Data processing: generic control systems or specific application – Specific application – apparatus or process – Product assembly or manufacturing
Reexamination Certificate
1997-11-07
2001-02-06
De Cady, Albert (Department: 2784)
Data processing: generic control systems or specific application
Specific application, apparatus or process
Product assembly or manufacturing
C700S049000, C706S023000
Reexamination Certificate
active
06185470
ABSTRACT:
TECHNICAL FIELD
The present invention relates generally to systems and methods for providing adaptive control of a dynamic system or structure and, more particularly, to a neural network based predictive controller.
BACKGROUND ART
Many dynamic, nonlinear systems exist which need adaptive forms of control. As an example, vibration and undesirable aeroelastic responses adversely affect various flexible structures (e.g., an aircraft wing). In turn, these adverse effects shorten the lifespans and increase the acquisition and maintenance costs of such structures. Thus, an active control system is desired for reducing vibration, alleviating buffet load and suppressing flutter of aircraft structures, providing adaptive hydraulic load control, reducing limit cycle oscillations of an aircraft store and the like.
U.S. Pat. Nos. 3,794,817, 4,358,822, 5,197,114 and 5,311,421, the entire disclosures of which are incorporated herein by reference, describe conventional controllers. In general, conventional adaptive control algorithms work almost entirely in the linear domain. Although U.S. Pat. No. 3,794,817 teaches a nonlinear adaptive controller, it requires that specific system knowledge about, for example, nonlinear deadband regions, be included for the controller to function.
Model-based predictive control systems, while sometimes adaptive, are generally linear and work with relatively large time constants (greater than one second). U.S. Pat. No. 4,358,822 discloses a typical adaptive predictive controller for use in a chemical process. In this instance, the controller is a linear adaptive model predictive controller with an eight minute time constant. Conventional controllers of this type generally use state space models for predicting future states.
Although some conventional controllers use neural networks as part of their control algorithm, such controllers typically include a separate controller in addition to the neural network. For example, U.S. Pat. No. 5,311,421 discloses such a process control method and system in which a neural network estimates certain parameters which are then used by a separate controller. Another use of neural networks in control systems is to learn control signal outputs from a conventional control algorithm or from a human operator as in U.S. Pat. No. 5,197,114.
Use of a neural network within a model predictive control scheme has been demonstrated but only for systems with relatively large time constants, such as controlling pH in a neutralization reactor.
For these reasons, a nonlinear adaptive controller which is not system specific and which learns nonlinearities in a neural network is desired. Further, such a controller is desired which has a relatively fast time constant of about one millisecond or faster and which does not need to copy the actions of another controller which must first be developed.
DISCLOSURE OF THE INVENTION
The invention meets the above needs and overcomes the deficiencies of the prior art by providing an improved system and method for adaptively controlling highly dynamic systems such as reducing undesirable vibration and undesirable aeroelastic responses associated with flexible structures such as aircraft wings. This is accomplished by a neural network adaptive controller which provides improved control performance over that of a conventional fixed gain controller. Such a neural network adaptive controller uses online learning neural networks to implement an adaptive, self-optimizing controller. In addition, such system is economically feasible and commercially practical and such method can be carried out efficiently and economically.
Briefly described, one embodiment of the invention is directed to a method of controlling a dynamic nonlinear plant. An input signal controls the plant and an output signal represents a state of the plant in response to the received input signal. The method includes the steps of storing the input and output signals corresponding to m consecutive past states of the plant and generating a set of trial control inputs. The trial control inputs represent the input signal corresponding to the next n consecutive future states of the plant. The method also includes predicting a set of future output states with a computer neural network. The future output states represent the output signal corresponding to the next n consecutive future states of the plant in response to the trial control inputs and are predicted based on the past input and output signals and the future trial control inputs. The method further includes determining a performance index as a function of the future output states. In this embodiment, the performance index is indicative of plant performance over time in reponse to the trial control inputs. The method also includes the step of modifying the input signal as a function of the trial control inputs for controlling the plant so that the performance index reaches a desired value.
A system embodying aspects of the invention includes a memory storing input and output signals corresponding to m consecutive past states of a plant to be controlled. A computer neural network predicts a set of future output states representative of the output signal corresponding to the next n consecutive future states of the plant in response to a set of trial control inputs. The trial control inputs represent the input signal corresponding to the next n consecutive future states of the plant. In this embodiment, the neural network predicts the future output states based on the past input and output signals and the future trial control inputs. The system also includes a processor for generating the input signal for controlling the plant. The processor generates the trial control inputs and determines a performance index, indicative of plant performance over time in response to the trial control inputs, as a function of the future output states. The processor also modifies the input signal as a function of the trial control inputs for controlling the plant so that the performance index reaches a desired value.
Other objects and features will be in part apparent and in part pointed out hereinafter.
REFERENCES:
patent: 3794817 (1974-02-01), Shinskey
patent: 4358822 (1982-11-01), Sáchez
patent: 5197114 (1993-03-01), Skeirik
patent: 5311421 (1994-05-01), Nomura et al.
patent: 5704011 (1997-12-01), Hansen et al.
patent: 5832468 (1998-11-01), Miller et al.
patent: 5909676 (1999-06-01), Kano
Draeger, A., Engell, S., and Ranke, H., Model Predictive Control Using Neural Networks, IEEE Control Systems, Oct. 1995, pp. 61-66.
Lichtenwalner, P.F., Little, G.R., and Scott, R.C., Adaptive Neural Control of Aeroelastic Response, Proceedings of the SPIE 1996 Symposium on Smart Structures and Materials, San Diego, CA 1996.
Pado, L.E., Damle, R.D. (1996), Predictive Neuro Control of vibration in Smart Structures, 1996 SPIE Smart Structures, 1996 SPIE Smart Structures and Materials Conference.
Litchenwalner, P.F., Little, G.R., Pado, L.E., and Scott, R.C., Adaptive Neural Control for Active Flutter Suppression, ASME 1996 IMEC&E Conference Proceedings, Atlanta, GA, Nov. 1996.
Lichtenwalner Peter F.
Pado Lawrence E.
Cady Albert De
Chase Shelly A
McDonnell Douglas Corporation
Senniger Powers Leavitt & Roedel
LandOfFree
Neural network predictive control method and system does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Neural network predictive control method and system, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Neural network predictive control method and system will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2580381