex GRSPF
AU IV RCO
AU III RCO
In Out
TK 803,804
In Out
TK 453,454
FCC feedFCC In
HCU feedHCU In
Gas
LPG
LN
HCU StreamsHCB
FCC Feed blend
Cold VGO ex GRSPF
Hot feed
Out
HCU Feed blend
Cold VGO ex GHC
Hot feed
Out
FPU II
GRSPF Area
OutIn
NG Block
OutIn
BH Block
Unit
In
AU III RCO
Tank
Blending
RCO
Unit
In
AU II RCO
Tank
Blending
RCO
GRSPF Area
VGO I
VGO
VGO
VGO
Cold VGO ex GHC
In
UnitIn
OutVGO
VGO III
HD
FPU II VR
FO_LSHS In
AU II RCO
RCO
VBU Feed
Out In
VBF
BIT
FPU I VR
FO_LSHS In
VBF
BIT
VDU VR
Others In
BIT
Feed
In
LCO
CLO
Tank
VGO IITank
VR ex 822
InFO
VBF
FCC Streams
HN
Feed
Kero
DieselHCU
HN
In
In Out
Grouped BH
DHDSIn Out
TK 823, 824
VDU HD
FCC LCO
HN
Gas oil
KSPL
KSPL Header
FPU I HD
FPU II HD
VDU HD
VDU Slop
TK 723
TK 724
TK 725
TK 731
TK 726
In
In
In
In
In
Out
Out
Out
Out
OutHSD BS II Header
AU I
GO
AU II
I GO
AU
IV H
GO
HC
U H
N
DH
DS
Die
sel
Out
HSD
Rail-BS II
TK 97
TK 99
TK 732
In
In
In
In
Out
Out
Out
Out
TK 906
TK 907
TK 910
TK 909
In
In
In
In
Out
Out
Out
Out
AU V
GO
AU II
GO
HC
U K
ero
HC
U D
iese
l
SRGO or LDOSRGO Rail
Euro III HSD Header
AU
I G
O
AU II
I GO
AU IV
GO
HC
U H
N
DH
DS
Die
sel
FPU
I H
D
Out
AU V
GO
AU
II G
OH
CU
Ker
o
HC
U D
iese
l
Army HSD Header
AU I GO
Out
AU II GO
AU V
HN
AU V
HN
K9 B
otto
mK9
Bot
tom
HPCL_BPCL PL
LAB
LMW
Army or LSHF BlendAU IV HGO
FPU I HD Out
LAB RB
VB GO
FPU II HD
FCC LCO
Army or LSHF Diesel
TK 730
In Out
HMW
FCC
LC
O
FCC
LC
O
AU II
I Ker
o
AU
IV K
ero
AU
I LA
BFS
AU II
LA
BFS
AU
III K
ero
AU IV
Ker
o
Rail
FPU I VR
FPU II VR
VDU VR
AU
AU I
ATF
AU
II A
TF
AU
III H
N
AU
IV H
N
AU
V K
ero
AU
IV L
GO
FPU
I H
D
FPU
II H
D
FPU
II H
D
AU
IV H
N
AU II
I HN
AU
I LA
BFS
AU II
LAB
FS
AU
I A
TF
AU II
ATF
AU V
Ker
o
FCC
HN
AU
IV L
GO
IV HGO
InArmy_Navy_FO
DHDS_HSD
IV LGODHDS_HSD
V Kero
InSKO_LABFS
HSD
700 TK Rail header
InOut
900 TK Rail header
InOut
IOTL_KSPL Header
InOut
TK 727,728,729
In Out
IOTL
PFRCBTT
AU V GO
FCC
HN
Army Diesel
AU IV LGO
OutAU IV HGO
Rail-Euro III
LDOTT
KRPLPL
VDU HD SRGO
KDPLPL
LDO Rail
V HN
In
V GO
In
II HD
In
II VR
In
I HD
In
I VR
In
HD
In
Slop
In
In
AU V Streams
In Out
TK 773
In Out
TK 705
In Out
TK 706
In Out
TK 707
In Out
TK 801
In Out
TK 802
In Out
TK 708,709,710,774
BBU
Cold VR ex GRE
TK 741
TK 742
TK 743
TK 747
TK 746
In
In
In
In
In
Out
Out
Out
TK 775
TK 744
TK 745
TK 749
TK 748
In
In
In
In
In
Out
Out
Out
Out
Out
Out
Out
Out
Bitumen60-70
80-100
Out
Bitumen Blend
BBU Prdt
VDU VR
FPU II VR
IFOIn
LSHS-Others Blend
FCC LCO
AU III RCO
VB Gas oil
FPU HD TK 101
TK 302
In
In Out
TK 102
TK 103
TK 734
In
In
In
In
Out
Out
O t
Out
Out
Out
LSHS-OthersRail
AU IV In
LPG
LN
Kero
LGO
Gas
AU V
AU V In
LPG
LN
Kero
Gas oil
Gas
HN
HGO
HN
AU IV Mix
SG
HS
Out
Out
LS
BH
AU V Mix
SG
HS
LS
BH
Out
In
In
In
Out
Out
Out
In
In
In
In
Out
Out
Out
Out
Vadinar Crudes
TK 701
TK 702
TK 772
TK 770
TK 901
TK 902
TK 903
Hot VR
Out
IFO Blend
AU III RCO
AU I RCO
AU II RCO
Vac Slop
TT
Vac Slop
FPU I_II VR
GRE Area
GRSPF Area
OutIn
LS Block
OutIn
HS Block
Unit
In
AU IV RCOTank
RCO GRE Area Cold VR ex GRE
LGO-HGO Swing
LS RCO ex GRE
30-40
AU IV HGO
VDU VR
LGO
OutIn
HGO
Out
In
VB Tar
FPU I VR
AU IV LGO
FPU I VGO
VDU VR Split
VBF
VRT
GRE RCO router
In Out
LSHS Group
In Out
TK 60-70
In Out
TK 80-100
In Out
TK 30-40
In Out
VD
U S
lop
FO Bl d
CLO
CLO
HN
VB Nap
VB Gas oil
TK 100
In
Naphtha re-run Streams
K9 Bottom
TK 205, 206
In Out
LABFS-NIRMA
TK 207, 208
In Out
PL
LABFS Blend
AU I ATF
AU II ATF Out
AU V Kero
TK 715
TK 716
TK 717
TK 718
TK 914
In
In
In
In
In
Out
Out
Out
Out
OutTK 739
TK 738
In
In
Out
Out
SKORail
SKO Blend
AU III and AU IV Kero Out
AU V Kero
HCU Kero
Back end (LAB Unit)
Kero
LAB
LAB RS
HABLABRS
In Out
TK 767,87,88(LAB LMW)
In Out
TK 557,558(HAB)
Benzene
In Out
TK 514
Front end (Rerun and Molex)
N-P
N-P
NIRMA RS
Out
LAB RB
Dumad PL
In Out
Dumad PL
BPCL_HPCL PL
Heavy AlkylateTT
VB Tar
In Out
TK 89,98(LAB HMW)
FPU
I V
R
LAB WN
In
LDO
FPU
II V
R
VD
U V
R
AU I LABFS
AU II ATF
AU II LABFS
AU I LABFS
AU II LABFS
LABFS WN
ASOJ PLTK 401,402,406 (PFRCB)
In OutFO Cutter Stock Blender
FPU I HD
AU IV Kero
VB Gas oil
LAB RB
Out
FCC LCO
AU IV LGO
AU IV HGO
Tk 95
In Out
Tk 96
In Out
In Out
HCU Kero
InATF
SKO_HSD
LCO
Army or Navy Diesel
DHDS_HSD
FO
In
AU IV Kero header
AU III Kero
AU IV Kero
Out
AU III and AU IV Kero
AU I ATF
HN
LAB RB
SRGO Header
AU V GO
LAB Section
RILPL In
FO
FO
GR
E R
CO
VR
GRE RCO
In
VB Nap
In
VB Gas Oil
In
In
HCU HN
In
HN
HCU Diesel
In
FCC HN
In
FCC CLO
In
55
GHC Area
TK 802
In
TK 451
In Out
TK 452
Out
VB Tar
LSHS-TEC Blend
AU III RCO
FPU II VR
AU I RCO
AU II RCO
TK 93,94
TK 301
In
In
Out
TK 218
TK 777
In
In Out
Out
Out
LSHS-TEC
Rail
TK 903
In Out
TK 904
Out
Unit
In
AU V RCOTank
RCO FCC CLO
FPU II HD
NG crude tanks
SG Crude tanks
LS Crude tanks
HS Crude tanks
BH Crude tanks
LSHS-TEC gradeLSHS-Other grades Bitumen 30-40
Bitumen 60-70 Bitumen 80-100
VR tanksRCO tanks VGO tanks
Bitumen - Total
Tank2
FO Blend Out
VB
U T
ar
FOTT
Rail
Benzene
To LABTK 303,304
In Out
Tk 733
In Out
FPU II HD
Tk 219
In Out
Tk 220HAB
Cut
ter S
tock
ATF Tanks MTO Tanks
SKO Tanks FO TanksHSD BS IIHSD Euro III
LABFS NirmaTK 823,824
SRGO or LDO Tanks
VB Tar
Fast-Moving Consumer Goods
» Includes an highly sequence-dependent edible oil-deodorizer (prize-collecting traveling salesman problem) and many types of edible oil customer demands.» Logistics-only has 130K rows, 57K columns, 8K binaries: 400-seconds.
56
Oil & Gas Processing
» Includes spheroids, spheres, flash drums, stabilizers, vaporizers, splitters, etc.» Logistics has 60K rows 30K columns 9K binaries: 15-seconds» Logistics has 60K rows, 30K columns, 9K binaries: 15 seconds.» Quality has : 70K L rows, 5K NL rows, 24K columns: 100-seconds.
57
Metals Casting
» Includes arcing-furnaces, decarburizers, cranes, ladles, batch and continuous casters – similar to open-shop/job-shop scheduling.» Logistics-only has 75K rows 24K columns 3 5K binaries: 200-seconds» Logistics only has 75K rows, 24K columns, 3.5K binaries: 200 seconds.
58
Petrochemical Cracking & Separating
» Includes cracking furnaces, compressors, distillation towers, cold-box, tanks, bullets, etc.» Logistics has 21K rows 10K columns 2 5K binaries: 50-seconds» Logistics has 21K rows, 10K columns, 2.5K binaries: 50 seconds.» Quality has 21K L rows, 2K NL rows, 8K columns: 5-seconds.
70FC808Hydrogen
H2 Pipeline
Cracked GasCompressor
Suction
Discharge
Feed
Cold BoxGas
70FC809Feed
70FC805
Tail Gas
Demethanizer
C2 Splitter
Charge
C2
70FC817
Eth Pipeline
Ethylene
70FI818
TG Pipeline
Furnaces
70FI001
RX Effluent
Primary Fractionator
RX Effluent
70FI501
70FI500Deethanizer
70FC805
C3plus
Ch
C3 C3 Splitter
Charge
70FC827
70FI828
Prop Pipeline
PropyleneOffsites [2]C2 Recycle
C3 Recycle
Furnace Fd Header
Feed Charge
Depropanizer
Charge
C4plus
Debutanizer
Charge
70FC830
70FC831
Offsites [1]
Offsites [3]
59Supply
Fuel Oil
Offsites [4]
Feed Naphtha
Mixed Butane
TK105Inlet Outlet
TK106Inlet Outlet
TK107Inlet Outlet
Ethylene Plant (1)
TK-118
TK-119
TK-120
TK-122
TK-123
05F118 05218
05F119
05F120
05F120
05F123
O5F219
05F220
05F222
05F223
C4 Header
Ethylene Plant (2)
NaphthaSupply
RailLoading
Naphtha Header
Inlet Outlet
Inlet Outlet
Loading
Ethylene Plant (3)
Ethylene Plant (4)
TK121Inlet Outlet
Pyrolysis GasolineJetty
Steam & Power Generating
» Includes boilers, turbo-generators, steam headers, turbines, relief-values, etc.» Logistics-only has 24K rows 13K columns 2K binaries: 30-seconds» Logistics only has 24K rows, 13K columns, 2K binaries: 30 seconds.
60
