Design of Active Sensor Network and Multilevel Data Fusion

Design of a sensing network, and subsequently, the interpretation of data acquired from sensors, are two critical issues in structural health monitoring (SHM). It is understood that the process to identify structural health status is basically an inverse problem, often mathematically ill-posed. This article has the goal of introducing the active sensor network, a spatially distributed sensing structure, designed to deal with the problem of SHM. Data fusion, especially multilevel data fusion, is then presented to face the challenges in data interpretation, which can be considered as a process to establish a posterior belief about a set of propositions like structural damage events on the basis of a set of prior beliefs that are possessed by sensors. A data fusion process is utilized to increase the robustness and reliability of structural condition identification algorithms by reducing imprecision, uncertainties, and incompleteness. There is a wide range of approaches in data fusion, including the algorithms based on voting scheme, Bayesian theory, Dempster–Schafer rules, fuzzy inference, and artificial neural network. Keywords: active sensor network; system design; data fusion; distributed sensing structure