image: Figure | Schematic of the proposed physics-informed DAS neural network paradigm. a Schematic diagram of the DAS system and its application. SMF single-mode fiber, RF radio frequency, AWG arbitrary waveform generator, AOM acousto-optic modulator, EDFA erbium-doped fiber amplifier, BPD balanced photodetector, ADC analog-to-digital converter, FUT fiber under test. The DAS system simultaneously detects both background and event signals in practical applications. b Traditional design process of DAS signal processing algorithms. ANN artificial neural network. A long-term data accumulation process is required to obtain a sufficient amount of event data, whether labeled or not. c Physics-informed DAS neural network paradigm. PIGN physics-informed generative network. The training of PIGN, noise-removal net, and classification network in the network paradigm does not require any real-world event data. PIGN is used to generate data, the noise-removal net is used to remove background noise, and the classification network is used for event recognition.
Credit: Yangyang Wan et al.
Distributed acoustic sensing offers high-resolution, long-range monitoring for applications including pipeline surveillance, seismic detection, and structural health assessment. Despite its promise, practical deployments face two major challenges: the scarcity of labeled event data and intense environmental noise that obscures target signals. Traditional data-driven networks struggle when real-world event data are limited, and conventional denoising methods often fail in complex environments.
In a new paper published in Light: Science & Applications, a team of scientists, led by Professor Zuyuan He from State Key Laboratory of Photonics and Communications, Department of Electronic Engineering, Shanghai Jiao Tong University, China, and co-workers have proposed a physics-informed neural network paradigm to address these challenges. By incorporating physical models of target events along with system constraints and expert knowledge, the framework generates synthetic DAS event data. A dedicated noise-removal network is trained using this generated data and easily obtainable background measurements, enabling accurate extraction of event signals. Finally, a classification network is trained with the denoised data to identify events in the field without relying on real-world event data.
The innovation lies in combining physical knowledge with AI-driven generative modeling and denoising. The physics-informed generative network can simulate diverse event scenarios while preserving essential temporal and spatial features. The denoising network effectively suppresses complex background noise, enhancing signal clarity. In experiments, the network achieved a 91.8% fault diagnosis accuracy in belt conveyor monitoring and demonstrated reliable event recognition in public DAS datasets. Its generalization capability allows rapid deployment across different sites, with minimal adaptation required.
This approach represents a significant step toward scalable, high-precision acoustic monitoring. By eliminating dependency on extensive real-world event datasets and overcoming noise limitations, the physics-informed neural network provides a robust solution for industrial safety, infrastructure monitoring, and other optical sensing applications.
Journal
Light: Science & Applications
Article Title
Towards a physics-informed network paradigm with data generation and background noise removal for different distributed acoustic sensing applications