Kiyoshi Yoneda

Job shop configuration with queueing networks and simulated annealing

A novel method for equipment planning is proposed for job type production systems, and computer experiment results obtained with a semiconductor fab model are given. The production system is modeled as an open queueing network (QN). The number of servers in the shops are then optimized by simulated annealing. The experiment demonstrated that the process considered is far faster than configuring the job shops manually, and that the configuration obtained is optimum, thus resolving the disadvantages of the traditional method. An advantage of the approach presented is that a wider range of parameter (e.g. product mix) optimization can be handled with the same combination of QN and annealing.<<ETX>>

Mixed-model assembly line scheduling using the Lagrangian relaxation technique

The increasing market demand for product variety forces the manufacturers to design a mixed-model assembly line on which different model products can be switched quickly to satisfy diversified demand. Scheduling is critical to generate a sequence of different products to meet due dates and to avoid possible component shortage or starvation. Because of the combinatorial nature of integer programming and large sizes of real problems, the scheduling for mixed-model assembly line has been recognized to be extremely difficult. In this paper, an optimization-base scheduling method for mixed-model assembly line is presented. The study is motivated by the design and implementation of a scheduling system for the compressor assembly lines of Toshiba. In the shop, an assembly line can be simplified as a single machine and the number of components can be supplied is limited within fixed length windows. A separable integer programming formulation is developed and a solution methodology based on a combined Lagrangian relaxation and heuristic can be applied to get near-optimal schedule. The method has been implemented using the object-oriented programming language C++, and numerical testing shows that the generated sequence will help on-time product delivery, low finished goods inventory and smooth usage of components.

Statistical Identification of Satellites' Heat Transfer Characteristics

A PROBLEM sometimes referred to as the backward analysis of heat transfer in a satellite is treated. Given heat quantities input and/or generated, and the component temperatures, the heat transfer coefficients (HTC) for conduction and radiation are sought. The mathematical formulation results in an ill-conditioned, noisy set of linear equations. Solution methods for the system of equations are discussed.