58480 - A Reduced Order Modeling Approach to Probabilistic Creep-Damage Predictions in Finite Element Analysis 

This paper introduces a computationally efficient Reduced Order Modeling (ROM) approach for the probabilistic prediction of creep-damage failure. Component-level probabilistic simulations are needed to assess the reliability and safety of high-temperature components. Current probabilistic creep-damage modeling in finite element (FE) analysis for large geometry is computationally costly. To that end, ROM is developed aiming at reducing computational cost with a controlled loss of accuracy. The present work is devoted to adoption of a Reduced Order Modeling (ROM) approach for component level simulation for larger geometry based on extremum prediction of a 1D geometry to predict the creep deformation, damage, and rupture. A probabilistic Sine-hyperbolic (Sinh) model is chosen and experimental creep data of alloy 304 stainless steel with replicates are gathered from the literature. The probabilistic model includes three sources of uncertainty: test condition (stress and temperature), initial damage (i.e. microstructure), and material properties. Probability density functions (pdfs) are tuned for each source based on ASTM E8 and E139 limits and Anderson-Darling (AD) goodness-of-fit test. The Sinh model programmed into ANSYS finite element software using the USERCREEP.F material subroutine. A series of 1D and 2D simulations are conducted to validate the accuracy and suitability of probabilistic Sinh model in FE analysis. The pdfs of the sources of uncertainties are modeled via well-tried Monte Carlo method. To assess the extremum condition, numerous validation simulations are conducted on 1D geometry. Herein, the concept of ROM is applied on 2D geometry with much smaller sample size for probabilistic prediction to reduce the memory needs and CPU time. The probabilistic creep deformation and damage predictions are compared between the 1D and 2D geometry and assess the implications of ROM approach on FE analysis. The accuracy of the probabilistic prediction employing ROM approach will potentially reduce the time and cost of simulating a 3D geometry. Future studies will introduce multi-stage Sinh, stochasticity, and spatial uncertainty for improved prediction.