Data processing: artificial intelligence – Neural network – Structure
Patent
1997-11-04
2000-08-01
Stamber, Eric W.
Data processing: artificial intelligence
Neural network
Structure
706 23, 706 25, 706 21, G06E 100
Patent
active
060980605
DESCRIPTION:
BRIEF SUMMARY
FIELD OF THE INVENTION
The present invention relates to a method for controlling a process, wherein manipulated variables for a plurality of actuators that influence the process are calculated in a control device from measured output quantities of the process. The present invention also concerns a corresponding device.
BACKGROUND INFORMATION
In controlling processes, a plurality of actuators are often available to influence certain output quantities of the process. The manipulated variables for the individual actuators are calculated in a control device as a function of guide variables whose values should assume the output quantities of the process. An example of such a process control is the regulation of the flatness of the strip in a rolling system, wherein the roll gap and thus the flatness of the strip are influenced by the sweep, bending, axial displacement and/or zone cooling of the rolls as a function of measured values for the flatness of the strip distributed over the width of the rolled strip.
For good regulation of the output quantities of the process, it is extremely important for the controlling device to be supplied with information regarding the efficiencies of the actuators, i.e., information on how the individual actuators act on the output quantities. The problem here is that little or no information about the efficiencies of actuators is usually available in industrial systems.
It is generally known that in rolling metal strips, the flatness of the strip can be improved by simultaneous operation of a plurality of actuators that influence the roll gap.
It is also known from Hitachi Review, vol. 41, no. 1 (1992), pages 31-38, that for regulating the flatness of the strip in a roll stand, each actuator of the roll stand can be assigned a typical sample profile of strip flatness of which it is known on the basis of empirical information about the rolling process that it can be improved by actuation of the respective actuator. Downstream from the roll stand, the actual profile of strip flatness is measured, and the measured values thus obtained are sent to a neural network that indicates as a network response the proportion of the individual sample profiles in the measured profile of strip flatness is composed of the individual sample profiles. The resulting proportions are linked in fuzzy rules to form actuating signals for the actuators. The known process is based to a very great extent on the empirical knowledge of the roll producer regarding the effect of the actuators, but that is comparatively general and inaccurate for optimum process control. Furthermore, the empirical knowledge introduced is usually limited to the respective system and cannot be readily transferred to other systems.
SUMMARY OF THE INVENTION
An object of the present invention is to provide a process and a device for controlling a process that will permit optimum process control, regardless of the available empirical knowledge about the process.
The present invention is directed to a method for controlling a process, wherein manipulated variables for a plurality of actuators that act on the process are calculated in a control device from measured values of output quantities of the process. The actuator efficiencies that describe the dependence of the changes in output quantities on the changes in the manipulated variables for each actuator are learned in a neural network. The learned actuator efficiencies are supplied to the control device to improve the calculation of the manipulated variables.
With the process according to the present invention and the corresponding device, the efficiencies of the actuators are learned automatically for the individual actuators on the basis of changes in the manipulated variables measured during the process and the output quantities of the process with normal process management and are sent to the control device to improve the calculation of the manipulated variables. The efficiencies of precisely those actuators being controlled by the control device are always learned. So i
REFERENCES:
patent: 5249954 (1993-10-01), Allen et al.
patent: 5282131 (1994-01-01), Rudd et al.
Berghs Andre
Yuan Hao
Brown Edward G.
Siemens Aktiengesellschaft
Stamber Eric W.
LandOfFree
Process controlling method and device does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Process controlling method and device, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Process controlling method and device will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-673452