|
White Paper: Simulation and Modeling Introduction A simulation model attempts to reproduce a system's core or important behaviors through the interconnections of intelligent components in a digital computer. Building the simulation model involves three important components: parameter estimation, computer model evolution, and model execution and analysis. The following figure illustrates the life cycle of building and running a computer simulation model.  Building the Simulation Model The crucial part of the design process is parameter estimation where we determine the actual nature and characteristics of the variables used in the model.
In an intelligent model, parameter estimation involves a knowledge acquisition phase
where the rules governing the system's behavior are extracted from both the SMEs and
the policies governing the procedures of the problem domain.
Discrete
Event Simulation Our approach to implementing a knowledge based simulation
is through a fusion of traditional stochastic modeling and advanced expert systems
technologies. This brings together the best worlds of quantitative analysis and artificial
intelligence. Our hybrid model approach has been used to solve many extremely difficult
problems that have not been particularly well served by conventional mathematical
techniques. The following figure shows the schematic organization of this hybrid model.  The Discrete Event Modeling Environment In the simulation model random number generators are used to create populations of objects or events common to the problem domain. A clock then starts the process running with a series of events. The clock continues until no more events are available. From this simulation we collect all the operational statistics and evaluate the model performance. Knowledge Based ApproachThe approach we take is through the fusion of both discrete event simulation and advanced techniques in computational intelligence (rule-based inference engines and fuzzy logic). Computational Intelligence, a broad form of artificial or machine intelligence, is concerned with capturing the behavior of a system through rules and semantics. Rules describe the behavior and semantics define the meanings of entities in the model. These semantics are captured in the form of fuzzy sets. Fuzzy sets allow us to represent and use imprecise or vague concepts in the model in a manner analogous to the way we think about processes in the real world. With our approach of developing a knowledge based fuzzy logic solution we avoid the harsh boundary changes that occur when Boolean parameters are used. Boolean parameters are well-known for there difficulty in representing semantic concepts and parameters, an area in which fuzzy logic excels. As a quite simple example, we can represent the concepts of a "lengthy manufacturing process" and a "high priority process" in two fuzzy sets (a complete discussion of the practical application of fuzzy logic can be found in Earl Cox's books on the subject). The following figure shows how these ideas are captured.  Semantic Concepts as Fuzzy Sets
Whitepaper content: © 1997 The Metus Systems Group, Used by Permission For more information
E-Mail: FuzzyQuery@Sonalysts.com © 2003 Sonalysts Inc. All rights reserved. |