Location: JHE 264
Process Systems Engineering Methods for Multi-Scale Chemical Process and Product Design
Dr. Mario Richard Eden
Department Chair and McMillan Professor of Chemical Engineering
Process and product design problems by nature are open ended and may yield many solutions that are attractive and near optimal. It is incumbent upon the process systems engineering community to help bridge the gap between fundamental science and engineering applications as new research areas continue to emerge. This presentation will highlight several novel methods for chemical process/product design, specifically: 1) Group contribution based synthesis of process flowsheets, 2) Efficient mixture/blend design using latent variable models; 3) Spatial molecular signature descriptors for generation of non-peptide mimetics; and 4) Modeling the dispersability of polydisperse nanoparticles in gas-expanded liquids.
A systematic group contribution based framework has been developed for synthesis of process flowsheets from a given set of input and output specifications. Analogous to the group contribution methods developed for molecular design, the framework employs process groups to represent different unit operations in the system. Feasible flowsheet configurations are generated using efficient combinatorial algorithms and the performance of each candidate flowsheet is evaluated using a set of flowsheet properties. The design variables for the selected flowsheet(s) are identified through a reverse simulation approach and are used as initial estimates for rigorous simulation to verify the feasibility and performance of the design.
Mixture design is a Design of Experiments (DOE) tool used to determine the optimum combination of chemical constituents that deliver a desired response (or property) using a minimum number of experimental runs. While the approach is sufficient for most experimental designs, it suffers from combinatorial explosion when dealing with the multi-component mixtures found in e.g. pharmaceutical excipients and polymer blends. The property clustering method can alleviate this limitation by transforming the properties to conserved surrogate property clusters described by property operators, which have linear mixing rules even if the operators themselves are nonlinear. Multivariate statistical methods, i.e. principal component analysis (PCA) and partial least squares (PLS), are utilized to generate the property operator models. The methodology is illustrated using several case studies including a polymer blend problem for the formulation of a thermoplastic.
A novel method for incorporating three-dimensional structural information in molecular design algorithms has been developed. The molecular signature descriptor provides a systematic way to encode the atom type and connectivity of a molecular structure, where the signature of a molecule is a linear combination of its atomic signatures. Our earlier works have shown that the signature descriptor is a useful platform for integrating different types of property models, e.g. topological (QSAR/QSPR) and group contribution methods. We have extended this method to include three-dimensional information in the form of a spatial molecular geometry matrix, which can be manipulated to provide several useful descriptors. The ability to include the spatial/topographical (3D) arrangements of the atoms in a molecule is particularly important in applications such as molecular recognition.
The precipitation and size-selective fractionation of nanoparticles is a crucial and exceedingly difficult part of post-synthesis nanomaterial processing. Due to the size-dependent properties of nanoparticles it is often necessary to fine-tune the materials for their intended application, e.g. contrast agents in medical imaging, drug delivery vehicles, highly selective catalysts, etc. Traditionally, these processing steps (particularly the size-selective fractionation) are somewhat trial-and-error in their application, and prediction of the size and size distribution of the recovered nanoparticle fractions is quite difficult. A thermodynamic model has been developed that can accurately predict (typically within 5%) the average size and size distribution of size-selectively fractionated nanoparticles. The application of the model is demonstrated using experimental data for a system of dodecanethiol-stabilized gold nanoparticles in hexane that are precipitated by the addition of CO2.
Return to list of all seminars