Abstract
In 90° elbows, abrupt turning induces strong secondary flow, separation, and turbulence, increasing pressure loss and degrading velocity uniformity. A hydrogen pipeline elbow is optimized by combining a nature-inspired cross-section with a guide vane, while tuning vane position/angle and geometric radii/offsets using a multi-objective genetic algorithm (MOGA). Three-dimensional CFD is performed for compressible gaseous hydrogen using the Peng–Robinson equation of state and the SST k–ω turbulence model. Design points are generated by Latin hypercube sampling, and response surface models based on non-parametric regression (NPR) and genetic aggregation (GA) guide the search. Relative to the reference elbow, the GA-based optimum improves velocity uniformity by 5.825% and reduces the total pressure-drop coefficient by 0.470%; the NPR-based optimum yields 4.021% and 0.229%, respectively. Flow-field analysis shows reduced separation area, axial vorticity, turbulent kinetic energy, and dissipation, indicating suppressed secondary flow and smoother turning. These gains translate to lower pumping power and enhanced energy efficiency, supporting cost-effective deployment of carbon-neutral hydrogen infrastructure.