DAMAGE DETECTING SYSTEM USING MULTI-PATTERN DIFERENTIATION NEURAL NETWORK WITH PARTICLE SWARM OPTIMIZATION AND APPLICATION FOR CONCRETE MATERIAL DEGRADATION

Abstract

This paper propses a new multi-pattern differentiation neural network system which can differentiate the multi-pattern mixed data by using Particle Swarm Optimization, and deal with its application for detecting fatigue damage of concrete material. The partical swarm optimization is served as both a pattern clacification and function fitting in this method instead of the back propagation method of conventional method. This modification make it possible to control the classification objective as one likes. This new system can detect the new pattern of input-output relationship caused by fatigue damage clearly compared to the conventional multi-pattern differentiation neural network system.