[in Japanese]

View full abstract

[in Japanese]

View full abstract

This paper proposes a scheme of approximate reasoning with analogical inference for purpose of flexible inference. In this method, we use truth value t ∈ [0,1] based on possibility and assign a truth value called certainty factor (cf) ∈ [0,1] to the implication A → B. This inference conclusion is derived as a product of the truth value T (A) of the fact A and certainty factor cf. We also introduce a scheme of analogical inference based on the concept "similar causes lead similar results". This analogical inference can be done by using a degree of equivalence as a similarity between statements. Moreover we describe the new combination rule of conclusions. It derives the supremum and the infimum of a conclusion and determines a meaningful value by using those bounds. Finally, it is clarified that our scheme can derive the inference result that is suited for our intuition.

View full abstract

Fuzzy Logic Programming is the theoretical foundation of Fuzzy Prolog which is a fuzzy Programming language based on fuzzy logic. In this paper, Fuzzy Logic Programming is defined such that (1) fuzzy definite clause which is a logic formula for a representation of uncertain assertion, (2)fuzzy logical consequence which is a fuzzy definite clause characterized to be correct conclusion, (3)SLD-resolution which is a procedure to obtain fuzzy logical consequences from a set of fuzzy definite clauses, (4)degree of resolution which is the value giving a fuzzy logical consequence with the highest confidence. The main results concern the soundness and completeness of SLD-resolution, that is, a fuzzy definite clause is Fuzzy logical consequence, if and only if it is resolved by SLD-resolution.

View full abstract

In this paper, a max-min solution is proposed based on possibility and necessity measures. This max-min solution is called an "extended max-min solution", since this solution is an extension of the max-min solution previously propoed by Nishizaki et al. A solution algorithm for the Nishizaki's max-min solution, has been proposed only when all components of fuzzy payoff matrices are triangular fuzzy numbers. It is shown that the extended max-min solution can be solved by the similar algorithm even when each component of fuzzy payoff matrices is an arbitrary L-R fuzzy number. Thus, the extended max-min solution can be obtained by the simplex method together with the bisection method. It has been considered that to illustrate the achivements of all fuzzy goals for a solution, n × K figures have been needed, where n is the number of pure strategies and K is the number of fuzzy goals. In this paper, it is shown that only K figures are needed to illustrate the achievements of all fuzzy goals. From this fact, it is easier to recognize all of the achievements of fuzzy goals. Thus, a decision maker can be easily checked whether his/her intended solution is obtained or not. In the last section of this paper, a numerical example is given to exemplify the proposed methods.

View full abstract

The purpose of this paper is to propose an additive fuzzy clustering model for similarity data and investigate its characteristic features. This method is regarded as a structural analysis of the similarity between the pair of objects. The cluster is defined, in this method, as the subset in which the objects share a common property. Then the similarity represents the degree of shared properties. Although, in a hard clustering method, the additive clustering models have been discussed, the number of clusters tend to increase describing the structure of observed similarities, because the models are discussed based on whether an object belongs to one cluster or not, namely, whether each object has the common property or not. Then we will show that it is rather natural to describe the model based on the fuzzy clusters because such a model will be able to explain the similarity by using fewer substantial clusters.

View full abstract

Computerized tomography using Radon and Fourier Transform has revolutionized medical X-ray imaging and non-destructive examination because of its ability to reconstruct the spatial distribution of X-ray attenuation over cross sections. In geophysical tomography, either electromagnetic energy or seismic energy is used and an iterative reconstruction method has been proposed. In this paper a method of computerized tomography using Neuro-Fuzzy model is proposed. With this method, low contrast pictures of the spatial distribution of attenuation or propagation velocity, whose changes are moderate, can be reconstructed. The Neuro-Fuzzy in this paper is an iterative learning algorithm using fuzzy models consist of Gaussian membership functions. The line integrals of the fuzzy model can be obtained in a simple manner and the spatial distribution is calculated from the line integrals along rays in a plane. This simple formula of the integration of Gaussian radial basis functions is applied to the computerized tomography with relatively small number of propagation paths. A straight-line rayoptics model is assumed for the propagation mechanism. A numerical example of computed tomography by the direct method of solution using the gradient descent method is presented.

View full abstract

An palpation is one of the most important means of the diagnosis. We have developed a master-slave micro-manipulator system for palpating even the impalpable part. By operating this system, an operator can perceive the stiffness of small part of tissue with his/her tactile sense. If an operator knows the value of tissue's stiffness evaluated by computer, he/she can judge the stiffness with more confidence. To this end, we propose the method of estimating the tissue's stiffness by the fuzzy reasoning in this paper. We incorporate the expert's subjective scale of stiffness based on their experience in their own field into this method. And, the estimation are represented by linguistic values, not by numerical values of the physical parameter that depends on the stiffness

View full abstract

In these days, many fuzzy database systems (FDBS) which can handle imprecise information by using fuzzy logic, have been proposed. But few FDBSs were designed considering the difference of uncertainty which we could observe among users' subjectivities. It is clear that each user has his/her unique subjectivity, and conventional FDBSs cannot be suitable for every users' subjective fuzziness. In this paper, we propose a FDBS which can automatically tune each user's membership functions which represent their subjective fuzziness. In addition, our system can handle nonnumerical fuzzy attributes, like "beauty", as well as numerical fuzzy attributes like "young".

View full abstract

This paper describes a follow-up system by fuzzy rules. In this system, fuzzy rules are modified by use of square-error between desired signal and estimated output of fuzzy system. To correct the error, fuzzy rules are modified by the system composed of two fuzzy controllers. The system compares two square-errors of two fuzzy controllers for modification of fuzzy rules. New modification value of fuzzy rules for the main controller is modification value provided to the controller that smaller square-error is obtained. A random value is utilized for defining the modification value of fuzzy rules for the subcontroller. It is clear that estimation for modification value of minimum error is possible by this system from simulation results.

View full abstract

Since fuzzy control was put into practical use, in Japan it has been applied in many industrial process control fields. The greatest problem in realizing fuzzy control is acquiring the operational knowledge. Therefore, many studies have been carried out on studies of automatic design and tuning methods on control rules and membership functions using optimization, neural network and other theory. However, introducing such new techniques produces some new problems, defecting advantages of fuzzy control such as easy design, understandable and linguistic representation, because the process model will be too complex to use in the practical fields. In this paper, we discuss the feature of fuzzy control for appling to process control. First, we show that it is necessary for a process control system to monitor the state and to operate it by operators, because of remaining uncertainties of plant characteristics and its changes. Therefore, we describe that it is impossible to get complete informations for control system design and it is required for operators to understand the fuzzy control system in order to monitor and operate it successfuly. Second, we suggest a design, tuning and estimation method of control rules and membership functions based on operational knowhows and data. And we discuss the function of fuzzy controller in case of the tuning of control rules and membership functions by operator for satisfying these requirements.

View full abstract

As an effective control structure for interconnected system, decentralized control scheme has been proposed, which distribute the control and information to each control station and accomplish the common goal by multiple controllers located at each control station. For the case that exact models of local subsystems can be obtained, the authors proposed a practical decentralized control scheme, which consists of hybrid controller, without exchanging information among its subsystems. In this paper, we consider identification and control problems for unknown subsystems by regarding interactions among subsystems as unknown disturbance for each subsystem. At first, we propose a robust adaptive algorithm, without assuming any statistical property of disturbance, which applies a fuzzy dead zone set based on the information about the upper bound of disturbance. Furthermore, we propose a fuzzy indentification system which uses a fuzzy dynamic linear model and a fuzzy dead zone set for the case that it is impossible to get any information about the upper bound of disturbance. It is shown that the proposed identification algorithm using the fuzzy dead zone set and the fuzzy identification model attains robustness and higher accuracy in identification compared with other methods.

View full abstract

In the operation of an overhead-traveling crane used for the loading work at ironworks and various factories, the load is required to arrive to the target point accurately without swing. At this cooperation of the opposite control aims, it is necessary to use skilled operator's technique. Then, there are many papers on the automatic operation using optimal control theory or fuzzy control logic. But, their algorithms are not on the cooperative ones and they are still on experimental scales, not practical use, or computer simulations. We have developed a cooperative method which is satisfied with positioning and swing prevention control of an overhead-traveling crane by using fuzzy control logic. After studing by simulation, this algorithm has been implemented on a programable controller and used to a practical crane. For the purpose of showing its effectiveness, we have carried out the experiments and compared with conventional methods. Then we have confirmed the fact that it is possible to reduce the swing of the load and arrive accurately at a goal position within 2.3 times of natural period of wire rope. Here, we discuss about these results.

View full abstract