Dr. Yu Sha | Deep Learning | Best Researcher Award

Yu Sha at Xidian University, China.

Yu Sha is a doctoral researcher specializing in artificial intelligence applications for cavitation detection and intensity recognition. He is pursuing a Doctor of Engineering at Xidian University, China, and was a visiting PhD student at the Frankfurt Institute for Advanced Studies, Germany. His research focuses on AI-driven fault detection in industrial systems, with multiple publications, patents, and academic honors to his name.

Professional Profile:

Scopus

Google Scholar

Education Background

1.  Xidian University, China (2019 – Present)

• • Ph.D. in Computer Science and Technology (College of Artificial Intelligence)
  • Research Focus: Cavitation detection and intensity recognition via deep learning
  • Anticipated Graduation: June 2024

2.  Frankfurt Institute for Advanced Studies, Germany (2020 – 2022)

• • Visiting PhD Researcher (Cavitation and leakage detection using AI)

3.  Lanzhou University of Technology, China (2015 – 2019)

• • B.Sc. in Information and Computing Science
  • Ranked 1st out of 54 students

Professional Development

Yu Sha has contributed to multiple research projects at Xidian University, including AI-driven battlefield situation analysis and decision-making. His work at the Frankfurt Institute for Advanced Studies focused on AI-based cavitation and leakage detection in large-scale pump and pipeline systems. His research expertise extends to deep learning, fault diagnosis in industrial systems, and reinforcement learning.

Research Focus

• AI-driven cavitation detection and intensity recognition
• Fault diagnosis and predictive maintenance in industrial systems
• Deep learning and reinforcement learning applications in engineering

Author Metrics:

• Publications: Articles accepted in high-impact journals like Machine Intelligence Research and Mechanical Systems and Signal Processing.
• Conferences: Research presented at ACM SIGKDD and other international venues.
• Patents: Multiple invention patents related to cavitation detection, face aging estimation, and heart rate estimation

Awards and Honors:

• Outstanding Doctoral Student, Xidian University (2021, 2022)
• Multiple Graduate Student Academic Scholarships (First & Second Level)
• National Encouragement Scholarship (2016, 2017)
• First Prize in multiple mathematical modeling and AI competitions, including MCM/ICM, MathorCup, and Teddy Cup Data Mining Challenge

Publication Top Notes

1. A Multi-Task Learning for Cavitation Detection and Cavitation Intensity Recognition of Valve Acoustic Signals

• Authors: Yu Sha, Johannes Faber, Shuiping Gou, Bo Liu, Wei Li, Stefan Schramm, Horst Stoecker, Thomas Steckenreiter, Domagoj Vnucec, Nadine Wetzstein, Andreas Widl, Kai Zhou
• Published In: Engineering Applications of Artificial Intelligence, Volume 113, August 2022, Article 104904
• DOI: 10.1016/j.engappai.2022.104904
• Publisher: Elsevier Ltd.
• Abstract: The paper proposes a novel multi-task learning framework using 1-D double hierarchical residual networks (1-D DHRN) for simultaneous cavitation detection and cavitation intensity recognition in valve acoustic signals. The approach addresses challenges such as limited sample sizes and poor separability of cavitation states by employing data augmentation techniques and advanced neural network architectures. The framework demonstrated high prediction accuracies across multiple datasets, outperforming other deep learning models and conventional methods.
• Access: The full paper is available at https://www.sciencedirect.com/science/article/pii/S0952197622001361

2. An Acoustic Signal Cavitation Detection Framework Based on XGBoost with Adaptive Selection Feature Engineering

• Authors: Yu Sha, Johannes Faber, Shuiping Gou, Bo Liu, Wei Li, Stefan Schramm, Horst Stoecker, Thomas Steckenreiter, Domagoj Vnucec, Nadine Wetzstein, Andreas Widl, Kai Zhou
• Published In: Measurement, Volume 192, June 2022, Article 110897
• DOI: 10.1016/j.measurement.2022.110897
• Publisher: Elsevier Ltd.
• Abstract: This study introduces a framework combining XGBoost with adaptive selection feature engineering (ASFE) for detecting cavitation in valves using acoustic signals. The methodology includes data augmentation through a non-overlapping sliding window, feature extraction using fast Fourier transform (FFT), and adaptive feature engineering to enhance input features for the XGBoost algorithm. The framework achieved satisfactory prediction performance in both binary and four-class classifications, outperforming traditional XGBoost models.
• Access: The full paper is available at https://www.sciencedirect.com/science/article/pii/S0263224122001798

3. Regional-Local Adversarially Learned One-Class Classifier Anomalous Sound Detection in Global Long-Term Space

• Authors: Yu Sha, Shuiping Gou, Johannes Faber, Bo Liu, Wei Li, Stefan Schramm, Horst Stoecker, Thomas Steckenreiter, Domagoj Vnucec, Nadine Wetzstein, Andreas Widl, Kai Zhou
• Published In: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, August 2022
• DOI: 10.1145/3534678.3539133
• Publisher: Association for Computing Machinery (ACM)
• Abstract: This paper introduces a multi-pattern adversarial learning one-class classification framework for anomalous sound detection (ASD) in mechanical equipment monitoring. The framework utilizes two auto-encoding generators to reconstruct normal acoustic data patterns, extending the discriminator’s role to distinguish between regional and local pattern reconstructions. A global filter layer is also presented to capture long-term interactions in the frequency domain without human priors. The proposed method demonstrated superior performance on four real-world datasets from different industrial domains, outperforming recent state-of-the-art ASD methods.
• Access: The full paper is available at https://dl.acm.org/doi/10.1145/3534678.3539133

4. A Study on Small Magnitude Seismic Phase Identification Using 1D Deep Residual Neural Network

• Authors: Wei Li, Megha Chakraborty, Yu Sha, Kai Zhou, Johannes Faber, Georg Rümpker, Horst Stöcker, Nishtha Srivastava
• Published In: Artificial Intelligence in Geosciences, Volume 3, December 2022, Pages 115-122
• DOI: 10.1016/j.aiig.2022.10.002
• Publisher: KeAi Publishing Communications Ltd.
• Abstract: This study develops a 1D deep Residual Neural Network (ResNet) to address the challenges of seismic signal detection and phase identification, particularly for small magnitude events or signals with low signal-to-noise ratios. The proposed method was trained and tested on datasets from the Southern California Seismic Network, demonstrating high accuracy and robustness in identifying seismic phases, thereby offering a valuable tool for seismic monitoring and analysis.
• Access: The full paper is available at https://www.sciencedirect.com/science/article/pii/S2666544122000284

5. Deep Learning-Based Small Magnitude Earthquake Detection and Seismic Phase Classification

• Authors: Wei Li, Yu Sha, Kai Zhou, Johannes Faber, Georg Ruempker, Horst Stoecker, Nishtha Srivastava
• Published In: arXiv preprint arXiv:2204.02870, April 2022
• DOI: N/A
• Publisher: arXiv
• Abstract: This paper investigates two deep learning-based models, namely 1D

Conclusion