Workshop on Sampling-Based RBDO
April 19, 2011, Detroit Arsenal, MI, USA

Prof. K. K. Choi, The University of Iowa
Hosted by Dr. David Lamb, U.S. Army TARDEC

Room 40 of Building 215

If you are not affiliated with TARDEC, you must contact Dr. Lamb ( if you wish to attend this workshop. Security clearances must be arranged in advance.
The agenda (pdf) is also available here.


Iowa research team have been working with the U.S. Army TARDEC to develop simulation-based reliability-based design optimization (RBDO) methods and software tools for minimization of the Army ground vehicle weight while improving durability and reliability.  The developed software tools were successfully applied to obtain reliable optimum designs with significantly reduced weight and improved fatigue life of U.S. Army High Mobility Trailer drawbar, Stryker A-arm, and HMMWV A-arm components.  With the success of the Iowa developed RBDO methods and software tools, engineers at TARDEC asked PI and his research team for possibility of extending the RBDO methods that can be used for broader applications beyond the durability.  At the Plenary speech at the Army sponsored Automotive Research Center (ARC) Conference at Ann Arbor on May 10, 2010, TARDEC Director Dr. Grace Bochenek pointed out that newly developed simulation-based design methods would be desirable to be able to support Army in obtaining designs for multi-functionality, such as durability and survivability (safety) for the vehicle structural design, etc.

The PI and his research team started an ARO project in 2009 to develop a sampling-based system level RBDO method to support broader applications.  For the sampling-based RBDO, the sequential sampling Dynamic Kriging (SS-DKG) method is developed for surrogate modeling.  Under the ARC funding at Iowa, a method for computing stochastic sensitivities of reliability and/or statistical moments with respect to design variables is developed using the score function.  The score function is derived from the copula, which have been used to model the joint distribution of the correlated Gaussian and non-Gaussian input random variables.  The stochastic sensitivity is calculated for the component reliability, system reliability, and/or statistical moments and their sensitivities by applying Monte Carlo simulation (MCS) to the surrogate model that is generated by SS-DKG.  For computational efficiency, a hyper-spherical local window for surrogate model generation, sample reuse, local window enlargement, filtering of constraints, and an adaptive initial point for the pattern search, have been developed.  To further improve computational efficiency of the sampling-based RBDO method for large-scale engineering problems, parallel computing method is developed.  Numerical examples verify that the proposed sampling-based RBDO finds the optimum designs very accurately and efficiently.  These codes (SS-DKG, Copula, and sampling-based RBDO) are integrated to develop the IOWA-RBDO software system that includes a very easy-to-use user interface.  The IOWA-RBDO software system can be used to for both component and system level reliability.  The workshop will provide tutoring of the IOWA-RBDO code for TARDEC engineers with their application specific modeling and simulation codes as black boxes.

Along with the ARO project, we are currently carrying out integration of Iowa developed durability analysis code DRAW with a commercial LS-DYNA FEA code for a large scale parallelization for durability analysis since TARDEC has unlimited license of LS-DYNA on TARDEC High Performance Computing (HPC) for safety analysis.  The DRAW and LS-DYNA will be integrated with the IOWA-RBDO so that, in the near future, TARDEC engineers can perform RBDO for durability and safety multifunctional optimum design ground vehicle structures.