Multi-Physics, Multi-scale Ground Vehicle Reliability Prediction

The proposed work will provide a complete vehicle reliability prediction approach and tool. The proposed approach integrates in a high-performance computing environment last advances in engineering analysis and stochastic modeling. It uses fast stochastic physics-based model simulation, including advanced finite element and progressive damage modeling, integrated with refined stochastic response approximation techniques. Bayesian framework is applied for handling modeling uncertainties, lack of data and model updating based on new evidence from test and field observations. A new conceptual “physics-based” Bayesian updating technique is included. The proposed approach includes the key multi-physics, including nonlinear behaviors and complex interactions between damage mechanisms, such LCF-HCF and corrosion-fatigue interactions that could affect significantly predicted failure risks. Multi-scale aspects are integrated from the overall vehicle behavior (using ADAMS), to subsystem and component behavior, and down to local meso-scale fracture mechanics behavior. The heart of the proposed reliability approach is an extremely fast parallel stochastic finite element technology that uses an efficient graph-partitioning and a parallel algebraic multi-grid preconditioner combined with gradient conjugate solvers. The proposed reliability approach will be demonstrated on the HMMWV front-wheel low-control arm subsystem. The effects of modeling uncertainties due to finite element modeling and limited data will be also considered.