Skip to Main Content
In the practical application of decouple control for MIMO system, precise mathematical model of controlled object is greatly depended on and cause unsatisfactory control effects. Complexity of algorithms based on large-scale neural network affects the practicability and real-time performance of control algorithms. A kind of MIMO decouple control system based on double-neuron adaptive predictive and static decouple algorithm was proposed. Algorithms of adaptive predicative controller based on neurons were expounded in detail. One neuron was used as adaptive controller and 3 input variables related with error were used. Gradient descent method was used in modifying weights in the neuron network. The other neuron was used as predictor. Tapped delay lines are used in the neuron predictor to input variables to the neuron. Gradient descent method was used in modifying weights for predictor neuron in online training. Former information was used to modify the predicted output values of predictor to improve prediction performance. Simulation research was done to the proposed algorithms. Satisfactory experimental results were achieved after simulation and on-site experiment, which shows the practicability and effectiveness of the controller.
Computational Intelligence and Industrial Application, 2008. PACIIA '08. Pacific-Asia Workshop on (Volume:2 )
Date of Conference: 19-20 Dec. 2008