Self-Organizing Neural Network for Identification of Natural Modes

Abstract
The application of artificial neural networks in civil engineering is gaining momentum. Most of the applications, however, use a feedforward network with a back-propagation algorithm. There is a variety of alternative network configurations and learning algorithms that should not be overlooked. This note draws attention to a self-organizing network that has potential applications in classification problems. The network is self-organizing and uses an unsupervised learning algorithm. Therefore, it continuously adapts to the input patterns. The network can be very useful in the solution of inverse problems that require reconstruction of the structure from measured response data (e.g., health monitoring of structures). In this note, the network has been employed to predict the natural mode shapes of building frames with a varying number of stories. Unlike the conventional tools that require a priori knowledge of the problem, the network classifies purely intuitively. Furthermore, the network is noise tolerant. This feature is extremely useful when the input is measured at site and may contain varying levels of noise.

This publication has 7 references indexed in Scilit: