EEMD assisted supervised learning for the fault diagnosis of BLDC motor using vibration signal
- PDF / 5,381,737 Bytes
- 10 Pages / 595.22 x 842 pts (A4) Page_size
- 99 Downloads / 207 Views
DOI 10.1007/s12206-020-2208-7
Journal of Mechanical Science and Technology 34 (0) 2020 Original Article DOI 10.1007/s12206-020-2208-7 Keywords: · Condition monitoring · Continuous wavelet transform · Ensemble empirical mode decomposition · Fault diagnosis · Principal component analysis
Correspondence to: Jang-Wook Hur [email protected]
Citation: Shifat, T. A., Hur, J.-W. (2020). EEMD assisted supervised learning for the fault diagnosis of BLDC motor using vibration signal. Journal of Mechanical Science and Technology 34 (0) (2020) ?~?. http://doi.org/10.1007/s12206-020-2208-7
Received April 10th, 2020 Revised
May 21st, 2020
Accepted June 1st, 2020 † This paper was presented at ICMR2019, Maison Glad Jeju, Jeju, Korea, November 27-29, 2019. Recommended by Guest Editor Insu Jeon
EEMD assisted supervised learning for the fault diagnosis of BLDC motor using vibration signal Tanvir Alam Shifat and Jang-Wook Hur Department of Mechanical Engineering, Kumoh National Institute of Technology, Gumi 39177, Korea
Abstract
Predictive maintenance (PdM) has become a major issue in system health monitoring, as machines are operating under more complex and diverse conditions nowadays. Besides minimizing the risk of a catastrophic failure, a proper maintenance scheme can amplify system yield as well as largely reduce production and maintenance costs. This paper presents a comprehensive study of a permanent magnet brushless DC (BLDC) motor’s fault diagnosis using vibration signals. Based on the degree of deviation from the normal operating condition, three health states are chosen from the entire lifecycle of motor. Acquired signals are decomposed using ensemble empirical mode decomposition (EEMD) and the appropriate intrinsic mode function (IMF) is selected based on the similarity index. Later, selected IMF is analyzed in time-frequency domain by using continuous wavelet transform (CWT) for better localization of fault frequencies. Several statistical features that indicate the health state of the motor are also extracted to diagnose different fault states. Later, feature dimensions were reduced using principal component analysis (PCA) technique and classified using a supervised machine learning technique named k-nearest neighbor (KNN). Extracted IMF from EEMD provides significant fault related information to detect and diagnose different fault states. Proposed method is effectively used to diagnose fault at the incipient stage as well as classify different fault states at incipient stage and severe stage.
1. Introduction
© The Korean Society of Mechanical Engineers and Springer-Verlag GmbH Germany, part of Springer Nature 2020
In many industrial applications, condition monitoring is regarded as a profit center due to the capability of early fault detection, predicting a system’s useful life and increasing efficiency [1]. Starting from the early preventive maintenance methods, health monitoring and prognosis of engineering systems has been an extensive concern for engineers to lower the maintenance cost and higher system
Data Loading...
