A Systematic Semi-Supervised Self-adaptable Fault Diagnostics approach in an evolving environment. (1st May 2017)

• Record Type: Journal Article
• Authors: Hu, Yang; Baraldi, Piero; Di Maio, Francesco; Zio, Enrico
• Is Part Of: Mechanical systems and signal processing. Volume 88(2017)
• Page Start: 413
• Page End: 427
• Publication Date: 2017-05-01
• Subjects: Evolving environment -- Feature selection -- Concept drift -- Drift detection -- Fault diagnostics -- Bearing faults; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2016.11.004
• Languages: English
• ISSNs: 0888-3270

Abstract

Abstract: Fault diagnostic methods are challenged by their applications to industrial components operating in evolving environments of their working conditions. To overcome this problem, we propose a Systematic Semi-Supervised Self-adaptable Fault Diagnostics approach (4SFD), which allows dynamically selecting the features to be used for performing the diagnosis, detecting the necessity of updating the diagnostic model and automatically updating it. Within the proposed approach, the main novelty is the semi-supervised feature selection method developed to dynamically select the set of features in response to the evolving environment. An artificial Gaussian and a real world bearing dataset are considered for the verification of the proposed approach. Highlights: A Systematic Semi-Supervised Self-adaptable Fault Diagnostics (4SFD) is proposed. 4SFD can detect the need for updating the diagnostic model in evolving environment. Fault features are dynamically updated in response to the evolving environment. Information of both labelled and unlabelled data are exploited in 4SFD.