We propose an approach to bridge structural monitoring that combines the advantages of a robust structural health monitoring paradigm with those of decentralizing the monitoring apparatus to fleets of vehicles that can continuously store or send data. The objective is to provide accurate, rapid, nearly continuous, and cost-effective assessments of several bridges. The new methodology envisions a set of moving vehicles equipped with sensors (which most of them already possess) able to capture the dynamic interaction between the vehicles and the bridge. As this interaction depends also on the modal characteristics of the bridge, we hypothesize that changes in the dynamic interaction can be inferred from damage related features of the bridge. The methodology couples the sensing system to multi-resolution signal processing and pattern recognition algorithms to capture, locate, and classify variations in structural dynamic properties, e.g. resonant frequencies, mode shapes, or localized stiffness. The proposed approach can be considered as indirect, since it acquires information about the bridge from sensor-equipped vehicles moving over the bridge. This approach needs no installation of equipment on the bridge or any traffic control measurements, being highly distributed and mobile. A key aspect of this project is to assess different classification techniques for determining the presence or absence, and extent of damage.
-
February 2012 - December 2013
-
The impact of our work thus far has been two-fold. First, this grant has allowed us to investigate new signal processing techniques which have advanced the sophisticated yet economical structural health monitoring technique our group has been developing. Second this grant has helped train civil engineering students to learn more about signal processing, while encouraging signal processing students, to examine applications in infrastructure. Our advances in signal processing will be communicated at two conferences this spring, at the 2013 Structures Conference in Pittsburgh PA, and at the 38th International Conference on Acoustics, Speech, and Signal Processing in Vancouver, Canada. The former paper presents the exceptionally good results we have found using sparse representation. The latter presents novel methods for dealing with data sets which are not well labeled, an approach known as semi-supervised learning. Both of these papers would not have been possible without the UTC funding, and both gratefully acknowledge the UTC support.
Name | Affiliation | Role | Position | |
---|---|---|---|---|
jbielak@cmu.edu | Bielak, Jacobo | CEE | PI | Faculty - Tenured |
Type | Name | Uploaded |
---|---|---|
Final Report | BridgeMonitoring.pdf | March 21, 2018, 8:14 a.m. |
No match sources!
No partners!