A correct modelling of rotor-bearing-foundation-system (RBFS) can provide effective rotor balancing and fault diagnostics, therefore, optimise the operating efficiency of the rotating machinery. One major modelling difficulty is to identify the equivalent foundation, which reproduces the same vibration responses over the operating speed range, when substituted for the actual foundation in the RBFS. This paper introduced an identification technique, which utilised the foundation vibration data excited by the existing rotor unbalance, to identify the equivalent foundation using modal parameters. The technique identified the modal parameters of each vibration modes independently by decoupling the modal analysis equations, thereby simplifying the identification procedure. In the illustrative example, the equivalent foundation of a numerical RBFS was identified. When substituted for the actual foundation, the equivalent foundation satisfactorily predicted the system unbalance response and the response prediction was still robust when using the noise corrupted data for the identification. Hence it is concluded that the proposed foundation identification technique is likely to be applicable in practice.