Frequency domain reflectometry (FDR) has been widely used for defect detection in cables. However, the reflection coefficient spectrum(RCF) is a non-periodic random signal, which leads to spectral leakage during the FFT process. Therefore, it is necessary for the staff to add window functions manually with experience to reduce the influence of spectral leakage on defect localization. In this paper, a fit method for defect localization of power cable based on Pisarenko harmonic decomposition (PHD) is proposed. In this method, a difference model of linearity usesd to fit the distribution of the RCF. Then, in order to estimate the frequency of each component, an auto-regression-moving average (ARMA) prediction model is built. Finally, the defects localization is completed by estimated the parameters of the ARMA model. In this paper, the intact cable model, the defective cable model, and the noisy environment are simulated. The simulation results show that the proposed method can accurately locate the defects and the total length of the cable, and has strong anti-noise capability.

reflection coefficient spectrum; Pisarenko harmonic decomposition; defect localization; frequency domain reflectometry.