Detailed Study on the Behavior of Improved Beam T-Junctions Modeling for the Characterization of Tubular Structures, Based on Artificial Neural Networks Trained with Finite Element Models

Abstract

The actual behavior of welded T-junctions in tubular structures depends strongly on the topology of the junction at the joint level. In finite element analysis, beam-type elements are usually employed due to their simplicity and low computational cost, even though they cannot reproduce the joints topologies and characteristics. To adjust their behavior to a more realistic situation, elastic elements can be introduced at the joint level, whose characteristics must be determined through costly validations. This paper studies the optimization and implementation of the validation data, through the creation of an optimal surrogate model based on neural networks, leading to a model that predicts the stiffness of elastic elements, introduced at the joint level based on available data. The paper focuses on how the neural network should be chosen, when training data is very limited and, more importantly, which of the available data should be used for training and which for verification. The methodology used is based on a Monte Carlo analysis that allows an exhaustive study of both the network parameters and the distribution and choice of the limited data in the training set to optimize its performance. The results obtained indicate that the use of neural networks without a careful methodology in this type of problems could lead to inaccurate results. It is also shown that a conscientious choice of training data, among the data available in the problem of choice of elastic parameters for T-junctions in finite elements, is fundamental to achieve functional surrogate models.

On the Application of Neural Networks Trained with FEM Data for the Identification of Stiffness Parameters of Improved Mechanical Beam Joints

“…This aspect was also analyzed in [2], in which the authors concluded that the beamtype elements "were simply not designed to take into account and reproduce the behavior of real joints, since all the elements that reach the same joint are joined in a single infinitely rigid node". The main drawback of beam-element-type modeling is represented in Figure 3, where the same beam T-junction model is obtained for both T1 and T2 real junctions, despite the fact that they have significant topological differences at the joint level with consequent behavioral differences between both junctions.…”

“…This aspect was also analyzed in [2], in which the authors concluded that the beam-type elements "were simply not designed to take into account and reproduce the behavior of real joints, since all the elements that reach the same joint are joined in a single infinitely rigid node".…”

“…In order to achieve realistic and feasible industrial development for this methodol ogy, various studies have estimated these stiffness values using a variety of methods such as structural FEM comparative analysis [3], experimental and FEM modal analysis [7] kriging regression models [8], or neural networks [2]. These methods have all resulted in satisfactory improvements in accuracy for T-junction models.…”

“…In this context, the use of ANNs, already explored by the authors in [2], presents a flexible and fast alternative for estimating the elastic elements, thereby avoiding the need for tests or joint-specific analysis.…”

“…The extensive analysis presented herein provides a valuable reference to be used as a starting point for similar works involving FEM structural beam models. In this context, the use of ANNs, already explored by the authors in [2], presents a flexible and fast alternative for estimating the elastic elements, thereby avoiding the need for tests or joint-specific analysis.…”