WSC 2003

WSC 2003 Final Abstracts

Simulation-Based Scheduling Track

Monday 10:30:00 AM 12:00:00 PM
Supply Chain Planning

Chair: Peter Lendermann (Singapore Institute of Manufacturing Technology)

Rolling Horizon Scheduling of Multi-Factory Supply Chains
Eunkyoung G. Cho, Kristin A. Thoney, Thom J. Hodgson, and Russell E. King (North Carolina State University)

The Virtual Factory is a job shop scheduling tool that was developed at NC State. It has been found to provide near-optimal solutions to industrial-sized problems in seconds. Recently, the Virtual Factory was expanded to include inter-factory transportation operations which enabled the detailed scheduling of entire multi-factory manufacturing supply chains. Separately, a rolling horizon procedure was developed to test the Virtual Factory for single factory problems. This procedure allowed us to more accurately predict how the Virtual Factory would perform in industry. Consequently, the rolling horizon procedure was extended to multi-factory settings to gauge industrial performance and eliminate transient effects found in previous multi-factory experimentation. Experimental results, under a variety of different scenarios, indicate that the Virtual Factory also performs well in multi-factory, rolling horizon settings.

A Reinforcement Learning Approach to Production Planning in the Fabrication/Fulfillment Manufacturing Process
Heng Cao (IBM T.J. Watson Research Center), Haifeng Xi (IBM T.J. Watson Research) and Stephen F. Smith (Carnegie Mellon University)

We have used Reinforcement Learning together with Monte Carlo simulation to solve a multi-period production planning problem in a two-stage hybrid manufacturing process (a combination of build-to-plan with build-to-order) with a capacity constraint. Our model minimizes inventory and penalty costs while considering real-world complexities such as different component types sharing the same manufacturing capacity, multi-end-products sharing common components, multi-echelon bill-of-material (BOM), random lead times, etc. To efficiently search in the huge solution space, we designed a two-phase learning scheme where “good” capacity usage ratios are first found for different decision epochs, based on which a detailed production schedule is further improved through learning to minimize costs. We will illustrate our approach through an example and conclude the paper with a discussion of future research directions.

Theory and Practice of Advanced Planner and Optimizer in Supply Chain Domain
Sam Bansal (International Management Consultant)

This paper describes the Supply Chain Management do-main of SAP. It further describes how the SAP APO, the Advanced Planner & Optimization tool set fits in the over-all domain of SCM. The founding principles of APO are also presented. Various algorithms used as part of planning & optimization are presented, as well as their relationships with simulation techniques.

Monday 1:30:00 PM 3:00:00 PM
Factory Scheduling

Chair: Juergen Potoradi (Infineon Technologies AG)

Simulation based Scheduling Using a Two-Pass Approach
Chin Soon Chong, Appa Iyer Sivakumar, and Robert Gay (Nanyang Technological University)

Bottleneck based scheduling is a popular approach in production scheduling, and it has achieved promising results in industry. To incorporate this approach in discrete event simulation tools is difficult since the approach requires multiple passes, forward and backward, to reach a good solution for the scheduling problem. In this paper, we propose a two-pass scheduling approach using discrete-event simulation that takes bottlenecks into consideration. In the first pass, a simulation run is performed and bottlenecks are determined. If significant bottlenecks are identified, a second-pass simulation is performed to reduce the loading on bottlenecks through specific scheduling strategies.

Implementing a Simulation-Based Scheduling System for a Two-Plant Operation
Jeffrey A. Joines, Andrew W. Sutton, Kristin Thoney, Russell E. King, and Thom J. Hodgson (North Carolina State University)

Scheduling any complicated job shop becomes increasingly more difficult when the cycle time is reduced. This paper will discuss the implementation of a simulation-based scheduling system that properly schedules parts in a two-plant operation. The system has allowed the company to reduce the cycle time by at least a week from two/three weeks to one/two weeks. As part of the project, the generation of the input data needed to drive the simulation is also discussed since this data did not exist in the correct form. The model generation, simulation development, and experimentation will be discussed. The system that is described is currently being used to generate the schedules.

Simulation-Based Finite Scheduling at Albany International
Juha-Matti Lehtonen and Patrik Appelqivst (Helsinki University of Technology) and Teemu Ruohola and Ilkka Mattila (Delfoi Ltd.)

Simulation-based production scheduling approaches are emerging as alternatives to optimization and simpler approaches such as priority rules. This paper presents an application of a simulation-based finite scheduling at Albany International, the largest manufacturer of paper machine clothing in the world. Simulation is used as a decision support tool for manual schedule creation. User experiences have been encouraging. We argue that an optimization-based approach is not necessarily the most economical and identify a number of tentative key enablers of a simulation-based solution. The case indicates that a simulation-based solution is a viable option when the production process does not include combination of materials and local sequencing is adequate. A simulation-based solution capitalizes on this existing source of tacit knowledge by giving expert human schedulers tools for testing and improving schedules.

Monday 3:30:00 PM 5:00:00 PM
Dynamic Scheduling I

Chair: Sam Bansal (International Management Consultant)

Real-Time Decision Making Using Simulation
Mukesh Dalal, Brett Groel, and Armand Prieditis (LookAhead Decisions Inc.)

Based on a discrete-event simulation model, Simulation-based Real-time Decision-Making (SRDM) is an innovative approach to real-time, goal-directed decision-making. When applied to a flexible manufacturing system, SRDM makes better decisions than most fixed policies, such as deterministic, stochastic and manual. SRDM even improves over machine-enhanced policies that have been optimized over several hours using a tool such as OptQuest. Compared to these fixed policies, SRDM shows greater improvement for more complex systems and is quite robust with respect to modeling errors. SRDM can handle the unexpected, because it avoids the rigidity and myopia caused by fixed policies. Since most real-time decisions in currently deployed manufacturing systems are made either manually or by using fixed policies, our results suggest that using SRDM instead could lead to significant improvement in operating performance.

Simulation-Based Scheduling for Dynamic Discrete Manufacturing
Chin Soon Chong, Appa Iyer Sivakumar, and Robert Gay (Nanyang Technological University)

A simulation-based real-time scheduling mechanism for dynamic discrete manufacturing is presented in this paper. Modified mean flow time performance for different scheduling approaches is compared through off-line simulation experiments, under dynamic manufacturing environments that are subjects to disturbances such as machine breakdowns. These experimental results are used as reference indices for the real-time scheduling mechanism to select the better scheduling approaches for further evaluation based on the actual manufacturing conditions. Discrete-event simulation is used on-line to evaluate the selected approaches and the corresponding schedules to determine the best solution. The selected schedule is used until the deviation of actual performance from the estimated one exceeds a given limit, or when a major event occurs. A new simulation is then performed with the remaining operations to select a new schedule.

Use of Discrete Event Simulation to Analyze Dispatch Policies of an Equipment Group in Semiconductor Fab
Raja Sunkara and Ramesh Rao (National Semiconductor Corp.)

This article describes a methodology to model complex operation and process driven practices using a discrete event simulator. This level of detail in the model is critical for the analyses and design of complex operation and process driven dispatch policies in a semiconductor fab. The modeling of these practices is typically not a part of the general set of rules and methods provided by commercially available simulation software. The methodology provides key information that simplifies the development of suitable dispatch policies subject to factory dynamics. The modeling philosophy plays a key role in the success of simulation as a culture. As an example, we present the modeling of complex floor practices to analyze the impact of setup changes subject to process restrictions.

Tuesday 8:30:00 AM 10:00:00 AM
Dynamic Scheduling II

Chair: Leon McGinnis (Georgia Tech)

Look-Ahead Strategies for Controlling Batch Operations in Industry – An Overview
Durk-Jouke van der Zee (University of Groningen)

Batching jobs in a manufacturing system is a very common policy in most industries. Main reasons for batching are avoidance of set ups and/or facilitation of material handling. Examples of batch-wise production systems are ovens found in aircraft industry and in semiconductor manufacturing. Starting from the early nineties much research efforts have been put in constructing strategies for the dynamic control of these systems in order to reduce cycle times. Typically, these so-called `look-ahead strategies' base their scheduling decision on the information on a few near future product arrivals. In this paper we give a literature overview of the developed strategies, consider basic insights in their construction and highlight issues for further research.

SIMUL8-Planner Simulation-Based Planning and Scheduling
Kieran H. Concannon, Kim I. Hunter, and Jillian M. Tremble (Visual8 Corporation)

This paper provides an introduction to the technique of simulation-based production planning and scheduling, a fast growing and popular area in the simulation industry. SIMUL8 and Visual8 Corporations have collaborated to develop a new software application called SIMUL8-Planner that assists in the development of this type of system. The following document outlines some of the requirements, advantages, and features within this exciting new product.

Fast Simulation Model for Grid Scheduling Using HyperSim
Sugree Phatanapherom and Putchong Uthayopas (Kasetsart University) and Voratas Kachitvichyanukul (Asian Institute of Technology)

To develop grid scheduling algorithms, a high performance simulator is necessary since grid is an uncontrollable and unrepeatable environment. In this paper, a discrete event simulation library called HyperSim is used as extensible building blocks for grid scheduling simulator. The use of event graph model for the grid simulation are proposed. This model is well supported by HyperSim which yields a very high performance simulation. The experiments are conducted to compare HyperSim with other several simulators in terms of speed and scalability. The result shows a significant simulation speed improvement over many widely used simulators. Furthermore, sample simulation results of basic job scheduling problem are shown to com-pare to well-known heuristics.

[ Return to Top | Return to Program ]