Probabilistic Neural Network Approaches for Autonomous Identification of Structural Dynamics

Abstract

This paper presents an application of Probabilistic Neural Networks (PNN) to identify dynamic responses of structures. The PNN architecture is described and its operation relative to other neural network types is discussed. The instant learning prop erty of the PNN is shown to be a critical advantage for applications where both adaptivity and autonomous operation are required. An example, PNN-based identification of struc tural responses from structural-member strain measurements is presented to illustrate op eration of the network. The network is shown to be capable of classifying dynamics in a spatial-frequency domain very quickly using a small number of active elements.