Abstract
To control countinuous process properly, it is very important to grasp process dynamics accurately. But that is not done satisfactorily. As a result, process control has not been executed properly. Experimental design that enables accurate grasping of process dynamics is proposed. Gaps between system theory and practice in system identification are invested. As a result, we have found that there are needs for experiment that is capable of producing high quality data for system identification. Experiment patterns under practical restricts of manipulated variable are proposed, and the condition that estimates parameters in the process model most accurately is derived. The best condition is then compared with what Box and Jenkins had given, and it is shown that the former can estimate as accurately as the latter with smaller range of manipulated variable.
