Optimal Statistical Feature Subset Selection for Bearing Fault Detection and Severity Estimation

Abstract

The performance of bearing fault detection systems based on machine learning techniques largely depends on the selected features. Hence, selection of an ideal number of dominant features from a comprehensive list of features is needed to decrease the number of computations involved in fault detection. In this paper, we attempted statistical time-domain features, namely, Hjorth parameters (activity, mobility, and complexity) and normal negative log likelihood for Gaussian mixture model (GMM) for the first time in addition to 26 other established statistical features for identification of bearing fault type and severity. Two datasets are derived from a publicly available database of Case Western Reserve University to identify the capability of features in fault identification under various fault sizes and motor loads. Features have been investigated using a two-step approach—filter-based ranking with 3 metrics followed by feature subset selection with 11 search techniques. The results indicate that the set of features root mean square, geometric mean, zero crossing rate, Hjorth parameter—mobility, and normal negative log likelihood for GMM outperforms other features. We also compared the diagnostic performance of normal negative log likelihood for GMM with the established feature normal negative log likelihood for single Gaussian. The selected set of statistical features is validated using ensemble rule-based classifiers and showed an average accuracy of 96.75% with proposed statistical features subset and 99.63% with all 30 features. F-measure and G-mean scores are also calculated to investigate their performance on datasets with class imbalance. The diagnostic effectiveness of the features was further validated on a bearing dataset obtained from an operating thermal power plant. The results obtained show that our newly proposed feature subset plays a major role in achieving good classification results and has a future potential of being used in a high-dimensional dataset with multidomain features.

Entropy Indicators: An Approach for Low-Speed Bearing Diagnosis

“…As shown in Equation (38), PerEn provides a scalar value. SepEn requires the generation of a signal based on PerEn, which is calculated using a certain data window.…”

“…The kurtosis, covariance, root mean square (RMS), and skewness are indicators that are widely used in the literature [ 34 , 35 ]. In addition, other indicators have been developed in the time domain, such as factors [ 36 , 37 , 38 ], and Hjorth parameters [ 39 , 40 , 41 ]. In the frequency domain, several statistical indicators have their counterparts from the time domain, which are helpful for frequency analysis [ 42 ].…”

Enhancing the Classification Performance of Machine Learning Techniques by Using Hjorth's and Other Statistical Parameters for Precise Tracking of Naturally Evolving Faults in Ball Bearings

“…33 Earlier, these parameters were more common in biomedical fields, 34 however, with recognized potential of fault characterization, the Hjorth's parameters have been successfully applied in some of the recent research works. [35][36][37][38][39][40][41] Caesarendra and Tjahjowidodo 35 studied the changes in statistical features, including Hjorth's parameters, as the natural fault develops in slow speed slew bearings, and found that impulse factor, margin factor, approximate entropy and largest Lyapunov exponent possess the fault indicative potentials. Among Hjorth's three parameters, only activity was found sensitive to the occurrence of fault.…”

Health Condition Estimation of Bearings with Multiple Faults by a Composite Learning-Based Approach

“…Similarly, Wan et al [ 23 ] compared the performance of different inputs in their research by using eight distinct types of inputs namely STFT, Constant-Q Gabor Transform, Instantaneous Frequency, Fast Kurtogram, HHT, Wigner–Ville Distribution, Fourier Synchro-squeezed Transform (FST), and CWT. Other academics have used statistical features for rotating machinery diagnostic systems [ 24 , 25 , 26 , 27 , 28 ]. Although the performance of fault diagnosis systems depends on the feature spaces used [ 28 ], it was observed that these feature spaces were used arbitrarily and not combined in such a way to achieve generalization between different conditions, that the bearings may be subjected.…”

“…Other academics have used statistical features for rotating machinery diagnostic systems [ 24 , 25 , 26 , 27 , 28 ]. Although the performance of fault diagnosis systems depends on the feature spaces used [ 28 ], it was observed that these feature spaces were used arbitrarily and not combined in such a way to achieve generalization between different conditions, that the bearings may be subjected. For instance, in [ 9 ] while the FFT would give unsatisfactory results with a signal whose frequency components changes with time, the wavelet transform is known to suffer from fixed scale resolution which would affect its real-life applications, and the HHT suffers from instability in its signal decomposition process.…”