NSGA-II for Multiobjective Optimization of Pervaporation Process: Removal of Volatile Organics from Water Gopal R. Nemmani, † Satyanarayana V. Suggala,* ,‡ and Prashant K. Bhattacharya § Department of Chemical Engineering, Bapatla Engineering College, Bapatla, India, Department of Chemical Engineering, J.NTU College of Engineering, Anantapur, India, and Department of Chemical Engineering, Indian Institute of Technology, Kanpur, India 208016 Pervaporation is fast emerging as a viable technique for the removal of VOC from wastewater. In our previous study (Satyanarayana, V. S.; Bhattacharya, P. K. Ind. Eng. Chem. Res. 2003, 42, 3118) we have carried out the minimization of the treatment cost for a fixed toluene removal fraction from multicomponent wastewater by single-stage pervaporation using shell and tube module without recycling permeate. In the present work, two objective optimization problem of minimization of the treatment cost with simultaneous maximization of percent removal of toluene is studied using an evolutionary algorithm of NSGA-II. The previously available model (Satyanarayana, V. S.; Bhattacharya, P. K. Ind. Eng. Chem. Res. 2003, 42, 3118) is employed including fiber diameter as an eighth decision variable. The study reveals that attractive trade-offs are available between the two objectives. Further, vacuum and condensation cost is found to be the major contributor to the treatment cost for the entire range of Reynold’s number and the contribution is found to be more dominating with decrease in Reynold’s number. 1.0. Introduction Removal of VOCs from air and water streams has been a global problem and billions of dollars are spent annually to remove these environmentally hazardous pollutants. With conventional separation techniques facing problems in removing VOCs from wastewater, the clean-technology alternatives like reverse osmosis, membrane contactors, and pervaporation are now being explored. In this regard, pervaporation, 2 considered to be under clean technology, is fast emerging as a viable unit process; particularly, when the process is overcoming its basic disadvantages of low flux and selectivity because of advent of newer materials, membrane preparatory techniques, and module designs. Industries, hitherto reluctant, now seriously consider membrane processes as alternatives 3 to their existing conven- tional processes like adsorption, distillation, advanced oxidation, and biological treatment, etc. Such utilities of the processes have drawn the attention of many for the removal of VOCs from wastewater as recently there has been significant improvements in the context of membrane materials, 4,5 module designs, 6,7 hybridizations, 8-10 and boundary layer formation, 11,12 etc. Pervaporation is proving to be a better alternative (especially for multicomponent mixtures 1 ) than reverse osmosis and membrane contactors; the former facing the limitation of high osmotic pressures and, in the membrane contactors, the VOCs are just transferred from one phase to another. Peng et al. 13 have reviewed the pervaporation-studies on the removal of VOCs from wastewater. Pervaporation is a membrane-based separation process in which a given feed mixture is brought in contact with a nonporous perm-selective barrier on one side with vacuum being applied on the other side to collect the permeate as vapor enriched with the preferentially permeating component on the down stream side. The vapor collected is often condensed and removed as liquid. For successful commercialization of the process, the effects of the process variables on performance are important. Lipski and Cote 14 have set up a mathematical model to study the effect of process variables like Reynold’s number, thickness of the fiber, and configuration of the modules on treatment cost. The model was then extended by Ji et al., 15 considering the effect of down stream pressure and variable volumetric feed flow rate in the fiber for multicomponent mixtures. Further, they also carried out single objective optimization of treatment cost for binary system as a function of process conditions using Powell’s conventional optimization technique. Later on an evolutionary algorithm of real coded simple genetic algorithm was used by Satyanarayana et al. 1 to determine the optimal conditions of the process. This was done for multicomponent mixture separa- tion with single objective optimization (minimization) of treat- ment cost. However, for VOC separation both objectives, that is, minimization of treatment cost and maximization of VOC removal are important. Therefore, the present work is carried out with the objective of multiobjective optimization problem utilizing previously 1 developed mathematical model. The basic concept of the multiobjective optimization 16 is to find a set of solutions which are called a nondominated set as none of the solutions are supposed to dominate any other solution in the set and there is no single solution which is the best with respect to all the objectives in the entire search space. In other words as we move from one point to the other of the nondominated solutions in the objective function space at least one of the objectives must be improving with simultaneous deterioration of at least one of the other objectives. This is also called the Pareto optimal solutions or Pareto optimal set. If the Pareto optimal set is such that if any other solution in the entire * To whom correspondence should be addressed. E-mail: svsatya7@ gmail.com. Tel: +91-9849509167. Fax: +91-8554-272098. † Bapatla Engineering College. ‡ J.NTU College of Engineering. § Indian Institute of Technology. Table 1. Genetic Parameters maximum number of generations 70.00 maximum population size 800 random seed 0.0625 probability of cross over 0.60 probability of mutation 0.06 distribution index for cross over 5.00 distribution index for mutation 80.00 Ind. Eng. Chem. Res. 2009, 48, 1543–1550 1543 10.1021/ie8005319 CCC: $40.75  2009 American Chemical Society Published on Web 01/08/2009