Development of Simulation Model Validation Framework for RBDO

Principal Investigator: K. K. Choi, University of Iowa, kyung-choi@uiowa.edu
Student: Min-yeong Moon, University of Iowa
Government: David Lamb, David Gorsich, James Sheng, Bob Garcia, U.S. Army TARDEC
Industry: Ren-Jye Yang, Ford Motor Company
Dmitry Tanenko, General Dynamics Land Systems

Uncertain predicted output PDFs and prob. of failuresThe central goal of design is to produce an economical product that satisfies all performance specifications and requirements set by design engineers. If they are not satisfied, the product can fail, potentially causing serious damage. Thus, reliability of the product is critical. In this context, reliability-based design optimization (RBDO) methods have been developed previously in ARC projects (most recently 3.1 & 3.2) to obtain optimal designs that satisfy the target reliability.

Existing RBDO methods require response evaluations, provided by computer simulations, at different design points during the design optimization process. More and more, design engineers seek to replace product testing with computer simulations because of the significant cost and time that the testing requires. Thus, reliable product designs have been obtained by using computer simulation methods instead of physical testing. However, it is well known that even physics-based simulation models are built on the basis of various assumptions and simplifications. Since the simulation-based RBDO method uses computer simulation models, the optimum design obtained from it may not actually satisfy the target reliability unless the simulation could correctly predict the test results statistically. In other words, the RBDO design without model validation may not be the true optimum.

The first objective of this project is to obtain a more accurate simulation model by developing a model validation framework. For this, we propose to carry out a two-step process instead of taking care of the bias characterization and model calibration simultaneously. To do that, it is necessary to figure out which uncertainty factor, model bias, or calibration parameter is dominant and to enhance the existing model improvement method. Once a model validation framework is developed, a more reliable optimum design can be obtained by integrating the design optimization process with the model validation framework. Therefore, the second objective of this project is to integrate model validation into the RBDO process. However, it is too expensive to validate the simulation model over the entire design domain. Hence, the third objective is to develop a local window concept that works for RBDO with model validation to minimize the test cost for model validation during the RBDO process.

Publications:

  • Cho, H., Bae, S., Choi, K.K., Lamb, D., and Yang, R-J., "An Efficient Variable Screening Method for Effective Surrogate Models for Reliability-Based Design Optimization," Structural and Multidisciplinary Optimization, DOI 10.1007/s00158-014-1096-9, 2014.
  • Choi, K.K., Gaul, N.J., Song, H., Cho, H., Lamb, D. and Gorsich, D., "Iowa Developed Reliability-Based Design Optimization (I-RBDO) - Technology Transfer," Modeling & Simulation, Testing and Validation (MSTV) Mini-Symposium, 2014 NDIA Ground Vehicle Systems Engineering and Technology Symposium, August 12-14, 2014, Novi, MI.
  • Cho, H., Choi, K.K. and Lamb, D., "Confidence-Based Method for Reliability-Based Design Optimization," 40th ASME Design Automation Conference, August 17-20, 2014, Buffalo, New York.
  • Moon, M., Choi, K.K., Cho, H., Gaul, N., Lamb, D., Gorsich, D., "Development of a Conservative Model Validation Approach for Reliable Analysis," 41th ASME Design Automation Conference, August 3-5, 2015, Boston, MA.
  • Cho, H., Choi, K.K., Lee, I., Lamb, D., "Design Sensitivity Method for Sampling-Based RBDO with Fixed COV," 41th ASME Design Automation Conference, August 3-5, 2015, Boston, MA.
  • Gaul, N., Cowles, M., Cho, H., Choi. K.K., Lamb, D., "Modified Bayesian Kriging for Noisy Response Problems for Reliability Analysis," 41th ASME Design Automation Conference, August 3-5, 2015, Boston, MA.