The interest in the ability to monitor a structure and detect, at the earliest possible stage, any damage to it has been pervasive through the Civil Engineering community, even before the catastrophic collapse of the I-35W Bridge over the Mississippi in the summer of 2007. This was driven largely by the fact that the current manual inspection and maintenance philosophy charged with preventing such failures cannot detect damage in its early stages, and the labor burdens associated with it are extremely heavy. In response, this dissertation proposed a two stage wireless structural health monitoring process, including damage detection and localization, to replace the manual and subjective paradigm. To enhance performance, this dissertation offers a network architecture that is organized into a multi-scale format, with data fusion of decentralized real-time damage decisions based on spatially distributed heterogeneous sensors, operating under a restricted activation scheme and within the computational constraints of the wireless platform with the objective of minimizing intrusion, enhancing the reliability of automated detection, maximizing network lifetime and eliminating the need for strict synchronization and transmission of large amounts of data. Thus the primary research tasks in this dissertation can be summarized as:  Develop a wireless sensor network philosophy that provides reliable data for detection and localization of damage in complex Civil Infrastructure, while maximizing the performance and lifetime of the hardware Develop an assessment framework suitable for damage detection and localization using data measured from a distributed wireless sensors and suitable for operation within said network, i.e., recognizing the computational resources, communications constraints, and power available to the network Analytically and experimentally verify, at various scales and levels of complexity, the proposed network philosophy and assessment framework.  The result of this effort is number of novel contributions achieved through the integrated development of the wireless sensing philosophy, network activation scheme and condition assessment framework to offset inherent limitations of the hardware and optimize performance for the challenging problem of output only, ambient vibration monitoring. These contributions include (1) a Bivariate Regressive Adaptive INdex (BRAIN) for damage detection that proves to be more robust and accurate than previous formats, (2) a Restricted Input Network Activation Scheme (RINAS) with a new image-based vehicle classification algorithm that not only reduces the size of reference databases and enhances detection reliability, but also relieves computational burdens and extends network lifetime and (3) an offline damage localization technique employing Dempster-Shafer Evidence Theory that is capable of effectively isolating damage positions even for minor loss levels. In total, this dissertation offers a definitive step in translating research to practice to advance the notion of ubiquitous sensing to address the 21st Centrury Infrastructure Challenges facing society.