Design and off-design performance improvement of a radial-inflow turbine for ORC applications using metamodels and genetic algorithm optimization

We describe the design and metamodel-based optimization of an ORC radial-inflow turbine for low-grade thermal energy applications using R245fa as a working fluid. Real gas properties were used in 1D and 3D CFD models, which agreed closely in predictions of total-to-total efficiency and net power. However, 3D analysis of the 1D design revealed non-uniform rotor blade loading. The rotor was optimized using Modefrontier (R) to maximize efficiency at the design point. The optimization method integrates CFD simulations, response surfaces constructed from Gaussian radial basis function and the NSGA II genetic algorithm. The wrap angle (theta) and the displacement in the X- and Y-directions of the intermediate control point of the shroud Bezier curve were chosen as the design variables. The performance characteristics and the 3D flow field at design and off-design operating conditions showed that the optimization resulted in better blade loading and higher efficiencies over the entire operating range studied. The new RBF-based metamodel method thus optimizes the performance of the radial turbine while taking into account the detailed 3D flow field at low computational cost.