The on-line supervision of the process variables in a steam distillation plant is the main objective of this work. For this end, a nonlinear observer is designed based on the mathematical model proposed by Cerpa et al. (2008). The model has three differential equations, which reproduce the dynamics of the oil mass throughout the three stages of the process: (i) Loss of oil from plant material, (ii) Biphasic (oil-water) layer, (iii) Colected essential oil. The purpose of this work is to estimate the mass flow of oil lost by the plant material and the mass flow in the two-phase layer called the aqueous layer. Both process variables are difficult to measure physically, but it is important to know them because they can have an impact on the performance of the process. However, the mass of oil collected is considered to be an available variable to measure. By taking this consideration into account, a nonlinear Luenberger type observer is designed that takes as available output the mass of oil collected, and from this variable is able to estimate the dynamics of the unavailable variables.