MEAN-LINE DESIGN AND OPTIMIZATION OF AXIAL-FLOW TURBINES BASED ON MIXED INTEGER NONLINEAR PROGRAMMING

This work proposes a novel equation-oriented approach for the mean-line design optimization of axial-flow turbines, in which turbine constitutive equations and design specifications are included as constraints of the optimization problem. The problem is formulated as a MINLP (Mixed Integer Nonlinear Programming), including either Soderberg or Traupel loss models. This model formulation allows the use of the gradient/Hessian information of both constraints and objective function, speeding-up the optimization process. Moreover, it allows using "Spatial Branch-and-Bound" optimization algorithms (e.g., BARON) which have guaranteed global optimality. After an exhaustive validation, the model is applied to the optimization of two case studies (air and sCO2 as working fluids), resulting in certified optimal solutions and computational time lower than the black-box approach.