The damage early warning of bridges is a mandatory issue if we want to guarantee the safety of transport lines. In principle, bridges should show a non-linear damaged phase which has enough remaining capacity to avoid brittle collapse, while permitting the identification of the decrease in stiffness, and therefore the warning about the efficiency of the structure (Farrar & Warden 2006). It is however to point out that the change detection of the bridge dynamic properties is a very complicated task, due to the variable environmental and loading conditions, and the small entity of change that by far severe damages produce (Magalhães et Al., .2012). In the paper the preliminary results of a two year long European project, granted by the Infravation Call, are presented. In this period, a consistent number of bridges all around the world have been monitored, and some specimen bridges thoroughly investigated in laboratory (see for instance the project description at the site http://www.infravation.net/projects/SHAPE). The performed investigation has shown that the continuous monitoring of bridges can be transformed in a sort of bridge behaviour intrinsic model which fixes the range of normal operation of the bridge itself. By this way, outliers can be identified, but a very important issue is the evaluation of the extent of the change that can start the warning activity. Among the possible strategies, it is shown that algorithms that evaluate the derivative of the change can be effective in pointing out the onset of a permanent damage (Nikowski & Jain, 2009).