(698e) Screening for Solvents in Extractive Distillation Considering Quantitative Risk Analysis

Extractive distillation systems are used to separate near-boiling and azeotropic mixtures. In addition to the design variables of conventional distillation problems, the selection of a solvent, or entrainer, is a major variable for the design of extractive distillation schemes. In general, the solvent selection is done taking into account its performance in terms of equipment cost and its capability to alter the relative volatility of the original mixture.¹ Other aspects that are important for the design implementation are usually analyzed after the conceptual design has been developed. One such aspect is related to the system safety, which in the case of extractive distillation systems is of major importance because of the potential risk inherent to the entrainer properties. Patel et al.² have reported an initial work that includes safety aspects during a system design, showing an application for the design of an extraction column.

This work presents an approach for solvent selection in an extractive distillation system. In addition to economic terms, safety measures are included through a Chemical Process Quantitative Risk Analysis (CPQRA). First, potential solvents that meet key properties specifications are found in a pre-selection step through the use of molecular design techniques. Then, the behavior of the potential solvents is tested in ASPEN Plus, and a multi-objective optimization problem is formulated, such that optimal tradeoffs between risk and cost criteria are found. Societal risk (SR) and total annual cost (TCA) are used to quantify the two objective functions.

In the pre-selection step, the ProCAMD tool within the ICAS Software, developed at the Technical University of Denmark, is used to find components that reach a set of specified solvent properties. In the optimization step, a CPQRA is formulated based on CCPS guidelines.³ SR is a measure to estimate risk for a group of people located to an affected area given a set of incidents. Risk is a function of probabilities of incidents and consequences. Incident probabilities are calculated through fault trees and event trees. Quantification of consequences implies source, dispersion, and incident effects calculations. All these estimations are represented by nonlinear models, which in addition to the equations for the distillation system provide a highly nonlinear model. To handle this situation, a Derivative Free Optimization (DFO) procedure was implemented within MATLAB, with ASPEN Plus simulations working as a black box within the DFO, and a genetic algorithm recursively used to find the set of optimal tradeoffs (or Pareto front) between both objectives. Results from the application to case studies are presented to show how the best potential solvents for a given separation task are identified through the use of the proposed approach.

1.            Luyben, W. L.; Chien, I. L., Design and Control of Distillation Systems for Separating Azeotropes. John Wiley & Sons: 2011.

2.            Patel, S. J.; Ng, D.; Mannan, M. S., Inherently safer design of solvent processes at the conceptual stage: Practical application for substitution. Journal of Loss Prevention in the Process Industries 2010, 23 (4), 483-491.

3.            CCPS, A. I. o. C. E., Guidelines for Chemical Process Quantitative Risk Analysis. 2 ed.; New York, New York, 2000; p 756.