Structural health monitoring and condition based fatigue damage prognosis of complex metallic structures.

摘要

Current practice in fatigue life prediction is based on assumed initial structural flaws regardless of whether these assumed flaws actually occur in service. Furthermore, the model parameters are often estimated empirically based on previous coupon test results. Small deviations of the initial conditions and model parameters may generate large errors in the expected dynamical behavior of fatigue damage growth. Consequently, a large degree of conservatism is incorporated into structural designs due to these expected uncertainties. The current research in the area of Structural Health Monitoring (SHM) and probabilistic fatigue modeling can help in improved fatigue damage modeling and remaining useful life estimation (RULE) techniques. This thesis discusses an integrated approach of SHM and adaptive prognosis model that not only estimates the current health, but can also forecast the future health and calculate RULE of an aerospace structural component with high level of confidence. The approach does not assume any fixed initial condition and model parameters. This dissertation include the following novel contributions. 1) A Bayesian based off-line Gaussian Process (GP) model is developed, which is the core of the present condition based prognosis approach. 2) Different passive and active SHM approaches are used for on-line damage state estimation. Applications of passive sensing are shown to estimate the time-series fatigue damage states both under constant and random fatigue loading. It is found that there is a good correlation between estimated damage states and optically measured damage states. In addition, applications for both narrow and broadband active sensing approaches are presented to estimate smaller incipient damage. It is demonstrated that the active sensing techniques not only can identify smaller incipient damage but also can quantify fatigue damage during all the three stages (stages I , II, and III) of fatigue life. 3) An integrated on-line SHM and off-line GP predictive model is developed for real-time condition based damage state estimation of complex Aluminum structures under fatigue loading. It is found that the proposed technique can forecast the future damage states well before the final failure.