CNN-Assisted Demodulation Approach for Fiber Bragg Grating Sensor Based on Specklegrams

This article presents a new demodulation method from the reflection specklegrams of fiber Bragg grating (FBG)-based sensors by employing convolutional neural networks (CNNs). To verify the feasibility of the fast and low-cost demodulation, temperature measurement was demonstrated using an FBG-based sensor working at the wavelength of ~850 nm, whose reflection specklegram was captured by a Raspberry PI camera to train and test the CNN model. With this method, characteristic information related to temperature change can be directly extracted from the specklegrams instead of spectral analysis. Since it makes full advantage of the characteristic information in specklegrams rather than peak tracking in traditional methods, it could improve the measurement accuracy of FBG, which can reach 99.95% with a mean square error (mse) of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$0.351~^{\circ }$ </tex-math></inline-formula>C2. The temperature measurement was repeated by an InGaAs CMOS camera and another FBG at 1550 nm under the same conditions, achieving similar demodulation accuracy, which confirms the effectiveness of the CNN-assisted demodulation at lower hardware cost. Furthermore, with the diffraction optimization by surface grating lines, the mse could reach <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$0.06~^{\circ }$ </tex-math></inline-formula>C2 and the RMSE is <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$0.24~^{\circ }$ </tex-math></inline-formula>C. The proposed method shows a cost-effective solution for the FBG-based sensing system and great potential for demodulation at 850 nm.