Robust design deals with the goal of achieving a targeted high performance when the system is subjected to exterior disturbances or fluctuations. Although several robust design schemes have been proposed by different researchers, they have all omitted the effects of statistically induced uncertainties associated with empirical design parameters (such as material properties and applied loads), on the design's performance. This work aims at developing an adaptive robust design approach for handling evolutionary statistical uncertainties present in the problem formulation via Bayesian statistics. By realizing the statistical nature of design parameters (i.e. the more experimental data collected. The closer the parameter estimates will be to their true values), the proposed robust design scheme allows a designer to adjust and update the empirical parameters continuously during the design progression, provided that new or additional (experimental) data are generated, measured, and collected in a prescribed sequence. A welded beam design problem is considered to illustrate the proposed approach, in which a Monte Carlo technique is employed to simulate the statistical feature of the (empirical) design parameters.
展开▼
