OPTIMIZATION OF PROCESS DESIGN PROBLEMS USING EVOLUTIOANRY ALGORITHMS

Population based search algorithms (also called Evolutionary Algorithms (EAs)) have shown potential for solving variety of complex problems in the field of engineering. Because EAs results in multiple solutions in a single run they are preferred over the traditional optimization algorithm for solving complex real world problems. The optimum solution in a single objective optimization is a single point where as in multi-objective optimization the decision maker is always interested in having multiple equally good solutions, the Pareto optimal front. However, due to the nonlinear and multi-dimensional search space of industrial problems and benchmark test problems, the attainment of optimum and global Pareto front with good diverse set of solutions is scarce. Various algorithms are applied successfully to find the optimum (for single objective optimization) and Pareto optimal set of solutions (for MOO). Though these algorithms have been successfully applied to some problems, they failed to give either optimal or global Pareto fronts with diverse set of solutions for many other problems. This project aims at simulating and optimizing certain complex real world chemical engineering process problems (in the fields of transportation design, chemical engineering, biochemical engineering, and other engineering fields). It is expected to gain a deeper knowledge of a given process using this approach, which is difficult to get otherwise. The performance and robustness (especially in terms of Pareto front and convergence and divergence) of the optimization results would be tested for the chosen benchmark chemical engineering process and transportation design problems.

Population based search algorithms (also called Evolutionary Algorithms (EAs)) have shown potential for solving variety of complex problems in the field of engineering. Because EAs results in multiple solutions in a single run they are preferred over the traditional optimization algorithm for solving complex real world problems. The optimum solution in a single objective optimization is a single point where as in multi-objective optimization the decision maker is always interested in having multiple equally good solutions, the Pareto optimal front. However, due to the nonlinear and multi-dimensional search space of industrial problems and benchmark test problems, the attainment of optimum and global Pareto front with good diverse set of solutions is scarce. Various algorithms are applied successfully to find the optimum (for single objective optimization) and Pareto optimal set of solutions (for MOO). Though these algorithms have been successfully applied to some problems, they failed to give either optimal or global Pareto fronts with diverse set of solutions for many other problems. This project aims at simulating and optimizing certain complex real world chemical engineering process problems (in the fields of transportation design, chemical engineering, biochemical engineering, and other engineering fields). It is expected to gain a deeper knowledge of a given process using this approach, which is difficult to get otherwise. The performance and robustness (especially in terms of Pareto front and convergence and divergence) of the optimization results would be tested for the chosen benchmark chemical engineering process and transportation design problems.