Temperature accuracy improvement of long-range RDS using Artificial Neural Network

In this paper, the enhancement of temperature accuracy of Raman Distributed Sensor (RDS) using Artificial Neural Network (ANN) is presented. A 50 km RDS is designed using optical time domain reflectometry (OTDR) technique for temperature measurement. The performance in terms of sensing accuracy of designed sensor is analyzed using numerical simulation technique. The temperature profile of the designed sensor is extracted employing Raman anti-Stoke backscattered power and Fourier deconvolution algorithm. The ANN is employed for denoising of the obtained temperature profile using Fourier deconvolution algorithm. The temperature accuracy of 4.21
⁰
K is achieved using ANN compared to the conventional Fourier deconvolution algorithm which shows 43.19
⁰
K at 50 km sensing distance. Moreover, the spatial resolution of the proposed sensor is calculated as 90 cm using ANN. Therefore, the proposed RDS system can be used for real time temperature measurement with high sensing accuracy and spatial resolution.