Optimization of an ORC supersonic nozzle under epistemic uncertainties due to turbulence models

International audienceOrganic Rankine Cycle (ORC) turbines usually operate in thermodynamic regions characterized by high-pressure ratios and strong non-ideal gas effects in the flow expansion, complicating their aerodynamic design significantly. This study presents the shape optimization of a typical 2D ORC turbine cascade (Biere), under epistemic uncertainties due to turbulence models (RANS). A design vector of size eleven controls the blade geometry parametrized with B-splines. The EQUiPS module integrated into the SU2 CFD suite, incorporating perturbations to the eigenvalues and the eigenvectors of the modeled Reynolds stress tensor, is used to evaluate the interval estimates on the predictions of integrated Quantity of Interest (QoI), performing only five specific RANS simulations. For a given blade profile, the QoIs total loss pressure and mass flow rate, are assumed to be independent uniform random variables, defined by those estimates. A global surrogate-based method allowing to propose different designs at each optimization step is used to solve the constrained mono-objective optimization problem. To illustrate the suitability of the method, several statistics of the total pressure are considered for the minimization, under the constraint that the mean of the mass flow rate to be within a range