The main goal of the raw meal blending control in cement industry is to maintain near the reference moduli values and to decrease the variation of the chemical composition rejecting the disturbances. For this purpose, Some commercial control systems are available today. But the control structure in these systems usually consists of a quality control. In this paper, the proposed constrained self-tuning controller deals with not only the quality control but quantity control as well. This is accomplished by inserting a energy term into the quadratic performance index. The self-tuning controller designed according to this performance index bases on a time-series model characterizing relationship between the feedrates and the moduli values. Because of the technological constraints, a optimization problem is formed to compute the control signal. The optimization problem is solved to obtain the least expensive and if possible the least sensitive mix of the raw materials that satisfies the quality requirements of the raw meal under incomplete and stochastic information of the different raw material component concentrations, and further more in real time.