On-line detection of fatigue cracks using an automatic mode tracking technique

Experimental fatigue tests usually require large testing times. In addition to the resulting increased time-to-market, the large fatigue test time also implies that any structural health monitoring technique that is used should be automatic. When using the modal parameters as damage indicators, an important amount of user interaction is still needed to separate physical poles from computational ones. In this paper, an experimental framework will be developed to automatically track the health of the structure on-line with the performance of fatigue tests. The modal parameters are tracked using a combination of the maximum likelihood estimator and an auto-regressive model. Since confidence levels on the modal parameter are available it is possible to detect if damage is present. In addition, the quasi-static stiffness with computed confidence levels is also used as a damage indicator. The proposed techniques are demonstrated on a steel beam with a propagating fatigue crack.