Optimization in Manufacturing Simulation using Numeric Methods and AI-Techniques as Aid-Tools

Univ.-Prof. Dr.-Ing. Peter Scharf, Dipl.-Ing. Georg Fretter, Universitaet-GH Siegen


Planning of modern flexible manufacturing systems (FMS) becomes increasingly complex. Higher integrated and automated production systems need a detail and careful analysis of the dynamic scenario to prevent misinvestments. Not only the planning object has to be optimal, the planning tools also should be appropriate and efficient to achieve best planning results. The methodology of computer simulation has been well developed in the last decade. Many publications demonstrate the usefulness of this planning aid.

Computer simulation in manufacturing is often used to find the optimum layout of a multi-station process or an entire factory for specific goals. Such goals are for example the minimum capacity of pre- or postprocess workpiece buffers or the minimum number of working machines, personnel and fixtures for a given production output. When using the methodology of computer simulation, these optimization procedures are iterative:

The simulationist develops a model, runs the simulation, analyzes the output figures; - and modifies model input data relatively to the goal, all based on his knowledge and experience. He runs the simulation cycles again and again until the results are sufficient.

The phase of carrying out such simulation experiments (model optimization phase) can take up to 20 per cent in relation to the workforce of a complete study, as experience shows. For an effective realization of this phase the simulationist can use numeric optimization techniques. Research institutes have developed knowledge based simulation methods for the configuration of manufacturing systems also. The "Institute for Manufacturing Engineering" of the Siegen University has developed the Computer program POSP (Program system to Optimize Simulated Production systems) to support the simulationist during the phase of optimization. The important view is, that the simulationist can choose between different optimization techniques to achieve best results in a given application. The implementation of the software prototype includes four alternative optimization techniques, which also may be used in cooperation:

These methods, formulated as algorithms in computer programs, are connected with the established simulation package SLAMSYSTEM, either WINDOWS- or OS/2-Version, as well as with FACTOR/AIM on personal computers for manufacturing simulation applications.

The aim of this paper is to show how the simulationist can reduce the time in the optimization phase when using the developed program. The practicable improvement is shown in a typical application of factory simulation.

Univ.-Prof. Dr.-Ing. Peter Scharf
Tel. 49 - (0) 271 / 740-2266
Universitaet Siegen, FB 11
Labor fuer Fertigungsautomatisierung
Paul-Bonatz-Str. 9-11
57368 Siegen
Fax. 49 - (0) 271 / 740-2542

About the authors:

Univ.-Prof. Dr.-Ing. Peter Scharf, born 1942. He got the Dr.-Ing. degree in 1975 at the University of Stuttgart. In the next 10 years he was manager of production planning and director of manufacturing in a German aircraft producing company, mainly involved in the production of the European "Airbus". In 1985 he was appointed as full professor for manufacturing automation in the department of mechanical engineering at the University of Siegen. He is mainly dealing with assembly automation, application of industrial robots, computer aided engineering in manufacturing and factory simulation.

Dipl.-Ing. Georg Fretter, born 1965, graduated in mechanical engineering in 1992 at the University of Siegen. Since thenhe has been working as a research assistant at the Institute for Manufacturing Engineering in Siegen, mainly in the area of simulation applications and program development.