Optimization of the neural RBF classifier for the diagnostics of electronic circuit

Bartosz Polok, Piotr Bilski

Abstract

The paper presents the detailed analysis of the Radial Basis Function (RBF) neural network application for diagnostics of analog systems. In most cases RBF networks are used in the approximation tasks. In this work the network is used as the fault classifier. Because RBF networks are known to be dependent on the size of training data, the procedure to minimize the amount of the learning examples was proposed. The classifier is tested on the model of the electronic filter, which, was also diagnosed by alternative methods, such as Multilayered Perceptron and Support Vector Machine. Experimental results show what are advantages and drawbacks of the RBF classifier compared to other solutions.

Download

IMEKO-TC10-2017-020.pdf

DOI

-

IMEKO TC

TC10 - Measurement for Diagnostics, Optimization and Control

Back to the proceedings