Data-driven prognostics method for turbofan engine degradation using hybrid deep neural network

Powerful sequence modeling capability for massive multi-sensor data enables deep-learning-based methods to obtain accurate remaining useful life (RUL) estimations. Hybrid neural networks, with learned representations based on various networks, have enhanced the prognostics accuracies than single net...

Full description

Saved in:
Bibliographic Details
Published in:Journal of mechanical science and technology 2021, 35(12), , pp.5371-5387
Main Authors: Xue, Bin, Xu, Zhong-bin, Huang, Xing, Nie, Peng-cheng
Format: Article
Language:English
Subjects:
Artificial neural networks
Control
Deep learning
Dynamical Systems
Engineering
Feature extraction
Industrial and Production Engineering
Machine learning
Mechanical Engineering
Neural networks
Original Article
Turbofan engines
Vibration
기계공학
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Powerful sequence modeling capability for massive multi-sensor data enables deep-learning-based methods to obtain accurate remaining useful life (RUL) estimations. Hybrid neural networks, with learned representations based on various networks, have enhanced the prognostics accuracies than single networks. However, assembly strategies that are limited to either parallel or serial, and insufficient utilization of single networks restrict the development of hybrid networks for more complex problems. This paper proposes a data-driven method using hybrid multi-scale convolutional neural network (MSCNN) and bi-directional long short-term memory (BLSTM) network (namely HMCB network) for RUL estimation. The framework of the network includes two parallel paths. One is composed of MSCNN and BLSTM in serial and the other is a BLSTM path. The HMCB network integrates the merits of multi-scale spatial feature extraction of MSCNN and sequence learning capacity of BLSTM. Validated by C-MAPSS dataset, the HMCB network demonstrates noticeably higher prognostic accuracy than other state-of-the-art methods.
ISSN:1738-494X
1976-3824
DOI:10.1007/s12206-021-1109-8