As in nature as in diverse anthropogenic processes, the multiple substances present in a specific system have a pattern of distribution among the diverse phases, regions with the same composition, delimited by interfaces. This pattern of distribution is ruled by conditions of pressure P, temperature T, global composition and characteristics of the intervening substances. In this project is proposed a method, based on the concept of artificial intelligence, as is the case of neural networks, to establish a relationship between the patterns of distribution and conditions/characteristics ruling the system. The neural network learns with examples (objects), systematically submitted to the network. Each object is constituted by two different parts, the input and the output, represented by conditions/characteristics ruling the system and the pattern of distribution, respectively. Previously to the learning process (train), the input is coded in a fixed length vector, in order to have generalizable models, independently of the number of components in the system. After training the network, the predictive ability is evaluated, comparing the experimental and predicted values of distribution pattern, for a set of objects that was not used during the training period (test set). In the next step, new systems/conditions are designed and submitted to the trained models to obtain predictions. The systems/conditions with higher potential of applicability in real life are tested experimentally in order to validate the trained networks. The models built include Ionic Liquids (ILs), and other compounds. The structuration of the project is modular, with progressively more generalizable models built, based on objects previously used in the construction of more specific trained networks. The construction of a database of objects will be initiated at the beginning of the project, being incorporated more data along the time. To construct this database will be used data from specialized literature on phase behaviour, where, usually is defined the composition of the phases in equilibrium at certain conditions of P and T. However the global composition usually is not explicit. This can be accessed by the definition of tie-lines and polygons, zones of equivalence where the global compositions will correspond to a constant pattern of distribution of the different components among the different phases in equilibrium. The characterization of the compounds of the mixture will be carried out using Chemoinformatic and Computational Chemistry methodology. The concretization of this generalizable methodology permits the interpretation of the results of phase behaviour and constitutes a novel tool with applicability on the screening of innovative systems/conditions in the field of unit operations at laboratorial, pilot and industrial scales.