A bayesian artificial neural network method to characterise laminar defects using dynamic measurements

This paper reports on the development of an artificial neural network (ANN) method to detect laminar defects following the pattern matching approach utilizing dynamic measurement. Although structural health monitoring (SHM) using ANN has attracted much attention in the last decade, the problem of how to select the optimal class of ANN models has not been investigated in great depth. It turns out that the lack of a rigorous ANN design methodology is one of the main reasons for the delay in the successful application of the promising technique in SHM. In this paper, a Bayesian method is applied in the selection of the optimal class of ANN models for a given set of input/target training data. The ANN design method is demonstrated for the case of the detection and characterisation of laminar defects in carbon fibre-reinforced beams using flexural vibration data for beams with and without non-symmetric delamination damage.

In situ TiB-reinforced Cu-based bulk metallic glass composites

Cu-based bulk metallic glass (BMG) composites containing in situ TiB particles were successfully fabricated. The reinforcing TiB particles with a size of 5-10 mu m are uniformly distributed in the amorphous matrix. The particles have a good bonding to the matrix with a reaction layer. The BMG composites exhibit an obvious ductility with a plastic strain of 2% for the 17.5 vol.% TiB sample due to the suppression of shear band propagation and the generation of multiple shear bands during compressive testing. The hardness of the materials is increased from Hv543 for monolithic BMG to Hv650 for 23.6 vol.% TiB-containing BMG composite. (c) 2006 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.