Abstract
This paper proposes a novel fault diagnosis approach based on generative adversarial networks (GAN) for imbalanced industrial time series where normal samples are much larger than failure cases. We combine a well-designed feature extractor with GAN to help train the whole network. Aimed at obtaining data distribution and hidden pattern in both original distinguishing features and latent space, the encoder-decoder-encoder three-sub-network is employed in GAN, based on Deep Convolution Generative Adversarial Networks (DCGAN) but without Tanh activation layer and only trained on normal samples. In order to verify the validity and feasibility of our approach, we test it on rolling bearing data from Case Western Reserve University and further verify it on data collected from our laboratory. The results show that our proposed approach can achieve excellent performance in detecting faulty by outputting much larger evaluation scores.
Highlights
Detection is of utmost importance for the reliability and safety of modern industrial systems [1]
COMPARE WITH OTHER METHODS 1) COMPARE WITH DBN, ANOGAN, AND BIGAN To ensures a fair comparison across different methods, hyper-parameters and architectures of each method are well tuned to obtain its best diagnosis performance for each dataset
2) RESULTS OF THE ROLLING BEARING DATASET FROM OUR LABORATORY For this dataset, we focus on fault signals acquired with lower signal to noise ratio (SNR)
Summary
Detection is of utmost importance for the reliability and safety of modern industrial systems [1]. For industrial anomaly detection [5], time series usually acts as the input data with which to train models. Taking the time series as input, a common anomaly detection framework often consists of two stages: feature extraction and fault recognition [6], [7]. Through feature extraction algorithms [7], time series is preprocessed to low dimensional feature vectors, which are fed into fault detector for fault detection. As a powerful pattern recognition tool for anomaly detection, machine learning algorithms have become the focus of attention [1], including Bayesian classifier [8], support vector machine (SVM) [9], [10], neural networks [10], [11], and deep learning methods [12]. Above methods are all struggling to attain high classification accuracies for imbalanced data because they are based on a class-balanced hypothesis [13]
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.