Shigehiro Masui

Design of quasi-optimal fuzzy controller by fuzzy dynamic programming

A design method is described for new fuzzy feedback systems which contains the idea of optimal control. The fuzzy optimal control rules are selected when the state variables are apart from the goal, and are switched to ordinary fuzzy proportional-integral-derivative (PID) control rules automatically near the goal to solve the problem of stability and offset. For optimal control, the phase plane is digitized and the optimal output of the controller is obtained by applying fuzzy dynamic programming which has linguistically explained goals and constraints to each state of the phase plane. Quasi-optimal feedback control rules are obtained by describing the optimal output of the controller for each state as fuzzy rules whose inputs are state variables. The effectiveness of the system designed with this method was verified by simulations. It was found that the characteristics of the system were similar to that of fuzzy sliding mode control.<<ETX>>

Man-machine system for triple inverted pendulum by fuzzy logic and genetic algorithm

We have studied a man-machine system: the co-operation of a man with a fuzzy controller on a very unstable control object. To decrease human operation, the design of a smart fuzzy logic controller is necessary. However, its membership function is so sensitive that its design is not easy. In this paper, we propose a method to tune the fuzzy logic controller by using genetic algorithms (GAs). The membership functions are made to evolve the controller gradually based on the evaluation standard such as the control error minimum.

Coloring of a landscape by fuzzy logic

An automatic coloring system where a colorless landscape drawn on a CRT is suitably painted by a computer when a human represents the image with a natural language is proposed. It is hard to represent the rules of coloring with mathematical equations, because color is not only a physical phenomenon but also a psychological one. The color, object, image; and coloring rules were expressed with natural language and changed to fuzzy sets. The selection of color was performed by approximate reasoning. As an example, a landscape was painted suitably when the human indicated the month and time.<<ETX>>

Manual control of an intrinsically unstable system and its modeling by fuzzy logic

Abstract The authors have shown in a previous paper that human operators could control stably such an intrinsically unstable object as a double inverted pendulum, when adequate information was given. In this paper, the authors suggest a model of human action in this control system by using fuzzy logic. This model consists of fuzzy and nonfuzzy controllers. The former represents three different modes of control action qualitatively, and the latter switches the control modes according to the situation of the controlled object. The membership functions of the fuzzy controllers are simple piecewise linear functions. Both the membership functions of the fuzzy controllers and the switching conditions of the nonfuzzy controllers are identified experimentally. The control action of this model is very similar to that of a human operator, and the double inverted pendulum can be controlled automatically by this model.

Decision making support system for house purchase by using fuzzy logic

Abstract In this paper, we suggest a decision making support system for house purchasers, using fuzzy inference and hierarchic structure of evaluation. Main part of this system consist of macro and micro evaluation. Essential factors are taken into account in macro evaluation, and unessential detailed factors are considered later in micro evaluation. By adopting this structure, many decision makers could get their most suitable result.

Fuzzy optimal control applying dynamic programming

The authors propose a design method for new fuzzy feedback systems which contain the idea of optimal control. The fuzzy optimal control rules are selected when the state variables are apart from the goal, and they are switched to ordinary fuzzy proportional-derivative (PD) control rules automatically near the goal to solve the problem of stability and off-set. For optimal control, the phase plane is digitized and the optimal output of the controller is obtained by applying fuzzy dynamic programming which

HUMAN ABILITY OF CONTROL AND ITS IMPROVEMENT BY INTERFACE

Abstract One of the aims of man-machine system is completing a high level job which is very difficult to perform by man or machine individually. The total performance of this system depends greatly on the interface, especially the information given to man is important. In this paper, we study the effects of interface by experiment in which the subjects are asked to keep stable such intrinsically unstable systems by manual control as the higher-order astatic system or the double inverted pendulum system. We show that these systems can be controlled stably, when some suitable information is displayed and the control devices are properly designed. The experimental results are collected, and some rules of human behavior and control devices are educed.

Human Ability of Control and its Improvement by Interface

Abstract One of the aims of man-machine system is completing a high level job which is very difficult to perform by man or machine individually. The total performance of this system depends greatly on the interface, especially the information given to man is important. In this paper, we study the effects of interface by experiment in which the subjects are asked to keep stable such intrinsically unstable systems by manual control as the higher-order astatic system or the double inverted pendulum system. We show that these systems can be controlled stably, when some suitable information is displayed and the control devices are properly designed. The experimental results are collected, and some rules of human behavior and control devices are educed.

Macro Fault Tree and its Application to Petro Chemical Plant-Accidents

Abstract FTA and ETA are popular method for analyzing the accidents of large scale process plants. While these are useful for the evaluation of rare but essential accidents of a specified plant, we can not use then effectively to such accidents that are caused by the combination of small and daily troubles. Most of them are based on human errors. In this paper, we introduce a new method which is effective for the analysis of this kind of troubles. We combine Fault Tree Analysis with Multi-variable Analysis by considering sets of events. The grade of occurrence and the transition of of macro-events are fuzzified in the calculation of MFT. We can obtain common characteristics of accidents through this analysis and the general counterplans are easily deduced. We apply this method to the analysis on real data of the plant-accidents happened in Kawasaki industrial area in 12 years and comfirm its effectiveness.