Compressed Sensing: On-Line Structural Damage Detection

I nearly missed that one:
Julien Cortial, Charbel Farhat, Leonidas Guibas and Manjunath Rajashekhar examined a way for comparing sensor data and computational models on the spot in Compressed Sensing and Time-Parallel Reduced-Order Modeling for Structural Health Monitoring using a DDDAS.

The abstract reads:

This paper discusses recent progress achieved in two areas related to the development of a Dynamic Data Driven Applications System (DDDAS) for structural and material health monitoring and critical event prediction. The first area concerns the development and demonstration of a sensor data compression algorithm and its application to the detection of structural damage. The second area concerns the prediction in near real-time of the transient dynamics of a structural system using a nonlinear reduced-order model and a time-parallel ODE (Ordinary Differential Equation) solver.

This is an interesting match between computational simulations, actual data and sensor network. As reported in the paper, the main issue is really that too many sensors would weigh too much for the plane example. While incipient fault detection is a subject of considerable interest and while it seems that this technique can detect it with less data than the full set, the compressed sensing step still requires the full set of data to be collected and the reconstruction step would take a looong time compared to the dynamics of the plane and may not be an optimal choice of strategy in this case. The hole in that wing reminded me of another one.