Ensuring the best possible solution while accounting for uncertainty in data, constraints and future conditions is of outmost importance in a wide range of industrial applications. In our research, we develop optimization algorithms that enhance the decision support system's (DSS) ability to provide reliable, adaptive, and efficient decision recommendations, ensuring resilience and robustness despite uncertain conditions. To this aim, we focus on efficient and accurate integrations of uncertainty quantification (UQ) methods in optimization under uncertainty, such as in Reliability-based Design Optimization (RBDO) and Robust Optimization (RO).

In optimization under uncertainty, the double loop method consists of two nested loops; an outer loop for the minimization of the objective function and an inner loop for the reliability assessment. It is the most direct and accurate approach, but also numerically expensive. Better efficiency can be achieved using a decoupled loop in which the reliability assessment is separated from the outer optimization loop and updated sequentially.  Decoupled loop methods generally have lower accuracy than double loop methods. The most efficient but generally the least accurate approach is the single loop method, in which the probabilistic constraint is approximated by a deterministic function. In the RMO lab, we are focusing on increasing the accuracy and efficiency of decoupled and single loop methods for optimization under uncertainty.