| Popis: |
In the framework of structural health monitoring, a ‘model-free’ approach has been gaining increasing attention to describe the structural behavior without building a numerical model but adopting the so-called artificial neural networks (ANNs) that must be trained on data (e.g., accelerations) recorded on the existing structure. However, the lack of complete ‘real-life’ applications is the biggest current shortcoming to the use of ANNs. Indeed, the application of ANNs is often limited to medium-scale structures. The few applications on full scale structures are rarely run for a time interval sufficient to recognize structural deterioration due to material degradation and/or damage due to external loads. Moreover, the formulation of a methodology to design the sensor network serving a neural network is necessary. |
