A simulation expertsystem based approach

[attachment=16170]

Introduction
This research is interested in one of the major design issues of production systems
(PSs): the resource sizing problem. It is defined as the specification of the number of
each type of resources to be used in a production process for a given time period (Miller
and Davis, 1977). Sizing is required while designing a new or expanding an existing
system (Feyzioglu et al., 2005). The approaches that tackled this problem can be
classified in two principal categories: analytical and simulation-based.


Proposed approach

Three main types of information are required for the application of the SESA: PS data,
demand pattern and performance limits (Table I).
The simulation tool uses the PS data and demand pattern to simulate the realization
of a typical set of MOs over a given planning horizon. Simulation results are then
considered as performance measures of the system. These results, in addition to the
performance limits and relevant PS and demand pattern data constitute the ES
required inputs. The ES is in charge of analyzing the PS situation. If the simulated
system performance is found to be improvable, the ES recommends a modification to
its resources in order to overcome the problem considered to be responsible, at the
largest extent, for the low performance. Consequently, a new cycle is run until the ES
becomes unable to suggest any modifications (Figure 1). Finally, it is worth noting
that the approach can be started from any initial PS configuration assuring the

feasibility of all MOs.

Production system modeling for simulation
The production of the typical MO pattern by the PS being sized was modelled for
simulation using the commercial tool arena (User’s Guide, 2002). The model involves
three main components discussed in the following subsections (Figure 2(a)).



Expert system
The developed ES is an object-oriented decision-making tool. It is composed of four
main parts. First, the object base is the “static knowledge” component. The objects are
organized hierarchically into classes and sub-classes representing all problem elements
such as global PS data and departments of machines (Abel and Abel, 1988). Thus, the
department i object is a sub-class of the class “Departments of machines” which is in
turn a sub-class of the “Resources” class. Besides, the rule base is the ES component
representing the “know-how”. This expertise is expressed in terms of inference rules of
the form: “IF [condition] THEN [action]” grouped in several packs, each representing
one of the main functions of the ES (Figure 3)