Deep learning for machine health prognostics using Kernel-based feature transformation

Shanmugasivam Pillai () and Prahlad Vadakkepat ()
Additional contact information
Shanmugasivam Pillai: National University of Singapore
Prahlad Vadakkepat: National University of Singapore

Journal of Intelligent Manufacturing, 2022, vol. 33, issue 6, No 6, 1665-1680

Abstract: Abstract Prognostic health management minimizes system downtime and improves overall equipment effectiveness. Accurate prediction of remaining useful life (RUL) is key to prognostics. Prominent machine learning algorithms implement handcrafted feature extraction to improve RUL prediction. Deep learning automates feature extraction from raw data but requires large datasets and computationally expensive fine-tuning. Data-specific handcrafting and fine-tuning limit the generalization capability of existing models. Proposed framework addresses these challenges using Temporal Multivariate 3D Convolutional Network (TM3C) and Kernel-based Transformation (KT) of features. KT generates 3D features that incorporate trendable degradation patterns from multivariate temporal relationship among sensor data. TM3C implements 3D convolutional layers with temporal filters for RUL prediction. KT is generalizable and improves feature relevance. Full-width filters in TM3C reduce number of tunable parameters and convolution operations. Proposed TM3C-KT capitalizes on the strength of deep learning while lowering the cost for feature discovery, parameter learning, and model fine-tuning. TM3C-KT is evaluated on three prognostics applications, (1) RUL prediction for turbofan engines, (2) Failure state estimation for hydraulic pumps, and (3) Component wear prediction for milling machines. Performance of the framework is comparable and better than benchmark methods in literature. Characteristics of the framework are reviewed on generalizability, prognosability and versatility metrics. Results and corresponding analysis demonstrate suitability of TM3C-KT for industrial applications of machine health prognostics.

Keywords: Feature transformation; Deep learning; Prognostic health management; Remaining useful life (search for similar items in EconPapers)
Date: 2022
References: View references in EconPapers View complete reference list from CitEc
Citations: View citations in EconPapers (1)

Downloads: (external link)
http://link.springer.com/10.1007/s10845-021-01747-6 Abstract (text/html)
Access to the full text of the articles in this series is restricted.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:joinma:v:33:y:2022:i:6:d:10.1007_s10845-021-01747-6

Ordering information: This journal article can be ordered from
http://www.springer.com/journal/10845

DOI: 10.1007/s10845-021-01747-6

Access Statistics for this article

Journal of Intelligent Manufacturing is currently edited by Andrew Kusiak

More articles in Journal of Intelligent Manufacturing from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().