The present study aims to minimize velocity disruption using rotational guide vane (RGV) before the 90° pipe bend where the curvature ratio is r/D = 1.0 at Re = 3 × 104. The combination of computational fluid dynamics and genetic algorithm is used as a tool for the geometric optimization of RGV. The SST k–ω turbulence model was preferred in flow analysis. The number of blade, the blade angle, and the location of RGV are the geometric parameters to be optimized. A new evaluation method of the velocity distribution after the bend is proposed. The objective function is defined as velocity distribution mean squared error that is based on the comparison of the velocity distribution on the cross section after the bend with the fully developed flow in a straight pipe under the same conditions. As a result of a series of optimization processes, velocity distribution at the 90° bend exit is compared between optimized RGV and without guide vane. The results show that the optimized RGV improves velocity distribution at the bend exit. Compared to the without guide vane case, the optimized RGV has delayed flow separation from α = 37° to 56°, and the flow reattachment point moves from x/D = 0.5 to 0 at the bend exit. Thanks to optimized RGV, the counter-rotating Dean vortices merged to form a single vortex at the center of the pipe. In addition, significant reduction in turbulent kinetic energy was observed, approximately 50% when using the optimized RGV compared to the without guide vane case.