摘要 |
PROBLEM TO BE SOLVED: To solve problems for example, when classifying failure causes by comparison of utilization data during machine failure, and utilization data of a past case example, in such a method, an amount of pieces of utilization data of the past case example affects a classification performance of the failure causes, and the utilization data similar to that of the past case example must exist for proper classification of the utilization data during machine failure occurrence time, therefore, in a conventional art, a sufficient amount of pieces of utilization data of the past case example, must be prepared.SOLUTION: The present invention analyzes sensor data during machine failure, then expressed by probability, time series variation of pieces of utilization data for every failure cause, thereby the pieces of utilization data which cannot be properly classified in a conventional art, can be properly classified.SELECTED DRAWING: Figure 1 |