Abstract
As failures in rotating machines can have serious implications, the timely detection and diagnosis of faults in these machines is imperative for their smooth and safe operation. Although deep learning offers the advantage of autonomously learning the fault characteristics from the data, the data scarcity from different health states often limits its applicability to only binary classification (healthy or faulty). This work proposes synthetic data augmentation through virtual sensors for the deep learning-based fault diagnosis of a rotating machine with 42 different classes. The original and augmented data were processed in a transfer learning framework and through a deep learning model from scratch. The two-dimensional visualization of the feature space from the original and augmented data showed that the latter’s data clusters are more distinct than the former’s. The proposed data augmentation showed a 6–15% improvement in training accuracy, a 44–49% improvement in validation accuracy, an 86–98% decline in training loss, and a 91–98% decline in validation loss. The improved generalization through data augmentation was verified by a 39–58% improvement in the test accuracy.
Highlights
Results andTransfer a customized convolutional neural network using the scalograms of the original data fromThe thedataset rotor system of Figure of Table is comprised of 1951 scenarios of 42 different classes, and each class is described in terms of eight signals obtained through the sensors
This section discusses the results of the transfer learning model of ResNet18 and cusThis section discusses the results of the transfer learning model of ResNet18 and tomized convolutional neural network (CNN) on the two scenarios of augmented data; a dataset augmented through customized CNN on the two scenarios of augmented data; a dataset augmented through eight virtual sensors, and a dataset augmented through 16 virtual sensors
This paper proposes a synthetic data augmentation scheme for the deep learning-based damage diagnosis of rotating machinery
Summary
Rotating machines are the backbone of a variety of modern applications, such as power turbines, pumps, automobiles, and oil/gas refineries [1,2,3,4,5]. These machines are vulnerable to unavoidable malfunction during their operation due to load fluctuations, and material degradation with time. Kolar et al [10] proposed a fault diagnosis strategy for rotary machinery using a convolutional neural network (CNN); the vibration signals of a three-axis accelerometer were supplied as high-definition images to the CNN, which automatically extracted the discriminative features, and classified the input data into four classes: normal, unbalance, cracked rotor, and bearing fault.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
