Converting strain to wavelength shift in FBG sensor development and determining its sensitivity is generally considered straightforward, especially for static measurements. However, the main challenge lies in accurately converting the strain-induced wavelength shift into acceleration for FBG accelerometers, particularly in dynamic measurement. This paper uses a cantilever-type FBG accelerometer with a tip mass to demonstrate the implementation of a neural network as an identifier to translate wavelength shift to acceleration. The neural network acts as an empirical data-based sensitivity identifier, allowing the conversion of the FBG's wavelength shift measurements to acceleration values even when different tip masses are employed than those used during training. The base acceleration is the input to the neural network with the range of frequency used 5–40 Hz, which is more than one-half of the resonant frequency of the cantilever beam, which is 48.9 Hz. The successful use of neural networks as sensitivity identifiers has paved the way for addressing the frequency-dependent behaviour of FBG sensors, particularly in dynamic measurements.
Empirical Data-Based Sensitivity Identification of a Cantilever-Type FBG Accelerometer
Lect.Notes Mechanical Engineering
International Conference on Mechanical, Automotive and Materials Engineering ; 2023 ; Da Nang, Vietnam December 20, 2023 - December 22, 2023
Proceedings of the 10th International Conference on Mechanical, Automotive and Materials Engineering ; Kapitel : 20 ; 231-241
25.09.2024
11 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Dual-Cantilever-Beam Accelerometer
NTRS | 1988
|Extrinsic Fabry–Perot Cantilever Accelerometer Based on Micromachined 45° Angled Fiber
British Library Online Contents | 2018
|Compact high-sensitivity accelerometer-seismometer
Online Contents | 2010
|Compact high-sensitivity accelerometer-seismometer
Springer Verlag | 2010
|A bi-directional accelerometer and flow sensor made using a piezoresistive cantilever
British Library Conference Proceedings | 1997
|