Parametric Shape Optimization of Pin-Fin Arrays Using a Surrogate Model-Based Bayesian Method

Abstract

Pin-fin arrays are generally present in trailing-edge regions of gas-turbine blades for internal cooling purposes. An elevated temperature is important to enhance the thermodynamic efficiency of gas-turbine cycles, and internal blade-cooling mechanisms are employed to mitigate material failure at these high temperatures. It is also important to have high heat transfer at a limited pressure drop to ensure that the overall cycle efficiency is not affected by power loss from driving the flow through internal cooling circuits. A pin-fin shape optimization has been carried out, with the array performance efficiency factor as the objective function, using surrogate modeling-based Bayesian methods. Surrogate modeling/machine learning approaches are instrumental in exploring global solutions to optimization problems using an exploration versus exploitation approach. This ensures prudent use of computational resources for computationally expensive computational fluid dynamics simulations. By varying geometric parameters of the pin fins, a three-dimensional design space has been explored to obtain an optimum pin-fin shape that has the maximum efficiency in the developing region of the flow. An improvement in the efficiency parameter of 1.4 times that of the baseline was obtained in a three-row pin-fin case, and the results were compared with those of genetic-algorithm-based optimization (nondominated sorted genetic algorithm–II). Computation time for optimization was saved by 52%, as an optimum was obtained at lower number of iterations when compared to the genetic algorithm approach, with a 1.2% higher objective function value. This shows an improvement in performance of the current novel approach compared to another stochastic method.