Eventos Anais de eventos
COBEM 2023
27th International Congress of Mechanical Engineering
Using an Inverse Method Based on Deep Learning to Obtain Shear Parameters to Optimize a Shear-Displacement Finite Element Solution
Submission Author:
Luana Claudia Bertoncello , PR
Co-Authors:
Luana Claudia Bertoncello, Francisco Augusto Aparecido Gomes, Robson Gonçalves Trentin, Paulo Rogério Novak
Presenter: Luana Claudia Bertoncello
doi://10.26678/ABCM.COBEM2023.COB2023-1389
Abstract
The application of computational simulation using finite element methods has been intensified to solve engineering problems. The technique adopted in this work is known as a solution based on an inverse problem. This technique is based on the optimization of certain input parameters, which will be used to minimize errors in the main physical solution. The objective of this work is to develop a computer simulation using the finite element method in conjunction with an algorithm based on deep learning to minimize solution errors. The application of deep learning in engineering has been growing in recent years, due to its ability to solve and minimize errors in complex problems. Through the use of artificial neurons, the main feature of the neural network is to simulate the functioning of human neurons that are interconnected by an artificial synapse. The numerical study will be based on the direct pullout test of a semi-smooth steel bar immersed in a concrete specimen, whose experimental data will be the input data of the neural network. In this case, input parameters such as shear stress and adhesion will be optimized from the neural network, in which shear will characterize concrete failure due to steel bar slippage. A validation of the numerical methodology will be presented to analyze the efficiency of the optimization method in relation to the accuracy of the physical solution, considering values of the bond stress between the steel bar and concrete (shear) and the sliding of the steel bar in relation to the concrete.
Keywords
Optimization, Inverse problem, Finite Element Method, Artificial neural networks
