Quantification and Management of Model Uncertainty in Simulation-Based Vehicle Design
Tuesday, February 23, 2016 - 4:35pm - 5:00pm
Zhen Jiang (Ford Motor Company)
Computer simulations, often much less expensive than physical tests, have been widely used for design and optimization in many engineering fields. However, computer simulations never agree completely with physical tests, because of the bias due to the physical phenomena not fully captured in the modeling process, and because of the test variations due to variability of parts and operational conditions. This talk will summarize the typical challenging issues associated with quantification of the aforementioned uncertainty for simulation-based vehicle design. Furthermore, model uncertainty should be investigated in the context of multidisciplinary design optimization (MDO), which is important and yet notably difficult due to the complexity of multidisciplinary simulations and interdisciplinary couplings. In that regard, this talk will discuss the impact of uncertainty in MDO collaboration, hoping to facilitate the decision making regarding additional data collection activities (usually referred to as “resource allocation”) for uncertainty reduction.