Acoustic Reflection Classification of PVDF Sensor Using Convolutional Neural Network

Abstract

Recently, polyvinylidene fluoride (PVDF) has been attracting attention in various fields due to its properties as a piezoelectric material with high thermal stability and flexibility. Therefore, in this experiment, we want to acquire the acoustic emission signal using a PVDF sensor that has the potential to be replaced with an AE sensor. Acoustic Emission (AE) signals are acoustic waves generated inside a material, and due to the internal geometry, there is a possibility that the source may be reflected and the signal information may be deformed. Therefore, to check whether the PVDF sensor acquires the acoustic emission signal like the AE sensor and can classify the reflected and transformed signal information, the acoustic signal most similar to the natural acoustic emission source was acquired through the Pencil lead break (PLB) test. The signal was acquired through the same two PVDF sensors, and it can be divided into an original signal (Direct Wave) and a converted signal (Reflected Wave) through a Convolution Neural Network (CNN) depending on the attachment location of the sensor. This shows that PVDF sensors can acquire and classify AE signals through deep learning and that PVDF can replace AE sensors in AE signal acquisition.