Multi-objective optimization of thermo-hydraulic performance for dimpled–protruded tubes of an intermediate heat exchanger in SCO2 gas fast reactor using machine learning and NSGA-II

The thermo–hydraulic performance of heat exchangers is governed by coupled interactions among geometry, flow conditions, and fluid properties. This study develops a hybrid framework integrating computational fluid dynamics (CFD), machine learning (ML) and multi-objective optimization framework to predict and optimize the thermo-hydraulic and exergy performance of supercritical CO 2 (SCO 2 ) flowing through dimpled-protruded tubes (DPTs) for intermediate heat exchangers in SCO 2 gas-cooled fast reactors. A CFD dataset comprising 504 cases was generated by varying Reynolds number ( Re ), Prandtl number ( Pr ), diameter of dimple-protrusion ( d / D ), and dimple-protrusion orientation parameter ( θ / ω ). Random forest (RF), extreme gradient boosting (XGB), and K-nearest neighbours (KNN) models were developed to predict Nusselt number ( Nu ), Poiseuille number ( fRe ), performance evaluation criterion ( PEC ), and exergy efficiency ( η W - S ). All models demonstrated good agreement with CFD results, with XGB achieving the highest predictive accuracy, followed by RF. At Re = 10 , 000 , d / D = 0.15 , and θ / ω = 0.33 , increasing Pr from 0.84 to 1.30 increased Nu by 56 % and PEC by 10 % . Increasing d / D from 0.10 to 0.25 enhanced Nu by 68 % but reduced PEC by 16 %, demonstrating the trade-off between heat-transfer enhancement and hydraulic performance. The best-performing XGB surrogate was coupled with the non-dominated sorting genetic algorithm II (NSGA-II) to simultaneously maximize Nu and η W - S while minimizing fRe . The resulting Pareto-optimal solutions identify parameter combinations balancing heat-transfer enhancement, exergy performance, and hydraulic losses. The proposed CFD–ML–MOOP framework provides an efficient data-driven approach for the thermo-hydraulic design and optimization of advanced heat exchangers for next-generation nuclear energy systems.