Detection of Impact Location and Magnitude for Isotropic Plates Using Neural Networks

Abstract

A neural network-based method of determining the location and magnitude of trans-verse impact events on isotropic plates is investigated experimentally. Time data from four sensors mounted in the corners of an aluminum plate was processed to provide inputs for two backpropaga-tion neural networks. The first neural network was responsible for detecting impact location. After 1 million iterations of training, this neural network was able to locate impacts with an average RMS error of 1.55 radial centimeters on a 58.5 centimeter by 36.8 centimeter (23 inch by 14.5 inch) fully-clamped plate. The second neural network was responsible for impact magnitude detection. This neural network was able to determine the impact magnitude with an average of 13.8% error.