Application of Neural Networks and Fuzzy Logic to the Calibration of the Seven-Hole Probe

Publisher Website
Google Scholar
Abstract
The theory and techniques of Artificial Neural Networks (ANN) and Fuzzy Logic Systems (FLS) are applied toward the formulation of accurate and wide-range calibration methods for such flow-diagnostics instruments as multi-hole probes. Besides introducing new calibration techniques, part of the work’s objective is to: (a) apply fuzzy-logic methods to identify systems whose behavior is described in a “crisp” rather than a “linguistic” framework and (b) compare the two approaches, i.e., neural network versus fuzzy logic approach, and their potential as universal approximators. For the ANN approach, several network configurations were tried. A Multi-Layer Perceptron with a 2-node input layer, a 4-node output layer and a 7-node hidden/middle layer, performed the best. For the FLS approach, a system with center average defuzzifier, product-inference rule, singleton fuzzifier, and Gaussian membership functions was employed. The Fuzzy Logic System seemed to outperform the Neural Network/Multi-Layer Perceptron.

This publication has 5 references indexed in Scilit: