Comparative GAN-based multi-objective optimization of aerothermal performance for an asymmetric cavity bottom squealer tip with two strategies: fluid-only versus conjugate heat transfer models

To investigate the impact of conjugate heat transfer (CHT) on the optimization design of cooled turbine squealer tips, this study employed two distinct optimization strategies: one based on a fluid-only domain model, and another incorporating a solid-fluid domain CHT model. A multi-objective optimization aimed at improving the aerothermal performance of an asymmetric cavity bottom blade squealer tip (ACBT) was carried out. The optimization framework utilized a Generative Adversarial Network (GAN) as the surrogate model. The training and test datasets were constructed through numerical simulations, whose methods were validated against measurements from a linear cascade. The Pareto fronts from the two optimization strategies reveal that achieving better cooling performance or lower heat transfer intensity inevitably leads to a deterioration in aerodynamic performance. Compared to the Baseline, nearly all optimized boundary samples from the two optimization strategies exhibit significant improvements in the film cooling effectiveness and overall cooling effectiveness, while achieving a reduction in the heat transfer coefficient and total pressure loss coefficient as well. The optimized asymmetric cavity bottom geometries obtained with and without CHT exhibit dramatically different trends. Under actual gas turbine operating conditions, computational verification shows that compared to the Baseline, the optimized ACBT obtained without CHT demonstrates better performance when assessed based on tip temperature. When evaluating performance based on tip heat load, the optimized ACBT obtained with CHT yields superior performance.