Compact convolutional transformer for bearing remaining useful life prediction

Jin, Zhongtian, Chen, Chong, Liu, Qingtao, Syntetos, Argyrios ORCID: https://orcid.org/0000-0003-4639-0756 and Liu, Ying ORCID: https://orcid.org/0000-0001-9319-5940 2024. Compact convolutional transformer for bearing remaining useful life prediction. Presented at: IWAR 2023- VII International Workshop on Autonomous Remanufacturing, Caserta, Italy, 18-19 October 2023. Published in: Fera, M., Caterino, M., Macchiaroli, R. and Pham, D.T. eds. Advances in Remanufacturing. IWAR 2023. Lecture Notes in Mechanical Engineering Cham, Switzerland: Springer, pp. 227-238. 10.1007/978-3-031-52649-7_18

Preview

PDF - Accepted Post-Print Version Download (553kB) | Preview

Abstract

An accurate prediction of bearing remaining useful life (RUL) has become in-creasingly important for equipment maintenance with the development of monitoring technology and deep learning (DL). Although Transformers are currently the most commonly used unique learning algorithms for sequential data, concerns about their computational efficiency and cost exist. In this re-gard, Compact Convolutional Transformers (CCT) have emerged as a promis-ing alternative that employs sequence pooling and replaces patch embedding with convolutional embedding to enhance computational efficiency while maintaining high prediction accuracy with smaller model sizes. This study proposes an RUL prediction modeling approach that utilizes the Continuous Wavelet Transform (CWT) to transform time-frequency domain features into images, subsequently fed into CCT to establish a highly accurate prediction model for the RUL of bearings. This study conducted experiments using the XJTU-SY rolling bearing dataset. The performance was evaluated in terms of root mean square error (RMSE) and maximum absolute error (MAE) by modi-fying the layer configuration and comparing with other state-of-the-art algo-rithms.

Item Type:	Conference or Workshop Item (Paper)
Date Type:	Published Online
Status:	Published
Schools:	Engineering Business (Including Economics)
Publisher:	Springer
ISBN:	978-3-031-52648-0
Date of First Compliant Deposit:	29 November 2023
Last Modified:	10 Nov 2024 03:30
URI:	https://orca.cardiff.ac.uk/id/eprint/164437

Actions (repository staff only)

Edit Item