This study primarily aims to develop a robust modelling approach to capture complex material behavior of CP-Ti, appeared by high anisotropy, differential hardening due to anisotropy evolution, and flow behavior sensitive to strain rate and temperature, using artificial neural networks (ANNs). Plasticity is characterized by uniaxial tension and in-plane biaxial tension tests at temperatures of 0°C and 20°C with strain rates of 0.001 /s and 0.01 /s, and the results are used to calibrate the non-quadratic anisotropic Yld2000-3d yield function with respect to the plastic work. In order to predict the intricate plastic deformation with the temperature and strain rate effects, two distinct ANN models are developed; one is to capture the strain hardening behavior and the other to predict the anisotropic parameters in the chosen yield function. The developed ANN models predict an unseen dataset well, which is intermediate testing conditions at a temperature of 10°C and strain rate of 0.005 /s. The ANN models, being computationally stable and adhering to conventional constitutive equations, are implemented into a user material subroutine for the ductile fracture characterization of CP-Ti sheet using the hybrid experimental-numerical analysis. The favorable agreement between experimental data and numerical predictions, particularly using the ANN models with evolving anisotropic material parameters for the Yld2000-3d yield function, underscores the significance of differential hardening effect on the ductile fracture behavior and highlights the capabilities of ANN models to capture the complex plastic behavior of CP-Ti. The key parameters including stress triaxiality, Lode angle parameter, and equivalent plastic strain at the fracture location are extracted from the simulations, enabling the calibration of ductile fracture models, namely Johnson-Cook, Hosford-Coulomb, and Lou-2014, and construction of fracture envelopes.
more »
« less
This content will become publicly available on May 1, 2025
Identification of Yld2000–2d anisotropic yield function parameters from single hole expansion test using machine learning
This study presents a novel machine learning approach for predicting the anisotropic parameters of the
Yld20002d non-quadratic yield function using a hole expansion test. Heterogeneous stress-strain fields during the test substitute for multiple experiments required in the conventional parameter identification approach.
An artificial neural network model for the parameter prediction is developed using a virtually generated training dataset composed of strains from hole expansion simulations, performed using randomly selected anisotropic parameters. The developed model predicts the Yld20002d parameters for AA6022-T4 based on the
measured strain field from a hole expansion experiment, and the parameter results are evaluated by comparing
anisotropy in uniaxial tension tests, the yield locus, and thinning variation in hole expansion test.
more »
« less
- Award ID(s):
- 1757371
- NSF-PAR ID:
- 10519338
- Publisher / Repository:
- Elsevier
- Date Published:
- Journal Name:
- CIRP Annals
- ISSN:
- 0007-8506
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract A numerical and experimental hybrid approach is developed to study the constitutive behavior of the central nervous system white matter. A published transversely isotropic hyperelastic strain energy function is reviewed and used to determine stress–strain relationships for three idealized, simple loading scenarios. The proposed constitutive model is simplified to a three-parameter hyperelastic model by assuming the white matter's incompressibility. Due to a lack of experimental data in all three loading scenarios, a finite element model that accounts for microstructural axons and their kinematics is developed to simulate behaviors in simple shear loading scenarios to supplement existing uniaxial tensile test data. The parameters of the transversely isotropic hyperelastic material model are determined regressively using the hybrid data. The results highlight that a hybrid numerical virtual test coupled with experimental data, can determine the transversely isotropic hyperelastic model. It is noted that the model is not limited to small strains and can be applied to large deformations.more » « less
-
The quantification of strain in three-dimensions is a powerful tool for structural investigations, allowing for the direct consideration of the localization and delocalization of deformation in space, and potentially, in time. Furthermore, characterization of the distribution of strain in three-dimensions may yield information concerning large-scale kinematics that may not be obtained through the traditional use of asymmetric fabrics. In this contribution, we present a streamlined methodology for the calculation of three-dimensional strain using objective approaches that allow for consideration of error assessment. This approach begins with the collection of suitable samples for strain analysis following either the Rf/ϕ or normalized Fry techniques. Samples are cut along three mutually perpendicular orientations using a set of jigs designed for use in a large oil saw. Cut faces are polished and scanned in high resolution. Scanned images are processed following a standard convention. The boundaries of objects are outlined as “Regions Of Interest” in the open-source program ImageJ and saved. A script reads the saved files of object outlines and statistically fits an ellipse to each digitized object. The parameters of fitted objects are then extracted and saved. Two-dimensional strain analyses are completed following the normalized Fry method or the Rf/ϕ technique following a bootstrap statistical approach. For the normalized Fry method, an objective fitting routine modified from Mulchrone (2013) is used to determine the parameters of the central void. For the Rf/ϕ method, an inverse straining routine is applied and tests the resulting object orientations to a random uniform distribution following a Kolmogorov–Smirnov test in order to obtain the sectional strain ratio and orientation. Bootstrap sampling of Fry coordinates or objects results in a distribution of possible sectional strains that can be sampled for fitting of strain ellipsoids following the technique of Robin (2002). As such, the parameters of three-dimensional strain including Lode parameter and octahedral shear strain can be contoured based on confidence intervals for each sample processed. The application of the objective approach is presented in a corresponding poster.more » « less
-
more » « less
We extend the doubly degenerate Cahn–Hilliard (DDCH) models for isotropic surface diffusion, which yield more accurate approximations than classical degenerate Cahn–Hilliard (DCH) models, to the anisotropic case. We consider both weak and strong anisotropies and demonstrate the capabilities of the approach for these cases numerically. The proposed model provides a variational and energy dissipative approach for anisotropic surface diffusion, enabling large‐scale simulations with material‐specific parameters.
-
more » « less
Understanding the yielding of complex fluids is an important rheological challenge that affects our ability to engineer and process materials for a wide variety of applications. Common theoretical understandings of yield stress fluids follow the Oldroyd–Prager formalism in which the material behavior below the yield stress is treated as solidlike, and above the yield stress as liquidlike, with an instantaneous transition between the two states. This formalism was built on a quasi-static approach to the yield stress, while most applications, ranging from material processing to end user applications, involve a transient approach to yielding over a finite timescale. Using stress-controlled oscillatory shear experiments, we show that yield stress fluids flow below their yield stresses. This is quantified through measuring the strain shift, which is the value about which the strain oscillates during a stress-controlled test and is a function of only the unrecoverable strain. Measurements of the strain shift are, therefore, measurements of flow having taken place. These experimental results are compared to the Herschel–Bulkley form of the Saramito model, which utilizes the Oldroyd–Prager formalism, and the recently published Kamani–Donley–Rogers (KDR) model, in which one constitutive equation represents the entire range of material responses. Scaling relationships are derived, which allow us to show why yield stress fluids will flow across all stresses, above and below their yield stress. Finally, derivations are presented that show strain shift can be used to determine average metrics previously attainable only through recovery rheology, and these are experimentally verified.
