KDnet-RUL: A Knowledge Distillation Framework to Compress Deep Neural Networks for Machine Remaining Useful Life Prediction

KDnet-RUL: A Knowledge Distillation Framework to Compress Deep Neural Networks for Machine Remaining Useful Life Prediction
Title:
KDnet-RUL: A Knowledge Distillation Framework to Compress Deep Neural Networks for Machine Remaining Useful Life Prediction
Other Titles:
IEEE Transactions on Industrial Electronics
Publication Date:
09 February 2021
Citation:
Q. Xu, Z. Chen, K. Wu, C. Wang, M. Wu and X. Li, "KDnet-RUL: A Knowledge Distillation Framework to Compress Deep Neural Networks for Machine Remaining Useful Life Prediction," in IEEE Transactions on Industrial Electronics, doi: 10.1109/TIE.2021.3057030.
Abstract:
Machine remaining useful life (RUL) prediction is vital in improving the reliability of industrial systems and reducing maintenance cost. Recently, long short-term memory (LSTM)-based algorithms have achieved state-of-the-art performance for RUL prediction, due to their strong capability of modeling sequential sensory data.In many cases, the RUL prediction algorithms are required to be deployed on edge devices to support real-time decision making, reduce the data communication cost and preserve the data privacy. However, the powerful LSTM-based methods which have high complexity cannot be deployed to edge devices with limited computational power and memory. To solve this problem, we propose a knowledge distillation framework, entitled KDnet-RUL, to compress a complex LSTM-based method for RUL prediction. Specifically, it includes a generative adversarial network based knowledge distillation (GAN-KD) for disparate architecture knowledge transfer, a learning-during-teaching based knowledge distillation (LDT-KD) for identical architecture knowledge transfer and a sequential distillation upon LDT-KD for complicated datasets. We leverage simple and complicated datasets to verify the effectiveness of the proposed KDnet-RUL. The results demonstrate that the proposed method significantly outperforms state-of-the-art KD methods. The compressed model with 12.8 times less weights and 46.2 times less total float point operations even achieves a comparable performance with the complex LSTM model for RUL prediction.
License type:
Funding Info:
This work is supported by the A*STAR Industrial Internet of Things Research Program under the RIE2020 IAF-PP Grant A1788a0023, and partially supported by the National Key Research and Development Program of China (under Grant 2017YFA0700900, 2017YFA0700903), and National Science Foundation of China (No. 61976200). (Zhenghua Chen is the corresponding authors.)
Description:
© 2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.
ISSN:
0278-0046
1557-9948
Files uploaded:
File Size Format Action
There are no attached files.