| Summary: | 碩士 === 大同大學 === 電機工程學系(所) === 99 === Variations in the values of the wind resistance, the friction coefficients of the spring and tensioner, DC motor, cable and other nonlinear dynamics may highly affect the performance of the platform screen door (PSD). A speed control of the PSD is implemented in this paper via an self-constructing type-2 fuzzy neural network (SCT2FNN) controller. The SCT2FNN controller is composed of a type-2 fuzzy neural network (T2FNN) controller, self-constructing learning algorithm and on-line learning algorithm. The T2FNN controller is main controller which computes the pulse-width modulation (PWM) duty of the DC motor to control the PSD. The structure and parameter learning are done automatic and online. The Mahalanobis distance (M-distance) method in the self-constructing learning algorithm is used to determine if the T2FNN rules are generated/eliminated or not. The on-line learning algorithm is based on the back-propagation method to update the parameters (means, standard deviations and weights) of the T2FNN using a delta law. Finally, a PSD speed control system considering wide resistance and friction coefficients is implemented in this paper to compare the other controllers and proposed SCT2FNN controller.
|
|---|
