11 September 2024 in IEEE Transactions on Instrumentation and Measurement
As one of the most common types of internal faults in power transformer, inter-turn short circuit faults can lead to a rise in temperature and damage the lifetime of the transformer. However, it is challenging to accurately localize fault in time with the existing techniques. Digital Twin (DT) with its capabilities of integrating and synchronizing the virtual and physical worlds, is an essential tool for accurately monitoring the internal state and health of transformers. It not only enables the timely detection and resolution of issues but also improves maintenance efficiency and reduces costs. In this paper, a multi-sensor data fusion method for capturing hot spot based on microthermal field DT is proposed to improve the fault localization accuracy and timeliness. First, a DT model of temperature distribution based on microthermal field is constructed by finite element analysis (FEA). The raw temperature signals in multi-sensors installed in the oil channel are collected by value game. Second, the key information for monitoring the spatiotemporal variation is obtained by conjoint analysis. A data fusion method of capturing hot spot temperature is proposed. Finally, through feature extraction, a diagnosis method based on Bayesian inference is developed for different degrees of faults according to different oil flow line symmetry characteristics. The simulations and practical results all show that the proposed method can not only ensures the accuracy of the localization, but also improves the timeliness of the detection. This work provides a novel solution for localization of inter-turn short circuit fault, and is of constructive idea for solving the fault diagnosis problem by physical fields DT.
2 December 2024 in IEEE Transactions on Instrumentation and Measurement
Given that the total labeled fault signals of rolling bearings are difficult to be collected, it is urgent to address the challenge of limited labeled vibration signals for intelligent fault diagnosis. In this case, the combination of artificial features and machine learning (ML) techniques is a potential way to handle this problem, because ML techniques have simple structures so that the large amounts of labeled signals are not demanded in the training process when compared to deep learning (DL)-based approaches. However, the artificial features usually have bias since they are based on experts’ subjective experience, which may leads to performance degradation of the ML-based approaches. To address this gap, a self-supervised framework of artificial feature bias rectification (SFAFBR) is proposed. Specifically, the considered artificial features of vibration signals are from wavelet packet (WP), which are combined with deep networks by serving as the guides to coax the deep networks learning. The learned features are closely relevant to the original artificial features but the bias is rectified due to the sparsely expression of deep model. Then, on the top of that, a ML-based classifier is constructed to perform the final diagnosis. The proposed framework achieved the state-of-the-art diagnosis accuracy of 97.79% and 94.84% on two bearing datasets with just 5 labeled samples. The result shows that using deep networks to assist in rectifying artificial feature bias is a promising way to address fault diagnosis problem of rolling bearings under limited labeled vibration signals.
24 July 2025 in IEEE Sensors Journal
Inter-turn short circuit (ITSC) faults represent a critical threat to the operational reliability of power transformers, often developing rapidly and eluding early detection. Existing diagnostic methods face challenges in fault localization due to limited observability, low sensitivity to internal disturbances, and high background interference. This paper presents a sensor-integrated Digital Twin (DT) framework for real-time ITSC fault localization, driven by micromagnetic field analysis. A finite element-based DT model is constructed to simulate the spatial distribution of leakage magnetic flux under various ITSC scenarios. Magnetic field signals are acquired from an optimized array of Hall-effect sensors deployed along the transformer winding surface. To bridge the gap between observed magnetic responses and internal fault evolution, a spatiotemporal data fusion and inversion strategy is developed. A regularized and gradient-informed two-dimensional particle swarm optimization algorithm is employed to estimate fault parameters by minimizing magnetic field mismatch under physical constraints, enabling accurate localization across fault positions and severities. Experimental validation demonstrates that the proposed DT-assisted method significantly improves the localization accuracy and responsiveness of sensor-based fault diagnostics. This approach provides a practical pathway for deploying real-time, non-invasive monitoring systems in transformer health management.