Fan Yang

Dr. Fan Yang


Research Focus & Metrics

Academic Career Length

1Year (2026-Present)

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

Representative Publications

View All Publications
5 August 2026 in Sensors
Verified
Journal Article
Open access
PDF file
HYHao Yin
Hao Yin
KZKai Zhang
Kai Zhang
YLYu Lu
Yu Lu
YCYunfeng Chang
Yunfeng Chang
YWYunhui Wu
Yunhui Wu
XPXinliang Pang
Xinliang Pang
Accurate classification of infrasound signals is essential for nuclear-test verification, natural-hazard warning, and geophysical monitoring. Conventional convolutional neural networks (CNN) applied to this task tend to overfit small, class-imbalanced datasets and cannot quantify predictive uncertainty. To address these limitations, we introduce a Bayesian CNN framework that treats network weights as probability distributions and performs inference by variational approximation. LeNet-5, AlexNet, and 4Conv3Fc network serve as baselines and are converted into Bayes LeNet-5, Bayes AlexNet, and Bayes 4Conv3Fc. The short-time Fourier transform (STFT) provides time–frequency spectrograms as model input. On a highly imbalanced dataset comprising nuclear tests, chemical explosions, volcanic eruptions, rocket launches, earthquakes, and lightning, Bayes 4Conv3Fc reaches an overall accuracy of 99.14% without any data augmentation. Relative to the deterministic baselines, precision, recall, and F1-score increase by up to 6.91, 7.12, and 7.20 percentage points, respectively, and Cohen’s Kappa coefficient by up to 8.86 percentage points. Against class-weighted cross-entropy, a standard imbalance-handling baseline, the Bayesian models yield consistently lower Brier scores, indicating that the gains stem from principled uncertainty modelling rather than loss re-weighting alone.. This study quantifies both epistemic and aleatoric uncertainty in an infrasound signal classification model, and the calibration analysis validates that these uncertainty estimates are reliable, providing a basis for evaluating prediction reliability and diagnosing potential failure modes, thereby contributing to improved model interpretability. Coupled with an event-level data partitioning strategy, the evaluation faithfully reflects the model’s generalization to unseen events and offers a promising direction toward uncertainty-aware infrasound monitoring.

20 May 2026 in Acoustics
Verified
Journal Article
Open access
PDF file
FYFan Yang
Fan Yang
HYHao Yin
Hao Yin
YLYu Lu
Yu Lu
XLXuexu Li
Xuexu Li
XPXinliang Pang
Xinliang Pang
The study of bubble pulsation from underwater explosions is critical for applications in marine resource exploration, underwater demolition, and offshore engineering. However, the existing research methods have significant limitations: Laboratory experiments struggle to replicate the dynamic decompression during the process of bubble rising. Field experiments in seas or lakes find it difficult to systematically cover complex parameter ranges. Furthermore, theoretical calculations face the problems of accurately coupling the bubble pulsation with its buoyancy-driven ascent. Therefore, this paper proposes a novel method for calculating the bubble pulsation period of underwater explosions. This method accurately simulates the pulsation and buoyancy-driven ascent of an underwater explosion bubble. Based on the bubble’s energy attenuation characteristics, it establishes the relationship between the pulsation period, TNT equivalent, and ambient hydrostatic pressure. To verify the accuracy of the method, we conducted underwater explosion experiments in the South China Sea with varying TNT equivalents and detonation depths. Abundant bubble pulsation period data of underwater explosions were obtained spatially by deploying hydrophone arrays at various depths. The close agreement between the theoretical predictions and the experimental results confirms the accuracy of the proposed method. By matching the measured values of the first pulsation period and the ratio of the second pulsation period to the first against a database of theoretical curves, a combination of depth and charge equivalent that satisfies both values can be identified, thereby enabling the inversion of the explosion parameters.

24 October 2026 in Applied Sciences
Verified
Journal Article
Open access
PDF file
HYHao Yin
Hao Yin
FYFan Yang
Fan Yang
XPXinliang Pang
Xinliang Pang
Infrasound waves attenuate weakly in the atmosphere and can propagate over regional, teleseismic, and global distances, which makes them valuable for monitoring artificial explosions, volcanic eruptions, earthquakes, and nuclear test-related events. Conventional infrasound localization based on the intersection of array back azimuths are computationally efficient, but it often approximates propagation paths as straight geometric rays and therefore cannot fully represent wind fields, temperature gradients, atmospheric waveguides, or Earth curvature. To address this limitation, this paper presents a two-step localization method that couples geometric preliminary localization with local travel-time residual optimization constrained by spherical-coordinate ray tracing. First, multi-station arrival azimuths are intersected, and k-means clustering with least-squares estimation is used to obtain a robust initial source position. Second, a local grid is constructed around the initial solution; for each candidate source, theoretical travel times to the recording stations are calculated using a ray-tracing model for a moving, inhomogeneous atmosphere in spherical coordinates. The final source location is selected by minimizing multi-station travel-time residuals. The method is tested using three representative events: the 2018 Ingolstadt refinery explosion, the 2019 Stromboli eruption, and the 2022 Hunga eruption. The final localization errors are 13.23 km, 5.93 km, and 6.98 km, respectively, corresponding to accuracy improvements of more than 84% relative to conventional geometric localization. The results show that incorporating atmospheric propagation constraints and local residual optimization can substantially reduce systematic bias while maintaining computational efficiency.