Search

• A two-stage network framework containing feature extraction from multiple domains. • A two-stage domain adaptation method based on global and class granularity. • A pseudo-tag filtering method based on high-confidence multi-classifier voting. The issue of restricted target domain tags and constrained information offered by a single source domain in the intelligent fault diagnosis may be successfully resolved by multi-source domain adaptation. Compared with traditional domain adaptation methods, multisource domain adaptation methods face more difficult challenges: the differences between domains are more complex. Hence, a two-stage domain alignment method for multi-source domain fault diagnosis is proposed. The method is accomplished by developing a common feature extractor and several domain feature extractors and classifiers, along with the two-stage distribution adaptation method. Finally, the classifier is used to forecast target samples, while the voting mechanism is utilized to construct the target samples' pseudo labels. With the resulting pseudo labels, the final model training is completed. [ABSTRACT FROM AUTHOR]

• A multi-scale convolutional neural network (MS-CNN) model for transfer learning is proposed. • A domain alignment method based on local central moment discrepancy in class subspace is proposed. • Compared with the central moment discrepancy (CMD), the local center moment discrepancy (LCMD) performs better in transfer diagnosis. Transfer learning is an excellent approach to deal with the problem that the target domain label can not be adequately obtained when rolling bearing cross-condition fault detection. A transfer learning fault diagnosis method of multi-scale CNN rolling bearings based on local central moment discrepancy is presented in this research. The method maps bearing vibration data to a shared space by building a shared multi-scale feature extraction structure and fully connected layers. The source domain label and target domain pseudo-label are used to divide the category subspace in the shared space. And then the local central moment discrepancy is used to match source and target domain in the category subspace to realize fault knowledge transfer under different conditions. The experimental findings reveal that multi-scale CNN migration diagnosis based on local central moment discrepancy has superior accuracy and stability in diverse diagnostic tasks when compared to classic transfer learning approaches. [ABSTRACT FROM AUTHOR]

• A novel diagnosis model(MFFRVM) is proposed for the deterioration state diagnosis of planetary gearboxes. • A new online wear debris index is extracted to evaluate the wear evolution of planetary gears with pitting faults. • Condition monitoring is conducted on a self-made planetary gearbox test rig with manufactured pitting faults. • Combining the MFFRVM model and ADIDC index can better diagnose the pitting fault and reflect the wear evolution of gears. The deterioration state of planetary gearboxes with pitting faults cannot be diagnosed and evaluated effectively with a single type of monitoring signal and traditional features. In this study, a new diagnose method was proposed based on the tribology and vibration signal of planetary gear systems. A new feature extraction approach was exploited to construct a multi-feature fusion relevance vector machine (MFFRVM) model for diagnosing the deterioration state of planetary gearboxes. Meanwhile, a novel index for online wear debris monitoring (ADIDC) was extracted to evaluate the wear evolution of planetary gears. Condition monitoring experiments were conducted on a planetary gearbox. Experimental results show that the diagnostic method combining the MFFRVM model and ADIDC index achieves better accuracy than other methods. This study provides guidance for the wear evolution evaluation of planetary gears with pitting faults and the early fault recognition of planetary gearboxes. [ABSTRACT FROM AUTHOR]

[Display omitted] • Bare-bone formant structure is used to extract syllable level speech features. • Formant skeleton carries cleaner information about emotional aspects, e.g., stress. • A small neural network outperforms the SOTA methods in cross-corpus tests. Speech emotion recognition systems have high computational requirements for deep learning models and low generalizability mainly because of the poor reliability of emotional measurements across multiple corpora. To solve these problems, we present a speech emotion recognition system based on a reductionist approach of decomposing and analyzing syllable-level features. Mel-spectrogram of an audio stream is decomposed into syllable-level components, which are then analyzed to extract statistical features. The proposed method uses formant attention, noise-gate filtering, and rolling normalization contexts to decrease contextual differences and increase focus the attention on the structure of a formant. A set of syllable-level formant features is extracted and fed into a single hidden layer neural network that makes predictions for each syllable as opposed to the conventional approach of using a sophisticated deep learner to make sentence-wide predictions. The syllable level predictions help to lower the aggregated error in utterance level cross-corpus predictions. The experiments on IEMOCAP (IE), MSP-Improv (MI), and RAVDESS (RA) databases show that the method archives better than the state-of-the-art cross-corpus unweighted accuracy of 47.6% for IE to MI and 56.2% for MI to IE. [ABSTRACT FROM AUTHOR]