Oliver Rose Dresden University of Technology Department of Computer Science Chair for Modeling and Simulation Benefits and Drawbacks of Simple Models for Complex Production Systems

2 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Dresden University of Technology  Dresden located in the state Saxony, often called “Silicon Saxony” (factories from Infineon, Qimonda, AMD, and ZMD; lots of suppliers)  Largest German University of Technology  Focus on Computer Science, Electrical Engineering, and Mechanical Engineering  About 35,000 students, 3,000+ students in Computer Science, 700+ beginners in winter semester 2005/2006  Faculty of Computer Science consists of 30 chairs  Institute of Applied Computer Science has 5 chairs, dealing with different aspects of factory planning, control, and automation

3 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Research Overview Complex Production System Model Simulation Dispatching Simulation- based Scheduling Measurement & Analysis Automation of Data Transfer High-level Production Control

4 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Cooperations  Industry  Infineon Technologies AG, Munich and Dresden  AMD Saxony LLC & Co. KG, Dresden  KBA Planeta, Radebeul (Offset Printing Machines)  Airbus Deutschland GmbH, Hamburg  Academia  Arizona State University, Tempe, AZ, USA  Singapore Institute of Manufacturing Technology  Georgia Institute of Technolgy, Atlanta, GA, USA  FernUni Hagen, Germany  Fraunhofer IPA, Stuttgart, Germany

5 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Introduction  Goal of the study: Estimate the transient behavior of a wafer fab after recovery from a catastrophic failure in order to improve the flow of material  Motivation: Infineon engineers reported that large amounts of WIP are present in the fab even weeks after end of repair  Goals of improvement:  reduction of WIP  reduction of cycle times  on-time delivery (reduction of variability of cycle times)

6 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Semiconductor Manufacturing Back End Front End Wafer Fab TestAssembly Probe

7 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Flow of material Fotos: Fullman-Kinetics, Varian, Sematech International

8 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Characteristics of wafer fabrication facilities  Large number of processing steps, typically several hundreds  Large number of tools of different types: photo equipment, ovens, etching equipment, ion implanters, …  Wafer are build up in layers: reentrant flow of material, jobshop type way of production  Machine breakdowns (typical availability: 70-90%)  Auxiliary resources, e.g., reticles (photo masks)  Batch tools with complex batching criteria  Sequence dependent setups

9 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Important performance measures  Cycle time: low  Output: high  Machine utilization: high  Inventory (work in process, WIP): low  Yield (percentage of good dies on a wafer): high  Cost per die: low Conflicting goals!

10 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Controllable factors  Factory load (wafer starts per week)  Product mix  Number of machines and operators  Preventive maintenance policies  Production planning & control policies:  scheduling vs. dispatching  lot release vs. shop-floor control  …

11 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Fab Model fixed number of cycles bottleneck workcenter delay unit (rest of the fab) finished? yes no lot release

12 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Modeling Details  Dispatching strategies:  FIFO  Critical Ratio  Slack Time  Delay time distributions of the delay unit  Lot release policy  Partial bottleneck breakdowns

13 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Modeling the Rest of the Fab ?

14 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Delay Time Distributions  Constant amount of time for sum of processing times  Distributions of different variability for sum of non- processing times (waiting times, transport times, etc.):  Constant (coefficient of variation: 0.0)  Erlang-5(coefficient of variation: 0.447)  Exponential(coefficient of variation: 1.0)  Independence of consecutive and parallel delays assumed

15 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Constant Delay FIFO Critical Ratio Slack Time time WIP

16 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Shifted Exponential Delay FIFO Critical Ratio Slack Time time WIP

17 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Shifted Erlang Delay FIFO Critical Ratio Slack Time time WIP

18 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Delay Distribution from Real Data

19 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Partial Bottleneck Workcenter Breakdowns ?

20 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Modeling of Partial Breakdowns number of broken machines duration of breakdowns

21 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models One Broken Machine for 200 Time Units time WIP FIFO Critical Ratio Slack Time

22 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models All machines broken for 50 time units FIFO Critical Ratio Slack Time time WIP

23 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Combination of two dispatch strategies ? Critical Ratio -> FIFO Slack Time -> FIFO

24 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Combination of Critical Ratio and FIFO FIFO Critical Ratio t =2500 FIFO time WIP

25 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Stop Lot Release During Breakdown Period

26 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Lot Releases Stopped FIFO (stopped) Critical Ratio (stopped) Slack Time (stopped) FIFO Critical Ratio Slack Time time WIP

27 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Estimation of Cycle Time Distributions Hours Full Simple Hours FIFO Full Simple Hours CR

28 Dresden University of Technology Department of Computer Science Oliver Rose Simple Models Conclusions  Simple models useful for understanding complex production systems  Simple models useful for the design of simple and robust scheduling heuristics  Not a tool for beginners  Pitfall of oversimplification  Simplification must not be the goal but only the method to reach the goal  Keep the model as simple as possible but not simpler!