To expedite the identification of a machine failure causes, this paper proposes a work instruction method based on expected value of working hours to identify the failure cause. Failure cause identification generally includes two types of work, which are diagnosis to narrow down the candidates of failure causes and confirmation to identify the actual failure cause. In this paper, a structure of interrelations among the works is modeled using a graphical model, on whose nodes working hours, probabilities concerning failure cause existence, and reliabilities concerning diagnostic results are defined. In the proposed methods, the graphical model is divided into small groups to reduce computational complexity. An optimal work instruction is decided to minimize the expected value of working hours for each group. After instructed work has been executed, the remaining work instructions, including the next optimal work, are updated. The method was evaluated using Monte Carlo simulation, in which actual failure causes and results of diagnostic work are decided with random numbers under defined probabilities and reliabilities. In the evaluation, the proposed method is compared with conventional methods that follow a flow chart of failure cause identification and instruct only the confirmation work with the expected value of working hours. As a result, it was confirmed that failure cause identification working hours of the proposed method are shorter than those of the conventional methods.
View full abstract