Download presentation
Presentation is loading. Please wait.
Published byAvis Blair Modified over 9 years ago
1
Berlin, December 11 th 2012 Faculty of Mechanical Engineering · Chair of Logistics Engineering Network Optimization prior to Dynamic Simulation of AMHS Christian Hammel, Technische Universität Dresden Matthias Schöps, Globalfoundries Dresden
2
Agenda Introduction Network model basics Optimization approach Application areas Case study: Introduction Simulation Results Berlin · Dec 11 th 2012 · Hammel, Schöpsslide 2
3
Berlin · Dec. 11 th, 2012 · Hammel, Schöps 3 Routing in complex AMHS Mainly based on shortest paths Mainly static as availability of information is insufficient for dynamic approach Risk of congestions even without failures shortest path
4
Berlin · Dec. 11 th, 2012 · Hammel, Schöps 4 Common Approach Manual adjustments of routing using dynamic simulations only in selected points expensively developed and tested No holistic approach feasible The bigger the system gets the more time-consuming and difficult this approach
5
Network Approach Transfer AMHS network model Shortest paths easy to find, sophisticated algorithms No dynamic behaviour Berlin · Dec 11 th 2012 · Hammel, Schöpsslide 5 tool queues tool ports ZFS AMHS track inter- section node source / sink information attached to links step 2
6
Track Utilization Average transports per unit of time transports as flows Idea: limit utilization, lower than technical limit because of dynamic behaviour If all tracks keep this limit: Congestions because of traffic should be rare Impacts of failures should be lower (higher robustness) Berlin · Dec 11 th 2012 · Hammel, Schöpsslide 6 sources sinks link utilization mid high low
7
Traffic Distribution Virtually adjusting lengths (=costs) of links enables manipulating routing with no or minor software changes (and without hardware changes) Analytic approach to keep all limits not feasible because of run time Iterative algorithm increasing costs of over-utilized links Berlin · Dec 11 th 2012 · Hammel, Schöpsslide 7 sources sinks link utilization mid high low
8
Algorithm Iteratively increase costs of over-utilized links Possibilities: One by one All over-utilized links at once Amount to increase depending on over-utilization Berlin · Dec 11 th 2012 · Hammel, Schöpsslide 8 + $ utilization mid high low
9
Simulation Network optimization prior to dynamic simulation of AMHS Gained insights from network analysis also help interpreting simulation behaviour and results Berlin · Dec 11 th 2012 · Hammel, Schöpsslide 9 = ?
10
Application New / adjusted transport layouts Evaluation of layout alternatives Analysis of max. TP / bottlenecks Existing transport systems Performance improvement without physical modification Case Study Berlin · Dec 11 th 2012 · Hammel, Schöpsslide 10 Large and complex transport networks
11
GF Fab1 Module1 Cleanroom area 14,000m² at level3 2,000m² at level1 (Test+metrology area) Tools direct deliverable by AMHS 740 at level3 15 at level1 AMHS is ~10 years old system from Murata ~6.5 km of track 280 Vehicle (235 then) ~850 intersections Berlin · Dec 11 th 2012 · Hammel, Schöpsslide 11 51 Stocker with 8120 storage bins ZFGs with up to 2850 storage bins
12
Iteration Process Calculate track utilization by adding shortest paths Increase costs of most used pieces of track (depending on amount of utilization lowering and of mean shortest path length increase) Berlin · Dec 11 th 2012 · Hammel, Schöpsslide 12 - 220 tph - 0 tph - 110 tph Iteratively changing cost factors
13
Validation by Simulation Model impact to AMHS by dynamic simulation Original setting Adjusted cost setting Berlin · Dec 11 th 2012 · Hammel, Schöpsslide 13 Change in average travel distance: + 4.8 % Change in 95-percentile of delivery time in sim.: +/- 0%.. – 20% Change in maximum throughput in simulation: + 10.9 %
14
Real System Implementation Impact on transport performance Berlin · Dec 11 th 2012 · Hammel, Schöpsslide 14 transports / hDT in mins date of change performance of AMHS transport load
15
Summary Network approach for traffic distribution in large transport systems Providing further insight into system behaviour More general system optimization possible because of Shorter run time than dynamic simulation Algorithm is distributing traffic by static routes Throughput increase by changing routes without physical system modification No negative impact to transport times Berlin · Dec 11 th 2012 · Hammel, Schöpsslide 15
16
Berlin, December 11 th 2012 Faculty of Mechanical Engineering · Chair of Logistics Engineering Network Optimization prior to Dynamic Simulation of AMHS Thank you for your attention! Christian Hammel, Technische Universität Dresden Tel.: +49 351 463 32539 E-mail:christian.hammel@tu-dresden.de Matthias Schöps, Globalfoundries Dresden Tel.: +49 351 277 3255 E-Mail:matthias.schoeps@globalfoundries.com
17
From existing system / simulation Based on technical limit and extent of dynamic behaviour Different limits might be necessary for different links Defining the limit Berlin · Dec 11 th 2012 · Hammel, Schöpsslide 17 arrivals per minute technical max average (both) little dynamics high dynamics
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.