Open AccessIdentification and location of PIM faults in radio‐frequency circuits with multiple coaxial connectors using a neural network approach
Author(s) -
Jin Qiuyan,
Gao Jinchun,
Flowers George Timothy,
Wu Yongle,
Bi Lingyu,
Ji Rui
Publication year - 2019
Publication title -
iet microwaves, antennas and propagation
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.555
H-Index - 69
eISSN - 1751-8733
pISSN - 1751-8725
DOI - 10.1049/iet-map.2018.5322
Subject(s) - electronic engineering , fault (geology) , intermodulation , coaxial , artificial neural network , harmonics , transmission line , electronic circuit , scattering parameters , feature (linguistics) , engineering , computer science , equivalent circuit , electrical engineering , artificial intelligence , voltage , amplifier , linguistics , philosophy , cmos , seismology , geology
Coaxial connectors are widely used in radio‐frequency circuits and are a major non‐linear source of passive intermodulation (PIM), especially when they are subjected to environmental degradation. Accordingly, an effective and accurate method to locate the position of the PIM source in a communication system is of substantial value. By analysing the characteristics of different connectors, equivalent circuit models were developed to generate dynamic harmonics powers. On the basis of transmission line theory and simulations, the features of the IM products (IM3, IM5, and IM7) in a series circuit with the selected PIM source points were investigated. Using this information, a PIM fault diagnosis method for coaxial connectors in a circuit was developed employing a neural network. The IM products powers were obtained as feature parameters using a scattering parameter simulation of series circuit for three different connectors. Simulations were then conducted to generate training and testing samples. Finally, using the back‐propagation neural network algorithm, the fault modes were classified and the results show diagnosis accuracy was 96.72%.