Neural-network (NN) modeling is most suitable for simulations of correlations that are hard to describe or cannot be accurately predicted by physical models. This article describes the principles and procedures of NN modeling. It discusses the use of NN modeling in general organization of software and graphical user interfaces. The article also provides information on the ways to improve and upgrade the NN models.