The monitoring and diagnosing of complex structures like vehicles, buildings, water plants, etc., involve investigation strategies for the accurate detection of anomalies such as defects and cracks.This problem is usually recast as the solution of an inverse problem. In this framework, state of the art methods that provide high performance in terms of resolution exist, but they often require non‐negligible computational resources. As a consequence, real-time investigations are very complex and the monitoring can be performed only periodically. Undesired anomalies may occur in between two consecutive tests and this is unaccepptable in a wide set of applications.
Alternative solutions are required to make the structural health monitoring (SHM) fast and effective, especially if embedded/low-power/low-cost hardware is adopted for data acquisition. Self-learning methods, such as learning-by-example (LBE) methods like Neural Network (NN), Support Vector Machine (SVM), Gaussian Process (GP), provide a fast (even if suboptimal) evaluation of the structural health status, triggering more deep analysis only if required. Such approaches can be profitably integrated on top of sensing hardware platforms, like wireless sensor network (WSN) nodes.
The WSN technology represents a valid solution for low-cost SHM system, thanks to its properties of scalability, flexibility, low-power, multi-sensing, and adaptability to a large number of different application fields. The ELEDIA@UniNAGA is involved in the study, design, and implementation of advanced SHM solutions, as well as in the testing and performance analysis of the investigated systems. Experimental setups and prototype are also under validation.