Expert systems for planning, design and control are considered useful for their easier handling of heuristic rules. These problems may have several candidates which maximize/minimize its criterion function at almost the same satisfaction. In order to obtain the most suitable one, a quantitative evaluation function is further required for expert systems.
We have been developed an expert system accompanied with a simulator for quantitative evaluation, especially suitable for a material handling problem. In this system, rules select several candidates. Then the simulator evaluates their usefulness and suggests the most appropriate one. This system configuration improves reliability of its solution and helps to add rules easier. From the point of maintainability, in the expert system, there are two knowledge bases according to their abstraction levels. These various methods improve practicability of this system.
As its application, the material handling problem for the billet conditioning line in the steel making process is solved.

抄録全体を表示

Hypermedia is a style of building systems for information representation and management around a network of multi-media nodes connected together by typed links. Such systems have recently become quite popular due to their potential for aiding in the organization and manipulation of irregularly structured information. Many hypermedia systems, however, provide a set of predetermined nodes and links that may not be amended. This paper describes a new method to develop the organizational information structure from the view of the connectionist model. The method is based on the idea of self-organization and growth. In this paper, we describe a way of acquiring complex hierarchical structures among nodes in the networks. The growth starts from an initial interrelationship between nodes, and the network is let to organize the whole hierarchical structure. The growth process is guided through the self-organizing mechanism. The hierarchical representation model is designed for providing the global index for each node in the network that represents the hierarchical position of each node in the network. The hierarchical structured learning algorithm provides the mechanisms to create and modify the global indexes of the generated or existing nodes. With this learning algorithm, generated nodes are self-organized into the pre-existing networks. We develop the hypernetworks with self-organization and growth and show that the proposed model may represent a highly structured design for self-organizing hypermedia systems.

抄録全体を表示

In this paper, we first propose an architecture of neural networks that have fuzzy weights and fuzzy biases. A neural network with the proposed architecture maps an input vector of real numbers to an output of fuzzy number. Therefore the neural network can be viewed as an approximator of non-linear fuzzy functions from real vectors to fuzzy numbers. Next we define a cost function using the fuzzy output from the neural network and the corresponding fuzzy target output. A learning algorithm can be derived from the cost function in a similar manner as the BP (Back-Propagation) algorithm. We also define two variations of the cost function that are based on the inclusion relation between the fuzzy output and the fuzzy target. Two learning algorithms can be also derived from the two cost functions. The derived learning algorithms are illustrated by computer simulations for numerical examples.

抄録全体を表示

The auditory system has flexible functions of voice recognition in noisy environment. This ability should be due to not only the linguistic function but also binaural sound localization.
Sound localization and voice separation based on fundamental 2-dimensional acoustical field model has been proposed formerly. However, the problem becomes more complicated according to increase of speakers.
In this paper, more realistic problem of 3-dimensional sound localization and voice separation is discussed theoretically.
Validity of the algorithm for solving the problem is comfirmed by simulation analysis. That is, mixed voices of three persons which would be detected by the microphones are synthesized using voice data base. Then, the individual voice is separated after estimating presetting speaker locations on the basis of the algorithm.
Satisfactory results are obtained in this simulation analysis.

抄録全体を表示

This paper gives four types of alarm subsystem configurations which include fault-alerting and safety-presentation types, where two.kinds of correspondence between sensor states and plant states are distinguished. For each configuration, we give a probabilistic analysis on fail-safe (FS) failures and fail-dangerous (FD) failures. We prove that we can have an optimal alarm subsystem which minimizes FS and FD failures probabilities simultaneously by choosing a human-machine interface configuration and an associated safety-control policy in an appropriate manner.

抄録全体を表示