Acquisition of Dynamic Material Properties in the Electrohydraulic Forming Process Using Artificial Neural Network

Abstract

Electrohydraulic forming is a high-velocity forming process that deforms sheet metals with velocities above 100 m/s and strain rates more than 100 s−1. This experiment was conducted in a closed space because of safety concerns related to the high-velocity conditions; therefore, we were not able to examine the deformation process of the sheet metal. To observe the electrohydraulic forming process in detail, we performed virtual numerical simulations using accurate material properties. Therefore, in this paper, we obtained the material property of a sheet metal from a numerical estimation by using a surrogate model based on the reduced order model and the artificial neural network. The Cowper–Symonds constitutive equation was selected for the Al 6061-T6 sheet metal, and two strain rate parameters were adopted as the unknown parameters. From the two sampling techniques, the training and test samples were extracted from the specific ranges of two unknown parameters, and a numerical simulation was performed for these samples by using the LS-DYNA program. The z-axis displacements of the deformed sheet metal were obtained from the results of the numerical simulation, and two basis vectors were extracted by using principal component analysis. In addition, to predict the weighting coefficients of the two basis vectors at the defined range of parameters, we used the artificial neural network technique as a surrogate model. By comparing the surrogate model and the experimental results and calculating the root mean square error value, we estimated the optimal parameter for Al 6061-T6. Finally, the reliability of the obtained material parameters was proved by comparing the experimental results, the surrogate model, and LS-DYNA.

Accurate Finite Element Simulations of Dynamic Behaviour: Constitutive Models and Analysis with Deep Learning

“…Here, σ represents the uniaxial flow stress, while (A + Bε n ) denotes the quasi-static flow stress, and D and q are material constants involved in the model. The specific constants for 6061 Al are provided in Table 2 [53]. The ZA model is derived based on dislocation mechanisms, which significantly influence the elastic behaviour and flow stress of metals under varying load conditions.…”

Numerical Estimation of Material Properties in the Electrohydraulic Forming Process Based on a Kriging Surrogate Model

“…A, B, and n are the quasistatic material parameters, and C and p are the high-strain-rate material parameters. e material parameters of Al 6061-T6 were determined in a quasistatic tensile test conducted on an Instron universal testing machine at 0.0007 s −1 , yielding A � 291.8 MPa, B � 451.5 MPa, and n � 0.666 [18]. Both C and p were unknown parameters θ ∈ R 2 to be found by inverse estimation in the Kriging surrogate model.…”

“…Previous study [18] described a parameter estimation method based on artificial neural network (ANN) method. However, ANN generates slightly different models according to internal weight factors and it constructs various models according to the number of nodes and hidden layers even at the same input and output, which means that ANN is suitable for the problem which should check the output tendency at given input, but it may not be appropriate when an accurate value should be obtained as in this paper.…”

A Parametric Physics-Informed Deep Learning Method for Probabilistic Design of Thermal Protection Systems

“…However, this approach necessitates calling the system model multiple times for computation, resulting in significant computational costs and reduced computational efficiency. Alternatively, the original high-precision complicated model can be substituted with a surrogate model to enhance computational efficiency while preserving accuracy [6][7][8][9][10]. A surrogate model refers to an approximation model that describes the relationship between the input and output of a system.…”