This paper proposes a method for calculating a health indicator (HI) for low-speed axial rolling element bearing (REB) health assessment by utilizing the latent representation obtained by variational inference using Variational Autoencoders (VAEs), trained on each speed reference in the dataset. Further, versatility is added by conditioning on the speed, extending the VAE to a conditional VAE (CVAE), thereby incorporating all speeds in a single model. Within the framework, the coefficients of autoregressive (AR) models are used as features. The dimensionality reduction inherent in the proposed method lowers the need of expert knowledge to design good condition indicators. Moreover, the suggested methodology allows for setting the probability of false alarms when encoding new data points to the latent variable space using the trained model. The effectiveness of the proposed method is validated based on two different datasets: from a workshop test of an offshore drilling machine and from an in-house test rig for axial bearings. In both datasets, the HI is exceeding the warning and alarm levels with a probability of false alarm (PFA) of 10 –6, and the method is most effective at lower shaft speeds.

Detecting bearing faults is very important in preventing non-scheduled shutdowns, catastrophic failures, and production losses. Localized faults on bearings are normally detected based on characteristic frequencies associated with faults in time and frequency spectra. However, missing such characteristic frequency harmonics in a spectrum does not guarantee that a bearing is healthy, or noise might produce harmonics at characteristic frequencies in the healthy case. Further, some defects on roller bearings could not produce characteristic frequencies. To avoid misclassification, bearing defects can be detected via machine learning algorithms, namely convolutional neural network (CNN), support vector machine (SVM), and sparse autoencoder-based SVM (SAE-SVM). Within this framework, three fault classifiers based on CNN, SVM, and SAE-SVM utilizing transfer learning are proposed. Transfer of knowledge is achieved by extracting features from a CNN pretrained on data from the imageNet database to classify faults in roller bearings. The effectiveness of the proposed method is investigated based on vibration and acoustic emission signal datasets from roller bearings with artificial damage. Finally, the accuracy and robustness of the fault classifiers are evaluated at different amounts of noise and training data.

This paper investigates the application of heterodyned Acoustic Emission (AE) compared to more conventional vibration measurements for Condition Monitoring (CM) of an offshore drilling machine, with a particular focus on the large, axial tapered roller bearing supporting the drill string weight in a top drive. The focus on cost reduction and operational uptime in the oil and gas industry motivates research on improved CM methods for fault detection, identification and ultimately prediction. However, bearing failure on this type of machines are currently responsible for a significant share of operational downtime on drilling rigs. In the experiment, a previously used and replaced bearing is compared to a new, healthy bearing with the purpose of identifying possible condition indicators (CI) from the vibration and AE measurements. AE root-mean-square values (RMS) was identified as a CI, being more consistent with the expected bearing health than vibration measurements and also less affected by operating speed. The AE measurements also show complementary forced frequency identification capabilities compared to the vibration measurements. The particular failure mode with bearing roller end damage is described and seen in conjunction with the results.

The industry is moving towards maintenance strategies that consider component health, which require extensive collection and analysis of data. Condition monitoring methods that require manual feature extraction and analysis, become infeasible on an industrial scale. Machine learning algorithms can be used to automatically detect and classify faults, however, obtaining sufficient data for training is required for deep learning and other data-driven classification approaches. Data from healthy machine operation is generally available in abundance, while data from representative fault- and operating conditions is limited. This limits both development and deployment of deep learning-based CM systems on an industrial scale. This paper addresses both the challenges of automated analysis and lack of training data. A deep learning classifier architecture utilizing 1-dimensional dilated convolutions is proposed. Dilation of the convolution kernel allows for analysis of raw vibration signals while simultaneously maintaining the receptive field of the classifier enough to capture temporal patterns. The proposed method performs classification in time domain on signal segments of 1 second or shorter. With knowledge of the bearing specification, artificial vibration signals with similar characteristics as an actual bearing fault can be created. In this work, generated fault signals are combined with healthy operational data to obtaintraining data for a deep classifier. Parameters of the vibration model is chosen as distributions rather than fixed values. By using a range parameters in the vibration model, the classifier learns to recognize temporal features from the training data that generalize to unseen data. The effectiveness of the proposed method is demonstrated by training classifiers on generated data and testing on real signals from faulty bearingsat both low and high speed. One dataset containing seeded faults and three run-to-failure tests are used for the demonstration.