Pengfei Pang

Pengfei Pang

Auto-generated profile, based on publicly available information

Pengfei Pang

Army Engineering University, Nanjing, China

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
24 January 2024 in IEEE Sensors Journal
Verified
Journal Article
Open access
PPPengfei Pang
Pengfei Pang
JTJian Tang
Jian Tang
JLJiqing Luo
Jiqing Luo
MCMiao Chen
Miao Chen
HYHui Yuan
Hui Yuan
LJLei Jiang
Lei Jiang
Previous 1D convolutional neural networks (1D CNN) models for vibration fault diagnosis have high computational complexity and poor interpretability, which cannot meet the higher requirements of model storage, computational efficiency and reliability for airborne and portable devices. Considering these challenges, an Explainable and Lightweight 1D CNN (ELCNN) model based on Square Global Average Pooling (S-GAP) and improved for vibration signals is proposed. The feature extraction and classification layers of 1D CNN are optimized to minimize the model parameters and computational complexity and improve interpretability while ensuring diagnostic accuracy. The model compresses the number of convolutional layers, removes unnecessary bias, activation function and pooling layer, and replaces the fully connected layer (FCL) with S-GAP. Improved 1D CNN models of different methods are evaluated and analyzed on public datasets of rolling bearings. Results show that the ELCNN improved for vibration signals is more lightweight, anti-noise and explainable than other models, and the diagnostic accuracy is further improved.

26 May 2025 in IEEE Sensors Journal
Verified
Journal Article
Closed access
PPPengfei Pang
Pengfei Pang
JTJian Tang
Jian Tang
JZJing Zhang
Jing Zhang
HFHusheng Fang
Husheng Fang
YHYinchuan Hou
Yinchuan Hou
BFBing Fang
Bing Fang
Aiming at the problem of fault diagnosis of rotating machinery under complex working conditions, a convolutional neural network model with improved convolutional layer based on wavelet transform is proposed. Firstly, multi-scale wavelets are employed to initialize the convolutional kernels, endowing the model with strong feature extraction capabilities and enhancing its multi-resolution properties. Subsequently, multiple adaptive Gaussian windows are introduced. These windows allow the convolutional kernels to concentrate on the central frequencies of faults by utilizing learnable gradients, effectively suppressing noise interference. Finally, the square function is used to suppress low-amplitude frequency components across the entire frequency band while enhancing the main frequency characteristics beneficial for classification. Through fault diagnosis experiments on the public datasets, the anti-noise performance of the model stably reaches over 90%, significantly outperforming similar models. In-depth mechanism analysis reveals how each improvement strategy operates and verifies the effectiveness of incorporating wavelets, Gaussian windows, and a square function into the CNN.

18 October 2025 in Data in Brief
Verified
Journal Article
Open access
YHYinchuan Hou
Yinchuan Hou
JTJian Tang
Jian Tang
JZJing Zhang
Jing Zhang
PPPengfei Pang
Pengfei Pang
HZHao Zhang
Hao Zhang
XLXinxin Li
Xinxin Li
Bearings and gears, as key rotating components, are prone to damage when operating under varying conditions with heavy loads, leading to equipment chain failures. In response to the demand for health monitoring based on vibration signals, this paper has constructed a vibration signal dataset for multiple fault types under varying operating conditions: the bearing dataset covers single faults such as inner ring, outer ring, rolling elements, and cage, as well as combined faults of inner and outer rings; the gearbox dataset includes typical defects such as tooth surface wear, broken teeth, and eccentricity; an additional mixed fault dataset of bearings and gearboxes is also provided. These datasets simulate real industrial scenarios through multiple dimensions, aiming to provide a reliable verification benchmark for intelligent diagnostic algorithms and improve the fault identification accuracy and model generalization ability under complex operating conditions.

1 January 2025 in Latin American Journal of Solids and Structures
Verified
Journal Article
Open access
XLXinxin Li
Xinxin Li
JTJian Tang
Jian Tang
JZJing Zhang
Jing Zhang
PPPengfei Pang
Pengfei Pang
JTJian Tang
Jian Tang
To address the issues of poor noise resistance and the lack of mechanistic analysis in the classification and diagnosis of fault vibration signals collected by sensors using existing deep learning models, this paper proposes a Transformer-based fault diagnosis model incorporating squared convolution under strong noise conditions, namely CS-Transformer. This model enhances the local feature representation of fault vibration signals through wide convolution kernels and squaring operations, improves the robustness of global features by leveraging global average pooling, and employs a single-layer Transformer encoder to uncover the correlations among global features, thereby further focusing on key fault features. Fault diagnosis experiments were conducted based on the CWRU and Paderborn bearing datasets. When the signal-to-noise ratio is -6 dB, the noise resistance of the model exceeds 91%, significantly outperforming other comparable models. This validates the superior classification performance and generalization ability of this model for bearing faults of varying degrees under strong noise conditions. Moreover, the analysis of the visualized envelope spectrum further confirms that this model can effectively enhance the target fault frequency and suppress the noise.

23 March 2026 in Advances in Mechanical Engineering
Verified
Journal Article
Open access
PPPengfei Pang
Pengfei Pang
JTJian Tang
Jian Tang
QYQin Yin
Qin Yin
JTJian Tang
Jian Tang
TRTing Rui
Ting Rui
XYXiuquan Yu
Xiuquan Yu
To address multiple faults and fault severity detection under complex operating conditions such as strong noise, strong time variation and cross-working conditions, traditional CNNs, and wavelet-CNN hybrids often suffer from low accuracy and poor generalization, as they either adopt offline wavelet denoising or use fixed wavelet kernels, leading to lost adaptability. WGS-CNN is proposed with three synergistic mechanism-level innovations, forming a closed-loop design rather than a simple component combination. Multi-scale wavelet initialization with retained backpropagation anchors time-frequency prior knowledge, while enabling adaptive learning, laying a targeted foundation for feature extraction. Building on this prior injection, adaptive Gaussian windows with learnable scale factors dynamically constrain convolutional kernels to align with varying fault features, breaking static filtering limitations. Finally, square function activation fused with Gaussian denoising inherits the constrained features to integrate power spectrum enhancement, strengthening weak fault signals. Experiments show WGS-CNN achieves over 87% F 1 across complex scenarios, outperforming traditional CNNs and existing wavelet-CNN hybrids in accuracy and lightweight performance, providing an effective end-to-end solution with fundamental innovations in wavelet-deep learning fusion.

23 March 2026 in Eksploatacja i Niezawodność – Maintenance and Reliability
Verified
Journal Article
Closed access
JTJian Tang
Jian Tang
JTJian Tang
Jian Tang
PPPengfei Pang
Pengfei Pang
QYQin Yin
Qin Yin
JZJing Zhang
Jing Zhang
HZHao Zhang
Hao Zhang
In response to the challenges of fault diagnosis for industrial equipment in high-noise environments, this study proposes an improved graph neural network model that combines envelope activation and cross-correlation function. This model first enhances single-channel feature extraction and anti-noise ability through an improved independent parallel one-dimensional convolution structure and envelope activation operation, achieving single-channel noise suppression and feature extraction; then, it uses the time-domain cross-correlation function to construct a topological graph structure between channels, highlighting the essential correlations of multi-channel signals, and realizes multi-channel feature fusion through graph convolution networks. Experimental results on the public bearing dataset show that proposed model has high classification accuracy, superior anti-noise performance, and good generalization ability in high-noise environments.