FUZZY REVERSE APPROACH TO SIMULATION METAMODELLING


Galina Merkuryeva, Department of Modelling and Simulation, Riga Technical University, Kalku Street 1, Riga, Latvia
gm@itl.rtu.lv

ABSTRACT:Metamodels in computer simulation are used to substitute a simulation model with computationally more efficient one or facilitate a simulation study using traditional analytical metamodels or advanced computer intelligence techniques such as neural networks, rule learning or fuzzy logic. A fuzzy reverse approach to simulation metamodelling is described in the paper. It provides fuzzy approximation of the input/output relations of the simulation model based only on the data set from simulation experiments. A fuzzy reverse metamodelling algorithm supposes a priori granularization of the output or dependent variables of the simulation model and following partitioning the domains of input variables by training on the simulation data set. Fuzzy reverse approach allows avoid needs of a priori background information about the model underlying functional dependencies. It is also computationally more efficient then so called a direct approach when the input granularization is defined a priori. A numerical example that demonstrates the main principles

KEYWORDS: Simulation metamodelling, fuzzy approximation, reverse approach.

---

Back to HMS2003 Home Page