22 January 2024 in Engineering Applications of Artificial Intelligence
25 December 2024 in IEEE Transactions on Instrumentation and Measurement
The operational efficiency of a permanent magnet synchronous motor (PMSM) is intrinsically linked to its stability, making the accurate diagnosis of common faults such as inter-turn short circuits (ITSC) crucial. ITSC can cause significant damage to the PMSM and its system. Traditional data-driven fault diagnosis methods often require a large amount of expert-labeled data to ensure accuracy, which is challenging to obtain in actual operations. To address this, this paper introduces an unsupervised learning method for ITSC fault diagnosis that extracts inherent data patterns from unlabeled samples. The approach begins with a unique data processing method and the Fast Fourier Transform to handle the original three-phase current data, thereby reducing data complexity. Following this, an unsupervised framework, utilizing Noise-Aware Density based Clustering with Auto-Tuning (NADCAT), is constructed, in which NADCAT provides pseudo-labels for deep neural networks, enabling precise mapping of features to ITSC faults. A PMSM experiment platform was established to validate the method proposed in this paper. The pseudo-labels obtained by proposed NADCAT achieved a 92.8% effective rate, and the diagnostic model attained an accuracy of 97.63%. These results demonstrate the effectiveness of the method proposed in diagnosing ITSC faults of the PMSM.
7 October 2024 in IEEE Sensors Journal
In the oil drilling environment, permanent magnet synchronous motors (PMSM) are constantly exposed to high temperatures, strong vibrations, and variable velocity. There is little real-world Inter-turn Short Circuit (ITSC) data from the oil well. These problems lead to most ITSC fault characterization types being invalid and no real-world ITSC data are available for diagnosis neural network model training. To address these challenges, we propose a zero real-world fault training data diagnosis (ZRFTD) framework. In this framework, the diagnosis network is trained by physical fault model data. Initially, the internal physical construction of PMSM-ITSC is modeled to generate trainset data. Secondly, the proposed pre-diagnosis mechanism accurately decides when to start the diagnosis neural network, avoiding unnecessary operation when the PMSM is healthy. And self-adjustment peak-to-peak alignment difference (APAD) method is proposed to extract the ITSC current characteristics stably. Thirdly we constructed a simplified local chunking parallel transformer neural network (SLCPTN) as a fault diagnosis network, greatly reducing computational consumption. Finally, an oil-drilling up-ground experimental platform is established to generate the testset data which are inputted into the trained SLCPTN for ITSC fault level diagnosis to assess the efficacy of ZRFTD. The related theoretical derivation and experimental validation prove the feasibility of the ZRFTD framework and achieve an accuracy of 99.2% in classifying real-world PMSM-ITSC data under an oil-drilling environment.
6 June 2023 in Measurement Science and Technology
Bearing remaining useful life (RUL) prediction has always been a central topic in the industry field, the aim of which is to optimize system safety and sustainability. The validity of prediction models and the accuracy of prediction results are affected by mid-term singularities and terminal mutations, under time-domain bearing vibration information. In this paper, a network structure-cascaded dilated convolution vision informer (CDC-Vii) is put forward to precisely forecast the RUL of bearings, which uses the time–frequency fault features as input. CDC-Vii breaks the limitation of the original Informer, which is only sensitive to time-series information. An adaptive fault frequency band selection algorithm is proposed, which can reduce training time while utilizing rich time–frequency information. Based on the Informer architecture, the attention mechanism is improved to form vision subsampling probsparse self-attention (VSPS). VSPS can precisely assign spatial attention weights and reduce computational complexity. At the same time, a truncated relative position encoding technique is proposed to strengthen the position dependence between attention information. Moreover, cascaded dilated convolution enhances the image contrast of faulty frequency bands while enlarging the use of the receptive field. Experiments on two extensively utilized bearing datasets reveal that CDC-Vii surpasses the advanced RUL prediction models.
25 February 2025 in IEEE Transactions on Instrumentation and Measurement
Deep learning technology based on large-scale labeled data is increasingly used to diagnose permanent magnet synchronous motor (PMSM) inter-turn short circuit (ITSC) faults. However, the presence of small fault samples context and the large amounts of unlabeled data are challenging problems in real-world industrial applications. To address these challenges, we propose an optimized Lipschitz generative network-gradient spectral generative adversarial network (GSGAN) within the semi-supervised mechanism. Then, we incorporate the graph attention mechanism to achieve accurate fault severity diagnosis. For the specific, firstly, the GSGAN utilizes gradient and parameter spectral information to strictly enforce the Lipschitz condition of the generative network for generating realisticlike ITSC fault data. Secondly, the semi-supervised mechanism enhanced by a temporal causal sparse fusion graph attention neural network (TCSFGAT) is proposed. The TCSFGAT obtains multi-sample level adjustable fault feature information which can clearly represent the graph features in temporal PMSM current fault information. In addition, TCSFGAT can effectively mitigate the impact of pattern confusion caused by mislabeled pseudo-labels in testset diagnostic process. Finally, based on the real industrial environment characterized by high temperatures and strong vibrations, experiments were conducted under 13 specific load and speed conditions to obtain raw fault data. Experimental results indicate the proposed framework achieves a remarkable 98.7% fault diagnosis accuracy.