The transport properties of a fluid show a complex dependence with temperature and pressure due to the combination of different phenomena occurring at the microscopic scale. The entropy scaling concept aims at describing this complex behavior by expressing reduced transport properties as one-variable functions of the Tv-residual entropy, a thermodynamic quantity that can be straightforwardly estimated with an equation of state (EoS). In this work, a reformulated version of Rosenfeld's original entropy scaling approach is proposed in order to calculate the thermal conductivities of pure fluids. A specifically developed reduced thermal conductivity expression was correlated to a function of the Tv-residual entropy that was recently proposed by our group to correlate viscosities and self-diffusion coefficients. The thermodynamic properties involved in the definition of the entropy-scaling variables (that are the residual entropies, densities, heat capacities) were estimated with either the I-PC-SAFT or the tc-PR equations of state (EoSs) thus leading to the definition of two different models. Each of them was validated against a large database of around 90,0 0 0 experimental thermal conductivities encompassing liquid, gas and supercritical states for 119 chemical species belonging to 11 chemical families such as n-alkanes, alkenes, alcohols, HFC-CFC etc. For each model, component-specific, chemical-family specific and universal parameters were proposed. Working with the I-PC-SAFT and tc-PR EoSs, the obtained MAPEs (Mean Absolute Percent Errors) are respectively 3.3% and 3.4% when the model parameters are considered as componentspecific, 9.7% and 5.6% when they are selected as chemical-family specific meanwhile they are 11.2% and 9.2% when they are assumed to be universal.