Design for Additive Manufacturing (DfAM) aims at optimizing the product design based on design rules defined thanks to Additive Manufacturing (AM) constraints. To elaborate a functional design, several parameters like part orientation and layer thickness influence the results of mechanical strength, roughness, and manufacturing time. Surface roughness is an important input of Fused Deposition Modeling which impacts the product functionalities. To participate in product design optimization and DfAM development, this research suggests studying roughness models in order to clarify the best functional design methodology. This analysis permits to propose a meta-model that provides a better estimation for the surface roughness of the FDM products. A new roughness model is developed by a combination of genetic programming and symbolic regression due to the experimental data which helps the manufacturer to predict the surface quality of the products before fabrication. It enables us to estimate the surface roughness of all the AM products fabricated in a different value of layer thickness for all possible orientations in the space regarding part building direction.