Ensemble clustering-based fault diagnosis method incorporating traditional and deep representation features

Abstract

The traditional and deep representation features have been successfully employed in the clustering-based fault diagnosis, however, current studies have ignored their heterogeneity and complementarity. Besides, these studies have rarely explored the superiority of ensemble clustering, leading to unstable diagnostic results. To address these issues, a novel ensemble clustering-based fault diagnosis method, namely ECFD, is proposed based on the collaboration of traditional and deep representation features. To the best of our knowledge, this is the first attempt to develop an ensemble clustering method with these types of features for fault diagnosis of rotating machinery. Initially, features are extracted by different approaches, among which the traditional features are obtained by time-domain, frequency-domain, time-frequency-domain approaches, and the deep representation features are extracted by the gated recurrent units-based autoencoders. Then, the heterogeneous and complementary information among these features is explored by the adaptive group regularization method, which is also developed to generate high-quality feature subspaces for constructing the base learners of the ensemble clustering. Finally, the clustering results of each base learner are integrated by a weighted voting method. The effectiveness of the proposed ECFD was verified by the experiments on the real-world bearing vibration dataset. The results indicate that the proposed ECFD can explore the heterogeneity and complementarity among different kinds of features and take advantage of the ensemble clustering, which is superior to other existing benchmark clustering methods in unsupervised fault diagnosis.

Contrastive learning via swapped cluster assignments for bearing fault diagnosis

“…Part of our work relies on prototype-based clustering methods [7,[9][10][11][12][13][14]. Caron et al showed that K-means could be used as pseudo-labels to learn the representations [13,15].…”

Hierarchical graph neural network with adaptive cross-graph fusion for remaining useful life prediction

“…Besides, the generalization ability of model-based methods is weak because of the different degradation mechanisms among machine systems. Owing to the rapid development of sensor techniques, data-driven methods based on massive monitoring data have become a new possibility when facing complex machine systems [9,10]. Since the data-driven methods are easy to apply without the need for prior expertise, some data-driven methods, such as support vector regression (SVR) and hidden Markov models have been proposed and widely applied in RUL prediction [11,12].…”

A novel complex network community clustering method for fault diagnosis

“…And these studies can be broadly summarized as model-driven methods, datadriven methods and hybrid-driven methods [3][4][5]. Further, based on the degree of dependence on data labels, fault diagnosis methods can also be broadly divided into supervised learning [6,7], semi-supervised learning [8,9] and unsupervised learning [10,11]. Despite the great progress in datadriven fault diagnosis, sufficient and high-quality labeled data are still scarce in practical industrial scenarios.…”