This paper deals with the identification and advanced control of the raw material blending process in cement industry. The process is multivariable and coupled one, because the feeder tanks do not contain homogeneous raw materials chemically. The time delays in the system are also considerable. The disturbances coming from the variations in the chemical compositions of the raw materials from long-term average compositions cause the changes of the system parameters. Therefore, for providing the target values of the oxide compositions of the raw meal determining the high quality of cement, the stochastic multivariable dynamic models are developed and model predictive controllers are designed to calculate the optimal feed ratios of the raw materials despite disturbances. This study consists of two parts; in the identification part, three different linear multivariable stochastic ARX models are proposed. The identification results show that these MISO and MIMO models are good models. In the control part, model predictive control strategy is applied. At the end of the simulation study, the output values reach the specified set points quickly. Also the significant decrease in the variance of controlled outputs is obtained. Copyright (C) 2004 John Wiley Sons, Ltd.