Assessment of bearing degradation by using intrinsic mode functions and k-medoids clustering

Abstract

Recently, the prognostic is much attention in the field of vibration-based bearing monitoring and it plays a significant role to avoid accidents. In rotary machines, the bearing failure is one of the major causes of machinery shutdown. The bearing degradation monitoring is a great concern for prevention of bearing failures. This paper presents an approach for the bearing degradation evaluation based on empirical mode decomposition and k-medoids clustering. The bearing fault features are extracted from vibration data using an intrinsic mode function of empirical mode decomposition process. The extracted features are then subjected to k-medoids clustering for obtaining normal and failure state. Assurance values curve, which is based on dissimilarity data of test object to the normal state is found and retained as degradation indicator for evaluation of bearing health. Experiment was conducted to verify and assess the effectiveness of proposed method for the evaluation of performance of bearing degradation. To justify the preeminence of recommended approach, the root mean square and kurtosis features of time domain, envelope analysis of diagnosis method, and degradation assessment classifiers, i.e. simplified fuzzy adaptive resonance theory map are commonly used in the bearing analysis compared with the proposed method. Early stage detection of degradation more accurately, the recommended method is better than the time-domain features and simplified fuzzy adaptive resonance theory map based on performance degradation assessment on bearing. Moreover, envelope analysis can be used to verify the early stage defect detected by the proposed method. In this study, it has been seen that the k-medoids clustering is an efficient tool to assess the performance of degradation of bearings.

Statistical and frequency analysis of vibrations signals of roller bearings using empirical mode decomposition

“…, c i ranges from high to low frequency bands. 14,36 Feature extractions extract from signals, easy to implement with low computation time. This work proposes the extraction and use of following statistical features from bearing time domain data and IMFs.…”

Diagnosis of Induction Motor Bearing Defect Using Discrete Wavelet Transform and Artificial Neural Network

“…In order to improve the accuracy of error detection results, various machine learning was developed, such as artificial neural networks, support vector machine, genetic algorithms and fuzzy logic. Among many clustering methods, the Artificial Neural Network (ANN) algorithm clustering is utilised because of its strong robustness, anti-noise ability, and its ability to handle abnormal values [12,13,14]. From the point of view of designing applications, the ANN algorithm also has a good convergence and time complexity, and the impact acquired in global inquiries is astounding.…”