A Novel Integrated Approach for a Resource-Efficient Design Validation Co-process

Project Team

Principal Investigator

Zissimos Mourelatos, Oakland University

Government

David Lamb, Amandeep Singh, David Gorsich, U.S. Army GVSC

Faculty

Dorin Drignei (co-PI), Oakland University

Michael Kokkolaras, University of Michigan

Industry

Ramesh Rebba, General Motors

Student

Vijitashwa Pandey, Oakland University

Monica Majcher, Igor baseski, Oakland University

Project Summary

Work began in 2012 and was completed in 2014.

design validation co-process schematic

This project provided a new paradigm in model validation and design optimization by performing both simultaneously using a resource-efficient (reduced number of required tests) co-process. At the same time, it provided a necessary model validation and verification (V&V) process, which was essential for the successful completion of core ARC projects of the topic: Time-Dependent Reliability/Durability Methodologies.

The vehicle systems design process as practiced today uses simulation models that have been validated a-priori (i.e., before the design process) and globally (i.e., in the multi-dimensional space defined by their inputs). However, due to practical reasons and limited resources, this validation is based on limited data obtained at an insufficiently small number of sample points in these very large spaces. In addition, the model validation process often includes a calibration stage to update the simulation model so that its predictions are statistically similar to available tests throughout the input space. This model validation/calibration paradigm can compromise local model accuracy and does not utilize available testing resources effectively and efficiently. In previous investigations, we demonstrated that design optimization using an a-priori, globally validated simulation model can yield a different and worse “optimal” design relative to the one obtained by using a model that is calibrated and validated when necessary as the optimization process progresses.

We developed a novel methodology that integrates design and validation into a co-process. This novel design validation paradigm ensures confidence that the designs generated during the exploration and optimization process behaves as expected when built. The approach i) ensures local model accuracy as the design optimization process progresses since model calibration is conducted whenever necessary, and ii) utilizes available testing resources more effectively by performing calibration-based validation within local domains when necessary.

Publications:

D. Drignei, Z.P. Mourelatos, M. Kokkolaras and V. Pandey, “Reallocation of Testing Resources in Validating Optimal Designs Using Local Domains,” Structural and Multidisciplinary Optimization, 50(5), 825-838, 2014.
Drignei, D., Baseski, I., Mourelatos, Z. P., & Kosova, E. (2016). A random process metamodel approach for time-dependent reliability. Journal of Mechanical Design, 138(1), 011403.
D. Drignei, I. Baseski, Z.P. Mourelatos, E. Kosova, “A Random Process Metamodel Approach for Time-Dependent Reliability,” ASME Journal of Mechanical Design, 138(1), 011403(9 pages), 2015.
Z.P. Mourelatos, M. Majcher, V. Pandey and I. Baseski, “Time-dependent Reliability Analysis Using the Total Probability Theorem,” ASME Journal of Mechanical Design, 137(3), 031405 (8 pages), 2015. Also in Proceedings ASME 2014 Design Engineering Technical Conferences, Paper DETC2014-35078, Buffalo, NY, Aug. 17-20, 2014.
I. Baseski, D. Drignei, Z.P. Mourelatos and M. Majcher, “A New Metamodeling Approach for Time-dependent Reliability of Dynamic Systems with Random Parameters Excited by Input Random Processes,” SAE Int. Journal of Materials and Manufacturing, 7(3):530-536, 2014, doi:10.4271/2014-01-0717. Also in Proceedings of SAE World Congress, Detroit, MI, Paper# 2014-01-0717, 2014.
D. Drignei, Z.P. Mourelatos, V. Pandey, I. Baseski, M. Kokkolaras, A. Singh and D. Lamb, “A Cost-driven Method for Design Optimization using Validated Local Domains,” Proceedings of SAE World Congress, Detroit, MI, Paper# 2013-01-1385, 2013.
D. Drignei, Z.P. Mourelatos, V. Pandey, M. Kokkolaras and D. Gorsich, “Accounting for Test Variability through Sizing Local Domains in Sequential Design Optimization with Concurrent Calibration-based Model Validation,” Proceedings ASME 2013 Design Engineering Technical Conferences, Paper DETC2013-12352, Portland, OR, Aug. 4-7, 2013.
M. Kokkolaras, G. Hulbert, P. Papalambros, Z.P. Mourelatos, R.-J. Yang, M. Brudnak and G. Gorsich, “Towards a Comprehensive Framework for Simulation-based Design Validation of Vehicle Systems,” International Journal of Vehicle Design, 61(1/2/3/4), 233-248, 2013.
J. Li, Z.P. Mourelatos, M. Kokkolaras, P.Y. Papalambros and D.J. Gorsich, “Maximizing Design Confidence in Sequential Simulation-based Optimization,” ASME Journal of Mechanical Design, 135(8), 081004 (8 pages), 2013.
D. Drignei, Z.P. Mourelatos, V. Pandey and M. Kokkolaras, “A Nonparametric Bootstrap Approach to variable-size Local-domain Design Optimization and Computer Model Validation,” Proceedings of SAE World Congress, Detroit, MI, Paper# 2012-01-0226, 2012.
D. Drignei, Z.P. Mourelatos, V. Pandey and M. Kokkolaras, “Concurrent Design Optimization and Calibration-based Validation using Local Domains Sized by Bootstrapping,” ASME Journal of Mechanical Design, 134, 100910 (8 pages), 2012. Also in Proceedings ASME 2012 Design Engineering Technical Conferences, Paper DETC2012-70423, Chicago, IL, Aug. 12-15, 2012.
D. Drignei, Z.P. Mourelatos, M. Kokkolaras and V. Pandey, “A Variable-size Local Domain Approach for Increased Design Confidence in Simulation-based Optimization,” Structural and Multidisciplinary Optimization, 46(1), 83-92, 2012.
D. Drignei, Z.P. Mourelatos, and R. Rebba, “Parameter Screening in Statistical Dynamic Computer Model Calibration Using Global Sensitivities,” ASME Journal of Mechanical Design, 134(8), 081001 (7 pages), 2012.