This paper presents a measurement system design methodology that incorporates damage scenarios. One of the goals of a measurement system is to detect damage for safety and for management of maintenance activities. This study builds upon previous research at EPFL in the area of model based diagnosis. An existing bridge in Switzerland is used to illustrate how damage scenarios are identified and a population of models are generated. A greedy algorithm is used to place sensors such that there is maximum separation between the predictions of the models in the population.