STW-project Effective Process Time: an overview

Transcription

1 12 STW-project Effective Process Time: an overview Ivo Adan, Onno Boxma, Pascal Etman, Ad Kock, Erjen Lefeber, Koos Rooda, Marcel van Vuuren Department of Mechanical Engineering & Department of Mathematics and Computer Science Eindhoven University of Technology Eurandom workshop June 19-20th 2006 Eindhoven 1/25

2 12Contents Effective process time concept STW project 2/25

3 12 3/25

4 Contents 12 Motivation Performance analysis of manufacturing systems. JJJNIII 4/25

5 12Motivation Performance analysis of manufacturing systems 5/25

6 12Models of manufacturing systems Performance measures: Throughput δ Flow time ϕ Work-in-process w Shopfloor realities: Machine downs, repairs, setup, product mix, operators, batching, product recipes, priority lots,... Modeling approaches: Queueing (network) analysis (exact/approximation) Discrete-event simulation 6/25

7 12Models of manufacturing systems Queueing (network) analysis (exact/approximation): Computationally efficient: ideal for what-if and optimization studies Limited applicability due to restrictive assumptions regarding shopfloor realities Discrete-event simulation: Shopfloor realities may be modeled in detail Computationally intensive Model parameters: Can we obtain all required data to set the model parameters? 7/25

8 12 8/25

9 12Example: a production flow line Station 1 Station 2 Station 3 Station 4 Flowline characterisics: Infinite buffers, single-server stations, single-lot machines. Shopfloor realities: Process mix, setup-times, recipes, operator availability, machine downs, repair,rework, hot-lots, sequencing, scheduling, shopfloor control,... 9/25

10 12Zooming in: single-server workstation Reality: arrival queueing M processing departure Natural processing Random failures (preemptive): e.g. machine downs Planned process delays (non-preemptive): e.g. setups Aggregation: Effective process time (EPT parameters t e and c 2 e may be estimated by adding natural processing and preemptive Queueing model: CT = c2 a +c 2 e 2 and nonpreemptive outages) u 1 u t e + t e, u = t e ta [Hopp, Spearman, 2000] 10/25

11 12Idea: Measure EPTs from arrival/departures Legend Setup Processing Queueing Waiting for operator Machine breakdown 11/25

12 12Idea: Measure EPTs from arrival/departures EPT 1 Legend Setup Processing Queueing Waiting for operator Machine breakdown 11/25

13 12Idea: Measure EPTs from arrival/departures EPT 1 EPT 2 Legend Setup Processing Queueing Waiting for operator Machine breakdown 11/25

14 12Idea: Measure EPTs from arrival/departures EPT 1 EPT 2 EPT 3 Legend Setup Processing Queueing Waiting for operator Machine breakdown 11/25

15 12Idea: Measure EPTs from arrival/departures EPT 1 EPT 2 EPT 3 EPT 4 Legend Setup Processing Queueing Waiting for operator Machine breakdown 11/25

16 12Idea: Measure EPTs from arrival/departures EPT 1 EPT 2 EPT 3 EPT 4 EPT 5 EPT i = d i max(a i, d i 1 ) Legend Setup Processing Queueing Waiting for operator Machine breakdown 11/25

17 12Generalization of the idea Event data Manufacturing system EPT algorithm Throughput Flow time Observed EPT realisations EPT parameters Distribution fit reject Validation accept Bottleneck analysis System optimisation Aggregate model Throughput Flow time Predicted Effective process time modeling framework 12/25

18 12 13/25

19 12Research objective and approach Goal: To build simple yet accurate models of manufacturing networks using operational factory data without the need to characterize all contributing disturbances and shopfloor realities. Approach: Effective process time paradigm in aggregate modeling and parameter identification Efficient queueing network approximations through iterative algorithms based on decomposition 14/25

20 12Challenges Machine: Single-lot, batching, conveyor, assembly, other types,... Workstation (multiple parallel servers sharing a buffer): Dispatching, recipes, unequal machines, lot overtaking,... Network: Blocking (finite buffers), assembly lines, re-entrant flow lines, general queueing network, production control,... 15/25

21 12 16/25

22 12Single-lot machine workstation EPT parameter estimation eqm 2 eqm lot 1 ept 1 lot 2 lot 3 ept 2 ept 3 eqm 2 eqm lot 1 ept 1 lot 2 ept 2 lot 3 ept A B C D E F G H I J K L M N 0 A B C D E F G H I J K L M N [Jacobs, Etman, Van Campen, Rooda, 2001, 2003] 17/25

23 12Batch machine workstation EPT parameter estimation lot 1 lot 2 lot 1 lot 3 lot 2 lot 4 lot 3 batch 1 batch 2 batch 1 batch 2 ept 1 2 ept 1 ept t e [-] [-] t c 2 e2 [-] c 0 [-] t [h] A B C D E F G H I A B C D E F G H I Equipment family Equipment family [Jacobs, Van Bakel, Etman, Rooda, 2006] 18/25

24 12Flow line subject to blocking EPT parameter estimation ept 0 ept 1 PA 0 =AA 0 PA 1 PD 0 =AD 0 =AA 1 PD 1 AD 1 time Queueing network approximation (talk Van Vuuren) * * *!! 5 > 5 > 5 >! 5! ) >, ) >,! )! >!,! [Kock, Wullems, Etman, Adan, Rooda, 2005] [van Vuuren, Adan, Resing-Sassen, 2005] 19/25

25 12Assembly machine EPT parameter estimation lot 0 lot 1 lot 2 lot 3 1SLMB-a events ept0 ept1 ept2 ept3 PD0 PD1 PD2 AD0 AD1 AD2 AA2.0 AA3.0 AA4.0 AA1.1 AA2.1 AA time Queueing approximation (talk Van Vuuren) 20/25

26 5 5 Contents 12Assembly machine EPT parameter estimation lot 0 lot 1 lot 2 lot 3 1SLMB-a events ept0 ept1 ept2 ept3 PD0 PD1 PD2 AD0 AD1 AD2 AA2.0 AA3.0 AA4.0 AA1.1 AA2.1 AA time Queueing approximation (talk Van Vuuren) 5 5! 5 " [Vijfvinkel, 2005] [van Vuuren, Adan, 2005,2006] 21/25

27 12Lithography cell Aggregate modeling and parameter estimation Head-tail model (talk Van der Eerden) Arrivals Buffer Head Tail Exit Inside detailed, outside aggregate (talk Kock) Aggregate modelling: Outside Machine Detailed modelling: Inside Machine r s e S a B b D c L d T f E [Van der Eerden, Saenger, Walbrick, Niesing, Schuurhuis, 2006] [Kock, Etman, Rooda, 2006] 22/25

28 12 23/25

29 12 Industry: Automotive industry (talk Nijsse) Semiconductor industry (talk Van der Eerden) (talk Van Campen) Consultancy (talk Resing/v. Doremalen)... Research: Control of manufacturing networks (talk Lefeber) In work by our students and colleagues Related research and work by others: See e.g. talks on Tuesday... 24/25

30 12Bibliography J.H. Jacobs, L.F.P. Etman, E.J.J. van Campen and J.E. Rooda (2001): Quantifying Operational Time Variability: the Missing Parameter for Cycle Time Reduction. In: 2001 IEEE/SEMI Advanced semiconductor manufacturing conference, April, Munich, pp J.H. Jacobs, L.F.P. Etman, E.J.J. van Campen and J.E. Rooda (2003): Characterization of operational time variability using effective process time. IEEE Transactions on Semiconductor Manufacturing, vol. 16 (3), pp M. van Vuuren, I.J.B.F Adan, S.A.E. Resing-Sassen (2003): Performance Analysis of Multi-Server Tandem Queues with Finite Buffers. In: Fourth International Conference on Analysis of Manufacturing Systems. pp A.A.A. Kock, F.J.J. Wullems, L.F.P. Etman, I.J.B.F. Adan and J.E. Rooda (2005): Performance Evaluation and Lumped Parameter Modelling of Single Server Flowlines subject to Blocking: an Effective Process Time Approach. In: Fifth International Conference on Analysis of Manufacturing Systems-Production Management. pp ( M. van Vuuren and I.J.B.F Adan (2005): Performance Analysis of Tandem Queues with Small Buffers. In: Proc. Fifth International Conference on Analysis of Manufacturing Systems-Production Management. pp ( M. van Vuuren and I.J.B.F. Adan (2005): Performance analysis of assembly systems. In: Proc. Fifth International Conference on Matrix Analytic Methods in Stochastic Models, June, Pisa. M. van Vuuren, I.J.B.F Adan, and S.A.E. Resing-Sassen (2005): Performance Analysis of Multi-Server Tandem Queues with Finite Buffers and Blocking. OR Spectrum, vol. 27, pp M. van Vuuren and I.J.B.F Adan (2005): Approximation the ΣGI/ G/ s Queue by using Aggregation and Matrix Analytic Methods. Stochastic Models, vol. 21, pp J.H. Jacobs, P.P. van Bakel, L.F.P. Etman and J.E. Rooda (2006): Quantifying variability of batching equipment using effective process times. IEEE Transactions on Semiconductor Manufacturing, vol. 19, pp A.A.A. Kock, L.F.P. Etman and J.E. Rooda (2006): Lumped Parameter Modelling of the Litho Cell In: Proc. INCOM 2006, May, St. Etienne. M. van Vuuren and I.J.B.F. Adan (2006): Performance analysis of assembly systems. In: Proc. Markov Anniversary Meeting, June, Charleston. M. van Vuuren and I.J.B.F Adan (2006): Approximating Multiple Arrival Streams by using Aggregation. Stochastic Models, to appear. 25/25