データベース駆動型モデリングによる応力推定と異常応力検知

抄録

If the stresses applied to a machine during operation can be estimated, the fatigue state and life can be calculated from the stress state. It is also possible to detect abnormal stresses that can lead to failure. In this paper, we propose a method for detecting abnormal stress in an excavator attachment, based on stress estimated through database-driven modeling. To suppress false detection, the reliability of the estimated stress is evaluated, and abnormal stress detection is performed for highly reliable stress. The proposed abnormality detection method employs an autoencoder with unsupervised learning, using wavelet coefficients obtained via wavelet transform as features. The effectiveness of the proposed method is demonstrated by validating it with actual data.