BAYESIAN KRIGING REGRESSION FOR THE ACCURACY IMPROVEMENT OF BEAM MODELED T-JUNCTIONS OF BUSES AND COACHES STRUCTURES WITH A METHODOLOGY BASED ON FEM BEHAVIORAL ANALYSIS |
F. B. ROMERO, J. M. MCWILLIAMS, E. A. FAZIO, F. A. IZQUIERDO |
University Institute of Automobile Research of the Madrid Community |
|
|
|
ABSTRACT |
This paper has been developed in the framework of the alternative beam T-junction solution previously propoused by the authors (Alcal? et al., 2013), with the scope of optimizing the behavior of buses and coaches upper structures modeled with beam type elements. The alternative beam T-junction model proposed by the authors, had a total of six elastic elements at the junction level allowing to modify the localized rigidity of any modeled T-junctions, therefore improving their behavior and avoiding the well known rigidity issue these elements have. A fundamental aspect behind the use of these alternative beam models is related to the necessity of correctly estimating the rigidity values of the elastic elements for each modeled T-junction. In this context, we propose applying a surrogate model for the reference calculations which is based on statistical Bayesian kriging predictors. Statistical predictions have the added value, with respect to deterministic solutions, of providing a quantification of uncertainty. The development of the kriging predictors has required the application of a statistical methodology including computer experiment design, computation of moments of inertia, graphical and ANOVA type sensitivity analysis and Monte Carlo computation of Bayesian inference. The results of the application have been a very satisfactory trade-off between accuracy of approximation (prediction) and computational cost. The surrogate kriging models also provide an useful tool for a better understanding of the input-output relationships involved in the reference computations. |
Key Words:
Structural optimization, Bus structural analysis, FEM, Kriging |
|