On the interpretation of deep learning models in bearing vibration diagnostics

In recent years, data-driven techniques such as deep learning (DL), have been widely represented in the literature in the field of bearing vibration condition monitoring. While these approaches achieve excellent performance in detecting bearing faults on controlled laboratory datasets, there is little information available on their applicability to more realistic working conditions. One challenge of these data-driven approaches is that they can learn non-classical features unrelated to the physical defect, making their generalizability debatable. To overcome the challenge of generalizability in DL models, we aim to first understand the underlying representation that the network uses to classify different bearing defects. Having an interpretable DL model may give us hints on how to increase its applicability by, e.g., data augmentation, changing input representations or adapting model architectures. To benefit from advances in interpretability in DL methods from computer vision, we first transform the vibration signal into an image. We evaluate a common input transformation, namely the spectrogram. Subsequently, the representations that the network has learnt are evaluated. We use the Grad-CAM algorithm together with signal modifications to evaluate which parts of the input signal contribute to class attribution. Our results show that the network learns signal features related to the transfer path, the physical properties of the test setup, rather than picking up classical features having a physical relation with the defect. Given that a transfer path is very machine specific, this could be an explanation for the lack of scalability of DL methods. To improve the generalizability of DL methods on bearing vibration analysis, the competing dominant machine specific features should be eliminated from the input representation. These results highlight the importance of combining domain expertise with data-driven approaches.

Beamforming Applied to Ultrasound Analysis in Detection of Bearing Defects

The bearings of rotating machinery often fail, leading to unforeseen downtime of large machines in industrial plants. Therefore, condition monitoring can be a powerful tool to aid in the quick identification of these faults and make it possible to plan maintenance before the fault becomes too drastic, reducing downtime and cost. Predictive maintenance is often based on information gathered from accelerometers. However, these sensors are contact-based, making them less attractive for use in an industrial plant and more prone to breakage. In this paper, condition monitoring based on ultrasound is researched, where non-invasive sensors are used to record the noise originating from different defects of the Machinery Fault Simulator. The acoustic data are recorded using a sparse microphone array in a lab environment. The same array was used to record real spatial noise in a fully operational plant which was later added to the acoustic data containing the different defects with a variety of Signal To Noise ratios. In this paper, we compare the classification results of the noisy acoustic data of only one microphone to the beamformed acoustic data. We do this to investigate how beamforming could improve the classification process in an ultrasound condition-monitoring application in a real industrial plant.

Convolutional Neural Networks for Automated Rolling Bearing Diagnostics in Induction Motors Based on Electromagnetic Signals

Bearing faults account for over 40% of induction motor faults, and for this reason, for several decades, much attention has been paid to their condition monitoring, through vibration measurements and, more recently, through electromagnetic signal analysis. Furthermore, in the last few years, research has been focused on evaluating deep learning algorithms for the automatic diagnosis of these faults. Therefore, the purpose of this study is to propose a novel procedure to automatically diagnose different types of bearing faults and load anomalies by means of the stator current and the external stray flux measured on the induction motor in which the bearings are installed. All the data were collected by performing experimental tests in the laboratory. Then, these data were processed to obtain images (scalograms and spectrograms), which were elaborated by a pre-trained Deep Convolutional Neural Network, modified through the transfer learning technique. The results demonstrated the ability of the electromagnetic signals, and in particular of the stray flux, to detect bearing faults and mechanical anomalies, in agreement with the recent literature. Moreover, the Convolutional Neural Network has been proven to be able to automatically discriminate bearing defects and with respect to the healthy condition.

Bearing Defect Detection with Unsupervised Neural Networks

Bearings always suffer from surface defects, such as scratches, black spots, and pits. Those surface defects have great effects on the quality and service life of bearings. Therefore, the defect detection of the bearing has always been the focus of the bearing quality control. Deep learning has been successfully applied to the objection detection due to its excellent performance. However, it is difficult to realize automatic detection of bearing surface defects based on data-driven-based deep learning due to few samples data of bearing defects on the actual production line. Sample preprocessing algorithm based on normalized sample symmetry of bearing is adopted to greatly increase the number of samples. Two different convolutional neural networks, supervised networks and unsupervised networks, are tested separately for the bearing defect detection. The first experiment adopts the supervised networks, and ResNet neural networks are selected as the supervised networks in this experiment. The experiment result shows that the AUC of the model is 0.8567, which is low for the actual use. Also, the positive and negative samples should be labelled manually. To improve the AUC of the model and the flexibility of the samples labelling, a new unsupervised neural network based on autoencoder networks is proposed. Gradients of the unlabeled data are used as labels, and autoencoder networks are created with U-net to predict the output. In the second experiment, positive samples of the supervised experiment are used as the training set. The experiment of the unsupervised neural networks shows that the AUC of the model is 0.9721. In this experiment, the AUC is higher than the first experiment, but the positive samples must be selected. To overcome this shortage, the dataset of the third experiment is the same as the supervised experiment, where all the positive and negative samples are mixed together, which means that there is no need to label the samples. This experiment shows that the AUC of the model is 0.9623. Although the AUC is slightly lower than that of the second experiment, the AUC is high enough for actual use. The experiment results demonstrate the feasibility and superiority of the proposed unsupervised networks.

Two Current-Based Methods for the Detection of Bearing and Impeller Faults in Variable Speed Pumps

The growing number of variable speed drives (VSDs) in industry has an impact on the future development of condition monitoring methods. In research, more and more attention is being paid to condition monitoring based on motor current evaluation. However, there are currently only a few contributions to current-based pump diagnosis. In this paper, two current-based methods for the detection of bearing defects, impeller clogging, and cracked impellers are presented. The first approach, load point-dependent fault indicator analysis (LoPoFIA), is an approach that was derived from motor current signature analysis (MCSA). Compared to MCSA, the novelty of LoPoFIA is that only amplitudes at typical fault frequencies in the current spectrum are considered as a function of the hydraulic load point. The second approach is advanced transient current signature analysis (ATCSA), which represents a time-frequency analysis of a current signal during start-up. According to the literature, ATCSA is mainly used for motor diagnosis. As a test item, a VSD-driven circulation pump was measured in a pump test bench. Compared to MCSA, both LoPoFIA and ATCSA showed improvements in terms of minimizing false alarms. However, LoPoFIA simplifies the separation of bearing defects and impeller defects, as impeller defects especially influence higher flow ranges. Compared to LoPoFIA, ATCSA represents a more efficient method in terms of minimizing measurement effort. In summary, both LoPoFIA and ATCSA provide important insights into the behavior of faulty pumps and can be advantageous compared to MCSA in terms of false alarms and fault separation.

Detection and evaluation of rolling stock wheelset defects using acoustic emission

Acoustic emission (AE) can be employed for the early fault detection of rolling stock wheelset components. Research has been carried out on the development of a remote condition monitoring (RCM) technology for monitoring online rolling stock wheelset defects. Railway axle bearings and wheels are critical components that can develop faults at any time when in service. AE is a reliable passive RCM technique that can be employed for the quantitative evaluation of the structural integrity of rolling stock wheelsets. The emphasis of this study is placed on the results obtained from experimental work performed under laboratory and field testing conditions. Several laboratory tests were carried out using different axle bearing defects. In addition, a customised online RCM system installed on the Chiltern Rail Line, adjacent to a hot box axle detector, was used for comparison purposes. Using this, AE signal analysis was carried out in order to detect potential rolling stock faults. Defect type evaluation and quantification can also be achieved, leading to effective diagnosis of the structural rolling stock integrity of rolling stock wheels and axle bearings.

Method of Limiting Shaft Voltages in AC Electric Machines

The article presents a new method of counteracting shaft voltages and currents in AC electrical machines. It is based on the use of an auxiliary winding located in the stator of the machine. The design of a test stand adapted to the measurement of shaft voltages of the machine, based on the prototype of a synchronous machine with permanent magnets, has been presented. The model was used to conduct a number of laboratory tests aimed at confirming the functionality of the auxiliary winding in various operating states of the machine (including no-load and load condition during generator operation). The article focuses on demonstrating the beneficial effect of the auxiliary winding on the level of induced shaft voltages in an electric machine. In order to confirm the close dependence of the circular flux in the stator yoke on the shaft voltage, shaft voltage measurement results for various cases of external power supply of auxiliary winding forcing a circular flux are presented. Regardless of the laboratory tests, a simulation model of a synchronous machine with permanent magnets, on which calculations were carried out to analyze the work of the auxiliary winding located in the stator yoke, was developed. The article is supplemented by a review of damage to electrical machines with a detailed description of bearing defects, as well as a brief de-scription of issues related to the mechanism of generating shaft voltages and currents in electrical machines and methods of counteracting them.