Anomaly Detection and Fault Disambiguation in Large Flight Data: A Multi-modal Deep Autoencoder Approach

Kishore K. Reddy, Soumalya Sarkar, Vivek Venugopalan, and Michael Giering
Submission Type: 
Full Paper
AttachmentSizeTimestamp
phmc_16_026.pdf1.27 MBAugust 26, 2016 - 8:54am

Flight data recorders provide large volumes of heterogeneous data from array of sensors on-board to perform fault diagnosis. Challenges like large volumes of data, lack of labeled data, and an increasing number of sensors (multiple modalities) add to the problems of being able to hand-craft the features needed for the state-of-the-art PHM algorithm to effectively perform system diagnosis. In this paper, the authors propose leveraging existing unsupervised learning methods based on Deep Autoencoders (DAE) on raw time series data from multiple sensors to build a robust model for anomaly detection. The anomaly detection algorithm analyzes the reconstruction error of a DAE modeled using nominal scenarios. The reconstruction error of individual sensors is examined to perform fault disambiguation. Training and validation are conducted in laboratory setting for various operating conditions. The proposed framework does not need any hand-crafted features and uses raw time series data. Our approach is tested on data from NASA open database and demonstrate high fault detection rates (97.8%) with zero false alarms. Our paper also demonstrates robust fault disambiguation on two different fault scenarios. Moreover, the paper provides a strong rationale for utlizing deep architecture (multi-hidden-layer neural network) via thorough comparison with a single hidden-layer DAE.

Publication Year: 
2016
Publication Volume: 
7
Publication Control Number: 
026
Page Count: 
8
Submission Keywords: 
deep learning; fault diagnostics; flight data; condition monitoring; electro-mechanical actuator;
Submission Topic Areas: 
Data-driven methods for fault detection, diagnosis, and prognosis
Industrial applications
Submitted by: 
  
 
 
 

follow us

PHM Society on Facebook Follow PHM Society on Twitter PHM Society on LinkedIn PHM Society RSS News Feed