Neural-network based structural health monitoring with wireless sensor networks

Wireless sensor networks, which enjoy increasing interests in the field of personal and ubiquitous computing, have recently been regarded as a promising candidate for structural health monitoring. This paper proposes a neural-network based structural health monitoring scheme by using wireless sensor networks, in which hundreds of sensor nodes perform distributed sensing and collaborative computing for structural health analysis. Since vibration frequencies of architectural or mechanical structures imply their health status, structural health problems can be detected through processing vibration frequency data collected by sensors. Artificial neural-network based machine learning algorithm is then employed for data processing. After constructing samples of training data in healthy and unhealthy status respectively, data coming from sensors are tested and classified into a certain category which indicates health status of monitoring structure. Simulation results demonstrate that the proposed scheme achieves not only higher accuracy of structural health monitoring but also more robust performance against environmental noises and interferences, compared with some existing methods.