A deep reinforcement learning-based fault diagnosis algorithm for unlabeled and imbalanced data

Mai Li; Wei Dong; Guohua Zhang; Weihua Yang

doi:10.1117/12.3026147

1 April 2024 A deep reinforcement learning-based fault diagnosis algorithm for unlabeled and imbalanced data

Mai Li, Wei Dong, Guohua Zhang, Weihua Yang

Proceedings Volume 13082, Fourth International Conference on Mechanical Engineering, Intelligent Manufacturing, and Automation Technology (MEMAT 2023); 130820F (2024) https://doi.org/10.1117/12.3026147
Event: 2023 4th International Conference on Mechanical Engineering, Intelligent Manufacturing and Automation Technology (MEMAT 2023), 2023, Guilin, China

Abstract

Aiming at the problems that most of the rotating machinery fault diagnosis algorithms are oriented to labeled data, the cost of label information collection is high and the number of fault samples is far less than that of normal samples, this paper proposes a deep reinforcement learning based rotating machinery fault diagnosis method for unlabeled and imbalanced data. The method leverages the relationship between samples and cluster centers to provide feedback in the form of reward information. It employs mechanical vibration signal samples as model state inputs and fault type selection as selectable actions for the agent. An interactive environment is constructed, allowing the agent to observe, act, and receive rewards in the absence of fault labels. Additionally, a deep neural network is utilized to approximate the Q function, which is then maximized to obtain the optimal policy, enabling fault diagnosis in the absence of labeled data. Through validation with rolling bearing fault data, the proposed method demonstrates a 15% improvement in diagnostic accuracy compared to the K-Means algorithm when dealing with imbalanced data.

(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Mai Li, Wei Dong, Guohua Zhang, and Weihua Yang "A deep reinforcement learning-based fault diagnosis algorithm for unlabeled and imbalanced data", Proc. SPIE 13082, Fourth International Conference on Mechanical Engineering, Intelligent Manufacturing, and Automation Technology (MEMAT 2023), 130820F (1 April 2024); https://doi.org/10.1117/12.3026147

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available