(153a) Industrial Multiscale Modeling: Methods and Applications | AIChE

(153a) Industrial Multiscale Modeling: Methods and Applications

Truly "multiscale" modeling has long been a goal of commercial modeling enterprises. The rationale being that challenges in industrial research and development are inherently multiscale: requiring investigation and understanding at length and time scales ranging from the subatomic to the bulk. Multiscale modeling constitutes a holistic approach to industrial research problems.

Over the past 20 years the spectrum of modeling scales has been completed particularly with the development of coarse-grained "mesoscale" simulation techniques. However, methods to connect the modeling scales have lagged behind. These methods are necessary to open the door to multiscale approaches.

Most attempts to provide multiscale modeling solutions use a top-down approach, where existing methodologies are brought together in a more or less ad hoc way to create a suite of tools that can be used to investigate real problems at multiple length and time scales. A disadvantage of this route is a lack of coordination between the methodologies leaving too much work for the end-user. Another drawback is that new methodologies must be added to this melange and the connection points with existing tools devised and delivered for each new module.

Within Culgi we have taken a bottom-up approach to this challenge. The starting point is a robust but flexible infrastructure into which state-of-the-art methodologies have been, and continue to be, incorporated. Since all methods share the same underpinning, new techniques are always as tightly integrated as existing modules, file formats are systematically related and communication between the scales is readily achieved.

The Culgi library provides a graphical programming environment (GPE) that allows the user to immediately perform standard modeling tasks (model building, energy minimization, etc.) and also to create arbitrarily complex, multiscale simulation scenarios within a script. The GPE acts as a guide to the creation of these scripts with contextual help, completeness checks, debuggers, etc. Even without prior knowledge of programming or script languages the scientist can rapidly create scripts tailored to their research needs. The script can call any of the algorithms included in the Culgi library, including: amorphous builders, a molecular dynamics engine with a range of force fields, Monte Carlo methods, Dissipative Particle Dynamics, dynamic density functional theory, Brownian dynamics, hybrid bead-field methods, etc.

To make movement between modeling scales seamless and transparent to the user, common file formats and ready communication are only part of the battle. Scientifically we must provide validated schemes to allow the user to replace a given system with an equivalent system at a different level of coarse-graining, ideally with minimal user-input. We provide "mappers" that can be used to move in either direction between an atomistic representation of a system and a coarse-grained bead-spring model. The details of these mappers will be presented.

Results of various case studies will be presented, to highlight the flexibility and breadth of application of the Culgi library. A multiscale study of asphaltene dispersion in crude oils, a model of the wet spinning of Kevlar involving high shear rates and studies involving controlled release formulations of drugs from structured delivery devices will be presented.

Culgi is an international software enterprise, with offices in USA, EU and China. Culgi participates in EU projects, and has customers in the pharmaceutical, chemical and petrochemical industries.