System Reliability-Based Design Optimization with Associated Confidence Level Under Simulation Model Uncertainties

This project began in 2009 and was completed by 2013.

The traditional sensitivity-based method is not broadly applicable to multi-disciplinary Reliability-Based Design Optimization (RBDO) due to the lack of sensitivity information when dealing with a number of Army ground vehicle designs. For example, survivability under explosion, advanced & hybrid powertrain, robotics, electric power & on-board electrification, vehicle multibody dynamic analysis, durability, and other ground vehicle design concerns. To successfully carry out multi-disciplinary RBDO in these applications, sampling-based RBDO methods that utilize surrogate models need to be used. However, surrogate models could require a very large number of compute-intensive multi-disciplinary simulations (i.e., samples) of large-scale models. In addition, in many engineering applications, only limited data on input variables could be available due to expensive testing/experimental costs to generate the data. The input statistical model estimated from the insufficient data will be inaccurate, which leads to either an unreliable optimum design or non-optimum over design.

To carry out RBDO with associated confidence level for these applications, a very efficient and accurate sampling-based RBDO method is needed. Our team developed a surrogate modeling method that can provide accurate limit states of the system performances based on a limited number of compute-intensive simulations. In addition, to have the RBDO design meet the target confidence level, a conservative surrogate model that does not yield a too-conservative design was developed. For successful application of a simulation-based multi-disciplinary RBDO method in obtaining reliable designs with desirable performances, we were able to predict the confidence level to certify the RBDO design meets the target reliability.

We developed and integrated an accurate and efficient surrogate model method with the RBDO process to obtain a reliable design with the associated target confidence level under the surrogate model and input model uncertainties. The basic research results obtained in this project were successfully integrated into the pre-commercial code I-RBDO that was delivered to the start-up company RAMDO Solutions, LLC in 2013 for commercialization.

Select Publications:

• Hyunkyoo Cho, Sangjune Bae, K. K. Choi, David Lamb, Ren-Jye Yang, “An efficient variable screening method for effective surrogate models for reliability-based design optimization”, Structural and Multidisciplinary Optimization, Volume 50, Issue 5, pp 717-738, November 2014. doi: 10.1007/s00158-014-1096-9
• Ikjin Lee, Jaekwan Shin, K. K. Choi, “Equivalent target probability of failure to convert high-reliability model to low-reliability model for efficiency of sampling-based RBDO”, Structural and Multidisciplinary Optimization, Volume 48, Issue 2, pp 235-248, August 2013. doi: 10.1007/s00158-013-0905-x
• Liang Zhao; K. K. Choi; Ikjin Lee; David Gorsich, “Conservative Surrogate Model Using Weighted Kriging Variance for Sampling-Based RBDO”, J. Mech. Des., 135(9):091003-091003-10, Paper No: MD-12-1356, 2013. doi: 10.1115/1.4024731
• Song, H., Choi, K.K., Lee, I., Zhao, L., and Gorsich, D., “Adaptive Virtual Support Vector Machine for Reliability Analysis of High-Dimensional Problems”, Structural and Multidisciplinary Optimization, Volume 47, Issue 4, pp. 479-491, 2013. doi: 10.1007/s00158-012-0857-6
• Lee, I., Choi, K.K., Noh, Y., Lamb, D., “Comparison Study between Probabilistic and Possibilistic Methods for Problems under a Lack of Correlated Input Statistical Information”, Structural and Multidisciplinary Optimization, Volume 47, Issue 2 , pp 175-189, 2013. doi:10.1007/s00158-012-0833-1