Prediction of Inconel 718 roughness with acoustic emission using convolutional neural network based regression

© 2019, Springer-Verlag London Ltd., part of Springer Nature.Acoustic signals have valuable information and can complement mechanical signals (e.g., effort, roughness, and optics) since both of them have a good correlation. Furthermore, acoustic signals have non-invasive nature. In this work, roughness characterization via acoustic emission, along with the subsequent roughness detection based on convolutional neural networks, is proposed. Results show reliable and adequate roughness measurement via acoustic emission, and convolutional neural networks performance reached an accuracy of 88 % with a mean square error of 3.35 %. The main contribution of this work is the demonstration of deep learning network feasibility on roughness identification, where no previous signal processing is required and which moves towards a highly robust pattern recognition system.

Prediction of Inconel 718 roughness with acoustic emission using convolutional neural network based regression Academic Article in Scopus