Title :
Decentralised neural network control of magnetic bearings
Author :
Zmood, R.B. ; Jiang, Yuhong
Author_Institution :
Dept. of Electr. Eng., R. Melbourne Inst. of Technol., Vic., Australia
Abstract :
This paper examines the application of artificial neural network techniques to the control of a magnetic bearing system. The application of the neural network control method to a magnetic bearing system with self-excited and forced disturbances is reviewed. In modelling the system, the shaft is first discretized into eighteen finite elements and then four levels of condensation are applied. This leads to a system with six masses and six compliant elements which can be described by twelve state variables. Two-layer neural networks have been used in the simulation work. The reinforcement, error-propagation, and temporal-difference methods have been used in the neural network controller. The simulation results show low sensitivity to external periodic disturbances can be achieved for speeds up to 3000 rpm using the proposed neural network controller
Keywords :
decentralised control; electric machines; feedforward neural nets; finite element analysis; learning (artificial intelligence); machine bearings; multilayer perceptrons; neurocontrollers; rotors; velocity control; condensation; decentralised neural network control; error-propagation; finite elements; forced disturbance; learning; magnetic bearing system; modelling; periodic disturbances; reinforcement method; self-excited disturbance; simulation; speed; state variables; temporal-difference; two-layer neural networks; Artificial neural networks; Computer simulation; Control systems; Error correction; Magnetic levitation; Magnetic variables control; Neural networks; Robust stability; Rotors; Shafts;
Conference_Titel :
Electrical and Computer Engineering, 1996. Canadian Conference on
Conference_Location :
Calgary, Alta.
Print_ISBN :
0-7803-3143-5
DOI :
10.1109/CCECE.1996.548086