Abstract
A critical issue for structural health monitoring (SHM) strategies based on pattern recognition models is a lack of diagnostic labels to explain the measured data. In an engineering context, these descriptive labels are costly to obtain, and as a result, conventional supervised learning is not feasible. Active learning tools look to solve this issue by selecting a limited number of the most informative observations to query for labels. This work presents the application of cluster-adaptive active learning to measured data from aircraft experiments. These tests successfully illustrate the advantages of utilising active learning tools for SHM, and they present the first application/adaptation of active learning methods to engineering data — a MATLAB package is available via GitHub: https://github.com/labull/cluster_based_active_learning.
Highlights
Structural health monitoring involves the observation of a structure or mechanical system over time using periodically spaced measurements [1]
Clustering techniques are a family of algorithms that work with unlabelled data by finding K groups/clusters of similar observations within the feature space
Results show that using the Dasgupta and Hsu (DH) learner provides a significant increase in classification performance, for lower query budgets
Summary
Structural health monitoring involves the observation of a structure or mechanical system over time using periodically spaced measurements [1]. As digital storage gets cheaper, and sensing devices develop, it has become much easier to collect large datasets that may be indicative of system health Using this resource enables the data-based approach to SHM, which focuses on machine learning and pattern recognition algorithms for black/grey-box modelling as a means of diagnosis and prognosis. While these datasets may be large, comprehensive labelling is rare; investigating diagnostic labels in an engineering context is often impractical and expensive, as it is infeasible to damage structures (such as bridges or wind turbines) to obtain labelled data for the damaged states of health. This work focusses on active learning as another variation of partially-supervised pattern recognition [2]
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
