This work presents a novel approach for the modelling of the regression rate in hybrid rocket engines loaded with paraffin-fuel. Two different motor thrust classes have been considered in this work to extend the validity of the numerical model on two motor geometries (200 N and 1 kN). The work consists of two parts. Firstly, a unique and general empirical law of the regression rate was found for both the motor configurations, which are governed by different consumption mechanisms due to the injection system. Ballistic reconstruction techniques have been introduced to reproduce the temporal trend of the regression rate during the burning time and to increase the number of experimental tests used for the development of accurate and reliable regression laws. Experimental observations in the larger motor suggest that the entrainment is dominated by mass diffusive transport phenomena, which is modelled here by a Fick-like relation. This model, coupled with a sophisticated gas-surface interaction treatment, allowed the evaluation of the regression rate in each point of the grain surface by computational fluid dynamic simulations. Then, the overall model has been validated on experimental and reconstructed data. Finally, the effect of the motor configuration on the fuel consumption is shown, thus revealing the crucial influence of the separation zone delivered by the axial injection.