Optimization of a Draft Tube Design Using Surrogate Modelling and Genetic Algorithm

A surrogate-model-based design optimization methodology using Genetic Algorithm to maximize the Static Pressure Rise (SPR) in conical draft tubes is presented. A set of accurate and computationally intensive data obtained from ANSYS-CFD simulations on space filling samples is used for developing a surrogate model. The methodology uses (i) A Latin hypercube experimental design for selecting space filling samples, (ii) Genetic Algorithm for determining parameters of a radial basis function based Kriging model formulated as minimization of a negative log-likelihood function and (iii) length of the straight portion (L), angle of divergence (αd) and inlet swirl angle (αsw) of draft tube as explanatory variables. Flow in the draft tube is characterized by the presence of wall separation, recirculation and axial vortex rope occurring under different inlet swirl and angle of divergence. The study shows that a flow consisting of a low intensity axial vortex rope near the exit of the draft tube is desirable for better distribution of flow in the radial direction for preventing the wall separation and recirculation in high area ratio draft tubes. It is found that the design variable that controls the development and structure of axial vortex rope is the inlet swirl. Verification using CFD analysis showed that the process of optimization has been able to fine-tune the inlet swirl angle that facilitated an optimum sized vortex rope at the center to cause uniform exit axial velocity and improved flow diffusion without wall separation, resulting in significant improvement in Static Pressure Rise. © 2021, The Institution of Engineers (India).