Contributions to better control of flotation columns may come from improvements in different areas: measurements, dynamic and functional models, control strategies and control algorithms. Lack of process knowledge, specifically on transient behaviour and interaction among variables, generally leads to low quality conventional distributed control and partially inhibits further improvements. In this work, discrete multivariate dynamic models of operating variables were experimentally obtained in a pilot column and then arranged to build a dynamic simulator prototype, for the air-water system. The main strategy was to use low order dynamic models between every independent and dependent variable, and to combine them linearly to predict the evolution of each variable of interest, by solving a set of difference equations. Given an initial steady state (i.e. feed flowrate, wash water flowrate, openings of tailing and air valves, froth depth, gas holdup and bias), the program running in a PC computer predicts the transient behaviour of the froth depth, the gas holdup and the bias, under any disturbance introduced in the independent variables over time. The program provides a friendly user interface to follow on-line the trends on screen and to recreate the whole experience later on from data previously stored in hard disk during the experiment Comparison of predicted and experimental responses under different disturbances showed the effectiveness of the simulator following the general trends, and where further work to include extreme nonlinear effects has to be done. The integration of dynamic process models and control algorithms in a computer program has proved to be very useful in evaluating control strategies. `