The production process in manufacturing has recently become highly complex. Therefore, it is difficult to solve problems in a process, by only using techniques that depend on the knowledge and know-how of engineers. Knowledge discovery in databases (KDD) techniques are supposed to assist engineers in extracting the non-trivial characteristics of a production process that are beyond their knowledge and know-how. However, the KDD process is basically a user-driven task and such a user-driven manner is not efficient enough for use in a manufacturing application. We developed an automated data-mining system designed for quality control in manufacturing. It has three features; periodical-analysis, storing the result and extracting temporal-variances of the result. We applied it to liquid crystal display fabrication and found that the data-mining system is useful for the rapid recovery from problems of the production process.