A rule-based fault-tolerant neurocontroller
Automation of operation and control functions in complex processes, unattended plants, and space stations must be reliable, robust, and fault-tolerant. These requirements have been realized in the design of a rule-based automatic controller by providing redundancy and diversification. The system comprises an adaptive controller, a rule-based fuzzy controller, and a self organizing tie-breaker. The controller has been successfully tested for individually components of a plant such as heat exchangers and centrifugal pumps. The limitations of the controller is that it relies on information sensed by individual transducers. By exploiting the neural network features of fault-tolerance, robustness and learning, a neurocontroller has been designed to simultaneously retrieve and process information to complete a closest-matched response as control action. Since the information is stored globaly, the associative memory is interrogated and emergently collected to deal with the stimuli from the distributed system. The adaptive control model carried a close resemblance to the neural controller. However, adaptive control depends on the signal received from the input and output sensors and the data acquisition system. A comparative analysis between the two controllers logic, demonstrated that they are comparable in performance. However, the neurocontroller has the capability of processing a large amount of simultaneously sensed data and adjusts by transforming and recognizing the feature patterns of the input/output signals. Furthermore, the neurocontroller has the capability of modulating the system dynamics by configuring a number of interconnected neurons. Hence, fault tolerance can be improved by the neural operations of fused data. In contrast to the neural controller, the fuzzy controller may require more than one processors.
- Research Organization:
- Technology International, Inc., P.O. Box 1749, LaPlace, LA (US)
- OSTI ID:
- 6275302
- Report Number(s):
- CONF-8809132-
- Journal Information:
- Neural Networks; (United States), Vol. 1:1; Conference: 1. International Neural Network Society annual meeting, Boston, MA, USA, 6 Sep 1988
- Country of Publication:
- United States
- Language:
- English
Similar Records
Adaptive Remote-Sensing Techniques Implementing Swarms of Mobile Agents
Adaptive Sensor Optimization and Cognitive Image Processing Using Autonomous Optical Neuroprocessors
Related Subjects
CENTRIFUGAL PUMPS
CONTROL SYSTEMS
COMPUTER NETWORKS
PERFORMANCE
DATA ACQUISITION SYSTEMS
DESIGN
HEAT EXCHANGERS
ARRAY PROCESSORS
AUTOMATION
DATA-FLOW PROCESSING
DYNAMICS
FAULT TOLERANT COMPUTERS
LEARNING
MEMORY DEVICES
NERVOUS SYSTEM
PATTERN RECOGNITION
SIMULATION
SUM RULES
COMPUTERS
DIGITAL COMPUTERS
EQUATIONS
MECHANICS
PROGRAMMING
PUMPS
990210* - Supercomputers- (1987-1989)