Monad is an object oriented language designed for building AI systems. AI language should provide the user with a means for organizing knowledge in a flexible and modular manner. Towards this end, it would also be useful to allow the user to write programs without worrying about control structures. To achieve these goals, we have incorporated into Monad the following features: (l) The user can program methods as rule based production systems. He can write easily not only non-deterministic programs but also deterministic programs. (2) The user can use patterns to catch messages. He can use such messages that contain two or more selectors and contain no selector (only arguments). The order of the selectors and arguments is optional. (3) The user can make use of multiple inheritance and specify searching path. The specification is done by special purpose reserved variable. It is also possible to combine methods defined in the super classes. (4) The user can define event-driven methods which will be invoked when a certain condition hold. By using this facility, it is easy to keep consistency of information in objects. We have developed the Monad interpreter in

-D on 1100 SIP and confirmed that it is effective for AI programming.

抄録全体を表示

Design methods for gear drive devices are well described in texbooks and handbooks, since the type of such design problems are recognized as routine design. Though the design methodology for individual machine elements has been formalized, it is still difficult and time-consuming to arrange total systems, such as gear drive devices, which are composed of a set of machine elements. In this paper, we develop a computer-aided design system for the above design problems. In the system, design knowledge for the individual machine elements is modularized as "functional modules" in the manner of object orientation, and the operation and control knowledge for design are integrated with it based on the machine structure defined by the designer. The system can deal with several classes of gear drive devices and automate the various design calculation in such design processes. Finally, we show examples of its application to design problems of gear reducers.

抄録全体を表示

This study is aimed at the investigation of the possibility of a problem-solving mechanism in practical engineering problems. Problems are composed of two classes of elements. The element in one class is defined as information flow, and the element in the other class is termed as a device. The information flow has static constraints on the relations in a state observed at one measuring point. The device has multiple constraints which describe relations between states. The information flow and the device have local problem-independent abilities which represent subproblems and their solutions. Problems are formulated by a graph in which an information flow and a device are adjacent to each other. The system of equations which governs the given problem is derived by using the fact that for one physical parameter, the value obtained by static constraints must be equal to the value deduced from multiple constraints in devices.

抄録全体を表示

The design procedure generally consists of many processes, from concept design to the release of parts and assembly drawings. Over all these procedures, designers always analyse various design requirements and conduct the information processing so as to have the most suitable result. Consequently, to rationally automatise such a design procedure, it is necessary to effectively process the design and its related information, e. g., decision making, by the computer system. In this paper, therefore, to establish a future CAD system of artificial intelligence type, the decision making in the basic layout design process of machine tools has been investigated and a decision making method using the multi-dimensional evaluation vectors has been proposed. The validity of the proposed decision making method has also been verified through the development of a system for selecting the suitable bearing type and arrangement for the main spindle bearing system from many candidates.

抄録全体を表示

A knowledge-based system for operation guidance of nuclear power plants is proposed The Dynamic Procedure Management System (DPMS) is designed and developed to assist human operators interactively by selecting and modifying predefined operation procedures in a dynamic situation. Unlike most operation guidance systems, DPMS has been built based on Generic Task Methodology, which makes the overall framework of the system perspicuous and also lets domain knowledge be represented in a natural way. This paper describes the organization of the system, the definition of each task, and the form and organization of knowledge, followed by an application example.

抄録全体を表示