Intelligent discrimination of failure modes in thermal barrier coatings: wavelet transform and neural network analysis of acoustic emission signals

Abstract

To identify failure modes in thermal barrier coatings (TBCs), we propose a method of processing acoustic emission signals based on the wavelet packet transform and neural networks. The results show that there are four typical failure modes in TBCs: surface cracks, sliding interface cracks, opening interface cracks, and substrate deformation. These failure modes can be discriminated by the wavelet energy coefficients that parameterize their characteristic frequency bands. By using the energy coefficient vector as an input, the back-propagation neural network has a self-learning ability to cluster signals with the same order features. In comparison with experiments, this processing method is effective for intelligently discriminating the failure modes of TBCs.