Special Session-2

Special Session Organizers:

Assoc. Prof. (Dr.) Peng Chen, Shantou University, China

Assoc. Prof. (Dr.) Changbo He, Anhui University, China

Dr. Junyu Qi, Karlsruhe Institute of Technology, Germany

Introduction and Topics:

Large-scale mechanical systems—such as industrial turbines, massive hydraulic networks, mining excavators, and offshore platforms—form the backbone of modern heavy industry. These “mega-systems” consistently operate under extreme loads, fluctuating speeds, and complex, non-stationary conditions. In such harsh environments, traditional contact-based sensing is often severely restricted by extreme temperatures, limited physical accessibility, and structural signal attenuation, making continuous health monitoring a profound challenge.

This special session explores the cutting-edge intersection of Advanced Acoustic Signal Processing and Artificial Intelligence (AI) to overcome these operational barriers. By shifting toward non-contact acoustic monitoring, engineers can safely capture the dynamic behaviors of heavy machinery. However, this introduces new hurdles, primarily intense industrial background noise and complex, overlapping failure modes.

This session aims to gather researchers and industry experts to present novel methodologies that leverage “acoustic intelligence” (e.g., sound field reconstruction, acoustic array processing, and beamforming) coupled with modern AI techniques (e.g., deep learning, transfer learning, and physics-informed models). Our overarching goal is to advance the state-of-the-art in Prognostics and Health Management (PHM), ultimately enhancing the safety, reliability, and lifespan of critical heavy infrastructure.

Potential Topics Include (but are not limited to):

• Detection of cavitation, turbulence, and internal leakage in large-scale hydraulic systems (e.g., axial piston pumps and valves) using fluid-borne and structure-borne noise.
• Cross-domain adaptation and transfer learning methods for machinery operating under variable speeds and widely differing environmental conditions.
• Advanced microphone array processing and sound source localization in high-noise, reverberant industrial environments.
• Physics-informed neural networks (PINNs) and explainable AI applied to acoustic emission data for robust fault diagnostics.
• Few-shot learning and anomaly detection techniques for heavy equipment with limited run-to-failure data.
• Multi-modal sensor fusion strategies integrating acoustic signals with vibration, thermal, or operational data for comprehensive PHM.
• Edge-computing frameworks and low-latency architectures for real-time acoustic monitoring in mega-systems.