The industrial trend toward the production of a wide variety of products in small lot sizes that has become pronounced in recent years has resulted in the replacement of economies of scale with economies of scope. The concept of relative annual profit is proposed as a generally comparable economy for process capability and capacity itself, and attempts made to theoretically establish the standard relationship between this and operation rate (utilization). This paper grasps the economy of process capability and capacity planning that will address the problem of product diversification, as the idea of relative annual profit, and presents discussions on the questions inherent in the search for such plans. Two conclusions have resulted from this discussion. First, the optimal searching direction is sought based on the "profitability of capability" and the "operation rate" that constitute this relative annual profit. Second, it has become clear it is possible to establish the aspiration level leading toward this direction through the use of the aforementioned standard relationships.

View full abstract

Investors are interested in how to hedge risk and earn return in capital markets. The pioneering paper in this field "Portfolio Selection" by H. Markowitz shows a very useful approach. But the theory is hard to adapt itself to the real markets when the number of stocks in a market is large. In this paper we discuss an adaptation for large size portfolio problems. Portfolio theory is discussed on the assumption that investors know the parameters of earnings distribution such as mean vector and covariance matrix. But in general those are estimated from historical data. Therefore the estimated covariance matrix is singular because the number of stocks is larger than that of sample. We propose two models for large size problems based on such nature. One is a model to calculate the efficient frontier quickly decreasing number of variables to covariance matrix's rank. The other is a model to reduce the number of stocks to be invested using the precision of forecasting the return of market index in the sence of back test.

View full abstract

When a shaft fits into a hole, the allowance between the diameters of shaft and hole is required to lies within some tolerance limits. In this paper, the probability of successful of fitting is discussed. When the lots of shaft and hole are divided into several (2-4) sub-lots according to their diameters, the probability of the success increases remarkably compared with that of non-division case. Further, the probability of fit depends largely on how to divide the original lots, and there exists the optimum division rule (especially, when sub-lots are three). Tables are presented showing the probability of accurate fit and how to divide the lots.

View full abstract

Inspection of finished products is necessary to ensure the required quality after all operations of jobs have been completed at workstations in a Flexible Manufacturing System. Since full inspection of finished products is ideal but too expensive to realize, we adopt a policy of partial inspection according to prespecified ratio (inspection ratio). Under this policy, the two kinds of loss caused by defective products should be considered : One is a cost for reprocessing a defective product detected by the inspection and the other is both the intangible penalty and the cost for replacing a defective product shipped without inspection. In addition, equipment costs are necessary for installing inspection devices and pallets. An optimal design problem of an inspection station is formulated to maximize the gross profit derived by subtracting the sum of the total loss and the total equipment cost from the total gain of finished products. An efficient solution method to obtain the optimal values for the number of inspection devices, inspection ratio and the number of pallets is proposed by exploiting three conjectures newly introduced.

View full abstract

A method for modeling conditional probabilities p(y|x) was suggested. The x is a number to discriminate simi1ar types in which entities in work groups were classified by the method. If an entity belongs to a similar type x, the probability that the entity attains a goal in designing work groups is p(1|x). In this paper, a method to construct the design criteria subject to the constraints in designing is suggested from the viewpoint of enhancing its effectiveness. The criteria consist of the models and effective schemes to form triplets of elements (worker, task and supervisor) in work groups.

View full abstract

Time series analysis is a representative method of demand forecasting. In time series analysis, we often replace demand with mathematical models, and estimate the values for the parameters involved. With this forecasting method, however, how we establish the mathematical models becomes a matter for debate. On the other hand, when input-output patterns are given, layered neural networks can by themselves change their inside states into optimum states to learn against the input-output patterns ; mathematical models need not be provided. In this paper, we consider demand forecasting as a kind of pattern recognition, and propose a demand forecasting procedure using layered neural networks. Moreover, we present an adaptive forecasting method that learns each time it obtains new demand data. The effectiveness of the proposed procedure is also discussed through three simulations.

View full abstract

Class scheduling is a complex combinatorial optimization problem which involves determining the most desirable schedule of lecture times and classes for each teacher under the constraints as to instructors, students, and facilities. In this paper, a class scheduling system by neurocomputing using the calicula at Hannan University as a model is presented. A quantitative comparison of the system with the previously developed rule-based system is made. It is shown that the neurocomputing-based approach is superior to the rule-based approach, especially in the case that several conflicting evaluation conditions are interwoven like the class scheduling in this study.

View full abstract

This paper discusses a class of job-shop control policies, i.e., joint policies of order-se1ection and switch-over. In a job-shop, arriving job-orders are accepted or rejected by using one of the two selection criteria : c_1 and c_2,and the job-orders accerted are processed by using one of the two available processing rates : μ_1 and μ_2(≥μ_1). A stochastic model is developed to compare the two typical job-shop control policies of periodic and dynamic type. The problem of maximizing the net reward (= price-cost) per unit time can be formulated as semi-Markov decision processes. Finally, a numerical example is given to show the structure of the optimal policies.

View full abstract

This paper describes the "causal relationship" among the system elements, and divides it into a "functional relation" that can be formulated with equation and an "impact relation" that can not be formulated, and focuses our attention on the "impact relation" since it must be considered when using the new dynamic simulation for analyzing the performance of a complex system such as social system. And in this paper, we propose several types of impact relations, the features and the transfer process of the impact action, and a calculation method of the impact force. This paper is the basis of the study on a new dynamic simulation for the system with a mixture of the impact relation and the functional relation.

View full abstract

In the factory, many kinds of materials are treated. These materials are stocked in many warehouses. It is an important problem that what items should be stocked in a warehouse, because the material-handling in the factory depends on this assignment. This problem is difficult to deal with because of many items and many points of supply ard delivery. In this paper, we try to reduce the problem scale and propose an algorithm based on the branch and bound method to assign various items to many warehouses for realizing their smooth handling. As a result, we were successful in decreasing the material distance by about 30%.

View full abstract