Freeform generative design of complex functional structures.

Generative machine learning is poised to revolutionise a range of domains where rational design has long been the de facto approach: where design is practically a time consuming and frustrating process guided by heuristics and intuition. In this article we focus on the domain of flow chemistry, which is an ideal candidate for generative design approaches. We demonstrate a generative machine learning framework that optimises diverse, bespoke reactor elements for flow chemistry applications, combining evolutionary algorithms and a scalable fluid dynamics solver for in silico performance assessment. Experimental verification confirms the discovery of never-before-seen bespoke mixers whose performance exceeds the state of the art by 45%. These findings highlight the power of autonomous generative design to improve the operational performance of complex functional structures, with potential wide-ranging industrial applications.

On the Mechanical Behaviour of Biomimetic Cornstalk-Inspired Lightweight Structures.

This paper presents an investigation on the stiffness and energy absorption capabilities of three proposed biomimetic structures based on the internal architecture of a cornstalk. 3D printing was used to manufacture specimens using a tough and impact-resistant thermoplastic material, acrylonitrile butadiene styrene (ABS). The structural stiffness, maximum stress, densification strain, and energy absorption were extracted from the compression tests performed at a strain rate of 10-3 s-1. A numerical model was developed to analyse the behaviour of the biomimetic structures under compression loading. Further, a damage examination was conducted through optical microscopy and profilometry. The results showed that the cornstalk-inspired biomimetic structure exhibited a superior specific energy absorption (SEA) capability that was three times higher than that of the other core designs as reported in the literature.

Lattice Boltzmann heat transfer model for permeable voxels.

We develop a gray-scale lattice Boltzmann (LB) model to study fluid flow combined with heat transfer for flow through porous media where voxels may be partially solid (or void). Heat transfer in rocks may lead to deformation, which in turn can modulate the fluid flow and so has significant contribution to rock permeability. The LB temperature field is compared to a finite difference solution of the continuum partial differential equations for fluid flow in a channel. Excellent quantitative agreement is found for both Poiseuille channel flow and Brinkman flow. The LB model is then applied to sample porous media such as packed beds and also more realistic sandstone rock sample, and both the convective and diffusive regimes are recovered when varying the thermal diffusivity. It is found that while the rock permeability can be comparatively small (order milli-Darcy), the temperature field can show significant variation depending on the thermal convection of the fluid. This LB method has significant advantages over other numerical methods such as finite and boundary element methods in dealing with coupled fluid flow and heat transfer in rocks which have irregular and nonsmooth pore spaces.

Grayscale lattice Boltzmann model for multiphase heterogeneous flow through porous media.

The grayscale lattice Boltzmann (LB) model has been recently developed to model single-phase fluid flow through heterogeneous porous media. Flow is allowed in each voxel but the degree of flow depends on that voxel's resistivity to fluid motion. Here we extend the grayscale LB model to multiphase, immiscible flow. The new model is outlined and then applied to a number of test cases, which show good agreement with theory. This method is subsequently used to model the important case where each voxel may have a different resistance to each particular fluid that is passing through it. Finally, the method is applied to model fluid flow through real porous media to demonstrate its capability. Both the capillary and viscous flow regimes are recovered in these simulations.

Effect of Uniformly Applied Force and Molecular Characteristics of a Polymer Chain on Its Adhesion to Graphene Substrates.

The force-induced desorption of a polymer chain from a graphene substrate is studied with molecular dynamics (MD). A critical force needs to be exceeded before detachment of the polymer from the substrate. It is found that for a chain to exhibit good adhesive properties the chain configuration should consist of fibrils-elongated, aligned sections of polymers and cavities which dissipate the applied energy. A fibrillation index is defined to quantify the quality of fibrils. We focus on the molecular properties of the polymer chain, which can lead to large amounts of fibrillation, and find that both strong attraction between the polymer and substrate and good solvency conditions are important conditions for this. We also vary the stiffness of the chain and find that for less stiff chains a plateau in the stress-strain curve gives rise to good adhesion however for very stiff chains there is limited elongation of the chain but the chain can still exhibit good fibrillation by a lamella-like rearrangement. Finally, it is found that the detachment time, t, of a polymer from the adsorbed substrate is inversely proportional to force, F (i.e., t ∝ F(-γ)), where exponent γ depends on the solvent quality, polymer-substrate attraction, and chain stiffness.

Polymers encapsulated in short single wall carbon nanotubes: pseudo-1D morphologies and induced chirality.

Molecular dynamics simulations are performed to investigate the stable morphologies of semi-flexible polymer chains within a single wall carbon nanotube (CNT). We characterize these morphologies with a variety of measures. Due to the different curvature inside the CNT to outside, there are increased numbers of polymer-CNT bead contacts for polymers which reside inside the CNT. A sufficiently long polymer chain first adsorbs on the exterior of the nanotube and subsequently moves inside the cavity of the nanotube. At equilibrium, the polymer configuration consists of a central stem surrounded by helically wrapped layers. Sections of the polymer outside the CNT have helical conformations (for CNTs of small radius) or circular arrangements (for CNTs of larger radius). Polymers encapsulated within the CNT have an increased chirality due to packing of the beads and this chirality is further enhanced for moderately stiff chains.

Lattice Boltzmann modeling of permeability in porous materials with partially percolating voxels.

A partial-bounce-back lattice Boltzmann model has been used to simulate flow on a lattice consisting of cubic voxels with a locally varying effective percolating fraction. The effective percolating fraction of a voxel is the total response to the partial-bounce-back techniques for porous media flow due to subvoxel fine structures. The model has been verified against known analytic solutions on two- and three-dimensional regular geometries, and has been applied to simulate flow and permeabilities of two real-world rock samples. This enables quantitative determination of permeability for problems where voxels cannot be adequately segmented as discrete compositions. The voxel compositions are represented as volume fractions of various material phases and void. The numerical results have shown that, for the tight-sandstone sample, the bulk permeability is sensitive to the effective percolating fraction of calcite. That is, the subvoxel flow paths in the calcite phase are important for bulk permeability. On the other hand, flow in the calcite phase in the sandstone sample makes an insignificant contribution to the bulk permeability. The calculated permeability value for the sandstone sample is up to two orders of magnitude greater than the tight sandstone. This model is generic and could be applied to other oil and gas reservoir media or to material samples.

Organization of polymer chains onto long, single-wall carbon nano-tubes: effect of tube diameter and cooling method.

We use molecular dynamics simulations to investigate the arrangement of polymer chains when absorbed onto a long, single-wall carbon nano-tube (SWCNT). We study the conformation and organization of the polymer chains on the SWCNT and their dependence on the tube's diameter and the rate of cooling. We use two types of cooling processes: direct quenching and gradual cooling. The radial density distribution function and bond orientational order parameter are used to characterize the polymer chain structure near the surface. In the direct cooling process, the beads of the polymer chain organize in lamella-like patterns on the surface of the SWCNT with the long axis of the lamella parallel to the axis of the SWCNT. In a stepwise, gradual cooling process, the polymer beads form a helical pattern on the surface of a relatively thick SWCNT, but form a lamella-like pattern on the surface of a very thin SWCNT. We develop a theoretical (free energy) model to explain this difference in pattern structures for the gradual cooling process and also provide a qualitative explanation for the pattern that forms from the direct cooling process.

Langevin dynamics simulations reveal biologically relevant folds arising from the incorporation of a torsional potential.

The conformational behaviour of polymer chains has been examined using Langevin dynamics simulation techniques. Polymer chains were modelled as "beads" undergoing Brownian motion in a defined potential that accounted for stretching, bending and solvation energies. As expected, the competition between chain stiffness and solvent interactions was found to yield standard swollen or collapsed configurations in good or poor solvents, respectively. However, when a torsional term was introduced into the model, additional biologically relevant conformations such as helices, sheets, turns and hairpins naturally arose.

Molecular dynamics simulations of polymeric fluids in narrow channels: methods to enhance mixing.

Mixing of shear thinning polymeric fluids in long channels with patterned boundary conditions is studied through molecular dynamics simulations. Patterned wettability was shown to induce spatially varying slip lengths at the channel walls which in turn induce mixing in the fluid. To quantify the amount of mixing for different wave lengths of patterns, transverse velocity profiles were evaluated. The transverse velocity profiles from the molecular dynamics simulations were then compared with predictions from continuum modeling and good quantitative agreement was found. Offsetting the pattern was shown to produce better mixing in the center of the channel. Transverse flow is found to increase when the radius of gyration of the chains is smaller than the pattern length. We also implement an oscillating (time dependent) body force and find that the transverse flow increases significantly. However, we do not find an increase in transverse flow with frequency of the oscillation as predicted from continuum modeling and we postulate reasons for this behavior.