Artificial Intelligence and Physics-of-Failure Prognostics Methods for Aircraft Maintenance and Reliability

##plugins.themes.bootstrap3.article.main##

##plugins.themes.bootstrap3.article.sidebar##

Published Oct 28, 2022
Thomas Pioger Marcia Baptista

Abstract

In recent years, Physics Informed Neural Networks (PINNs) have shown promising results in Prognostics and Health Management (PHM). A PINN relies on the physical bias inside a Neural Network (NN) to predict engineering equipment’s Remaining Useful Life (RUL). This physical integration can assume three forms: observational bias, inductive bias, or learning bias. This doctoral thesis will focus on inductive bias by creating physical layers inside an NN. This research aims to create a Convolutional Neural Network (CNN) with a modular structure composed of different “lego” layers. Each layer will have a different structure and goal. We will also study how the CNN structure can be modified and optimized according to the case study.

How to Cite

Pioger, T., & Baptista, M. (2022). Artificial Intelligence and Physics-of-Failure Prognostics Methods for Aircraft Maintenance and Reliability. Annual Conference of the PHM Society, 14(1). Retrieved from http://papers.phmsociety.org/index.php/phmconf/article/view/3412
Abstract 2501 |

##plugins.themes.bootstrap3.article.details##

Keywords

PINNs, physical layers, CNN, Machine learning, CNN structure

Section
Doctoral Symposium Summaries