posted on 2022-12-16, 13:40authored byXiaodong Zhang, Lu Jin, Hao Lyu, Jie He, Jinsheng Sun
The optimal synthesis of distillation sequences has been
a popular
research topic in the field of chemical process system engineering
for many years. In most works, the optimization procedure first generates
a set of information about the presence or absence of stream states
or equipment through some specific rules, usually represented by explicit
binary variables, and then corresponds to a specific process structure,
i.e., “encoding → structure”. The purpose of
encoding information is to “draw” the corresponding
process structures and evaluate their performance. Typically, only
the fully stacked structures with exactly N –
1 columns are considered for N-component separation.
The current work developed an array formulation to completely consider
the complete search space of the state-equipment-network distillation
sequence superstructure. The presence or absence of stream states
or equipment is determined by the calculation results of the previous
devices, which then corresponds to a specific encoding form, i.e.,
“structure → encoding”. The purpose of encoding
information on the “already drawn” nonstacked process
structures is to evaluate the performance of their corresponding various
stacked structures. As an alternative, the features of this inverse
procedure are that the binary variables representing the on/off state
of equipment, and the logical constraints between different process
alternatives are implicitly included in the “Selector”
module of the process simulator. A close-boiling industrial organosilicon
case, which was abstracted as a quaternary zeotropic system, was surveyed
using the proposed method. The optimal configuration along with corresponding
operating parameters was found from 63 flowsheet equivalents in 4.2
CPU hours. Furthermore, five high-quality solutions including both
stacked and nonstacked configurations were compared and analyzed to
demonstrate the optimality of the final solution.