Preprocessing of Massive Flight Data Based on Noise and Dimension Reduction

Abstract

The integrated modular avionics system is an important part of modern aircraft. It will generate massive and diverse types of flight data during its operation. When we want to extract useful information from flight data for aircraft system fault diagnosis and life prediction, the noise and data redundancy are the first problem to confront and it will affect the selection of feature parameters. For this reason, this article chooses the wavelet threshold de-noising method to reduce the noise of raw data, and then uses the principal component analysis method to reduce the dimensionality. The results illustrate that the effects of de-noising and dimensionality reduction are effective.