Detection of Induced Damage in Medium-Density Fiberboard Panels Using a Neural Network Method

Abstract

This research assessed the feasibility of using a neural network to detect induced and interior damage to small samples of medium-density fiberboard (MDF). The neural network was a 3-layer back-propagation network. The undamaged stress wave frequency spectrum patterns were used to train the neural network. In a previous study, we successfully used the trained patterns to evaluate low levels of damage in samples of MDF onto which various percentages of their estimated failure loads were applied. In this experiment, after introduction of grooves on the surface or a hole through the center of the samples, a small change in the wave patterns occurred. The neural network has the unique ability to train itself using data to recognize spectral patterns and was successfully used to detect structural damage.