This paper is devoted to the problem of digital control design to keep the output variables of the controlled process in a given range. Such a problem is of particular importance in control practice if it is only necessary to maintain some variables of a dynamic process in a stated range, but not to track the reference signal. The control design approach is developed for a nonlinear digital model. This approach is based on the predictive model and a special introduced cost functional that is minimized over the prediction horizon. This functional include two terms: the first term represent intensity of the control actions, the second term is a penalty for violating the specified range. It is shown that the optimal control design at each instant of discrete time is reduced to nonlinear programming problem. This problem always has a solution due to the addmissible set, which includes additional variables that guarantee the existence of the solution even in the case of violating the given range. The application of the proposed approach for oil refining in the distillation column is given. The examples of process simulation are presented and discussed.