Jie and Enjamamul presented at Engineering Mechanics Institute Conference

CMML members Jie and Enjamamul presented talks at the 2021 Engineering Mechanics Institute Conference:

2021-05-26
Minisymposium #213 Machine Learning-based Computational Modeling for Civil Engineering Applications
#2032317 Data-Driven Modeling of Multi-Parameter Mechanical Problems in Heterogeneous Media
Authors: Jie Hou, Jun Li, Alfa Heryudono

Predicting full-field responses of multi-parameter mechanical problems in heterogeneous material systems is of fundamental importance and has a variety of applications in design optimizations, uncertainty quantification, and structural health monitoring. Physics-based simulations such as the finite element method (FEM) provide high-fidelity predictions but can be computationally expensive and challenging in scenarios of real-time interactive design evaluations and decision-making. On the other hand, data-driven approaches encoded with physical constraints have the promise to rapidly predict reliable results for real-time application scenarios. In this study, a sequence of numerical examples with multi-dimensional parameters of heterogeneous material distributions is considered. Model order reduction techniques of proper orthogonal decomposition (POD) and proper generalized decomposition (PGD) are utilized to reduce the high-dimensional field response variables, respectively. POD is a posteriori method based on the snapshot matrix of field results, while PGD builds on a priori separated representation of the objective field to counter the curse of dimensionality. In POD based framework, an artificial neural network was employed to predict field responses based on POD reduced modes. In PGD based framework, the field responses are approximated by PGD reduced modes in a separated form with extra-coordinates of material variations. Both approaches are capable to predict full-field responses with satisfactory accuracy at a lower computational cost.

 

Minisymposium #255 Physics Informed Machine Learning for Data-driven Modeling and Discovery of Complex Systems
#2032312: Deep-Learning based Stress-Field Prediction of Heterogeneous Media
Authors: Enjamamul Hoq, Osama Aljarrah, Jun Li, Alfa Heryudono, Jing Bi

Rapid and accurate stress-field prediction in heterogeneous material systems is critical for a variety of applications, including design optimization, uncertainty quantification, structural monitoring, and predictive maintenance. The purpose of this research was to design and evaluate several deep learning-based frameworks for capturing the high-dimensional stress field responses of heterogeneous media. The first framework employs a combination of model order reduction and artificial neural networks (ANN). Using correct orthogonal decomposition, the stress fields are first projected to a low-dimensional representation (POD). Following that, ANN is used to restore the dimensionality of the minimized predicted fields. The second framework is based on a deep Convolutional Neural Network (CNN), while the third framework is based on a conditional Generative Adversarial Network (GAN) (cGAN). Two numerical examples were analyzed to determine the efficacy of the suggested frameworks. The first is a panel containing a variety of heterogeneous material inclusions that vary in terms of position and size. In the second example, a plate with holes of varied positions and sizes is considered. Both CNN and cGAN models accurately caught stress concentrations and whole stress fields. The suggested frameworks based on deep learning shown remarkable potential for modeling whole stress fields for a variety of heterogeneous materials.

Congratulations Osama on passing PhD thesis defense!

Congratulations Osama on successfully passing his PhD thesis defense! His doctoral thesis title is “Prediction, Optimization, and Simulations of Additive Manufacturing Process using Surrogate Modeling”. He will start his new position as a tenure-track assistant professor in Industrial and Systems Engineering at Youngstown State University. Best wishes for his new journey!

Congratulations Rojin on passing PhD thesis defense!

 

Congratulations Rojin on passing her PhD thesis defense! Her doctoral thesis title is “Computational Modeling of Anisotropic Fracture in 3D Printed Polymers and Patterned Structures”. In this study, both extended finite element method (XFEM) and phase field fracture method (PFFM) are developed to model anisotropic fractures in 3D printed polymers as well as patterned structures for enhanced fracture resistance.

 

Congratulations Osama on his paper published in ASTM Smart Sustain Manuf Syst

Congratulations Osama! His 2nd paper “A Self-Organizing Evolutionary Method to Model and Optimize Correlated Multiresponse Metrics for Additive Manufacturing Processes” was accepted for publication in ASTM Journal of Smart and Sustainable Manufacturing Systems. Here is the abstract:

A Self-Organizing Evolutionary Method to Model and Optimize Correlated Multiresponse Metrics for Additive Manufacturing Processes

Osama Aljarrah1, Jun Li1*, Wenzhen Huang1, Alfa Heryudono2, and Jing Bi3

1. Department of Mechanical Engineering, University of Massachusetts Dartmouth, Dartmouth, MA 02747

2. Department of Mathematics, University of Massachusetts Dartmouth, Dartmouth, MA 02747

3. Dassault Systemes SIMULIA Corp, Johnston, RI 02919

The use of robust multi-response constrained optimization techniques, where multiple objective responses are involved, is becoming a crucial part in additive manufacturing (AM) processes. Common and popular techniques, in most cases, rely on the assumption of independent responses. In practice, however, many of the desired quality characteristics can be correlated. In this work, we propose a technique based on three ingredients: hybrid self-organizing (HSO) method, desirability function (DF), and evolutionary algorithms (EA) to analyze, model and optimize the multiple correlated responses for the fused deposition modeling (FDM) process, one of the most popular AM technologies. The multi-objective functions are formulated by employing the HSO method and DF, where structural integrity, and process efficiency metrics are considered for the data-driven correlated multi-response models. Subsequently, layer thickness, nozzle temperature, printing speed, and raster angles are taken as process parameters (decision variables). The operational settings and capabilities for the FDM machine are defined as boundary constraints. Different EA algorithms, the non-dominated sorting genetic algorithm (NSGA-II) and the multi-objective particle swarm optimization (MOPSO) method, are then deployed to model the AM criteria accordingly to extract the Pareto-front curve for the correlated multi-response functions. FDM experimental design and data collection for the proposed method are provided and used to validate our approach. This study sheds light on formulating robust and efficient data-driven modeling and optimizations for additive manufacturing processes.

 

Congratulations Rojin on her paper published in Addit Manuf

Congratulations Rojin! Her paper “Extended finite element method (XFEM) modeling of fracture in additively manufactured polymers” was accepted for publication in the journal of Additive Manufacturing. Here is the abstract (the paper link):

Extended finite element method (XFEM) modeling of fracture in additively manufactured polymers

R. Ghandriz1, K. Hart2, J. Li1*

1Department of Mechanical Engineering, University of Massachusetts, Dartmouth, MA 02747

2 Milwaukee School of Engineering, Milwaukee, WI 53202

The fracture of additively manufactured polymer materials with various layer orientations is studied using the extended finite element method (XFEM) in an anisotropic cohesive zone model (CZM). The single edge notched bending (SENB) specimens made of acrylonitrile-butadiene-styrene (ABS) materials through fused filament fabrications with various crack tip/layer orientations are considered. The XFEM coupled with anisotropic CZM is employed to model the brittle fracture (fracture between layers), ductile fracture (fracture through layers), as well as kinked fracture behaviors of ABS specimens printed with vertical, horizontal, and oblique layer orientations, respectively. Both elastic and elastoplastic fracture models, coupled with linear or exponential traction-separation laws, are developed for the inter-layer and cross-layer fracture, respectively. For mixed inter-/cross- layer fracture, an anisotropic cohesive zone model is developed to predict the kinked crack propagations. Two crack initiation and evolution criteria are defined to include both crack propagation between layers (weak plane failure) and crack penetration through layers (maximum principal stress failure) that jointly determine the zig-zag crack growth paths. The anisotropic cohesive zone model with XFEM developed in this study is able to capture different fracture behaviors of additively manufactured ABS samples with different layer orientations.

Congratulations Prosenjit on his paper published in Acta Mechanica

Congratulations Prosenjit! His paper “Numerical Prediction of Orthotropic Elastic Properties of 3D Printed Materials using Micro-CT and Representative Volume Element” was accepted for publication in Acta Mechanica. Here is the abstract (the paper link):

Numerical Prediction of Orthotropic Elastic Properties of 3D Printed Materials using Micro-CT and Representative Volume Element

P. Biswas1, S. Guessasma2, J. Li1*

1. Department of Mechanical Engineering, University of Massachusetts, Dartmouth, MA 02747

2. INRA, UR1268 Biopolymères Interactions Assemblages, F-44300 Nantes, France

The mechanical property of 3D printed components often exhibits anisotropic behaviors and a strong dependence on printing orientations and process parameters. In this study, computational models based on microstructures of 3D printed ABS polymers are developed using micromechanics of representative volume element (RVE) to investigate the orthotropic elastic properties. Two finite element (FE) models, based on Micro-CT or CAD geometry, with different raster angles: 0°, 30°, 45°, and 60° (corresponding to the layups of 0°/90°, 30°/-60°, 45°/-45°, and 60°/-30°), are considered. The Micro-CT model used the realistic geometry of a 3D printed cube reconstructed from Micro-CT scans. The CAD model used periodic layers and filaments with sizes specified according to the dimensional statistics from the Micro-CT model, including bond width, layer height, filament width, and total porosity. All models are subjected to six independent load cases of macroscopically uniform boundary conditions (BCs) admitted by the Hill-Mandel condition to obtain fully orthotropic elastic stiffness matrix. Scale-dependent bounds from those BCs were used to establish the convergence of RVE responses. The theoretical method of Mori-Tanaka homogenization was also employed to verify numerical predictions. The results of both Micro-CT and CAD models are close to each other and agree well with experimental results in the literature. Parametric studies are further performed in CAD models by varying layer height, filament width, and bond width to investigate their effects on the orthotropic elastic properties. It was found that the porosity plays a significant role and more porosity leads to larger elastic anisotropy in 3D printed materials.

Congratulations Prosenjit on passing his MS thesis defense!

Congratulations Prosenjit on successfully passing his MS thesis defense! His thesis title is “Prediction of Elastic and Strength Properties of 3D Printed Materials using Microstructure-based Representative Volume Element”. He will start his PhD in Materials Science and Engineering at the University of Virginia. Farewell, Prosenjit. We wish you all the best on your new journey!

Congratulations Osama on his paper published in Int J Adv Manuf Tech

Congratulations Osama! His paper “ARIMA-GMDH: A low order integrated approach for predicting and optimizing the additive manufacturing process parameters” was accepted for publication in the International Journal of Advanced Manufacturing Technology. Here is the abstract (the paper link):

ARIMA-GMDH: A low order integrated approach for predicting and optimizing the additive manufacturing process parameters

Osama Aljarrah1, Jun Li1*, Wenzhen Huang1, Alfa Heryudono2, and Jing Bi3

1. Department of Mechanical Engineering, University of Massachusetts Dartmouth, Dartmouth, MA 02747

2. Department of Mathematics, University of Massachusetts Dartmouth, Dartmouth, MA 02747

3. Dassault Systemes SIMULIA Corp, Johnston, RI 02919

This paper proposes a novel data-driven approach for predicting and optimizing the additive manufacturing process parameters. The integrated scheme consists of three popular algorithms: (1) group method for data handling (GMDH) as the engine of neural networks, (2) autoregressive integrated moving average (ARIMA) for characterizing spatial collinearity of the multiple response, and (3) indirect optimization on the basis of self-organization (IOSO) to adopt the emerged correlated multi-response optimization problem. As a numerical case study: a computer-generated fused deposition modeling data tested the introduced algorithms. The finite element (FE) simulation model consists of the multi-layer residual stresses as targets, in respect of printing speeds as process parameters. The residual stresses predicted by the low order Integrated ARIMA-GMDH variants correlate well with the FE simulations. This approach provides a viable data-driven alternative for computationally-based rapid prototyping and additive manufacturing processes.

1 2 3