Design of tailor-made chemical blend using a decomposition-based computer-aided approach

Computer aided techniques form an efficient approach to solve chemical product design problems such as the design of blended liquid products (chemical blending). In chemical blending, one tries to find the best candidate, which satisfies the product targets defined in terms of desired product attributes (properties). The systematic computer-aided technique first establishes the search space, and then narrows it down in subsequent steps until a small number of feasible and promising candidates remain. At this point, experimental work may be conducted to verify if any or all the candidates satisfy the desired product attributes. Alternatively, rigorous modeling could also be used in this final step. In other words, the candidates are quickly generated and screened until a small number is left for final selection and evaluation by experiments and/or rigorous modeling. This paper presents a design methodology for blended liquid products that identifies a set of feasible chemical blends. The blend design problem is formulated as a Mixed Integer Nonlinear Programming (MINLP) model where the objective is to find the optimal blended gasoline or diesel product subject to types of chemicals and their compositions and a set of desired target properties of the blended product as design constraints. This blend design problem is solved using a decomposition approach, which eliminates infeasible and/or redundant candidates gradually through a hierarchy of (property) model based constraints. This decomposition method reduces the search space in a systematic manner and the general blend design problem is decomposed into two stages. The first stage investigates the mixture stability where all unstable mixtures are eliminated and the stable blend candidates are retained for further testing (note that all blends must be stable liquid mixture). In the second stage, the blend candidates have to satisfy a set of target properties that are ranked according to a specified priority. Finally, a short list of candidates, ordered in terms of specified performance criteria, is produced for final testing and selection. The application of this systematic and computer-aided approach is illustrated through a case study involving the design of blends of gasoline with oxygenated compounds resulting from degradation and fermentation of biomass for use in internal combustion engines. Emphasis is given here on the concepts used and on the validation of the property models, mainly, the Reid vapor pressure model and the liquid phase stability tests.