TX 75083-3836, U.S.A., fax 01-972-952-9435. AbstractThe accurate prediction of the downhole pressures and the returning flow rates in low-head drilling (LHD) and underbalanced drilling operations (UBD) is a major concern in the oil industry. The present work shows an original formulation of a dynamic two-phase flow model based on the classic driftflux set of conservation equations. However, differently from the traditional approach, the closure of the system is obtained by using measured data acquired during the execution of the operation. The innovative concept consists of utilizing mechanistic models that are dependent on unknown parameters. These parameters and the model state are calculated and updated on a regular basis to minimize the differences between model predictions and measured data.The potential of this new approach is discussed by presenting some transient examples of application. The current study represents a continuation of previous work on real time data interpretation 1 , and a step in the development of a methodology for the introduction of learning while drilling process in the hydraulic design and follow-up of LHD and UBD operations.