In this study, we focus on a phenomenon in which a manufacturing system suddenly begins yielding many defective products during production processes. Especially, we consider a statistical method of estimating mean shift for a fraction defective of population of the products. In order to treat this type of problems, the CUSUM(cumulative sum) method is widely known as a traditional method for this estimation problem. The method enables us to find when the mean shift phenomenon occurs under several assumptions and conditions. In contrast, from the view point of stochastic processes, this estimation problem can be dealt with a Markov process as follows. The process is assumed that it has two states, one is good (fraction defective is low) and the other bad(fraction defective is high),and starts in the good state with probability one. Then we are interested in judging when the state has moved to the bad state by analyzing the observed results. (i.e. the product is good or bad). Under this modeling, HMM(hidden Markov model) can grately help us to find the time point of mean shift occurrence. Also in our previous work, we have already found that these two estimation methods have advantage points and disadvantage ones vice versa. Thus the main goal of this study is to propose a new hybrid method based on the CUSUM method and HMM,and assess its performance by using a simulation approach.

114 工学部 (旧) ・工学研究科 ＞ 学位論文 ＞ 修士論文