In the production process of making material pulp through the ball grinder, water-content is a very important parameter. In order to make the water-content in material pulp reaches the target value, we adopt the Theory on Self-tuning Predictive Compensating Control to predict and controll so that we can find out the best value promptly, and realize the best controlling on-line. To do so, the average value of the send outs can be equal to the set value. In the thesis, the author gives examples to make his analysis and does computational numerical simulation. As a result, it turns out that we can do predicting to the send out to time ‘k+d’ when the time is ‘k’, and thus we can control water-content effectively. In this way, the ultimate water-content in the kiln basin reaches the target value, and we can achieve our goal of automatically measuring and controlling water-content.