Abstract
<title>Abstract</title> <p>The convergence of artificial intelligence (AI) and digital twin (DT) technology has opened new possibilities for the real-time thermal management of high-performance mechanical systems. This paper proposes an AI-driven digital twin framework that integrates high-fidelity computational fluid dynamics (CFD) with physics-informed surrogate models to optimize the design and operation of internal cooling channels in gas turbine blades. A novel hybrid bifurcated-serpentine cooling channel geometry with staggered micro-ribs is introduced, combining the flow-reattachment benefits of serpentine passages with the localized heat-transfer augmentation of rib turbulators and the flow-distribution advantages of bifurcated branch paths. Four coolant flow arrangements straight-through, serpentine, counter-flow, and the proposed hybrid bifurcated configuration are systematically compared under identical boundary conditions. A multi-objective optimization using NSGA-II coupled with an artificial neural network (ANN) surrogate is performed to balance cooling effectiveness against pressure penalty, yielding a Pareto-optimal design frontier. The CFD baseline is validated against published experimental Nusselt number and friction factor data for rib-roughened channels, with deviations below 8%. A sensitivity analysis examines the influence of Reynolds number, coolant inlet temperature, rib pitch-to-height ratio, and heat flux on overall thermal-hydraulic performance. The proposed hybrid channel is expected to yield approximately 23% improvement in overall thermal performance factor compared with a conventional ribbed serpentine channel at comparable pressure drop, based on literature-consistent estimates. The AI surrogate is projected to achieve prediction speeds several orders of magnitude faster than full CFD, as demonstrated by comparable surrogate models in published literature. This work establishes a methodological pathway toward adaptive, self-optimizing cooling systems that respond to changing operating conditions in real time.</p>