Using genetic algorithms to optimize an active sensor network on a stiffened aerospace panel with 3D scanning laser vibrometry data

With the increasing complexity of aircraft structures and materials there is an essential need to continually monitor the structure for damage. This also drives the requirement for optimizing the location of sensors for damage detection to ensure full damage detection coverage of the structure whilst minimizing the number of sensors required, hence reducing costs, weight and data processing. An experiment was carried out to investigate the optimal sensor locations of an active sensor network for detecting adhesive disbonds of a stiffened panel. A piezoelectric transducer was coupled to two different stiffened aluminium panels; one healthy and one with a 25.4mm long disbond. The transducer was positioned at five individual locations to assess the effectiveness of damage detection at different transmission locations. One excitation frequency of 100kHz was used for this study. The panels were scanned with a 3D scanning laser vibrometer which represented a network of 'ideal' receiving transducers. The responses measured on the disbonded panel were cross- correlated with those measured on the healthy panel at a large number of potential sensor locations. This generated a cost surface which a genetic algorithm could interrogate in order to find the optimal sensor locations for a given size of sensor network. Probabilistic techniques were used to consider multiple disbond location scenarios, in order to optimise the sensor network for maximum probability of detection across a range of disbond locations.