In Hall-Heroult process for aluminium production, estimating the alumina concentration and the anode-cathode distance (ACD) remains a challenge. One of the difficulties arises from the fact that it is not possible to measure those quantities continuously during the pot operation. This article presents a novel approach for on line estimating alumina concentration and ACD in a regular aluminum-reduction pot cell using a Linear Kalman Filter. This is done by using an appropriate dynamical model for the pot, which is obtained by combining the first principle modeling and experimental identification of alumina concentration behavior from irregularly sampled data. Moreover, a dynamical model for the pot resistance is identified as a function of the alumina concentration and ACD data. The proposed approach is validated on an industrial platform.