No abstract

Selecting appropriate features from the vibration condition monitoring data of ball-bearings is one of the main challenges in the application of data-driven methods for remaining useful life prediction purpose. In this article, a new feature based on the high-frequency vibration of ball-bearings is proposed. The feed forward neural network will be used for training and prediction. The experimental data of the bearing accelerated life in the PROGNOSTIA test (published in PHM 2012 IEEE conference) are used to verify the method. The results obtained by applying new features are compared with those of two popular features in the time domain (RMS and kurtosis) for prognostic purpose. Applying the proposed feature shows more accurate estimation of the bearings’ remaining useful life.

One of the challenges in predicting the remaining useful life (RUL) of rolling element bearings (REBs) is determining a proper failure threshold (FT). In the literature, the FT is usually assumed to be a constant value of an extracted feature from the vibration signals. In this study, a degradation indicator was extracted to describe damage to REBs by applying principal component analysis (PCA) to their run-to-failure data. The relationship between this degradation indicator and the vibration peak was represented through a joint probability distribution using statistical copula models. The FT was proposed as a probability distribution based on the fluctuation increase in the vibration trend. A set of run-to-failure tests was conducted. Applying the proposed method to this dataset led to various FTs for the different failure modes that occurred. It is shown that, for inner race degradation, a higher FT can be assumed than for rolling element degradation. This could help extend the lives of REBs regarding the degrading elements. A dataset for an industrial machine was also analyzed and it is shown that the proposed model estimated a reasonable and proper FT in an actual case study.

No abstract

The objective of this research is to diagnose an inaccessible rolling bearing by indirect vibration measurement. In this study, a shaft supported with several bearings is considered. It is assumed that the vibration for at least one bearing is not recordable. The purpose is to diagnose inaccessible bearing by the recorded data from the sensors located on the other bearings. To achieve this goal, the continuous wavelet transform is used to detect weak signatures in the available vibration signals. A new criterion for adjusting the scale parameter of continuous wavelet transform is proposed based on the amplitude of the bearing characteristic frequencies. In this criterion, the optimal scale is selected to maximize the amplitude of bearing characteristic frequencies in comparison with the amplitude of the other frequencies. The results of the proposed method are compared with a popular method, energy-to-entropy ratio criterion, using two different sets of run-to-failure experimental data. Results indicate that the proposed method in this article is more effective and efficient for extracting the weak signatures and diagnosing inaccessible bearings from the recorded vibration signals.