ABSTRACT
ABSTRACT The state of health of a structure, with control devices, can be assessed from sensor measure ments, but its estimate is very sensitive to any sort of uncertainty. This paper investigates the potentialities offered by neural computing for building a different way of manipulating sensor measurements. The main goals are: 1) to avoid mathematical models for the definition , in the state variable space, of the region where the actuators may remain inactive; 2) to give the system control ’brain’ the possibility of learning from the experience of service.