Published May 2004
As discussed in PEP Review 2001-2, the engineering aspects of process design attempt to maximize the economic impact and define the best possible process. However, the most underrated aspect of such efforts is the determination of the underlying physical properties used to define the process. Where the previous review focused on gas solubility, this review examines the impact of vapor-liquid equilibria (VLE) on process design.
As with gas solubility, early VLE assumptions tend to be defined by a favored property system with ideal gas behavior unless there is a priori knowledge of the system. When pressed for answers from simulation results, an engineer rarely has the luxury of researching the strengths and weaknesses of alternate property models, nor to evaluate the depth and breadth of parameters available for the chosen property model.
For an existing process, that does not necessarily pose a problem, since some effort will typically be expended on such issues during the development and commercialization of the process. However, inadequate property studies for existing processes and the potential complete lack of appropriate property information for new processes do present difficulties to the engineer. The consequences are illustrated using same ammoximation process from the previous review employing ammonia gas and aqueous hydrogen peroxide to convert cyclohexanone to cyclohexanone oxime.
In the previous review, ammonia was treated as a Henry's Law component in one case and as a normal component for VLE in the next case. The Henry's Law case handles the ammonia easily, while the VLE case struggles with unusual behavior from the ammonia. The cases are repeated in this review to emphasize implications on the t-butanol distillation column. Since ammonia is not supercritical at the process conditions, the VLE route should technically be favored. In both cases, too much water is sent to the distillate, exposing the cyclohexanone oxime to excessive heat. A third case explores the t-butanol / water binary and finds better parameters for simulation and correcting the azeotrope position, but the case fails to solve the ammonia problem. The final case suppresses ammonia-containing binaries to circumvent erroneous parameters caused by simplistic assumptions in group contribution estimation methods.
This review is intended to show the problems and limitations with physical property data and the impact on simulation results, as well as the subsequent process design.
