Data Based Causal Modelling of Manufacturing Plants using Transfer Entropy

In complex manufacturing plants it is generally difficult to define first principle models to describe the system behaviour mathematically. In this paper we show how to generate data based models, whose input structure allows a physical interpretation. Therefore we applied the transfer entropy method in order to find cause-effect relationships in a given or measured data set usually from industrial plants and multivariate correlation analysis to decrease the dimension of the basic problem of data based identification. One main problem of such data is the poor information which is sometimes contained in the data. We present results of the proposed method dealing with this problems by using data of real applications.