Principal Component Analysis and Artificial Neural Network Framework for Damage Detection Strategy under Varying Operational Loading Conditions

One of the key challenges in developing a robust damage detection system for aircrafts in service is the capability to discriminate the effects of operational and environmental variations from damage-sensitive features. This work aims to shed some light in normalizing the vibration test data under the effects of operational loading variables analyzed in frequency response function (FRF) by means of standard Principal Component Analysis (PCA) and fused in with Artificial Neural Network (ANN) framework. PCA is applied in the vibration data set acquired from a wing box structure with an attached liquid tank in aim to reduce the dimensionalities as well as for visualization technique. The step is essential in interest to examine the correlation between both load and damage parameters with the key point of damage and load identification. To develop a robust damage novelty detector technique, the PCA is incorporated with the forward feed two-layer Artificial Neural Network (ANN) with an optimum numbers of perceptron. The output from incorporating both PCA and twolayer ANN has shown encouraging and positive novelty damage detection results under varying load and damage parameters.