Jiaxian Zhu

Dr. Jiaxian Zhu

orcid

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

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

Representative Publications

View All Publications
19 November 2025 in IEEE Internet of Things Journal
Verified
Journal Article
Closed access
JLJunjie Liu
Junjie Liu
JZJiaxian Zhu
Jiaxian Zhu
WBWeihua Bai
Weihua Bai
HZHuibing ZHANG
Huibing ZHANG
LWLianghai Wu
Lianghai Wu
Keqin LiKeqin Li
Keqin Li
Industrial IoT (IIoT) sensing nodes for bearing monitoring often operate in high-noise environments, where acoustic and mechanical interference masks weak fault signatures and undermines diagnostic reliability. To address this challenge, we propose a lightweight multi-modal fusion framework for robust fault diagnosis under extreme noise. Our method first applies multi-resolution decomposition with selective reconstruction and adaptive enhancement to preserve fault-related components while suppressing interference. The enhanced signals are transformed from 1D time series into two-dimensional representations to jointly capture temporal dynamics and spectral characteristics. We then design a tri-branch multi-head attention architecture that integrates a multiscale recurrence plot network, a Gramian angular field network, and a lightweight residual network. Learnable attention weights enable adaptive fusion of complementary cross-modal features with low computational overhead. Extensive experiments on the CWRU benchmark show superior robustness from 0 dB to -6 dB SNR, with a mean accuracy above 99.6% and consistent gains over eight state-of-the-art methods. Additional tasks on single-domain diagnosis, cross-condition fault type recognition, and fault degree discrimination (T1–T3) confirm strong generalization and multi-scale adaptability, with average improvements of 2–4% over the second-best baseline and stable variance under noise. The compact architecture and high noise immunity indicate practical suitability for IIoT sensing nodes and edge deployment in complex industrial scenarios.

13 November 2025 in IEEE Transactions on Neural Networks and Learning Systems
Verified
Journal Article
Closed access
JLJunjie Liu
Junjie Liu
YWYu Wang
Yu Wang
JZJiaxian Zhu
Jiaxian Zhu
WBWeihua Bai
Weihua Bai
HZHuibing ZHANG
Huibing ZHANG
Keqin LiKeqin Li
Keqin Li
Traffic flow prediction is fundamental to traffic information services, control, and guidance. The challenge to accurately model the traffic flow is to comprehensively capture dynamic, global spatial, and local temporal correlations. To address challenges in spatiotemporal dependencies, global similarity, and local dynamics, we propose a multilayer spatiotemporal correlation-aware graph attention network (MSTC-GAT) for traffic flow prediction. Our model contains a multilayer spatial structure-aware module [spatial graph attention network (S-GAT)] using a spatial GAT with hierarchical attention masks and a path-based node correlation matrix to effectively capture local and global spatial dependencies. The temporal structure-aware module [temporal graph attention networks (T-GATs)] constructs a short-term similarity matrix of nodes for the temporal GAT to capture local dynamic temporal dependencies. Finally, a spatiotemporal Transformer (ST-Transformer) fuses weighted spatiotemporal node embeddings to capture global dynamic dependencies for accurate prediction. We conduct extensive experiments on four public benchmark datasets compared with 10 state-of-the-art models. The experimental results demonstrate that the MSTC-GAT outperforms all comparisons for short- and long-term predictions.

23 September 2022 in Symmetry
Verified
Journal Article
Open access
PDF file
WBWeihua Bai
Weihua Bai
JZJiaxian Zhu
Jiaxian Zhu
JZJialing Zhao
Jialing Zhao
WCWenwei Cai
Wenwei Cai
Keqin LiKeqin Li
Keqin Li
The intelligent electrical power system is a comprehensive symmetrical system that controls the power supply and power consumption. As a basis for intelligent power supply control, load demand forecasting in power system operation management has attracted considerable research attention in energy management. In this study, we proposed a novel unsupervised multi-dimensional feature learning forecasting model, named MultiDBN-T, based on a deep belief network and transformer encoder to accurately forecast short-term power load demand and implement power generation planning and scheduling. In the model, the first layer (pre-DBN), based on a deep belief network, was designed to perform unsupervised multi-feature extraction feature learning on the data, and strongly coupled features between multiple independent observable variables were obtained. Next, the encoder layer (D-TEncoder), based on multi-head self-attention, was used to learn the coupled features between various locations, times, or time periods in historical data. The feature embedding of the original multivariate data was performed after the hidden variable relationship was determined. Finally, short-term power load forecasting was conducted. Experimental comparison and analysis of various sequence learning algorithms revealed that the forecasting results of MultiDBN-T were the best, and its mean absolute percentage error and root mean square error were improved by more than 40% on average compared with other algorithms. The effectiveness and accuracy of the model were experimentally verified.

18 April 2025 in Symmetry
Verified
Journal Article
Open access
PDF file
JZJiaxian Zhu
Jiaxian Zhu
JCJiahong Chen
Jiahong Chen
HHHuiyang He
Huiyang He
WBWeihua Bai
Weihua Bai
Teng ZhouTeng Zhou
Teng Zhou
The citrus industry plays a pivotal role in modern agriculture. With the expansion of citrus plantations, the intelligent detection and prevention of diseases and pests have become essential for advancing smart agriculture. Traditional citrus leaf disease identification methods primarily rely on manual observation, which is often time-consuming, labor-intensive, and prone to inaccuracies due to inherent asymmetries in disease manifestations. This work introduces CBACA-YOLOv5, an enhanced YOLOv5s-based detection algorithm designed to effectively capture the symmetric and asymmetric features of common citrus leaf diseases. Specifically, the model integrates the convolutional block attention module (CBAM), which symmetrically enhances feature extraction across spatial and channel dimensions, significantly improving the detection of small and occluded targets. Additionally, we incorporate coordinate attention (CA) mechanisms into the YOLOv5s C3 module, explicitly addressing asymmetrical spatial distributions of disease features. The CARAFE upsampling module further optimizes feature fusion symmetry, enhancing the extraction efficiency and accelerating the network convergence. Experimental findings demonstrate that CBACA-YOLOv5 achieves an accuracy of 96.1% and a mean average precision (mAP) of 92.1%, and improvements of 0.6% and 2.3%, respectively, over the baseline model. The proposed CBACA-YOLOv5 model exhibits considerable robustness and reliability in detecting citrus leaf diseases under diverse and asymmetrical field conditions, thus holding substantial promise for practical integration into intelligent agricultural systems.

26 April 2026 in Mathematics
Verified
Journal Article
Open access
PDF file
JZJiaxian Zhu
Jiaxian Zhu
CZChuanbin Zhang
Chuanbin Zhang
ZSZhaoyin Shi
Zhaoyin Shi
HCHang Chen
Hang Chen
Zhizhe LinZhizhe Lin
Zhizhe Lin
Teng ZhouTeng Zhou
Teng Zhou
The mathematical characterization of non-stationary signals remains a significant challenge, particularly when impulsive components are obscured by high-dimensional noise and structural coupling. This paper proposes an application-driven mathematical methodology for a learnable discrete wavelet transform (LDWT) that combines classical multi-resolution analysis with task-optimized data-driven adaptivity. Rather than introducing entirely new foundational theory, our approach strategically relaxes constraints from orthogonal wavelet theory within the non-perfect reconstruction filter bank framework, enabling controlled spectral decomposition optimized for supervised fault diagnosis. We introduce a specialized regularization term based on the half-band property to ensure spectral complementarity and minimize cross-band correlation, while a Jacobian-based stabilization approach is formulated to ensure the convergence of filter coefficients during optimization. The proposed algorithmic architecture, LDBRFnet, features a dual-branch encoder system designed to capture the mathematical synergy between sub-band-level global statistics and time-domain local morphology. This dual-view representation effectively mitigates feature leakage and overconfidence in classification. Theoretical analysis and numerical experiments demonstrate that the learned filters satisfy the frequency-shift property and maintain robust spectral partitioning even under low signal-to-noise ratios. Validation on complex vibration datasets confirms that the framework achieves superior diagnostic accuracy (over 95.5%) and computational efficiency, reducing model parameters by 96.7% compared to state-of-the-art baselines. This work provides a generalizable mathematical approach for adaptive signal decomposition and robust pattern recognition in interdisciplinary applications.

5 November 2022 in Mathematics
Verified
Journal Article
Open access
PDF file
JZJunxi Zheng
Junxi Zheng
JZJuan Zeng
Juan Zeng
XWXinyang Wang
Xinyang Wang
GLGang Li
Gang Li
JZJiaxian Zhu
Jiaxian Zhu
DTDeyu Tang
Deyu Tang
Recently, research on detecting SNP interactions has attracted considerable attention, which is of great significance for exploring complex diseases. The formulation of effective swarm intelligence optimization algorithms is a primary resolution to this issue. To achieve this goal, an important problem needs to be solved in advance; that is, designing and selecting lightweight scoring criteria that can be calculated in O(m) time and can accurately estimate the degree of association between SNP combinations and disease status. In this study, we propose a high-accuracy scoring criterion (HSICCR) by measuring the degree of causality dedicated to assessing the degree. First, we approximate two kinds of dependencies according to the structural equation of the causal relationship between epistasis SNP combination and disease status. Then, inspired by these dependencies, we put forward this scoring criterion that integrates a widely used method of measuring statistical dependencies based on kernel functions (HSIC). However, the computing time complexity of HSIC is O(m2), which is too costly to be an integral part of the scoring criterion. Since the sizes of the sample space of the disease status, SNP loci and SNP combination are small enough, we propose an efficient method of computing HSIC for variables with a small sample in O(m) time. Eventually, HSICCR can be computed in O(m) time in practice. Finally, we compared HSICCR with five representative high-accuracy scoring criteria that detect SNP interactions for 49 simulation disease models. The experimental results show that the accuracy of our proposed scoring criterion is, overall, state-of-the-art.