Content area
Abstract
The petroleum and petrochemical processes are being subject to ever-increasing scrutiny as to not only performance properties but also environmental properties. Thus, the new process modeling paradigm is to track each molecule in the feed and product through the process. Achieving these modeling goals is now possible because of two enabling advances. First, recent developments in analytical chemistry now permit the direct or at least indirect measurement of the molecular structures in complex feedstocks. Second, the explosion in computational power allows for the necessary bookkeeping to track the fate of all the molecules during both reaction and separation processes.
Both pathways and mechanistic modeling approaches describe a large number of species (often O(10$\sp3{-}10\sp4)),$ reactions, and associated rate constants. This can render detailed kinetic modeling tedious, time consuming, and, thus, expensive. The essential conflict to be resolved, then, is the need for a detailed molecular representation and detailed kinetic information, on the one hand, and the formidable size of these models, on the other hand. This motivates the automation of the model building process.
To this end, graph theoretic concepts were used to develop an automated model building capability by representing molecules as graphs and reactions by matrix operations. The reaction kinetics were described in terms of linear free energy relationships (LFER's), which constrained members within families of similar reactions to follow a family-specific quantitative structure-reactivity relationship. The abstract representation of a model, then, was the composition of the molecules in each compound class and a small set (O(10)) reaction matrices and a rule list as the set of model approximations. The approach was used to develop model builders and models for complex chemistries including pyrolysis, catalytic reforming and fluid catalytic cracking.
It was realized that the different model builders for each complex process were various combinations of three fundamental chemistries, i.e. thermal, metal, and acid, at different process conditions. Each process chemistry was realized as an associated set of reaction rules, derived from kinetics approximations, user insights or experiences. A rule-based approach was thus developed for distinguishing various complex processes.
