Structural health monitoring (SHM) can be defined as the process of developing and implementing structural damage detection strategies. Ideally, this detection should be carried out in real-time before damage reaches a critical state and impairs structural performance and safety. Hence, it must be based on sensorial systems permanently installed on the target structures and on fully automatic detection methodologies.
The ability to detect damage in real-time is vital for controlling the safety of old structures or for post-retrofitting/post-accident situations, where it might even be mandatory for ensuring a safe service. Under these constraints, SHM systems and strategies must be capable of conducting baseline-free damage identification, i.e. they must not rely on comparing newly acquired data with baseline references in which structures must be assumed as undamaged.
The present paper describes an original strategy for baseline-free damage detection based on the application of artificial neural networks and clustering methods in a moving windows process. The proposed strategy was tested on and validated with numerical and experimental data obtained from a concrete cable-stayed bridge and proved effective for the automatic detection of small stiffness reductions in single stay cables as well as the detachment of neoprene pads in anchoring devices, requiring only a small number of inexpensive sensors.