Reliability-Based Design Optimization of Systems with Deterministic Degradation Using a Meta-Model of the System Reliability  
Author Gordon J. Savage


Co-Author(s) Young Kap Son


Abstract A major hurdle in the application of reliability-based design optimization (RBDO) to degrading systems is the need to ensure reliability policies are met while moving the design point to optimize some objective function. Herein, the efficiency of RBDO of degrading systems is greatly improved by essentially uncoupling the time-variant reliability problem from the optimization problem. That is, time-variant reliability is related to the design space through a simple meta-model. A design of experiments paradigm helps select a few training sets of the design parameters that cover the design space. Monte Carlo Simulation serves as the most general method to find the corresponding system reliability accurately. Finally, the meta-model is used to quickly evaluate any functions of reliability in the optimization process. The novel meta-model method is applied to the seminal roller clutch with wear-related degradation. The case-study shows means and tolerances are found that correctly minimize a monetary cost objective and yet ensure reliability policies are met. The meta-model approach is simple, accurate and very fast, suggesting an attractive means for RBDO of time-variant systems in general.


Keywords Optimization, design, Path degradation, System reliability, Monetary costs, Meta-models
    Article #:  RQD25-78
Proceedings of 25th ISSAT International Conference on Reliability & Quality in Design
August 1-3, 2019 - Las Vegas, NV, U.S.A.