Initial column profile with composition-dependent relative volatility

When solving the large scale, highly nonlinear, equation systems of steady state multistage distillation processes modelled with theoretical plates, convergence depends on the initial values. The conventionally applied linear initial profiles frequently give rise to divergence when azeotrope is present. The usual practice in such cases is initializing with engineering insight, i.e. anticipating the results. Composition-dependent relative volatility model is developed in the present work, and applied successfully to initialize the composition profiles of distillation columns. The model describes the implicit temperature dependence through composition dependence. Equlibrium plots computed with the new model well approximate the measured data of strongly non-ideal, and even azeotropic, mixtures. The initial column profiles computed with the new model are rather similar to the final solution, and thus accelerate the computation. Processes can be modelled with the new initial profile when the solution algorithms do not converge with the conventional initialization.

Research highlights▶ When the solution algorithms do not converge with the conventional initialization, they have a greater chance to converge if the new relative volatility-based initial profile estimation is applied. ▶ Our model describes the implicit temperature dependence through composition dependence. ▶ Our new composition-dependent relative volatility model is regressed once before any column calculation, and applied as a simplified thermodynamic model to calculate a good initial column profile. ▶ The initial profiles produced this way are very good approximations to the final profiles obtained by rigorous calculation.