Simulation and testing of a noise-limited demodulation system for a fiber-optic hydrophone system based on a Michelson interferometer

Hu, Jiangfei; Li, Duanming

doi:10.1109/coa.2016.7535780

Cited by 4 publications

(1 citation statement)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…FOHs do not need recalibration [37] and present higher bandwidths than conventional piezoelectric hydrophones [38]. Interferometric techniques can be used to improve their sensitivity [39], resolution, bandwidth, dynamic range, and signal-to-noise ratio [40]. A twisted pair made with single-mode fibers has been recently published as a high-sensitivity, broadband ultrasound sensor able to work from 20 kHz to 94.4 MHz [41].…”

Section: Introductionmentioning

confidence: 99%

Fiber-Optic Hydrophone Based on Michelson’s Interferometer with Active Stabilization for Liquid Volume Measurement

Duque

Díaz

Leal‐Junior

et al. 2022

Sensors

View full text Add to dashboard Cite

Sensing technologies using optical fibers have been studied and applied since the 1970s in oil and gas, industrial, medical, aerospace, and civil areas. Detecting ultrasound acoustic waves through fiber-optic hydrophone (FOH) sensors can be one solution for continuous measurement of volumes inside production tanks used by these industries. This work presents an FOH system composed of two optical fiber coils made with commercial single mode fiber (SMF) working in the sensor head of a Michelson’s interferometer (MI) supported by an active stabilization mechanism that drives another optical coil wound around a piezoelectric actuator (PZT) in the reference arm to mitigate external mechanical and thermal noise from the environment. A 1000 mL glass graduated cylinder filled with water is used as a test tank, inside which the sensor head and an ultrasound source are placed. For detection, amplitudes and phases are measured, and machine learning algorithms predict their respective liquid volumes. The acoustic waves create patterns electronically detected with resolution of 1 mL and sensitivity of 340 mrad/mL and 70 mvolts/mL. The nonlinear behavior of both measurands requires classification, distance metrics, and regression algorithms to define an adequate model. The results show the system can determine liquid volumes with an accuracy of 99.4% using a k-nearest neighbors (k-NN) classification with one neighbor and Manhattan’s distance. Moreover, Gaussian process regression using rational quadratic metrics presented a root mean squared error (RMSE) of 0.211 mL.

show abstract