摘要 |
In the case of known processes for the control of a process in a control system, the system is preset at the beginning of each process sequence as a function of a predicted process parameter which has a dependency, caused by the system, on faulty input variables. In this arrangement, the description of the dependency is carried out with the aid of a model of the process which is adapted in the course of the process. In order not to be dependent on the construction of models which are as a rule inexact, before the start of the process the input variables are fed to a neural network (9) having variable network parameters for the prediction of the process parameter (Fn); during the course of the process, on the basis of measurements of the input variables (xn sim) and of the process parameter (Fn), these variables are simulated and used for the adaptation of the network parameters. <IMAGE>
|
申请人 |
SIEMENS AG, 80333 MUENCHEN, DE |
发明人 |
GRAMCKOW, OTTO, DR., 91052 ERLANGEN, DE;MARTINETZ, THOMAS, DR., 80686 MUENCHEN, DE;POPPE, THOMAS, DIPL.-ING., 80689 MUENCHEN, DE |