This paper represents the second part of a work devoted to an innovative version of the traditional extended corresponding states technique for the development of thermal conductivity equations on the whole lambda, rho, T surface for individual fluids. In the first part, the theoretical aspects of modeling the thermal conductivity surface of a pure fluid in the new extended corresponding states-neural networks format have been discussed; the fundamental characteristics of the proposed method have been tested using thermal conductivity values generated from dedicated thermal conductivity equations. The promising results formerly obtained suggest to move from generated to experimental data correlation in order to determine dedicated thermal conductivity equations for a number of fluids in real conditions of development. A suitable base of data is required for each fluid of interest. In this second part two fluids, the carbon dioxide and the haloalkane R152a, are studied. The haloalkane R134a is chosen as reference fluid for the model. The absolute average deviations obtained from primary data for the new thermal conductivity models in the extended corresponding states-neural networks format are 1.45% for R152a and 1.13% for carbon dioxide, with an improvement of these figures with respect to the existing equations in the literature. Considering that the expected experimental uncertainty for thermal conductivity data is generally not better than 2%, the obtained results seem to be very promising. (c) 2006 Elsevier B.V. All rights reserved.