Self-Supervised Learning for data scarcity in a fatigue damage prognostic problem

• Published in: Engineering applications of artificial intelligence Vol. 120; p. 105837
• Main Authors: Akrim, Anass, Gogu, Christian, Vingerhoeds, Rob, Salaün, Michel
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.04.2023; Elsevier
• Subjects: Data scarcity; Deep Learning (DL); Engineering Sciences; Mechanics; Prognostics and Health Management (PHM); Remaining Useful Life (RUL); Self-Supervised Learning (SSL); Structural mechanics; Data scarcity; Self-Supervised Learning (SSL); Prognostics and Health Management (PHM); Deep Learning (DL); Remaining Useful Life (RUL); Data Scarcity
• ISSN: 0952-1976; 1873-6769; 1873-6769
• DOI: 10.1016/j.engappai.2023.105837

With the increasing availability of data for Prognostics and Health Management (PHM), Deep Learning (DL) techniques are now the subject of considerable attention for this application, often achieving more accurate Remaining Useful Life (RUL) predictions. However, one of the major challenges for DL techniques resides in the difficulty of obtaining large amounts of labelled data on industrial systems. To overcome this lack of labelled data, an emerging learning technique is considered in our work: Self-Supervised Learning, a sub-category of unsupervised learning approaches. This paper aims to investigate whether pre-training DL models in a self-supervised way on unlabelled sensors data can be useful for RUL estimation with only Few-Shots Learning, i.e. with scarce labelled data. In this research, a fatigue damage prognostics problem is addressed, through the estimation of the RUL of aluminium alloy panels (typical of aerospace structures) subject to fatigue cracks from strain gauge data. Synthetic datasets composed of strain data are used allowing to extensively investigate the influence of the dataset size on the predictive performance. Results show that the self-supervised pre-trained models are able to significantly outperform the non-pre-trained models in downstream RUL prediction task, and with less computational expense, showing promising results in prognostic tasks when only limited labelled data is available. •Self-supervised learning is investigated to address data scarcity in prognostics.•Fatigue crack propagation prognostics in structures based on strain data.•Synthetic dataset composed of multivariate run-to-failure strain time series.•Self-supervised pre-trained models improve RUL estimation performance. [Display omitted]