Empirical Mode Decomposition-Based Detection of Bend-Induced Error and Its Correction in a Raman Optical Fiber Distributed Temperature Sensor

Abstract

The calibration of Raman scattering-based optical fiber distributed temperature sensor (OFDTS) is performed using temperature of the integrated reference (calibration) loop located at the start of the sensing fiber. OFDTS measures distributed temperature profile assuming that sensing fiber is free from any discontinuity, so that anti-Stokes (AS) and Stokes (St) lights have uniform decay. However, in real cases, the fiber loss may get affected by the bend in fiber which causes discontinuity in AS and St signals. If the distributed temperature profile is still calibrated by using the same calibration loop, temperature profile of the fiber zone that exists after the bend will be highly erroneous. Therefore, detection of the bend, temperature error caused by that bend, and compensation of this error are of utmost importance. It is difficult for the user to visually identify the presence and location of the bend from AS and St signals. This paper presents the empirical mode decomposition-based automatic technique to dynamically detect the presence of the bend and its location using area parameter of analytic intrinsic mode functions (IMFs). We demonstrate that the measure of area parameter for the analytic IMFs of St signal can serve as a feature for automatic detection of bend. The utilization of second calibration loop after the detected bend makes it possible to use rest of the fiber for correct temperature profiling. © 2015 IEEE.