Abstract
This research quantitatively confirmed an acoustical anisotropy that exists in a pure titanium plate from the signal of ultrasonic flow detection and suggested a new way to evaluate the acoustical anisotropy by inputting acquired characteristic of ultrasound signal into the neutral network. Using the fact with the suggested method that the characteristic of ultrasound signal is shown differently depending on the pure titanium plate's rolling direction, the neural network was constructed by extracting the characteristic that can decide each direction of $0^{\circ}$, $45^{\circ}$, and $90^{\circ}$ with waveform analysis program. As a result of inputting the characteristic of ultrasound signal acquired from a random rolling direction into the neural network that was built like this, it showed a pattern recognition rate higher than 95% on directions of $0^{\circ}$, $45^{\circ}$, $90^{\circ}$.
