The design of tension roofs is always striving towards technological and structural innovation in order to build ever lighter structures to cover equally ever larger spans. Tension structures are in line with the new contemporary architectural concept of space, allowing for the construction of large open spaces where the flexibility and the modifiability of the structures are the most important prerequisites. The geometrical shape that is often used to the design of tension roofs with a cable structure is the hyperbolic paraboloid. In Italy, this constructive typology and this geometrical shape have so far been largely neglected and to a certain extent even ignored by building regulations. In fact no mandatory standards have been set for the design and construction of these structures, except for temporary constructions. Especially, there is a lack of information about aerodynamic coefficients to evaluate the wind action. In fact, while there are precise indications for shells, slopes and domes, there are currently no wind regulations for the hyperbolic paraboloid shape in Italy, but also in Europe (Eurocode) and American (ASCE) regulations. No prescriptions are provided, also, in the new Italian CNR-2008, for the calculation of wind action, either. The purpose of this research is to evaluate aerodynamic loads on hyperbolic paraboloid shaped roofs with a simplify procedure. In order to study the aerodynamic behaviour of this shape are performed Wind tunnel tests and CFD analyses. A database of pressure coefficients is established for different parameters of hyperbolic paraboloid function. In particular, the study followed three consecutive steps: the first was aimed at developing an optimized procedure of preliminary design for the cable structure in order to identify the optimal geometrical configurations; the second focussed on the study of the wind action on these configurations with wind tunnel test and the third phase presents some comparisons between wind tunnel tests and a CFD analysis. Moreover, interesting data are evaluated with the Proper Orthogonal Decomposition method.