作者: Na Qiu , Yunkai Gao , Jianguang Fang , Guangyong Sun , Qing Li
DOI: 10.1016/J.TWS.2018.05.002
关键词: Surrogate model 、 Crash simulation 、 Optimal design 、 Uncertainty quantification 、 Reliability (statistics) 、 Control theory 、 Crashworthiness 、 Polynomial 、 Noise 、 Computer science
摘要: Abstract Due to the expensive cost of full-scale tests, more and designs rely on simulation. For highly nonlinear crash simulation, numerical uncertainty is an inherent by-product, which refers oscillation results when simulation repeated at same design or variables are slightly changed. This directly influences quality reliability optimal design. paper shows how these issues can be addressed by proposing a simple quantification method for (noise) surrogate model (error) in optimization process. Three engineering problems, tube crush example, automotive front-rail example multi-cell structure used illustrate this method. Firstly, level quantified terms noise frequency amplitude, convergence study two criteria employed determine appropriate data size describe noise. Secondly, estimation considering both error proposed based prediction variance polynomial response surface. Finally, front rail structures optimized according It was found that sources uncertainty, reliable than deterministic solutions.