Eventos Anais de eventos
COBEM 2017
24th ABCM International Congress of Mechanical Engineering
Analysis and Classification of Welded Joint Defects by Ultrassonic Testing and Artificial Neural Networks
Submission Author:
Tadeu Mansur Barroso , RJ
Co-Authors:
José Flávio Silveira Feiteira, Rodrigo Pinto de Siqueira, Tadeu Mansur Barroso
Presenter: Tadeu Mansur Barroso
doi://10.26678/ABCM.COBEM2017.COB17-1162
Abstract
In this work, the main objective is to evaluate the structural integrity of welded joints in steel components through the ultrasonic nondestructive test, applying computational-mathematical methods in order to reduce the time needed to evaluate the integrity of a joint and improve the production rate. Thus, artificial neural networks and mathematical concepts will combine to detect the higher variance points, such as principal component analysis. In other words, the aim is to get the best rate of accuracy compared to the computational cost needed, also other possible architectures to the neural network and the mathematical concepts to reduce inspection time.
Keywords
Artificial neural networks, Ultrasonic testing, Welding defects, Welding profile
