S. J. Hudak, M. P. Enright, R. C. McClung, Southwest Research Institute, San Antonio, TX; H. Millwater, University of Texas at San Antonio, San Antonio, TX
The development and implementation of Prognosis Systems has the potential to significantly enhance the reliability and readiness of high-value assets, while concurrently decreasing sustainment costs. This process includes the acquisition and fusion of on-line sensor information, combined with physics-based models for damage accumulation, and higher order reasoning for decision making. This paper will summarize and demonstrate several prognosis-enabling technologies. First, an advanced fracture mechanics model, which explicitly treats crack nucleation, small crack propagation, and large crack propagation, is described. The utility of this model is demonstrated for predicting fatigue life in the presence of high stress gradients associated with beneficial surface treatments, including shot peening and low-plasticity burnishing. The development and laboratory demonstration of a novel thin-film magnetostrictive sensor for detection and monitoring crack propagation is also summarized. Results from laboratory experiments that monitored the propagation of a surface crack in a Ti-6Al-4V specimen using the thin-film sensor are used to formulate a statistical model for sensor uncertainty. This sensor uncertainty model is also combined with probabilistic simulation to assess the potential benefits of embedded sensors for on-line detection and monitoring of defects, as compared to the more traditional depot inspections. The benefits of using Bayesian statistics to fuse continual sensor inputs with life prediction models for predicting the current state of fatigue damage in fracture-critical components is also demonstrated.
Summary: A new probabilistic fatigue life prediction model that includes crack nucleation, small crack growth and large crack growth is presented. The model has advantages for situations where steep stress gradients are present -- beneficial surface treatments, contact fatigue, and thermal stresses. The development and demonstration of a novel thin-film sensor is also summarized and laboratroy measurements with the sensor are used to quantify measurement uncertainty.Probabilistic simulations are used to demonstrate the benefits of fusing results from embedded sensing and life prediction to forecast component reliability.
