ABSTRACT
The engineering community has recognized that structural optimization not considering uncertainty will result in catastrophic failure consequences. Often, such failures cause the loss of many lives. Thus, this issue is attempted by the researchers by exploring various approaches to optimization under uncertainty. Among these, Robust Design Optimization (RDO) is the most popular one, which ensures reliability as well as the least deviation of structural performance under uncertainty. One of the conventional ways of accomplishing RDO is to apply Monte Carlo Simulation (MCS), which is associated with large computational time. Often, surrogate-assisted optimization schemes are adopted to circumvent this computational challenge. But such approaches are hinged on the conventional least squares method, which is often observed to be a source of error in the existing literature. Thus, the issues of either enormous error 304or large computational time prohibit the potential use of RDO in industry. Hence, in the present chapter, a new and prudent method of moving the least squares method (MLSM) is applied in the RDO. The new approach applies a local and moving regression. The proposed method is illustrated by examples that show its accuracy over the conventional approach. At the same time, the results achieved are least sensitive to input uncertainty. The MLSM-based RDO has been observed to be accurate as well as computationally efficient.
