Addressing the Issue of Insufficient Information in Data-based Bridge Health Monitoring

Download Final Report

CAIT project no.: CAIT-UTC-042

Fiscal Year: 2012/2013

Status: Final

Rutgers-CAIT Author(s): Patrick Szary, Ph.D., Rutgers' CAIT

External Author(s): Raimondo Betti, Ph.D., Columbia University

Sponsor(s): Parsons Transportation Group


This project endeavored to develop, investigate, and quantitatively validate techniques that address issues and limitations of a scarcity of measured data in data-based bridge health monitoring.

One of the most efficient ways to solve the damage-detection problem using statistical pattern recognition is to exploit the methods of outlier analysis. Cast within the pattern recognition framework, damage detection assesses if patterns of damage-sensitive features that are extracted from the system response under unknown conditions depart are different from those drawn by the features extracted from the system response in a healthy state.

The metric dominantly used to measure the testing feature’s departure from the trained model is the Mahalanobis Squared Distance (MSD). Evaluation of MSD requires the use of the inverse of the training population’s covariance matrix. It is known that when the feature dimensions are comparable to the number of observations, the covariance matrix is ill-conditioned and numerically problematic to invert. When the number of observations is smaller than the feature dimensions, the covariance matrix is not even invertible.

In this work, four alternatives to the canonical damage-detection procedure were investigated: 1) data compression through Discrete Cosine Transform, 2) use of pseudo-inverse of the covariance matrix, 3) use of shrinkage estimate of the covariance matrix, and 4) a combination of the three techniques.

The performance of the four methods was first studied for solving the damage identification problem on simulated data from a four DOFs shear-type system, and on experimental data recorded on a four-story steel frame excited at the base using the shaking table facility available at the Columbia University’s Carleton Laboratory.

Finally, the proposed techniques also were investigated in the context of damage location applications on simulated data from a bridge deck model.