A Neural Network Approach to Bearing Health Assessment

Abstract

Vibration measurement has been widely applied to bearing condition monitoring and health assessment. To device a method for signal interpretation and automate the process of defect severity classification under varying operating conditions, a multilayer feed-forward neural network has been developed. A health index based on the Weibull theory has been proposed for defect severity assessment. Feature vectors extracted from the wavelet transform and spectral postprocessing of the vibration data were used as inputs to the neural network. The designed neural network has shown to be able to effectively differentiate faulty bearings from a comparatively "healthy" bearing, identify defective elements, and classify the defect severity using a corresponding health index value. A classification rate of 99% and 97% were achieved for defects in the inner and outer raceways, respectively. The results encourage further exploration of various neural network structures for automated bearing health diagnosis under varying operating conditions.

Robust bearing performance degradation assessment method based on improved wavelet packet–support vector data description

“…Yan and Lee [19] and Yan et al [20] utilized logistic regression to realize machine performance assessment. Gao et al [21] studied bearing health assessment under varying operating condition using neural network. Ocak et al [22] developed a new robust scheme based on WD and HMM for tracking the severity of bearing faults, and got the result that the probabilities of the normal bearing HMM kept decreasing as the bearing damage progressed toward bearing failure.…”

Wavelet domain principal feature analysis for spindle health diagnosis

“…Once the suitable feature set was chosen from the extracted features, the bearing condition is evaluated by means of a status classifier. Previous research [34][35][36][37][38] explored the use of various types of neural networks as classifiers for machine health diagnosis, such as in rotating machine unbalance and rub faults classification, bearing defect severity assessment, induction motor fault identification, and reciprocating compressor damage detection. In this study, a multiplayer perception (MLP) neural network was chosen as the classifier due to its good performance in bearing defect severity assessment.…”

“…In this study, a multiplayer perception (MLP) neural network was chosen as the classifier due to its good performance in bearing defect severity assessment. 36 The MLP is trained in a supervised manner with error back-propagation, based on the error-correction learning rule. Given that various ratios for the training and testing data were suggested for neural network-based classifier design in the literature, 39,40 no particular ratio was preferred; two thirds of the data sets in each condition were used in this study for training the classifier, and one third for performance checking, from a total of 540 collected data sets.…”

Self-Supervised Health Index Curve Generation for Condition-Based Predictive Maintenance

“…To counteract the human selection bias, more extensive HI estimation approaches have emerged. These either construct the HI based on autonomous evaluation of multiple traditional features [5] or use data-driven supervised learning methods to extract expertindependent features from the raw data [6]. Supervised machine learning methods have been heavily used in predictive maintenance [7][8][9][10].…”