Hybrid Twin applied to Structural Health Monitoring

To ensure the proper functioning of a structure, a monitoring during its life cycle is necessary, with the objective of detecting in time any possible anomalies or damage of the structure. To accomplish this, high fidelity numerical models that correctly capture the physics of the system should be developed, so that this model can be further used to achieve a desired design goal, as well as to ensure the longevity of the structure. However, the complexity of real phenomena often makes it impossible for physics-based models to deliver a correct prediction of reality, which limits their use. To overcome this limitation, one solution consists in building a model based on experimental data to ensure correct predictability. Nevertheless, this imposes technical limitations, since obtaining data is often scarce due to limited number of sensors or high costs of experimental campaigns. In this context, hybrid twins emerge as an attractive solution to this problem. Hybrid twins consists in enriching a physics-based model by building an ignorance model, which corrects the predictions of the numerical model. This allows to build a representative model of reality by using a limited number of sensors, since the global behavior of the system is reproduced by the physical model, making the ignorance model to be constructed in a coarse way. In this sense, the present work shows the implementation of a hybrid twin, applied to the monitoring of a structure using Structural Health Monitoring techniques. The performance of the developed hybrid twin is tested on synthetic data, where the hybrid twin built from a simplified physics-based model allows to correct the latter and can be used later to accurately predict damage location on a more complex structure.