Peneloux's volume translation is found of not equal benefit depending on pressure and temperature ranges for which volumetric properties are represented (some ranges being unfortunately worse described after correction than before). Other expressions for volume translations have been proposed up to now. They are slightly better than original Peneloux's one. However, the main default of most of these published modifications contain translation parameters that are not pressure dependent or, if they are, lead to time consuming non-cubic equations of state. It is clearly demonstrated herein that parameters must be pressure dependent. It is also clearly demonstrated that pressure dependence can be conveniently, easily and accurately represented using neural networks. Through volume translations using temperature and pressure dependent parameters, we show it is possible to represent simultaneously VLE and volumetric properties with accuracy degree similar to that obtained using Lee-Kessler-Plocker equation. The new model (cubic equation of state + neural network based volume translation) is very competitive in terms of accuracy with respect to more complex equations of state based on virial developments for VLE and volumetric properties. It is easier to use with mixtures, as mixing rules are quite straightforward. Simplicity use of cubic equations of state is strictly maintained using neural networks. Furthermore, we show that weights for mixtures comes from weights adjusted on pure compound PVT data, no new adjustment is required to treat mixtures provided pure component weights are available. Pure component weights related to neural networks are conveniently adjusted on PVT data obtained through the vibrating tube experimental method. (c) 2007 Elsevier B.V. All rights reserved.