Structural health monitoring (SHM) involves continuously surveilling the performance of structures to identify progressive damage or deterioration that might evolve over time. Recently, machine learning (ML) algorithms have been successfully employed in various SHM applications, including damage detection. However, supervised ML algorithms often require labelled data for multiple possible damage states of the structure for successful damage identification. Although it may be feasible to gather such data for low-value structures, obtaining damage data for expensive structures such as aircraft could be highly challenging. Herein, this data insufficiency is addressed by combining Finite Element (FE) models with domain adaptation, specifically transfer component analysis (TCA) and joint domain adaptation (JDA). The proposed methodology is showcased in two case studies, a Brake–Reuß beam, where damage scenarios correspond to different torque settings on a lap joint and a wingbox laboratory structure where damage is introduced as saw-cuts. Supervised learning algorithms in the form of Artificial Neural Networks (ANNs) and K-Nearest Neighbours (KNNs) are trained based on FE data after domain adaptation is applied and are then tested with the experimental data. It is shown that even though the performance of classifiers in distinct scenarios of dual, three, four and five-class cases is sensitive to choices in the training stage, the use of TCA or JDA allows for the use of FE data for training and significantly reduces the need for expensive experimental damage data to be used for training. These results can pave the way for a broader use of ML algorithms in SHM of critical and/or expensive structures.

Machine learning has been successfully applied to many structural health monitoring (SHM) projects. However, it relies heavily on data from structures. Particularly, if supervised learning approaches are employed, then data from all possible damaged states of the structure will be required. For inexpensive structures, destructive means of acquiring those data under laboratory conditions may be possible, but for more expensive structures it may become prohibitively expensive, and other approaches will be required. Recently, generative machine learning models have been used to create synthetic data to create or augment databases and provide an alternative solution to the lack of training data. The current paper explores the use of generative adversarial networks (GANs) for the creation of synthetic data from different damaged states and their suitability for SHM. The approach is applied to a laboratory structure, a nonlinear Brake-Reuß beam where the damage scenarios correspond to different torque settings in the bolts of a lap-joint.

Structural Health Monitoring (SHM), is a very active research field, aimed at producing methodologies for the periodic, and often online, assessment of structures. While there are different approaches, and perhaps definitions for SHM, there is one common necessary element if applied in practice: data acquired from sensors. In an ideal scenario, sensors could be deployed anywhere on a structure of interest, and could be added retrospectively, if not included in the original design. In practice, there are limitations to the availability and installation of sensors, such as physical access and cost. Virtual sensing has thus been proposed as a solution to the problem of sensor availability and different approaches have been applied in various fields. This paper explores and compares two approaches for virtual sensing in structural dynamics: Kalman filters and Gaussian process regression - with the ultimate purpose of assessing their suitability for SHM. The approaches are demonstrated on data from a laboratory structure: a nonlinear Brake-Reuß beam.