Abstract: Robust design is aimed at determining the optimal design factors settings to make products less sensitive to noise factors and reduce quality variation effectively. Design of experiments for noise factors becomes important, which is also called outer array design. Noise factors cannot be controlled, but sometimes they can be observed and the distribution can be known. The prior information can be effectively used to optimize outer array design. To get representative samples of noise factors, an outer array design approach based on quantile of noise distribution and uniform design is proposed. Appropriate quantiles of noise distribution are chosen as non-random samples of noise factors. The quantiles are regarded as levels of noise factors, which are used for outer array design combined with uniform design tables. Illustrative example shows the approach can arrange experiments reasonably, simulate the noise factors and quality variation effectively. The proposed approach can also reduce experiment times, improve estimation precision and find robust solution with low cost and high efficiency.

Key words: Noise distribution, Outer array design, Quantile, Robust design, Uniform design, Genetic algorithm, Quality of workspace, Workspace, Cable-Spring Mechanism