This
article deals with the phase stability analysis of multicomponent
systems via the determination of all the stationary points of the
so-called tangent plane distance function, which is known as an important
but difficult problem of the thermodynamics of phase equilibrium,
where high nonlinearities are inherent aspects of the different thermodynamic
models commonly used in the description of this distance function.
To analyze phase stability described from different models, we combine
three basic procedures here that were selected in order to attack
points that we deem of relevant difficulty. To this end, we encapsulate
in a single iterative algorithm the following steps: (i) polarization
of a merit function associated with first-order stationary conditions
of the phase stability problem, in order to avoid (or at least minimize)
the repetition of stationary points previously calculated; (ii) a
stochastic optimization method, which (at each iteration of the encapsulation
algorithm) minimizes m times the same merit function
polarized with the predetermined stationary points, using m different seeds for a reliable random number generator
(where m is a given integer), in order to calculate
new stationary points; (iii) a change of variables designed to allow
the numerical method to work in the unconstrained optimization framework,
providing the location of stationary points near the boundary of the
original feasible set. Using the NRTL and UNIQUAC models for liquid–liquid
equilibria at low pressures and the Soave–Redlich–Kwong
and Peng–Robinson cubic equations of state for vapor–liquid
equilibria at high pressures, we analyze 9 multicomponent systems
studied in the literature, equipped with different feed compositions,
totaling 32 tested mixtures, whose component numbers ranged between
3 and 12. We show 14 new stationary points obtained here that were
not detected by methods previously used by other authors.