PI: Shamim N. Pakzad, Lehigh University
Co-PI: Martin Takac, Lehigh University
Mobile sensing is an alternative paradigm which offers numerous advantages compared to the conventional stationary sensor networks. Mobile sensors have low setup costs, collect spatial information efficiently, and require no dedicated sensors to any particular structure. Most importantly, they can capture comprehensive spatial information using few sensors. The advantages of mobile sensing combined with the ubiquity of smartphones with the internet of things (IoT) connectivity have motivated researchers to think of cars and smart phones as large-scale sensor networks that can contribute to the health assessment of structures. Working with mobile sensors has several challenges. The signals collected within vehicles’ cabin is contaminated by the vehicle suspension dynamics; therefore, the extraction of bridge vibration from signals collected within a vehicle is not an easy task. Additionally, mobile sensors simultaneously measure vibration data in time while scanning over a large set of points in space which creates a different data structure compared with fixed sensors. Since collected data is mixed in time and space, it contains spatial discontinuities. When these challenges are addressed, mobile sensing is a promising data resource enabling crowdsourcing and an opportunity to extract information about infrastructure conditions at an unprecedented rate and resolution. This project proposes a deep learning framework specific to mobile sensing to perform system identification and life-cycle assessment of bridge structures.