Papers »

Spatial damage detection and localization using compressive sampling

Recent advances in sensing technology facilitates application of dense sensor arrays in structural health monitoring (SHM) projects. Today, wired and wireless contact sensors as well as noncontact digital camera lenses can be utilized to measure structural response quantities with high resolution in time and space. In order to infer about structural health condition, these measured signals are processed through various SHM algorithms, one group of which is data-driven damage detection methods. Therefore, along with the hardware improvements in the SHM field, it is vital to enhance the efficiency and scalability of SHM algorithms as well. In an effort toward this goal, this presentation covers development of a data-driven damage diagnosis algorithm based on compressive sampling of damage sensitive features extracted from subsets of sensors available in the network. Scalability of this algorithm is ensured through feature extraction on the basis of neighboring sensors, as well as efficient methods of data sampling. In the proposed framework, a group of sensors are sampled over the entire sensor grid. The data measured through the selected sensors are processed through change point analysis and support vector machines to provide an initial estimate for location of damage. This estimate forms the basis for efficient selection of more sensors from the network. As more sensor nodes are added to the processing subset, a new estimate for damage boundary is provided by updating the previous model of support vectors. Performance of the proposed methodology is evaluated using a finite element model of a gusset plate connection with damage simulated in form of cracking. The gusset plate is subjected to seismic loading and strain distribution of the plate pre- and post-damage is generated and used for damage identification and localization.

Author(s):

S. Golnaz Shahidi    
Lehigh University
United States

Shamim N. Pakzad    
Lehigh University
United States

Jamie M. Hudson    
Lehigh University
United States

M. Mohsen Moarefdoost    
Lehigh University