Diagonal flow fans offer substantial energy-saving potential and find broad application across various sectors. Their performance relies heavily on factors like outlet guide vanes and spacing relative to moving blades. However, research into enhancing fan performance through optimized guide vanes and spacing remains limited. In this study, we focus on improving the accuracy of predicting the internal flow field of diagonal flow fans. This paper incorporate rotation and curvature effects using the Large Eddy Simulation (LES) model and introduce stress terms with helicity constraints to create a non-linear subgrid-scale model. This refined model enables more precise numerical simulations. By employing accurate simulations, we optimize the outlet guide vane configuration and conduct sensitivity analysis. We utilize a Radial Basis Function (RBF) model coupled with the Sobol method for this purpose. The optimized guide vane design exhibits enhanced resistance to airflow separation compared to the original, resulting in notable reductions in flow losses within the grille channel. Experimental tests are performed on the diagonal flow fan both before and after optimization. At the specified operating point, the second guide vane optimization leads to a 1.28 m3/min increase in fan flow, a 4.33% rise in total pressure efficiency, and a 2.2 dB noise reduction. These findings underscore the accuracy of the helicity correction model in predicting diagonal flow fan behavior. The multi-objective optimization approach, combining the RBF proxy model with the Sobol method, proves highly reliable. It offers valuable design insights for similar fans and establishes a credible design methodology.