Due to the lack of experimental data regarding two-phase flow through safety relief valves, a number of different calculation methods are presently available in the literature for their sizing. All these models mainly refer to the flow through an ideal nozzle, so that, in order to match the measured values, a discharge coefficient is to be introduced in the calculation, the coefficient thus depending on the model adopted. Furthermore, most of the available data are referred to a few operating conditions. As a result, none of the available models is presently considered sufficiently accurate to be used in a wide range of operating conditions. In the present paper, new data are produced on a steam/water flashing system through a real valve with different values of the main operating parameters (vapour quality, inlet pressure, mass flow rate, and backpressure). The measurements are compared with the predictions of a commonly used homogeneous equilibrium model (HEM), the so-called omega method, and show that the model markedly underestimates the mass flow rate in the whole range of conditions investigated. This unexpectedly implies the introduction of a two-phase discharge coefficient much higher than the vapour coefficient. Finally, a new correlation for the discharge coefficient as a function of the main operating parameters is proposed. (c) 2005 Elsevier Ltd. All rights reserved.