Defect Identification by Sensor Network Under Uncertainties

This paper presents a theoretical framework for identification of defects by a sensor network under uncertainties. While location of sensors are not known due to their inspection due to limited knowledge on the structure to be inspected, existing inspection methods do not take uncertainties of sensor locations into account for the localization of defects. The proposed theoretical framework formulates the uncertainties of sensor states stemming from both motion and measurement and allows stochastic identification of defects using recursive Beyesian estimation. Multi-sensor belief fusion further allows a network of sensors to jointly identify defects and improve the accuracy of identification. Parametric studies and application to practical defect identification have shown the validity of the proposed framework.