Neural networks based feature selection approaches for prognostics of aircraft engines
Abstract
The majority of the Prognostics and Health Management (PHM) models proposed during the past few years have shown a significant increase in the amount of data-driven deployments. While more complex data-driven models are often associated with higher accuracy, there is a corresponding need to reduce model complexity. One possible way to reduce the complexity of the model is to use the feature (attribute or variable) selection and dimensionality reduction methods prior to the model training process. In this work, the effectiveness of multiple filter and wrapper feature selection methods (correlation analysis, relief forward/backward selection and etc.), along with Principal Component Analysis (PCA) as a dimensionality reduction method, was investigated. A basis algorithm of deep learning, Feedforward Artificial Neural Network (FFNN), was used as a benchmark modeling algorithm. As a case study, the prognostics of aircraft engine data from NASA Ames prognostics data repository was used to test the effectiveness of the filter and wrapper feature selection methods. The findings show that applying feature selection methods helps to improve model accuracy and significantly reduced the complexity of the models. This study hopes to be a baseline assumption for future complexity reduction solutions that can help in improving the deep learning approaches employed in the PHM domain.
Authors: Phattara Khumprom, Nita Yodo, David Grewell