The range hood is crucial in kitchens during cooking activities. Inside the residential houses, cooking activities are one of the main sources of particle emission, which decreases the air quality level. Furthermore, multiple studies found a strong correlation between the particles emitted from cooking activity and chronic obstructive pulmonary disease (COPD), lung cancer and diabetes. The use of an efficient range hood is essential to maintain a healthy air quality level inside the house. The fan is the main component inside a range hood. Most of the range hoods are equipped with an axial fan or a one rotor squirrel cage fan. In the present study, a powerful double-intake and rotor squirrel cage fan is designed and optimized by using a developed optimization process loop based only on open source libraries. Dakota is used to achieve the sampling and build the surrogate surfaces, Salome to generate the geometry and the mesh grid and OpenFoam for the calculations. More than eleven design parameters are selected in the impellers, blades and volute regions. The two objective functions: total efficiency and the generated noise are improved by maximizing and minimizing their values, respectively. The Latin Hypercube Sampling (LHS) method is selected to achieve sampling over more than 363 design parameters set. In order to model the turbulent flow, a 3D incompressible simpleFoam solver is used and coupled to the Multiple Reference Frame (MRF) approach. The Kriging and the quadratic polynomial response surface are used to expand the design space and improve the objective functions. The total efficiency is improved by 12 % and the noise is reduced by 2 sones compared to the initial design. The Kriging Metamodel predicts with less than 2 % the total efficiency and 1% the generated noise compared to the OpenFoam calculation. A large 3D coherent structure is observed in the volute region with a scattered turbulent region near the outlet. The optimal design is validated at the design point against the produced prototype, with an error of 2.8 % and 1.3 % on the total efficiency and generated noise, respectively.