The dynamic behaviour of an industrial five-effect evaporator in sugar manufacturing is determined using the parameter identification approach. The purpose is to obtain an accurate mathematical model that can be used for advanced control design. The evaporation process in sugar industry is highly complex. This process is characterized by strong disturbances with nonlinearities, many constrains and large time delays. Experiments were conducted at sugar factory. Data acquisition involve measurements of syrup brix (sugar concentration) as output, juice flow and steam flow as inputs, syrup flow and vacuum at the fifth effect as measurable disturbances. Different Multi-input single-output ARMAX models were tried using several model orders, combination of inputs and time delays. An extended recursive least squares identification with forgetting factor was used to accurately estimating the model parameters of the evaporation process from experimental data. Reasonable model structure was obtained by an appropriate data filter and an adequate choice of identification design parameters. The obtained model proved to be representative of the considered plant.