NEURO-FUZZY CONTROLLER BASED ON CEREBELLAR MODEL ARTICULATION AND LEAST SQUARE ESTIMATION

Abstract

Magnetic levitation system has attracted the attention of the control engineers and researchers due to its non linear and unstable characteristics. Magnetic fields are conservative forces and therefore system has no built-in damping. This permits vibration to exist in the motion of the ball and makes the system unstable. Since no mechanical support is given to the ball damping of motion of the ball is done using electromagnets controlled by electronics. Neuro-fuzzy control technique is a novel way to control and stabilize the complex nonlinear system. Neuro-fuzzy refers to combinations of artificial neural networks and fuzzy logic. In this dissertation, Cerebellar modal articulation control (CMAC) technique along with least square estimation (LSE) technique is used as the learning algorithm for neural network. Neuro Fuzzy controller based on Cerebellar model Articulation and Least Square estimation (NFCALS) is implemented to stabilize the system. CMAC provides robust control in the presence of noise, table based computation helps in performing same task multiple times and very fast. Fully learned CMAC weights help in damping out the vibration making the system dynamically stable. LSE keeps system up-to date using recursive adaptation of model parameters by distributing them normally. This creates a net force to push back the magnetic object if any small displacement occurs making the system static stable.