A hybrid ensemble scheme for diagnosing new class defects under non-stationary and class imbalance conditions
Document Type
Conference Proceeding
Publication Date
12-9-2017
Publication Title
Proceedings - 2017 International Conference on Sensing, Diagnostics, Prognostics, and Control, SDPC 2017
Volume
2017-December
First Page
355
Keywords
Fault diagnosis, Hybrid ensemble, imbalanced data, Incremental learning, induction motor, New class defects, nonstationary environments
Last Page
360
Abstract
Strategic necessities to design and implement practical diagnostic systems are the abilities of incremental learning and diagnosing new class defects under non-stationary and class imbalance conditions. In this work, a hybrid ensemble scheme, named Learn++NCS, is adopted for diagnosing bearing defects in induction motors. This diagnostic scheme includes a feature extraction module and a hybrid ensemble scheme. The former intends to extract discriminant features from the vibrational signals. The latter collects various class imbalance sets of samples chunk by chunk from a non-stationary environment, constructs a hybrid ensemble by means of a consultation and voting mechanism, incrementally learns novel features-defects relations and diagnoses new class defects. Experimental results present the effectiveness of the proposed hybrid scheme.
DOI
10.1109/SDPC.2017.74
ISBN
9781509040209
Recommended Citation
Razavi-Far, Roozbeh; Farajzadeh-Zanjani, Maryam; Saif, Mehrdad; and Palade, Vasile. (2017). A hybrid ensemble scheme for diagnosing new class defects under non-stationary and class imbalance conditions. Proceedings - 2017 International Conference on Sensing, Diagnostics, Prognostics, and Control, SDPC 2017, 2017-December, 355-360.
https://scholar.uwindsor.ca/electricalengpub/143