Weisheng Cheng

Mr. Weisheng Cheng

orcid
work life

Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

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

Representative Publications

View All Publications
30 June 2025 in Photonics
Verified
Journal Article
Open access
PDF file
WCWeisheng Cheng
Weisheng Cheng
Zexiao LiZexiao Li
Zexiao Li
XWXuanzong Wu
Xuanzong Wu
SYShuangxiong Yin
Shuangxiong Yin
BZBo Zhang
Bo Zhang
XZXiaodong Zhang
Xiaodong Zhang
Silicon (Si) and silicon carbide (SiC) are second- and third-generation semiconductor materials with excellent properties that are particularly suitable for applications in scenarios such as high temperature, high voltage, and high frequency. Si/SiC wafers face warpage and bending problems during production, which can seriously affect subsequent processing. Fast, accurate, and comprehensive detection of thickness, thickness variation, and flatness (including bow and warpage) of SiC and Si wafers is an industry-recognized challenge. Frequency scanning interferometry (FSI) can synchronize the upper and lower surfaces and thickness information of transparent parallel thin wafers, but it is still affected by multiple interfacial harmonic reflections, reflectivity asymmetry, and phase modulation uncertainty when measuring SiC thin wafers, which leads to thickness calculation errors and face reconstruction deviations. To this end, this paper proposes a high-precision facet reconstruction method for SiC/Si structures, which combines harmonic spectral domain decomposition, refractive index gradient constraints, and partitioning optimization strategy, and introduces interferometric signal “oversampling” and weighted fusion of multiple sets of data to effectively suppress higher-order harmonic interferences, and to enhance the accuracy of phase resolution. The multi-layer iterative optimization model further enhances the measurement accuracy and robustness of the system. The flatness measurement system constructed based on this method can realize the simultaneous acquisition of three-dimensional top and bottom surfaces on 6-inch Si/SiC wafers, and accurately reconstruct the key parameters, such as flatness, warpage, and thickness variation (TTV). A comparison with the Corning Tropel FlatMaster commercial system shows that this method has high consistency and good applicability.

9 June 2026 in Measurement Science and Technology
Verified
Journal Article
Closed access
WCWeisheng Cheng
Weisheng Cheng
Zexiao LiZexiao Li
Zexiao Li
JYJuncheng Yu
Juncheng Yu
FSFuqi Sun
Fuqi Sun
GLGuoyuan Lyu
Guoyuan Lyu
XZXiaodong Zhang
Xiaodong Zhang
Automated optical inspection (AOI) of silicon wafers plays a crucial role in ensuring yield and reliability in semiconductor manufacturing. Tiny surface defects can degrade device performance and even cause failure. As manufacturing processes become increasingly complex and throughput requirements continue to rise, AOI systems face significant challenges, including diverse defect types, complex optical responses, and the difficulty in balancing inspection speed and accuracy. This study addresses three key limitations of traditional AOI methods: limited adaptability to variations in illumination and reflectance caused by varying process conditions; increased algorithmic complexity due to the wide variety of defect types; and insufficient comprehensive multi-surface inspection capabilities. To overcome these issues, we propose an intelligent inspection method inspired by human thought and behavior that enables efficient and comprehensive surface quality assessment at under the tested silicon wafer production-line conditions. The proposed method incorporates three core innovations: a bio-inspired defect detection framework based on visual perception mechanisms, providing a bionic-inspired intelligent inspection framework for complex defect identification; a multi-degree-of-freedom biomimetic inspection system that mimics human eye-hand coordination to improve path adaptability and inspection coverage; and a neural network architecture that combines a self-attention mechanism with multimodal data fusion to showed improved robustness inspection accuracy and robustness. Field validation in a real semiconductor production environment has demonstrated the robust performance, flexibility, and practical reliability of this approach. The system demonstrates can improve silicon wafer surface defect detection under varying illumination and acquisition conditions.

29 January 2026 in Measurement
Verified
Journal Article
Closed access
BZBo Zhang
Bo Zhang
ZLZexiao Li
Zexiao Li
WCWeisheng Cheng
Weisheng Cheng
XZXiaodong Zhang
Xiaodong Zhang

11 October 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
FSFuqi Sun
Fuqi Sun
WCWeisheng Cheng
Weisheng Cheng
CXChenghong Xiang
Chenghong Xiang
YTYuxiao Tang
Yuxiao Tang
NLNana Li
Nana Li
XZXiaodong Zhang
Xiaodong Zhang
Industrial visual anomaly detection is crucial for ensuring system reliability, safety, and production efficiency. However, most existing methods rely on static image modeling and ignore the temporal characteristics of industrial data, making it difficult to capture subtle and dynamically evolving anomalies in complex environments. To address this issue, we propose an online industrial visual anomaly detection method based on temporal aggregation representations. This method first converts consecutive image frames into temporally ordered spatiotemporal representations using a dimension-transformation-based temporal aggregation strategy, thereby suppressing factory environmental noise and enhancing subtle anomalous features. Subsequently, a Multiscale Wavelet-based Salient Attention (MWSA) module is designed to enhance the discriminability of subtle anomalous features through low-frequency structural enhancement and high-frequency noise suppression. Building upon this, a memory bank of normal features is constructed, and anomaly detection and localization are achieved based on nearest-neighbor distances. Experimental results demonstrate that the proposed framework achieves high-precision visual anomaly detection and localization, while maintaining stable performance under varying camera conditions and complex industrial environments. Actual deployment results demonstrate that the method exhibits excellent environmental robustness, achieving an event-level recall rate of 94.83\% and effectively reducing false triggers caused by complex background interference, thereby providing an efficient and robust solution for quality control in industrial automation.