Robust process model identification with model-based control techniques

摘要

A robust method of the generation of process models for use in the production of controllers such as, for example, in the production of mpc - controls the process data noise is added to this, which in the process of the model generation is collected and are used. In particular, a robust method for producing a parametric model first process outputs on the basis of known test input signals or sequences, the collected process data noise is added and then used as a standard - or known art, in order to provide a process model to determine from the collected process data. In contrast to existing techniques for the elimination of noise, which is located on the elimination of random noise in front of the generating a process model, makes possible the addition of random noise with the expected value zero to the process data, in many cases, the generation of an acceptable parametric process model in situations in which in any other way convergence of the process model parameters was obtained. Furthermore, by means of this art process models produced generally greater confidence intervals and thus provide a model, the process operates adequately in many situations, without the necessity of the model to be changed manually or graphically.