Abstract

Soft computing methods such as fuzzy logic and artificial neural networks (ANN) have gained popularity in solving engineering problems. Particularly, fuzzy logic and especially developments in uncertainty assessment, enable us to construct and validate precise geoid models. We investigate three different point densities, five variations in the numbers of subsets and five different membership functions when forming the fuzzy model to calculate the geoid heights in the Istanbul (Turkey) area. The results of the fuzzy model are compared with geoid heights obtained using GPS and leveling. The fuzzy model has been verified against the test points. The results indicate that constructing the fuzzy model with a point density of at least one point in 25 km² , carefully selected number of subsets in accordance with point density and a Gaussian membership function, gives superior performance.

Key words: ANFIS, fuzzy logic, type of membership function, number of fuzzy subsets, point density, geoid height.