A Wireless Mobile Sensor Platform for Structural Health Monitoring

Modern system identification (SID) procedures rely on fixed sensor networks for data collection. Ideally, sensors are fixed at locations and contain profitable structural responses, however, such sensing areas are often limited by the accessibility of the structure and environmental hazards. Not only are fixed sensor placements limited, the data contains restricted spatial information. Mobile sensors simultaneously record data in time while moving in space, so that few sensors collect data containing dense, less-restricted spatial information, providing a more cost-effective solution than a dense array of static sensors. Mobile sensor data contain fundamentally different attributes than fixed sensor data. Such data can be classified as dynamic sensor network (DSN) data, which inherently include spatial discontinuities whenever sensors change position. Despite this challenge, such data can be processed for identification using an iterative machine learning technique Structural Identification using Expectation Maximization (STRIDE). Furthermore, the preservation of spatial information in mobile sensor networks has been quantified throughout simulations: given the same number of sensors, a mobile sensor network produces superior spatial information when compared to a static sensor network. In this paper, ambient vibrations of a simple beam test-bed are measured by a wireless mobile sensor network which includes four parallel lines of motor-driven belts that tow an array of sensor carts along the span of the beam. Feedback between the step motor and a computer was established to achieve a precise spatial grid for the mobile sensors. Modal identification results are presented, documenting the accuracy and feasibility of a realworld mobile sensor network for SID.