Fiber Bragg grating-based acoustic sensing system enabled by ML-trained, sub-picometer-tunable hybrid III-V/SiN lasers

Abstract

Distributed acoustic emission (AE) sensing is critical for early detection of structural degradation, yet conventional electrical sensors are difficult to scale and fiber-based approaches are limited by interrogation complexity and resolution. Here, we report an intelligent fiber Bragg grating (FBG) sensing system enabled by machine learning (ML)-trained hybrid III-V/SiN tunable lasers that achieve uniform, mode-hop-free, sub-picometer wavelength control. A supervised gradient-descent algorithm is used to learn the nonlinear electro-thermal tuning space of Vernier-based external-cavity lasers, enabling continuous tuning with <0.1 pm resolution and <0.5 dB power variation. This capability allows precise alignment to FBG reflection slopes for high-sensitivity acoustic detection. We demonstrate a four-laser interrogation system monitoring 16 FBG sensors distributed across multiple metallic structures, operating over a 35 nm wavelength span. The system autonomously identifies sensor resonances, dynamically tracks spectral shifts, and reconfigures interrogation wavelengths in response to localized acoustic events. Using pencil-lead break tests as calibrated AE sources, we show simultaneous multi-channel detection and adaptive spatial localization. The combination of narrow linewidth (<10 kHz), wide tunability, and ML-driven calibration enables robust, scalable, and high-resolution sensing. This approach establishes a pathway toward fully autonomous, distributed photonic sensing networks for real-time structural health monitoring.