Abstract

In this paper, we first sketch conventional ("crisp") approaches to obtain performance values for variables such as utilization, number in the system, throughput, response time, and lost requests (when relevant). Two calculation schemes are followed, one from arrival- and service-rates and another from transition probability matrices. These values are to be associated with cost and benefit factors and, with multiple runs over a finite set of choices (dictated by the practical situation), optimizations in outcome become possible. After these sketches, fuzziness is introduced. A desired goal is to compute results from fuzzy computations using conventional simulation. The first goal sought is to have the conventional simulations produce results quicker. A second goal is to get results that are either identical or which represent respectable approximations. Since this goal has proved somewhat elusive, a third goal seeks steps to improve on the situation and results show a high degree of success but with some remaining issues.

Key Words

Fuzzy probability, fuzzy transition matrices, arrival rates, service times, "fuzzification time."