This paper deals with modeling preference of a person when there exists ignorance resulting from the lack of belief. We introduce Dempster-Shafer theory of probability into a measurable value function under uncertainty, where the degree of pessimism is included in our preference model. The descriptive model proposed in this paper could handle the set element of the results in the alternatives. Therefore, the model could distinguish between “lack of belief” and “disbelief” in the preference. Consequently, our model could properly account for Ellsburg paradox which has been hard to explain using the previous utility models. This paper also describes how to determine an appropriate pessimism function.
To realize a low-order thickness control system for tandem cold mills, a multivariable control algorithm based on decentralized technique has been studied. In the algorithm, the static interaction and the dynamic one are treated separately. The static interaction due to interstands tension produces steady state thickness errors at neighboring two stands. While, the dynamic interaction between thickness and tension control weaken the effect of each other. The steady state thickness is optimally controlled using a multivariable control method. The dynamic response at each stand is improved by a new thickness and tension control sub-system in which the interaction is eliminate a by the non-interaction control. The low-order system based on decentralized concept facilitates parameters adjustment and system maintenance. This system has been applied to the tandem cold mill and the off-gauge length corresponding to non-permissible thickness deviation is reduced by more than 20 % compared with that before application.
The classical facility location problem is “Given a finite set of locations of potential facilities which produce a single product, and a set of customers with known demands, find which location should be built so as to minimize the total cost.” The total cost consists of transportation cost between facilities and customers, and fixed cost. In this paper, the problem is extended to allow the selection of different facility scale at each location, and to consider operation cost associated with facilities. In addition to constraints for classical facility location problem, a transportation constraint is imposed, in which each customer must be assigned exclusively to a single facility. This problem is first formulated as a pure 0-1 integer nonlinear programming. A solution technique based on the branch and bound method is developed. A procedure for finding good lower bounds is presented, and the lower bound is obtained by solving a relaxation problem using column generation technique. Computational results are also reported.
This paper deals with utilization of visual speech information as the first step to realize a system that combines visual and acoustic information. In order to grasp how humans utilize visual information for speech recognition, an experiment on the visual word recognition is performed with the hearing-impaired subject who generally practices lip-reading. Based on the experimental result, a lip-reading system is developed : some feature values are extracted from visual speech information using techniques-of image processing, and they are used for word recognition. It is found that the visual speech information is effective as a means of selecting the word candidate.