The paper describes a fast simulation component for a Petri net tool. The approach is based on the recently developed RESTART method, which is applicable for a wide range of simulation models and has the potential to reduce the simulation overhead by some orders of magnitude if the measure of interest corresponds to a rare event in the model.
The paper presents a selection technique for thresholds, which are the most important input parameters of RESTART. The technique comprises of a pilot study combined with a heuristic for the selection of thresholds according to an optimum criterion. Examples show speedups up to one million and investigate the sensitivity if non-optimal parameters are chosen.
Address:Technische Universitaet Berlin
Franklinstr. 28/29 (Sekr. FR2-2)
fax: +49.30.314 21 116
phone: +49.30.314 73 116
Technische Universitaet Berlin, Department of computer science
diploma degree in electrical engineering (1992)