شبیه سازی گسسته مدل کاهش رویداد : رویکرد علت و معلولی

Discrete event simulation is an important system analysis technique. But for today’s demand for speed, the time to complete a simulation study is considered to be long, even with current developments in hardware and simulation software. In this scenario, simplification methods for simulation models could play a key role. This paper proposes a technique for reducing the complexity of a discrete event simulation model at the conceptual phase of simulation modeling that can be fully automatized through a computer program. We applied this technique on some problems which demonstrate the feasibility of this approach.

Current simulation studies are being performed by means of more complex models. This is allowed by the development of powerful computer hardware and software (a more detailed discussion of some of these factors is provided in [1]). Complex and huge simulation models forces modelers to face problems they were not used to, referred by Page et al. [2] as “Problems of Scale”. Despite of the growth of more complex models, several authors reinforced the importance of simpler simulation models [3], [4], [5], [6], [7], [8] and [9]. Salt [10] asserts that “simplification is the essence of simulation” and Pidd [11] is conclusive in his declaration: “Complicated models have no divine right of acceptance”. Unfortunately according to Brooks and Tobias [12] and [13] there is a scarcity of simulation research in the field of simplification of simulation models. These authors also mentioned that the majority of work in this field does not constitute a general methodology to simplify a given simulation model. The aim of this paper is to propose a technique for reducing the complexity of simulation model at the conceptual phase of simulation i.e. when the results of simulation runs are yet known. This paper is organized as follows: Section 2 makes a brief review in literature regarding model reduction or simplification. Section 3 defines the reduction problem; Section 4 gives a brief explanation on the chosen model representation technique for the reduction process. In Section 5, we explain how the reduction algorithm works, and in Section 6, we present a simple application example while in Section 7 a more complex example is shown. Finally, Section 8 concludes this paper, adding directions for future work.

This work is an attempt to search for a formal methodology for the Model Reduction Problem at the conceptualization phase of simulation modeling. Besides the simple example shown here, we have made other tests with the algorithm which led to the following conclusions: (1) The potential to reduce an existing simulation model does not depend on its complexity, but on simulation objectives and additional hypotheses made on the model, refuting the common sense idea that a more complex model has more reduction potential. (2) As shown in the example, there are some cases that no reduction is achieved, because all attributes considered in the model can affect the desired performance measures. The reduction algorithm proposed by this work follows the evolutive/reductionist approach, i.e. starts with a more complex model and attempts to simplify it. We found this approach feasible most to non-expert modelers which tend to build complex models first [1]. Still the reduction algorithm obeyed the four principles listed in Section 3: (i) it depends upon the simulation objectives, (ii) it is performed at conceptual phase of simulation modeling, (iii) it guarantees the validity of the simpler model since it eliminates only the attributes that do not interfere with the desired measures, being classified into type II reduction process. Finally, it can be performed on a computer (although here it was manually made for didactic purposes, the algorithm could be written in any programming language, since it consists of well-defined sequential steps). By this work, we demonstrate the feasibility of the reduction algorithm to a discrete event simulation model described in Condition Specification Technique. In this case we reduced the complexity of the model by omission i.e. skipping some parts of the model that do not interfere with simulation objectives. This is however one possibility of reduction. There are, according to Pedgren et al. [9], other possibilities like substitution and aggregation. Future work will search for techniques that could also deal with these possibilities.