Efficient ultrasonic data processing for crack detection in concrete structures

When subjected to high stress over a long period of time, concrete structures may deteriorate, thus having their mechanical characteristics worsened due to the growing of microcracks within them. Detecting the potential presence of such cracks is of fundamental importance to check both their state of health and the conservation one. To this end, non-destructive testing techniques are extremely useful, as they can detect the possible presence of inner defects from the sample’ surface, thus aiding to estimate the mechanical characteristics of the structure at a given moment. Among these techniques, those able to detect cracks when they first appear are highly desired, as they allow consolidation work to be carried out in situations that avoid any danger to the structure and to the users. Our study leverages the structure non-linear response to an ultrasonic excitation due to the presence of cracks: detecting this non-linear behavior with highsensitivity techniques allows for the early detection of the damage growing. This paper addresses this problem with the goal of realizing a net of detection systems, placed in a distributed manner on the structures and endowed with a local processing capacity. This allows the possible occurrence of non-linear behavior to be analyzed and to point it out at an early stage. In the system we propose, each intelligent sensor will therefore be equipped with a system for modelling non-linearities and a neural decision maker. The proposed technique is described in detail and is verified on a set of data collected experimentally in laboratory.