genetic algorithm (production scheduling)

Post on 12-Jan-2017

40 Views

Category:

Engineering

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

Improving the ADACOR2 Supervisor Holon Scheduling Mechanism with Genetic Algorithm

by

(José Barbosa, Paulo Leitão, Emmanuel Adam and Damien Trentesaux)

on PRODUCTION SCHEDULING

presented by

Ay ob a mi A to la g b e (2015 434 4)20 DECEMBER 2016

• Abstract• Introduction• Heuristic algorithm as scheduling

techniques• An evolution manufacturing control

architecture• Experiment & Result• Conclusion

Keywords to note

• Genetic Algorithm: is  a  method  for  solving  both constrained  and  unconstrained  optimization  problems based  on  a  natural  selection  process  that  mimics biological evolution.

• Scheduling• Manufacturing Control

Case Problem• Use of GA techniques to improve the

existing fast, non-optimal scheduling techniques and improvement of the overall system processing execution

Bio-Inspired Algorithms

Ant colony optimization

algorithm (Ant)

Particle Swarm Optimization (Birds & Fish)

Genetic Optimization

(Natural Evolution)

SELECTIONMUTATION

CROSSOVER

Heuristic Algorithm As Scheduling Techniques

An Evolution Manufacture Control ArchitechADACOR & ADACOR2

• PRODUCT HOLONPH • TASK HOLONTH • OPERATION HOLONOH • SUPERVISOR 

HOLONSH

ADACOR2 + GA based Algorithm

(Barbosa, J., Leitão, P., Adam, E., & Trentesaux, D. 2015)

Experiment Test on AIP-PRIMECA Flexible Manufacturing System

(FMS)• Composed of a Shuttle Transport Conveyor System

& 7 work station.• Categorised into Loading/Unloading Station,

Automated Inspection Unit, Skilled operation and Manual Recovery

• Use of different batch sizes (ranging from A0 to F0)

Experiment • A value of 6 was used (Batch sizes)• 6 initially scheduling solution were generated• Algorithm runs iteratively 6 times

Result

(Barbosa, J., Leitão, P., Adam, E., & Trentesaux, D. 2015)

Considering TIME and the OUTPUT RESULT

• GA OVERALL TIME improves the “old” by 24.81%

• Possible increase based on BATCH ORDER increases, GA improvements also rise, giving 34.77% for scenario F0

Conclusion• Experimental results have shown that even with a simple version of the 

GA it is still possible to increase deeply the actual scheduling algorithm.

• Future work will be devoted to incorporate a dedicated scheduling tool in the SH. Tools such IBM ILOG or the Choco API are good candidates for this integration, being the last one fully compliant with Java. With these it is expected to greatly improve the GA calculation speed and the GA results.

THANK YOU

top related