On-Line Life Prediction of a Structural Hotspot Available to Purchase

Current aerospace practice follows an engineering model based on damage-tolerant reliability whereby structural components are regularly inspected and replaced. Under this practice, engineering designs are generally based on a physics-based fracture mechanics approach, in which the life of structural component is estimated using an assumed initial damaged condition. However, in a real time environment, keeping track of the damage condition of a complex structural component manually is quite difficult and requires automatic damage state estimation. The real-time damage state information can be regularly fed to a prognosis model to update the residual useful life estimation in event of a new prevailing situation. The present paper discusses the use of an adaptive hybrid prognosis model, which estimates the residual useful life of a structural hotspot using information on the damage condition obtained in real time. The hybrid prognosis model has two modules: an off-line prognosis module that forecasts the future damage state, and an on-line state estimation module, which regularly predicts the current damage state and feeds into the off-line module in real time. Both the off-line and on-line modules are probabilistic models and use the concept of Bayesian inference based on input-output mapping through a Gaussian process.