High Spatial Resolution OFDR System Based on Independent Component Analysis Algorithm for Long-Range Distributed Strain Measurement

In this study, a high spatial resolution optical frequency domain reflectometry (OFDR) based on independent component analysis (ICA) algorithm is proposed and experimentally demonstrated. In the proposed sensing system, two-dimensional (2D) images with wavelength shift information induced by applied strains as a function of fiber position are constructed by utilizing the data arrays obtained after cross-correlation processing of the reference signal and measurement signals of each sensing fiber segment. The ICA algorithm, as an effective 2D image denoising technique, is applied to the constructed 2D images to remove random noise and improve the sensing accuracy of the system so as to realize long-range distributed strain measurement with high spatial resolution. Compared with traditional 2D image denoising methods including the Gaussian filtering (GF) method and the wavelet denoising (WD) method, ICA method makes full use of the independence of image source information and noise information, which is able to effectively suppress the intensification of noise without compromising the source information. With no modification on the OFDR hardware system, strain gradient information is successfully extracted over an effective sensing distance of 75 m with a spatial resolution up to 2 mm by using the ICA method. The calculated mean strain measurement error for the ICA method is 5.45 μϵ, which is significantly improved compared to the error of 72.73 μϵ when no image denoising method is applied and reduced by approximately half compared to the errors when traditional GF method and WD method are used. The mean standard deviations of the measured strain gradient along the sensing fiber length for the proposed method, the GF method and the WD method are reduced by 93.42%, 76.99%, and 84.56%, respectively, compared to the raw data without any denoising processing, showing excellent smoothness of the recovered strain profiles. The signal-to-noise ratio (SNR) for the proposed method is improved by 23.17 dB, which is better than 14.8 dB of GF and 19.1 dB of WD. The experimental results show that the proposed method provides a new solution for OFDR system on achieving long-distance distributed strain sensing with high spatial resolution.