15 August 2023 in IEEE Transactions on Instrumentation and Measurement
Deep supervised learning-based fault diagnosis methods require a large amount of labeled data, which frequently contradicts the typical engineering scenario, in which numerous samples are available but only a small portion are labeled. To conduct a fault diagnosis in this case, unsupervised and semi-supervised representation learning have recently been proposed. However, most of these methods have been designed without much consideration for downstream classification tasks and thus are insufficiently relevant for providing sufficient targeted assistance. Therefore, this study proposes a semi-self-supervised method that consists of two learning stages. In the self-supervised learning stage, two pretext tasks are implemented. Using two auxiliary feature extractors and classifiers, the global and local features of the unlabeled samples can be learned by the feature extractor. In the supervised learning stage, the feature extractor parameters are frozen, and the classifier is trained with the limited number of labeled samples available. The performance of the proposed method was verified using the CWRU open dataset and a self-built experimental dataset. Experiments on these two datasets demonstrated that the proposed method outperformed existing diagnosis methods for a few labeled samples.
14 September 2022 in IEEE Transactions on Instrumentation and Measurement
In recent years, methods based on meta-learning have been widely used in cross-domain fault diagnosis, and promising results can be obtained even with limited target training data. However, data scarcity problems can exist not only in the target domain but also in the source domain, which puts a damper on the meta-knowledge learning process since the source domain cannot provide sufficient source tasks. In this study, a novel adversarial one-shot cross-domain network named AOCN for fault diagnosis is proposed, which requires only a few source samples and as low as one labeled target sample per class. The main idea of AOCN is to learn domain invariant embedding and generate domain invariant prototypes without causing overfitting problems. AOCN consists of two modules: a feature generator and a sample-pair discriminator with four outputs. The optimization process is divided into three steps. The first step is the meta-learning of the feature generator. The second step is the pretraining of the sample-pair discriminator to distinguish four groups of sample pairs that are generated by the pairing strategy. The third step is the adversarial learning of the two modules to confuse the features between homogeneous pairs and the features between heterogeneous pairs, respectively. Experiment results on two datasets show that AOCN can achieve more satisfactory performance than the existing methods compared.
1 June 2021 in Journal of Physics: Conference Series
In order to study the influence of multi-point leakage on the flow field in the bellows and the quantitative evaluation method of air leakage based on temperature, the Fluent software was used to consider the fluid-solid coupling heat transfer in the bellows and the convection and radiation heat dissipation on the outer wall of the bellows. The multi-point leakage environment was set up, and the numerical simulation model of single-side sintered bellows-branch pipe section was established. The temperature distribution and flow field distribution of the fluid on the outer wall and inside the bellows are solved. Infrared thermal imaging was used to extract the temperature trend of the local leakage point of 5# bellows, and compared with the simulation results, it was found that the closer to the leakage point, the faster the temperature dropped, and the two had a strong correlation, which proved that it was feasible to use numerical simulation to study the air leakage of bellows. Through the analysis of different leakage forms, it is found that the influence range and intensity of single-point leakage and multi-point leakage on the field quantity are different. The research method lays a solid foundation for the in-depth study of temperature-based detection algorithm.
1 January 2026 in IEEE Access