A number of problems in River Engineering are governed by the unsteadiness of the flow hydrograph imposed by the natural hydrologic regime and eventually, anthropic alterations such as dams operation. A main issue in the realm of hydraulic research deals with the similitude of conditions typically achieved at the laboratory scale with those occurring in real rivers. In this context, the generation of hydrographs in a laboratory flume, having real-like shapes is of crucial importance for research and technologies related to the design of waterworks. This paper approaches the problem of generation of hydrographs in a laboratory flume, reproducing and controlling a desired flow and level curve on time. A hydrograph was used as a reference signal for the controllers. A model of the system, a structure of two inputs and two outputs (TITO) was proposed, whose parameters were obtained by means of a PRBS signal designed according to the frequency range of the hydrograph, with which it was possible to replicate the dynamic characteristics of interest at a point of operation. Given the non-linearity of the system, a robust decentralized predictive proportional integral control (PPI) strategy is proposed, using two SISO controllers. The selection of input pairs was made by RGA, a decoupler stage with a delta adjustment parameter was implemented, PPI decentralized controllers were designed and implemented by means of lambda tunning and in a similar way PI controllers as a benchmark. Finally, the results of using PI or PPI controllers, with or without uncoupler, are presented. Which were able to reproduce the hydrographs, however it is determined that the PPI controller with decoupler presents a greater decrease of the ERMS and a lower control effort for the tracking of variable references in a non-linear plant, i.e.: replicating a real-like shape of an hydrograph in the laboratory flume similar to that occuring in a river.Regular 2015FONDECYTFONDECYT