Multiscale CFD modeling for design, simulation and stability of distributed chemical process systems: Application to carbothermic aluminium production

Multiscale modeling is a powerful idea for accurate and efficient simulation of challenging chemical processes characterized by significant complexity at several length and time scales. A plethora of emerging and future chemical processes is of extreme industrial importance and cannot be satisfactorily studied using the arsenal of standard modeling simplifications. Accurate process representations need rely on nonlinear partial differential equation systems that exhibit spatiotemporal variation and fluid flow, not allowing for model order reduction. A variety of further challenges may significantly perplex reliable process modeling efforts: poor understanding of underlying physics inherently limits the scope and potential of models; computational expense is frequently prohibitive in terms of cost, CPU time and applicability; outsourcing and intellectual property restrictions result in distributed, inadequate knowledge; poorly documented legacy computer codes with are not easily interfaced with modern tools; finally, costly validation experiments have to be conducted at extreme operating conditions, minimizing the feedback margin between process modeling and experimental investigation. The notion of scale integration is closely related to the concept of multiscale modeling and is proposed as a strategy for efficiently addressing classes of complex modeling problems: it involves construction and coordination of a series of individual models for the various scales of interrelated phenomena, rather than an aggregate large-scale model of extreme complexity. The suitable partitioning of phenomena and the seamless integration of computational tools is then the challenge, since proper decomposition largely depends on the nature of each process.; This doctoral dissertation focuses on illustrating and quantifying multiscale modeling concepts that address the design and simulation of a new distributed chemical process system. The body elaborates on the construction of a multiscale model of the conceptual ARP reactor, a new multistage and multiphase prototype developed for carbothermic aluminium production (a potentially feasible but technically complex alternative electrochemical process whose cornerstone principle is to exploit the endothermic reduction reaction between carbon and alumina instead of resorting to the electrolysis of aluminium oxide towards pure aluminium). Model construction and validation is organized around three distinct complementary levels: the first level consists of a CSTR series resembling the carbothermic reactor configuration, the second level uses a detailed CFD model solving multiple PDE equilibria in each section and the third level can probe thermodynamic equilibrium via Gibbs free energy minimization. An MINLP design and optimization study and multiple reactor CFD simulation studies focus on simultaneous solution of charge, heat, momentum, mass and gas volume fraction balances and succeed in accurate state variable distributions confirmed by experimental measurements. Multiscale modeling is therefore suitable for efficient complex process design and simulation.