Handling Noise-Contaminated Data and Nonunique Identification Results in Wireless Sensor Networks for Structural Health Monitoring

University Of Oklahoma Norman Campus, Norman OK

Investigators

Abstract

PI: Jin-Song Pei, University of Oklahoma This SGER will support exploratory research on wireless sensor networks for structural health monitoring (SHM) with focus on handling noise-contaminated data measurements and nonunique system identification results, the two main issues that have been identified with the most critical and urgent needs in the context of the rapidly growing sensor networks, especially wireless sensor networks. Untested aggressive novel ideas are presented in the proposal while analytical development as well as numerical and experimental validation will be carried out in the project. Therefore, the project is considered as exploratory and high risk. The project research will be undertaken through collaboration between the University of Oklahoma (OU) and Massachusetts Institute of Technology (MIT). The PI at OU will advise a PhD student at MIT to conduct key parts of the research efforts. Experimental study will be conducted at OU with the participation of the PI's team and the student from MIT during each summer while analytical and numerical work will be carried out throughout the two academic years. The main research ideas stem from the PI's successful PhD work at Columbia University on transparent and engineered Artificial Neural Networks (ANNs) with a development plan expanded drastically in both breadth and depth. Serving as a proof-of-concept for the PI's novel ideas on sensor network design and data interpretation, the results will be used to build the PI's academic credentials for her future submission for NSF CAREER grant and participations in future major NSF solicitations such as .Sensors and Sensor Networks.. As an individual member and an institution alternate representative of OU at NEES Consortium, the PI will apply the ideas and methods developed in this project not only to SHM but also to earthquake engineering, especially those NSF NEES related activities. The intellectual merit of the proposed activity includes advancing knowledge on sensor technology and information technology for civil infrastructure health monitoring. Central to this research is to handle real-world situations and uncertainties, namely noise-contaminated data and nonunique system identification results. The attempts are two folds: In terms of data processing and interpretation, novel and ambitious ideas including several types of transparent and engineered Artificial Neural Networks (ANNs) and other techniques (such as improved ERA/OKID, demystified HHT/EMD) will be tested to 1) identify structural nonlinearities, 2) de-noise and 3) overcome nonuniqueness. In terms of sensors and sensor network configurations, existing commercially available sensors will be compared in terms of performance for civil engineering applications with the focus on practical issues related to wireless MEMS sensors applications. New sensor network design methodology will be developed with concentration on solving noise contamination and nonuniqueness. With respect to broader impacts, this SGER project will facilitate cross-institutional and cross-regional communications and partnership by bringing together researchers at OU and MIT. Research results will be disseminated widely in publications and academic conferences. Development of algorithms will be carried out in a generic form so that the research will impact a wide range of applications involving data mining. The proposed research will promote applications of state-of-the-art technologies to traditional fields such as civil engineering. This research will help strengthen the safety of our nation's civil infrastructures and improve preparedness for natural and man-made hazards.