Statistical robust design of a complex system through a sequential approach

Abstract

This paper presents a new robust design approach for a complex system that has multiple candidate scopes of robustness enhancement. This approach captures the robust design project for the system as a sequence of investments under uncertainty, and enhances its economic performance by properly choosing the scopes in which experiment-based robustness enhancement should be performed. It first identifies the set of the scopes in which spending money on experiments can be economically justified. If the set is not empty, it next selects the most appropriate scope from the set or determines not to conduct further experiments according to the economic impacts as well as the associated risks. The approach replicates this process as many times as necessary, and finally establishes an appropriate noise source control policy. An illustrative example confirms that performing robustness enhancement in some scopes can be economically justified, but applying it in every possible scope may not be the most economical option and furthermore can be even counter-economical. Thus, the target scopes of robustness enhancement must be carefully determined. The example also shows that the proposed approach is capable of properly guiding this critical decision-making and obtaining a higher economic performance for the robust design project.

Keywords:

• Complexity management
• Design of experiments
• Robust design
• Six sigma
• Taguchi methods
• Variation reduction