Abstract
In this work, performance evaluation of surrogate models in a wire-wrapped fuel assembly shape optimization is carried out. In addition to three basic models, i.e., Response Surface Approximation (RSA), Kriging (KRG) and Radial Basis Neural Network (RBNN), the multiple surrogate model Press Based Averaging (PBA) is also tested. Two design variables are selected to enhance the performance of wire wrapped fuel assembly and design points are selected using Latin Hypercube Sampling (LHS). Optimization problem has been stated as maximization of the objective function, defined as a linear combination of heat transfer and friction loss related terms with a weighing factor. Among the three basic models Kriging performs better while among the all models multiple surrogate model, PBA performs the best. Use of multiple surrogate PBA gives more robust approximation than individual surrogate
