18 April 2024 in Applied Intelligence
30 April 2024 in IEEE Transactions on Instrumentation and Measurement
Maintaining the operational stability of an aero-engine heavily relies on the health of the rotor bearings. Data-driven fault diagnosis approaches have achieved significant success due to their superior ability to identify faults. However, the application of conventional deep learning methods, which heavily rely on a huge amount of vibration signals, is unrealistic in most aero-engine-related cases due to the limited number of fault samples. To tackle this problem, a multitask Siamese network (MTSN) is proposed, which combines data augmentation (DA) and metric learning into a unified framework. First, in the multiclassification task, a limited number of samples are augmented using Mixup and learned by a convolutional neural network (CNN). Then, in the metric learning task, the similarity between sample pairs is learned by the shared encoder in the Siamese network. To ensure effective learning in both tasks, the loss functions of the two learning tasks are balanced by introducing uncertainty. Based on two public bearing fault datasets, we conducted a series of experiments. The experimental results show that the learning tasks mutually reinforce each other, and the MTSN exhibits good fault diagnosis performance with limited data.
25 September 2024 in IEEE Transactions on Industrial Informatics
In data-driven mechanical fault diagnosis, domain adaptation is used to solve the problem of different distribution of collected data under varying operating conditions. However, the assumption of consistent label space between source and target domains is difficult to satisfy due to the types of faults that occur under different operating conditions not being completely identical. To address the problem of label inconsistency in partial domain adaptation, we propose a feature separation weight rectified network (FS-WRN). Shared domain features are extracted by feature separation, and the source and target domain distributions are aligned using the rectified local maximum mean discrepancy. The source-domain features are weight-rectified by capturing changes in the target domain distribution during training to minimize the effect of negative transfer due to private classes. Three public datasets are used, and the results show that FS-WRN has promising fault diagnosis performance under complex partial domain adaptation tasks.
1 January 2025 in IEEE Transactions on Instrumentation and Measurement
1 January 2026 in IEEE Transactions on Instrumentation and Measurement