Research in object-oriented manufacturing simulations: an assessment of the state of the art

IIE Transactions, Sept, 1998 by S. Narayanan, D.A. Bodner, U. Sreekanth, T. Govindaraj, L.F. McGinnis, C.M. Mitchell

1. Introduction

Advanced manufacturing systems are capital-intensive due to hardware and software requirements. As a result, it is essential that they achieve high levels of flexibility and productivity compared to traditional manufacturing systems. Modeling and analysis to gain a better understanding of the system complexities and to predict system performance are critical in the system design stage, and often valuable for system management.

Modern manufacturing systems tend to be tightly coupled. They are characterized by a high degree of automation, low levels of work-in-process inventory, and various forms of supervisory control. These systems are difficult to analyze using purely analytic models, such as queueing theory or queueing networks [1]. Simulation is an indispensable tool for their design and operational performance analyses.

The rapid growth of the simulation software market provides evidence for the popularity of simulation. Not only are there a number of general purpose simulation languages such as SIMAN (ARENA), SLAM II (AWESIM), MODSIM and GPSS, but also specialized languages and simulators for manufacturing, such as AutoMod, ProModel, and Factor [2]. Simulation languages provide ready-made modeling constructs, and are generally compiled into a standard programming language, such as Fortran or C.

One fundamental problem in applying simulation languages is that of modeling, i.e., using the abstractions of the language to describe the system being analyzed. Recent interest in object-oriented programming (OOP) applied to manufacturing simulation has led to a significant body of related research in the modeling of manufacturing systems [3-8]. There is significant compatibility between the OOP paradigm and the discrete-event worldview formalism [9-11]. To be truly useful, though, research in OOP applied to simulation modeling must have an impact on one or more of the key aspects of simulation modeling and analysis: abstraction, implementation, application, verification, and validation.

One reason for the tremendous appeal of OOP is a major change in abstraction, moving from the traditional 'seize-hold-release' paradigm of simulation languages to a more 'natural mapping' paradigm. This is made possible by the object construct, which allows a one-to-one mapping between objects in the manufacturing system being modeled (process equipment, material handling equipment, programmable controllers, cell controllers, etc.) and their abstractions in the simulation model [6, 12]. OOP also has a major effect on implementation through its facilitation of modular design and software reusability [5, 6, 13, 14]. When exploited, OOP features such as encapsulation, inheritance, and polymorphism facilitate code reuse and programming efficiency [15]. (Refer to Appendix A for a glossary of commonly used terms in object-oriented programming.) For building simulation models, the idea is to design and implement reusable classes and store them in a software library, thus facilitating rapid model development. The concept of a reusable class library therefore aids in application of the modeling architecture to specific scenarios. Extending this idea, one can build a model base, or collection of models, that can be used to facilitate model development further. The OOP paradigm has other potential benefits in verification. The feature of encapsulation in OOP enforces structured code development. Together with modular systems design principles, OOP also offers potential advantages in the incremental development and verification of simulations of large-scale systems [14]. In addition, OOP provides a natural environment for graphical user interface development [7]. Graphical animation of simulation output has been found to be useful in model verification and analysis, Finally, natural mappings offer the modeler fundamental abstractions that correspond closely to the manufacturing world being simulated, thus facilitating bottom-up validation of models. Thus, backed by a design purpose, OOP applied to manufacturing simulation modeling and analysis can be potentially very beneficial.

The natural appeal of OOP in manufacturing simulation has motivated many researchers, whose work is published in a broad range of journals. The objective of this paper is not to provide a comprehensive review of this rapidly expanding literature. Instead, this paper focuses on a subset of the literature representing large-scale, persistent efforts that have addressed fundamental architectural issues, i.e., how we should think about manufacturing in order to create simulation models, and computing tools to implement that way of thinking. The motivation is to make this research available to the widest possible audience, and to identify the current directions of research and future research opportunities. To this end, we focus on the following research efforts:

* BLOCS/M - University of California, Berkeley [3, 4, 13, 16-18]. BLOCS/M is the first manufacturing-specific object-oriented simulation architecture, applied primarily to problems in semiconductor fabrication;