abstract
This article presents the development, characterization, and analysis of optical fiber integrated sensor systems for structural health monitoring (SHM) in oil tanks. The sensors are based on fiber Bragg gratings (FBGs) embedded in nitrile rubber strips through a vulcanization method. Two sensor arrays (with five FBGs each) are fabricated and installed at different regions of the oil tank, namely the metallic side wall and polycarbonate display of the tank. The sensors were characterized as a function of the strain in the nitrile rubber strips, where all analyzed sensors presented a high linearity, since a determination coefficient ( R2 ) higher than 0.999 is obtained. Tests with different oil levels inside the tank indicate the feasibility of the sensors in the shape reconstruction and real-time strain monitoring of the structural element, where even the bending in the sidewall of the tank is estimated. Such results indicated a correlation between the wavelength shift of the FBGs and the liquid level inside the tank, which leads to the possibility of measuring the liquid level from the tank strain distribution in a noncontact approach using a Random Forest algorithm for data processing. The proposed noncontact and machine-learning-enable liquid-level assessment resulted in errors of around 1.07%. Thus, the proposed approach indicates the feasibility of predictive maintenance of the oil tank through real-time monitoring of strain and possible structural flaws of the oil tanks in conjunction with a noncontact approach of oil processing monitoring that can be useful in data fusion as well as multiparameter approaches for oil monitoring. © 2001-2012 IEEE.
authors
Lazaro R.C.; Souza E.; Frizera A.; Marques C.; Leal-Junior A.