A universal approach to phenomenological compartment models of unit operations

A compartment model describes the transmission of materials and/or energies through a unit operation, as a network of flow connected sub-volumes. Each sub volume is a well-mixed compartment, formed based on the identification of negligible gradients in the system properties of interest. Ordinary differential equations describe the temporal phenomenological and flow effects imposed on the variables (species mass and compartment enthalpy) of the system. Along with the associated initial values of the system, the variable ODE’s are numerically solved over time. Compartment modelling is widely used in chemical engineering as it provides a balance between flow and phenomena resolution, and solution times. From the profusion of compartment models in literature, the model development and thus solutions for this approach are both bespoke. Models are either hard coded ODE’s or built through the improvised use of available non-domain-specific tools; the former is especially error prone, and the latter restricts the model development to the capability of the tool used. For full modelling flexibility, modellers are required to have knowledge of software design for implementing and solving ODE’s with many variables. CompArt - A universal compartment modelling tool for unit operations has been developed in this work, this is formed of (i) a universal input language used to describe unit operation compartment models, (ii) complemented by an interpretation algorithm for the conversion of the model description into ODE’s for solving (utilising a universal compartment modelling equation set developed in this work) and, (iii) the wrapping of choice numerical solvers targeting stiff non-linear problems. This addition to the field circumvents the need for modelers to have specialised skills to utilise this modelling approach allows focus upon their domain of model development to take priority. The universal compartment modelling system, CompArt is validated against a benchmark set of 20 models ranging in structural make-up and applied phenomena.