Qualifying and Segmentation of Historical Process Data Using Optimal Experiment Design Techniques for Supporting Parameter Estimation

With the wide-spread application of process models and simulators, estimation of model parameters becomes a crucial project. In chemical industry the processes are mostly highly non-linear which makes the identification of model parameters difficult. In the practice the process simulators are not just for design but optimization of operating plants in numerous cases various sets of process data are available to determine the necessary model parameters. With further examination of the historical process data, a new possibility becomes applicable: some time-series segments can provide more information about the estimated model parameters than other parts of the recorded time-series. Since the tools of Optimal Experiment Design (OED) are for maximizing the information content of the experiments regarding to the unknown model parameters, the applicability of these tools in qualifying the recorded process data is obvious. In this paper the connection of classical time-series segmentation and OED tools will be examined throughout a simple polymerization example to prove the efficiency of integration of these tools to support the parameter identification of process models.