Elsevier

Computers in Industry

Volume 58, Issue 7, September 2007, Pages 644-655
Computers in Industry

Simulation-based assessment of machine criticality measures for a shifting bottleneck scheduling approach in complex manufacturing systems

https://doi.org/10.1016/j.compind.2007.05.010Get rights and content

Abstract

In this paper, we describe adaptation techniques for a hierarchically organized multi-agent-system (MAS) applied to production control of complex job shops. The system architecture of the production control system is based on three different control layers. The mid layer implements a distributed shifting bottleneck type solution procedure. The shifting bottleneck heuristic decomposes the overall scheduling problem into scheduling problems for parallel machines. The sequence of solving the resulting scheduling problems for parallel machines is determined by machine criticality measures. We can adapt this solution scheme in a situation dependent manner by choosing appropriate machine criticality measures. Furthermore, the performance of the shifting bottleneck scheme is also influenced by the selection of a proper subproblem solution procedure for each of the parallel machine scheduling problems. The subproblem solution procedures typically are given by heuristics. A situation dependent parameterization of these heuristics is highly desirable. In this paper, we sketch an overall concept for adaptation of our hierarchically organized multi-agent-system. We present results of computational experiments based on the simulation of a dynamic environment for the appropriate selection of machine criticality measures.

Introduction

Complex job shops are characterized by parallel machines, sequence dependent set-up times, a mix of different process types, for example batch processes, prescribed due dates and reentrant flows (cf. Ovacik and Uzsoy [1] and Mason et al. [2] for the notation of complex job shops). Semiconductor wafer fabrication facilities (wafer fabs) are examples for complex job shops. As indicated by Schömig and Fowler [3], today it seems that better operational strategies are the main key in order to reduce costs and improve overall efficiency. New planning, scheduling, and dispatching methods are required in order to reach the goal of better operational performance. The improved software and hardware capabilities have to be taken into account during the development of more sophisticated algorithms.

Manufacturing systems are changing over time because of the global competition and unforeseeable markets. Today's manufacturing systems are customer demand driven and have to face with a large number of different products and an over time changing product mix in contrast to mass production type manufacturing systems of former decades. Therefore, meeting customer demands and especially due dates is extremely important. The changing customer demands result in a changing base system and process. Here, we define the base system as the set of machines (tools) that form the manufacturing system. The base process is given by the routes of the jobs and by a work in process (WIP) distribution as an initial condition. An over time changing base system and process leads to an over time changing production control system and also to a modified control process. The production control system is given by the control algorithms and the software and hardware to run these algorithms. Therefore, production control algorithms that are able to adapt to different situations are highly desirable. In this paper, we describe some steps towards reaching this goal by considering a hierarchically organized multi-agent-system (MAS) for production control of complex job shops. We show how we can obtain, in principle, an adaptive behaviour of the production control system. Furthermore, we describe how we can assess the performance of the production control system by emulation of the manufacturing system via discrete-event simulation.

The paper is organized as follows. In the next section, we describe the considered problem. Then, we discuss related literature. In Section 4, we present a concept for adaptation of the MAS used for production control of complex manufacturing systems. Section 5 describes our benchmarking architecture and explains the experimental setting. Finally, in Section 6, we present the results of the simulation-based benchmarking efforts.

Section snippets

Multi-agent-system architecture and solution approach for production control

In this section, we discuss very briefly a hierarchically organized MAS used for production control of complex manufacturing systems. Then we give a short outline of the shifting bottleneck heuristic and discuss its potential for adaptation.

Related literature

The shifting bottleneck heuristic for job shops is originally described for the make span performance measure in [6]. Extensions to other performance measures and more realistic process conditions are presented, for example, by Ovacik and Uzsoy in [1] and by Ivens and Lambrecht [11]. For this paper, the modifications carried out by Mason et al. [2] are the most important. Basically, these modifications allow for the usage of shifting bottleneck type algorithms to schedule wafer fabs. However,

Adaptation concept for the hierarchically organized multi-agent-system

In this section, we describe first some general principles for adaptive production control. Then, we apply these principles to our hierarchically organized MAS. We present details for the solution of problem P1.

Benchmarking issues

In this section, we describe first a simulation-based benchmarking approach applied to the solution of problem P1. Then, we describe the design of simulation experiments.

Computational experiments

In a first scenario, we consider Model A with high and with very high load and tight due dates, i.e., we use FF = 1.4. We show the corresponding computational results in Table 2. All values are the ratio of total weighted tardiness values obtained by the shifting bottleneck approach with the criticality measure given by Eq. (6) and with a shifting bottleneck heuristic with the TWT based criticality measure discussed in Section 4.2.

It turns out that for both high and very high load of the system

Conclusions and future research

In this paper, we discuss a concept for adaptation of a hierarchically organized MAS to different system conditions. We describe a general adaptation architecture for production control systems. The basic idea consists in constructing a mapping between certain situation and parameterization attributes. Then, we apply this architecture to our hierarchically organized MAS. We describe our benchmarking architecture and explain how we may implement an adjustment component as a special staff agent.

Lars Mönch is a Professor for Enterprise-wide Software Systems in the Department of Mathematics and Computer Science at University of Hagen, Germany. He received a master's degree in applied mathematics and a PhD in the same subject from the University of Göttingen, Germany. He earned a Habilitation degree in Information Systems from the Technical University of Ilmenau in 2005. He worked in the area of object-oriented software development for 2 years after getting his PhD degree. His current

References (35)

  • A. Schömig et al.

    Modelling semiconductor manufacturing operations

  • L. Mönch et al.

    FABMAS—an agent based system for semiconductor manufacturing processes

  • L. Mönch et al.

    The FABMAS multi-agent-system prototype for production control of waferfabs: design, implementation, and performance assessment

    Production Planning & Control

    (2007)
  • J. Adams et al.

    The shifting bottleneck procedure for job shop scheduling

    Management Science

    (1988)
  • L. Mönch et al.

    A distributed shifting bottleneck heuristic for complex job shops

    Computers and Industrial Engineering

    (2005)
  • M. Woolridge

    An Introduction to Multiagent Systems

    (2002)
  • H. Van Brussel et al.

    Refernce architecture for holonic manufacturing systems: PROSA, computers in industry

    Special Issue on Intelligent Manufacturing Systems

    (1998)
  • Cited by (22)

    • A survey of scheduling with parallel batch (p-batch) processing

      2022, European Journal of Operational Research
    • Modeling and simulation in intelligent manufacturing

      2019, Computers in Industry
      Citation Excerpt :

      And the discrete event simulation method was also used to select process improvement initiatives [29,30]. Studies on online simulation-based scheduling include simulation-based dynamic planning and scheduling [31,32], dispatching rules-based simulation [33,34], simulation-based hybrid backwards scheduling [35], machine criticality-based simulation [36], etc. With the rapid increase of the computing power, the simulation running time is greatly reduced, which creates conditions for online simulation technology.

    • Flexible manufacturing system with industry 4.0

      2023, AIP Conference Proceedings
    View all citing articles on Scopus

    Lars Mönch is a Professor for Enterprise-wide Software Systems in the Department of Mathematics and Computer Science at University of Hagen, Germany. He received a master's degree in applied mathematics and a PhD in the same subject from the University of Göttingen, Germany. He earned a Habilitation degree in Information Systems from the Technical University of Ilmenau in 2005. He worked in the area of object-oriented software development for 2 years after getting his PhD degree. His current research interests are in simulation-based production control of semiconductor wafer fabrication facilities, applied optimization, multi-agent-systems, and artificial intelligence applications in manufacturing. He is a member of GI (German Chapter of the ACM), GOR (German Operations Research Society), SCS and INFORMS. He has been a member of the Intelligent Manufacturing Systems Network of Excellence (IMS-NoE) of the European Community, SIG 4—Benchmarking and Performance Measurement of Online-Scheduling-Systems since 2004.

    Jens Zimmermann is a PhD student at the Chair of Enterprise-wide Software Systems at the University of Hagen, Germany. He received a master's degree in information systems from the Technical University of Ilmenau. He is interested in semiconductor manufacturing, simulation, multi-agent-systems, and machine learning. He is a member of GI. He has been a member of the Intelligent Manufacturing Systems Network of Excellence (IMS-NoE) of the European Community, SIG 4—Benchmarking and Performance Measurement of Online-Scheduling-Systems since 2004.

    View full text