Optimization and Engineering – Louvain-La-Neuve – Wednesday May 24th 2006 Frédéric Bair, ANAST...

Optimization and Engineering – Louvain-La-Neuve – Wednesday May 24th 2006

Frédéric BairFrédéric Bair, ANAST (University of Liege), Belgium, , ANAST (University of Liege), Belgium, F.Bair@ulg.ac.beF.Bair@ulg.ac.beYves LangerYves Langer, n-Side, Belgium, Yves.Langer@ulg.ac.be, n-Side, Belgium, Yves.Langer@ulg.ac.bePhilippe RigoPhilippe Rigo, ANAST (University of Liege), Belgium, , ANAST (University of Liege), Belgium, Ph.Rigo@ulg.ac.bePh.Rigo@ulg.ac.be

Modelling, Simulation and Modelling, Simulation and Optimization of a Optimization of a

Shipbuilding WorkshopShipbuilding Workshop

Optimization and Engineering – Louvain-La-Neuve – Wednesday May 24th 2006

AAAA

Summary Summary ::

Welding Shipbuilding Workshop (ALSTOM, Welding Shipbuilding Workshop (ALSTOM, « Chantiers de l’Atlantique »)« Chantiers de l’Atlantique »)

Software used: eM-Plant (Tecnomatix)Software used: eM-Plant (Tecnomatix)

Treatment of the huge database (Access)Treatment of the huge database (Access)

Very detailed modelVery detailed model

StochasticStochastic

Optimization: Genetic AlgorithmOptimization: Genetic Algorithm

Optimization and Engineering – Louvain-La-Neuve – Wednesday May 24th 2006

Welding Workshop

Optimization and Engineering – Louvain-La-Neuve – Wednesday May 24th 2006

Side ASide A- Assembly 1- Assembly 1- Assembly 2- Assembly 2- Assembly 3- Assembly 3- Assembly 4- Assembly 4- Assembly 5- Assembly 5- …- …- Assembly n-1- Assembly n-1- Assembly n- Assembly n

- …- …- …- …- …- …- …- …- …- …- Assembly m- Assembly m

Kit 1

Kit 2

Kit i

Group 1

Group 2

SideSide B B- - Assembly m+1Assembly m+1

- Assembly m+2- Assembly m+2

- Assembly m+3- Assembly m+3

- Assembly m+4- Assembly m+4

- Assembly m+5- Assembly m+5

- …- …

- …- …

- …- …

- …- …

- …- …

- …- …

- …- …

- …- …Group k

Group k+1

Kit j

Kit j+1

Group k+2

Group l

Optimization and Engineering – Louvain-La-Neuve – Wednesday May 24th 2006

Objective: minimize production timewith following characteristics :

Assemblies belong to fixed Kit

Kits belong to fixed Group

Orders of groups fixed

Can optimize only position of kits among Groups

Sequence optimization

Useful for Genetic Algorithm (GA)

Optimization and Engineering – Louvain-La-Neuve – Wednesday May 24th 2006

2 different sequences (functioning): half-workshop sequence avoid stack up

Half-Cell 1

Half-Cell 2

Half-Cell 3

Optimization and Engineering – Louvain-La-Neuve – Wednesday May 24th 2006

Half-cell sequence better to equilibrate both sides (saturation of welding robot)

Half-Cell 1

Half-Cell 2

Half-Cell 3

Optimization and Engineering – Louvain-La-Neuve – Wednesday May 24th 2006

Operation Time for each Kit : Preparation – Tacking – Welding –

Finishing

Optimization and Engineering – Louvain-La-Neuve – Wednesday May 24th 2006

Other possible results of one simulation:

Occupation Times of welding robots, mechanized gripper, crane bridges, workers

Gantt diagram (for each half-cell)

Total times of each operation (by assembly, by kit, by group)

Storage Times

…

Optimization and Engineering – Louvain-La-Neuve – Wednesday May 24th 2006

Optimization:

Genetic Algorithm : 1 individual by a given sequence

Low variance

Not necessary to multiply simulations

Half-cell sequence better than half-workshop sequence

Optimization and Engineering – Louvain-La-Neuve – Wednesday May 24th 2006

Evolution of the population :

Gain > 8% between a random sequence and best solution

Optimization and Engineering – Louvain-La-Neuve – Wednesday May 24th 2006

Conclusion and perspectives:

Simulation useful even without

optimization

Optimization development Time <

Modelling development Time

Not optimal solution but interesting gains

Automatic Kits creation