Reliability of Deterministic Optimization and Limits of RBDO in Application to a Practical Design Problem

Abstract

A practical engineering design problem is used to examine the over-conservativeness of designs obtained using deterministic optimization with worst-case parameter assumptions and a safety factor. Additionally, an attempted application of reliability-based design optimization (RBDO) demonstrates the limits of RBDO for practical problems. The design problem considered here is TESCO's Internal Casing Drive System (ICDS), which is used in feeding pipeline, or casing, into predrilled holes. After developing a finite element model of the ICDS, experimental data is used to successfully validate modeling methods and assumptions. The validated model is then subjected to multiple analyses to determine an appropriate design configuration to be used as the starting point for optimization. Worst-case, safety factor-based design optimization (SFBDO) is then applied considering two and three design variables, and is successful in increasing the critical load of the ICDS, Pcrit, by 35% and 45%, respectively. An efficient and recognized RBDO method, Sequential Optimization and Reliability Assessment, is selected for application to the design problem to determine an optimum design based on reliability. Due to the optimization formulation, however, SORA cannot be applied. The ICDS design problem represents a practical example that demonstrates the difficulties and limits in applying RBDO to practical engineering design problems. To evaluate the over-conservativeness of worst-case SFBDO, structural reliability analysis is performed on the deterministic optimum designs. It is found that the value of Pcrit for both the two and three variable optimum designs can be increased by 53% while maintaining acceptable probability of failure, demonstrating the over-conservativeness of the worst-case SFBDO.