Hao Yu

Hao Yu

Auto-generated profile, based on publicly available information

Hao Yu

School of Engineering, The University of Edinburgh, Edinburgh, U.K.

Research Focus & Metrics

Academic Career Length

23Years (2004-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
18 October 2023 in Computer Methods and Programs in Biomedicine
Verified
Journal Article
Open access
HYHao Yu
Hao Yu
HLHaoyu Liu
Haoyu Liu
ZLZhe Liu
Zhe Liu
ZWZeyu Wang
Zeyu Wang
JJJiabin Jia
Jiabin Jia
Electrical impedance tomography (EIT) has gained considerable attention in the medical field for the diagnosis of lung-related diseases, owing to its non-invasive and real-time characteristics. However, due to the ill-posedness and underdetermined nature of the inverse problem in EIT, suboptimal reconstruction performance and reduced robustness against the measurement noise and modeling errors are common issues. This study aims to mine the deep feature information from measurement voltages, acquired from the EIT sensor, to reconstruct the high-resolution conductivity distribution and enhance the robustness against the measurement noise and modeling errors using the deep learning method. A novel data-driven method named the structure-aware hybrid-fusion learning (SA-HFL) is proposed. SA-HFL is composed of three main components: a segmentation branch, a conductivity reconstruction branch, and a feature fusion module. These branches work in tandem to extract different feature information from the measurement voltage, which is then fused to reconstruct the conductivity distribution. The unique aspect of this network is its ability to utilize different features extracted from various branches to accomplish reconstruction objectives. To supervise the training of the network, we generated regular-shaped and lung-shaped EIT datasets through numerical calculations. The simulations and three experiments demonstrate that the proposed SA-HFL exhibits superior performance in qualitative and quantitative analyses, compared with five cutting-edge deep learning networks and the optical image-guided group sparsity (IGGS) method. The evaluation metrics, relative error (RE), mean structural similarity index (MSSIM), and peak signal-to-noise ratio (PSNR), are improved by implementing the SA-HFL method. For the regular-shaped dataset, the values are 0.119 (RE), 0.9882 (MSSIM), and 31.03 (PSNR). For the lung-shaped dataset, the values are 0.257 (RE), 0.9151 (MSSIM), and 18.67 (PSNR). Furthermore, the proposed network can be executed with appropriate parameters and efficient floating-point operations per second (FLOPs), concerning network complexity and inference speed. The reconstruction results indicate that fusing feature information from different branches enhances the accuracy of conductivity reconstruction in the EIT inverse problem. Moreover, the study shows that fusing different modalities of information to reconstruct the EIT conductivity distribution may be a future development direction.

2 May 2022 in IEEE Transactions on Instrumentation and Measurement
Verified
Journal Article
Closed access
HYHao Yu
Hao Yu
ZZZhixi Zhang
Zhixi Zhang
YGYang Gao
Yang Gao
JJJiabin Jia
Jiabin Jia
Oil-water two-phase flow as a typical two-phase flow type widely exists in various industrial processes and the accurate measurement of oil volume fraction plays a significant role in transporting and separating oil-water mixture in the processes. Electrical Impedance Tomography (EIT) as a merging technology with the advantages of non-invasive, low cost and real-time measurement is widely applied in the industrial field to measure the volume fraction for different types of two-phase flows. However, the measurement process of taking homogeneous reference voltages is time-consuming and costly. To cope with the problem, in the paper, by establishing an end-to-end mapping between measurement voltages and volume fraction, we propose an Attention UNet-Fully Connected (AU-FC) architecture. Relying on the attention mechanism, the reconstructed voltages having a strong correlation or a weak correlation with volume fraction are highlighted or suppressed respectively. Oil-water two-phase flow experiment was conducted in the NEL facility to collect EIT voltage data. Compared with six state-of-the-art and existing machine learning methods, the proposed method performs better in predicting volume fraction. The results indicate that the proposed AU-FC architecture can accurately and real-time predict the volume fraction of oil-water two-phase flow, which improves the application potential of EIT combined with deep learning method in the industrial field.

30 June 2020 in IEEE Transactions on Magnetics
Verified
Journal Article
Closed access
HYHao Yu
Hao Yu
Guodong YuGuodong Yu
Guodong Yu
Yongxiang XuYongxiang Xu
Yongxiang Xu
JZJibin Zou
Jibin Zou
Due to stator magnetic saturation and inherent cogging torque, the slotted limited angle torque motor (LATM) performs large torque ripple within its operation range, resulting in vibration noise and complex control problem. The novelty of this paper is to not only firstly apply 1J22 soft magnetic alloy (SMA) in the LATM, but also propose a multi-objective optimization method to improve torque performances including torque ripple, cogging torque, and output torque, based on genetic algorithm (GA). Firstly, 1J22 soft magnetic alloy is applied to LATM to reduce the torque ripple caused by magnetic field saturation. Then, through the sensitivity analysis, weight coefficients of the related parameters of stator tooth, needed for the optimization process, are obtained. Besides, in order to optimize the torque performances, the multi-objective optimization process is carried out, and the GA is used to obtain the desired objective. Finally, the finite element analysis (FEA) is used to validate the optimization design.

12 August 2025 in IEEE Transactions on Industrial Electronics
Verified
Journal Article
Closed access
HYHao Yu
Hao Yu
shaobin lishaobin li
shaobin li
Yongxiang XuYongxiang Xu
Yongxiang Xu
ZZZihao Zhang
Zihao Zhang
HMHaoyi Mu
Haoyi Mu
WHWei Han
Wei Han
This article presents a novel method for diagnosing interturn short circuits (ITSC) in permanent magnet synchronous motors (PMSMs) utilizing a transformer-based hybrid architecture. The influence of ITSC on dq-axis currents at different fault severity levels is first analyzed mathematically, providing a foundation for extracting distinctive spectral features. Building on this analysis, a motor short-circuit transformer (MSCFormer) is proposed, employing a dual-branch framework to effectively integrate and exploit faulty features from both the time and frequency domains. This dual-domain approach enhances diagnostic accuracy and robustness by utilizing the complementary perspective provided by frequency-domain spectral features, such as the 2nd, 4th and other harmonics. Additionally, the network incorporates an adaptive feature-level fusion strategy, lightweight optimization, and the multihead attention mechanism to mitigate noise interference and improve generalization. Comprehensive experimental evaluations demonstrate the effectiveness and superiority of the proposed method in terms of accuracy and network complexity, achieving 99.79% overall accuracy (OA) and 99.79% F1-score, with only 0.156 million parameters and 1.77 million floating point operations (FLOPs). Code is available at https://github.com/uoe-haoyu/PMSM_Fault_Diagnosis_DL.

28 October 2022 in IEEE Transactions on Instrumentation and Measurement
Verified
Journal Article
Closed access
HYHao Yu
Hao Yu
XWXingchen Wan
Xingchen Wan
ZDZhongxu Dong
Zhongxu Dong
ZZZhixi Zhang
Zhixi Zhang
JJJiabin Jia
Jiabin Jia
Electrical impedance tomography (EIT), as a merging technology, has been widely used in the industrial and clinical fields. However, the causes of the uncertainty of measuring reference voltages, which are affected by medium temperature, measurement errors, and reference conductivity distribution that varies with the patient’s posture, have brought obstacles to applying EIT in industry and medicine. In the paper, two methods: the Multiple Measurements (MM) method and the deep learning method, convolutional neural network (CNN) are proposed to establish the nonlinear mapping between measurement voltages and reference voltages. The novelty of the article is firstly adopting the deep learning method to estimate the reference voltages from measurement voltages for the time-difference EIT. Both static experiments—water tank experiments and dynamic experiments—two-phase flow experiments were carried out. Compared with the two existing estimation methods: best homogeneous (BH) approximation and measurement-scale feature (MSF), and the proposed MM method, the deep learning method shows excellent results in quantitative analysis of the relative errors of reference voltages and ground truth. In addition, the CNN method also displays a better performance in qualitative analysis in terms of the reconstructed tomographic images. The study shows the potential to real-time estimate the reference voltages for time-difference EIT in the industrial and medical fields.

14 February 2025 in IEEE Transactions on Industrial Electronics
Verified
Journal Article
Closed access
NLNing Li
Ning Li
HYHao Yu
Hao Yu
SFStephen Finney
Stephen Finney
PJPaul D. Judge
Paul D. Judge
Conventional long-horizon finite-control-set model predictive control (FCS-MPC) for power electronics has two main real-time implementation problems: 1) calculation burden and latency; and 2) high hardware resource requirements on real-time platforms such as FPGA. To solve these problems, machine learning models have been proposed using model predictive control results as offline training data to train an artificial neural network (ANN), which will be implemented in the real-time controller instead of the original model predictive control model. In this way, the ANN reduces the online calculation burden and the hardware resource requirements. In this article, an FPGA-based 1-D convolution neural network (CNN) for long-horizon FCS-MPC is proposed. To simplify the network model and reduce the hardware resources further, model compression techniques of pruning, fine-tuning, and quantization are implemented. The proposed model is verified using a parallel hybrid converter as a case study converter, with performance and hardware resource requirements compared to conventional FCS-MPC and multilayer perceptron (MLP) models. The results show that the proposed compressed CNN model can reduce FPGA DSP resource requirements by 75% compared to a FCS-MPC model and an MLP models, while also achieving comparable or superior performance.