430 likes | 1.17k Vues
OPTIMIZING HELICOPTER TRANSPORT OF OIL RIG CREWS AT PETROBRAS. Ayberk Göksenin ÜLKER Samet AKÇA Feyza KESKİN. OUTLINE. Introduction Current Process Problem Definition Related Work Project Description Model Solution Methodology Evaluation and Benefits Additional Examples and Comparison
E N D
OPTIMIZING HELICOPTER TRANSPORT OF OIL RIG CREWS AT PETROBRAS Ayberk Göksenin ÜLKER Samet AKÇA Feyza KESKİN
OUTLINE • Introduction • Current Process • Problem Definition • Related Work • Project Description • Model • Solution Methodology • Evaluation and Benefits • Additional Examples and Comparison • Questioning and Conclusion • References
INTRODUCTION • Petrobras • Founded in 1953 • Major oil producer of Brazil • Under management of government • 25,000 workers • Oil production of Brazil: 2 million barrels/day, 13th in world • 90% by Petrobras • Milestone: 1974 - Campos basin explored
CURRENT PROCESS • 80 offshore oil-production platforms • 1,900 workers to be transported by helicopter • Between platforms and 4 mainland bases • 2-weeks shift, 3-weeks rest • Largest non-military helicopter operations
CURRENT PROCESS • Macaé: • 65 daily flights • 33 helicopters • São Tomé: • 30 daily flights • 7 helicopters • Jacarepaguá & Vitória • 15 daily flights • 5 helicopters
CURRENT PROCESS • Flight and passenger assignments done manually, based on • Travel demands • Departure time and destination • Selected from a fixed timetable by passengers • Helicopter availability
PROBLEM DEFINITION • Complexities • Limited number of available helicopters • Strict operational rules • 8 types of helicopter with different • Operational characteristics • Capacity • Cost
PROBLEM DEFINITIONObjectives • Output required each day at each airport including • Flight scheduling • Helicopter routing • Assignment of workers to flights • Output required within 1 hour
PROBLEM DEFINITIONObjectives • Satisfy all demands • Improve safety • Reduce number of landings • Minimize costs • Helps decreasing flight time
PROBLEM DEFINITIONConstraints • Flights start and finish at same base • Max 5 fligths/day for each helicopter • Inspection time between flights • Limited number of landings for each flight • Limited number of legs for each passenger • Limited number of helicopters visiting same platform for each departure time • Lunch stops • Helicopter capacity (determined by route length)
RELATED WORKin Petrobras • Investments in IT to assist manual operation • Attempts to implement a decision support system, by Galvão & Guimarães (1990) • Routes for fixed departure times • Unsuccessful due to worker resistance • Not fully automated, still required manual input
RELATED WORKin Literature • Helicopter-scheduling studies • Timlin & Pulleyblank (1992) • Heuristics, not concerned with time factor • Tjissen (2000) • SDVRP, constant capacity • Hernadvolgyi (2004) • Single helicopter
PROJECT DESCRIPTION • Contract signed with Gapso • Operational version of scheduling system (2005) – 50 weeks • IT functionality (2006) – 6 months • MPROG • 2005 – São Tomé • 2006 – Macaé • 2008 – Vitória & Jacarepaguá • 5 years contract for support and improvement • Training and assistance
MODEL • Billions of variables • NP-hard • Generalization of SDVRP
Solution Method • Column-generation • Network flow formulation assign passengers to previously selected routes, employs heuristics • Which variables to use for a good solution • Challenge: maximum possible number of passengers being picked for each demand adhf= qd or remaining capacity required columns cannot be generated • Solution: Disaggregating demands adhf = 1 if corresponding passenger is on the flight
Solution Method • Column-generation sub-problem • Dual variables: , • Computation of reduced cost of : Determine h and f with minimum and satisfy landing number constratins NP-hard (prize collecting TSP)
Solution Method • Heuristic Procedure • Most departure times in timetable serve small number of platforms • Max 5 landings in each flight • For each departure time and helicopter, seeking profitable flights, with fixed number of landings
Solution Method • Heuristic Procedure • Generate all possible routes with 1 or 2 landings • Generate routes with 3,4 and 5 landings by neighborhood search • For each route, compute • Solve minimum-cost-flow problem to assign passengers • Sum and (optimal value of MCF) to find • Incorporate with negative ’s into the restricted integer program • Stop local search when a column with negative reduced cost is found
Solution Method • MCF network: • Stop nodes: bases & platforms • Demand nodes: passengers • Optimum flow value: = (computed before)
Solution Method • Main Algorithm • Decompose the problem: Generation of flights & assembly • Assembly done by integer programming model
Solution Method • To ease the solution of MIP constraints are relaxed • Equations to ≥ inequalities • Allowing demand to be oversatisfied • Postoptimization:
Evaluation and Benefits • 18% fewer landings • 8% less flight time • 14% reduction in costs • Annual saving: ~ $24 million • Scheduling process improved • In afternoon, schedules of next day can be generated • Time for analysis and adjustments if necessary • Human factor eliminated
Evaluation and Benefits • Before (manual method observation for 354 days): • On 255 days: landings on same platform limit violated • On 202 days: inspection between flights violated • On 212 days: capacity was exceeded • In Macaé savings of $50,000/day estimated, compared to manual schedules. • Safety level increased
Additional Examples and Comparison • Turkey: Hierarchicalanalysis of helicopterlogistics in disasterreliefoperationsby Gülay Barbarosoğlu, Linet Özdamarand Ahmet Çevik • Aim: scheduling helicopter activities in a relief disaster operation • Assigning, scheduling and routing of pilots, flights and helicopters • Mixed Integer Programming was developed with makespan minimization objective
Additional Examples and Comparison • Abroad: Routing helicopters for crew exchanges on off-shore locations (North Sea-Holland) • Aim: determining a flight schedule for helicopters andexchanging crew withminimizing the costof flights. • Determined as Split Delivery Vehicle Routing Problem(SDVRP) • Column generation procedure was used
Conclusion • MPROG is used at Campos basin rigs • Planned integration with flight and passenger control systems • 5 years contract, still in use • Dynamic development and changes required due to variabilities of recent reserve discoveries • 2009 finalist in the Wagner Prize, an INFORMS award for the best cases of practical use of Operational Research
References • http://www.gapso.com.br/en/the-gapso-solution-could-save-us-24-million-per-year-in-aircraft-operations/ • http://www.petrobras.com/en/about-us/