You are here:
USING GENETIC ALGORITHMS TO DESIGN ENVIRONMENTALLY FRIENDLY PROCESSES
Citation:
Smith*, R L. AND K. J. Kim**. USING GENETIC ALGORITHMS TO DESIGN ENVIRONMENTALLY FRIENDLY PROCESSES. Presented at EPA Science Forum, Washington, DC, 05/5-6/2003.
Description:
Genetic algorithm calculations are applied to the design of chemical processes to achieve improvements in environmental and economic performance. By finding the set of Pareto (i.e., non-dominated) solutions one can see how different objectives, such as environmental and economic performance, are related. Other objectives that are important in the design of processes can be included, such as flexibility and/or controllability. These multidimensional objectives are difficult to optimize because of the tradeoffs between individual objectives. Thus, it is appealing to be able to visualize the tradeoffs between the various objectives, and this can be accomplished through finding the Pareto solutions.
Genetic algorithms are an appealing calculation technique for finding multiobjectional solutions because they search possible solutions without getting stuck in local maximums, as do methods that climb the nearest (short) "mountain". The genetic algorithm creates solutions based on a chromosome (ie., a bit string or series of numbers, similar to a string of DNA, that represent the blueprint for a particular process design). By recombining and mulating a set of chromosomes various designs with selectively better fitness characteristics (i.e., the Pareto solutions) are obtained. The use of recombination and mutation means that a genetic algorithm can search a very large and poorly understood space of solutions without getting stuck in local maximums.
One aspect of genetic algorithms that is being exploited in the current research is the need to create a number of chromosome solutions in parallel, so that fit chromosomes can be recombined with each other to find better solutions. To exploit this need the calculations have been performed on a parallel computing machine made up of used PCs. Each PC can evaluate a design by itself, so that calculation speeds are increased dramatically. The current work focuses on designing a batch distillation column that can be used to recycle used solvents. In the future, much larger design problems are anticipated, and a quick solution to these large, poorly understood problems could revolutionize the way engineers do process design, including the early and simple incorporation of environmental performance.