A Terahertz Identification Method for Internal Interface Structures of Polymers Based on the Long Short-Term Memory Classification Network

Abstract

Polymers are used widely in the power system as insulating materials and are essential to the power grid’s security and stability. However, various insulation defects may occur in the polymer., which can lead to severe insulation accidents. Terahertz (THz) detection is a novel non-destructive testing (NDT) method that is able to detect the interface structures inside polymers. The large quantity of information in the THz waveform has potential for the identification of interface types, and the long short-term memory (LSTM) network is one of the most popular artificial intelligence methods for time series data like THz waveform. In this paper, the LSTM classification network was used to identify the internal interfaces of the polymer with the reflected THz pulses of the internal interfaces. The experiment verified that it is feasible to identify and image the void interfaces and impurity interfaces in the polymer using the proposed method.