Jianbo Zheng

Dr. Jianbo Zheng

orcid

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

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

Representative Publications

View All Publications
3 January 2025 in IEEE Transactions on Transportation Electrification
Verified
Journal Article
Closed access
JZJianbo Zheng
Jianbo Zheng
CYChao Yang
Chao Yang
TZTairui Zhang
Tairui Zhang
BJBin Jiang
Bin Jiang
XFXuhui Fan
Xuhui Fan
Haidong ShaoHaidong Shao
Haidong Shao
Cross-speed bearing fault diagnosis based on multiple source domains and their data enables high-performance condition monitoring for variable-speed equipment, such as engines and turbines. Current multi-source methods typically employ a fixed-length sampling strategy to construct samples and then align the distributions of these samples from different domains. However, these methods neglect the inherent periodic characteristics of bearing data, resulting in incomplete or redundant periodic features in the samples. To address this challenge, we propose a periodic-based framework, PeriodicMFD, for multi-source cross-speed fault diagnosis, which ensures complete periodic information. Our PeriodicMFD framework begins with a periodic sampling strategy designed to construct periodic samples that effectively capture periodic features while maintaining their periodic integrity. Nevertheless, periodic samples from different domains exhibit inconsistencies at both the sample and domain levels. To reconcile these inconsistencies, we introduce sample-level matching to address inconsistencies in feature dimensions and fault patterns among samples from various domains. Additionally, we propose domain-level alignment to handle inconsistencies in space and distribution across different domains. Extensive experiments across three datasets highlight the effectiveness of the PeriodicMFD framework, with a stable average accuracy of 99.55%.

8 June 2026 in Electronics
Verified
Journal Article
Open access
PDF file
JXJiaoyang Xia
Jiaoyang Xia
JZJin Zhang
Jin Zhang
JZJianbo Zheng
Jianbo Zheng
Hypergraph Neural Networks (HGNNs) have demonstrated exceptional capability in modeling high-order correlations; however, their vulnerability to adversarial attacks remains inadequately addressed due to the limited scope of existing security investigations. The prevailing white-box structural attack, HyperAttack, relies exclusively on gradient-derived information and overlooks the semantic affinities between nodes and hyperedges. This oversight limits attack efficacy because gradient signals can be noisy or ambiguous under certain conditions (e.g., saturated regions or local optima), whereas semantic similarities provide complementary cues that help identify hyperedges whose perturbation more reliably alters the target node’s representation. To mitigate this limitation, this paper introduces a semantic enhanced adversarial attack framework for hypergraph neural networks, termed SE-HyperAttack. Specifically, hyperedge features are first aggregated, and semantic similarity scores are computed based on the feature similarity between target nodes and their incident hyperedges to capture latent semantic correlations. These semantic similarity scores are subsequently integrated with integrated gradient scores via a weighted summation scheme, refining the precision of hyperedge selection. Extensive experiments on two datasets demonstrate that the proposed SE-HyperAttack achieves an optimal average attack success rate (ASR) of 79.4%, showing an improvement of 2.6% over HyperAttack. Ablation studies further ascertain that a semantic weight of 30% yields peak performance, beyond which degradation is observed. Notably, the proposed approach preserves computational efficiency commensurate with HyperAttack, incurring negligible additional overhead. These findings substantiate that the integration of semantic information effectively enhances adversarial attack effectiveness on hypergraph neural networks without compromising efficiency.

1 January 2026 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Verified
Journal Article
Open access
XLXiaoge Li
Xiaoge Li
JZJin Zhang
Jin Zhang
PYPengfei Yuan
Pengfei Yuan
LMLiang Meng
Liang Meng
JZJianbo Zheng
Jianbo Zheng

19 January 2024
Verified
Conference Paper
Open access
JZJianbo Zheng
Jianbo Zheng
QLQiao Li
Qiao Li
BJBin Jiang
Bin Jiang
The rapid development of the industrial Internet has boosted the research of using sensor data to realize intelligent fault diagnosis of industrial equipment. Especially, the deep learning methods, such as auto-encoder (AE) and convolutional neural network(CNN), have achieved promising results in the intelligent diagnosis of rolling bearing faults. However, the existing deep learning methods still have the shortcomings of inconspicuous fault features in sensor data, large information loss in the process of feature reconstruction, and poor classification ability of AE-based methods, which limit the further improvement of fault diagnosis algorithms. To overcome the weaknesses, this paper proposes a two-stage fault diagnosis method, namely CNNECAE, to implement efficient feature reconstruction and fault classification. First, a feature reconstruction method based on convolutional auto-encoder (CAE) is proposed to solve the problem that the fault features of vibration data are difficult to capture, and a multi-channel feature coding with prominent fault features and less information loss can be obtained. Then, according to the special input format requirements of multi-channel feature coding and the shortage of poor classification ability of AE-based methods, CNN with strong classification ability is used to classify the faults of multi-channel feature encoding. In this way, better fault diagnosis performance is achieved through the collaboration of the CAE and CNN. Extensive experiments on the CWRU dataset demonstrate the effectiveness of the CNNECAE method. Keywords: intelligent maintenance, convolutional neural network, convolutional auto-encoder