PITA FY2014 Projects

High Density Joint Sensing And Compression Design For Structural Damage Detection

PI:Shamim Pakzad, Civil and Environmental Engineering

University:Lehigh University

Co-PI(s):Parvathinathan Venkitasubramaniam, Electrical and Computer Engineering

Industry Affiliate(s):Trilion Quality Systems

Summary:The emergence of very dense instrumentation techniques has provided yet another set of opportunities and challenges in health monitoring of structures for the engineering and scientific communities. Whereas only very important structures were instrumented with sporadic sensor networks for a very high cost only 20 years ago, dense sensor networks today provide the opportunity to install hundreds or thousands of sensors in any structure at very little cost. An example of such technology is Digital Image Correlation systems that provide a nearly continuous sensing grid for measuring response of static and dynamic systems. These very dense sensing systems produce large volumes of data that cannot be processed using the existing analytical methods. The density of information in the data is relatively low, and a lot of structural redundancy exists in them. The existing methods will require memory and computational capacity that is simply not available, or very costly to apply. The objective of this proposal is to develop algorithms for compression of such data for identifying damage in a structural system.

The sensing and processing of measured data in existing research is conducted independently, motivated by the hardware-software distinction in the design. In applications such as Structural Health Monitoring (SHM), however, the large volume and high dimensionality of the data coupled with limited processing power necessitates a new approach that studies sensing and compression jointly. The objective of such an approach is neither to maximize the number or accuracy of sensed measurements, nor to optimally reconstruct compressed data, but to instead infer an underlying phenomenon that drives the measurements, such as the presence and location of a crack in a structure. Our approach is motivated by the fact that in dense sensor networks for structural health monitoring applications, the information about damage is expressed through changes in spatial correlation across sensor measurements. Our objectives in the proposed research are threefold: to derive theoretical performance benchmarks for the proposed designs, to optimize the key parameters of the design subject to the energy constraints and the desired performance level of the underlying damage inference problem, and implement the optimized design on a test bed.