A feedforward, multilayer neural network was used for simulation of complex liquid-vapour equilibrium in an industrial process of urea synthesis from ammonia and carbon dioxide. It was found that the properly selected and trained neural network rendered precisely, qualitatively and quantitatively, the thermodynamic character of the dependence of total pressure, partial pressures and mole fractions of gaseous reactants on the urea system state parameters. Artificial neural networks are an effective tool for modelling complex phase equilibria in chemical technology.