Seizo Fujii

A Discrete Model Reference Adaptive Control Using an Autoregressive Model with Dead Time of the Plant

Abstract A new method for designing a discrete model reference adaptive control system is presented. This method avoids the restriction of the minimum phase system by the adaptive controller designed for an autoregressive model with dead time of the plant. First, a stable adaptive control system is designed for an AR model with dead time, based on Lyapunovs direct method. The derived adaptation control algorithm can be implemented using only the available input-output data of the plant. Next, in order to illustrate the effectiveness of this model for many different types of plants, the results of computer simulation are given. Finally, the design parameters of the model are discussed.

Discrete Time Multivariable Adaptive Control for Non-Minimum Phase Plants with Unknown Dead Time

Abstract A new method for designing a discrete time multivariable adaptive control system is presented. The controlled plant is a multi-input, multi-output stable, non-minimum phase plant with unknown dead time. In the proposed method an adaptive control is carried out using a controller designed by a certain decomposed representation of the unknown plant. This method requires no a priori knowledge regarding the dead time of the plant and the plant structure, and may in many cases require less computation than other methods which have been proposed for non-minimum phase systems. The stability of the proposed adaptive control system is considered. Moreover, the relation between the proposed method and the conventional model reference adaptive control or adaptive pole placement method is clarified. Finally, the results of computer simulation of adaptive control, applied to minimum phase plant without dead time and non-minimum phase plant with dead time, are included to illustrate the effectiveness of the proposed method.

Identification of Air Conditioner as a Multi-Dimensional Auto-Regressive Model

Abstract An air conditioner consists of both the refrigerant system and the air conditioning system. The former has a compressor, heat exchangers and an expansion valve as its major components and forms by the circulating refrigerant in the components. The latter is the load of the refrigerant system. It is very difficult to construct the physical model which describes the dynamic characteristics of the air conditioner by means of analytical methods. Because there are the complexity of its configulation, unknown factors of components and the movement of material with phase transition. The authors identified the system as the multi-dimentional auto-regressive model. The obtained model provides the good fitting of real behaviors of the system and we got the good result in design of control system of products by using this model.

The Tracking Control of Mechanical Systems by an Adaptive Pole Placement Controller Combining with a Neural Network

Abstract In this paper, an adaptive controller of a plant with unknown parameters and unknown nonlinear frictional force is proposed and applied to the control of a single link robot. The measurement for control is only the motor angle. The proposed adaptive controller is composed of an adaptive pole placement controller and a neural network. The neural network is designed to compensate the nonlinear frictional force and linearize the robot plant. To improve the efficiency of learning of the neural network, an inverse model is introduced. Several considerations are made to improve the robustness of the controller. Many experiments are carried out. The experimental results show the effectiveness of developed control system.

Improvement of Convergence Rate of the Transformation Method for Constrained Optimization Problems

A transformation method with two parameters is described. A new correction of the Lagrange multiplier is presented and its theoretical results are proved. Based on the results one modified algorithm to improve the rate of convergence and make the existence of the saddle point possible is proposed. The present method is an efficient one for nonconvex function problems with constraints, does not suffer from numerical difficulties and has a rapid convergence rate. The method is applicable successfully to many nonlinear programming problems with a variety of nonconvexities and nonlinearities.