HiDeNN-FEM: A seamless machine learning approach to nonlinear finite element analysis.

The hierarchical deep-learning neural network (HiDeNN) (Zhang et al, Computational Mechanics, 67:207-230) provides a systematic approach to constructing numerical approximations that can be incorporated into a wide variety of Partial differential equations (PDE) and/or Ordinary differential equations (ODE) solvers. This paper presents a framework of the nonlinear finite element based on HiDeNN approximation (nonlinear HiDeNN-FEM). This is enabled by three basic building blocks employing structured deep neural networks: 1) A partial derivative operator block that performs the differentiation of the shape functions with respect to the element coordinates, 2) An r-adaptivity block that improves the local and global convergence properties and 3) A materials derivative block that evaluates the material derivatives of the shape function. While these building blocks can be applied to any element, specific implementations are presented in 1D and 2D to illustrate the application of the deep learning neural network. Two-step optimization schemes are further developed to allow for the capabilities of r-adaptivity and easy integration with any existing FE solver. Numerical examples of 2D and 3D demonstrate that the proposed nonlinear HiDeNN-FEM with r-adaptivity provides much higher accuracy than regular FEM. It also significantly reduces element distortion and suppresses the hourglass mode.

Data-driven discovery of dimensionless numbers and governing laws from scarce measurements.

Dimensionless numbers and scaling laws provide elegant insights into the characteristic properties of physical systems. Classical dimensional analysis and similitude theory fail to identify a set of unique dimensionless numbers for a highly multi-variable system with incomplete governing equations. This paper introduces a mechanistic data-driven approach that embeds the principle of dimensional invariance into a two-level machine learning scheme to automatically discover dominant dimensionless numbers and governing laws (including scaling laws and differential equations) from scarce measurement data. The proposed methodology, called dimensionless learning, is a physics-based dimension reduction technique. It can reduce high-dimensional parameter spaces to descriptions involving only a few physically interpretable dimensionless parameters, greatly simplifying complex process design and system optimization. We demonstrate the algorithm by solving several challenging engineering problems with noisy experimental measurements (not synthetic data) collected from the literature. Examples include turbulent Rayleigh-Bénard convection, vapor depression dynamics in laser melting of metals, and porosity formation in 3D printing. Lastly, we show that the proposed approach can identify dimensionally homogeneous differential equations with dimensionless number(s) by leveraging sparsity-promoting techniques.

Universal scaling laws of keyhole stability and porosity in 3D printing of metals.

Metal three-dimensional (3D) printing includes a vast number of operation and material parameters with complex dependencies, which significantly complicates process optimization, materials development, and real-time monitoring and control. We leverage ultrahigh-speed synchrotron X-ray imaging and high-fidelity multiphysics modeling to identify simple yet universal scaling laws for keyhole stability and porosity in metal 3D printing. The laws apply broadly and remain accurate for different materials, processing conditions, and printing machines. We define a dimensionless number, the Keyhole number, to predict aspect ratio of a keyhole and the morphological transition from stable at low Keyhole number to chaotic at high Keyhole number. Furthermore, we discover inherent correlation between keyhole stability and porosity formation in metal 3D printing. By reducing the dimensions of the formulation of these challenging problems, the compact scaling laws will aid process optimization and defect elimination during metal 3D printing, and potentially lead to a quantitative predictive framework.

Microscale Structure to Property Prediction for Additively Manufactured IN625 through Advanced Material Model Parameter Identification.

Challenge 4 of the Air Force Research Laboratory additive manufacturing modeling challenge series asks the participants to predict the grain-average elastic strain tensors of a few specific challenge grains during tensile loading, based on experimental data and extensive characterization of an IN625 test specimen. In this article, we present our strategy and computational methods for tackling this problem. During the competition stage, a characterized microstructural image from the experiment was directly used to predict the mechanical responses of certain challenge grains with a genetic algorithm-based material model identification method. Later, in the post-competition stage, a proper generalized decomposition (PGD)-based reduced order method is introduced for improved material model calibration. This data-driven reduced order method is efficient and can be used to identify complex material model parameters in the broad field of mechanics and materials science. The results in terms of absolute error have been reported for the original prediction and re-calibrated material model. The predictions show that the overall method is capable of handling large-scale computational problems for local response identification. The re-calibrated results and speed-up show promise for using PGD for material model calibration.

Macroscale Property Prediction for Additively Manufactured IN625 from Microstructure through Advanced Homogenization.

Design of additively manufactured metallic parts requires computational models that can predict the mechanical response of parts considering the microstructural, manufacturing, and operating conditions. This article documents our response to Air Force Research Laboratory (AFRL) Additive Manufacturing Modeling Challenge 3, which asks the participants to predict the mechanical response of tensile coupons of IN625 as function of microstructure and manufacturing conditions. A representative volume element (RVE) approach was coupled with a crystal plasticity material model solved within the Fast Fourier Transformation (FFT) framework for mechanics to address the challenge. During the competition, material model calibration proved to be a challenge, prompting the introduction in this manuscript of an advanced material model identification method using proper generalized decomposition (PGD). Finally, a mechanistic reduced order method called Self-consistent Clustering Analysis (SCA) is shown as a possible alternative to the FFT method for solving these problems. Apart from presenting the response analysis, some physical interpretation and assumptions associated with the modeling are discussed.

Data-driven characterization of thermal models for powder-bed-fusion additive manufacturing.

Computational modeling for additive manufacturing has proven to be a powerful tool to understand physical mechanisms, predict fabrication quality, and guide design and optimization. Varieties of models have been developed with different assumptions and purposes, and these models are sometimes difficult to choose from, especially for end-users, due to the lack of quantitative comparison and standardization. Thus, this study is focused on quantifying model uncertainty due to the modeling assumptions, and evaluating differences based on whether or not selected physical factors are incorporated. Multiple models with different assumptions, including a high-fidelity thermal-fluid flow model resolving individual powder particles, a low-fidelity heat transfer model simplifying the powder bed as a continuum material, and a semi-analytical thermal model using a point heat source model, were run with a variety of manufacturing process parameters. Experiments were performed on the National Institute of Standards and Technology (NIST) Additive Manufacturing Metrology Testbed (AMMT) to validate the models. A data analytics-based methodology was utilized to characterize the models to estimate the error distribution. The cross comparison of the simulation results reveals the remarkable influence of fluid flow, while the significance of the powder layer varies across different models. This study aims to provide guidance on model selection and corresponding accuracy, and more importantly facilitate the development of AM models.

Cell and nanoparticle transport in tumour microvasculature: the role of size, shape and surface functionality of nanoparticles.

Through nanomedicine, game-changing methods are emerging to deliver drug molecules directly to diseased areas. One of the most promising of these is the targeted delivery of drugs and imaging agents via drug carrier-based platforms. Such drug delivery systems can now be synthesized from a wide range of different materials, made in a number of different shapes, and coated with an array of different organic molecules, including ligands. If optimized, these systems can enhance the efficacy and specificity of delivery compared with those of non-targeted systems. Emerging integrated multiscale experiments, models and simulations have opened the door for endless medical applications. Current bottlenecks in design of the drug-carrying particles are the lack of knowledge about the dispersion of these particles in the microvasculature and of their subsequent internalization by diseased cells (Bao et al. 2014 J. R. Soc. Interface 11, 20140301 (doi:10.1098/rsif.2014.0301)). We describe multiscale modelling techniques that study how drug carriers disperse within the microvasculature. The immersed molecular finite-element method is adopted to simulate whole blood including blood plasma, red blood cells and nanoparticles. With a novel dissipative particle dynamics method, the beginning stages of receptor-driven endocytosis of nanoparticles can be understood in detail. Using this multiscale modelling method, we elucidate how the size, shape and surface functionality of nanoparticles will affect their dispersion in the microvasculature and subsequent internalization by targeted cells.

Automatised selection of load paths to construct reduced-order models in computational damage micromechanics: from dissipation-driven random selection to Bayesian optimization.

In this paper, we present new reliable model order reduction strategies for computational micromechanics. The difficulties rely mainly upon the high dimensionality of the parameter space represented by any load path applied onto the representative volume element. We take special care of the challenge of selecting an exhaustive snapshot set. This is treated by first using a random sampling of energy dissipating load paths and then in a more advanced way using Bayesian optimization associated with an interlocked division of the parameter space. Results show that we can insure the selection of an exhaustive snapshot set from which a reliable reduced-order model can be built.

Shape effect in cellular uptake of PEGylated nanoparticles: comparison between sphere, rod, cube and disk.

The size, shape, surface property and material composition of polymer-coated nanoparticles (NPs) are four important parameters in designing efficient NP-based carriers for targeted drug delivery. However, due to the complex interplay between size, shape and surface property, most studies lead to ambiguous descriptions of the relevance of shape. To clarify its influence on the cellular uptake of PEGylated NPs, large scale molecular simulations have been performed to study differently shaped convex NPs, such as sphere, rod, cube and disk. Comparing systems with identical NP surface area, ligand-receptor interaction strength, and grafting density of the polyethylene glycol, we find that the spherical NPs exhibit the fastest internalization rate, followed by the cubic NPs, then rod- and disk-like NPs. The spherical NPs thus demonstrate the highest uptake among these differently shaped NPs. Based on a detailed free energy analysis, the NP shape effect is found to be mainly induced by the different membrane bending energies during endocytosis. The spherical NPs need to overcome a minimal membrane bending energy barrier, compared with the non-spherical counterparts, while the internalization of disk-like NPs involves a strong membrane deformation, responsible for a large free energy barrier. Besides, the free energy change per tethered chain is about a single kBT regardless of NP shape, as revealed by our self-consistent field theory calculations, where kB and T denote Boltzmann constant and temperature, respectively. Thus, the NP shape only plays the secondary role in the free energy change of grafted PEG polymers during internalization. We also find that star-shaped NPs can be quickly wrapped by the cell membrane, similar to their spherical counterparts, indicating star-shaped NPs can be used for drug delivery with high efficacy. Our findings seem to provide useful guidance in the molecular design of PEGylated NPs for controllable cellular uptake and help establish quantitatively rules in designing NP-based vectors for targeted drug delivery.

Endocytosis of PEGylated nanoparticles accompanied by structural and free energy changes of the grafted polyethylene glycol.

Nanoparticles (NPs) are in use to efficiently deliver drug molecules into diseased cells. The surfaces of NPs are usually grafted with polyethylene glycol (PEG) polymers, during so-called PEGylation, to improve water solubility, avoid aggregation, and prevent opsonization during blood circulation. The interplay between grafting density σp and grafted PEG polymerization degree N makes cellular uptake of PEGylated NPs distinct from that of bare NPs. To understand the role played by grafted PEG polymers, we study the endocytosis of 8 nm sized PEGylated NPs with different σp and N through large scale dissipative particle dynamics (DPD) simulations. The free energy change Fpolymer of grafted PEG polymers, before and after endocytosis, is identified to have an effect which is comparable to, or even larger than, the bending energy of the membrane during endocytosis. Based on self-consistent field theory Fpolymer is found to be dependent on both σp and N. By incorporating Fpolymer, the critical ligand-receptor binding strength for PEGylated NPs to be internalized can be correctly predicted by a simple analytical equation. Without considering Fpolymer, it turns out impossible to predict whether the PEGylated NPs will be delivered into the diseased cells. These simulation results and theoretical analysis not only provide new insights into the endocytosis process of PEGylated NPs, but also shed light on the underlying physical mechanisms, which can be utilized for designing efficient PEGylated NP-based therapeutic carriers with improved cellular targeting and uptake.

USNCTAM perspectives on mechanics in medicine.

Over decades, the theoretical and applied mechanics community has developed sophisticated approaches for analysing the behaviour of complex engineering systems. Most of these approaches have targeted systems in the transportation, materials, defence and energy industries. Applying and further developing engineering approaches for understanding, predicting and modulating the response of complicated biomedical processes not only holds great promise in meeting societal needs, but also poses serious challenges. This report, prepared for the US National Committee on Theoretical and Applied Mechanics, aims to identify the most pressing challenges in biological sciences and medicine that can be tackled within the broad field of mechanics. This echoes and complements a number of national and international initiatives aiming at fostering interdisciplinary biomedical research. This report also comments on cultural/educational challenges. Specifically, this report focuses on three major thrusts in which we believe mechanics has and will continue to have a substantial impact. (i) Rationally engineering injectable nano/microdevices for imaging and therapy of disease. Within this context, we discuss nanoparticle carrier design, vascular transport and adhesion, endocytosis and tumour growth in response to therapy, as well as uncertainty quantification techniques to better connect models and experiments. (ii) Design of biomedical devices, including point-of-care diagnostic systems, model organ and multi-organ microdevices, and pulsatile ventricular assistant devices. (iii) Mechanics of cellular processes, including mechanosensing and mechanotransduction, improved characterization of cellular constitutive behaviour, and microfluidic systems for single-cell studies.

Dynamic structure of unentangled polymer chains in the vicinity of non-attractive nanoparticles.

Using coarse-grained molecular dynamics simulation, we study the motion of unentangled polymer chains dynamically confined by non-attractive nanoparticles (NPs). Both normal mode and dynamic structure factor S(q, t) analysis are adopted to analyze chain's dynamics. Relaxation behaviors of chains are found to be significantly slowed down by NPs. The relaxation times of chain's normal modes are monotonically increasing with the NP volume fraction ϕ. At the same time, chains' dynamics are becoming non-Gaussian. Inspection of S(q, t) reveals that chain's dynamics can be attributed to two 'phases', a bulk polymer phase and a confined polymer phase between NPs. The dynamics of a confined polymer is slower than that of a bulk polymer, while still exhibiting high mobility. The amount of the bulk polymer phase is found to exponentially decay with increasing ϕ. With this figure at hand, we establish a simple relationship between NP and confined/interphase polymer volume fractions. This work seems to provide the first quantitative prediction on the relationship between NP and confined/interphase polymer volume fractions.

Ternary and senary representations using DNA double-crossover tiles.

The information capacity of DNA double-crossover (DX) tiles was successfully increased beyond a binary representation to higher base representations. By controlling the length and the position of DNA hairpins on the DX tile, ternary and senary (base-3 and base-6) digit representations were realized and verified by atomic force microscopy. Also, normal mode analysis was carried out to study the mechanical characteristics of each structure.

Synthesis of nanodiamond-daunorubicin conjugates to overcome multidrug chemoresistance in leukemia.

Nanodiamonds (NDs) are promising candidates in nanomedicine, demonstrating significant potential as gene/drug delivery platforms for cancer therapy. We have synthesized ND vectors capable of chemotherapeutic loading and delivery with applications towards chemoresistant leukemia. The loading of Daunorubicin (DNR) onto NDs was optimized by adjusting reaction parameters such as acidity and concentration. The resulting conjugate, a novel therapeutic payload for NDs, was characterized extensively for size, surface charge, and loading efficiency. A K562 human myelogenous leukemia cell line, with multidrug resistance conferred by incremental DNR exposure, was used to demonstrate the efficacy enhancement resulting from ND-based delivery. While resistant K562 cells were able to overcome treatment from DNR alone, as compared with non-resistant K562 cells, NDs were able to improve DNR delivery into resistant K562 cells. By overcoming efflux mechanisms present in this resistant leukemia line, ND-enabled therapeutics have demonstrated the potential to improve cancer treatment efficacy, especially towards resistant strains. FROM THE CLINICAL EDITOR: The authors of this study demonstrate superior treatment properties of resistant leukemia cell lines by utilizing nanodiamond vectors loaded with daunorubicin, paving the way to clinical studies in the hopefully not too distant future.

Quantifying uncertainties in the microvascular transport of nanoparticles.

The character of nanoparticle dispersion in the microvasculature is a driving factor in nanoparticle-based therapeutics and bio-sensing. It is difficult, with current experimental and engineering capability, to understand dispersion of nanoparticles because their vascular system is more complex than mouse models and because nanoparticle dispersion is so sensitive to in vivo environments. Furthermore, uncertainty cannot be ignored due to the high variation of location-specific vessel characteristics as well as variation across patients. In this paper, a computational method that considers uncertainty is developed to predict nanoparticle dispersion and transport characteristics in the microvasculature with a three step process. First, a computer simulation method is developed to predict blood flow and the dispersion of nanoparticles in the microvessels. Second, experiments for nanoparticle dispersion coefficients are combined with results from the computer model to suggest the true values of its unknown and unmeasurable parameters-red blood cell deformability and red blood cell interaction-using the Bayesian statistical framework. Third, quantitative predictions for nanoparticle transport in the tumor microvasculature are made that consider uncertainty in the vessel diameter, flow velocity, and hematocrit. Our results show that nanoparticle transport is highly sensitive to the microvasculature.

Efficient prediction of protein conformational pathways based on the hybrid elastic network model.

Various computational models have gained immense attention by analyzing the dynamic characteristics of proteins. Several models have achieved recognition by fulfilling either theoretical or experimental predictions. Nonetheless, each method possesses limitations, mostly in computational outlay and physical reality. These limitations remind us that a new model or paradigm should advance theoretical principles to elucidate more precisely the biological functions of a protein and should increase computational efficiency. With these critical caveats, we have developed a new computational tool that satisfies both physical reality and computational efficiency. In the proposed hybrid elastic network model (HENM), a protein structure is represented as a mixture of rigid clusters and point masses that are connected with linear springs. Harmonic analyses based on the HENM have been performed to generate normal modes and conformational pathways. The results of the hybrid normal mode analyses give new physical insight to the 70S ribosome. The feasibility of the conformational pathways of hybrid elastic network interpolation (HENI) was quantitatively evaluated by comparing three different overlap values proposed in this paper. A remarkable observation is that the obtained mode shapes and conformational pathways are consistent with each other. Our timing results show that HENM has some advantage in computational efficiency over a coarse-grained model, especially for large proteins, even though it takes longer to construct the HENM. Consequently, the proposed HENM will be one of the best alternatives to the conventional coarse-grained ENMs and all-atom based methods (such as molecular dynamics) without loss of physical reality.

On the near-wall accumulation of injectable particles in the microcirculation: smaller is not better.

Although most nanofabrication techniques can control nano/micro particle (NMP) size over a wide range, the majority of NMPs for biomedical applications exhibits a diameter of ~100 nm. Here, the vascular distribution of spherical particles, from 10 to 1,000 nm in diameter, is studied using intravital microscopy and computational modeling. Small NMPs (≤100 nm) are observed to move with Red Blood Cells (RBCs), presenting an uniform radial distribution and limited near-wall accumulation. Larger NMPs tend to preferentially accumulate next to the vessel walls, in a size-dependent manner (~70% for 1,000 nm NMPs). RBC-NMP geometrical interference only is responsible for this behavior. In a capillary flow, the effective radial dispersion coefficient of 1,000 nm particles is ~3-fold larger than Brownian diffusion. This suggests that sub-micron particles could deposit within diseased vascular districts more efficiently than conventional nanoparticles.

A mass weighted chemical elastic network model elucidates closed form domain motions in proteins.

An elastic network model (ENM), usually Cα coarse-grained one, has been widely used to study protein dynamics as an alternative to classical molecular dynamics simulation. This simple approach dramatically saves the computational cost, but sometimes fails to describe a feasible conformational change due to unrealistically excessive spring connections. To overcome this limitation, we propose a mass-weighted chemical elastic network model (MWCENM) in which the total mass of each residue is assumed to be concentrated on the representative alpha carbon atom and various stiffness values are precisely assigned according to the types of chemical interactions. We test MWCENM on several well-known proteins of which both closed and open conformations are available as well as three α-helix rich proteins. Their normal mode analysis reveals that MWCENM not only generates more plausible conformational changes, especially for closed forms of proteins, but also preserves protein secondary structures thus distinguishing MWCENM from traditional ENMs. In addition, MWCENM also reduces computational burden by using a more sparse stiffness matrix.

Dielectrophoretic concentration of low-abundance nanoparticles using a nanostructured tip.

Electric field-induced concentration has the potential for application in highly sensitive detection of nanoparticles (NPs) for disease diagnosis and drug discovery. Conventional two-dimensional planar electrodes, however, have shown limited sensitivity in NP concentration. In this paper, the dielectrophoretic (DEP) concentration of low-abundance NPs is studied using a nanostructured tip where a high electric field of 3 × 10(7) V m(-1) is generated. In experimental studies, individual 2, 10, and 100 nm Au NPs are concentrated to a nanotip using DEP concentration and are detected by scanning transmission and scanning electron microscopes. The DEP force on Au NPs near the end of a nanotip is computed according to the distance, and then compared with Brownian motion-induced force. The computational study shows qualitative agreement with the experimental results. When the experimental conditions for DEP concentration are optimized for 8 nm-long oligonucleotides, the sensitivity of a nanotip is 10 aM (10 attomolar; nine copies in a 1.5 µl sample volume). This DEP concentrator using a nanotip can be used for molecular detection without amplification.