[in Japanese]

View full abstract

[in Japanese]

View full abstract

A human's subjective evaluation of some given objects, such as real things or abstract concepts, can be viewed as a regression process on the class of clustered concepts formed on the basis of knowledge and context. As the first step of this approach, we restrict our concerns not to how to understand, manipulate and represent the concepts formed on the basis of his/her knowledge and context, but to how to cluster the given attributes into macro attributes and then regress on the clustered macto attributes. In this paper, we propose a clusterwise regression-type model for the subjective evaluation process using non-monotonic fuzzy measures and the Croquet integral. A heuristic algorithm using AIC (Akaike's Information Criterion) and properties of coveringe os inclusion and exclusion are devised to identify an optimal model. We apply the model to some data sets obtained from real sensory evaluation. The experimental results of sensory evaluation of this data support the approach that subjective evaluation can be viewed as a process of regression on a hierachical structure formed on the basis of knowledge and context.

View full abstract

This paper presents the extended fuzzy relation. In conventional fuzzy relation equiations, elements of matrices are expressed by real numbers. This expression prevents us fromdealing with vague phenopmenons. In this paper, elements of the matrices are expressed by fuzzy variables. Yherefore the inverseproblems of fuzzy relation equation are extended to be capable of deal with fuzziness. Generally, the process of getting fuzzy solutions are very complicated. Converting a fuzzy matrices to several matrices whose element are real numners, the easy method of getting fuzzy solutions of these inverse problems arerealized.

View full abstract

Fuzzy Flexible Interpretive Structural Modeling (FISM/fuzzy) is a fuzzy version of Flexible Interpretive Structural Modeling (FISM). The fuzzy transitive embedding is a problem of how to efficiently fill the fuzzy reachability matrix. In this paper a process and strategy for the fuzzy transistive embedding of FISM/fuzzy are proposed.To perform the fuzzy transitive embedding logically and effectively, a fuzzy partially filled reachability matrix (FPR-matrix) is proposed. FPR-matrix is an extension of a fuzzy reachability matrix ans has great utility in the process of developing a fuzzy reachability matrix. The implication theory for FPR-matrix is applied to perform the fuzzy transitive embedding of FISM/fuzzy.In this paper, the algorithm and the strategy for the fuzzy transitive embedding are proposed. Use of the algorithm and strategy makes it possible to do a flexible and an efficient modeling for complex and fuzzy systems.

View full abstract

It is a well know problem of the conventional fuzzy reasoning that the fuzziness of inferred results gradually increases accprding to the progress of the inference. In this paper, we discuss a combination function which can reduce the fuzziness measured as the degree of crisp. Essential problems are (1) the combined result obtained with a conventional combination function becomes close to one of the two non-fuzzy values, that is, grade 0 or 1,and never approaches to the other non-fuzzy value, and (2) lack of reinforcement property. We proposed a new combination function which resolves the problems by introducing equilibrium E and dependency factors a and b. In this paper, first, the basic idea and the calculation flow are briefly explained. Aecondly, semantics of the parameters and the coverage of the function are described as important features. Finally, the learning process of the parameters is demonstrated by using the fuzzy inference software environment called FINEST which is being developed at the Laboratory for International Fuzzy Engineering Research in Japan.

View full abstract

This paper proposes a learning algorithm of fuzzy weights that are given as non-symmetric trapezoid fuzzy numbers in three-layer feedforward fuzzy neural networks. In the proposed learning algorithm, adjustment rules for the four parameters of each fuzzy weight are derived from a cost function defined for the level sets of fuzzy number outputs and fuzzy number targets. Since non-symmetric trapezoid fuzzy numbers include real numbers, intervals and triangular fuzzy numbers as special cases, the learning algorithm proposed in this paper can be viewed as a generalization of our former studies on fuzzy neural networks and interval neural networks. It is demonstrated by computer simulations that the ability of fuzzy neural networks to implement fuzzy if-then rules is drastically improved by such a generalization.

View full abstract

Recently, genetic algorithms, a new lerning paradigm that models a natural evolution mechanism, have received a great deal of attention regarding their potential as optimization techniques for solving combinatorial optimization problems. In this paper, we focus on multiobjective 0-1 programming problems and propose an interactive fuzzy satisficing method by incorporating the desirable features of both the interactive fuzzy programming methods and genetic algorithms. By considering the imprecise nature of human judgements, we assume that the decision maker (DM) may have a fuzzy goal for each of the objective functions. Having elicited the corresponding linear membership functions, if the DM specifies the reference membership levels for all the membership functions, the corresponding Pareto optimal solution which is, in the minimax sense, nearest to the requirement can be obtained by solving the minimax problem. In order to generate Pareto optimal solutions by applying the proposed genetic algorithm which was modified to generate only feasible solutions, we further revise the algorthm by providing some new mechanism for forming an initial population after the first interaction with the DM. Illustrative numerical examples demonstrate the both feasibility and efficiency of the proposed methods

View full abstract

In this paper, we present a sytem for generation of brush characters. The system converts kanji characters which user draws with a mouse or a pen into brush-style kanji characters. The user can operate the system with ease, and can generate natural-looking characters. The brush characters are output on the display of a workstation. kanji characters are fundamentally composed of strokes. The database of the system has 46 strokes. The user can freely create kanji characters by a combination of several strokes. In short, the system converts one-stroke-at-a-time input to brush strokes. To recognize an input stroke, the system uses a neural network. The outout brush stroke reflects individual information of the user (the position of an input stroke, the size, the angle, and the writing speed). The writing speed is evaluated by fuzzy interpretation and expressed as scratchiness and bluriness in output brush characters.

View full abstract

This paper deals with the construction of consumer's decision support system which is based on the analysis of consumer's purchase behavior. This system is developed from the data retrieval system discussed in the previous paper. In controls the number of recommended items for the sake of consumer's convenience, automatically arranging the retrieval condition. As an example, the used car consumer's decision support system is presented.

View full abstract

In the state synthesis evaluation of the plant, even if each state can be recognized correctly, the meaning of the state changes depending on other states that exist at the same time, that is, the context information. This makes the intefrated evaluation method suitable for the state interpretation important as well as the state interpretation method. This paper proposes the intellectual evaluation method that achieves, in a hierarchical structure, state recognition with the language label of numerical information, state interpretation using special knowledge of the experienced cases, and state synthesis evaluation by the scenario interation.The characteristic of this method is to decrease the vagueness of the operator and the information on the evaluation object. This paper proposes "state interpretation by fuzzy memory-based reasoning" and "state synthesis evaluation by the identification of the scenario synthesis evaluation standard and fuzzy measure", and makes a basic examination of the state synthesis evaluation proposed, by imitating the mechanism by which the operator evaluates and judges the plant.In this paper, we apply the method to adjustment of the design specifications of a certain cooling system plant as an example, showing that even a small number of cases are adequate to acquire a correct state synthesis evaluation and to extract the knowledge to adjust the design specifications. Furthermore, we show that the state synthesis evaluation proposed is effective to the field where the evaluation and the judgement of operator's decision making are accompanied.

View full abstract

A testing method of fuzzy control knowledge including reles and membership-functions based on measured cases has been proposed. It is used to validate fuzzy control knowledge for applying the knowkedge to plant control, and makes it possible to expose wrong or hardly-hitted rules. It reduces not only the time of inference but also the maintenance cost that plays an important role in real application using rule-base methodology including fuzzy control.Requirements of sophisticated fuzzy control knowledge are made clear first. Then five indices of degrees of reference, covering, effectiveness, contradiction and overlapping are introduced. A fuzzy control application to chemical addition in a water purification plant has been done to confirm the improvement of the control knowledge (detection of unknown situations by even professionals and so on) and man-day saving (from expert's seven days to beginner's five days).

View full abstract

Fuzzy control has a distinguishing feature in that it can incorporate experts' control rules using linguistic expressions. One of the main problems of the fuzzy control is the difficulty of acquiring the fuzzy rules and tuning the membership functions.The conventional control theory to design controllers using models of controlled objects has been established. Many research works, also, have been done on the design of fuzzy control systems using fuzzy models of controlled objects.This paper presents a linguistic design method of the fuzzy controller using the fuzzy model of the controlled object. The fuzzy model is identified by a fuzzy neural network (FNN). Although the designed fuzzy control rules are not exact, the dynamic behavior of the designed controller can be described and checled withe the fuzzy labels of the fuzzy controller and the fuzzy model. As a result, the designer of the controller can use the designed controller to the real system safely. Modifications of the control rules are also made possible from the known behavior of the control system. An adaptive tuning of the control rules of the FNN is studied. Simulations were done to verify the proposed control system.

View full abstract

This paper presents active and semi-active controls of suspension for compact passenger cars from the view-points of improvements of ride quality and vibration y using fuzzy reasoning. The cars are modelled by systems with four degrees of freedom, in which the restoring forces by suspension spings and the damping forces by shock absorbers are assumed to have nonlinear relations of suspension strokes and their derivatives, respectively. The strokes at the front/rear suspension locations are measured at the sampling instants. In the fuzzy reasoning, the measurements and their derivatives are treated as input variables in if-then fuzzy control rules. While, in the design fo active controls, the output variables are denoted as the control forces toe be inserted in the suspension locations, and in the design of semi-active controls, those are denoted as the damping forces of shock absorbers. The active and semi-active controls are determined so as to minimize the accelerations and displacements of vehicle body. That is, in the tuning of control policies, fuzzy control rules are assumed at the first step and secondly scale factors of membership functions characterizing fuzzy sets are modified, where two kinds of product-max gravity methods and two kinds of product-sum ones are considered as the defuzzyfier ones. The simulation results show that the proposed active and semi-active controls are more improved in suspension performance than passive control.

View full abstract

The stepping motors are used as the driver for positioning in industry applications. It is well-known that the transient response of stepping motor is generally oscillatory in the case of open-loop operation. The rotor oscillation must be suppressed in quicl positioning applications. Accordingly, many researches about rotor positioning and suppression of rotor oscillation of stepping motors have been reported in recent years.In this paper, the suppression control of the rotor oscullation of variable reluctance stepping motors are studied. The supprssion control is acheved by the inverse-phase excitation damping using fuzzy reasoning. The availability of the proposed control is shown by numerical simulations and experiments. As the result, it has been found that the rotor oscillation effectively vanishes by simple excitation sequence. Therefore, the quick positioning control of rotor position is possible. Since the rule type fuzzy reasoning of the proposed control is made up easily, the construction cost of the proposed damping control would be inexpensive.

View full abstract